PHONETICS and PHONOLOGY

Editors

STEPHEN R. ANDERSON PATRICIA A . K E A T I N G

Department of Cognitive Science Department of Linguistics

The Johns Hopkins University University of California, Los Angeles
Baltimore, Maryland 21218 Los Angeles, California 90024

A list of titles in this series appears at the end of this book.

PHONETICS and PHONOLOGY
VOLUME 4
Studies in Lexical Phonology

Edited by

Sharon Hargus
Ellen M. Kaisse
Department of Linguistics
University of Washington
Seattle, Washington

ACADEMIC PRESS, INC.

A Division of Harcourt Brace & Company
S a n D i e g o N e w York Boston
London Sydney Tokyo Toronto
This book is printed on acid-free paper. @

Copyright © 1 9 9 3 by A C A D E M I C P R E S S , I N C .
All Rights Reserved.
No part of this publication may be reproduced or transmitted in any form or by any
means, electronic or mechanical, including photocopy, recording, or any information
storage and retrieval system, without permission in writing from the publisher.

Academic Press, Inc.

1250 Sixth Avenue, San Diego, California 92101-4311

United Kingdom Edition published by

Academic Press Limited
24-28 Oval Road, London NW1 7DX

Library of Congress Cataloging-in-Publication Data

Studies in Lexical phonology / edited by Sharon Hargus and Ellen M. Kaisse.

p. cm. - (Phonetics and phonology ; v. 4)
Includes index.
ISBN 0-12-325070-6 (Hardcover)
ISBN 0-12-325071-4 (Paperback)
1. Lexical phonology. I. Hargus, Sharon. II. Kaisse, Ellen M.
III. Series.
P217.62.L49 1993
414-dc20 92-23535
CIP

PRINTED IN THE UNITED STATES OF AMERICA

93 94 95 96 97 98 QW 9 8 7 6 5 4 3 2 1
CONTRIBUTORS

Numbers in parentheses indicate the pages on which the authors' contributions begin.

G e e r t B o o i j (23), V a k g r o e p T a a l k u n d e , Vrije Universiteit, 1007 M C A m s t e r d a m ,

The Netherlands
T o n i B o r o w s k y ( 1 9 9 ) , D e p a r t m e n t of L i n g u i s t i c s , University of Sydney, S y d n e y ,
N e w South Wales 2006, Australia
B . E l a n D r e s h e r ( 3 2 5 ) , D e p a r t m e n t of Linguistics, University of T o r o n t o , T o -
ronto, Ontario, Canada M 5 S 1 A l
S h a r o n H a r g u s ( 1 , 4 5 ) , D e p a r t m e n t of Linguistics, University of W a s h i n g t o n ,
Seattle, W a s h i n g t o n 9 8 1 9 5
L a r r y M . H y m a n ( 2 3 5 ) , D e p a r t m e n t of Linguistics, University of California,
Berkeley, California 9 4 7 2 0
S h a r o n I n k e l a s (75), D e p a r t m e n t of L i n g u i s t i c s , University of California, B e r k e -
ley, California 9 4 7 2 0
G r e g o r y K. I v e r s o n ( 2 5 5 ) , D e p a r t m e n t of Linguistics, University of W i s c o n -
sin—Milwaukee, Milwaukee, Wisconsin 53201
E l l e n M . K a i s s e ( 1 , 3 4 3 ) , D e p a r t m e n t of Linguistics, University of W a s h i n g t o n ,
Seattle, W a s h i n g t o n 9 8 1 9 5
P a u l K i p a r s k y ( 2 7 7 ) , D e p a r t m e n t of L i n g u i s t i c s , Stanford University, Stanford,
California 9 4 3 0 5
R o c h e l l e L i e b e r (23), D e p a r t m e n t of E n g l i s h , University of N e w H a m p s h i r e ,
Durham, New Hampshire 03824
D a v i d O d d e n ( 1 1 1 ) , D e p a r t m e n t of L i n g u i s t i c s , O h i o State University, C o l u m -
bus, Ohio 43210
W i l l i a m J. P o s e r ( 3 1 5 ) , D e p a r t m e n t of Linguistics, Stanford University, Stan-
ford, California 9 4 3 0 5

xi
xii Contributors

K e r e n D . R i c e (145), D e p a r t m e n t of Linguistics, University of Toronto, Toronto,

O n t a r i o , C a n a d a M 5 S 1A1
R i c h a r d S p r o a t ( 1 7 3 ) , Linguistics R e s e a r c h D e p a r t m e n t , A T & T Bell L a b o r a t o -
ries, M u r r a y Hill, N e w Jersey 0 7 9 7 4
D r a g a Z e e ( 3 6 5 ) , D e p a r t m e n t of M o d e r n L a n g u a g e s and Linguistics, C o r n e l l
University, Ithaca, N e w York 14853
PREFACE

W h a t d o lexical p h o n o l o g i s t s believe a b o u t p h o n o l o g y ? To w h a t u s e s are they

putting t h e t h e o r y ? W h a t c o r e of t h e o r e m s r e m a i n s from the " c l a s s i c a l " version
of lexical p h o n o l o g y , a n d w h a t h a s b e e n w i d e l y rejected or seriously c h a l l e n g e d ?
A l t h o u g h w e c o n s i d e r o u r s e l v e s to b e well within the lexical p h o n o l o g y fold, w e
h a v e n o n e t h e l e s s found ourselves u n c e r t a i n a b o u t these points a n d a b o u t others as
well. At the t i m e w e first v o i c e d these uncertainties to e a c h other, in late 1989,
there h a d b e e n n o m a j o r c o n f e r e n c e s o n lexical p h o n o l o g y for m a n y years, a n d
the last sizable collection of p a p e r s with an e m p h a s i s on lexical p h o n o l o g y h a d
a p p e a r e d in 1985 in Phonology Yearbook 2. W i t h a recent w o r k s h o p o n the p h o -
n o l o g y - s y n t a x interface as an inspiration (see Inkelas a n d Z e e , 1990), w e d e c i d e d
to apply to the N a t i o n a l S c i e n c e F o u n d a t i o n for funding to host a w o r k s h o p o n
lexical p h o n o l o g y . W h e n funding w a s a p p r o v e d , w e o r g a n i z e d this w o r k s h o p ,
w h i c h t o o k p l a c e at the University of W a s h i n g t o n in J u n e , 1990. T h e p r e s e n t vol-
u m e c o n t a i n s m a n y of the p a p e r s p r e s e n t e d at that w o r k s h o p , since revised in the
light of c o m m e n t s from participants a n d referees, p l u s additional p a p e r s from lexi-
cal p h o n o l o g i s t s w h o did not attend. W e trust that other p h o n o l o g i s t s w h o h a v e
b e e n w o n d e r i n g the s a m e things that w e w o n d e r e d in 1989 will find this collection
of articles useful.
F o r general i n t r o d u c t i o n s to lexical p h o n o l o g y , the r e a d e r is referred to Booij
a n d R u b a c h ( 1 9 9 1 ) , K a i s s e a n d S h a w ( 1 9 8 5 ) , and w o r k s cited therein. O u r intro-
duction c o n c e n t r a t e s o n subjects that call for special explication a n d on topics
treated in several articles in t h e v o l u m e . O u r original call for p a p e r s s u g g e s t e d that
authors c o n c e n t r a t e o n t w o of o u r favorite subfields: p h o n o l o g y - m o r p h o l o g y in-
teraction a n d the application of lexical p h o n o l o g y to d i a c h r o n i c investigations.
T h e r e a d e r will therefore notice s o m e bias t o w a r d these areas, but a l m o s t e v e r y
d i m e n s i o n of lexical p h o n o l o g y receives s o m e attention in the d o z e n or so p a p e r s
in this v o l u m e .
xiii
xiv Preface

W e t h a n k the N S F for p r o v i d i n g funding ( N S F grant # B N S - 8 9 1 9 4 7 5 ) . W e also

t h a n k Siri Tuttle for d o i n g m o s t of the n u t s - a n d - b o l t s o r g a n i z a t i o n w h i c h resulted
in a successful w o r k s h o p .
W e e a c h c o n t r i b u t e d equally to the task of editing; o u r n a m e s are printed b e l o w
in alphabetical order.

Sharon Hargus
Ellen M . K a i s s e

REFERENCES

Booij, G., and Rubach, J. (1991). Lexical phonology. In International Encyclopedia of

Linguistics (W. Bright, ed.), 2, pp. 3 2 7 - 3 3 0 . Oxford University Press, New York.
Inkelas, S., and Zee, D. (1990). The Phonology-Syntax Connection. CSLI Publications
and University of Chicago Press, Chicago.
Kaisse, E. M, and Shaw, P. (1985). On the theory of lexical phonology. Phonology Year-
book!, 1-30.
INTRODUCTION

ELLEN M. KAISSE
SHARON HARGUS
Department of Linguistics
University of Washington
Seattle, Washington 98195

1. T H E O R E T I C A L C O M M O N D E N O M I N A T O R 1

It m a y at t i m e s s e e m to the o b s e r v e r that there is n o c o m m o n c o r e of beliefs

that all lexical p h o n o l o g i s t s a d h e r e to. A n d indeed, the p a p e r s in this v o l u m e re-
flect differences o n practically every tenet of " c l a s s i c a l " lexical p h o n o l o g y , in-
c l u d i n g the s o u r c e of w o r d - i n t e r n a l p h o n o l o g i c a l d o m a i n s , p h o n o l o g y - m o r p h o l -
o g y interaction, structure p r e s e r v a t i o n , a n d the strict c y c l e c o n d i t i o n . Yet there is
a p e r v a s i v e similarity in t h e sorts of q u e s t i o n s that lexical p h o n o l o g i s t s ask w h e n
they d e s c r i b e a l a n g u a g e , q u e s t i o n s w h i c h m i g h t not h a v e e v e n b e e n t h o u g h t of
ten years a g o . N o t m a n y p h o n o l o g i s t s in the early eighties w o u l d h a v e tried as a
m a t t e r of c o u r s e to find out w h e t h e r a rule they w e r e studying applied b e t w e e n
w o r d s ; w h e t h e r it w a s sensitive to syntactic, p r o s o d i c , or m o r p h o l o g i c a l d o m a i n s ;
w h e t h e r it h a d e x c e p t i o n s ; w h e t h e r it applied only in m o r p h o l o g i c a l l y derived
c o n t e x t s ; w h e t h e r the rule w a s n e u t r a l i z i n g or rather supplied a l l o p h o n i c features;
w h e t h e r it w a s c y c l i c ; w h e t h e r it w a s crucially o r d e r e d with a rule already k n o w n
to b e lexical or postlexical; a n d so forth. E v e n if s o m e of these q u e s t i o n s m i g h t
h a v e b e e n a d d r e s s e d individually, there w o u l d h a v e b e e n n o o v e r a r c h i n g theoreti-
cal v i e w p o i n t that w o u l d m a k e s u c h p i e c e s of information interrelated a n d critical
to the overall p i c t u r e of h o w the rule fit into the g r a m m a r as a w h o l e .
In addition to a g r e e i n g on the essential information that b e l o n g s in p h o n o l o g i -
cal description, m a n y lexical p h o n o l o g i s t s c o n t i n u e to a g r e e on the following g e n -

1
Phonetics and Phonology, Volume 4 Copyright © 1993 by Academic Press, Inc.
Studies in Lexical Phonology All rights of reproduction in any form reserved.
2 Ellen M. Kaisse and Sharon Hargus

eral points. First, l a n g u a g e s h a v e rules with different clusters of characteristics (as

p o s i t e d b y Kiparsky, 1 9 8 3 , for e x a m p l e ) , usually associated with the labels LEXI-
CAL a n d POSTLEXICAL. W e will see, h o w e v e r (Section 7 ) , that this rule t y p o l o g y is
not as simple and straightforward as w a s first suggested.
S e c o n d , l a n g u a g e s m a y h a v e w o r d - i n t e r n a l p h o n o l o g i c a l d o m a i n s w h i c h are
derived from m o r p h o l o g i c a l structures in various w a y s b u t d o n o t necessarily cor-
r e s p o n d to either m o r p h o l o g i c a l or m e t r i c a l p h o n o l o g i c a l structures. T h i s hy-
p o t h e s i s , discussed in greater detail in Section 3 , is n o w generally r e g a r d e d as
s u p e r s e d i n g the level o r d e r i n g h y p o t h e s i s of Siegel ( 1 9 7 9 ) a n d A l l e n ( 1 9 7 8 ) , with
i m p r o v e d e m p i r i c a l a n d theoretical c o n s e q u e n c e s .
Finally, l a n g u a g e s d o n o t allow postlexical rules, b e they syntactic or p h o n o -
logical, to refer to w o r d - i n t e r n a l structure. In the past, this h a s b e e n e x p l a i n e d b y
positing a b r a c k e t i n g e r a s u r e c o n v e n t i o n , w h i c h erases w o r d - i n t e r n a l brackets at
s o m e p o i n t in the lexical derivation. A g a i n , the issue is n o t as s i m p l e as it first
a p p e a r e d . T h e invisibility of lexical j u n c t u r e s h a s b e e n c h a l l e n g e d b y the exis-
t e n c e of postlexical rules w h i c h a p p e a r to refer to w o r d - i n t e r n a l p h o n o l o g i c a l d o -
m a i n s (Sproat, this v o l u m e , and Section 2.1 of this I n t r o d u c t i o n ) , a n d an alterna-
tive theory ( " p r e c o m p i l a t i o n " ) h a s b e e n d e v e l o p e d b y H a y e s (1990) to h a n d l e
such cases as well.
W e see then that there is a g r e e m e n t c o n c e r n i n g the gross characteristics of the
m o d e l and a b o u t the k i n d s of q u e s t i o n s that are of interest in arriving at a p h o n o -
logical analysis. In m u c h of the rest of this introduction, w e discuss m o r e c o n t r o -
versial aspects of the theory.

2. M O R P H O L O G I C A L Q U E S T I O N S

A large n u m b e r of articles in this v o l u m e (Booij and Lieber, B o r o w s k y , H a r g u s ,

Inkelas, O d d e n , Sproat) deal with q u e s t i o n s c o n c e r n i n g the interaction of p h o -
n o l o g y and m o r p h o l o g y .

2.1 B r a c k e t E r a s u r e

L i k e Inkelas ( 1 9 9 0 ) , S p r o a t (this v o l u m e ) suggests that reference to m o r p h o -

logical b r a c k e t i n g m a y b e b l o c k e d b y locality c o n d i t i o n s o n p h o n o l o g i c a l rules.
E n g l i s h l\l h a s " l i g h t " a n d " d a r k " a l l o p h o n e s , the major d e t e r m i n a n t s of light-
n e s s and d a r k n e s s b e i n g the length of a p r e b o u n d a r y r i m e a n d the d e g r e e of t o n g u e
d o r s u m retraction. B e c a u s e of the n o n c a t e g o r i c a l nature of its structural c h a n g e ,
/1/-Darkening is c o n s i d e r e d to b e a postlexical rule of p h o n e t i c i m p l e m e n t a t i o n ,
yet, as Sproat s h o w s , the d e g r e e of d a r k e n i n g d e p e n d s o n the kind of (lexical)
b o u n d a r y (%) w h i c h follows / ! / in the c o n t e x t /V1%V7. Laterals following + and
Introduction 3

# b o u n d a r i e s pattern together, but c o m p o u n d b o u n d a r i e s are treated differently:

i.e., in n o n s e n s e w o r d s beel-ic and beel-ing III is lighter than it is in heel equator.
S p r o a t s u g g e s t s that the increased d a r k n e s s of l\l before the c o m p o u n d b o u n d a r y
m i g h t b e attributed to indirect visibility of w o r d - i n t e r n a l structure via m e t r i c a l or
p h o n o l o g i c a l d o m a i n c o n s t r u c t s , raising t h e q u e s t i o n of w h e t h e r or not there is
any e r a s u r e of m o r p h o l o g i c a l b r a c k e t s .

2.2 I n t e r a c t i o n i s m

T h e classical t h e o r y of lexical p h o n o l o g y w a s b a s e d in p a r t o n a s e e m i n g l y well

established n u m b e r of a n a l y s e s in w h i c h p h o n o l o g i c a l rules apply cyclically (e.g.,
B r a m e , 1974; C h i e r c h i a , 1 9 8 3 ; C o h n , 1989; M a s c a r o , 1976; R u b a c h , 1984; etc.;
see C o l e , 1 9 9 1 , for a recent o v e r v i e w of the p h o n o l o g i c a l cycle). F o l l o w i n g M a s -
c a r o ( 1 9 7 6 ) , cyclic rules w e r e c o n s i d e r e d to b e typically n e u t r a l i z i n g , rather t h a n
a l l o p h o n i c . T h e h y p o t h e s i s a d o p t e d b y P e s e t s k y ( 1 9 7 9 ) and K i p a r s k y ( 1 9 8 2 )
w a s that the distinction b e t w e e n cyclic a n d postcyclic rules c o i n c i d e d with the
l e x i c a l - p o s t l e x i c a l distinction. T h e e x p l a n a t i o n for this correlation w a s that cy-
clicity followed from the interleaving of p h o n o l o g i c a l a n d m o r p h o l o g i c a l rules,
w i t h the b o l d theoretical c o n s e q u e n c e that p h o n o l o g i c a l rules w e r e then p r e d i c t e d
to b e able to p r e c e d e m o r p h o l o g i c a l o n e s .
H o w e v e r , as d i s c u s s e d in the n e x t section, it is n o w firmly established that m o r -
p h o l o g i c a l a n d p h o n o l o g i c a l (cyclic) structure n e e d n o t b e i s o m o r p h i c , b u t that
there m a y b e a m o r e indirect c o r r e s p o n d e n c e b e t w e e n t h e m . F u r t h e r m o r e , the
h y p o t h e s i s that all lexical rules are cyclic h a s b e e n c h a l l e n g e d recently b y Booij
a n d R u b a c h ( 1 9 8 7 ) w h o p r o p o s e , largely o n the basis of t h e o r y - i n t e r n a l c o n s i d -
erations, that the last set of rules in the lexical c o m p o n e n t applies noncyclically.
B o r o w s k y (this v o l u m e ) p r o p o s e s an even m o r e restrictive version of this hy-
p o t h e s i s , in w h i c h the lexical c o m p o n e n t c o n t a i n s t w o p h o n o l o g i c a l d o m a i n s , the
s t e m level a n d t h e w o r d level, w i t h w o r d - l e v e l p h o n o l o g i c a l rules a p p l y i n g n o n -
cyclically to the o u t p u t s of the stem level a n d also to t h e word-level m o r p h e m e s
p r i o r to their c o n c a t e n a t i o n with the o u t p u t of the stem level.
T h e r e m a i n i n g h y p o t h e s i s of the cluster of related h y p o t h e s e s within the " c l a s -
s i c a l " lexical p h o n o l o g y p o s i t i o n — t h a t p h o n o l o g i c a l rules m a y p r e c e d e m o r p h o -
logical o n e s in s o m e c a s e s — i s w h a t h a s c o m e to b e k n o w n as INTERACTIONISM,
the focus of a n u m b e r of p a p e r s in this v o l u m e . I n t e r a c t i o n i s m is d e f e n d e d by
H a r g u s (this v o l u m e ) , w h o presents a t y p o l o g y of a n a l y s e s w h i c h s u p p o r t the
interactionist position, i n c l u d i n g m o r p h o l o g i c a l rules w h i c h refer to d e r i v e d p h o -
n o l o g i c a l p r o p e r t i e s (e.g., stress-sensitive affix a l l o m o r p h y in F i n n i s h , D u t c h ,
G e r m a n , etc.), a n d p h o n o l o g i c a l rules w h o s e d o m a i n s crucially e x c l u d e s o m e
p h o n o l o g i c a l p r o c e s s (e.g., infixation rules in Sanskrit and S u n d a n e s e w h o s e out-
p u t s n o r m a l l y b l o c k or fail to u n d e r g o the p h o n o l o g i c a l rules in q u e s t i o n ) . Booij
a n d L i e b e r (this v o l u m e ) similarly c o n c l u d e that an interactionist m o d e l is to b e
4 Ellen M. Kaisse and Sharon Hargus

preferred over the m o r e restrictive noninteractionist m o d e l (below) or alternative

m o d e l s p r o p o s e d b y A n d e r s o n ( 1 9 8 8 ) a n d Aronoff ( 1 9 8 8 ) . H o w e v e r , there are
certain facts a b o u t the k i n d s of a n a l y s e s w h i c h s u p p o s e d l y s u p p o r t interactionism
w h i c h classical lexical p h o n o l o g y fails to explain, as p o i n t e d out to us b y D o n c a
S t e r i a d e (personal c o m m u n i c a t i o n ) . First, if an affix s u b c a t e g o r i z e s for a b a s e with
certain derived p h o n o l o g i c a l p r o p e r t i e s , t h o s e properties are a l m o s t a l w a y s supra-
s e g m e n t a l (e.g., stress). S e c o n d , if a p h o n o l o g i c a l rule fails to apply to the output
of a w o r d formation rule, that m o r p h o l o g y is a l m o s t always n o n c o n c a t e n a t i v e
(e.g., reduplication or infixation). I n t e r a c t i o n i s m w o u l d b e better s u p p o r t e d if a
b r o a d e r r a n g e of types of m o r p h o l o g y a n d p h o n o l o g y interacted in the m a n n e r
predicted.
A return to the C h o m s k y a n d Halle ( 1 9 6 8 ) noninteractionist m o d e l , in w h i c h
all m o r p h o l o g y p r e c e d e s all p h o n o l o g y (albeit a version of the p h o n o l o g i c a l c o m -
p o n e n t w h i c h is e n r i c h e d to i n c l u d e w o r d - i n t e r n a l p h o n o l o g i c a l d o m a i n s ) , has
recently b e e n p r o p o s e d b y H a l l e a n d V e r g n a u d (1987a, 1987b), and also adopted
b y H a l l e , H a r r i s , a n d V e r g n a u d ( 1 9 9 1 , H H V ) a n d O d d e n (this v o l u m e ) . O d d e n
p r o v i d e s an interesting e m p i r i c a l a r g u m e n t against interactionism using d a t a from
M a l t e s e . A n o r d e r i n g p a r a d o x arises in an interactionist analysis, w h i c h c a n b e
avoided if all m o r p h e m e s are c o n c a t e n a t e d before any p h o n o l o g i c a l rules apply.
B u t the noninteractive analysis is itself s o m e w h a t o d d , requiring that o n e of the
p h o n o l o g i c a l rules in q u e s t i o n b e a n a l y z e d as precyclic, a p p l y i n g before the d e -
m o n s t r a b l y cyclic stress rules ( B r a m e , 1974); p r e c y c l i c p h o n o l o g i c a l rules are not
o t h e r w i s e k n o w n to exist. L e s s controversially, the M a l t e s e data s e e m to support
H a y e s ' s (1990) theory of p r e c o m p i l a t i o n . A p r e c o m p i l a t i o n analysis of the data
involves neither precyclicity n o r the o r d e r i n g p a r a d o x of the interactive analysis.
A p o i n t w h i c h is m a d e b y several authors ( H H V ; H a r g u s , this v o l u m e ; O d d e n ,
this v o l u m e ) is that s u p p o r t for interactionist a n d noninteractionist positions in
s o m e cases relies on o n e ' s particular c o n c e p t i o n of m o r p h o l o g y with respect to
p h o n o l o g y . C o n s i d e r the S p a n i s h rule of la-el R e p l a c e m e n t , discussed by H H V . 2

T h i s dissimilatory rule a c c o u n t s for alternations b e t w e e n la, the feminine singular

definite article, and el, a f o r m of this article w h i c h is used if a n o u n b e g i n s with
stressed a: la agudda 'the w a t e r s u p p l y ' , la almeja 'the c l a m ' , but el dgua 'the
w a t e r ' , el alma 'the s o u l ' . C o m p l i c a t i o n s arise w h e n el o c c u r s instead of la even
before certain u n s t r e s s e d a's: el aguita ' t h e w a t e r ' (diminutive), el aguanieve 'the
s l e e t ' . To a c c o u n t for t h e s e facts, H H V n o t e that forms with u n s t r e s s e d a w h i c h
take el instead of la a l w a y s alternate with f o r m s with stressed a. T h e y suggest
that these are forms w h i c h either contain " a n o n c y c l i c derivational affix" (di-
m i n u t i v e -it- v s . cyclic -ad- in la agudda) o r o c c u r as the first c o m p o n e n t of a
c o m p o u n d . T h u s H H V posit a crucial distinction b e t w e e n cyclic and n o n c y c l i c
affixes and also a s s u m e that la-el R e p l a c e m e n t p r e c e d e s Conflation (a rule w h i c h
r e m o v e s crucial stresses). T h u s the input to la-el R e p l a c e m e n t will b e la agudda
( w h i c h surfaces with la) versus la aguita ( w h i c h surfaces with el). T h e theoretical
Introduction 5

point w h i c h arises from this discussion is, as m e n t i o n e d a b o v e , that la-el Replace-

m e n t , a rule w h i c h o b v i o u s l y refers to a d e r i v e d p h o n o l o g i c a l p r o p e r t y (stress),
c o u l d b e a n a l y z e d in o n e of t w o w a y s ; as a p h o n o l o g i c a l rule (1), as they c h o o s e
(p. 153), or as " a contextually d e t e r m i n e d lexical c h o i c e , " as s u g g e s t e d b y
Harris (1989).

(1)

la -> //[ N [a
N

T h e i r rationale for f o r m a l i z i n g la-el R e p l a c e m e n t as in (1) is as follows (HHV,

p . 157).

[If formalized as in (1),] then this phenomenon involves no interaction between morpho-
logical rules and those of the phonology. . . . On the other hand, if the phenomenon is
treated by multiple listings in the lexicon, then the separateness of morphology from pho-
nology is compromised (insofar as this type of lexical choice is a matter of morphology).

T h i s sort of analytical uncertainty can b e found in other l a n g u a g e s , as p o i n t e d out

b y H a r g u s (this v o l u m e ) a n d O d d e n (this v o l u m e ) .
L i k e Booij and Lieber, Inkelas (this v o l u m e ) a r g u e s for a m o d e l in w h i c h p h o -
n o l o g i c a l s t r u c t u r e is c o n s t r u c t e d in t a n d e m w i t h m o r p h o l o g i c a l structure, as
3

d i s c u s s e d in m o r e detail in the n e x t section. W h i l e I n k e l a s ' s theory incorporates

v a r i o u s h y p o t h e s e s , h e r p r o p o s a l that the distribution of cyclic and n o n c y c l i c rules
is p r e d i c t a b l e p r o v i d e s general s u p p o r t for i n t e r a c t i o n i s m . A further a s s u m p t i o n
m a d e b y a n u m b e r of noninteractionists (e.g., H a l l e a n d V e r g n a u d , 1987a, 1 9 8 7 b ;
cf. also H a l l e a n d M o h a n a n , 1985) is that w o r d - i n t e r n a l p h o n o l o g i c a l d o m a i n s are
u n p r e d i c t a b l y cyclic or n o n c y c l i c . Inkelas p r o p o s e s that cyclicity b e f o r m a l i z e d
as the a u t o m a t i c application of p h o n o l o g i c a l rules u p o n c o n s t r u c t i o n of a n e w
P - C O N S T I T U E N T ( n o n m e d i c a l p h o n o l o g i c a l d o m a i n ) , a p r o p o s a l w h i c h is directly
s u p p o r t e d b y C o n n ' s ( 1 9 8 9 ) cyclic analysis of stress a s s i g n m e n t in I n d o n e s i a n .
I n k e l a s ' s m o d e l a c c o u n t s for the e x i s t e n c e of n o n c y c l i c p h e n o m e n a (or the n o n -
application of cyclic rules) in t w o p l a c e s within the lexicon. First, cyclic rules m a y
fail to apply p r i o r to the first p h o n o l o g i c a l c y c l e , either b e c a u s e of a universal
prohibition against cyclic rules a p p l y i n g to b o u n d r o o t s or b e c a u s e there is c r o s s -
linguistic variation in the o r d e r i n g of affixation w i t h respect to p-structure c o n -
struction. Secondly, a l a n g u a g e m a y h a v e a set of n o n c y c l i c rules w h i c h a p p l y at
the e n d of the lexical c o m p o n e n t (a la Booij a n d R u b a c h , 1987). Inkelas suggests
that the e x i s t e n c e of the latter m a y follow from w h e t h e r or not the l a n g u a g e h a s
m o r p h e m e s w h i c h b e l o n g to the last lexical p h o n o l o g i c a l d o m a i n . T h e s e relatively
n e w p r e d i c t i o n s should b e further e x a m i n e d b u t for the m o m e n t a p p e a r to offer
general s u p p o r t for interactionism.
It s h o u l d b e noted that even the h y p o t h e s i s that p h o n o l o g i c a l rules apply cycli-
cally h a s b e e n controversial t h r o u g h o u t the history of g e n e r a t i v e p h o n o l o g y ,
6 Ellen M. Kaisse and Sharon Hargus

partly b e c a u s e cyclic a n a l y s e s of individual l a n g u a g e s tend to b e b a s e d o n abstract

m o r p h o l o g i c a l a n d / o r p h o n o l o g i c a l analysis, a n d partly b e c a u s e theoretical refine-
m e n t s h a v e r e n d e r e d individual a n a l y s e s obsolete. T h e c y c l e in S p a n i s h p r o v i d e s
a g o o d e x a m p l e of the latter. B r a m e ( 1 9 7 4 ) a r g u e d that stress w a s a s s i g n e d cycli-
cally in S p a n i s h on the b a s i s of the stress-conditioned raising of e to i in a d v e r b s
d e r i v e d from the p a s t participles of s e c o n d conjugation verbs: conocidamente <
conoc-e-r ' k n o w i n g l y ' , debidamente < deb-e-r ' j u s t l y ' . In order for stresses
to a p p e a r in the p r o p e r locations, stress a s s i g n m e n t m u s t p r e c e d e suffixation of
-mente; that is, stress m u s t b e a s s i g n e d cyclically. T h e need for the cycle dis-
a p p e a r s , however, if -mente is a n a l y z e d as constituting a separate p h o n o l o g i c a l
d o m a i n , as d o e s S u n e r ( 1 9 7 5 ) (cf. also Lantolf, 1977), w h o p r o v i d e s m o r p h o l o g i -
cal a n d p h o n o l o g i c a l e v i d e n c e for this analysis.

3. W O R D - I N T E R N A L P H O N O L O G I C A L D O M A I N S

A s m e n t i o n e d a b o v e , lexical p h o n o l o g i s t s generally r e c o g n i z e that, in s o m e

l a n g u a g e s , it is n e c e s s a r y to posit o n e or m o r e p h o n o l o g i c a l d o m a i n s w h i c h are
smaller than the p h o n o l o g i c a l w o r d (the latter b e i n g a d o m a i n w h i c h m a y or m a y
not e x c l u d e clitics; see Z e e , this v o l u m e ; Z e e a n d Inkelas, 1991). T h e r e h a v e b e e n
t w o recent a n d related d e v e l o p m e n t s in the t h e o r y of word-internal p h o n o l o g i c a l
d o m a i n s w h i c h are reflected in m a n y of the p a p e r s in this v o l u m e . T h e first c o n -
c e r n s the r e c o g n i t i o n a n d e x p l a n a t i o n of the fact that word-internal p h o n o l o g i c a l
d o m a i n s m a y exhibit v a r y i n g d e g r e e s of i n d e p e n d e n c e from m o r p h o l o g i c a l repre-
sentations. T h e s e c o n d c o n c e r n s the viability of the level o r d e r i n g h y p o t h e s i s .
D u e to w o r k b y Booij ( 1 9 8 5 ) , Booij a n d R u b a c h ( 1 9 8 4 ) , Inkelas ( 1 9 9 0 , this
v o l u m e ) , a n d S p r o a t ( 1 9 8 5 ) , there is n o w s o m e t h i n g of a c o n s e n s u s that w o r d -
internal p h o n o l o g i c a l d o m a i n s ( d u b b e d P-STRUCTURE b y Inkelas) m a y b e derived
from, b u t n e e d n o t b e i s o m o r p h i c with, m o r p h o l o g i c a l or m o r p h o s y n t a c t i c struc-
ture. Polish, as a n a l y z e d b y Booij a n d R u b a c h ( 1 9 8 4 ) , p r o v i d e s an e x a m p l e of this
n o n i s o m o r p h i s m . T h e s e authors a r g u e that [rozijimitsa] 'truce m a k e r ' has the fol-
l o w i n g p h o n o l o g i c a l a n d m o r p h o l o g i c a l structures.

(2) [[[rozi [jim]] its] a] m o r p h o l o g i c a l representation

[rozi [[[jim] its] a]] p h o n o l o g i c a l representation

T h e m o r p h o l o g i c a l l y m o t i v a t e d structure incorporates their observation that " p r e -

fixed stems may, w h i l e inflected f o r m s m a y not, function as b a s e s for further
w o r d - f o r m a t i o n " (p. 19). Yet the cyclic derivation required for the correct appli-
cation of the p h o n o l o g i c a l rules of Yer L o w e r and Yer Deletion is o n e in w h i c h
the prefix / r o z i - / m u s t b e l o n g to the last cycle. Booij a n d R u b a c h p r o p o s e that
both m o r p h o l o g i c a l prefixes and inflected stems in Polish are p h o n o l o g i c a l w o r d s
Introduction 7

( m e m b e r s of t h e c a t e g o r y mot), a n d that t h e m o r p h o l o g i c a l w o r d c o r r e s p o n d s to
a h i g h e r - o r d e r p h o n o l o g i c a l constituent, t h e p h o n o l o g i c a l c o m p o u n d , or mot':

(3) mot'

mot mot

[rozi] [[Dim] its] a]

S u c h data, o n c e c o n s i d e r e d u n u s u a l a n d p u z z l i n g , h a v e t u r n e d u p in a variety of
l a n g u a g e s , a n d a n a l y s e s similar t o that p r o p o s e d b y Booij a n d R u b a c h for Polish
can b e found in N e s p o r a n d Vogel ( 1 9 8 6 ) , C o h n ( 1 9 8 9 ) , K a n g ( 1 9 9 2 ) , a n d Inkelas
( 1 9 9 0 , this v o l u m e ) . F o r e x a m p l e , K a n g ( 1 9 9 2 ) s h o w s that A n n ' s ( 1 9 8 5 ) four-level
analysis of the lexical p h o n o l o g y of K o r e a n , w h i c h required the l o o p , yields to a
p r o s o d i c analysis with n o e x t r a o r d i n a r y theoretical devices a n d only o n e w o r d -
internal p h o n o l o g i c a l d o m a i n o n c e t h e i n d e p e n d e n c e of m o r p h o l o g i c a l r e p r e s e n -
tations a n d p-structure is r e c o g n i z e d . F o l l o w i n g Inkelas ( 1 9 9 0 ) , the version of
lexical p h o n o l o g y w h i c h r e c o g n i z e s this i n d e p e n d e n c e m a y b e called p r o s o d i c
lexical p h o n o l o g y ( P L P ) .
T h e level o r d e r i n g h y p o t h e s i s ( L O H ) , o n c e c o n s i d e r e d a tenet of classical lexi-
cal p h o n o l o g y , a t t e m p t e d to explain an o b s e r v e d correlation b e t w e e n affix o r d e r
a n d p h o n o l o g i c a l rule d o m a i n s b y p o s i t i n g a strict o r d e r i n g b e t w e e n affixes w h i c h
b e l o n g to different p h o n o l o g i c a l rule d o m a i n s ( t e r m e d LEVELS or STRATA in the
theory of level o r d e r i n g ) . Yet in recent years a n u m b e r of authors h a v e c h a l l e n g e d
the descriptive a d e q u a c y of the correlation in a n u m b e r of l a n g u a g e s (see, e.g.,
Aronoff a n d Sridhar, 1 9 8 7 ; H a r g u s , 1 9 8 8 ; M o h a n a n , 1986; Strauss, 1982). F o r
e x a m p l e , Aronoff a n d S r i d h a r n o t e for E n g l i s h that t h e (stress-affecting) level 1
affix -ity attaches q u i t e p r o d u c t i v e l y to w o r d s c o n t a i n i n g t h e (stress-neutral)
level 2 affix -able, c o u n t e r e x e m p l i f y i n g the L O H . M o r e o v e r , S p e n c e r ( 1 9 8 9 ) h a s
d e m o n s t r a t e d that m u c h of t h e w o r k attributed t o t h e L O H in t h e r e g u l a t i o n of
affix o r d e r in E n g l i s h m u s t b e a c c o m p l i s h e d b y m o r e restrictive m o r p h o l o g i c a l
s u b c a t e g o r i z a t i o n frames. In PLP, with its l o o s e r c o n n e c t i o n s b e t w e e n m o r p h o -
logical a n d p h o n o l o g i c a l r e p r e s e n t a t i o n s , there is n o r o l e for t h e L O H . Instead,
like word- external p h o n o l o g i c a l d o m a i n s , w o r d - i n t e r n a l p-structure in P L P is a s -
s u m e d to b e nested, r e g u l a t e d b y t h e strict layer h y p o t h e s i s ( S L H ) (Selkirk, 1984;
N e s p o r a n d Vogel, 1986), in w h i c h s m a l l e r p h o n o l o g i c a l d o m a i n s are strictly c o n -
tained within larger o n e s .
It is legitimate to a s k w h e t h e r or n o t the S L H is simply a reincarnation of the
L O H . W h i l e t h e o r d e r i n g restrictions i m p o s e d b y t h e L O H a n d t h e S L H are b o t h
hierarchical, P L P a l l o w s that affixes n o t b e u n d e r l y i n g l y specified for d o m a i n b u t
m a y instead CLITICIZE to the p h o n o l o g i c a l d o m a i n of an adjacent m o r p h e m e ,
t h e r e b y a c c o m m o d a t i n g w h a t w e r e p r e v i o u s l y c o n s i d e r e d to b e level o r d e r i n g v i o -
lations. A n Aronoff a n d S r i d h a r - s t y l e analysis of E n g l i s h w o r d s like ungrammat-
8 Ellen M. Kaisse and Sharon Hargus

icality a n d turnability w o u l d h a v e the following c o m p o n e n t s . First, affixes within

the d o m a i n of the E n g l i s h Stress R u l e , like -ity, w o u l d b e u n d e r l y i n g l y specified
as stem-level or level 1, w h e r e a s affixes like -able- and un- w o u l d b e unspecified
for d o m a i n a s s i g n m e n t . S e c o n d , s o m e sort of a l g o r i t h m w o u l d be required for
c o n v e r t i n g m o r p h o l o g i c a l r e p r e s e n t a t i o n s ( w h i c h include u n d e r l y i n g affixal d o -
m a i n a s s i g n m e n t , if a n y ) into p h o n o l o g i c a l r e p r e s e n t a t i o n s . T h e following is a
p a r a p h r a s e of o n e such a l g o r i t h m p r o p o s e d b y Aronoff a n d Sridhar.

(4) T h e e d g e of a p h o n o l o g i c a l w o r d ( W ) is the e d g e of any e l e m e n t w h i c h is a

m e m b e r of a m a j o r lexical category, unless that e d g e is adjacent to a s t e m -
level affix.

T h i s a l g o r i t h m predicts t h e following m a p p i n g b e t w e e n m o r p h o l o g i c a l and p h o -

nological representations for w o r d s like ungrammaticality and turnability*

j | stem
W W w

N [AdjM t u r n
]v a b l e
l Adj itylN
| j stem
w w w

T h e Aronoff a n d S r i d h a r - s t y l e a l g o r i t h m predicts that n o W - b o u n d a r y can b e

inserted b e t w e e n grammatical a n d -ity, even t h o u g h grammatical is a m e m b e r of
a major lexical category, b e c a u s e -ity is a stem-level affix. Similarly, -able and
-ity are correctly p r e d i c t e d to f o r m a p h o n o l o g i c a l unit. N o t e further that the p r e -
c e d i n g derived p h o n o l o g i c a l r e p r e s e n t a t i o n s are in accord with the S L H , w h e r e a s
representations in w h i c h -able a n d un- are c o n s i d e r e d level 2 m o r p h e m e s violate
the L O H .

(6) [[un[grammatical]] ity]j 2

[[[turn] a b i l ] i t y ] !2

To s u m m a r i z e , then, the relationship b e t w e e n m o r p h o l o g i c a l and p h o n o l o g i c a l

representations is less restrictive in several w a y s : structures such as the p h o n o -
logical w o r d a n d the m o r p h o l o g i c a l w o r d are not required to b e identical, and
affixes m a y h a v e e m p t y d o m a i n specifications. O n e current, u n r e s o l v e d issue in
P L P c o n c e r n s the extent to w h i c h p-structure is predictable. Inkelas ( 1 9 9 0 , this
v o l u m e ) a n d B o r o w s k y (this v o l u m e ) a d o p t the traditional v i e w that rule d o m a i n s
are lexically specified information i n c l u d e d in the subcategorization frames of
affixes. T h e alternative position, b a s e d on Selkirk ( 1 9 8 6 ) , is that word-internal
p-structure is basically p r e d i c t a b l e , a position a d o p t e d b y K a n g ( 1 9 9 2 ) and b y
R i c e (this v o l u m e ) . K a n g ( 1 9 9 2 ) s h o w s that the p h o n o l o g i c a l w o r d in K o r e a n can
b e d e t e r m i n e d from the left e d g e of a m o r p h o l o g i c a l constituent, thereby correctly
Introduction 9

p r e d i c t i n g that the left b r a n c h e s of c o m p o u n d s a n d prefixed w o r d s are i n d e p e n -

d e n t p h o n o l o g i c a l constituents. R i c e (this v o l u m e ) p r o p o s e s three w o r d - i n t e r n a l
p h o n o l o g i c a l rule d o m a i n s for Slave, an A t h a b a s k a n l a n g u a g e , the SMALL W O R D ,
W O R D , a n d PHONOLOGICAL PHRASE, the latter b e i n g the p h o n o l o g i c a l d o m a i n usu-
ally t h o u g h t of as the entire " v e r b w o r d " in p r e v i o u s a n a l y s e s of A t h a b a s k a n
l a n g u a g e s . R i c e s u g g e s t s that the w o r d - i n t e r n a l m o r p h o l o g i c a l structure of verbs
in Slave is basically syntactic, rather than m o r p h o l o g i c a l , a n d that Slave h a s in
fact very little true m o r p h o l o g y . B a s i n g h e r analysis o n earlier w o r k b y S p e a s
( 1 9 9 0 ) , R i c e a r g u e s that this m o r p h o s y n t a c t i c constituent structure is not deter-
m i n e d b y p h r a s e structure rule b u t is d e r i v e d from universal n o t i o n s of syntactic
SCOPE. T h u s , in R i c e ' s analysis, the e d g e s of p h o n o l o g i c a l d o m a i n s follow from
the particular m o r p h o s y n t a c t i c structure a d o p t e d .

4. T H E S T R I C T C Y C L E C O N D I T I O N : I S I T D E R I V A B L E
FROM UNDERSPECIFICATION?

T h e strict c y c l e c o n d i t i o n ( S C C ) , a d e v e l o p m e n t of p r e v i o u s c o n c e p t s intro-
d u c e d b y K i p a r s k y over t w o d e c a d e s , such as the (revised) alternation c o n d i t i o n
( R A C ) , is a c o n s t r a i n t o n the application of certain rules w h i c h p r e v e n t s t h e m from
a p p l y i n g w i t h i n a m o r p h e m e or from a p p l y i n g in e n v i r o n m e n t s w h i c h w e r e al-
r e a d y available o n a p r e v i o u s cycle. L e x i c a l p h o n o l o g i s t s generally a g r e e that
there is a r o b u s t set of p h e n o m e n a w h i c h m o t i v a t e s o m e t h i n g r o u g h l y like t h e
S C C or t h e R A C . In his c o n t r i b u t i o n to this v o l u m e , K i p a r s k y a r g u e s that the S C C
suffers from e m p i r i c a l difficulties w h i l e the R A C suffers from theoretical w e a k -
n e s s e s a n d a r g u e s instead that the p h e n o m e n a they c o v e r c a n b e d e r i v e d from
s e g m e n t a l underspecification. T h e applicability of the S C C to a rule w a s a clear
indication of its lexical status in classical lexical p h o n o l o g y , b u t this h y p o t h e s i s is
u n d e r attack from all directions, i n c l u d i n g that of its originator, for K i p a r s k y n o w
a r g u e s that true lexical rules, word-level rules, a n d postlexical rules m a y all e x -
hibit or fail to exhibit such effects. Iverson, in his c o n t r i b u t i o n , a r g u e s for a return
to the position that strict c y c l e effects are a characteristic s i m p l y of neutralization
rules, w h i l e n o n n e u t r a l i z i n g rules m a y fail to exhibit t h e m even if lexical. B o t h
K i p a r s k y ( 1 9 8 5 ) a n d B o r o w s k y ( 1 9 8 6 , this v o l u m e ) h a v e a r g u e d that w h a t e v e r t h e
precise f o r m of the c o n d i t i o n b l o c k i n g rule application in n o n d e r i v e d e n v i r o n -
m e n t s , it s h o u l d turn off at the last level of the lexicon, the w o r d level, rather than
postlexically. Finally, P o s e r s u g g e s t s in this v o l u m e that the d e r i v e d e n v i r o n m e n t
condition s h o u l d b e d i v i d e d into constraints o n a r e q u i r e m e n t for c o n c a t e n a t i o n
versus a r e q u i r e m e n t of p r e v i o u s application of p h o n o l o g i c a l rule.
K i p a r s k y ( 1 9 8 5 , this v o l u m e ) h a s i n t r o d u c e d a sort of c o n t e x t u a l underspecifi-
cation w h i c h h e uses in his a t t e m p t to a c c o u n t for n o n d e r i v e d e n v i r o n m e n t effects.
10 Ellen M. Kaisse and Sharon Hargus

S i n c e this version of underspecification figures in Iverson's article as well, and

b e c a u s e it m a y b e less familiar to o u r r e a d e r s , w e s u m m a r i z e it briefly here. W e
h a v e m a d e u p a s i m p l e e x a m p l e below, but t h e e x p l a n a t i o n is transferable to m o r e
c o m p l i c a t e d cases such as Trisyllabic S h o r t e n i n g in English, and Iverson's K o r e a n
Palatalization as well.
C o n s i d e r a l a n g u a g e w h i c h h a s n a s a l i z e d v o w e l s u n d e r l y i n g l y b u t also h a s a
natural sort of p h o n o l o g i c a l p r o c e s s w h i c h nasalizes v o w e l s before nasals. Vowel
N a s a l i z a t i o n is thus neutralizing or " s t r u c t u r e p r e s e r v i n g . " N o w a s s u m e that
Vowel N a s a l i z a t i o n s h o w s the d e r i v e d e n v i r o n m e n t effect, so that within a single
m o r p h e m e , it fails to a p p l y — t h e r e are m o n o m o r p h e m e s with a n o n n a s a l i z e d
vowel p r e c e d i n g a nasal.
Recall n o w K i p a r s k y ' s ( 1 9 7 1 ) d i s c u s s i o n of o p a q u e rules. P a r a p h r a s i n g , a rule
is o p a q u e to the extent that there are surface f o r m s w h i c h contradict it. C o n t r a d i c -
tion can take the form of strings w h e r e the rule should h a v e applied but has not,
a n d cases w h e r e the rule should not h a v e applied, but h a s . M o r e technically, as
K i p a r s k y p u t s it, a rule P, A —» B/C D, is o p a q u e to the extent that there are
strings CAD in its o u t p u t (in other w o r d s , u n d e r a p p l i c a t i o n ) ; or strings CBD not
d e r i v e d from P.
W e can n o w see that a neutralizing rule subject to the derived e n v i r o n m e n t
condition results in both k i n d s of opacity. W e h a v e strings that l o o k like they
should h a v e u n d e r g o n e the rule b u t d o not ( a p p a r e n t u n d e r a p p l i c a t i o n ) , in the form
of m o r p h e m e - i n t e r n a l oral v o w e l plus nasal s e q u e n c e s . A n d w e h a v e strings that
l o o k like they h a v e u n d e r g o n e the rule b u t shouldn't have, in the form of under-
lying nasalized v o w e l s followed b y n o n n a s a l c o n s o n a n t s (apparent overapplica-
tion). O p a q u e f o r m s are m a r k e d . K i p a r s k y ' s m e t h o d of underspecifying therefore
r e s e r v e s t h e u n m a r k e d value for v o w e l s w h i c h u n d e r l y i n g l y a c c o r d with w h a t the
rules of the l a n g u a g e will p r o d u c e . T h a t is, the m o r e natural, m o r p h e m e - i n t e r n a l
s e q u e n c e s of nasal v o w e l p l u s nasal c o n s o n a n t a n d of n o n n a s a l vowel plus non-
nasal c o n s o n a n t will b e u n m a r k e d . T h o s e o p a q u e cases of nasal vowel followed
b y n o n n a s a l c o n s o n a n t a n d n o n n a s a l vowel followed b y nasal c o n s o n a n t will b e
marked.

(7) Nonnasalized vowels: [ - nasal] before nasals in the s a m e m o r p h e m e ;

[Onasal] e l s e w h e r e
Nasalized vowels: [ + n a s a l ] before n o n n a s a l s ;
[Onasal] e l s e w h e r e

T h u s the c a s e of [ V N ] within a m o r p h e m e , the c a s e w h e r e the n o n d e r i v e d envi-

r o n m e n t effect a p p e a r s , is achieved b y h a v i n g the v o w e l prespecified for non-
nasality. T h e 'overapplication' c a s e , [ V T ] , is d e r i v e d b y prelinking the vowel to a
[ + n a s a l ] specification. T h e m o r p h e m e - i n t e r n a l and -external cases w h e r e a vowel
nasalizes before a nasal are b o t h h a n d l e d by Vowel Nasalization filling in the
feature o n the u n d e r l y i n g [Onasal] vowel. Implicit in this system is the claim that
Introduction 11

there is n o p o s s i b l e representational contrast b e t w e e n m o r p h e m e - i n t e r n a l [ V N ]

d e r i v i n g from an u n d e r l y i n g nasal v o w e l a n d an u n d e r l y i n g n o n n a s a l v o w e l . T h e
e n v i r o n m e n t is o n e in w h i c h the vowel is s i m p l y [Onasal]. T h u s w e r e d i s c o v e r the
R A C , w h i c h p r e v e n t e d a neutralization rule from a p p l y i n g to every instance of a
m o r p h e m e . B y the R A C , Sanskrit s before m o r p h e m e - i n t e r n a l r u, k, or / h a d to
}

b e a real u n d e r l y i n g s a n d o u r V before nasal h a d to b e a real u n d e r l y i n g nasalized

vowel. N o w the s a n d the V m u s t b e u n m a r k e d for retroflexion and nasality re-
spectively a n d t h u s c a n n o t contrast w i t h an u n d e r l y i n g / s / or IVI w h i c h u n d e r g o e s
the ruki rule or N a s a l i z a t i o n .
T h e final step in K i p a r s k y ' s disposal of the S C C is to say that rules delinking
features, that is, s t r u c t u r e - c h a n g i n g rules, are m a r k e d . R u l e s b u i l d i n g structure,
that is, filling in features such as [ + n a s a l ] or [ — nasal] on previously u n d e r s p e c i -
fied s e g m e n t s , are u n m a r k e d . T h u s , unless the l a n g u a g e in o u r nasalization ex-
a m p l e h a s an u n u s u a l , specific rule r e m o v i n g prespecified linkages, w e should not
e x p e c t u n d e r l y i n g p r e n a s a l , m o r p h e m e - i n t e r n a l oral v o w e l s to nasalize, since they
are already linked to [ — n a s a l ] . H o w e v e r , a m o r p h e m e - f i n a l oral v o w e l will h a v e
n o specification for nasality. T h u s w h e n it is prefixed to a nasal-initial m o r p h e m e ,
spreading can o c c u r w i t h o u t any m a r k e d d e l i n k i n g rule b e i n g p r e s e n t in the lan-
g u a g e , a n d w e achieve the d e r i v e d e n v i r o n m e n t effect.

5. S T R U C T U R E P R E S E R V A T I O N

T h e basic c o n c e p t of structure p r e s e r v a t i o n is a s i m p l e o n e , t h o u g h w e shall see

that its p r o p e r technical instantiation m a y b e a n y t h i n g b u t that. T h e general idea
is that the prototypical lexical rule p r e s e r v e s the basic u n d e r l y i n g s e g m e n t and
tonal inventory of the l a n g u a g e and the basic a r r a n g e m e n t of strings of s e g m e n t s
as well. T h e r e will b e a lexical v o c a b u l a r y of s e g m e n t s a n d tones, w h i c h d o e s n o t
alter d u r i n g the derivation of w o r d s , a n d an e n r i c h e d array of a l l o p h o n e s , w h i c h
is derived postlexically.
W e h a v e already seen that B o r o w s k y and K i p a r s k y (1985) b o t h p r o p o s e that
structure p r e s e r v a t i o n turns off a bit earlier than this simple d i c h o t o m y of lexical
versus postlexical rules w o u l d predict. In w o r k in this v o l u m e a n d in h e r 1986
dissertation, B o r o w s k y suggests that n e w a l l o p h o n e s can b e created within the
lexicon, at the w o r d level. E x t e n d i n g the p o s s i b l e relevance of structure preser-
vation in the other direction, H y m a n (this v o l u m e ) c o n s i d e r s the effects of struc-
ture preservation in postlexical applications of a r i g h t w a r d H i g h T o n e S p r e a d in
the G u r l a n g u a g e D a g b a n i . H e c o n c l u d e s that there are t w o very similar H i g h
T o n e S p r e a d rules. H T S - 1 o b e y s structure preservation, for it d o e s not create c o n -
tour tones, w h i c h are o t h e r w i s e p r o h i b i t e d in the l a n g u a g e . H T S - 2 , on the other
hand, d o e s not obey structure p r e s e r v a t i o n — i t creates c o n t o u r tones. T h e straight-
12 Ellen M. Kaisse and Sharon Hargus

forward expectation in s t a n d a r d lexical p h o n o l o g y w o u l d h a v e b e e n that H T S - 1

applies in an earlier s t r a t u m than H T S - 2 . O n e m i g h t b e lexical, the other postlex-
ical; o n e m i g h t b e an early lexical rule, the other a word-level rule; or, if there are
t w o postlexical levels, as a r g u e d b y K a i s s e ( 1 9 8 5 ) a n d b y M o h a n a n ( 1 9 8 6 ) , w e
m i g h t e x p e c t that H T S - 1 a p p l i e d at the earlier of the postlexical strata, called P I
b y K a i s s e , w h e r e lexical characteristics persist.
T h e r u b is that, as H y m a n d e m o n s t r a t e s , BOTH H T S - 1 and H T S - 2 m u s t b e
utterance-level rules. In a theory like K a i s s e ' s ( 1 9 8 5 , 1990), the first postlexical
level, P I , should c o r r e s p o n d loosely to the small p r o s o d i c d o m a i n k n o w n as
the p h o n o l o g i c a l p h r a s e . R u l e s restricted to the p h o n o l o g i c a l p h r a s e will apply,
roughly, b e t w e e n the h e a d of a syntactic p h r a s e and its various a r g u m e n t s . S u c h
rules, a p p l y i n g i m m e d i a t e l y after lexical rules, should b e e x p e c t e d to exhibit m o r e
lexical characteristics than later, so-called P 2 rules. B u t utterance-level rules, cor-
r e s p o n d i n g to P 2 rules, apply b e t w e e n all w o r d s in the s a m e s e n t e n c e , or even
b e t w e e n closely b o u n d s e n t e n c e s . T h u s w e are forced to the c o n c l u s i o n , if H y m a n
is correct, that even within a single c o m p o n e n t (in this case, the last postlexical
stratum, P 2 ) , structure p r e s e r v a t i o n m a y arbitrarily b e or n o t b e a characteristic of
a rule. Typically, H y m a n c o n c l u d e s , structure preservation will c h a r a c t e r i z e the
application of lexical rules, w h i l e failure to p r e s e r v e structure will b e a h a l l m a r k
of postlexical o n e s , as in the standard theory. B u t this is only a tendency. Since an
utterance-level rule is clearly P 2 , w e d o n o t h a v e t w o separate c o m p o n e n t s h e r e to
segregate the rules into. T h e e x a m p l e is w e a k l y consistent w i t h K a i s s e ' s and K i -
p a r s k y ' s theories in the sense that in these theories, n o postlexical rule is REQUIRED
to create novel s e g m e n t s or s e q u e n c e s , so that H T S - 1 's inability to create a c o n t o u r
tone d o e s n o t m e a n it c a n n o t b e a P 2 rule. T h i n g s c o u l d also b e w o r s e in that at
least H T S - 1 p r e c e d e s H T S - 2 , b u t the p h e n o m e n o n is p u z z l i n g n o n e t h e l e s s , for it
s h o w s that structure p r e s e r v a t i o n m a y turn off arbitrarily, rather than at a well-
defined level p r o v i d e d b y the lexical p h o n o l o g y .
R e c e n t w o r k b y M c F a r l a n d a n d P i e r r e h u m b e r t ( 1 9 9 1 ) h a s the potential to m a k e
irrelevant m a n y i n v o c a t i o n s of structure preservation as a b l o c k to the application
of lexical rules. W e d i s c u s s their p r o p o s a l h e r e , for H y m a n ' s results m a y turn on
w h e t h e r w e accept M c F a r l a n d a n d P i e r r e h u m b e r t ' s interpretation. T h e s e authors
c o n t e n d that the p r o p e r interpretation of structure preservation, as spelled out in
K i p a r s k y ( 1 9 8 5 ) , d o e s n o t rule out t h e lexical derivation of s e g m e n t s w h i c h are
novel b u t derived b y feature s p r e a d i n g . In such cases, the n u m b e r and configura-
tion of association lines m a y b e critically different from w h a t a lexical m a r k i n g
condition rules out. A strict application of H a y e s ' s ( 1 9 8 6 ) linking constraint and
of the t h e o r y of feature g e o m e t r y s h o w s that rules s p r e a d i n g a feature from o n e
s e g m e n t to a n o t h e r create structures with d o u b l e linkages of the assimilated
feature. T h e s e d o u b l e linkings d o not m a t c h the single linkages of lexical c o n d i -
tions intended to rule out the association of a s e g m e n t t y p e with a certain feature.
C o n s i d e r K i p a r s k y ' s ( 1 9 8 5 ) o w n use of structure preservation to prevent the lexi-
Introduction 13

cal participation of R u s s i a n s o n o r a n t s in voicing assimilation. M a c F a r l a n d and

P i e r r e h u m b e r t p o i n t out that the lexical prohibition against (8)

(8) [+son]

I
[avoice]

d o e s n o t rule out t h e lexical spread of v o i c i n g o n t o a sonorant, b e c a u s e in a form

d e r i v e d b y s p r e a d i n g , [voice] is linked to T W O association lines, a n d lines of as-
sociation m u s t b e interpreted exhaustively.

(9) [+son] [-son]

[—voice]

T h e r e l e v a n c e of M c F a r l a n d a n d P i e r r e h u m b e r t ' s m e t i c u l o u s interpretation of
structure p r e s e r v a t i o n to H y m a n ' s D a g b a n i c a s e m a y n o w b e a p p a r e n t to the
reader. T h e H i g h T o n e S p r e a d rules, as their n a m e s imply, of c o u r s e c r e a t e linked
structures. If the lexical constraint in D a g b a n i rules out c o n t o u r tones as in (10);
tbu = ' t o n e - b e a r i n g u n i t ' ,

(10) * Tone Tone

tbu

it will not rule out the structure d e r i v e d b y s p r e a d i n g (11),

(11) H L

M
tbu tbu

b e c a u s e in this latter structure, t h e first t o n e is d o u b l y linked. It is only singly

linked in the lexical constraint. B y the strict interpretation of structure p r e s e r v a -
tion, therefore, n o t o n e spreading rules can b e p r o h i b i t e d from a p p l y i n g in D a g -
b a n i m e r e l y as a c o n s e q u e n c e of the fact that D a g b a n i h a s n o u n d e r l y i n g c o n t o u r
t o n e s . W e will h a v e to h a v e a specific lexical c o n d i t i o n ruling out t o n e - b e a r i n g
units with m o r e than o n e t o n e attached to t h e m , r e g a r d l e s s of w h e t h e r t h o s e tones
h a v e additional lines of association.

(12) * T T

AA
(tbu) tbu (tbu)

In the a b s e n c e of t h e r e l e v a n c e of structure p r e s e r v a t i o n to s p r e a d i n g rules, such

an u n p a l a t a b l e c o n d i t i o n is i n d e e d n e e d e d , since there are lexical rules w h i c h
w o u l d o t h e r w i s e p r o d u c e c o n t o u r tones.
T h e strict interpretation of the applicability of structure p r e s e r v a t i o n gets us out
14 Ellen M. Kaisse and Sharon Hargus

of the difficulty H y m a n ' s c a s e raises, t h o u g h at significant cost, b o t h to the sim-

plicity of the g r a m m a r of D a g b a n i a n d p e r h a p s to p h o n o l o g i c a l theory in general.
R a t h e r than w o r r y i n g w h y structure p r e s e r v a t i o n turns off at an arbitrary point in
D a g b a n i g r a m m a r , w e will say it w a s never r e l e v a n t — s p r e a d i n g rules of this form
are n o t b l o c k e d b y structure p r e s e r v a t i o n . Instead w e i m p o s e a condition on d e -
rived c o n t o u r tones that d o e s n o t s t e m organically from D a g b a n i ' s lack of under-
lying c o n t o u r t o n e s . A n d w e m u s t still stipulate that that condition is applicable to
H T S - 1 but not to H T S - 2 .
M c F a r l a n d and P i e r r e h u m b e r t c o n c l u d e their article b y p o i n t i n g out that the
strict a n d consistent interpretation of structure p r e s e r v a t i o n will rule out m a n y
o t h e r w i s e a p p e a l i n g analyses w h i c h used that p r i n c i p l e to prevent assimilations
within the lexical p h o n o l o g y . T h e s e cases i n c l u d e s o m e of K i p a r s k y ' s o w n , such
as R u s s i a n Voicing A s s i m i l a t i o n a n d F i n n i s h Vowel H a r m o n y , as well as n u m e r -
o u s others in the literature of lexical p h o n o l o g y . W e agree with M c F a r l a n d and
P i e r r e h u m b e r t ' s a s s e s s m e n t : p h o n o l o g i s t s m u s t either revise structure preserva-
tion to allow it to rule out t h o s e assimilations that it a p p e a r e d to prevent so ap-
pealingly, or they m u s t consistently interpret structure preservation in the light of
the results of feature g e o m e t r y a n d the linking constraint a n d r e w o r k the analyses
that incorrectly i n v o k e d it. W e c a n n o t h e l p b u t feel that the m o s t r i g o r o u s inter-
pretation of structure p r e s e r v a t i o n leaves it t o o w e a k to deal w i t h the very cases it
w a s d e s i g n e d to explain.

6. A P P L Y I N G L E X I C A L P H O N O L O G Y T O D I A C H R O N I C C H A N G E

T h r e e c o n t r i b u t i o n s to this v o l u m e , those of Dresher, K a i s s e , and Z e e , deal not

so m u c h with controversies within the t h e o r y as with the uses of its basic p r e m i s e s
to elucidate d i a c h r o n i c c h a n g e . T h e addition, generalization, and m o v e m e n t of
rules h a s long b e e n the b a s i s for d e s c r i b i n g the b u l k of historical c h a n g e s within
generative p h o n o l o g y . B e c a u s e lexical p h o n o l o g y h a s d e v e l o p e d a sophisticated
v i e w of the position o r d o m a i n of rules in a g r a m m a r , it allows o n e to u n d e r s t a n d
the p r o g r e s s of a rule over t i m e in a m u c h m o r e detailed and revealing fashion.
B o t h Kaisse a n d Z e e flesh out the i d e a that rules are a d d e d postlexically and
gradually m o v e u p t h r o u g h the levels. Z e e traces the d e v e l o p m e n t of a rule spread-
ing high tones leftward in S e r b o - C r o a t i a n . T h e g r a m m a t i c i z a t i o n of S p r e a d i n g is
seen b y Z e e as the w i t h d r a w a l of S p r e a d i n g from successive l o w e r (i.e., later)
levels of the g r a m m a r . Different dialects h a v e frozen the p r o g r e s s i o n of the rule at
various stages. In addition, c h a n g e s in the level at w h i c h p h o n o l o g i c a l w o r d s (the
mots discussed in Section 3) are created a c c o u n t s for another dialect divergence.
K a i s s e looks at rules w h i c h m a y h a v e g o n e even farther along the path of
c h a n g e . S h e suggests that o n c e rules h a v e b e c o m e lexical, a further possible path
Introduction 15

of d e v e l o p m e n t is for t h e m to again b e g i n s p r e a d i n g or m o v i n g their d o m a i n s to

later levels a n d into the postlexical c o m p o n e n t , w h e r e they h a v e a c c e s s to r e d u n -
d a n t features a n d t h u s a p p e a r to h a v e generalized. I n d e e d , she c l a i m s that it m a y
b e p o s s i b l e to d o a w a y altogether w i t h the notion of rule generalization as an
i n d e p e n d e n t m e c h a n i s m of historical c h a n g e a n d s i m p l y to d e r i v e generalization
from underspecification. Recall that K i p a r s k y p r o p o s e d to derive the S C C from
underspecification as well. It is a p p a r e n t that w h i l e n o o n e is yet able to agree on
exactly h o w underspecification w o r k s , it is a powerful tool.
D r e s h e r a r g u e s that a p p a r e n t s y n c h r o n i c a n d d i a c h r o n i c o r d e r i n g p r o b l e m s in-
v o l v e d in t h e analysis of the rule of A n g l i a n S m o o t h i n g c a n b e r e s o l v e d if the rule
w a s at s o m e p o i n t r e a n a l y z e d as a constraint h o l d i n g o v e r the lexicon but turning
off at the w o r d level. A s D r e s h e r r e m a r k s , the c h r o n o l o g i c a l status of S m o o t h i n g
w a s a l w a y s o b v i o u s from the O l d E n g l i s h texts, b u t earlier r e s e a r c h e r s d o u b t e d
the c h r o n o l o g y n o n e t h e l e s s , for their t h e o r i e s c o u l d n o t a c c o u n t for the facts.
A g a i n , the p o w e r of lexical p h o n o l o g y to r e s o l v e l o n g s t a n d i n g difficulties in dia-
c h r o n i c c h a n g e is m a d e evident.
T h e a r g u m e n t s of K a i s s e a n d Z e e rest in p a r t o n the notion that historical
c h a n g e is profitably v i e w e d n o t only as t h e p r o g r e s s i o n of rules from o n e c o m p o -
n e n t to another, b u t that it is p o s s i b l e for a rule to r e m a i n in the d o m a i n in w h i c h
it b e g a n w h i l e e x t e n d i n g its application to a n o t h e r d o m a i n or c o m p o n e n t . T h i s
v i e w of rule p r o g r e s s i o n is implicitly c h a l l e n g e d b y Iverson's c o n t r i b u t i o n to this
v o l u m e . I v e r s o n w i s h e s to return to the early versions of lexical p h o n o l o g y , in
w h i c h rules h a d well-defined d o m a i n s — t h e y b e l o n g e d to particular strata. In lexi-
cal p h o n o l o g y d a t i n g from 1984 a n d b e y o n d , K i p a r s k y ( 1 9 8 4 ) a r g u e s in fact that
the natural state of a rule is not to b e restricted to a particular c o m p o n e n t or stra-
t u m (the s t r o n g d o m a i n h y p o t h e s i s ) . R u l e s will apply at Level 1 a n d t h r o u g h o u t
the lexicon a n d will e x t e n d into the postlexical c o m p o n e n t as well, unless they are
t u r n e d off b y a specific s t a t e m e n t in the g r a m m a r . B u t t h e characteristics of the
e x t e n d e d rule will differ from c o m p o n e n t to c o m p o n e n t , in a c c o r d a n c e w i t h t h e
p r e c e p t s of the theory. K i p a r s k y ' s best e x a m p l e s , C a t a l a n N a s a l A s s i m i l a t i o n a n d
R u s s i a n D e v o i c i n g , are c h a l l e n g e d in Iverson's article. Booij a n d R u b a c h ( 1 9 8 7 )
also c h a l l e n g e the c h a n g e in the t h e o r y w h i c h allowed, i n d e e d e x p e c t e d , rules to
apply in several strata a n d c o m p o n e n t s ; this r e m a i n s t h e n a controversial modifi-
cation. Historical studies should p r o v i d e a g o o d a v e n u e for r e s o l v i n g the q u e s t i o n
of the u n m a r k e d e x t e n s i o n of a rule's d o m a i n of application.

7. R U L E T Y P O L O G Y

A m o n g the m o r e c o m p e l l i n g h y p o t h e s e s of the early lexical p h o n o l o g y m o d e l

w a s the p r o p o s a l that p h o n o l o g i c a l a n d m o r p h o l o g i c a l rules w h i c h b e l o n g e d to
16 Ellen M. Kaisse and Sharon Hargus

different c o m p o n e n t s of g r a m m a r h a d different characteristics. C o n s i d e r the s u m -

m a r y of such characteristics p r o v i d e d b y K i p a r s k y ( 1 9 8 3 ) .

(13) LEXICAL POSTLEXICAL

a. word-bounded not w o r d - b o u n d e d
b. access to w o r d - i n t e r n a l structure access to p h r a s e
a s s i g n e d at s a m e level only structure only
c. p r e c e d e all postlexical rules follow all lexical rules
d. cyclic apply o n c e
e. disjunctively o r d e r e d with respect conjunctively o r d e r e d
to other lexical rules with respect to
lexical rules
f. apply in derived e n v i r o n m e n t s apply across the b o a r d
g. structure-preserving not structure-preserving
h. apply to lexical c a t e g o r i e s only apply to all categories
i. may have exceptions automatic

W i t h nearly a d e c a d e of s u b s e q u e n t w o r k , w e n o w k n o w that m a n y of these char-

acteristics c a n n o t b e c o n s i d e r e d diagnostic of the lexical o r postlexical status of
a rule.
It h a s b e e n suggested, s o m e w h a t controversially ( H a y e s , 1990; b u t cf. Kaisse,
1990), that not all lexical rules are w o r d - b o u n d e d (13a) or even restricted to infor-
m a t i o n p r o v i d e d within their o w n p h o n o l o g i c a l d o m a i n (13b) (as seen in M a l t e s e ;
O d d e n , this v o l u m e ) . A s d i s c u s s e d in Section 2.2, it h a s b e e n s u g g e s t e d that not
all lexical rules are cyclic ( 1 3 d ) , a n d the e x i s t e n c e of lexical postcyclic rules simi-
larly m a k e s characteristic (13f) p r o b l e m a t i c . In s o m e l a n g u a g e s , structure preser-
vation (13g) a p p e a r s to h o l d of postlexical rules ( H y m a n , this v o l u m e ) , w h e r e a s
in other l a n g u a g e s , s o m e rules w h i c h are clearly lexical (albeit word-level) m a y
not b e structure-preserving ( B o r o w s k y , this v o l u m e ) . In contradiction to (13h),
there are postlexical rules w h i c h m a y b e indirectly or directly sensitive to syntactic
c a t e g o r y (Kaisse, 1 9 8 5 ; N e s p o r a n d Vogel, 1986; a m o n g others), a n d s o m e post-
lexical rules h a v e e x c e p t i o n s ( K a i s s e , 1986) (13i). T h e role of characteristic (13e)
s e e m s to h a v e b e e n p r i m a r i l y to regulate w o r d formation, enforcing (via the else-
w h e r e condition) the b l o c k i n g of r e g u l a r derivatives like * borer ' o n e w h o b o r e s '
by the existence of s p e c i a l / e x c e p t i o n a l f o r m s like bore. However, since it w a s also
c l a i m e d that there w a s n o postlexical w o r d formation, the utility of this character-
istic in d i a g n o s i n g postlexical versus lexical m o r p h o l o g y d o e s not s e e m to us to
h a v e b e e n very great.
Yet it is striking that m a n y lexical a n d postlexical rules d o h a v e at least s o m e of
the characteristics given in (13) a b o v e . A c o n c e i v a b l e alternative to (13) w o u l d b e
to w e a k e n the o r d e r i n g prediction (13c) a n d to r e m o v e the labels on the c o l u m n s
so that these characteristics are n o t rigidly associated with particular c o m p o n e n t s
b u t that for e a c h erstwhile lexical characteristic, a rule in a particular l a n g u a g e
w i t h that p r o p e r t y p r e c e d e s a rule with the c o m p l e m e n t a r y postlexical character-
Introduction 17

istic. Alternatively, the list in (13) could b e v i e w e d as constituting the u n m a r k e d

set of characteristics of lexical a n d postlexical rules.
T o a suitably w e a k e n e d version of the list it m a y also b e p o s s i b l e to a d d a few
m o r e characteristics.
(13') "LEXICAL" "POSTLEXICAL"
j. n o t transferred to a s e c o n d transferable to L 2
language
k. o u t p u t s subject to lexical subject to N e o g r a m m a r i a n
diffusion sound change
1. apply categorically m a y have gradient outputs
W e t a k e these m o r e recently p r o p o s e d characteristics from R u b a c h ( 1 9 8 4 ) , Kipar-
sky ( 1 9 8 8 ) , a n d K i p a r s k y ( 1 9 8 5 ) , respectively.

8. A N I N V I T A T I O N

T o the r e a d e r w h o feels the n e e d of m o r e certainty or at least m o r e c o n c r e t e

detail than w e h a v e b e e n able to p r o v i d e in this introduction, w e suggest t u r n i n g
directly to the p a p e r s offered b y o u r c o n t r i b u t o r s . T h e y p r e s e n t their cases m o r e
c o g e n t l y than w e c o u l d h o p e to d o here.

NOTES

The order of names is arbitrary; authorship was shared equally. Our remarks owe much
1

to discussion by workshop participants; in particular, Geert Booij, Elan Dresher, Jim

Harris, Paul Kiparsky, and Bill Poser. We also thank Jurek Rubach and Donca Steriade for
helpful comments.
2
An earlier version of this article was presented by Harris at the 1990 Workshop on
Lexical Phonology.
3
Inkelas (1990, this volume) distinguishes metrical and nonmetrical phonological
domains.
4
We assume, following Cohn (1989) and Inkelas (1990), that the morphological struc-
ture is more basic.

REFERENCES

Ahn, S.-C. (1985). The Interplay of Phonology and Morphology in Korean. Doctoral dis-
sertation, University of Illinois, Urbana-Champaign.
Allen, M. (1978). Morphological Investigations. Doctoral dissertation, University of Con-
necticut, Storrs.
18 Ellen M. Kaisse and Sharon Hargus

Anderson, S. R. (1988). Inflection. In Theoretical Morphology (M. Hammond and M.

Noonan, eds.), pp. 2 3 - 4 3 . Academic Press, San Diego.
Aronoff, M. (1988). Head operations and strata in reduplication: a linear treatment. Year-
book of Morphology 1, 1-15.
Aronoff, M., and Sridhar, S. (1987). Morphological levels in English and Kannada. In
Rules and the Lexicon (E. Gussmann, ed.), pp. 9 - 2 2 . Katolickiergo Uniwersytetu
Lubelskiego, Lublin, Poland.
Booij, G. (1985). Coordination reduction in complex words: A case for prosodic phonol-
ogy. In Advances in Non-linear Phonology (H. van der Hulst and N. Smith, eds.),
pp. 143-160. Foris, Dordrecht.
Booij, G., and Rubach, J. (1984). Morphological and prosodic domains in lexical pho-
nology. Phonology Yearbook 1, 1 - 2 8 .
Booij, G., and Rubach, J. (1987). Postcyclic versus postlexical rules in lexical phonology.
Linguistic Inquiry 18, 1 - 4 4 .
Borowsky, T. (1986). Topics in the Lexical Phonology of English. Doctoral dissertation,
University of Massachusetts, Amherst.
Brame, M. (1974). The cycle in phonology: Stress in Palestinian, Maltese, and Spanish.
Linguistic Inquiry 5, 3 9 - 6 0 .
Chierchia, G. (1983/1986). Length, syllabification, and the phonological cycle in Italian.
Journal of Italian Linguistics 8, 5 - 3 4 .
Chomsky, N., and Halle, M. (1968). The Sound Pattern of English. Harper and Row, New
York.
Cohn, A. (1989). Stress in Indonesian and bracketing paradoxes. Natural Language and
Linguistic Theory 7, 167 - 2 1 6 .
Cole, J. (1991). Arguing for the phonological cycle: A critical review. In Proceedings of
the First Meeting of the Formal Linguistics Society of Midamerica. Department of
Linguistics, University of Wisconsin, Madison.
Halle, M., Harris, J., and Vergnaud, J.-R. (1991). A reexamination of the stress erasure
convention and Spanish stress. Linguistic Inquiry 22, 141 - 1 5 9 .
Halle, M., and Mohanan, K. P. (1985). Segmental phonology of Modern English. Linguis-
tic Inquiry 16, 5 7 - 1 1 6 .
Halle, M., and Vergnaud, J.-R. (1987a). An Essay on Stress. MIT Press, Cambridge, Mass.
Halle, M., and Vergnaud, J.-R. (1987b). Stress and the cycle. Linguistic Inquiry 1 8 , 4 5 - 8 4 .
Hargus, S. (1988). The Lexical Phonology ofSekani. Garland, New York.
Harris, J. (1989). The stress erasure convention and cliticization in Spanish. Linguistic
Inquiry 20, 3 3 9 - 3 6 4 .
Hayes, B. (1986). Inalterability in CV phonology. Language 62, 321 - 3 5 2 .
Hayes, B. (1990). Precompiled phrasal phonology. In The Phonology-Syntax Connection
(S. Inkelas and D. Zee, eds.), pp. 8 5 - 1 0 8 . CSLI Publications and University of Chi-
cago Press, Chicago.
Inkelas, S. (1990). Prosodic Constituency in the Lexicon. Garland, New York.
Kaisse, E. M. (1985). Connected Speech: The Interaction of Syntax and Phonology. Aca-
demic Press, Orlando.
Kaisse, E. M. (1986). Locating Turkish devoicing. Proceedings of the West Coast Confer-
ence on Formal Linguistics 5, 119 - 1 2 8 .
Kaisse, E. M. (1990). Toward a typology of postlexical rules. In The Phonology-Syntax
Introduction 19

Connection (S. Inkelas and D. Zee, eds.), pp. 1 2 7 - 1 4 3 . CSLI Publications and Uni-
versity of Chicago Press, Chicago.
Kang, O. (1992). Word-internal prosodic words in Korean. Proceedings of the North East-
ern Linguistics Society 22, 2 4 3 - 2 5 7 .
Kiparsky, P. (1971). Historical linguistics. In A Survey of Linguistic Science (W. O. Ding-
wall, ed.), pp. 5 7 6 - 6 4 9 . University of Maryland, College Park.
Kiparsky, P. (1982). Lexical morphology and phonology. In Linguistics in the Morning
Calm (I.-S. Yang, ed.), pp. 3 - 9 1 . Hanshin Publishing Co., Seoul.
Kiparsky, P. (1983). Word formation and the lexicon. Proceedings of the 1982 Mid-
America Linguistics Conference, pp. 3 - 2 9 .
Kiparsky, P. (1984). On the lexical phonology of Icelandic. In Nordic Prosody III: Papers
from a Symposium (C. C. Elert et al., eds.), pp. 135-164. University of Umea.
Kiparsky, P. (1985). Some consequences of lexical phonology. Phonology Yearbook 2,
85-138.
Kiparsky, P. (1988). Phonological change. In Linguistics: The Cambridge Survey (F. J.
Newmeyer, ed.), 1, 3 6 3 - 4 1 5 .
Lantolf, J. (1977). Stress subordination: Evidence from Spanish. General Linguistics 17,
8-19.
McFarland, T., and Pierrehumbert, J. (1991). On ich-laut, ach-laut and structure preserva-
tion. Phonology 8, 171 - 1 8 0 .
Mascaro, J. (1976). Catalan Phonology and the Phonological Cycle. Doctoral dissertation,
Massachusetts Institute of Technology, Cambridge.
Mohanan, K. P. (1986). The Theory of Lexical Phonology. Reidel, Dordrecht.
Nespor, M., and Vogel, I. (1986). Prosodic Phonology. Foris, Dordrecht.
Pesetsky, D. (1979). Russian Morphology and Lexical Theory. Unpublished manuscript,
Massachusetts Institute of Technology, Cambridge.
Rubach, J. (1984a). Cyclic and Lexical Phonology: The Structure of Polish. Foris,
Dordrecht.
Rubach, J. (1984b). Rule typology and phonological interference. In Theoretical Issues in
Contrastive Phonology (S. Eliason, ed.), pp. 3 7 - 5 0 . Julius Groos, Stuttgart.
Selkirk, E. O. (1984). Phonology and Syntax: The Relation Between Sound and Structure.
MIT Press, Cambridge, Mass.
Selkirk, E. O. (1986). On derived domains in sentence phonology. Phonology Yearbook 3,
371-405.
Siegel, D. (1979). Topics in English Morphology. Garland, New York.
Spencer, A. (1989). Morphological Theory: An Introduction to Word Structure in Genera-
tive Grammar. Blackwell, Oxford.
Speas, P. (1990). Functional Heads and the Mirror Principle. Unpublished manuscript,
University of Massachusetts, Amherst.
Sproat, R. (1985). On Deriving the Lexicon. Doctoral dissertation, Massachusetts Institute
of Technology, Cambridge.
Strauss, S. (1982). Lexicalist Phonology of English and German. Foris, Dordrecht.
Suner, M. (1975). Spanish adverbs: Support for the phonological cycle? Linguistic Inquiry
6, 6 0 2 - 6 0 5 .
Zee, D., and Inkelas, S. (1991). Clitic groups and the prosodic hierarchy. Proceedings of
the West Coast Conference on Formal Linguistics 5, 5 0 5 - 5 1 9 .
ON THE SIMULTANEITY OF MORPHOLOGICAL
AND PROSODIC STRUCTURE

GEERT BOOIJ*
ROCHELLE LIEBER
* Vakgroep Taalkunde
Vrije Universiteit
1007 MC Amsterdam
The Netherlands

^Department of English
University of New Hampshire
Durham, New Hampshire 03824

1. I N T R O D U C T I O N

In recent years, m u c h attention h a s b e e n d e v o t e d to the internal organization of

c o m p l e x w o r d s a n d to their p r o s o d i c structure. L e s s attention h a s b e e n d e v o t e d to
the relationship b e t w e e n p r o s o d i c a n d m o r p h o l o g i c a l structure. In this article w e
e x p l o r e this relationship in s o m e detail, a r g u i n g that there is g o o d reason to b e -
lieve that m o r p h o l o g i c a l a n d p r o s o d i c structure are built in t a n d e m a n d are avail-
able s i m u l t a n e o u s l y . W e s h o w further that it m u s t b e p o s s i b l e to m a k e reference
1

to the t w o coexisting structures of a single string b o t h in p h o n o l o g i c a l rules and

in the lexical entries of affixes. T h e theoretical benefits that w e derive from this
p r o p o s a l are large a n d c o n c e r n several o u t s t a n d i n g p r o b l e m s in m o r p h o l o g i c a l
theory, i n c l u d i n g h e a d o p e r a t i o n s (Aronoff, 1988), b r a c k e t i n g p a r a d o x e s (Peset-
sky, 1985; Sproat, 1985; a m o n g m a n y others), and the status of clitics.
T h e t h e o r y of m o r p h o l o g y w e a s s u m e h e r e is that of L i e b e r ( 1 9 8 9 , 1 9 9 2 ) , w h i c h
shares with p r e v i o u s w o r k in m o r p h o l o g y the n o t i o n that c o m p l e x w o r d s are hier-
archically structured and with L i e b e r ( 1 9 8 0 ) the idea that m o r p h o l o g i c a l struc-

23
Phonetics and Phonology, Volume 4 Copyright © 1993 by Academic Press, Inc.
Studies in Lexical Phonology All rights of reproduction in any form reserved.
24 Geert Booij and Rochelle Lieber

tures are built from the b o t t o m u p , as follows. A c c o r d i n g to this theory, all mor-
p h e m e s have lexical entries w h i c h indicate their c a t e g o r y and subcategorization
( w h a t category, if any, they attach to, and in w h a t direction), as well as their p h o -
nological representations, lexical c o n c e p t u a l structures ( L C S s ) , and predicate ar-
g u m e n t structures ( P A S s ) . M o r p h e m e s are put t o g e t h e r a c c o r d i n g to their m o r -
2

phological s u b c a t e g o r i z a t i o n r e q u i r e m e n t s , and hierarchical structure is projected

from lexical information a n d labeled using general principles of X - b a r theory and
feature percolation.
W i t h respect to p r o s o d i c c a t e g o r i e s , w e a s s u m e the following. P h o n o l o g i c a l
s e g m e n t s are g r o u p e d into a n u m b e r of hierarchically o r g a n i z e d p r o s o d i c c a t e g o -
ries. It is relatively uncontroversial to i n c l u d e a m o n g these p r o s o d i c categories the
syllable a , the foot F, the p h o n o l o g i c a l w o r d W d , and the p h o n o l o g i c a l p h r a s e c|>.
M c C a r t h y and P r i n c e ( 1 9 8 6 ) a r g u e that reference is s o m e t i m e s n e c e s s a r y as well
to particular sorts of s y l l a b l e s — t h e light syllable a ^ , the h e a v y syllable c r ^ , a n d
the c o r e syllable a (that is, a constituent consisting of a s i m p l e C V ) — a n d to a
c

constituent w h i c h they refer to as the m i n i m a l w o r d ( W d , w h i c h is equal to a

M I N

foot (see M c C a r t h y a n d Prince, 1 9 8 6 : 8 , for technical details). N e s p o r and Vogel

( 1 9 8 6 ) also a r g u e for a n u m b e r of p r o s o d i c constituents a b o v e the level of the
w o r d . For our p u r p o s e s it is not n e c e s s a r y to d e t e r m i n e w h a t the exact inventory
of p r o s o d i c constituents is. W e will b e m o s t c o n c e r n e d with constituents at or
b e l o w the level of the p r o s o d i c w o r d : a (with variants cr^ a ^ , a ) , F ( = W d c ),
M I N

and Wd.
A n o t h e r p o i n t in p r o s o d i c theory that w e take to b e uncontroversial is that m o r -
p h o l o g i c a l structure a n d p r o s o d i c structure n e e d not a l w a y s b e i s o m o r p h i c . Syl-
lable a n d foot b o u n d a r i e s d o n o t a l w a y s c o i n c i d e with m o r p h e m e b o u n d a r i e s , n o r
d o e s the p h o n o l o g i c a l w o r d a l w a y s m a t c h exactly w i t h the m o r p h o l o g i c a l w o r d
(see B o o i j , 1 9 8 5 ; Booij and R u b a c h , 1984).
It is at this point, however, that w e part c o m p a n y with the a b o v e m e n t i o n e d
theories of p r o s o d i c p h o n o l o g y . B o t h Selkirk ( 1 9 8 4 ) a n d N e s p o r a n d Vogel ( 1 9 8 6 )
a s s u m e that p r o s o d i c structure is built only after construction of w o r d s and sen-
tences has b e e n c o m p l e t e d . Selkirk ( 1 9 8 4 : 8 2 ) d u b s this a SYNTAX-FIRST a p p r o a c h .
Prosodic structure is created in t w o stages. B e l o w the w o r d level, prosodic struc-
ture is built after all m o r p h o l o g i c a l o p e r a t i o n s h a v e b e e n c o m p l e t e d . A b o v e the
w o r d level, p r o s o d i c structure is built as part of the postlexical p h o n o l o g y . N e s p o r
and Vogel ( 1 9 8 6 ) are s o m e w h a t less explicit than Selkirk about the overall orga-
nization of the g r a m m a r , b u t the picture that e m e r g e s from their w o r k is o n e in
w h i c h all p r o s o d i c structure is created as part of the postlexical phonology.
W e argue in w h a t follows that neither of t h e s e m o d e l s is correct. Rather, there
is g o o d reason to believe that m o r p h o l o g i c a l a n d p r o s o d i c structure are built at the
s a m e t i m e , from the b o t t o m u p , so that representations of w o r d s consist of t w o
s i m u l t a n e o u s structurings coexisting on distinct p l a n e s . T h i s a s s u m p t i o n has al-
w a y s b e e n m a d e in the standard version of the theory of lexical p h o n o l o g y , as
Morphological and Prosodic Structure 25

p r o p o s e d in K i p a r s k y ( 1 9 8 2 , 1985), b u t not a l w a y s very explicitly. It is o u r a i m to

s h o w that this a s s u m p t i o n is correct and that there is substantial e v i d e n c e that
b e l o w the w o r d level, m o r p h o l o g y and p r o s o d i c p h o n o l o g y interact and apply in
tandem.

2. E V I D E N C E F O R S I M U L T A N E I T Y

In this section w e a r g u e that lexical entries of m o r p h e m e s m a y refer simulta-

n e o u s l y to b o t h syntactic a n d p r o s o d i c r e q u i r e m e n t s on their e n v i r o n m e n t , and
that therefore the syntactic a n d p r o s o d i c structuring of s e g m e n t a l strings m u s t b e
d e r i v e d in t a n d e m .
A first e x a m p l e c o m e s from D u t c h . In this l a n g u a g e , there are a n u m b e r of
p r o d u c t i v e n o n n a t i v e suffixes that d e r i v e adjectives from n o n n a t i v e n o u n s e n d i n g
in -ie [i], a m o n g t h e m -isch / i s / a n d -ief/iv/. T h e c h o i c e b e t w e e n these t w o suf-
fixes with respect to b a s e n o u n s in -ie crucially d e p e n d s on the stress patterns of
the b a s e w o r d s : -isch is selected if the final syllable of the b a s e n o u n b e a r s m a i n
stress, w h e r e a s -ief is the correct suffix for n o u n s in u n s t r e s s e d -ie. T h i s is illus-
trated in (1).

(1) sociologie 'sociology' sociologisch 'sociological'

blasfemie 'blasphemy' blasfemisch 'blasphemous'
allergie 'allergy' allergisch 'allergic'
b. preventie 'prevention' preventief 'preventive'
constructie 'construction' constructief 'constructive'
integrdtie 'integration' integratief 'integrating'

T h e final s e g m e n t [i] of the b a s e n o u n is s u b s e q u e n t l y deleted before the initial [i]

of the suffix.
T h e t w o suffixes are also different in that -ief only p r o d u c t i v e l y attaches to
n o u n s in u n s t r e s s e d -ie, w h e r e a s -isch also o c c u r s with n o u n s that d o not e n d in
-ie. In the latter c a s e , there is n o stress c o n d i t i o n involved.

(2) profeet 'prophet' profetisch 'prophetical'

algebra 'algebra' algebraisch 'algebraic'
organisdtor 'organizer' organisatorisch 'organizational'

T h a t is, the stress p r o p e r t i e s of the b a s e n o u n are only relevant in the d o m a i n in

w h i c h t h e t w o suffixes c o m p e t e .
G i v e n these facts, the lexical entries for the m o r p h e m e s -ief and -isch must
c o n t a i n the following subcategorization.

(3) -ief (i)

°"W]N 1A
-isch ]N ] A
26 Geert Booij and Rochelle Lieber

W e a s s u m e that a t t a c h m e n t of an affix with a m o r e specific subcategorization

takes p r e c e d e n c e o v e r that of a c o m p e t i n g affix with a m o r e general subcategori-
zation (the e l s e w h e r e principle, cf. also van M a r i e , 1985). Therefore, it suffices to
m e n t i o n the p r o s o d i c c o n d i t i o n in the lexical entry of -ief. T h i s lexical entry then
requires b o t h the m o r p h o l o g i c a l and the p r o s o d i c properties of the b a s e w o r d to
b e available. N o t e that stress p r o p e r t i e s of w o r d s are to b e e x p r e s s e d in t e r m s of
s t r o n g / w e a k labeling of p r o s o d i c categories such as the syllable and the foot.
Therefore, a w o r d m u s t b e prosodified before the stress rules can assign a p r o m i -
nence p a t t e r n . 3 4

T h e relevance of p r o s o d i c information for m o r p h o l o g y is not restricted to in-

formation c o n c e r n i n g stress. In Polish, the c h o i c e b e t w e e n o n e of the t w o allo-
m o r p h s of both the c o m p a r a t i v e a n d the i m p e r a t i v e suffix a p p e a r s to d e p e n d on
a n o t h e r p r o s o d i c p r o p e r t y of the b a s e w o r d , namely, w h e t h e r its final c o n s o n a n t
can b e syllabified b y the syllabification a l g o r i t h m of Polish, or r e m a i n s extrasyl-
labic. T h e facts are as follows ( w e b a s e ourselves h e r e o n the analysis in R u b a c h
a n d B o o i j , 1990). T h e c o m p a r a t i v e m o r p h e m e is either sz [s] or ejsz [ejs]. T h e
general form is sz, a n d the a l l o m o r p h ejsz has to b e selected w h e n the stem e n d s
in an extrasyllabic c o n s o n a n t . For instance, in the following e x a m p l e s the stem
e n d s in a cluster of an o b s t r u e n t followed b y a s o n o r a n t c o n s o n a n t , w h i c h is an
i m p o s s i b l e c o d a b e c a u s e it violates the universal sonority s e q u e n c i n g generaliza-
tion (Selkirk, 1984) (4a), or by a cluster of t w o sonorant c o n s o n a n t s (4b), an ill-
f o r m e d c o d a in Polish, a n d therefore, the final c o n s o n a n t of these stems r e m a i n s
unsyllabified (-y is the n o m i n a t i v e singular e n d i n g ; the / before ejsz indicates pala-
talization of the p r e c e d i n g c o n s o n a n t ) .

(4) ADJECTIVE: COMPARATIVE:

a. podl-y ' m e a n ' podl-ejsz-y
szczodr-y ' g e n e r o u s ' szczodrz-ejsz-y
b . czarn-y ' b l a c k ' czarn-iejsz-y
ogoln-y ' g e n e r a l ' ogoln-iejsz-y
skromn-y ' m o d e s t ' skromn-iejsz-y
fajn-y ' n i c e ' fajn-iejsz-y

Therefore, the lexical entry for the m o r e specific c o m p a r a t i v e a l l o m o r p h ejsz will

b e as follows, w h e r e C * indicates an extrasyllabic c o n s o n a n t .

(5) ejsz C * ] A ] A

A s above, w e a s s u m e that in the c a s e of c o m p e t i n g affixes, the m o r e specific o n e

takes p r e c e d e n c e over the m o r e general, unrestricted o n e . 5

N o r m a l l y , the imperative m o r p h e m e of Polish d o e s not surface directly, but

only indirectly, in the form of palatalization of the stem-final c o n s o n a n t . T h e r e -
fore, it is a s s u m e d that it consists of a so-called yer, a floating s e g m e n t that only
surfaces phonetically in specific c o n t e x t s . For o u r p u r p o s e s it suffices to point out
Morphological and Prosodic Structure 27

here that w e also find an a l l o m o r p h in w h i c h the yer is p r e c e d e d b y t h e s e q u e n c e

ij [ij]. T h i s a l l o m o r p h only o c c u r s w h e n t h e final c o n s o n a n t of t h e stem is extra-
syllabic, as is illustrated in (6). A s in t h e p r e v i o u s case, there are t w o types of c o d a
clusters that give rise to extrasyllabic c o n s o n a n t s : clusters that violate the univer-
sal sonority s e q u e n c i n g generalization ( 6 a ) , a n d clusters that violate t h e Polish
prohibition o n clusters of s o n o r a n t s ( 6 b ) .

(6) V E R B A L STEM: IMPERATIVE:

a. nagl- 'to hurry' nagl-ij
spulchn- 'to m a k e soft' spulchn-ij
b. zwoln- 'to cover' zwoln-ij
utajn- 'to cover u p ' utajn-ij

H e n c e , t h e a l l o m o r p h / i j E / (E stands for t h e yer) is s u b c a t e g o r i z e d as follows.

(7) /ijE/ C*] v ] v

In s u m , for t h e selection of t h e p r o p e r a l l o m o r p h of b o t h t h e c o m p a r a t i v e a n d t h e
i m p e r a t i v e m o r p h e m e it is crucial that b o t h t h e m o r p h o l o g i c a l a n d t h e p r o s o d i c
structuring of t h e stem b e available. T h e s e facts thus s u p p o r t both t h e theory of
lexical p h o n o l o g y that c l a i m s that p h o n o l o g y a n d m o r p h o l o g y are interspersed,
and t h e c l a i m that is t h e subject of this article, t h e simultaneity thesis.
T h e r e q u i r e m e n t of simultaneity n o t only manifests itself in t h e subcategoriza-
tions of b o u n d m o r p h e m e s in t h e lexicon, b u t also in t h e fact that there are p h o -
nological rules that refer s i m u l t a n e o u s l y t o b o t h types of structuring. L e t us call
s u c h rules B I P L A N A R R U L E S . 6

H a y e s ( 1 9 8 2 ) p r o p o s e d such a b i p l a n a r rule for E n g l i s h , n a m e l y t h e rule of

Adjective Extrametricality. T h i s rule states that in E n g l i s h adjectives t h e final syl-
lable is e x t r a m e t r i c a l . T h u s , w e g e t correct stress a s s i g n m e n t s such as magnani-
mous a n d reluctant instead of t h e incorrect * magnanimous a n d * reluctant. T h i s
rule is a typically b i p l a n a r rule, b e c a u s e it refers to b o t h m o r p h o s y n t a c t i c infor-
m a t i o n ( t h e notion " a d j e c t i v e " ) a n d to p r o s o d i c structure (the notion " e x t r a p r o -
sodic s y l l a b l e " ) .
A s e c o n d e x a m p l e of such a rule is t h e stress rule for D u t c h n o m i n a l c o m -
p o u n d s . T h i s rule assigns m a i n stress to t h e left constituent of such c o m p o u n d s
(Visch, 1 9 8 9 : 8 4 ) .

(8) D U T C H C O M P O U N D STRESS R U L E :
In a configuration [ A B ] , A is strong.N

V i s c h correctly restricts this rule to n o m i n a l c o m p o u n d s b e c a u s e adjectival c o m -

p o u n d s such as reuze-sterk ' v e r y s t r o n g ' a n d donker-groen ' d a r k green' clearly
h a v e a different stress pattern in w h i c h both constituents are felt to b e equally
stressed. Therefore, rule (8) m u s t refer to m o r p h o s y n t a c t i c information, t h e cate-
g o r y label N . O n the o t h e r h a n d , this rule clearly refers to p r o s o d i c structure, since
28 Geert Booij and Rochelle Lieber

the constituents that receive the labels " s t r o n g " and " w e a k " are p r o s o d i c cate-
gories (usually called PROSODIC WORDS) w h i c h d o m i n a t e p r o s o d i c categories like
syllable and foot.
T h e stress rule for n o m i n a l c o m p o u n d s of D u t c h is a typical lexical rule, b e -
c a u s e it can also h a v e e x c e p t i o n s (cf. B o o i j , 1977). T h a t is, it c a n n o t simply
b e p a r t of the m a p p i n g p r o c e d u r e that m a p s m o r p h o s y n t a c t i c structure into p r o -
sodic structure. It is, therefore, an instance of a lexical p h o n o l o g i c a l rule that re-
fers s i m u l t a n e o u s l y to the t w o kinds of hierarchical structuring of w o r d s dis-
cussed here.
A final e x a m p l e of a b i p l a n a r rule is the G e r m a n rule of S c h w a Insertion in
n o u n s (Hall, 1989; W i e s e , 1988). T h i s rule inserts the G e r m a n default vowel
s c h w a before an extrasyllabic c o n s o n a n t . F o r instance, the u n d e r l y i n g form of
Uebel ' e v i l ' is / y b l / . W h e n w e syllabify this u n d e r l y i n g form, the III r e m a i n s ex-
trasyllabic, b e c a u s e a c o d a cluster lb\l w o u l d violate the sonority s e q u e n c i n g g e n -
eralization. A s c h w a is then inserted to " s a v e " the HI. A s Hall ( 1 9 8 9 : 8 3 5 ) points
out, this rule only applies to n o u n s : S c h w a Insertion before c o n s o n a n t s also o c c u r s
in adjectives, b u t at a later level, a n d not only before extrasyllabic c o n s o n a n t s .
Therefore, the structural description of this rule h a s to refer s i m u l t a n e o u s l y to
the m o r p h o s y n t a c t i c c a t e g o r y " n o u n " and the p r o s o d i c notion " e x t r a s y l l a b i c
consonant."
T h e c o n c l u s i o n of this section is that b o t h s u b c a t e g o r i z a t i o n s of m o r p h e m e s
and p h o n o l o g i c a l rules s o m e t i m e s h a v e to refer s i m u l t a n e o u s l y to m o r p h o l o g i c a l
a n d p r o s o d i c information, and b o t h thus h a v e a b i p l a n a r character. In the next
section w e s h o w h o w the c o n c e p t of biplanarity can b e u s e d to solve a n u m b e r of
theoretical p r o b l e m s with respect to the interaction of p h o n o l o g y and m o r p h o l o g y .

3. T H E O R E T I C A L C O N S E Q U E N C E S

3.1. Head Operations

T h e first p r o b l e m w e c o n s i d e r c o n c e r n s the e x i s t e n c e of w h a t Aronoff (1988),

following H o e k s e m a ( 1 9 8 5 ) , calls h e a d o p e r a t i o n s . H o e k s e m a (1985) defines the
7

n o t i o n H E A D O P E R A T I O N as in (9).

(9) F is a h e a d operation if F(Y) = Z and W = XY ( w h e r e Y is the h e a d of W)

together i m p l y that F(W) = X + F(Y) = X + Z.

(9) says simply that a m o r p h o l o g i c a l rule is a h e a d operation if it r e a c h e s into a

w o r d W to p e r f o r m an operation o n its h e a d Y, c h a n g i n g Y to Z. Aronoff applies
the notion of h e a d operation to several recalcitrant cases of reduplication, a m o n g
t h e m a classically p r o b l e m a t i c c a s e in T a g a l o g . T a g a l o g h a s a prefix pang- w h i c h
Morphological and Prosodic Structure 29

attaches to n o u n s . A s the d a t a in (10a) show, [rj] plus a following stop a p p e a r s in

the derived form as a single nasal h o m o r g a n i c with the u n d e r l y i n g stop.

(10) a. atip 'roofing' pang-atip 'that used for roofing'

pu: tul 'cut' pa-mu: tul 'that used for c u t t i n g '
b. pa-mu-mu: tul ' a cutting in q u a n t i t y '

T h e e x a m p l e in (10b) s h o w s further that w h e n the s e c o n d form in (10a) is re-

d u p l i c a t e d , the r e d u p l i c a t i n g stem s h o w s the effects of h a v i n g already u n d e r g o n e
affixation; the stem-initial [p] h a s b e c o m e [m] prior to reduplication. T h i s analysis
is of c o u r s e p r o b l e m a t i c in traditional f r a m e w o r k s in w h i c h m o r p h o l o g y strictly
p r e c e d e s p h o n o l o g y ; in s u c h c a s e s the sandhi rule o p e r a t i n g b e t w e e n prefix and
stem s e e m s to h a v e " o v e r a p p l i e d . " T h e o r d e r i n g of the p h o n o l o g i c a l rule with
respect to reduplication is not necessarily p r o b l e m a t i c in f r a m e w o r k s w h e r e m o r -
p h o l o g i c a l rules c a n apply to t h e o u t p u t of p h o n o l o g i c a l rules and vice versa, as
in the theory of lexical p h o n o l o g y w e a s s u m e h e r e . N e v e r t h e l e s s , even in frame-
w o r k s in w h i c h p h o n o l o g i c a l a n d m o r p h o l o g i c a l rules can be interspersed, it m u s t
still b e e x p l a i n e d w h y the reduplication rule s e e m s to reach into an already
prefixed w o r d .
Aronoff suggests that the derivation of the form in (10b) involves a head o p e r a -
tion. After affixation of pang-, w h i c h triggers sandhi, reduplication r e a c h e s into
the w o r d to c o p y the first t w o s e g m e n t s of the stem. T h e notion that certain m o r -
p h o l o g i c a l o p e r a t i o n s m u s t b e " h e a d o p e r a t i o n s " is a p r o b l e m a t i c o n e . It is not at
all clear that the item o p e r a t e d on by the " h e a d o p e r a t i o n " is actually the h e a d of
the w o r d . L i e b e r ( 1 9 9 2 ) s h o w s that T a g a l o g w o r d formation is largely left-headed;
the majority of T a g a l o g prefixes are c a t e g o r y - c h a n g i n g . In the c a s e outlined a b o v e
as well, it is very likely that it is the o u t e r m o s t prefix rather than the stem w h i c h
is the head. Specifically, a c c o r d i n g to S c h a c h t e r and O t a n e s ( 1 9 7 2 ) , pang- attaches
to n o u n or verb s t e m s to form adjectives. A l t h o u g h Aronoff's g l o s s e s , taken from
Bloomfield ( 1 9 3 3 ) , suggest that the pang- f o r m s are n o u n s , a native s p e a k e r of
T a g a l o g confirms that they are adjectives instead with the glosses 'for roofing' and
'for c u t t i n g ' , in c o n f o r m i t y to S c h a c h t e r and O t a n e s ( 1 9 7 2 ) . R e d u p l i c a t i o n then
8

c h a n g e s the pang- adjective to a n o u n . A n d if the reduplicative affix c h a n g e s cate-

gory, it m u s t be the head. T h e stem therefore c a n n o t b e the head, a n d the operation
c a n n o t be a head operation.
W e therefore suggest that T a g a l o g reduplication and other similar cases are not
head o p e r a t i o n s . Rather, they a p p e a r to involve w h a t B r o s e l o w and M c C a r t h y
( 1 9 8 4 ) and M c C a r t h y and Prince ( 1 9 8 6 ) call AFFIXATION T O A PROSODIC C O N -
S T I T U E N T . In fact, w e p r o p o s e that the T a g a l o g reduplication p r o c e s s sketched
a b o v e is o n e in w h i c h the reduplicative m o r p h e m e is s u b c a t e g o r i z e d for both m o r -
p h o l o g i c a l and p r o s o d i c constituents, and that this s i m u l t a n e o u s biplanar subca-
tegorization gives rise to n o n i s o m o r p h i c p r o s o d i c and m o r p h o l o g i c a l structures in
this case.
30 Geert Booij and Rochelle Lieber

Let us first illustrate the notion of b i p l a n a r subcategorization with a s o m e w h a t

simpler c a s e . M c C a r t h y a n d P r i n c e ( 1 9 8 6 : 1 2 ) s h o w that it is s o m e t i m e s n e c e s s a r y
to s u b c a t e g o r i z e affixes to attach to p r o s o d i c constituents (e.g., W d , W d , etc.), M I N

rather than to p u r e l y m o r p h o l o g i c a l constituents (that is, an X ° of s o m e s o r t ) . 9

T h e y argue, for e x a m p l e , that the reduplicative affix in the Australian l a n g u a g e

Y i d i n y ( D i x o n , 1977) m u s t attach to W d , rather than simply to N ° . C o n s i d e r
M I N

the e x a m p l e s in (11).

(11) YIDINY NOMINAL REDUPLICATION : 1 0

mulari 'initiated m a n ' mulamulari 'initiated m e n '

kintalpa 'lizard s p . ' kintalkintalpa 'lizards'

T h e Y i d i n y reduplicative prefix is, a c c o r d i n g to M c C a r t h y and P r i n c e ( 1 9 8 6 ) , the

p r o s o d i c constituent W d ( w h i c h is to say a f o o t — t w o syllables in Y i d i n y ) . If
M I N

the W d M I N
prefix w e r e to attach to the m o r p h o l o g i c a l constituent N ° , with con-
c o m i t a n t c o p y i n g of the p h o n e m i c m e l o d y of this constituent, w e w o u l d expect
the derivations illustrated in (12). N o t e that in (12) p r o s o d i c structure is illustrated
a b o v e the p h o n e m i c melody, m o r p h o l o g i c a l structure below.

(12) a. Wd M I N
Wd M I N
b. Wd M I N
Wd M I N

A A A A
o" (j a a a era a a a
AAA AAA
kintalpa m ul a r i

N N N ^ ^ ^ ^

\ Copying and Association \ Copying and Association

Wd M I N
Wd M I N
Wd M I N
Wd M I N

A A A A
era CTO-CF a a a a a

AA AAA AA AAA
kin tal kintalpa = kintalkintalpa mular mulari = *mularmulan

T h e derivations in (12) are p r e s u m e d to g o as follows. In both (12a) and (12b) the

reduplicativeN prefix W N d is attached to the n o u n , and m
M I N
^ ^o\ ryp/ h o l o g i c a l structure

is built. T h e p h o n e m i c m e l o d y of the verb stem is c o p i e d and the p r o s o d i c affix

incorporates as m u c h of the p h o n e m i c m e l o d y as can b e fitted into its t w o syl-
lables. T h e result is correct for the c a s e in (12a); kintalpa reduplicates as kintalkin-
talpa. B u t (12b) is not; reduplication b a s e d on the w h o l e n o u n stem yields *raw-
Morphological and Prosodic Structure 31

larmulari, rather than t h e correct mulamulari. T h e question raised is thus h o w t o

get t h e / of kintalpa t o reduplicate w i t h o u t also getting t h e r of mulari s h o w i n g
u p in t h e reduplicative prefix.
M c C a r t h y a n d Prince a r g u e that this pattern of facts follows if the reduplicative
prefix W d attaches t o t h e p r o s o d i c constituent W d
M I N
rather than simply to M I N

N ° , a n d if w e m a k e t h e following crucial a s s u m p t i o n : O N L Y T H E PHONEMIC

M E L O D Y O F T H E PROSODIC C O N S T I T U E N T T O W H I C H T H E R E D U P L I C A T I V E AFFIX
A T T A C H E S is A V A I L A B L E FOR C O P Y I N G . F o r t h e e x a m p l e in ( 1 2 a ) , t h e p r o s o d i c
constituent W d w h i c h is copied is kintal, since t h e / forms t h e c o d a of t h e
M I N

s e c o n d syllable of t h e W d . B u t for ( 1 2 b ) t h e constituent w h i c h is copied is

M I N

mula, t h e r b e i n g t h e onset of t h e third stem syllable, a n d therefore n o t part of t h e

Wd . T h i s is illustrated in the derivations in (13), w h e r e the p l a n e of m o r p h o -
M I N

logical structure is again b e n e a t h the m e l o d y a n d that of p r o s o d i c structure a b o v e

the melody.

(13) a. W d M I N
Wd M I N
b. W d M I N
Wd w

A A
AAA
kin tal pa

N N

I Copying and Association \ Copying and Association

Wd Wd

W d M I N W d M I N w d M I N w d M I N

A A A A
O" O" CTCTCT O" O" (J (J (J

AA AAA AA AAA
kin tal pa

The pattern of reduplication illustrated in Yidiny can thus be accounted for if

we assume that a reduplicative affix can sometimes be subcategorized for a pro-
sodic constituent alongside a morphosyntactic constituent. In the theory of Lieber
( 1 9 9 2 ) , the Yidiny reduplicative prefix will therefore have the biplanar subcate-
gorization in ( 1 4 ) .

(14) YIDINY REDUPLICATION:

WdMIN / [ n [ N / W d M , N
32 Geert Booij and Rochelle Lieber

T h e notation in (14) should b e interpreted as follows. T h e reduplicative prefix is

a Wd M I N
w h i c h attaches to a W d M I N
in p r o s o d i c structure and to a N ° in m o r p h o -
logical structure. B o t h m o r p h o l o g i c a l structure and p r o s o d i c structure m u s t o b v i -
ously b e p r e s e n t s i m u l t a n e o u s l y for such a subcategorization to b e met.
T h e notion of b i p l a n a r s u b c a t e g o r i z a t i o n m a y n o w b e u s e d to a c c o u n t for the
T a g a l o g case in (10). W e a s s u m e that the particular reduplicative prefix in q u e s -
tion is a core syllable (that is, C V ) , a c in the notation of M c C a r t h y a n d P r i n c e
( 1 9 8 6 ) , and that it h a s the s u b c a t e g o r i z a t i o n in (15).

(15) TAGALOG REDUPLICATION:

a c / [N [ /Wd
A
M I N

(15) says that the reduplicative prefix a attaches m o r p h o l o g i c a l l y to an A and

c
0

prosodically to the W d ( = a foot in T a g a l o g ) . L e t us see w h a t h a p p e n s w h e n

M I N

this prefix is attached. W e a s s u m e , first of all, that the prefix pang- is attached to
a n o u n or verb stem, triggering the p h o n o l o g i c a l rule of sandhi a n d giving rise to
the s i m u l t a n e o u s m o r p h o l o g i c a l a n d p r o s o d i c structure illustrated in (16).

(16) a. F Wd M I N

I A
o" a a

pan puu tul

A V

I Sandhi and Resyllabification

MIN
F W d

I Ao~ (J
A
pa
AA
muu tul

A V

W h e n w e try to insert the reduplicative prefix, however, w e find that w e c a n n o t

fulfill the m o r p h o l o g i c a l and p r o s o d i c subcategorizations simultaneously. If w e
insert the CT to the left of pa-, in (16b), the CT will n o t b e adjacent to the W d
C c , M I N

as illustrated in (17a). B u t if w e try to insert the reduplicative prefix so that it is

adjacent to the W d , it will not b e adjacent to the A, as s h o w n in ( 1 7 b ) ; indeed,
M I N
Morphological and Prosodic Structure 33

it is u n c l e a r h o w m o r p h o l o g i c a l structure c o u l d b e projected at all in this structure,

since to d o so w o u l d involve creation of m o r p h o l o g i c a l structure on top of already
existing m o r p h o l o g i c a l structure.
,MIN
(17) a.* b.* F

In o r d e r to get the reduplicative prefix in T a g a l o g to fulfill its m o r p h o l o g i c a l

a n d p r o s o d i c s u b c a t e g o r i z a t i o n s simultaneously, w e n e e d to m a k e o n e further as-
s u m p t i o n . It is clear that the lexical entry in (15) c o n t a i n s (at least) t w o sorts of
r e q u i r e m e n t s , b o t h of w h i c h m u s t b e met. T h e reduplicative prefix consists of
p h o n o l o g i c a l information (it is a c o r e syllable w i t h o u t any inherent s e g m e n t a l
c o n t e n t ) a n d m o r p h o s y n t a c t i c information (it is a b o u n d n o u n , w h i c h p r e s u m a b l y
carries all of the m o r p h o s y n t a c t i c features of n o u n s in T a g a l o g ) . G i v e n the dual
c o n t e n t of the reduplicative prefix in T a g a l o g , w e a s s u m e that the following o c -
curs. S i n c e it is not p o s s i b l e to satisfy its subcategorization if the prefix r e m a i n s
intact, w e a s s u m e that a split o c c u r s in the lexical representation of the prefix in
o r d e r to m e e t b o t h p h o n o l o g i c a l a n d m o r p h o s y n t a c t i c r e q u i r e m e n t s : the p h o n o -
logical material is inserted into the tree in (16b) adjacent to the W d M I N
, thus
satisfying the p h o n o l o g i c a l part of the subcategorization, and the m o r p h o s y n t a c t i c
part (the c a t e g o r y features for N plus c o n c o m i t a n t m o r p h o s y n t a c t i c features) is
adjoined to the A , t h u s satisfying the m o r p h o s y n t a c t i c part of the s u b c a t e g o r i z a -
tion. T h i s is illustrated in (18).

(18) a. b.

r-
A A A A
pa muutul muu tul

- pa-mu-mu:tul
34 Geert Booij and Rochelle Lieber

(18a) s h o w s the splitting of the p h o n o l o g i c a l and m o r p h o s y n t a c t i c parts of the

entry. T h i s is then followed b y the c o p y i n g of the p h o n e m i c m e l o d y of the W d M I N

and association to the a . W e a s s u m e that the cr prefix is then incorporated into

c c

the existing p r o s o d i c structure by b e i n g a b s o r b e d into the p r e c e d i n g F. T h i s is

illustrated in (18b).
N o t e that w e are not p r o p o s i n g that the reduplicative affix in T a g a l o g MOVES
from o n e part of the w o r d structure to another, b u t rather that the dual s u b c a t e g o -
rization r e q u i r e m e n t forces the lexical entry of the prefix to split u p o n insertion,
so that the syllable t e m p l a t e is severed from its categorial signature. T h e o u t e r m o s t
layer of structure in (18b) d o e s not contain a trace or an e m p t y e l e m e n t of any
sort, since there is n o m o v e m e n t involved h e r e ; it m e r e l y carries the categorial
signature of the prefix. A s s u m i n g that m o r p h o l o g i c a l and p r o s o d i c structure are
built in t a n d e m , and also that s u b c a t e g o r i z a t i o n of m o r p h e m e s m u s t s o m e t i m e s
satisfy both m o r p h o l o g i c a l a n d p r o s o d i c r e q u i r e m e n t s thus allows us to explain
the apparently o d d b e h a v i o r of the reduplicative prefix in T a g a l o g w i t h o u t invok-
ing the special d e v i c e of h e a d o p e r a t i o n s . W e will see in the next section that other
theoretical benefits follow from these a s s u m p t i o n s as well.

3.2. Bracketing Paradoxes

In this section w e a r g u e that a n u m b e r of w e l l - k n o w n bracketing p a r a d o x e s can

b e m a d e to d i s a p p e a r if the simultaneity of m o r p h o l o g i c a l and p r o s o d i c structure
is taken into account, and specifically if affixes are p e r m i t t e d to h a v e both m o r -
p h o l o g i c a l a n d p r o s o d i c s u b c a t e g o r i z a t i o n s , as p r e v i o u s l y a r g u e d . W e begin with
a discussion of the w e l l - k n o w n b r a c k e t i n g p a r a d o x of the E n g l i s h c o m p a r a t i v e
form unhappier (see also Booij and R u b a c h , 1984, a n d C o h n , 1989, for discussion
of similar p a r a d o x e s in I n d o n e s i a n ) .
T h e p r o b l e m p r e s e n t e d b y unhappier is as follows: the E n g l i s h c o m p a r a t i v e
suffix -er can n o r m a l l y only b e attached to adjectival b a s e s consisting of o n e syl-
lable, or consisting of t w o syllables of w h i c h the s e c o n d o n e is light, a character-
istic e x a m p l e of a p r o s o d i c c o n d i t i o n on w o r d formation. Pesetsky (1985) o b -
serves not only that happy allows for -er affixation, but also that it is possible to
affix -er to the derived adjective unhappy, a l t h o u g h it consists of three syllables.
T h e so-called b r a c k e t i n g p a r a d o x is therefore that from the m o r p h o l o g i c a l point
of view unhappier is d e r i v e d from unhappy, w h e r e a s , given the p r o s o d i c c o n d i -
tion on the c o m p a r a t i v e m o r p h e m e , unhappier s e e m s to b e derived from happier.

(19) morphology: [[un[happy]]er]

phonology: [un[[happy]er]]

Booij and R u b a c h (1984) p r o p o s e to solve this p r o b l e m b y a s s u m i n g that the

p r o s o d i c condition on -er-affixation d o e s not pertain to the w h o l e w o r d , but rather
to the p r o s o d i c w o r d to w h i c h -er is attached. T h e p r o s o d i c structure of happy
Morphological and Prosodic Structure 35

consists of t w o syllables that t o g e t h e r form o n e p r o s o d i c w o r d . Unhappy, on the

other h a n d , consists of t w o p r o s o d i c w o r d s , un a n d happy.
W e p r o p o s e to treat the c o m p a r a t i v e affix -er as an affix w h i c h h a s s i m u l t a n e o u s
m o r p h o l o g i c a l a n d p r o s o d i c s u b c a t e g o r i z a t i o n . T h e lexical e n t r y for the English
c o m p a r a t i v e m o r p h e m e -er thus l o o k s like ( 2 0 ) .

(20) er ]A ] A

A
[ a (a )]
c w d

N o t e that w e d o n o t n e e d to stipulate h e r e that the p r o s o d i c restriction to o n e or

t w o syllables that -er is subject to p e r t a i n s to the last p r o s o d i c w o r d only. W e
a s s u m e that s u b c a t e g o r i z a t i o n requires strict locality. A n affix s u b c a t e g o r i z e d to
attach to a p r o s o d i c c o n s t i t u e n t X m u s t attach to the closest X . In the c a s e of -er,
this is the last p r o s o d i c w o r d of the b a s e w o r d , since -er, like all c o h e r i n g suffixes
(i.e., suffixes that d o n o t f o r m a p r o s o d i c w o r d of their o w n ) fuses prosodically
w i t h the last p r o s o d i c w o r d of the w o r d to w h i c h it is attached, w i t h c o n c o m i t a n t
^ s y l l a b i f i c a t i o n . In other w o r d s , p r o s o d i c s u b c a t e g o r i z a t i o n s of m o r p h e m e s can
only see the p r o s o d i c a l l y adjacent m a t e r i a l .

A related b r a c k e t i n g p a r a d o x is the c a s e of ungrammaticality a n d similar w o r d s

in -ity a n d -ation. M o r p h o l o g i c a l l y , ungrammaticality is to b e c o n s i d e r e d as a
derivation from ungrammatical. H o w e v e r , p h o n o l o g i c a l l y it s h o u l d b e seen as a
c a s e of prefixation of un- to grammaticality, b e c a u s e in the c u r r e n t a n a l y s e s of
lexical p h o n o l o g y the stress-neutral prefix un- s h o u l d b e a d d e d after (i.e., at a later
level t h a n ) the stress-shifting suffix -ity. T h i s is a p r o b l e m for m o r p h o l o g y , b e -
c a u s e un- is s u b c a t e g o r i z e d for adjectives, n o t for n o u n s .
A s Booij and R u b a c h ( 1 9 8 4 ) p o i n t out, this p r o b l e m can b e solved by realizing
t w o t h i n g s . First, the d o m a i n of the W o r d Stress rule of E n g l i s h is n o t the m o r p h o -
logical w o r d b u t rather the p r o s o d i c w o r d . T h u s in c o m p o u n d s , w h i c h consist of
36 Geert Booij and Rochelle Lieber

at least t w o p r o s o d i c w o r d s , the W o r d Stress rule applies in at least t w o d o m a i n s .

N o t e also that the W o r d Stress rule is a rule that specifies p r o m i n e n c e relations
b e t w e e n syllables within a p r o s o d i c w o r d . Secondly, as pointed out a b o v e , the
prefix un- can b e a s s u m e d to form a p r o s o d i c w o r d of its o w n . T h i s implies that
there is n o p h o n o l o g i c a l p r o b l e m created a n y m o r e b y the correct m o r p h o l o g i c a l
structure [[un[grammatical]]ity]. T h e relevant d o m a i n s for the a s s i g n m e n t of
w o r d stress are (un) a n d (grammaticality). In other w o r d s , a l t h o u g h m o r p h o l o g i -
cally -ity attaches to the w h o l e b a s e w o r d ungrammatical, prosodically it is only
attached to the last p r o s o d i c word, with w h i c h it fuses: (grammatical). T h e lexical
entry for -ity will therefore b e as follows.

(22) -ity h
N o t e that it is not n e c e s s a r y to s u b c a t e g o r i z e -ity for a p r e c e d i n g p r o s o d i c word,
b e c a u s e n o r m a l l y suffixes b e c o m e p a r t of the p r e c e d i n g p r o s o d i c w o r d .
T h e representation of ungrammaticality will thus b e as follows (the asterisk
indicates the d e s i g n a t e d t e r m i n a l e l e m e n t of the p r o s o d i c w o r d grammaticality).

(23)

T h e s a m e analysis c a n b e applied to similar cases such as underestimation and

extrametricality, since under- a n d extra- c a n also b e c o n s i d e r e d to b e p r o s o d i c
w o r d s of their o w n . N o t e , b y the way, that w e also h a v e to specify p r o m i n e n c e
relations within so-called stress-neutral prefixes; both in under- a n d extra- the first
syllable is strong, in c o n f o r m i t y with the w o r d stress rule. T h a t is, it is i m p o s s i b l e
to a c c o u n t for the so-called stress-neutral character of E n g l i s h prefixes even b y
o r d e r i n g prefixation after the w o r d stress rules, since polysyllabic prefixes c o n -
form to the p a t t e r n s of m e t r i c a l structure a s s i g n m e n t that w e find for w o r d s , and
therefore they h a v e to u n d e r g o the rule for (prosodic!) w o r d stress.
It should b e o b s e r v e d that the solution to this particular b r a c k e t i n g p a r a d o x ,
a l t h o u g h it s h o w s the necessity for an analysis in w h i c h the t w o sorts of structur-
ing are available, d o e s not necessarily require these t w o structures to b e present
simultaneously. N e v e r t h e l e s s , w e deal with these p h e n o m e n a h e r e for t w o r e a s o n s .
Morphological and Prosodic Structure 37

First, the h y p o t h e s i s of biplanarity p r o v i d e s us with a natural solution for this kind

of b r a c k e t i n g p a r a d o x . Secondly, this analysis d o e s i m p l y that rules m a y h a v e to
refer to b o t h p l a n e s : w h e r e a s in E n g l i s h c o m p o u n d s consisting of t w o p r o s o d i c
w o r d s the C o m p o u n d Stress rule correctly predicts the first to b e strong, the situ-
ation is j u s t the o p p o s i t e in prefixed structures w h e r e the s e c o n d p r o s o d i c w o r d is
the strongest o n e . T h a t is, the metrical rules that assign p r o m i n e n c e relations
a b o v e t h e level of the p r o s o d i c w o r d are sensitive to m o r p h o l o g i c a l information,
n a m e l y the difference b e t w e e n n o m i n a l c o m p o u n d s a n d prefixed c o m p l e x n o u n s .
T h u s , this c a s e is parallel to the D u t c h o n e d i s c u s s e d a b o v e c o n c e r n i n g the stress
differences b e t w e e n n o m i n a l and adjectival c o m p o u n d s in D u t c h .
W e therefore c o n c l u d e that by m a k i n g use of the b i p l a n a r nature of the struc-
turing of w o r d s , there are n o b r a c k e t i n g p a r a d o x e s that h a v e to d o w i t h a conflict
b e t w e e n p h o n o l o g y a n d m o r p h o l o g y , a n d w e d o not n e e d to i n t r o d u c e multiple
levels of r e p r e s e n t a t i o n and rules relating t h e s e levels in m o r p h o l o g y , as p r o p o s e d
by P e s e t s k y ( 1 9 8 5 ) and S p r o a t ( 1 9 8 5 , 1 9 8 8 ) .
11

3 . 3 . Clitics

Clitics f o r m classical e x a m p l e s of the n o n i s o m o r p h y b e t w e e n m o r p h o s y n t a c t i c

a n d p r o s o d i c structure. T h i s can b e seen m o s t clearly in the c a s e of so-called
s i m p l e clitics ( Z w i c k y , 1977) that h a v e the s a m e syntactic distribution as their
nonclitic c o u n t e r p a r t s b u t are prosodically d e p e n d e n t o n either the following p r o -
sodic w o r d (proclisis) or the p r e c e d i n g p r o s o d i c w o r d (enclisis). In this section w e
a r g u e that s i m p l e clitics are e l e m e n t s that h a v e only p r o s o d i c s u b c a t e g o r i z a t i o n
but n o m o r p h o l o g i c a l subcategorization a n d that they are distinct from affixes,
w h i c h d o h a v e m o r p h o l o g i c a l subcategorization. T h e p r e s e n t f r a m e w o r k therefore
m a k e s available a c o n v e n i e n t t y p o l o g y in w h i c h clitics can b e distinguished from
other b o u n d m o r p h e m e s .
W e illustrate this with the D u t c h third p e r s o n singular clitic p r o n o u n ie [i] that
is syntactically e q u i v a l e n t to its strong c o u n t e r p a r t hij ' h e ' . Ie is an enclitic
1 2

b e c a u s e it a l w a y s fuses prosodically with t h e p r e c e d i n g p r o s o d i c w o r d , w h i c h

functions as its host. T h i s host p r o v i d e s the n e c e s s a r y p r o s o d i c support. T h e fol-
l o w i n g s e n t e n c e s illustrate the syntactic e q u i v a l e n c e of hij and ie.

(24) a. Komthij? /Komtie?

lit. ' C o m e s h e ? D o e s he c o m e ? '
b. dat hij komt / dat ie komt
'that h e c o m e s '
c. wat hij doet / wat ie doet
'what he does'

T h a t ie f o r m s o n e p r o s o d i c w o r d with the p r e c e d i n g w o r d is clear from the sylla-

bification p a t t e r n s (kom)^ (tie) , (da) (tie) , and (wa) (tie) , w h i c h s h o w that
iT a a a a
38 Geert Booij and Rochelle Lieber

the syntactic b o u n d a r y before ie d o e s not create a p r o s o d i c word b o u n d a r y [com-

p a r e komt aan ' c o m e s at (i.e., a r r i v e s ) ' with the syllabification pattern (komt)„
(aan) ].
a N o t e also that ie is a typical clitic in that it c o m b i n e s with w o r d s of
c o m p l e t e l y different syntactic c a t e g o r i e s , n a m e l y verbs, c o m p l e m e n t i z e r s , and
relative (or interrogative) p r o n o u n s . W e can e x p r e s s this p r o s o d i c p r o p e r t y of the
clitic ie by a s s i g n i n g the following p r o s o d i c subcategorization to its lexical entry.

(25) ie N , 3rd p e r s . sing. ] W d

T h i s lexical e n t r y for -ie states that ie c a n only b e inserted after a p r o s o d i c

w o r d . T h i s clearly requires that at the level of lexical insertion the p r o s o d i c struc-
turing of w o r d s u p to the w o r d level is already available, and this is exactly w h a t
is predicted by o u r view of the role of p r o s o d i c structure in the lexical p h o n o l o g y :
since m o r p h o l o g i c a l a n d p r o s o d i c structure are derived simultaneously, both k i n d s
of information are available at the level of lexical insertion. W e also a s s u m e that,
like affixes, such clitics b e c o m e p a r t of the p r o s o d i c c a t e g o r y for w h i c h they are
subcategorized. But, u n l i k e affixes, they d o n o t h a v e a syntactic subcategorization,
and h e n c e they c o o c c u r with w o r d s of different syntactic categories.
This p r o s o d i c s u b c a t e g o r i z a t i o n of ie also correctly predicts that ie c a n n o t o c -
c u r at the b e g i n n i n g of a s e n t e n c e , b e c a u s e in that case there is n o host available.

(26) Hij komt. I*le komt

'He comes.'

T h a t is, the e x c l u s i o n of ie from the sentence-initial position d o e s not n e e d to b e

a c c o u n t e d for by a special stipulation in the syntax, but simply follows from its
p r o s o d i c s u b c a t e g o r i z a t i o n . Similarly, ie c a n n o t b e u s e d as a o n e - w o r d sentence
13

(for instance as an a n s w e r to a q u e s t i o n ) b e c a u s e in that case it w o u l d also lack a

prosodic host. F r o m this w e m a y c o n c l u d e that the c o n c e p t of " p r o s o d i c subcate-
g o r i z a t i o n " is not only n e c e s s a r y for e x p r e s s i n g p r o s o d i c conditions in m o r -
phology, but also to a c c o u n t for the b e h a v i o r and distributional restrictions
of simple clitics. M o r e o v e r , this analysis s u p p o r t s o u r view that p r o s o d i c and
m o r p h o s y n t a c t i c properties of m o r p h e m e s a n d w o r d s m u s t b e simultaneously
available.
T h e c o n c e p t of p r o s o d i c s u b c a t e g o r i z a t i o n can also b e u s e d in a c c o u n t i n g for
the o b s e r v a t i o n s c o n c e r n i n g clitics m a d e by K l a v a n s ( 1 9 8 5 ) . T h e m a i n t h e m e of
this interesting article is the i n d e p e n d e n c e of syntax and p h o n o l o g y in cliticiza-
tion. For instance, the following situation obtains in N g a n h c a r a , an Australian
l a n g u a g e : the clitics ngku ' y o u ' a n d nhcara ' u s ' o c c u r either before or after the
verb, w h i c h is a l w a y s sentence-final (otherwise, w o r d order is free in this lan-
g u a g e ) . Therefore, K l a v a n s c o n s i d e r s the verb as the syntactic host of these clitics.
However, phonologically, these clitics are always attached to the p r e c e d i n g word.
This is a p h o n o t a c t i c necessity, b e c a u s e N g a n h c a r a d o e s not allow for the c o n s o -
nant clusters ngk and nhc in word-initial position. Therefore, the first c o n s o n a n t
Morphological and Prosodic Structure 39

of the cluster h a s to f o r m a syllable with the final v o w e l of the p r e c e d i n g w o r d .

T h i s is illustrated by the following s e n t e n c e t a k e n from K l a v a n s ( 1 9 8 5 : 1 0 4 ) .

(27) nhila pama-ng nhingku ku?a=ngku wa:

he.NOM man.ERG 2sg.DAT child.DAT—2sg.DAT g i v e = D A T
T h e m a n g a v e a d o g to you.'

T h e enclitic n a t u r e of ngku is indicated b y ' = ' . K l a v a n s ( 1 9 8 5 : 9 8 ) r e m a r k s that

the direction of p h o n o l o g i c a l a t t a c h m e n t is a p r o p e r t y of the clitic itself. In o u r
a n a l y s i s , this c a n b e e x p r e s s e d by p r o v i d i n g t h e lexical e n t r y for s u c h clitic p r o -
nouns with the prosodic subcategorization ] w d W e also a s s u m e that, j u s t
as in the c a s e of the E n g l i s h c o m p a r a t i v e suffix dealt w i t h a b o v e , clitics that
are s u b c a t e g o r i z e d for a p r o s o d i c w o r d b e c o m e p a r t of that p r o s o d i c w o r d b y
convention.
A l t h o u g h K l a v a n s ' s o b s e r v a t i o n s a b o u t the b e h a v i o r of clitics a p p e a r to b e cor-
rect, h e r o w n f o r m a l i z a t i o n of t h e enclitic p r o p e r t y of s u c h p r o n o u n s is i n a d e -
q u a t e . S h e p r o p o s e s to c o n s i d e r clitics as " p h r a s a l affixes," that is, as w o r d s that
are s u b c a t e g o r i z e d (in the sense of Lieber, 1980) for a p h r a s a l host. F o r i n s t a n c e ,
the g e n e r a l f o r m of the s u b c a t e g o r i z a t i o n frame of clitics that she p r o p o s e s
(p. 117) is as follows.

(28) X ' [ lx' clitic

= e n

proclitic = - [ ]
x x

N o t e , h o w e v e r , that s u c h a s u b c a t e g o r i z a t i o n frame is i m p o s s i b l e in t h o s e c a s e s
w h e r e an enclitic is s u b c a t e g o r i z e d for a syntactic host on its right side, unless w e
also a l l o w for s u b c a t e g o r i z a t i o n frames of the following t y p e , with the b o u n d a r y
s y m b o l " = " n o n a d j a c e n t to the c a t e g o r y for w h i c h the clitic is s u b c a t e g o r i z e d .

(29) = enclitic , [ x ] .
x

X ' [ ] ' proclitic =

X

T h i s a m o u n t s to u s i n g the s y m b o l " = " as a diacritic for t h e p r o s o d i c r e q u i r e m e n t

" f o l l o w s / p r e c e d e s a p r o s o d i c h o s t . " T h a t is, s u b c a t e g o r i z a t i o n f r a m e s of t h e f o r m
p r o p o s e d by K l a v a n s d o n o t m a k e it p o s s i b l e to a c c o u n t for the difference b e t w e e n
the p r o s o d i c host a n d the syntactic host of a clitic, w h i c h are not n e c e s s a r i l y iden-
tical, as K l a v a n s h a s a r g u e d c o n v i n c i n g l y [cf. ( 2 7 ) ] . T h i s is only p o s s i b l e b y m a k -
ing use of a separate p r o s o d i c s u b c a t e g o r i z a t i o n .
In fact, it is u n l i k e l y that w e n e e d syntactic s u b c a t e g o r i z a t i o n at all for clitics.
In c a s e s such as the D u t c h clitic d i s c u s s e d a b o v e , the p r o n o m i n a l clitic s h o w s u p
only in p l a c e s w h e r e the i n d e p e n d e n t l y n e e d e d p h r a s e structure p r i n c i p l e s of
D u t c h w o u l d a l l o w p r o n o u n s . Similarly, in the c a s e s of the G r e e k definite article
ho a n d the K w a k w a l a d e t e r m i n e r particles that K l a v a n s d i s c u s s e s , the clitics s h o w
u p only w h e r e the p h r a s e structure p r i n c i p l e s of t h e s e l a n g u a g e s w o u l d i n d e p e n -
dently allow articles a n d d e t e r m i n e r particles. S i n c e the syntactic p o s i t i o n s of
40 Geert Booij and Rochelle Lieber

these clitics follow from the p h r a s e structure rules of the l a n g u a g e s in question, it

w o u l d b e superfluous (and incorrect) to s u b c a t e g o r i z e t h e m for syntactic phrasal
hosts, as K l a v a n s p r o p o s e s to d o . Clitics are prosodically, not syntactically, d e p e n -
dent, and w e p r o p o s e to e x p r e s s this p r o s o d i c d e p e n d e n c e t h r o u g h p r o s o d i c
subcategorization.
N o t e that there are, however, items w h i c h w e w o u l d c o n s i d e r to b e b o n a fide
phrasal affixes, that is, b o u n d m o r p h e m e s w h i c h s u b c a t e g o r i z e for a phrasal host.
L i e b e r ( 1 9 9 2 ) gives a n u m b e r of e x a m p l e s of phrasal affixes, including the E n g l i s h
p o s s e s s i v e m a r k e r -s w h i c h K l a v a n s a s s u m e s to b e a clitic. W h e r e a s clitics can
h a v e w o r d s of different syntactic c a t e g o r i e s as p r o s o d i c hosts, a real phrasal affix
such as the p o s s e s s i v e suffix -s o c c u r s only with p h r a s e s of a specified type, in
this c a s e N P ; prosodically it is a b s o r b e d into the closest p h o n o l o g i c a l w o r d to its
left, as m o s t suffixes w i t h o u t special p r o s o d i c subcategorization are.
W e c o n s i d e r then that it is correct to c h a r a c t e r i z e simple clitics as items w h i c h
are syntactically i n d e p e n d e n t , but p r o s o d i c a l l y d e p e n d e n t , and therefore that a
theory that allows s i m u l t a n e o u s reference to p r o s o d i c a n d m o r p h o l o g i c a l structure
is superior to o n e that d o e s not.

4. CONCLUSION

W e h a v e a r g u e d in this article that there are a n u m b e r of r e a s o n s to believe that

p r o s o d i c structure and m o r p h o l o g i c a l structure m u s t b e built in t a n d e m . T h e r e are
p h o n o l o g i c a l rules that m u s t refer to b o t h sorts of structure s i m u l t a n e o u s l y and
affixes w h o s e s u b c a t e g o r i z a t i o n s m u s t b e b i p l a n a r as well. A s s u m i n g simultaneity
of p r o s o d i c and m o r p h o l o g i c a l structure allows us to e l i m i n a t e the notion of " h e a d
o p e r a t i o n " from m o r p h o l o g y , to a c c o u n t simply for several sorts of b r a c k e t i n g
p a r a d o x e s , and to c h a r a c t e r i z e s i m p l e clitics in an a p p r o p r i a t e way. Finally, by
using the notions of m o r p h o l o g i c a l a n d p r o s o d i c subcategorization w e can arrive
at a t y p o l o g y of m o r p h e m e s that allows us to distinguish clitics from both free
m o r p h e m e s and affixes.

ACKNOWLEDGMENT

The second author would like to acknowledge the generous support of the NWO, the
Netherlands Organization for Scientific Research, during the time that this article was
written.
Morphological and Prosodic Structure 41

NOTES

1
Cf. Anderson (1975), Booij (1988), Booij and Rubach (1984), and Cohn (1989). Simi-
lar ideas have been developed in an unpublished dissertation by Inkelas (1989).
2
See Jackendoff (1987), Levin and Rappaport (1986), and Rappaport and Levin (1988)
for discussion of LCS, PAS, and the relationship between them.
3
Note that even the grid-only theory of word stress requires that information about the
syllabification of words be available.
4
It is probably useful at this point to discuss some conceivable alternatives to the analy-
sis proposed here. First, note that the difference in stress pattern between, e.g., sociologie
and preventie cannot be predicted on the basis of the segmental composition of these
words. All present analyses of Dutch stress (e.g., Van der Hulst, 1984; Kager, 1989) assume
that in the normal case main stress falls on the penultimate syllable of words ending in -ie,
and therefore words in -ie with final stress have to be marked diacritically with a feature,
say [ + F ] , that takes care of this. Note, however, that we cannot make use of this feature
[ + F ] instead of stress to select the proper suffix, since it is the distributionally more re-
stricted suffix -ief thai requires that its class of base words be characterized, whereas the
words in -ie that are marked by the feature [ + F ] are those that cooccur with the more
general suffix -isch (note that there is no evidence in Dutch that the distribution of -ief is
determined by a diacritic feature [latinate]).
Observe, furthermore, that we cannot derive the adjectives from nominal stems without
-ie such as sociolog- and prevent-, because in that case the property that distinguishes the
bases of -ief and -isch would not be available, since it is located on the final syllable with
[i]. That is, this is a typical case of word-based morphology.
Another conceivable analysis is based on the idea expressed in Chomsky and Halle
(1968) that morphology precedes phonology, as suggested more recently by Halle and
Vergnaud (1987). The facts discussed here might be analyzed within such as theory as
follows. The morphology attaches both -ief and -isch to nouns in -ie. Prosodic structure is
created cyclically on the basis of the morphological structure of the complex words, and
there is a filter that states that words in which the suffix -ief is preceded by a syllable with
main stress are ill-formed. Note, however, that the final [i] of the base noun that bears main
stress before the suffix is added is deleted by rule before suffixes beginning with [i]. There-
fore the filter could only do its work if it applied before the application of the [i]-deletion
rule. Similarly, the filter would also have to apply before the application of the stress rules
that derive the stress pattern of the adjectives, because otherwise the crucial information
would get lost. That is, the filter cannot function as a prosodic well-formedness condition
on the surface form of these adjectives, as one would expect from filters. One could of
course envision a theory in which filters could be cyclic checking mechanisms, but such a
theory would be far less restrictive than the theory of lexical phonology we assume here; it
would, for example, leave the way open for the ordering of filters after particular rules in a
cycle. Thus, the filter approach that one is forced to accept here, if one rejects the basic
tenet of lexical phonology, seems to be completely ad hoc.
A final alternative analysis of the -iefl-isch facts might seem to be the following. We
might assume a surface filter at the end of the lexicon for checking the stress patterns of
42 Geert Booij and Rochelle Lieber

words with these suffixes, with a postlexical rule deleting [i] before [i]. This alternative is
not correct either, however; such a postlexical filter would incorrectly apply to words end-
ing in [i] followed by the clitic ie [i], for example, wie-ie is 'who he is' (note that -ie forms
one prosodic word with the preceding word).
Alternatively, one might assume a phonological rule that inserts ej in this context.
5

Note, however, that this cannot be a general rule of epenthesis, because ej is only inserted
in comparatives. It is therefore more natural to analyze these facts in terms of two compet-
ing suffixes. If one preferred to assume a phonological insertion rule here, this would still
make the point that lexical phonological rules have to refer to both morphological and
prosodic structure.
Parallel to the discussion above with respect to Dutch, one might consider an alternative
analysis in which a filter forbids the long allomorph to occur after a syllabified consonant.
Again, such a filter could not be a condition on the surface form of these words, because at
the surface all consonants will be syllabified due to the recursive application of syllabifi-
cation procedures.
Note that the examples which we discuss below provide direct evidence against the
6

claim in Cohn (1989:197) that, in languages which have prosodic structure not isomorphic
with morphological structure, the phonology will not refer to morphological structure.
This section is adapted from Lieber, Deconstructing Morphology. Word Formation in
7

Syntactic Theory, with permission from the publisher, the University of Chicago Press.
Copyright © 1992 by the University of Chicago.
Thanks to Patrocinio Schweikart for the Tagalog data. Further evidence that pang-
8

forms are adjectives is that they can occur in the position of modifiers of nouns, as in papel
pang-sulat 'paper for writing'.
McCarthy and Prince (1986) do not state the facts below in terms of morphological
9

subcategorization, so here we are taking the liberty of translating their basic idea into the
morphological framework we have adopted.
M c C a r t h y and Prince (1986) label this reduplication "Verbal Reduplication," but in
10

Dixon (1977) these examples are given as examples of Nominal Reduplication.

See also Hoeksema (1987) for a critical appraisal of Pesetsky's (1985) proposal.
1 1

See also Booij and Rubach (1987) and the references cited there for data concerning
1 2

Dutch clitics.
The general distribution of ie, as with other pronouns, is accounted for by syntactic
1 3

principles such as X-bar theory, 9-theory, case theory, and so on.

REFERENCES

Anderson, S. (1975). On the interaction of phonological rules of various types, Journal of

Linguistics 11, 3 9 - 6 2 .
Aronoff, M. (1988). Head operations and strata in reduplication. Yearbook of Morphology
1, 1-15.
Bloomfield, L. (1933). Language. Holt, New York.
Morphological and Prosodic Structure 43

Booij, G. E. (1977). Dutch Morphology: A Study of Word Formation in Generative Gram-

mar. Foris, Dordrecht.
Booij, G. E. (1985). Coordination reduction in complex words: a case for prosodic pho-
nology. In Advances in Non-linear Phonology (H. van der Hulst and N. Smith, eds.),
pp. 1 4 3 - 1 6 0 . Foris, Dordrecht.
Booij, G. E. (1988). On the relation between lexical and prosodic phonology. In Certamen
Phonologicum. Papers from the 1987 Cortona Phonology Meeting (P. M. Bertinetto
and M. Loporcaro, eds.), pp. 6 3 - 7 5 . Rosenberg & Selier, Turin.
Booij, G. E., and Rubach, J. (1984). Morphological and prosodic domains in Lexical Pho-
nology. Phonology Yearbook 1, 1-28.
Booij, G. E., and Rubach, J. (1987). Postcyclic versus postlexical rules in Lexical Pho-
nology. Linguistic Inquiry 18, 1 - 4 4 .
Broselow, E., and McCarthy, J. (1984). A theory of internal reduplication. The Linguistic
Review 3, 2 5 - 8 8 .
Chomsky, N., and Halle, M. 1968. The Sound Pattern of English. Harper & Row, New
York.
Cohn, A. (1989). Stress in Indonesian and bracketing paradoxes. Natural Language and
Linguistic Theory 7, 167-216.
Dixon, R. M. W. (1977). A Grammar of Yidiny Cambridge University Press, Cambridge.
Hall, T. (1989). German syllabification, the velar nasal, and the representation of schwa.
Linguistics 27, 8 0 7 - 8 4 2 .
Halle, M., and Vergnaud, J.-R. (1987). An Essay on Stress. MIT Press, Cambridge, Mass.
Hayes, B. (1982). Extrametricality and English stress. Linguistic Inquiry 13, 2 2 7 - 2 7 6 .
Hoeksema, J. (1985). Categorial Morphology Garland, New York.
Hoeksema, J. (1987). Relating word structure and logical form. Linguistic Inquiry 18,
119-126.
Hulst, H. van der. (1984). Syllable Structure and Stress in Dutch. Foris, Dordrecht.
Inkelas, S. (1989). Prosodic Constituency in the Lexicon. Doctoral dissertation, Stanford
University, Stanford, Calif.
Jackendoff, R. (1987). The status of thematic relations in linguistic theory. Linguistic In-
quiry 18, 3 6 9 - 4 1 2 .
Kager, R. W. J. (1989). A Metrical Theory of Stress and Destressing in English and Dutch.
Doctoral dissertation, University of Utrecht, Utrecht, Netherlands.
Kiparsky, P. (1982). Lexical morphology and phonology. In Linguistics in the Morning
Calm (Linguistic Society of Korea, ed.), pp. 3 - 9 1 . Hanshin, Seoul.
Kiparsky, P. (1985). Some consequences of lexical phonology. Phonology Yearbook 2,
85-138.
Klavans, J. (1985). The independence of syntax and cliticization. Language 61, 9 5 - 1 2 0 .
Levin, B., and Rappaport, M. (1986). The formation of adjectival passives. Linguistic In-
quiry 17, 6 2 3 - 6 6 2 .
Lieber, R. (1980). On the Organization of the Lexicon. Doctoral dissertation, Massachu-
setts Institute of Technology, Cambridge.
Lieber, R. (1989). On percolation. Yearbook of Morphology 2, 9 5 - 1 3 8 .
Lieber, R. (1992). Deconstructing Morphology: Word Formation in Syntactic Theory. Uni-
versity of Chicago Press, Chicago.
44 Geert Booij and Rochelle Lieber

Marie, J. van. (1985). On the Paradigmatic Dimension of Morphological Creativity. Foris,

Dordrecht.
McCarthy, J., and Prince, A. (1986). Prosodic Morphology. Unpublished manuscript, Uni-
versity of Massachusetts, Amherst, and Brandeis University, Waltham, Mass.
Nespor, M., and Vogel, I. (1986). Prosodic Phonology. Foris, Dordrecht.
Pesetsky, D. (1985). Morphology and logical form. Linguistic Inquiry 16, 1 9 3 - 2 4 8 .
Rappaport, M., and Levin, B. (1988). What to do with 0-roles. In Syntax and Semantics 21
(W. Wilkins, ed.), pp. 7 - 3 6 . Academic Press, New York.
Rubach, J., and Booij, G. (1990). Syllable structure assignment in Polish. Phonology 7,
121-158.
Schachter, P., and Otanes, F. (1972). Tagalog Reference Grammar. University of California
Press, Berkeley.
Selkirk, E. (1984). Phonology and Syntax: The Relation between Sound and Structure.
MIT Press, Cambridge, Mass.
Sproat, R. (1985). On Deriving the Lexicon. Doctoral dissertation, Massachusetts Institute
of Technology, Cambridge.
Sproat, R. (1988). Bracketing paradoxes, cliticization, and other topics. In Morphology and
Modularity (M. Everaert, A. Evers, R. Huybregts, and M. Trommelen, eds.), pp. 3 3 9 -
360. Foris, Dordrecht.
Visch, E. (1989). The Rhythm Rule in English and Dutch. Doctoral dissertation, University
of Utrecht, Utrecht, Netherlands.
Wiese, R. (1988). Silbische und lexikalische Phonologie. Studien zum Chinesischen und
Deutschen. Niemeyer, Tubingen.
Zwicky, A. (1977). On Clitics. Distributed by the Indiana University Linguistics Club,
Bloomington.
MODELING THE PHONOLOGY-MORPHOLOGY
INTERFACE

SHARON HARGUS
Department of Linguistics
University of Washington
Seattle, Washington 98195

1. I N T R O D U C T I O N

T h i s article offers a defense of interactionism, the h y p o t h e s i s that p h o n o l o g i c a l

rules m a y p r e c e d e m o r p h o l o g i c a l rules. A s d i s c u s s e d b e l o w a n d also b y K a i s s e
and H a r g u s (this v o l u m e ) , i n t e r a c t i o n i s m h a s b e e n c h a l l e n g e d b y a n u m b e r of
recent authors, including Halle a n d V e r g n a u d (1987a, 1987b) and O d d e n (this
v o l u m e ) . In this article I h a v e tried to a s s e m b l e the best available e v i d e n c e for
interactionism, s u m m a r i z e d in S e c t i o n 3 . In Section 4 , 1 c o n s i d e r the theoretical
implications of the a n a l y s e s d i s c u s s e d in Section 2 a n d p o s s i b l e r e a n a l y s e s of
t h e s e data. I begin w i t h a brief d i s c u s s i o n of interactionism and n o n i n t e r a c t i o n i s m .

2. R E C E N T T H E O R I E S O F P H O N O L O G Y - M O R P H O L O G Y
INTERACTION

2.1. Lexical Phonology M a y Precede A n y Morphology

A s is well k n o w n , the h y p o t h e s i s that p h o n o l o g y a n d m o r p h o l o g y apply in the

s a m e c o m p o n e n t is traceable to an u n p u b l i s h e d p a p e r by Pesetsky ( 1 9 7 9 : 4 8 ) (cf.
Pesetsky, 1985): W e p r o p o s e that the p r o c e s s of w o r d formation consists of the
following s t e p s : . . . A p p l y an affix . . . to a b a s e . . . A p p l y all cyclic p h o n o l o g i c a l
45
Phonetics and Phonology, Volume 4 Copyright © 1993 by Academic Press, Inc.
Studies in Lexical Phonology All rights of reproduction in any form reserved.
46 Sharon Hargus

rules, subject to the Strict C y c l e C o n d i t i o n . . . E r a s e inner b r a c k e t s , according to

the B E C . F o l l o w i n g t h e recognition that p h o n o l o g i c a l rules c a n b e broadly d i -
vided into t w o t y p e s , lexical a n d postlexical, P e s e t s k y ' s h y p o t h e s i s c o u l d b e r e -
stated as in (1).

(1) L e x i c a l p h o n o l o g i c a l rules m a y p r e c e d e m o r p h o l o g i c a l rules.

P e s e t s k y ' s m o d e l m i g h t b e called the " s t a n d a r d " theory of p h o n o l o g y - m o r p h o l -

o g y interaction in lexical p h o n o l o g y , incorporating several h y p o t h e s e s — l e x i c a l
p h o n o l o g i c a l rules apply cyclically in t h e u n m a r k e d case, t h e lexical p h o n o l o g y
m a y p r e c e d e ( s o m e ) w o r d formation, a n d word-internal structure is invisible t o
certain later p r o c e s s e s .
H y p o t h e s i s (1) predicts that any lexical p h o n o l o g i c a l rule w h i c h b e l o n g s t o t h e
s a m e o r an earlier p h o n o l o g i c a l d o m a i n as a m o r p h o l o g i c a l rule should b e able to
p r e c e d e that m o r p h o l o g i c a l rule. S o m e w h a t m o r e restrictive versions of this hy-
p o t h e s i s h a v e b e e n p r o p o s e d b y K i p a r s k y ( 1 9 8 5 ) , Aronoff ( 1 9 8 8 ) , and B o r o w s k y
(this v o l u m e ) .
A c c o r d i n g t o K i p a r s k y ( 1 9 8 5 ) , there a r e universally only t w o word-internal
p h o n o l o g i c a l d o m a i n s , L E V E L 1 o r t h e STEM L E V E L , a n d t h e W O R D L E V E L , w i t h
the stem level b e i n g cyclic a n d the w o r d level b e i n g n o n c y c l i c . K i p a r s k y ' s version
of the m o d e l thus p r e d i c t s that stem-level p h o n o l o g i c a l rules m a y p r e c e d e stem-
and word-level m o r p h o l o g i c a l rules, but that word-level p h o n o l o g i c a l rules should
follow word-level m o r p h o l o g i c a l rules, since t h e latter d o m a i n is n o n c y c l i c .
B o r o w s k y (this v o l u m e ) p r o p o s e s not only that the word-level p h o n o l o g y is n o n -
cyclic, b u t that word-level p h o n o l o g y p r e c e d e s word-level m o r p h o l o g y , b a s i n g
this c o n c l u s i o n o n a n u m b e r of studies of different dialects of E n g l i s h in w h i c h
m o n o m o r p h e m i c f o r m s a n d t h o s e with word-level suffixes pattern together as o p -
p o s e d t o f o r m s with stem-level affixes.
T h e m o d e l p r o p o s e d b y Aronoff ( 1 9 8 8 ) , a slightly evolved version of Aronoff
( 1 9 7 6 ) , is very similar t o these m o d e l s . Aronoff ( 1 9 7 6 ) , a s s u m i n g the distinction
b e t w e e n t w o kinds of p h o n o l o g i c a l rules, cyclic a n d word-level ( = postcyclic,
last-cyclic), p r o p o s e d in C h o m s k y a n d H a l l e (1968) (henceforth SPE), h a d sug-
gested that p h o n o l o g i c a l and m o r p h o l o g i c a l rules m a y interact with e a c h other in
limited w a y s , with m o r p h o l o g i c a l rules restricted t o a p p l y i n g in o n e of three
p l a c e s in the p h o n o l o g y : "first, before t h e [cyclic] p h o n o l o g y . . . ; second, before
the word-level rules; third, after t h e p h o n o l o g y " ( p . 7 3 ) , with t h e first t y p e of
m o r p h o l o g i c a l rule a p p l y i n g in a b l o c k p r i o r t o the cyclic p h o n o l o g y .
Aronoff ( 1 9 8 8 ) is p r i m a r i l y c o n c e r n e d with s h o w i n g h o w cases of a p p a r e n t
" m i s a p p l i c a t i o n " of p h o n o l o g i c a l rules (under- a n d over-application) in redupli-
cation can b e a c c o u n t e d for w i t h o u t the nonlinear, s i m u l t a n e o u s m o d e l s of redu-
plication that have b e e n p r o p o s e d b y C l e m e n t s ( 1 9 8 5 ) , M e s t e r ( 1 9 8 8 ) , a n d U h r -
b a c h ( 1 9 8 7 ) . H e m a i n t a i n s that m i s a p p l i c a t i o n can b e a c c o u n t e d for b y allowing
" a restricted t y p e of m o r p h o l o g i c a l rule t o follow certain p h o n o l o g i c a l r u l e s "
(p. 4 ) . A s in t h e m o d e l of K i p a r s k y ( 1 9 8 5 ) , t w o word-internal p h o n o l o g i c a l d o -
Phonology-Morphology Interface 47

m a i n s are posited as universal (stem- and w o r d - l e v e l ) , a n d m o r p h o l o g i c a l p r o -

c e s s e s are classes as stem-level or word-level d e p e n d i n g o n their p h o n o l o g i c a l
d o m a i n a s s i g n m e n t . Putting aside the question of the e x i s t e n c e of m o r p h o l o g y
w h i c h follows word-level p h o n o l o g y , Aronoff's m o d e l (both 1976 a n d 1988 ver-
sions) m a k e s t w o p r e d i c t i o n s .

1. L a n g u a g e s m a y contain m a x i m a l l y t w o w o r d - i n t e r n a l p h o n o l o g i c a l d o -
m a i n s , to w h i c h m o r p h o l o g i c a l p r o c e s s e s are assigned.
2. Stem-level (cyclic, lexical) p h o n o l o g i c a l rules m a y p r e c e d e word-level, but
not stem-level, m o r p h o l o g y .

T h e m a i n difference b e t w e e n K i p a r s k y ' s and B o r o w s k y ' s lexical p h o n o l o g y m o d -

els o n the o n e h a n d , versus that of Aronoff ( 1 9 8 8 ) , is that Aronoff's m o d e l predicts
the a b s e n c e of a n a l y s e s in w h i c h stem-level p h o n o l o g y p r e c e d e s stem-level m o r -
phology, w h e r e a s the m o d e l s of K i p a r s k y a n d B o r o w s k y at least implicitly predict
that they d o e x i s t . 1

2.2. P h o n o l o g y M a y N o t P r e c e d e A n y M o r p h o l o g y

A n u m b e r of p h o n o l o g i s t s (Halle, H a r r i s , a n d V e r g n a u d , 1 9 9 1 ; H a l l e a n d Verg-
n a u d , 1987a, 1987b; O d d e n , this v o l u m e ; S z p y r a , 1987) h a v e recently e s p o u s e d a
resurrection of t h e p r e v i o u s " s t a n d a r d " t h e o r y of p h o n o l o g y - m o r p h o l o g y inter-
action, that p r o p o s e d in SPE a n d further e l a b o r a t e d b y H a l l e ( 1 9 7 3 ) . T h o u g h dif-
fering in details, all t h e s e " n o n i n t e r a c t i o n i s t " m o d e l s posit a m o r p h o l o g i c a l c o m -
p o n e n t w h i c h p r e c e d e s the (possibly level-ordered) p h o n o l o g i c a l c o m p o n e n t . T h e
noninteractionist position is well s u m m a r i z e d b y H a l l e a n d V e r g n a u d ( 1 9 8 7 a : 7 8 ) .

We make the traditional assumption that [morphological] rules are the province of a spe-
cial module, the morphology. In our theory, then, as in SPE, morphology is distinct and
separate from the phonology. Morphology interacts with phonology in that it creates the
objects on which the rules of phonology operate.

O d d e n (this v o l u m e ) p r o v i d e s t h e b e s t d e f e n s e of n o n i n t e r a c t i o n i s m , e x a m i n i n g
m u c h of the often-cited e v i d e n c e for lexical p h o n o l o g y and c o n c l u d i n g that this
e v i d e n c e can b e r e a n a l y z e d in a w a y w h i c h d o e s n o t support it. O d d e n also offers
an e m p i r i c a l a r g u m e n t against interactionism from M a l t e s e .
However, w h i l e s o m e of the e v i d e n c e usually cited for interactionism can a n d
s h o u l d b e r e a n a l y z e d , I s u g g e s t in the following section that not all the e v i d e n c e
can easily b e d i s m i s s e d .

3. P H O N O L O G Y P R E C E D I N G M O R P H O L O G Y

M c C a r t h y and P r i n c e ( 1 9 9 0 ) d i s c u s s a n u m b e r of a n a l y s e s of m o r p h o l o g i c a l
rules w h i c h apply to a " p r o s o d i c a l l y c i r c u m s c r i b e d " portion of a m o r p h o l o g i c a l
48 Sharon Hargus

b a s e , with the additional restriction that the prosodically defined b a s e b e identical

to the M i n i m a l W o r d in a particular l a n g u a g e . T h e usual c o n c e p t i o n of the M i n i -
m a l W o r d is that of m e t r i c a l foot, a p h o n o l o g i c a l l y defined unit. M c C a r t h y and
P r i n c e a d d u c e a n u m b e r of m o r p h o l o g i c a l rules from various l a n g u a g e s w h i c h
o p e r a t e o n a prosodically c i r c u m s c r i b e d b a s e in this manner. This w e l l - s u p p o r t e d
c o n c l u s i o n p r e s u p p o s e s and thus p r o v i d e s support for interactionism; that is, p h o -
nologically defined c i r c u m s c r i p t i o n of a b a s e p r e c e d e s s o m e m o r p h o l o g i c a l p r o -
cess in a n u m b e r of l a n g u a g e s .
B e l o w I s u m m a r i z e other types of analyses in w h i c h a p h o n o l o g i c a l rule m u s t
p r e c e d e a m o r p h o l o g i c a l o n e , segregating analyses into three tables. In Table 1 I
include the better k n o w n t y p e of analysis, in w h i c h a m o r p h o l o g i c a l p r o c e s s refers
to a d e r i v e d p h o n o l o g i c a l p r o p e r t y [English, Finnish, Koasati, H e b r e w , G e r m a n
(Case 1), D u t c h ( C a s e 1)], followed b y a brief discussion of o n e of these types of
cases. T a b l e 2 c o n t a i n s a short list of l a n g u a g e s ( K i h e h e , T a g a l o g , M e n d e ) in
w h i c h p h o n o l o g i c a l rules are k n o w n to overapply in reduplicative structures.
W h i l e interactionism c a n easily a c c o m m o d a t e these rules, alternative representa-
tional a n a l y s e s are also available. Table 3 contains analyses in w h i c h the d o m a i n
of a p h o n o l o g i c a l p r o c e s s crucially e x c l u d e s a ( p r e s u m a b l y later) m o r p h o l o g i c a l
p r o c e s s , w h i c h , unlike the Table 1 analyses, d o e s not refer to the output of the
p h o n o l o g i c a l rule [ S u n d a n e s e , Sanskrit, Javanese, L u i s e n o , Icelandic, D a n i s h ,
G e r m a n ( C a s e 2), D u t c h ( C a s e 2 ) ] . Since this t y p e of e v i d e n c e for interactionism
is not as well k n o w n as t h o s e s u m m a r i z e d in Tables 1 and 2 , 1 illustrate this type
with several l a n g u a g e s — S u n d a n e s e , G e r m a n , L u i s e n o , and J a v a n e s e .

3.1. Morphology W h i c h Refers to a Derived Phonological Property

T h e p h o n o l o g i c a l p r o c e s s e s in Table 1 are diverse, r a n g i n g from metrical struc-

ture c o n s t r u c t i o n or alteration (stress, syllabification, s y n c o p e , e p e n t h e s i s ) to
c h a n g e s in feature c o n t e n t ( m u t a t i o n ) to m e t a t h e s i s . T h e m o r p h o l o g i c a l o p e r a t i o n s
are equally varied, i n c l u d i n g affixation (English), a l l o m o r p h selection (Finnish,
L a p p i s h , G e r m a n , D u t c h ) , a n d r h y m e deletion (Koasati).

3.1.1. GERMAN

Let us n o w c o n s i d e r o n e of the cases from Table 1 in m o r e detail, stress and

-eil-erei a l l o m o r p h y in G e r m a n . T h e s e data involve the distribution of a l l o m o r p h s
of the deverbal, n o m i n a l i z i n g (with slightly pejorative c o n n o t a t i o n s ) , stem-level
suffix -eil-erei.
Hall ( 1 9 9 0 ) a n a l y z e s t h e distribution of the a l l o m o r p h s of this suffix as follows
(cf. also G i e g e r i c h , 1987): -erei is a d d e d to verb s t e m s with final stressed syl-
lables, including verb s t e m s derived b y the addition of -ier; -ei is a d d e d to other
verb s t e m s . 2
Phonology-Morphology Interface 49

TABLE 1
MORPHOLOGY REFERRING TO A DERIVED PHONOLOGICAL PROPERTY

Language Phonology Morphology Reference

English
Case 1 Stress Expletive infix McCarthy (1982)
Case 2 Stress -ize Marchand (1969), Hayes
(personal communication)
Finnish Stress Illative allomorphy Kiparsky (this volume),
Kanerva (1987), Keyser and
Kiparsky (1984)
Lappish Stress Various Bergsland(1976)
allomorphy
Koasati Syllabification Plural formation Kimball (1982), Martin (1988)
Hebrew Metathesis Extraction Bat-El (1986)
German
Case 1 Stress, -erZ-erei Hall (1990), Giegerich (1987)
epenthesis allomorphy
Dutch
Case 1 Stress -ie/isch Booij and Lieber (this volume)
allomorphy

(2) INFINITIVE NOMINAL

sing-en Sing-erei 'singing'
lauf-en Lauf-erei 'running'
ess-en Ess-erei 'eating'
bdck-en Back-erei 'baking'
widm-en Widm-erei 'dedicating'
ordn-en Ordn-erei 'arranging'
lack-ieren Lackier-erei 'lacquering'
spion-ier-en Spion-ier-erei 'spying'
drbeit-en Arbeit-erei 'working'
trompet-en Trompet-erei 'trumpeting'

T h e v e r b s t e m is p r o v i d e d in the infinitive c o l u m n : the f o r m s in (2) consist of the

verb stem, t h e infinitival suffix /-n/, a n d an e p e n t h e t i c v o w e l (Hall, 1987).
Insofar as the distribution of -erei refers to stress, w h i c h is a d e r i v e d p h o n o -
logical p r o p e r t y in G e r m a n (although, as Hall n o t e s , stress is sensitive to a n a t i v e /
n o n n a t i v e lexical distinction, as are other rules of G e r m a n p h o n o l o g y ) , these d a t a
are p r o b l e m a t i c for O d d e n ' s noninteractionist m o d e l . T h e p r e c e d i n g d a t a require
an analysis in w h i c h stress a s s i g n m e n t p r e c e d e s the selection of -ei or -erei.
Selection of the a l l o m o r p h -ei is illustrated by the liquid-final verb s t e m s in (3),
w h i c h also u n d e r g o a stem-level rule of preliquid s c h w a e p e n t h e s i s ( G i e g e r i c h ,
1987; Hall, 1987):
50 Sharon Hargus

(3) /se:gl/ segel-n Segel-ei 'sailing'

/tro: dl/ trodel-n Trodel-ei 'loitering'
/bii: g l / biigel-n Bugel-ei 'ironing'
/plaudr/ plduder-n Plauder-ei 'chatting'

Both the e p e n t h e s i s rule a n d t h e rule assigning stress to the f o r m s in (3) m u s t

apply prior to the a l l o m o r p h y rule, since e p e n t h e s i s inserts a s c h w a before the
stem-final unsyllabified liquid, t h e r e b y creating a final unstressed syllable w h i c h
d e t e r m i n e s that the -ei a l l o m o r p h b e selected.
A n alternative, p h o n o l o g i c a l a c c o u n t of this a l l o m o r p h y h a s b e e n s u g g e s t e d
to m e b y D o n c a Steriade (personal c o m m u n i c a t i o n ) . S u p p o s e the suffix h a s the
u n d e r l y i n g representation -/arai/. F o l l o w i n g its a t t a c h m e n t to /l/-final s t e m s ,
/ 1 + 9 / —> syllabic [1] ( [ s i ] ) , w h i c h then triggers a dissimilatory deletion of Ixl fol-
l o w i n g the syllabic [1]: / s e : g l + o r a i / —> [ s e : galrai] —> [ s e : g a l a i ] . W h i l e this analy-
sis s e e m s r e a s o n a b l e , Hall (personal c o m m u n i c a t i o n ) notes that there is n o other
e v i d e n c e for liquid dissimilation in G e r m a n , a n d that t h e directionality of t h e rule
/l3/ —» [al] is the o p p o s i t e of that r e q u i r e d to a c c o u n t for other [3] ~ 0 alternations
in G e r m a n (Hall, 1987). Finally, I n o t e that a p h o n o l o g i c a l analysis of these data
s o m e w h a t u n d e r m i n e s t h e notion of m o r p h o l o g y , since the alternation discussed
is confined to a single m o r p h e m e .

3.1.2. L A P P

In L a p p there are suffixal a l l o m o r p h y p r o c e s s e s w h i c h refer to w h e t h e r or not

the final syllable of t h e stem is s t r e s s e d . I m e n t i o n t w o such cases here, allo-
3

m o r p h s of the illative plural suffix a n d a l l o m o r p h s of a n o m i n a l d i m i n u t i v e suf-

fix. All information about L a p p p r e s e n t e d h e r e c o m e s from B e r g s l a n d ' s (1976)
grammar.

3.1.2.1. Stress. A c c o r d i n g to B e r g s l a n d , m a i n stress usually falls o n the leftmost

syllable, with every o t h e r syllable to the right r e c e i v i n g stress as well, with the
exception of word-final syllables, w h i c h are never stressed, even in w o r d s with
odd-numbered syllables. 4

(4) [oap.pa.haed.dji rek.ke.nas.ti.goat.ta] ' t h e teacher starts to c o u n t '

(5) [boar.ra.seb.mu.sat gud.ne.jat.tu.juv.vu.jit] ' t h e oldest are h o n o r e d '

Clearly, L a p p c o n t a i n s an alternating stress rule. I suggest that the lack of w o r d -

final stresses is d u e to a general prohibition against m o n o s y l l a b i c feet, a n d that
these stray, footless syllables as in (5) are adjoined to the foot to their left follow-
ing w o r d formation. W h i l e B e r g s l a n d also n o t e s s o m e w o r d s with exceptional,
lexically specified m o n o - o r trisyllabic feet, w h i c h are said to b e mainly loan-
w o r d s , c o m p o u n d s , a n d w o r d s with frozen derivational suffixes, the regular al-
Phonology-Morphology Interface 51

ternating stress rule a p p e a r s to b e the n o r m . T h u s stress is basically p r e d i c t a b l e

in L a p p .

3.1.2.2. Illative Plural Allomorphy. B e r g s l a n d notes that it is very i m p o r t a n t to

distinguish b e t w e e n w o r d s w h i c h contain o d d - or e v e n - n u m b e r e d syllables in the
description of L a p p w o r d formation. M a n y suffixes h a v e o n e a l l o m o r p h that is
used after stems e n d i n g in an o d d n u m b e r of syllables, a n d a n o t h e r after even
n u m b e r s of syllables. F o l l o w i n g H a y e s ' s ( 1 9 8 2 b ) analysis of Yidiny, I suggest that
these a l l o m o r p h y rules are sensitive to foot s t r u c t u r e — i n particular, w h e t h e r or
not the final syllable of the b a s e e n d s in a foot. S i n c e L a p p h a s an alternating stress
rule, it s e e m s that the suffixal a l l o m o r p h s that B e r g s l a n d refers to are selected in
a c c o r d a n c e with w h e t h e r or not the final syllable of the stem b e l o n g s to a foot.
O n e such a l l o m o r p h y rule is that w h i c h d e t e r m i n e s the s h a p e of the illative
plural suffix. A c c o r d i n g to B e r g s l a n d , the a l l o m o r p h -ide is used after stems with
even n u m b e r s of syllables (i.e., u n s t r e s s e d stem-final syllables), w h e r e a s -ida is
u s e d after s t e m s with o d d n u m b e r s of syllables (i.e., stressed stem-final syllables).

(6) -ide ALLOMORPH:

[cie.ga-ide] ' c o r n e r (ill.pl.)'
[boa.lo-ide] ' b u t t o n (ill.pl.)'
[rerj.ku-ide] 'stool (ill.pl.)'

(7) -ida ALLOMORPH:

[mal.la.si-ida] 'feed (ill.pl.)'
[baed.na.gi-ida] ' d o g (ill.pl.)'

3.1.2.3. Diminutives. O n e t y p e of n o m i n a l d i m i n u t i v e in L a p p is f o r m e d b y add-

ing a suffix w h o s e s h a p e also varies a c c o r d i n g to b a s e syllable count. T h e f o r m of
the suffix is -(a)s in the singular. In the plural, the suffix h a s the s h a p e -zat follow-
ing e v e n - n u m b e r e d s t e m s (also r e q u i r i n g the w e a k s t e m g r a d e a n d a stem-internal
vowel height alternation) (8), b u t -(a)zzat (with the strong g r a d e of the n o u n stem)
following o d d - n u m b e r e d s t e m s (9).

(8) -zzat ALLOMORPH:

[bae.na.ga-zzat] 'dogs (dim.)'
[us.ti.ba-zzat] 'friends ( d i m . ) '

(9) -zat ALLOMORPH :

ful.ke.zat] 'relatives ( d i m . ) '
[ja.ga-zat] 'rivers ( d i m . ) '
[viel.lja-zat] 'sisters ( d i m . ) '

3.1.2.4. Summary. In the analysis p r o p o s e d here, the a p p a r e n t reference to syl-

lable c o u n t in the a b o v e - m e n t i o n e d a l l o m o r p h y rules is interpreted as a reference
52 Sharon Hargus

to foot structure. T h u s stress a s s i g n m e n t to stems m u s t p r e c e d e selection of the

a l l o m o r p h s of the suffixes d i s c u s s e d h e r e . M a n y m o r e a l l o m o r p h y rules of this
sort (e.g., attributive adjective suffix a l l o m o r p h y ) are found in L a p p .

3.1.3. HEBREW

M o d e r n H e b r e w as d e s c r i b e d b y B a t - E l ( 1 9 8 6 ; cf. also Bat-El, 1989) also p r o -

vides s u p p o r t for interactionism. T h e relevant m o r p h o l o g i c a l p r o c e s s involves a
s o m e w h a t typologically u n u s u a l p r o c e s s B a t - E l t e r m s Extraction, a p r o d u c t i v e
w a y of d e r i v i n g m e m b e r s of o n e lexical c a t e g o r y from another, as in the following
d e n o m i n a l or deadjectival verbs.

(10) [koxav] 'star'

[kixev] ' t o star'

(11) [telefon] 'telephone'

[tilfen] 'to telephone'

(12) [varod] 'pink'

[hivrid] 'to become pink'

Extraction is d e s c r i b e d as an essentially d e l i n k i n g p r o c e s s , o n e w h i c h separates

the c o n s o n a n t a l root tier from the syllable structure. F o l l o w i n g E x t r a c t i o n of the
c o n s o n a n t a l tier from a lexical item, such as a n o u n , the c o n s o n a n t i s m is r e a s s o -
ciated with a verb t e m p l a t e .
Bat-El s h o w s that the c o n s o n a n t i s m of derivational, but not inflectional, affixes
is E x t r a c t e d a n d s u b s e q u e n t l y reassociated to a p p e a r in derived f o r m s . In the fol-
l o w i n g e x a m p l e , both the initial and final /t/'s in 'pattern' are affixes; b o t h a p p e a r
in the d e r i v e d verb.

(13) [tavnit] 'pattern'

[tivnet] 'to structure'

(14) [toxnit] 'plan'

[tixnen] ' t o plan'
[tixnet] ' t o p r o g r a m (a c o m p u t e r ) '

In the s e c o n d e x a m p l e , the older d e n o m i n a l v e r b ' t o plan' is E x t r a c t e d from the

n o m i n a l root w i t h o u t the f e m i n i n e l-il suffix, w h e r e a s the inclusion of the l-il
suffix in the m o r e n e w l y d e r i v e d verb c o n t r i b u t e s to the s e m a n t i c difference b e -
tween 'to plan' and ' t o p r o g r a m ' .
Of greater interest for present p u r p o s e s , Bat-El also s h o w s that the output of a
p r o d u c t i v e p h o n o l o g i c a l rule of M e t a t h e s i s is also reflected in Extracted forms.
M e t a t h e s i s , the i n t e r c h a n g e lil a n d / s / , applies in m o r p h o l o g i c a l l y c o m p l e x forms
only. T h e failure of M e t a t h e s i s to apply in n o n d e r i v e d f o r m s can b e seen in (15)
and (16).
Phonology-Morphology Interface 53

(15) [tsisa] 'fermentology'

(16) [hitsis] 'to ferment'

T h e viability of M e t a t h e s i s as a s y n c h r o n i c rule can b e illustrated with forms

w h i c h contain the verbal t e m p l a t e / h i t C a C ( C ) e C / . Bat-El notes that n e w l y derived
verbs w h i c h m a t c h this t e m p l a t e u n d e r g o M e t a t h e s i s , as s h o w n in (17), for
example.

(17) [zanav] 'tail'

/hit-zanev/ —> [hizdanev] 'to plod along'

T h e following e x a m p l e illustrates that E x t r a c t i o n applies to the o u t p u t of M e t a -

thesis, since in the d e r i v e d a g e n t i v e n o u n the o r d e r of the c o n s o n a n t s is [st]:

(18) /ski/ (root)

/ h i t - s a k e l / —> [histakel] 'to observe'
[staklan] 'observer'

M e t a t h e s i s m u s t apply to the verb ' o b s e r v e ' , rather than to the derived n o u n , on

a c c o u n t of the m o n o m o r p h e m i c status of t h e latter. S o m e w h a t m o r e controver-
5

sially, B a t - E l also s u g g e s t s that E x t r a c t i o n applies to the output of a p h o n o l o g i c a l

rule of Spirantization.

3.2. O v e r a p p l i c a t i o n of P h o n o l o g i c a l R u l e s

T h e s e c o n d class of c a s e s w h i c h are potentially supportive of interactionism are

relatively well k n o w n . T h e s e are c a s e s in w h i c h p h o n o l o g i c a l rules OVER APPLY in
reduplicative structures; for instance, T a g a l o g / p a r j + p u t u l / —> ( p h o n o l o g y : Nasal
Substitution) [ p a m u t u l ] —> ( m o r p h o l o g y : r e d u p l i c a t i o n ) [ p a m u m u t u l ] . O n l y a
small s a m p l e of these sorts of cases are listed in Table 2.

TABLE 2

OVERAPPLICATION OF PHONOLOGICAL RULES

Language Phonology Morphology Reference

Tagalog Nasal Substitution, Various reduplication Carrier (1979),

syncope Carrier-Duncan
(1984), French
(1988)
Mende Mutation Reduplication Innes (1971), Hayes
(1990)
Kihehe Glide Formation Reduplication Odden (this volume),
Odden and Odden
(1986)
54 Sharon Hargus

T h e rules in Table 2 constitute w e a k e r e v i d e n c e for interactionism, since there

are c o m p e t i n g , r e p r e s e n t a t i o n - b a s e d e x p l a n a t i o n s for overapplication in this c o n -
text, p r o p o s e d b y M e s t e r ( 1 9 8 8 ) and C l e m e n t s ( 1 9 8 5 ) .
C l e m e n t s ' s p r o p o s a l is that reduplication is a m u l t i s t a g e operation, s o m e of
w h o s e structural c h a n g e s are m o r p h o l o g i c a l and s o m e p h o n o l o g i c a l [in particular,
s e q u e n c i n g of the reduplicative affix with r e s p e c t to the b a s e (left, right, or inter-
nally) w o u l d b e c o n s i d e r e d p h o n o l o g i c a l ] . M e s t e r ' s m o d e l is similar. A d o p t i n g
this t y p e of a p p r o a c h , O d d e n (this v o l u m e ) a r g u e s that the a p p a r e n t overapplica-
tion of a p h o n o l o g i c a l rule of g l i d e formation in K i h e h e reduplicative f o r m s can
b e a c c o u n t e d for in a noninteractionist m o d e l if transfer of the b a s e m e l o d y to the
reduplicative t e m p l a t e is c o n s i d e r e d to b e a p h o n o l o g i c a l operation. T h u s the glide
formation p r o c e s s d o e s not crucially p r e c e d e the m o r p h o l o g i c a l operation. H o w -
ever, it s e e m s p r e m a t u r e to disallow all the c a s e s in Table 2, as there s e e m s to
b e s o m e d i s a g r e e m e n t o n the best n o n l i n e a r representation of reduplication (cf.
Steriade, 1988; H a y e s and A b a d , 1989; Aronoff, 1988).

3 . 3 . T h e D o m a i n of a P h o n o l o g i c a l R u l e E x c l u d e s a M o r p h o l o g i c a l P r o c e s s

C o n s i d e r n e x t a third class of c a s e s w h i c h p r o v i d e e v i d e n c e for interactionism

(Table 3). A s p o i n t e d out b y S t e r i a d e (personal c o m m u n i c a t i o n ) , many, but not
all, of these c a s e s involve p r o s o d i c , n o n c o n c a t e n a t i v e t y p e s of m o r p h o l o g y , such
as r e d u p l i c a t i o n o r infixation, a p o i n t I return to in Section 4 . 1 n e x t review several
of these c a s e s in m o r e detail.

TABLE 3

DOMAIN OF PHONOLOGICAL R U L E EXCLUDES MORPHOLOGICAL PROCESS

Language Phonology Morphology Reference

Sundanese Nasal Harmony Plural infix Robins (1957), Cohn

(1989)
Sanskrit Ruki rule 'Augment' infix Murti(1984), Kiparsky
(1982)
Luiseno Spirantization Various reduplication Munro and Benson (1973),
Davis (1976)
Javanese A-Raising, Elative, reduplication Dudas(1974, 1975)
Laxing
Icelandic Syllabification Subtraction Kiparsky (1984), Stong-
Jensen(1987)
Danish Lengthening Subtraction Anderson (1975), Sobel
(1981)
German
Case 2 Nasal Assim., Various suffix Hall (1990)
g-Deletion
Dutch
Case 2 Shwa Deletion -eur/-ris allomorphy Booij (1981)
Phonology-Morphology Interface 55

3.3.1. SUNDANESE

I b e g i n with a fairly s i m p l e case. In S u n d a n e s e , there is a rule of r i g h t w a r d -

s p r e a d i n g N a s a l H a r m o n y , w h i c h s p r e a d s nasality from nasal c o n s o n a n t s to
v o w e l s , b l o c k e d only by s u p r a l a r y n g e a l c o n s o n a n t s . T h e following d a t a are from
Conn (1989).

(19) /jiiar/ [Jtfar] 'seek (active)'

/jiaian/ [jiaian] ' w e t (active)'
/mihak/ [mihak] ' t a k e sides (active)'
/rjatur/ [rjatur] 'arrange (active)'
/rjuliat/ [rjuliat] 'stretch ( a c t i v e ) '
/marios/ [marios] 'examine (active)'
/rj a r a h i t a n / [rj arahitan] 'wound (active)'

A s n o t e d b y C o h n , N a s a l H a r m o n y b o t h p r e c e d e s a n d follows a rule of -al-I

-ar- plural infixation.

(20) STEM SINGULAR PLURAL

/jiiar/ [rjlar] [jialiar] ' s e e k (active)'
/niis/ [nTTs] [narl?ls] 'relax in a cool p l a c e '
/jiaian/ [jiaian] [jiaraian] 'wet (active)'

C o m p a r i n g t h e t w o sets of data, n o t e that t h e o c c u r r e n c e of nasal v o w e l s follow-

ing the liquid of the plural suffix is u n u s u a l . H o w e v e r , as C o h n suggests, the dis-
tribution of nasal v o w e l s is easily a c c o u n t e d for if N a s a l H a r m o n y p r e c e d e s ( a n d
follows) infixation.

(21) /jiaian/
Nasal Harmony jiaian
Plural infixation ji-ar-alan
Nasal Harmony ji-ar-aian

T h e p l a c e m e n t of Sanskrit in Table 3 is for c o m p l e t e l y parallel r e a s o n s , involv-

ing the w e l l - k n o w n p h o n o l o g i c a l ruki rule a n d various infixation rules.

3.3.2. GERMAN

T h e following d a t a (from H a l l , 1 9 9 0 ; cf. also B o r o w s k y , this v o l u m e ) involve

forms c o n t a i n i n g t h e velar nasal, w h i c h is traditionally a n a l y z e d as derived from
an / N g / cluster via rules of N a s a l A s s i m i l a t i o n a n d ^ - D e l e t i o n . I suggest that t h e
latter rule m u s t follow t h e suffixation of certain vowel-initial m o r p h e m e s b u t p r e -
c e d e t h e suffixation of other vowel-initial m o r p h e m e s .
T h e m o n o m o r p h e m i c f o r m s in (22) contain surface [rjg].

(22) Tango [tarjgo] 'tango'

Evangelium [evarjgelium] 'gospel'
56 Sharon Hargus

Ganges [garjges] 'Ganges (River)'

Singular [zirjgulae] 'singular'
Kongo [korjgo] 'Congo'
Ungarn [urjgarn] 'Hungary'
Singapur [zirjgapurj] 'Singapore'
Angina [arjgfna] 'angina'

[g] surfaces in these forms b e c a u s e the rule of g-Deletion is a n a l y z e d as apply-

ing only to syllable-final Igl ( p r e c e d e d b y [rj]), a n d in t h e forms in (22), [g] occurs
in an onset. In t h e forms in (23), [g] is also p r e s e r v e d b e c a u s e a stem-level vowel-
initial suffix h a s b e e n a d d e d to t h e root, thereby c a u s i n g t h e Igl to b e syllabified
as t h e onset of t h e suffix syllable.

(23) tang-ier-en [tarjgfiran] 'to t o u c h '

fing-ier-en [firjgiiran] 'to f a k e '
prolong-ier-en [prolorjgiiran] 'to prolong'
laryng-al [larurjgail] 'laryngeal'
Laryng-itis [larurjgf.-tis] 'laryngitis'
Angl-ist [arjglist] 'anglicist'
angl-isier-en [arjglizfrren] 'anglicize'

H o w e v e r , w h e n a word-level vowel-initial suffix is added, [g] is deleted even

w h e r e it w o u l d form a p e r m i s s i b l e onset, as in the forms in ( 2 4 ) .

(24) Spreng-ung [Jprerjurj] 'explosion'

lang-lich [lerjlig] iongish'
Jung-ling jiirjlirj] 'youth'
hungr-ig [hurjric] 'hungry'

Clearly, if all surface instances of t h e velar nasal are to b e derived from / N g / ,

then N a s a l A s s i m i l a t i o n a n d g-Deletion m u s t apply to the stem-final / N g / se-
q u e n c e s in (25) prior t o t h e addition of t h e word-level suffixes.

(25) //pre N g/
Nasal Assimilation
g-Deletion 0
word-level m o r p h o l o g y -ung

Hall (1990) a r g u e s that g - D e l e t i o n is a cyclic, stem-level rule, a n d that m o r -

phological roots are n o t cyclic d o m a i n s [thereby prohibiting g-Deletion from
a p p l y i n g to t h e f o r m s in ( 2 3 ) ] . H o w e v e r , the rule fails to apply to stem-final / N g /
s e q u e n c e s in diphthongier-en [drftorjgiiran] ' d i p h t h o n g i z e ' a n d monophthon-
gier-en [monoftorjgiiran] ' m o n o p h t h o n g i z e ' , w h i c h are derived with stem-level
m o r p h e m e s . B o r o w s k y (this v o l u m e ) a n a l y z e s G e r m a n a n d E n g l i s h g-Deletion a s
a word-level rule, w h e r e in h e r m o d e l , all word-level p h o n o l o g y p r e c e d e s word-
level m o r p h o l o g y . T h e crucial aspect of this analysis is that g-Deletion p r e c e d e s
s o m e suffixation rules.
Phonology-Morphology Interface 57

3.3.3. LUISENO

In L u i s e n o , there is a p h o n o l o g i c a l rule of Spirantization, the p r o b l e m a t i c un-

derapplication of w h i c h h a s attracted c o n s i d e r a b l e attention from theoretical p h o -
nologists ( M u n r o a n d B e n s o n , 1 9 7 3 ; A n d e r s o n , 1 9 7 5 ; Aronoff, 1976; M a r a n t z ,
1982; Aronoff, 1 9 8 8 ) . 6

B y m o s t a c c o u n t s , L u i s e n o c o n t a i n s the c o n s o n a n t inventory given in (26).

N o t e that /v 6 / h a v e [ — cont] a l l o p h o n e s ; their status as u n d e r l y i n g fricatives is
debatable.

(26) p t c k kw q qw ?
s S (s) X x w
v a'
mn rj
1 ry w h
O n e i m p o r t a n t p o i n t on w h i c h analysts d i s a g r e e is the p h o n e m i c status of [s], a
m a r g i n a l p h o n e m e at best. M u n r o a n d B e n s o n ( 1 9 7 3 ) and D a v i s (1976) a n a l y z e
[c] and [s] as a l l o p h o n e s w h i c h are in c o m p l e m e n t a r y distribution. To a c c o u n t for
the distribution of [ c ] / [ s ] , M u n r o a n d B e n s o n f o r m u l a t e the rule of Spirantization
given in (27).
(27)

I will refer to t h e first subrule of Spirantization as W O R D - F I N A L S P I R A N T I Z A T I O N

and the s e c o n d subrule as DISSIMILATORY S P I R A N T I Z A T I O N . T h e rule a c c o u n t s not
only for the distribution of [c] and [s] b u t also for w i d e s p r e a d alternations, s h o w n
in ( 2 8 ) - ( 2 9 ) .

(28) [s]/ #
/qe:nic7 [qe:nis] ''sqi
squirrel'
/qe:nic-um/ [qe:nicum] ''sqi
squirrels'
/ki:-ca] [kf:ca] ''hhoou s e ( n o m . ) '
/ki:-c/ [kis] ''hoi
house (acc.)'
(29) [§]/ [-cont]
/capomkat/ [capomkat] iiar'
/cacapomkat-um/ [caspumkatum] iiars'
/coka: yla-c/ [ c o k a : ylas] ' w a l k i n g stick (abs.)'
/no-coka: yla/ [ n o s k a : yla] (construct f o r m )
/pu:ci-l/ [pu:cil] 'eye ( a c c . ) '
/ p u : ci-la/ [ p u : sla] 'eye (nom.)'
/yo: vi-c-um-i/ [yo:vismi] 'meadow mice (acc.)'
/po-curo?a/ [pusro?ax] 'his leveling'
/curo?a/ [curo?a]- ' t o level'
58 Sharon Hargus

/ c i k w i : -la/ [cikwi :1a] 'to be sad'

/cikwi: -cikwi: / [cikwiskwi] ' t o suffer'

In s o m e cases, Spirantization is fed b y a stress-sensitive S y n c o p e rule. Stress d o e s

not a p p e a r to b e c o m p l e t e l y p r e d i c t a b l e .
A s noted by M a r a n t z ( 1 9 8 2 ) , m o r p h e m e - i n t e r n a l l y , [c] and [s] are not in per-
fect c o m p l e m e n t a r y distribution, [c] o c c u r s a l m o s t exclusively before [ + c o n t ]
segments.

(30) [pu?eeva] 'his left h a n d '

[wacxat] 'shoe'

[cacwumal] 'Gilia capitata ssp. staminea' (a type of flower)

B u t there are a few e x a m p l e s of [c] before [4-son] s e g m e n t s .

(31) [poxecla] 'its p o i n t (of an a r r o w ) '
[cacmis] ' a stone t o o l '

N o n - a l t e r n a t i n g [s] is rare, b u t a few i n s t a n c e s are s h o w n in (32).

(32) a. [masxai] 'isn't i t ? '

b. [tosnu]- 'to order'
c. [sox] ( e x c l a m a t i o n , indicative of surprise)

(32c) is apparently the only r e c o r d e d form in w h i c h [s] o c c u r s word-initially b e -

fore a vowel. N o t e that in (32a) a n d (32c) [s] o c c u r s before a [ + c o n t ] s e g m e n t ,
not as p r e d i c t e d b y Spirantization. M a r a n t z c o n c l u d e s that an u n d e r l y i n g contrast
b e t w e e n Icl a n d III m u s t b e r e c o g n i z e d , a n d that Spirantization is thus a neutral-
ization rule, restricted to d e r i v e d c o n t e x t s , not simply a rule of allophony. I will
follow M a r a n t z o n this point.
N e x t c o n s i d e r the d o m a i n of Spirantization, w h i c h M u n r o and B e n s o n (1973)
suggest is a late rule, apparently not o r d e r e d before any other p h o n o l o g i c a l rules.
C o m p a r a t i v e e v i d e n c e also supports their view. K r o e b e r a n d G r a c e (1960) note
that a c o n t i g u o u s l a n g u a g e to the north, J u a n e n o ( s p o k e n in w h a t is n o w O r a n g e
C o u n t y ) , " i s little m o r e than a dialect [of L u i s e n o ] a n d m u s t have b e e n largely
intelligible" to its speakers (p. 1; cf. also Miller, 1961). In J u a n e n o ( B e a n and
Shipek, 1978), there is n o rule of Spirantization.

(33) JUANENO LUISENO

[qe: ?ec] [qe: ?es] ' M i s s i o n S a n Luis R e y '
[?axacme] [?axasmay] ' t o w n of M i s s i o n San Luis R e y '

However, there is other e v i d e n c e that Spirantization is not a late rule, in that

Spirantization a p p e a r s to p r e c e d e several m o r p h o l o g i c a l p r o c e s s e s . K r o e b e r and
G r a c e , M u n r o and B e n s o n , and D a v i s (1976) all note certain constructions in
w h i c h Spirantization a p p e a r s not to apply. T h e best s u m m a r y of these is provided
by D a v i s , from w h i c h the e x a m p l e s in s u b s e q u e n t sections are taken.
Phonology-Morphology Interface 59

Adjectival (deverbal) reduplication is the m o s t celebrated e x a m p l e of this (it is

the c a s e d i s c u s s e d b y A n d e r s o n , 1975, for e x a m p l e ) but is by n o m e a n s the only
e x a m p l e . D a v i s r e m a r k s (p. 2 0 1 ) :

The surprising thing about the Munro/Benson article is that the authors approach the
subject as if the failure of the c to s rule is found only in connection with one particular
form of reduplication and furthermore only when this produces adjectives.

Adjectival reduplication h a s b e e n (successfully, I think) r e a n a l y z e d b y M a r a n t z

( 1 9 8 2 ) as reduplicative suffixation, and the n o n a p p l i c a t i o n of Spirantization in
these m o r p h e m e s can b e a c c o u n t e d for b y positing the a p p r o p r i a t e suffixal t e m -
plate. H o w e v e r , w h i l e M a r a n t z ' s solution will a c c o u n t for adjectival reduplication,
it will a p p a r e n t l y not w o r k for the following c a s e s . T h e s e are noted b y M a r a n t z ,
w h o , significantly, suggests that their analysis necessarily involves s o m e sort of
"boundary phenomenon."

Reduplicative Plurals

Plurals m a y b e f o r m e d from n o m i n a l or verbal roots via reduplication of the

initial C V - of the b a s e . K r o e b e r and G r a c e n o t e that this sort of reduplication
a p p e a r s to b e frozen a n d only found on p e r s o n a l n o u n s . O n e of the regular plural
suffixes, -m, -um, or -am, also o c c u r s on these f o r m s .

(34) /surja:-l/ [surja.l] 'woman'

/su-srja-l-um/ [susrjalum] 'women'

(35) /ca-capomka-t-um/ [cacpomkatum] 'liars'

/capomka-t/ [capomkat] 'liar'

Reduplicative Protracted Action Verbs

Protracted action verbs are f o r m e d via reduplication of the first C V C - of the

verbal b a s e (Davis, 1 9 7 6 : 2 0 1 ) .

(36) /nec-neci-q/ [necniciq] ' p a y s in dribs and d r a b s '

/neci/- [neci] 'to pay'

(37) /nuc-nuci-q/ [nucnuciq] 'keeps going and squashing things'

/nuci/ [nuci]- 'to s q u a s h '

C a u s a t i v e s -ki, -xami, -kixa, -kixani

T h e s e data from K r o e b e r a n d G r a c e c o u l d not b e reelicited by D a v i s .

(38) /hakwaci-kixa/ [hakwackixa] 'to h u r r y s o m e o n e ( c a u s . ) '

/hakwaci/ [hakwaci] 'to h u r r y s o m e o n e '

(39) /tuc-kixa/ [tuckixa] 'to tie u p ( c a u s . ) '

/tu:ci/ [tu:ci] 'to tie u p '
60 Sharon Hargus

(40) [non poi n e c k i x a n i q ] T got h i m to m a k e ( s o m e o n e else) pay u p '

[neci] 4
to pay up'

Agentive-forming and Adjectival [-kawut]/[-ku:t]

(41) / m i c i - k u : t/ [ m i c k u : t] 'strangled
/mici/- [mici]- 'to strangle'

(42) /neci-kawut/ [neckawut] 'one who pays'

/neci/- [neci]- 'to pay'

(43) / t u : c i - k u : t/ [tuckurt] ' ( s o m e t h i n g ) w h i c h often gets

entangled'
/tu: c-kawut/ [tuckawut] 'often e n t a n g l e d '
/tuici/ [tu:ci] ' t o tie u p '

T e n s e / A s p e c t M a r k e r s -q(a), -qat, -qus

(44) /wac-qa/ [wacqa] ' a r e a few (of t h i n g s ) '

/wac-qat/ [wacqat] ' w e r e a few ( y e s t e r d a y ) '
/wac-qus/ [wacqus] ' u s e d to b e a f e w '

A n Analysis

I suggest that a solution to the u n d e r a p p l i c a t i o n of Spirantization in the p r e c e d -

ing m o r p h o l o g i c a l c o n t e x t s is available in a theory w h i c h incorporates interaction-
ism. A n o r d e r i n g solution, similar to t h o s e s u g g e s t e d by M u n r o a n d B e n s o n
( 1 9 7 3 ) a n d b y A n d e r s o n ( 1 9 7 5 ) , is s k e t c h e d in (45).

(45) S T E M - L E V E L : Intensive R e d u p l i c a t i o n c —> sI [ - cont]

W O R D - L E V E L : Plural R e d u p l i c a t i o n c —» s I ]word

Protracted R e d u p l i c a t i o n
a g e n t i v e a n d adjectival -kawut/-ku:t
c a u s a t i v e suffixes -ki, -kixa, -xami, -kixani
t e n s e / a s p e c t suffixes -q(a) -qat, -qus
y

T h e various m o r p h o l o g i c a l p r o c e s s e s d i s c u s s e d a b o v e are assigned to o n e of

t w o d o m a i n s , a n d the t w o subrules of Spirantization apply in different d o m a i n s ,
7

as predicted by general p r i n c i p l e s . T h e crucial aspect of this analysis is that dis-

similatory Spirantization, a stem-level rule, p r e c e d e s the word-level m o r p h o l o g y .

3.3.4. JAVANESE

L i k e L u i s e n o , J a v a n e s e a p p e a r s to distinguish t w o word-internal p h o n o l o g i c a l
d o m a i n s , referred to here as stem- a n d word-level. I p r o p o s e an analysis in w h i c h
Phonology-Morphology Interface 61

a n u m b e r of p h o n o l o g i c a l rules m u s t b e assigned to the stem level since there are

s o m e (word-level) m o r p h e m e s that d o n o t u n d e r g o t h e s e rules.
J a v a n e s e h a s b e e n insightfully d e s c r i b e d and a n a l y z e d by D u d a s ( 1 9 7 4 , 1975),
with later r e a n a l y s e s of s o m e of the d a t a p r o v i d e d b y K e n s t o w i c z ( 1 9 8 6 ) , M e s t e r
( 1 9 8 8 ) , and S c h l i n d w e i n ( 1 9 8 9 ) . U n l e s s o t h e r w i s e noted, all forms are from D u -
d a s . A small n u m b e r of f o r m s are from S u h a r n o ( 1 9 8 2 ) . M y transcriptions differ
8

slightly from t h o s e of b o t h D u d a s a n d S u h a r n o .
S i n c e m o s t of the rules d i s c u s s e d here involve the vocalic p h o n o l o g y of Java-
n e s e , the vowel i n v e n t o r y is given in (46).

(46) i u
e d o
a

T h e rules to b e d i s c u s s e d apply in a variety of m o r p h o l o g i c a l c o n t e x t s , p r i m a r -

ily suffixal.

(47) -[ku] first p e r s o n p o s s e s s i v e

-[mu] second person possessive
-Kn)e] third p e r s o n p o s s e s s i v e or d e m o n s t r a t i v e
-[an] substantive or v e r b - f o r m i n g
-[?ake] causative
-W i m p e r a t i v e , subjunctive
-K?)no] causative imperative
-[(n)ono] locative i m p e r a t i v e

T h e prefixes are generally uninteresting for p r e s e n t p u r p o s e s . T h e only v o w e l s

w h i c h o c c u r in prefixes are [a] a n d [a], a n d [a] d o e s n o t alternate in quality.

3.3.4.1. Mid Vowel Laxing. L a x variants [e 0] of the m i d v o w e l s o c c u r in three

well-defined c o n t e x t s .

1. A m i d v o w e l is lax in a stem-final c l o s e d syllable a n d r e m a i n s lax r e g a r d l e s s

of w h e t h e r later suffixation results in a surface o p e n or closed stem-final
syllable.

T h i s is illustrated in the following m o r p h o l o g i c a l l y related forms [cf. (47)].

(48) [kate?] 'monkey'

[kate?e]
[kate?ku]

(49) [dorjerj] 'story'

[dorjerjku]
[dorjerje]
62 Sharon Hargus

2. Stem-internally, before a h i g h vowel in an o p e n syllable (which will surface

as tense), m i d v o w e l s are lax.

(50) [kleru] [klerune] 'mistaken'

[edum] [edume] 'shady, s h e l t e r e d '

(51) [kopi] [kopine] 'coffee'

[tonus] [tomise] 'rice-accompanying dish'

3. Before a stem-final syllable c o n t a i n i n g [a] ( w h i c h d o e s not o c c u r in o p e n

syllables stem-finally), m i d v o w e l s are lax.

(52) [bosan] 'tired o f

(53) [empar] 'resemblance'

I refer to these three rules collectively as M i d Vowel L a x i n g ( M V L ) .

A n o t h e r source of m i d lax v o w e l s is the p h e n o m e n o n D u d a s calls M i d Vowel
H a r m o n y ( M V H ) . A m i d v o w e l in a final c l o s e d syllable is lax b y M V L , and a
vowel in a p e n u l t i m a t e syllable a g r e e s in laxness with the final v o w e l b y M V L .
However, as s u g g e s t e d b y K e n s t o w i c z ( 1 9 8 6 ) a n d M e s t e r ( 1 9 8 8 ) , this appears to
b e a c a s e of sharing of v o w e l m e l o d i e s b y separate v o w e l s , rather than h a r m o n y .

(54) [bodo] 'stupid'

[bodol] ' c o m e out (hair)'
[dede] 'sun oneself
[dedel] 'rip'

3.3.4.2. High Vowel Laxing. In a p r o c e s s related to M V L , lax variants [iu] of

the high v o w e l s o c c u r in c l o s e d syllables, as d i s c u s s e d b y D u d a s ( 1 9 7 5 : 5 9 ) .

(55) V —> [ — a d v a n c e d t o n g u e root] / C] s y l l

[+high]

T h e high s t e m v o w e l s in (56) alternate b e t w e e n tense and lax, w h e r e a s those in

(57) and (58) invariably fall in o p e n or c l o s e d syllables respectively and are thus
always tense or lax respectively.

(56) [api?] [api?e] 'good, nice'

[j-upu?] [jiupu?o] 'go get'
[kluwurj] [kluwurje] 'rainbow'
[wiwit] [wiwitan] 'beginning'

(57) [ibu] [ibune] 'mother'

[tuku] [nukuo] 'buy'

(58) [jamur] [jamurku] 'mushroom'

[munt] [muntku] 'student'
[tandu?] [tandu?mu] 'actions'
Phonology-Morphology Interface 63

D u d a s notes that e x c e p t i o n s to H i g h Vowel L a x i n g are found in loan w o r d s .

(59) [bensin] 'gas'

[parsis] 'precise'
[ k o r n e t bif] 'corned b e e f

3.3.4.3. a-Raising. B o t h D u d a s a n d S u h a r n o n o t e that there are basically n o

J a v a n e s e m o r p h e m e s e n d i n g in word-final a. D u d a s defends the rule given in (60),
hereafter called ^ - R a i s i n g (cf. also K e n s t o w i c z , 1986; Mester, 1988).

(60) a-^ol #

a - R a i s i n g a c c o u n t s for w i d e s p r e a d alternations, w i t h stem-final / a / a p p e a r i n g

as [a] w h e n n o n - w o r d - f i n a l (in suffixed f o r m s ) .

(61) [jiwo] [jiwaku] ' s o u l , spirit'

[kano] [rpna?ake] 'can, may'
[mej-o] [mejamu] 'table'
[atmo] [atmane] ' s o u l , spirit'
[brasto] [mbrastani] 'wipe out'
[swargo] [swargane] 'heaven'

a - R a i s i n g also applies to suffixal / a / as well as to stem-final /a/.

(62) -[o] i m p e r a t i v e , subjunctive

~[(?)no] causative imperative
-[(n)ono] locative i m p e r a t i v e

T h e i m p e r a t i v e suffixes a p p e a r to b e the only suffixes w h i c h contain suffix-final

/ a / in an o p e n syllable. E x a m p l e s of the first of t h e s e suffixes are given in (63).

(63) IMPERATIVE
[turn] [turuo] 'sleep' (S22)
[ginan^ar] [ginanjaro] 'rewarded' (S22)

D u d a s d o e s n o t p r o v i d e e x a m p l e s of the c a u s a t i v e i m p e r a t i v e a n d locative i m p e r a -
tive, a n d these suffixes are not in S u h a r n o .
D u d a s notes that " t h e I m p e r a t i v e is the only f o r m a t i o n in the l a n g u a g e w h e r e
a suffixed stem-final / a / a p p e a r s on the surface with [o] as its final s e g m e n t "
(p. 110).

(64) IMPERATIVE
[lurjo] [rjlurjoo] 'go away'
[tako] [nakoo] 'come'

In all other f o r m a t i o n s t h r o u g h o u t the l a n g u a g e , stem-final / a / d o e s not a p p e a r as

[o] on the surface unless it is in absolute word-final p o s i t i o n . 9
64 Sharon Hargus

S u h a r n o and D u d a s n o t e various e x c e p t i o n s to ^ - R a i s i n g , m a i n l y in l o a n w o r d s
and place names.

(65) [kolera] ~ [kolerah] 'cholera'

[ora] 'no'
[Jakarta] 'Jakarta' (S6)
[jayapura] 'Jayapura' (S6)
[jakaria] 'Zakaria' (S6)

D u d a s suggests that the p h o n o l o g i c a l rules d i s c u s s e d a b o v e m u s t p r e c e d e a

n u m b e r of m o r p h o l o g i c a l rules.

3.3.4.4. Elative Formation. Elative formation ( E F ) derives intensive forms of

adjectives. For p r e s e n t p u r p o s e s , the m o s t i m p o r t a n t m a r k of E F is an obligatory
s e g m e n t a l c h a n g e , the raising of the r i g h t m o s t v o w e l of the s t e m to [i] o r [u],
d e p e n d i n g o n the b a c k n e s s of the final s t e m v o w e l . 1 0

(66) PRIMARY ELATIVE

[arjel] [anil] 'hard, difficult'
[luwe] [luwi] 'hungry'
[abot] [abut] 'heavy, h e a r d '
[adoh] [aduh] 'far'
[ijo] [iju] 'green'
[rindi?] [rindi?] 'slow'
[wani] [wani] 'bold, daring'
[alus] [alus] 'refined, s m o o t h '
[lugu] [lugu] 'ordinary'

T h e r e is a c o m p l i c a t i o n to E F w h i c h arises w i t h stem-final /a/. T h e elative

forms of adjectives with stem-final / a / vary a c c o r d i n g to w h e t h e r / a / o c c u r s in a
closed syllable or an o p e n syllable.

(67) [lararj] [larirj] ' h i g h in c o s t '

[gamparj] [gampirj] 'easy'
[karas] [karis] 'hard, h a r s h '
/rosa/ [roso] [rosu] 'strong'
/kamba/ [kambo] [kambu] insipid, w i t h o u t spirit'

If / a / occurs in an o p e n syllable, then the elative form contains [u]. O t h e r w i s e , / a /

raises to [i] in the elative.
D u d a s suggests that E F should b e a n a l y z e d as a p p l y i n g to the output of a-
R a i s i n g , w h i c h creates a b a c k vowel from a vowel w h i c h is not clearly m a r k e d for
b a c k n e s s (/a/). H i g h Vowel L a x i n g m u s t also p r e c e d e E F . T h i s counter-feeding
o r d e r is required b e c a u s e the o u t p u t of E F is a tense vowel, even if it occurs in a
closed syllable.
Phonology-Morphology Interface 65

3.3.4.5. Doubling. A s e c o n d m o r p h o l o g i c a l p r o c e s s w h i c h lies outside the d o -

m a i n of the a b o v e p h o n o l o g i c a l rules is a kind of reduplication w h i c h D u d a s calls
D o u b l i n g . T h i s reduplication of entire s t e m s generally indicates plurality of o b -
j e c t s or actions, but as D u d a s n o t e s , D o u b l i n g is also used in conjunction with
affixes w h i c h c h a n g e the m e a n i n g of the b a s e .

(68) /meja/ [mep] [mep-mep] 'table'

/kodo?/ [kocb?] [kDdo?-kodo?] 'frog'
/abur/ [abur] [abur-abur] 'flight'

D u d a s notes (p. 2 1 0 ) that " t h e operation of D o u b l i n g m u s t b e deferred until

after nearly all p h o n o l o g i c a l rules in the g r a m m a r h a v e a p p l i e d . " a - R a i s i n g ap-
parently p r e c e d e s D o u b l i n g , since, as c a n b e seen in ' t a b l e ' in (68), the derived
vowel [o] appears in both p o r t i o n s of the d o u b l e d form. T h e s e f o r m s c o u l d b e
easily h a n d l e d b y a n a l y z i n g c o m p o u n d s as consisting of t w o p h o n o l o g i c a l w o r d s
(cf. N e s p o r a n d Vogel, 1986). H o w e v e r , there are f o r m s w h i c h h a v e u n d e r g o n e
b o t h D o u b l i n g a n d suffixation, w h i c h indicate that a - R a i s i n g is a stem-level p r o -
cess, w h e r e a s D o u b l i n g m u s t b e word-level.

(69) [dorp] [dorp-dorp] [dorja-dorjane] 'prayer'

[dowo] [dowo-dowo] [dawa-dawane] 'long'
[mej-o] [mej-o-mej-o] [meja-mejane] 'table' 1 1

A s n o t e d by S c h l i n d w e i n ( 1 9 8 9 ) , D o u b l i n g m u s t distinguish b e t w e e n stem a n d
suffix, since only the stem r e d u p l i c a t e s in ^ - R a i s i n g . T h i s is easily a c c o m p l i s h e d
via p r o s o d i c c i r c u m s c r i p t i o n of the b a s e ( M c C a r t h y a n d Prince, 1990). However,
the fact that the final vowel of the D o u b l e d form contains [a], rather than [o],
indicates that a - R a i s i n g m u s t counterfeed D o u b l i n g .

(70) /dorja/ /dorja/

STEM-LEVEL: suffixation — dorjane
a-Raising D —
WORD-LEVEL: Doubling dorjo-dorjo dorja-dorjane

If D o u b l i n g p r e c e d e d <3-Raising, the D o u b l e d form * [ d o r p - d o r j a n e ] w o u l d result.

/dorja/
M O R P H O L O G Y : suffixation dorjane
Doubling dorja-dorja dorja-dorj ane
PHONOLOGY: ^-Raising o o o
dorjo-dorjo * dorjo-dorjane

Similarly, H i g h Vowel L a x i n g also p r e c e d e s D o u b l i n g .

(72) [abur] [abur-abur] [abur-abure] 'flight'

[api?] [api?-api?] [api?-api?e] 'good, nice'
66 Sharon Hargus

[dudu?] [dudu?-dudu?] [dudu?-dudu?e] 'place'

[gilik] [gilik-gilik] [gilig-gilige] 'cylindrical'
[munt] [munt-munt] [murid-muride] 'student'

A g a i n , following D u d a s ( 1 9 7 5 ) , I suggest that suffixation p r e c e d e s ^ - R a i s i n g ,

which precedes D o u b l i n g . 1 2

3.3.4.6. An Analysis. A n analysis of the facts p r e s e n t e d is given in (73).

(73) RULES SUFFIXATION

STEM-LEVEL: High Vowel Laxing ~[(n)e] 3rd p . p o s s e s s i v e ,
a-Raising demonstrative
Consonant -[ku] 1st p . p o s s e s s i v e
Neutralization -[mu] 2nd p. possessive
Mid Vowel Laxing -[?ake] simple causative
/j-Deletion -[(n)i] simple locative
-[o] imperative,
subjunctive
-[?no] causative imperative
-[(n)ono] locative i m p e r a t i v e
W O R D - L E V E L : M i d Vowel L a x i n g Elative F o r m a t i o n
Doubling

In s u p p o r t of the early application of a - R a i s i n g , D u d a s notes (p. 106) that " t h e r e

are a n u m b e r of things a b o u t t h e b e h a v i o r of [ a - R a i s i n g ] w h i c h indicate that it
m a y b e a very early rule of the g r a m m a r . " In s u p p o r t of the relatively late location
of E F in the g r a m m a r of J a v a n e s e , I n o t e that it is p r e c e d e d b y T i e r Conflation,
since shared m e l o d i e s are split u p in this p r o c e s s : /lararj/, elative [larirj] ' h i g h
in c o s t ' .

3.4. Summary

A n u m b e r of s e e m i n g l y interactive analyses d o not a p p e a r in Tables 1 - 3 .

A m o n g p r o p o n e n t s of n o n i n t e r a c t i o n i s m , O d d e n (this v o l u m e ) in particular h a s
c h a l l e n g e d the validity of a n u m b e r of the c a s e s w h i c h h a v e b e e n a d d u c e d to
support the interactionist lexical p h o n o l o g y m o d e l on a n u m b e r of g r o u n d s , as
discussed in Section 4 . O n e objection with w h i c h I a m in a g r e e m e n t c o n c e r n s
analyses w h i c h involve u n p r o d u c t i v e m o r p h o l o g y , h e n c e lexically listed or n o n -
rule-governed morphology.
F o r this reason, I h a v e not i n c l u d e d E n g l i s h deverbal -al suffixation in Table 1,
although, as is well k n o w n , -al only a p p e a r s on verbs with final stress on their
b a s e s . T h e a t t a c h m e n t of -al m u s t b e c o n s i d e r e d a lexical property of the b a s e s to
w h i c h -al is attached, as indicated by m i n i m a l pairs like arrival versus *derival
Phonology-Morphology Interface 67

(derivation); refusal versus *confusal (confusion); recital versus *incital (incite-

ment); a n d so on. For this reason, it m a k e s n o sense to a n a l y z e the suffixation of
-al as crucially following the a s s i g n m e n t of stress to v e r b s . 1 3

It should also b e noted that the strength of individual cases w h i c h h a v e b e e n

i n c l u d e d in Tables 1 - 3 , such as D u t c h -eurl-ris allomorphy, relies on information
not p r o v i d e d b y m y sources. It is p o s s i b l e that the s a m e objection raised above for
E n g l i s h -al c o u l d b e raised with respect to the D u t c h -eurlris alternation. Simi-
larly, the validity of the analysis c a n b e q u e s t i o n e d if the p h o n o l o g i c a l p r o c e s s
turns out to b e u n p r e d i c t a b l e . If stress in D u t c h is " n o t transparently p r e d i c t a b l e , "
as s u g g e s t e d b y O d d e n (this v o l u m e ) , then there is n o stress rule a n d h e n c e n o
p h o n o l o g i c a l p r o c e s s w h i c h p r e c e d e s the selection of the -iefox -isch a l l o m o r p h s .
Finally, s o m e a p p a r e n t o r d e r i n g cases can b e r e d u c e d to n o n o r d e r i n g given a p -
p r o p r i a t e representations. T h i s c a n be illustrated w i t h a n o t h e r e x a m p l e from
Javanese.
T h e habitual-repetitive (hereafter H a b - R e p , following D u d a s ) , is a t y p e of re-
duplication that is a c c o m p a n i e d b y obligatory c h a n g e s to the v o c a l i s m of the b a s e .
T h e s e m a n t i c s of the H a b - R e p is ' b e continually d o i n g V or ' b e continually say-
ing N ' (see D u d a s , 1 9 7 5 : 2 2 9 ) .
T h e s h a p e of the H a b - R e p is d e t e r m i n e d b y three t e m p l a t e s , s u m m a r i z e d and
exemplified in ( 7 4 ) . 1 4

(74) a. C V ^ C V ^ C ) , w h e r e V is not [a] - > C V C f l ( C ) - C V C V ( C )

2 1 1 2

[bodo] [boda-bodo] 'stupid'

[dede] [deda-dede] 'sun oneself
[elirj] [elarj-elirj] 'remember'
b. C t f C V ( C ) - » C t f C V ( C ) - C a C V ( C )
2 2 2

[adoh] [odah-adoh] 'far'

[bali] [bola-bali] 'return'
[adus] [odas-adus] 'take a bath'
c C V j C a C -> C V j C a C - C V i C c C
[gombal] [gombal-gombel] 'rag'
[dolan] [dolan-dolcn] ' e n g a g e in recreation'
[edan] [edan-eden] 'crazy'
[rewarj] [rewarj-rewerj] 'servant'

D u d a s argues that ^ - R a i s i n g p r e c e d e s H a b - R e p formation, b e c a u s e f o r m s with

stem-final, o p e n - s y l l a b l e / a / [o] d o not select t e m p l a t e c, as d o forms with stem-
final closed-syllable / a / [a], b u t instead they select t e m p l a t e a.

(75) /j-iwa/ [jiwo] [j-iwa-jiwo] *[j-iwa-^iwe] 'soul, spirit'

/dorja/ [dorp] [dorja-dorp] etc. 'prayer'
/meja/ [mej-o] [meja-mep] 'table'
/sida/ [sido] [sida-sido] ' s u c c e e d in d o i n g '
68 Sharon Hargus

M o r e o v e r , if b o t h s t e m v o w e l s are /a/, a n d the r i g h t m o s t o c c u r s in an o p e n rather

than closed syllable, these f o r m s select t e m p l a t e a, rather than t e m p l a t e b or c, as
w a s the c a s e for / a / in a closed stem-final syllable.

(76) /lara/ [loro] [lora-loro] 'ill, painful'

/dawa/ [dowo] [dowa-dowo] 'long'
/rasa/ [roso] [rosa-roso] '(the) taste ( o f ) '

D u d a s suggests that the distinction in H a b - R e p f o r m s b e t w e e n closed a n d o p e n

syllable stem-final / a / b a s e s is best a c c o u n t e d for b y a s s u m i n g that their b a s e s are
in fact phonetically distinct at the t i m e that H a b - R e p formation h a s applied, as
s h o w n in the d e r i v a t i o n s in (77).

(77) /lawas/ 'old' /dawa/ 'long'

a-Raising o o
Hab-Rep dowa-dowo
a.
b. lowas-lawas
c. lawas-lew es
M i d Vowel L a x i n g dowa-dowo
also, [lowas-lewes]

D u d a s n o t e s a n d a r g u e s against a syllable-based alternative to this analysis. If

t e m p l a t e c w e r e altered so that it referred only to s t e m s with final-syllable / a /
w h i c h O N L Y o c c u r r e d in c l o s e d syllables, then in the c a s e of stems like / d a w a / ,
only t e m p l a t e b w o u l d b e applicable, with H a b - R e p formation resulting in [dowa-
d a w a ] , followed b y the application of a - R a i s i n g , resulting in [ d o w a - d o w o ] .
D u d a s ' s a r g u m e n t against this alternative analysis is that the syllable structure of
the b a s e is not relevant for d e t e r m i n i n g the s h a p e of the output t e m p l a t e for any
of the other H a b - R e p f o r m s , and that this condition duplicates exactly the effect
of a - R a i s i n g . W h i l e the syllable structure condition d o e s e n c o d e the effect of
a - R a i s i n g , w e n o w k n o w that reduplicative or other t e m p l a t e s m a y indeed refer
to the syllable structure of the b a s e ( M c C a r t h y and P r i n c e , 1990), and it s e e m s
best to disregard J a v a n e s e H a b - R e p formation as p r o v i d i n g an a r g u m e n t for
interactionism.

4. T H E O R E T I C A L I M P L I C A T I O N S

A noninteractionist m i g h t raise either of t w o r e m a i n i n g objections to the cases

i n c l u d e d in Tables 1 - 3 .
T h e first issue c o n c e r n s the separation b e t w e e n m o r p h o l o g y a n d p h o n o l o g y ;
that is, w h e t h e r a given p h e n o m e n o n is in fact m o r p h o l o g i c a l or p h o n o l o g i c a l .
Phonology-Morphology Interface 69

T h u s , H a l l e , H a r r i s , a n d V e r g n a u d (1991) a s s u m e that the rule of S p a n i s h w h i c h

a c c o u n t s for alternant f o r m s la a n d el of the f e m i n i n e singular definite article is
p h o n o l o g i c a l ( a l t h o u g h they n o t e that it c o u l d in fact b e a n a l y z e d a s m o r p h o l o g i -
cal). Similarly, O d d e n (this v o l u m e ) p r o p o s e s that t h e subtractive m o r p h o l o g y
found in D a n i s h , Icelandic, and Koasati (from Tables 1 and 3) a n d in K i m a t u u m b i
(Locative T r u n c a t i o n , d i s c u s s e d b y O d d e n ) b e r e a n a l y z e d a s p h o n o l o g y (with
m o r p h o l o g i c a l c o n d i t i o n i n g ) rather than m o r p h o l o g y . F o r the D a n i s h alternation,
O d d e n posits the rule in (78).

(78) 9-^0 / L
[IMPER]

O d d e n h y p o t h e s i z e s that m o r p h o l o g i c a l o p e r a t i o n s are restricted to c o n c a t e n a t i o n

through a PRINCIPLE OF M O R P H O L O G Y - P H O N O L O G Y SEGREGATION ( P M P S ) , and
thus that a rule like (78) m u s t b e c o n s i d e r e d p h o n o l o g i c a l . F o r this reason, O d d e n
a r g u e s that a rule of C h i m w i i n i w h i c h c h a n g e s stem-final dental HI to a n alveolar
in p a s s i v e f o r m s is a p h o n o l o g i c a l rule w h i c h is c o n d i t i o n e d b y the m o r p h o l o g i c a l
c a t e g o r y p a s s i v e . Similarly, a noninteractionist w o u l d n e e d t o a n a l y z e the G e r m a n
-eil-erei a l l o m o r p h y rule a s p h o n o l o g i c a l dissimilation a n d ^ s y l l a b i f i c a t i o n , a s
s k e t c h e d a b o v e in Section 3. S c h l i n d w e i n ( 1 9 8 9 ) a r g u e s that reduplication in Java-
n e s e is p h o n o l o g i c a l , rather than m o r p h o l o g i c a l . H o w e v e r , h e r c o n c l u s i o n is b a s e d
o n t h e controversial a s s u m p t i o n that the b r a c k e t i n g erasure c o n v e n t i o n is cyclic
(see M o h a n a n , 1986, for discussion). T h e fact r e m a i n s that the P M P S is simply a
h y p o t h e s i s , a n d in fact there exists a c o m p e t i n g p r o p o s a l that the p o s s i b l e o p e r a -
tions in m o r p h o l o g y are exactly t h o s e found in p h o n o l o g y (Martin, 1988). I n d e e d ,
if a m o r p h o l o g i c a l c a t e g o r y has a consistent p h o n o l o g i c a l realization, albeit of a
n e g a t i v e (subtractive) nature, it s e e m s p l a u s i b l e t o c o n s i d e r this a m o r p h o l o g i c a l
regularity, a consistent pairing of p h o n o l o g y a n d s e m a n t i c s .
Secondly, a s p o i n t e d out b y Steriade (personal c o m m u n i c a t i o n ) , nearly all t h e
cases p r e s e n t e d in Tables 1 - 3 h a v e certain u n e x p l a i n e d similarities. M o s t of the
Table 1 cases involve p h o n o l o g i c a l rules w h i c h assign s u p r a s e g m e n t a l p r o p e r t i e s
(e.g., stress, syllable structure), w h e r e a s m o s t of the Tables 2 a n d 3 cases involve
m o r p h o l o g i c a l rules w h i c h are p r o s o d i c (e.g., reduplicative, infixational), exten-
sively m a n i p u l a t i n g p h o n o l o g i c a l structure. W e m i g h t e x p e c t a w i d e r r a n g e of
m o r p h o l o g i c a l a n d p h o n o l o g i c a l rules t o o c c u r in all of Tables 1 - 3 if p h o n o l o g y
can g e n u i n e l y p r e c e d e m o r p h o l o g y .
A l t e r n a t i v e e x p l a n a t i o n s for the Tables 2 a n d 3 rules m i g h t i n v o k e o n e of the
ideas in the t h e o r y of reduplication p r o p o s e d b y C l e m e n t s ( 1 9 8 5 ) d i s c u s s e d a b o v e
in Section 3.2. H o w e v e r , not all t h e Table 3 m o r p h o l o g i c a l rules involve p r o s o d i c
m o r p h o l o g y . T h e G e r m a n word-level m o r p h o l o g y that follows g - D e l e t i o n is suf-
fixational. I n a n o n i n t e r a c t i o n i s t analysis of t h e s e data, it w o u l d b e n e c e s s a r y t o
posit either u n d e r l y i n g velar nasals o r a separate rule of g - D e l e t i o n , w h i c h w o u l d
apply t o Igl before v o w e l s in word-level suffixes. M o r e o v e r , even s o m e of t h e
70 Sharon Hargus

m o r p h o l o g y that d o e s a p p e a r to b e p r o s o d i c d o e s not uniformly fail to u n d e r g o a

given p h o n o l o g i c a l rule, as seen in L u i s e n o . In a noninteractionist analysis, the
m o r p h o l o g y w h i c h fails to u n d e r g o Spirantization w o u l d h a v e to b e represented
on a p l a n e separate from the m o r p h o l o g i c a l b a s e , and Spirantization w o u l d h a v e
to apply before p l a n e conflation. H o w e v e r , this analysis w o u l d not explain w h y
s o m e n o n c o n c a t e n a t i v e m o r p h o l o g y , such as Intensive Reduplication, d o e s un-
d e r g o Spirantization. P r e s u m a b l y the output of Intensive R e d u p l i c a t i o n w o u l d b e
r e p r e s e n t e d in the s a m e w a y that (e.g.) Protracted R e d u p l i c a t i o n is. T h e noninter-
actionist m o d e l w o u l d still n e e d to r e c o g n i z e that m o r p h o l o g i c a l p r o c e s s e s m a y
b e l o n g to separate p h o n o l o g i c a l d o m a i n s , a solution w h i c h is similar to that p r e -
dicted b y lexical p h o n o l o g y . P l a n a r segregation of m o r p h e m e s appears to b e ir-
relevant to the analysis of L u i s e n o .
Finally, a l t h o u g h all of the Table 1 cases involve the m a n i p u l a t i o n of supraseg-
m e n t a l information (stress, syllable s t r u c t u r e — H e b r e w M e t a t h e s i s ) , there is n o
o b v i o u s e x p l a n a t i o n for this. It w a s an oft-noted fact in the history of the p h o n o -
logical cycle that m u c h of the best e v i d e n c e for cyclicity involved stress rules, but
this fact has to m y k n o w l e d g e resisted explanation.

ACKNOWLEDGMENTS

I thank the following for comments on earlier versions of this paper: Ewa Czaykowska-
Higgins, Tracy Hall, Ellen Kaisse, Paul Kiparsky, Ove Lorentz, David Odden, Bill Poser,
Keren Rice, Yasushi Sato, Pat Shaw, Donca Steriade, Margaret Strong-Jensen, and an
anonymous reviewer.

NOTES

Aronoff notes that his prediction is empirically supported, at least by the reduplicative
1

cases he discusses in the 1988 paper: "To my knowledge, . . . there are no cases in the
literature where a cyclic phonological rule can be shown to misapply in a demonstrably
stem-level reduplication" (p. 6).
The following forms are cited in German orthography, with stress marked with an acute
2

accent.
1 thank Paul Kiparsky and Ove Lorentz for drawing Lapp to my attention.
3

The following words are cited in Bergsland's transcription system, which uses, among
4

other diacritics, acute accents to mark vowel quality distinctions. Stress is therefore indi-
cated by underlining the stressed nuclei. I have also added [.] to mark syllable division (not
provided in Bergsland's transcriptions).
The /h/ fails to be extracted in the above example because it is a tense affix, one of a
5

set of inflectional affixes which fail to appear in Extracted forms.

Phonology-Morphology Interface 71

Additional primary sources consulted on Luiseno are Bright (1965), Kroeber and Grace
6

(1960), and Davis (1976).

Adjectival reduplication could be assigned to either domain, given the analysis pro-
7

posed by Marantz.
These are marked S with a page number.
8

Other word-internal instances of the output of a-Raising are due to sharing of vowel
9

features by more than one vowel, as seen above with MVL.

(i) [gowo] [ngawani] 'bring'
[loro] [larane] 'ill, painful'
In support of this, Dudas notes that a-Raising results in violations of her rule of MVH.

(ii) [godo] [godane] 'temptation'

[polo] [polane] 'design, pattern'
[roso] [rosane] 'strong'
[sogo] [sogane] 'bark of the indigo tree'
T h e elative is also optionally prosodically marked by lengthening the final vowel and/
1 0

or raising its pitch.

Doubling must retain stem brackets since only the stem reduplicates in affixed forms.
1 1

Word-internal structure persists from an earlier level in these forms.

Dudas also shows that Doubling must follow two additional phonological rules, Con-
1 2

sonant Neutralization and /z-Deletion.

Moreover, given (e.g.) Hayes's (1982a) analysis of English stress, the stress contours
1 3

of -al nominals could be correctly assigned even if stress followed -al suffixation.
There is a complication to template c which will not be discussed here. See Dudas for
1 4

details.

REFERENCES

Anderson, S. R. (1975). On the interaction of phonological rules of various types. Journal

of Linguistics 11, 3 9 - 6 2 .
Aronoff, M. (1976). Word Formation in Generative Grammar. MIT Press, Cambridge,
Mass.
Aronoff, M. (1988). Head operations and strata in reduplication: A linear treatment. Year-
book of Morphology 1, 1-15.
Bat-El, O. (1986). Extraction in Modern Hebrew Morphology. Master's thesis, University
of California, Los Angeles.
Bat-El, O. (1989). Phonology and Word Structure in Modern Hebrew. Doctoral disserta-
tion, University of California, Los Angeles.
Bergsland, K. (1976). Lappische Grammatik mit Lesestucken (Veroffentlichungen der So-
cietas Uralo-Altaica, 11.) Harrasowitz, Wiesbaden.
Bean, L. J., and Shipek, F. C. (1978). Luiseno. In California (R. F. Heizer, ed.) (Hand-
book of North American Indians, 8), pp. 5 5 0 - 5 6 3 . Smithsonian Institution, Washing-
ton, D.C.
72 Sharon Hargus

Booij, G. (1981). Rule ordering, rule application, and the organization of grammars. In
Phonologica 1980 (W. U. Dressier et al., eds.), pp. 4 5 - 5 6 . Innsbrucker Beitrage zur
Sprachwissenschaft.
Bright, W. (1965). Luiseno phonemics. International Journal of American Linguistics 31,
342-345.
Carrier, J. (1979). The Interaction of Morphological and Phonological Rules in Tagalog:
A Study in the Relationship between Rule Components in Grammar. Doctoral disser-
tation, Massachusetts Institute of Technology, Cambridge.
Carrier-Duncan, J. (1984). Some problems with prosodic accounts of reduplication. In Lan-
guage Sound Structure (M. Aronoff and R. Oehrle, eds.), pp. 2 6 0 - 2 8 6 . MIT Press,
Cambridge, Mass.
Chomsky, N., and Halle, M. (1968). The Sound Pattern of English. Harper and Row, New
York.
Clements, G. N. (1985). The problem of transfer in nonlinear morphology. In Cornell
Working Papers in Linguistics 5, 3 8 - 7 3 .
Cohn, A. (1989). Phonetic evidence for configuration constraints. In Proceedings of the
North Eastern Linguistics Society 1 9 , 6 3 - 7 7 .
Davis, J. F. (1976). Some notes on Luiseno phonology. International Journal of American
Linguistics 42, 192-216.
Dudas, K. (1974). A case of functional opacity: Javanese elative formation. Studies in the
Linguistic Sciences 4, 91 - 1 1 1 .
Dudas, K. (1975). The Phonology and Morphology of Modern Javanese. Doctoral disser-
tation, University of Illinois, Champaign-Urbana.
French, K. M. (1988). Insights into Tagalog Reduplication, Infixation and Stress from Non-
linear Phonology. Summer Institute of Linguistics, Arlington, Tex.
Giegerich, H. (1987). Zur Schwaepenthese im Standarddeutschen. Linguistische Berichte
112,449-469.
Hall, T. (1987). Schwa-Zero Alternations in German. Master's thesis, University of Wash-
ington, Seattle.
Hall, T. (1990). Syllable Structure and Syllable-related Processes in German. Doctoral
dissertation, University of Washington, Seattle.
Halle, M. (1973). Prolegomena to a theory of word-formation. Linguistic Inquiry 4, 3 - 1 6 .
Halle, M., Harris, J., and Vergnaud, J.-R. (1991). A reexamination of the stress erasure
convention and Spanish stress. Linguistic Inquiry 22, 141 - 1 5 9 .
Halle, M., and Vergnaud, J.-R. (1987a). An Essay on Stress. MIT Press, Cambridge, Mass.
Halle, M., and Vergnaud, J.-R. (1987b). Stress and the cycle. Linguistic Inquiry 1 8 , 4 5 - 8 4 .
Harris, J. (1989). The stress erasure convention and cliticization in Spanish. Linguistic
Inquiry 20, 3 3 9 - 3 6 4 .
Hayes, B. (1982a). Extrametricality and English stress. Linguistic Inquiry 13, 2 2 7 - 2 7 6 .
Hayes, B. (1982b). Metrical structure as the organizing principle of Yidiny phonology. In
The Structure of Phonological Representations, part 1 (H. v.d. Hulst and N. Smith,
eds.), pp. 97 - 1 1 0 . Dordrecht, Foris.
Hayes, B. (1990). Precompiled phrasal phonology. In The Phonology-Syntax Connection
(S. Inkelas and D. Zee, eds.), pp. 8 5 - 1 0 8 . CSLI Publications and University of Chi-
cago Press, Chicago.
Hayes, B., and Abad, M. (1989). Reduplication and syllabification in Ilokano. Lingua 77,
331-374.
Phonology-Morphology Interface 73

Innes, G. (1971). A Practical Introduction to Mende. School of Oriental and African Stud-
ies, London.
Kanerva, J. (1987). Morphological integrity and syntax: The evidence from Finnish pos-
sessive suffixes. Language 6 3 , 4 9 8 - 5 2 1 .
Kenstowicz, M. (1986). Multiple linking in Javanese. In Proceedings of the North Eastern
Linguistics Society 16, 2 3 0 - 2 4 8 .
Keyser, S. J., and Kiparsky, P. (1984). Syllable structure in Finnish phonology. In Language
Sound Structure (M. Aronoff et al., eds.), pp. 7 - 3 1 . MIT Press, Cambridge, Mass.
Kimball, G. (1982). Verb pluralization in Koasati. In 1982 Mid-America Linguistics Con-
ference Papers (F. Ingemann, ed.), pp. 4 0 1 - 4 1 1 . Department of Linguistics, Univer-
sity of Kansas, Lawrence.
Kiparsky, P. (1982). Lexical morphology and phonology. In Linguistics in the Morning
Calm (I.-S. Yange, ed.), pp. 3 - 9 1 . Hanshin Publishing Co., Seoul.
Kiparsky, P. (1984). On the lexical phonology of Icelandic. In Nordic Prosody III: Papers
from a Symposium (C. C. Elert et al., eds.), pp. 1 3 5 - 1 6 4 . University of Umea, Umea,
Sweden.
Kiparsky, P. (1985). Some consequences of lexical phonology. Phonology Yearbook 2,
85-138.
Kroeber, A. L., and Grace, G. (1960). The Sparkman Grammar of Luiseno (University of
California Publications in Linguistics, 16). University of California Press, Berkeley
and Los Angeles.
Marantz, A. (1982). Re reduplication. Linguistic Inquiry 1 3 , 4 3 5 - 4 8 2 .
Marchand, H. (1969). The Categories and Types of Present-Day English Word-Formation,
2nd ed. Beck, Munich.
Martin, J. (1988). Subtractive morphology as dissociation. Proceedings of West Coast Con-
ference on Formal Linguistics 8, 2 2 9 - 2 4 0 .
McCarthy, J. (1982). Prosodic structure and expletive infixation. Language 58, 5 7 4 - 5 9 0 .
McCarthy, J., and Prince, A. (1990). Foot and word in prosodic morphology: The Arabic
broken plural. Natural Language and Linguistic Theory 8, 2 0 9 - 2 8 4 .
Mester, R.-A. (1988). Studies in Tier Structure. Garland, New York.
Miller, W. (1961). Review of Kroeber and Grace, The Sparkman Grammar of Luiseno.
Language 37, 1 8 6 - 1 8 9 .
Mohanan, K. P. (1986). The Theory of Lexical Phonology. Reidel, Dordrecht.
Munro, P., and Benson, J. (1973). Reduplication and rule ordering in Luiseno. International
Journal of America! Linguistics 39, 1 5 - 2 1 .
Murti, M. S. (1984). An Introduction to Sanskrit Linguistics. D. K. Publications, Delhi.
Nespor, M., and Vogel, N. (1986). Prosodic Phonology. Foris, Dordrecht.
Odden, D., and Odden, M. (1986). Ordered reduplication in Kihehe. Linguistic Inquiry 16,
497-503.
Pesetsky, D. (1979). Russian Morphology and Lexical Theory. Unpublished manuscript,
Massachusetts Institute of Technology, Cambridge.
Pesetsky, D. (1985). Morphology and logical form. Linguistic Inquiry 16, 193-246.
Robins, R. H. (1957). Vowel nasality in Sundanese: A phonological and grammatical study.
In Studies in Linguistic Analysis, pp. 8 7 - 1 0 3 . Blackwell, Oxford.
Schlindwein, D. (1989). Reduplication in lexical phonology: Javanese plural reduplication.
In Arizona Phonology Conference, vol. 2 (S. L. Fulmer et al., eds.) (Coyote Papers,
9), pp. 1 1 6 - 1 2 4 . Department of Linguistics, University of Arizona, Tucson.
74 Sharon Hargus

Sobel, C. P. (1981). A Generative Phonology of Danish. Doctoral dissertation, City Uni-

versity of New York.
Steriade, D. (1988). Reduplication and syllable transfer in Sanskrit and elsewhere. Pho-
nology 5, 7 3 - 1 5 5 .
Strong-Jensen, M. (1987). Lexical overgeneration in Icelandic. Journal of Nordic Linguis-
tics 10, 1 8 1 - 2 0 5 .
Suharno, I. (1982). A Descriptive Study of Javanese. Department of Linguistics, Australian
National University, Canberra.
Szpyra, J. (1987). Inputs to WFRs—Phonological, intermediate or phonetic? The case of
verbs and deverbal nouns in Polish. In Rules and the Lexicon (E. Gussman, ed.),
pp. 1 6 9 - 2 0 3 . Katolickiergo Uniwersytetu Lubelskiego, Lublin, Poland.
Uhrbach, A. (1987). A Formal Analysis of Reduplication and Its Interaction with Phono-
logical and Morphological Processes. Doctoral dissertation, University of Texas,
Austin.
DERIVING CYCLICITY

SHARON INKELAS
Department of Linguistics
University of California
Berkeley, California 94720

1. I N T R O D U C T I O N

T h e a i m of this article is to p r e s e n t a particular m o d e l of the m o r p h o l o g y -

p h o n o l o g y interface from w h o s e p r i n c i p l e s it follows that every p r o c e s s of w o r d
formation will trigger cyclic p h o n o l o g i c a l rules. I call this m o d e l p r o s o d i c lexical
p h o n o l o g y ( P L P ) ; o n e of its properties is that it derives cyclicity as the c o n -
s e q u e n c e of a regulated c o r r e s p o n d e n c e b e t w e e n t w o distinct levels of lexical
representation: m o r p h o l o g i c a l l y b a s e d constituent structure, a n d p h o n o l o g i c a l l y
b a s e d constituent structure.

2. TERMINOLOGY

A n o t o r i o u s l y a m b i g u o u s e l e m e n t in the literature o n p h o n o l o g y a n d m o r -
p h o l o g y is the t e r m PROSODIC, w h i c h c r o p s u p also in the n a m e of the theory
discussed here. O n e application of this t e r m has b e e n to metrical constituents,
such as the m o r a , the syllable, the foot, a n d the w o r d tree. N o t a b l e e x a m p l e s of
this u s a g e are f o u n d in Ito's ( 1 9 8 6 ) p r i n c i p l e of p r o s o d i c licensing, w h i c h requires
every s e g m e n t a l string to be incorporated into units of h i g h e r metrical structure;
in w o r k on p r o s o d i c m o r p h o l o g y by M c C a r t h y and Prince ( 1 9 8 6 , 1990); and in

75
Phonetics and Phonology, Volume 4 Copyright © 1993 by Academic Press, Inc.
Studies in Lexical Phonology All rights of reproduction in any form reserved.
76 Sharon Inkelas

the article b y Booij and L i e b e r (this v o l u m e ) o n m o r p h o l o g i c a l sensitivity t o p r o -

sodic structure.
T h e t e r m " p r o s o d i c " is also c o m m o n l y i n v o k e d t o s u b s u m e stress a n d tone,
and, m o s t generally, all a u t o s e g m e n t a l effects, r a n g i n g over tone, vowel h a r m o n y ,
a n d the representation of length.
A third t y p e of entity t o w h i c h " p r o s o d i c " h a s b e e n e x t e n d e d is the represen-
tation of rule d o m a i n s . T h e t h e o r y of p r o s o d i c p h o n o l o g y , d e v e l o p e d b y Selkirk
( 1 9 7 8 , 1 9 8 0 , 1 9 8 6 ) a n d N e s p o r a n d Vogel ( 1 9 8 2 , 1 9 8 6 ) , posits a p r o s o d i c hierar-
c h y c o n t a i n i n g constituents such a s t h e p h o n o l o g i c a l p h r a s e a n d p h o n o l o g i c a l
w o r d , w h o s e m a i n function is to delimit t h e strings within w h i c h p h o n o l o g i c a l
rules apply. T h e s e relatively abstract constituents differ from the metrical constitu-
ents in lacking a universal p h o n o l o g i c a l o r p h o n e t i c reflex.
In this w o r k , I take t h e position, d e v e l o p e d in m o r e detail in Section 3 , that
there is a crucial distinction b e t w e e n p h o n o l o g i c a l rule d o m a i n s , the m a i n topic of
the paper, a n d the p h o n o l o g i c a l structure created b y p h o n o l o g i c a l rules applying
within t h o s e d o m a i n s . T o refer t o p h o n o l o g i c a l l y derived constituents such a s the
foot, the syllable, a n d the m o r a , I u s e the t e r m METRICAL CONSTITUENT. TO refer
to rule d o m a i n s such a s t h e p h o n o l o g i c a l p h r a s e , I u s e t h e t e r m s P-STRUCTURE
a n d P - C O N S T I T U E N T ( S e l k i r k , 1 9 8 6 ) . M - S T R U C T U R E a n d ITI-CONSTITUENT refer t o
m o r p h o l o g i c a l constituent structure.

3. P R O S O D I C L E X I C A L P H O N O L O G Y

P L P (Inkelas, 1 9 8 9 ) is a m o d e l of the p h o n o l o g y - m o r p h o l o g y interface w h i c h

c a p t u r e s the e m p i r i c a l generalization that word-internal lexical rule d o m a i n s cor-
r e s p o n d (roughly) t o m o r p h o l o g i c a l constituents, w h i l e a d h e r i n g t o the restrictive
position that p h o n o l o g i c a l rules d o n o t access m - s t r u c t u r e o r syntactic p h r a s e
structure directly.

3.1. Prosodic Hierarchy Theory

T h e latter c l a i m u n d e r l i e s a n influential t h e o r y of the interface b e t w e e n syn-

tax a n d t h e postlexical c o m p o n e n t of the p h o n o l o g y , alternately called prosodic
p h o n o l o g y o r t h e p r o s o d i c h i e r a r c h y theory (Selkirk, 1 9 7 8 , 1 9 8 6 ; N e s p o r a n d
Vogel, 1 9 8 2 , 1 9 8 6 ; H a y e s , 1 9 8 9 ) . T h e postlexical units of this hierarchy, that is,
of p-structure, a r e w i d e l y a s s u m e d t o include t h e following, from w h i c h I h a v e
o m i t t e d t h e controversial clitic g r o u p ( H a y e s , 1 9 8 9 ; N e s p o r a n d Vogel, 1 9 8 6 ;
Vogel, 1 9 8 9 ) . 1
Deriving Cyclicity 77

(1) P O S T L E X I C A L PROSODIC H I E R A R C H Y :

Utterance
i

Intonational p h r a s e

Phonological phrase

Phonological word
T h e appeal of this theory c o m e s from its ability to constrain in a natural m a n n e r
the a m o u n t of information m a d e available to e a c h c o m p o n e n t a b o u t the other. A n
i m p o r t a n t step in this direction is the indirect reference h y p o t h e s i s (2), w h i c h p r e -
vents p h o n o l o g i c a l rules from seeing any structure other than the string delimited
by a particular p-constituent.
(2) INDIRECT REFERENCE HYPOTHESIS: p h o n o l o g i c a l rules h a v e access only to
p-structure (i.e., not to m - s t r u c t u r e or c-structure).

A l o n g w i t h s o m e version of the indirect reference h y p o t h e s i s , m o s t i m p l e m e n -

tations of the p r o s o d i c h i e r a r c h y theory h a v e i n c o r p o r a t e d further c o n d i t i o n s on
the k i n d of syntactic information m a d e available to the a l g o r i t h m s w h i c h g e n e r -
ate p-structure (see, e.g., B i c k m o r e , 1990; C h o , 1990; N e s p o r a n d Vogel, 1986;
Selkirk, 1986; Z e e a n d I n k e l a s , 1990). T h e m o s t restrictive p r o p o s a l s limit a l g o -
r i t h m s to information a b o u t the configurationality of syntactic structure (with the
p o s s i b l e inclusion of X ' level), specifically i n c l u d i n g information such as sister-
h o o d and syntactic b r a n c h i n g n e s s , w h i l e e x c l u d i n g reference to syntactic c a t e g o r y
or other m o r p h o s y n t a c t i c features.
D e s p i t e the p r e v a l e n c e of this general p e r s p e c t i v e o n the p h o n o l o g y - s y n t a x
interface, the o p p o s i t e view is a s s u m e d — a n d h a s generally p r o v e d u n p r o b l e m -
a t i c — i n m o s t theories of the interaction b e t w e e n m o r p h o l o g y a n d p h o n o l o g y ,
a b o u t w h i c h the p r o s o d i c h i e r a r c h y t h e o r y says relatively little. L e x i c a l p h o n o l o g y
(Kiparsky, 1982; M o h a n a n , 1986; Pesetsky, 1979), for e x a m p l e , m a k e s a certain
a m o u n t of internal m o r p h o l o g i c a l b r a c k e t i n g accessible to p h o n o l o g i c a l rules, but
the d o m a i n s of rule application t h e m s e l v e s are p r o v i d e d b y the interleaving of
m o r p h o l o g i c a l o p e r a t i o n s a n d p h o n o l o g i c a l rule application. A n d e r s o n ( 1 9 8 6 ,
1988) also a s s u m e s a m o d e l in w h i c h the o u t p u t of e a c h m o r p h o l o g i c a l rule is fed
directly to p h o n o l o g i c a l rules.
A few significant e x c e p t i o n s to this g e n e r a l o u t l o o k o c c u r in the w o r k of Booij
and R u b a c h ( 1 9 8 4 ) ; Booij ( 1 9 8 5 ) ; S p r o a t ( 1 9 8 5 , 1986); Booij a n d R u b a c h ( 1 9 8 7 ) ;
and C o h n ( 1 9 8 9 ) . T h e s e authors p r o p o s e that W I T H I N T H E LEXICON there exist
t w o distinct, c o p r e s e n t structures, o n e m o t i v a t e d b y the m o r p h o l o g y a n d the other
b y the p h o n o l o g y . T h e latter (crucially distinct from metrical structure) o r g a n i z e s
78 Sharon Inkelas

m o r p h e m e s into the constituents utilized as the d o m a i n s of p h o n o l o g i c a l rule

application.
W h a t is the nature of these lexical p - c o n s t i t u e n t s ? It w a s initially a s s u m e d by
Selkirk ( 1 9 7 8 ) , and, following her, b y N e s p o r and Vogel ( 1 9 8 2 , 1986), that the
p h o n o l o g i c a l p h r a s e , the p h o n o l o g i c a l word, the foot, and the syllable w e r e di-
rectly related. S i n c e the latter t w o constituents are smaller than m o s t p h o n o l o g i c a l
w o r d s , a natural c o n c l u s i o n w a s that t h e s e m e t r i c a l units f o r m e d the lexical e n d
of the p r o s o d i c hierarchy. However, as foreseen b y N e s p o r a n d Vogel (p. 18) and
Selkirk (p. 385) in their respective 1986 expositions of the theory, this position
h a s faltered u n d e r the m o r e recent focus o n lexical rules. M e t r i c a l constituents are
sufficient in neither n u m b e r n o r size to d e s c r i b e a d e q u a t e l y all the d o m a i n s of
lexical rules. W o r s e , they d o not even form a consistent h i e r a r c h y with the lexical
rule d o m a i n s that are attested (Inkelas, 1989; Selkirk, 1986; Z e e , 1988).
Significantly, in e a c h of t h e w o r k s cited a b o v e in w h i c h p-structure is i n v o k e d
to d e s c r i b e lexical p h o n o l o g i c a l rule d o m a i n s that d o not c o i n c i d e exactly with
the c o r r e s p o n d i n g m o r p h o l o g i c a l constituent structure, it is the p-units at the w o r d
level a n d h i g h e r — n o t t h e s u p p o s e d l y lexical m e t r i c a l u n i t s — w h i c h are utilized.
In a p a p e r on D u t c h and G e r m a n c o m p o u n d s , Booij (1985) s h o w s that the ellip-
sis u n d e r c o o r d i n a t i o n of certain e l e m e n t s of these c o m p o u n d s can b e explained
only u n d e r the a s s u m p t i o n that their m - s t r u c t u r e a n d p-structure are distinct. T h e
o m i s s i b l e e l e m e n t s of c o m p o u n d s c o r r e s p o n d not to m o r p h o l o g i c a l constituents,
but rather to p h o n o l o g i c a l w o r d s (p. 151).
S p r o a t ( 1 9 8 6 ) a r g u e s a l o n g similar lines that the distinct t r e a t m e n t b y p h o n o -
logical rules of s u b c o m p o u n d s and c o c o m p o u n d s in M a l a y a l a m is best h a n d l e d
b y treating the f o r m e r as a single p h o n o l o g i c a l w o r d and the latter as t w o p h o n o -
logical w o r d s . A g a i n , the a p p e a l is to this relatively abstract m e m b e r of the
p-hierarchy, rather than to o n e of its h y p o t h e s i z e d lower-end constituents.
Finally, C o h n ( 1 9 8 9 ) a r g u e s persuasively that the cyclic application of stress
rules in I n d o n e s i a n yields t h e correct results only w h e n p-structure is p e r m i t t e d to
depart from m - s t r u c t u r e . T h e c o m p l e x i t y of C o h n ' s data requires reference to t w o
different types of lexical p-constituents. E v e n so, h e r analysis c o n f o r m s to those
of Booij and S p r o a t in i n v o k i n g only e l e m e n t s of the p-hierarchy at and above the
level of the p h o n o l o g i c a l w o r d .
T h e generalization e m e r g i n g from the w o r k of B o o i j , R u b a c h , Sproat, and C o h n
is that as it b e c o m e s clearer that units of the p-hierarchy m u s t b e i n v o k e d to h a n d l e
lexical m i s m a t c h e s b e t w e e n m - s t r u c t u r e and p h o n o l o g i c a l rule d o m a i n s , it b e -
c o m e s c o r r e s p o n d i n g l y o b v i o u s that the lexical units offered b y standard p r o s o d i c
hierarchy theory are not equal to the task. T h e only units that prove relevant in the
lexicon are the NONMETRICAL units of that hierarchy.
If the only n e e d e d lexical p-constituent is the p h o n o l o g i c a l word, then the p r o b -
l e m is not severe: Booij and R u b a c h ( 1 9 8 7 ) h a v e argued that b o t h the lexicon and
the postlexical c o m p o n e n t p o s s e s s word-level rules. Z e e ( 1 9 8 8 , this v o l u m e ) ar-
Deriving Cyclicity 79

g u e s that the p h o n o l o g i c a l w o r d straddles the b o u n d a r y b e t w e e n the lexical and

the postlexical c o m p o n e n t s a n d is thus accessible to both.
H o w e v e r , e v i d e n c e such as C o n n ' s that m o r e than o n e p - c o n s t i t u e n t is n e e d e d
lexically p o s e s a m o r e serious p r o b l e m . O n e p o s s i b l e solution, a d o p t e d in C o h n ' s
article, is to d r a w the postlexical e l e m e n t s of the hierarchy (in h e r case, the clitic
g r o u p ) into the lexicon. H o w e v e r , d o i n g so m a k e s t h e s e p-constituents unavailable
for postlexical u s e and also w e a k e n s the cross-linguistic c l a i m s that h a v e b e e n
m a d e as to the syntactic correlates of the v a r i o u s levels in the p-hierarchy.
T h e other t y p e of solution is simply to j e t t i s o n the useless (in this context)
metrical units from the p - h i e r a r c h y a n d r e p l a c e t h e m with new, g e n u i n e l y lexical
p-constituents w h i c h will function as the lexical p - d o m a i n s called for in the w o r k
of B o o i j , Sproat, C o h n , a n d others. T a k i n g this idea to its logical e x t r e m e , I
p r o p o s e d (Inkelas, 1989) to s u p p l e m e n t the lexical e n d of the p - h i e r a r c h y suffi-
ciently that it can e x t e n d to all lexical p h o n o l o g i c a l rules. W i t h i n the lexicon, j u s t
as outside it, p h o n o l o g i c a l rules never apply to m o r p h o l o g i c a l strings directly.
Rather, they a l w a y s apply within d o m a i n s d e s c r i b e d b y p-constituents. M i s m a t c h
p h e n o m e n a like those a n a l y z e d b y B o o i j , Sproat, and C o h n are special not b e -
c a u s e the relevant rules refer to p - c o n s t i t u e n t s , b u t rather b e c a u s e the relevant
p-constituents d o not m a t c h t h e c o r r e s p o n d i n g m o r p h o l o g i c a l structure e x a c t l y . 2

T h i s a p p r o a c h t h u s unifies the theoretical t r e a t m e n t of lexical and postlexical

rule application, w h i l e o b v i a t i n g the theoretically a n d empirically a w k w a r d posi-
tion of forcing lexical rules to apply w i t h i n m e t r i c a l constituents or within post-
lexical p - d o m a i n s .
In the two-structure m o d e l , the u n m a r k e d representation of a m o r p h o l o g i c a l l y
c o m p l e x w o r d will thus look s o m e t h i n g like that in (3).

(3) m-structure p-structure

icon ic ity icon ic ity

D e s p i t e its c o m p l e x i t y , the m o d e l as illustrated in (3) still d o e s not address

C o h n ' s a r g u m e n t that m o r e t h a n o n e t y p e of lexical p - c o n s t i t u e n t is n e e d e d . T h i s
c o n c l u s i o n , f o r m e d on the basis of I n d o n e s i a n , patterns with a larger b o d y of
e v i d e n c e a c c u m u l a t e d from other l a n g u a g e s that the lexicon m a y be stratified into
m o r e than o n e level, with c o r r e s p o n d i n g differences in p h o n o l o g i c a l rule appli-
cation (see, e.g., B o r o w s k y , 1986; H a r g u s , 1988; Kiparsky, 1982; M o h a n a n , 1982;
Pesetsky, 1979; Z e e , 1988).
To a c c o m m o d a t e this c o m p l e x i t y , I p r o p o s e d (Inkelas, 1989) that lexical
80 Sharon Inkelas

p-constituents m a y c o m e in m o r e than o n e type. Lexical stratification of p h o n o -

logical rules is thus c o m p a r a b l e theoretically to postlexical stratification—the lat-
ter b e i n g m o r e c o m m o n l y r e c o g n i z e d as the various layers in the (postlexical)
p r o s o d i c hierarchy. Sublexical strata m a y simply b e interpreted as different lexical
p - c a t e g o r i e s , a r r a n g e d in a fixed hierarchy d o m i n a t e d b y the p h o n o l o g i c a l word.
A s s u m i n g , for e x a m p l e , the stratification of the E n g l i s h lexicon into t w o lev-
els (Kiparsky, 1985), w e m u s t posit t w o c o r r e s p o n d i n g sublexical layers in
p-structure. In the figure in (4), the c a t e g o r y a c o r r e s p o n d s to stratum 1 a n d the
category (3, to stratum 2.

(4) P-HIERARCHY (English):

Utterance

I
Intonational p h r a s e Post-lexical

Phonological phrase

I
— Phonological word
I
P Lexical

a
T h e parallel b e t w e e n these sublexical constituent types and traditional levels is
close. Just as affixes select for a t t a c h m e n t at a given level in standard lexical p h o -
nology, affixes s u b c a t e g o r i z e for a t t a c h m e n t to a given constituent type in the
m o d e l p r o p o s e d h e r e . (See Section 3.2.3 and Inkelas, 1989, for m o r e details.)
T o s u m u p , the p r o p o s a l is that lexical p h o n o l o g i c a l rule d o m a i n s are to b e
represented neither as metrical constituents n o r as m o r p h o l o g i c a l o n e s , b u t rather
as e l e m e n t s of p-structure, a u n i q u e h i e r a r c h y w h o s e c o v e r a g e generally c o r r e -
s p o n d s to that of m o r p h o l o g i c a l constituents. M e t r i c a l structure exists in a hier-
archy distinct from p-structure. It is a different level of representation, and its units
obey different constraints from those that g o v e r n p-structure. Just as phrasal stress
is assigned b y rules a p p l y i n g within the p h o n o l o g i c a l p h r a s e , so metrical constitu-
ents are built b y p h o n o l o g i c a l rules applying within p-constituents. M e t r i c a l c o n -
stituents are built b y p h o n o l o g i c a l rules, w h e r e a s p-constituents are built b y m o r -
phologically sensitive a l g o r i t h m s — a p r o c e s s to w h i c h I n o w turn.
3.2. L e x i c a l P - C o n s t i t u e n t s

T h e c o r r e s p o n d e n c e b e t w e e n m o r p h o l o g i c a l structure and p h o n o l o g i c a l rule

d o m a i n s that lexical p h o n o l o g y takes to b e obligatory is treated in P L P as
Deriving Cyclicity 81

the u n m a r k e d case. T h e simple a l g o r i t h m in (5) m a p s from m - s t r u c t u r e into

p - s t r u c t u r e , p r o v i d i n g for a perfect m a t c h b e t w e e n the t w o . 3

(5) P - C O N S T I T U E N T F O R M A T I O N ALGORITHM ( P C F ) :

Wm -> <-*>m Wp
T h e application of P C F , parallel to an u n m a r k e d p h o n o l o g i c a l p h r a s i n g a l g o -
r i t h m o p e r a t i n g o n syntactic constituents, p r e s u p p o s e s the g e n e r a t i o n of the input
m - s t r u c t u r e . T h o u g h I c a n n o t p r e s e n t t h e full details of a m o r p h o l o g i c a l theory
h e r e , I a s s u m e the e x i s t e n c e of a parallel m - c o n s t i t u e n t f o r m a t i o n w h i c h serves a
function c o m p a r a b l e to Selkirk's ( 1 9 8 2 ) rewrite rules for m o r p h o l o g y .

(6) M - C O N S T I T U E N T FORMATION ALGORITHM ( M C F ) :

x -> (x) m

In (6), the i n p u t x is a stem m o r p h e m e w h o s e insertion is g o v e r n e d b y the

m o r p h o l o g i c a l c o m p o n e n t of the g r a m m a r . A derivation illustrating the s i m p l e
interaction b e t w e e n m - s t r u c t u r e a n d p-structure f o r m a t i o n is given in (7).

(7) INPUT: stem

MCF (stem),
m
PCF [stem]p
PHONOLOGICAL RULES: [stem] p

L i k e theories in w h i c h p h o n o l o g i c a l rules refer directly to syntactic structure

(e.g., K a i s s e , 1985), past t h e o r i e s in w h i c h p h o n o l o g i c a l rules apply to m o r p h o -
logical d o m a i n s directly h a v e restricted the a c c e s s of those rules to the internal
structure of the d o m a i n s that they apply within. In lexical p h o n o l o g y , such restric-
tions h a v e taken t h e f o r m of b r a c k e t e r a s u r e c o n v e n t i o n s a p p l y i n g to m o r p h o l o g i -
cal structure (Kiparsky, 1982; M o h a n a n , 1982; Pesetsky, 1979). A l t h o u g h less
structure is available to p h o n o l o g i c a l rules in PLP, it is still desirable to prevent
p h o n o l o g i c a l rules from a c c e s s i n g all of the internal p-structure g e n e r a t e d in an
e x t e n d e d derivation b y i n v o k i n g s o m e notion of b r a c k e t e r a s u r e . S p r o a t (this vol-
u m e ) and Inkelas (1989) h a v e p r o p o s e d to view b r a c k e t e r a s u r e not as a transfor-
m a t i o n a l p r o c e s s d e s t r o y i n g structure, b u t rather as a locality c o n s t r a i n t o n w h a t
p h o n o l o g i c a l rules m a y access. T h e c l a i m w o u l d thus b e that p h o n o l o g i c a l rules
m a y only refer to the highest n o d e in p-structure, thus limiting their access to
internal c o m p l e x i t y . 4

G i v e n this or any version of b r a c k e t erasure, it will b e n e c e s s a r y for p h o n o -

logical rules to apply i m m e d i a t e l y u p o n the creation of e a c h n e w p-constituent.
O t h e r w i s e internal p-structure w o u l d b e lost, a n d n o word-internal rule applica-
t i o n — o r , e x t e n d i n g t h e c o n v e n t i o n to the postlexical p h o n o l o g y , n o rule applica-
tion b e l o w the level of the u t t e r a n c e — c o u l d occur.

(8) P h o n o l o g i c a l rules apply automatically u p o n the c o n s t r u c t i o n of a n e w

p-constituent.
82 Sharon Inkelas

S o far I have d i s c u s s e d the g e n e r a t i o n of p-structure only in abstract t e r m s . B u t

insofar as w e find it n e c e s s a r y to p r o v i d e distinct t y p e s of lexical p-constituents in
a level-ordered lexicon, the th eo r y will n e e d a m e c h a n i s m for g e n e r a t i n g t h e m in
the a p p r o p r i a t e order. T h e m o s t straightforward m e t h o d is to adjust the existing
P C F a l g o r i t h m such that instead of simply building g e n e r i c p-structure, it creates
p-constituents of the i m m e d i a t e l y h i g h e r category. T h e index / in the revised al-
g o r i t h m (9) r a n g e s over the o r d e r e d set of c a t e g o r y labels, with a b e i n g the lowest.

(9) P-CONSTITUENT FORMATION ALGORITHM (revised)I

W m / Wp,-, -> <*>m, Mp,

C o r r e s p o n d i n g revisions are m a d e to the M - C F A .

(10) M-CONSTITUENT FORMATION ALGORITHM (revised):

<*> . m -» (x) mi+l

A s a m p l e derivation s h o w i n g the a s s i g n m e n t of level 1 (a), level 2 (P), and w o r d

level ((o) structure to a s i m p l e stem follows.

(11) U n d e r l y i n g representation: s t e m

Level 1 MCF <stem) m a

PCF t s t e m
]p«
p h o n o l o g i c a l rules [stem] p a

Level 2 MCF <stem) m p

[ s t e m
]p a

PCF [stem] p p

p h o n o l o g i c a l rules [stem] p p

W o r d level MCF (stem) m w

[stem] p p

PCF [stem] p a )

p h o n o l o g i c a l rules

I n s t a n c e s in w h i c h P C F supplies p-structure to a representation that formerly

lacked it, as seen in the s a m p l e derivation in (7) and in the first step of the deri-
vation in (11), are simply the special c a s e in w h i c h the index in the a l g o r i t h m is
instantiated as a .

3.3. M i s m a t c h e s — o r , P C F as t h e " E l s e w h e r e C a s e "

E l s e w h e r e (Inkelas, 1989) I h a v e a r g u e d in favor of P L P on the g r o u n d s that it

is e m i n e n t l y suited to d e s c r i b e in an insightful w a y the attested set of m i s m a t c h e s
that o c c u r b e t w e e n m o r p h o l o g i c a l c o n s t i t u e n c y and p h o n o l o g i c a l rule d o m a i n s .
5

T h o u g h they m a y c o r r e s p o n d exactly, as the p r e c e d i n g a l g o r i t h m w o u l d suggest,

Deriving Cyclicity 83

the very a s s u m p t i o n that p-structure a n d m - s t r u c t u r e are represented separately

predicts that m i s m a t c h e s should b e possible. M i s m a t c h e s b e t w e e n p-structure
and syntactic structure are, of c o u r s e , celebrated, f o r m i n g o n e of the fundamental
a r g u m e n t s in favor of the p r o s o d i c hierarchy theory ( N e s p o r and Vogel, 1986;
Selkirk, 1978, 1980). A basic tenet of PLP, in w h i c h sublexical p-structure n o w
h a s m a n y of the s a m e properties as postlexical p-structure, is that m i s m a t c h e s
should o c c u r in the lexicon as well.

3.3.1. COMPOUNDS

O n e kind of lexical m i s m a t c h is already quite well k n o w n in the literature on

the p r o s o d i c hierarchy, n a m e l y that o c c u r r i n g in a certain t y p e of c o m p o u n d . In
s o m e l a n g u a g e s , the t w o m e m b e r s of a c o m p o u n d w o r d act as separate p h o n o -
logical rule d o m a i n s . T h i s is the c a s e , for e x a m p l e , in Sanskrit (Selkirk, 1980); in
D u t c h and G e r m a n (Booij, 1985); in M a l a y a l a m (Sproat, 1986); in I n d o n e s i a n
( C o h n , 1989); and in Italian, as d e s c r i b e d by N e s p o r and Vogel ( 1 9 8 6 ; see also
Nespor, 1984). F o r e x a m p l e , a l t h o u g h all other w o r d s in Italian m a y contain at
m o s t o n e p r i m a r y stress, c o m p o u n d s c o n t a i n t w o . In addition, a w o r d - b o u n d e d
rule of nasal assimilation applies within each m e m b e r of a c o m p o u n d b u t is
b l o c k e d across the c o m p o u n d - i n t e r n a l b o u n d a r y . A third rule, intervocalic voic-
ing, also voices s w o r d - m e d i a l l y — b u t not if its e n v i r o n m e n t contains a c o m -
p o u n d b o u n d a r y ( N e s p o r and Vogel, 1 9 8 6 ) . 6

S o m e c o m p o u n d s d o b e h a v e as a unit for p h o n o l o g i c a l p u r p o s e s . F o r e x a m p l e ,
c o m p o u n d s in G r e e k ( N e s p o r and Vogel, 1986) form a unitary d o m a i n for stress
rules such that the surface stress of a c o m p o u n d n e e d not reflect the surface stress
of either c o m p o n e n t w o r d . T h e internal m o r p h o l o g i c a l b o u n d a r y of G r e e k c o m -
p o u n d s d o e s not interfere with the d o m a i n of the stress rule.
T h i s split in the b e h a v i o r of c o m p o u n d s for p h o n o l o g i c a l p u r p o s e s has led to
the h y p o t h e s i s that w h i l e c o m p o u n d s of the type found in G r e e k c o r r e s p o n d to a
single p-constituent (12a), c o m p o u n d s like those in Italian c o r r e s p o n d to t w o
constituents (12b).

(12) a. <tosta) m [t6sta] p 'toaster'

(pane) m [pane] p 'bread'
(tDstapane)^ [ t 6 s t a ] [pane]p
p ' b r e a d toaster'
b. (kukla> m [kukla] p 'doll'
<spiti) m [spiti] p 'house'
(kuklaspiti) r [kuklaspiti] p 'doll's h o u s e '

T h u s , the array of p h o n o l o g i c a l c o m p o u n d types p r o v i d e s e v i d e n c e in favor of

allowing m i s m a t c h e d m o r p h o l o g i c a l and p-structure.
84 Sharon Inkelas

3 . 3 . 2 . INVISIBILITY

A different type of m i s m a t c h involves t h e p h e n o m e n o n characterized by Poser

( 1 9 8 4 ) as INVISIBILITY. Often referred to in individual cases as extrametricality
(Harris, 1983; H a y e s , 1981), extratonality (Pulleyblank, 1986), or e x t r a p r o s o d i c -
ity (Kiparsky, 1985), invisibility involves t h e e x c l u s i o n of s o m e part of the p h o -
nological string from t h e d o m a i n of p h o n o l o g i c a l rules.
In A m e l e , a l a n g u a g e of P a p u a N e w G u i n e a (Roberts, 1987), there is an allo-
p h o n i c voicing alternation a m o n g labial a n d velar stops. T h e s e stops are voiced
word-initially a n d intervocalically (13a), b u t voiceless word-finally ( 1 3 b ) ( R o b -
erts, 1 9 8 7 : 3 3 3 ) .

(13) a. bee 'today' b . gcelcep 'body ornament'

cebd 'brother' bolop 'trap'
gce:d 'crazy' bcemik 'his scrotum'
ogol 'tree s p e c i e s ' celok 'raven'

However, j u s t in c a s e t h e w o r d is m o n o s y l l a b i c , a final labial or velar stop will

instead b e voiced.

(14) a. sib 'rubbish' b . ?og 'frog'

nce:b 'termite' gug 'basis'
sub 'comb' lig 'shrub species'

If w e a s s u m e that these stops, for w h i c h voicing is n o t p h o n e m i c , u n d e r g o a w o r d -

level context-free rule inserting [ + v o i c e ] , then w e m a y a c c o u n t for the voiceless
status of word-final labial a n d velar stops with a rule of final c o n s o n a n t invisibility.
Invisible c o n s o n a n t s are e n c l o s e d in a n g l e d brackets.

(15) INPUT: polop kuk

INVISIBILITY: polo(p) —
VOICING: bolo(p) gug
OUTPUT: [bolop] [gug]
'trap' 'basis'

Invisibility h a s often b e e n treated diacritically in the past (though see Poser,

1984), a theoretical b l e m i s h h a n d l e d with exception features (e.g., [ + e x t r a m e t r i -
cal]) w h o s e restricted distribution is largely unexplained. In Inkelas (1989) I
argued, to the contrary, that invisibility is an integrated facet of p-constituent for-
mation: in particular, 'invisibility effects' result w h e n certain e l e m e n t s of a m o r -
phological constituent are e x c l u d e d from the c o r r e s p o n d i n g p-constituent.

(16) <polop> m [pol3] P p

T h e rule responsible for generating a p-representation like that in (16) is formu-

lated in (17) a n d s h o w n applying in ( 1 8 ) . 7
Deriving Cyclicity 85

(17) F I N A L C O N S O N A N T INVISIBILITY: [.... C] p -> [...] p C

(18) [pobp] -» p [poio] p -> [bololpp ->

p [b5bp] p

invisibility voicing other rules

T h e first a d v a n t a g e to treating invisibility as the adjustment of the p-constituent

e d g e is that w e can explain the b l o c k a g e of invisibility in m o n o s y l l a b i c forms: re-
m o v i n g the final c o n s o n a n t from the p-constituent w o u l d r e n d e r these f o r m s m o n -
o m o r a i c , p r e s u m a b l y in violation of a m i n i m a l size constraint in the l a n g u a g e . 8

A similar a c c o u n t can b e given for any of the w e l l - k n o w n e x a m p l e s of invisi-

bility in the literature; final syllable extrametricality in E n g l i s h n o u n s ( H a y e s ,
1981), for e x a m p l e , will d e r i v e f o r m s of the sort depicted in (19).

(19) <Pamela) m [Pame] lap

A s in A m e l e , E n g l i s h invisibility a s s i g n m e n t is b l o c k e d w h e n the resulting

form w o u l d dip b e l o w m i n i m a l size constraints: it is well k n o w n that m o n o s y l -
labic n o u n s d o not u n d e r g o the extrametricality rule, the best e v i d e n c e b e i n g that
they a l w a y s surface with stress (Hayes, 1981).

(20) (dog) m [dog] p (*[] dog)

p

A c c o u n t s in w h i c h extrametricality is a diacritic feature h a v e h a d to stipulate

its failure to b e a s s i g n e d to f o r m s w h i c h it w o u l d e x h a u s t (Hayes, 1981). But this
prohibition follows naturally from the p - c o n s t i t u e n t a p p r o a c h , in w h i c h invisibil-
ity a s s i g n m e n t directly affects the size of the p - d o m a i n .
In addition to o b v i a t i n g the n e e d for an invisibility diacritic a n d e x p l a i n i n g
m i n i m a l i t y c o n d i t i o n s , the p-constituent t r e a t m e n t of invisibility also explains the
best k n o w n condition on invisibility, n a m e l y its distribution: invisibility is found
only at the e d g e (Buckley, 1992; H a r r i s , 1 9 8 3 ; H a y e s , 1 9 8 1 ; Poser, 1984; Pulley-
blank, 1986).
T h i s well k n o w n PERIPHERALITY CONDITION follows directly from the repre-
sentation of invisible material as external to the p - c o n s t i t u e n t — t h u s necessarily
in the e n v i r o n m e n t of a p - c o n s t i t u e n t e d g e . It is n o m o r e p o s s i b l e on this a p p r o a c h
to represent m e d i a l invisibility than it is to represent a d i s c o n t i n u o u s p-constituent.
T h e r e is thus n o s o u r c e for m e d i a l invisibility in the theory.
In s u m , the h y p o t h e s i z e d e x i s t e n c e of i n d e p e n d e n t m- a n d p-structure yields a
m o r e restrictive t h e o r y of w h a t information p h o n o l o g i c a l rules can see directly.
B u t p o s s i b l y m o r e importantly, it also p r o v i d e s a uniform t r e a t m e n t of c o m p o u n d -
ing and invisibility m i s m a t c h e s , treating these o t h e r w i s e t r o u b l e s o m e p h e n o m e n a
as c o m p l e m e n t a r y facets of the r e g u l a r p r o c e s s of p-constituent formation.
R e p r e s e n t i n g invisibility and p - c o m p o u n d s as instances of p-structure m i s -
m a t c h e s has an i m p o r t a n t c o n s e q u e n c e for o u r interpretation of the P C F a l g o -
r i t h m p r e s e n t e d earlier. I n t e n d e d to c a p t u r e the generalization that m - a n d p -
constituency c o r r e s p o n d , this a l g o r i t h m w o u l d clearly o v e r g e n e r a t e if it applied in
86 Sharon Inkelas

all the cases w e h a v e seen: it w o u l d incorrectly neutralize the distinction b e t w e e n

c o m p o u n d types, a n d it w o u l d o b s c u r e representations of invisibility. W e m u s t
therefore e n s u r e that P C F applies only in the regular c a s e and not in the e x c e p -
tional c a s e . T h i s can b e a c c o m p l i s h e d by c o n s t r u i n g the P C F as an e l s e w h e r e case,
treating it as a p u r e l y structure-filling a l g o r i t h m w h i c h applies only in the a b s e n c e
of the structure it w o u l d i n s e r t . 9

3.3.3. P-SUBCATEGORIZATION

Positing separate structures in the lexicon m a k e s p o s s i b l e the description of

m i s m a t c h e s . It also e n h a n c e s t h e representation of d e p e n d e n t m o r p h e m e s , result-
ing in a principled classification of that c o m p l e x set. Presented in ( 2 1 ) , d e p e n d e n t
( b o u n d ) m o r p h e m e s are defined as those w h i c h are u n a b l e to stand alone. In this
they contrast with underlyingly free s t e m s .

(21) FREE BOUND

Stem Affix
B o u n d root
Clitic

G r e a t differences separate b o u n d roots, affixes, and clitics. M o s t salient is that

clitics form their o w n syntactic t e r m i n a l s , their d e p e n d e n c e s e e m i n g l y p h o n o -
logical in nature. B y contrast, affixes a n d b o u n d roots display d e p e n d e n c e of a
m o r e m o r p h o l o g i c a l ilk, r e q u i r i n g a m o r p h o l o g i c a l sister in o r d e r to constitute
well-formed w o r d s c a p a b l e of entering the syntax. Yet affixes a n d roots differ
p h o n o l o g i c a l l y in that affixes typically trigger a c y c l e of p h o n o l o g i c a l rules upon
insertion a n d can b e sensitive to p h o n o l o g i c a l p r o p e r t i e s of sister m o r p h e m e s —
properties lacking in b o u n d r o o t s . 1 0

T h e s e differences can b e s c h e m a t i z e d as in ( 2 2 ) , w h e r e the starting a s s u m p t i o n

is the distinctness of m - a n d p - s t r u c t u r e s , a n d the relevant p a r a m e t e r is best char-
acterized as d e p e n d e n c e .

(22) M-DEPENDENT NOT INDEPENDENT

P-DEPENDENT: Affix Clitic
N O T P-DEPENDENT: Root Stem

T h e cells in ( 2 2 ) m a t c h the four c a t e g o r i e s of m o r p h e m e s depicted in ( 2 1 ) . W e

h a v e thus factored apart the d e p e n d e n c e properties of stems, roots, affixes, and
clitics.
T h e question n o w arises as to h o w to e n c o d e t h e p r o p e r t y of d e p e n d e n c e in an
enlightening and useful fashion. At least in the c a s e of m o r p h o l o g i c a l d e p e n d e n c e ,
w e can m a k e u s e of existing p r o p o s a l s by L i e b e r ( 1 9 8 0 ) , Selkirk ( 1 9 8 2 ) , K i p a r s k y
( 1 9 8 3 ) , Sproat ( 1 9 8 5 ) , and others to e n c o d e the r e q u i r e m e n t of a m o r p h o l o g i c a l
sister in a subcategorization frame.
Deriving Cyclicity 87

(23) A F F I X (e.g., the E n g l i s h suffix -ity): N ( ( )

A m a ity) m a

R O O T (e.g., the E n g l i s h root -mit): v\[+Lat]( >m„ m i t

) «
m

A logical a p p r o a c h w o u l d b e to p r o p o s e a parallel notion for p - s t r u c t u r e . 11

(24) A F F I X (e.g., the E n g l i s h suffix -ity): [[ ] p a ity] p a

CLITIC (e.g., t h e E n g l i s h auxiliary - s): 9

[[ ] p < 0 z] p w

Several a d v a n t a g e s follow from the p r o p o s a l to assign distinct p - a n d m-frames

in the lexicon. O n e is the potential for a p r i n c i p l e d division of labor b e t w e e n the
t w o types of entities. W o r k i n g in a f r a m e w o r k w h e r e m - f r a m e s a n d p-frames c o n -
strain the s a m e constituents, S p r o a t ( 1 9 8 5 ) p r o p o s e d that m - f r a m e s c o u l d e n c o d e
d o m i n a n c e relations in m - s t r u c t u r e , w h i l e p-frames w o u l d e n c o d e linear p r e c e -
d e n c e relations. W h i l e in o u r c u r r e n t p r o p o s a l m - f r a m e s a n d p-frames refer to
different constituent hierarchies, the idea that the responsibility for linear p r e c e -
d e n c e m i g h t b e c o n s i g n e d to p-frames is still applicable, thus e l i m i n a t i n g s o m e
potential r e d u n d a n c y from the theory.
A further role that p - f r a m e s alone w o u l d p e r f o r m u n d e r this a p p r o a c h w o u l d b e
e n c o d i n g sensitivity of the b o u n d f o r m to p h o n o l o g i c a l p r o p e r t i e s of its sisters.
F o r e x a m p l e , affixes w h i c h are sensitive to the stress or syllabification or s e g m e n t
structure of the b a s e w o u l d state these r e q u i r e m e n t s in p-frames. T h i s m a k e s the
prediction that b o u n d roots, w h i c h lack such frames, will not i m p o s e this type of
r e q u i r e m e n t , a prediction w h i c h as far as I k n o w is c o r r e c t . 12

A third a d v a n t a g e to p r o p o s i n g distinct p - a n d m - s u b c a t e g o r i z a t i o n frames is

the resulting ability to c h a r a c t e r i z e exceptionality in o n e d o m a i n w h i c h d o e s n o t
carry o v e r into the other. F o r e x a m p l e , the u n p r e d i c t a b l e invisibility inherent to
particular suffixes in E n g l i s h , J a p a n e s e , and other l a n g u a g e s can b e stated purely
in t e r m s of their lexical p - f r a m e s , w i t h o u t c o m p l i c a t i n g their perfectly n o r m a l
m o r p h o l o g i c a l representation. If, like m - f r a m e s , p-frames e n c o d e b a s i c informa-
tion a b o u t the sister a n d the m o t h e r of the relevant affix, then the contrast b e t w e e n
lexically invisible a n d ( n o r m a l ) lexically visible affixes b e c o m e s o n e of the rela-
tive location of the e d g e of the mother, relative to the s u b c a t e g o r i z i n g affix.

(25) a. U N D E R L Y I N G REPRESENTATION
OF A VISIBLE SUFFIX.' = [[ ] p Suffix] p

b. U N D E R L Y I N G REPRESENTATION
O F A N I N V I S I B L E SUFFIX: =[[ ] ] p Suffix

O u r earlier decision to interpret the P C F a l g o r i t h m as a purely structure-filling

e l s e w h e r e a l g o r i t h m saves the invisible suffix in (25b) from b e i n g i m m e d i a t e l y
regularized in the derivation. M o r e specifically, the information in the lexical
frame of the affix a b o u t the p-status of its m o t h e r overrides the default a l g o r i t h m
and survives.
88 Sharon Inkelas

To exemplify, (26) p r o v i d e s a partial derivational history, in p-structure, of the

suffixed form historic-al. S h o w n b y H a y e s (1981) to b e inherently invisible,
the suffix -al will b e associated with the t y p e of entry s c h e m a t i z e d in (25b). ( T h e
partial derivation takes place at stratum 1—that is, the p-constituents in question
are of type a . I h a v e o m i t t e d c a t e g o r y labels t o i m p r o v e legibility.)

(26) INPUT: [historic] p

A F F I X A T I O N O F [[ ] ] al:
p p [ [ h i s t o r i c ] ] al
p p

PCF: —
PHONOLOGICAL RULES APPLY TO: [historic] p al

A s c a n b e seen in the a b o v e derivations, the b r a c k e t erasure, o r locality c o n v e n -

tion, is a s s u m e d t o b e a u t o m a t i c , defining the a m o u n t of p-structure p h o n o l o g i c a l
rules c a n access at a n y given p o i n t in the derivation.

4. A F F I X A T I O N A N D C Y C L I C I T Y I N P R O S O D I C
LEXICAL PHONOLOGY

T h u s far w e h a v e identified t h e units of representation in a theory w h i c h con-

strues lexical rule d o m a i n s as units of p-structure a n d utilizes t h e distinction b e -
t w e e n m - a n d p-structure to c a p t u r e certain a s y m m e t r i e s that have b e e n o b s e r v e d
to obtain b e t w e e n t h e t w o . W e h a v e also seen t w o sources for p-structure in t h e
lexicon: P C F , a n d p - s u b c a t e g o r i z a t i o n . E a c h of these is m o t i v a t e d indepen-
13

dently of a n y discussion of cyclicity. P C F is required t o supply p - c o n s t i t u e n c y t o

n o n d e r i v e d s t e m s , a n d p - f r a m e s a r e n e e d e d t o represent direction of attachment,
p h o n o l o g i c a l constraints o n p o s s i b l e sisters, a n d p-structure irregularities of af-
fixes (and, in principle, clitics). W e h a v e also a s s u m e d that p h o n o l o g i c a l rules
apply automatically t o the d o m a i n described b y t h e u p p e r m o s t n o d e in p-structure
i m m e d i a t e l y u p o n the creation of that n o d e .
T h e relevance of this d i s c u s s i o n to the issue of cyclic rule application is that the
very e l e m e n t s of the t h e o r y p r e s e n t e d thus far c o n s p i r e t o predict cyclicity at all
word-internal stages of the derivation.

(27) A l l m o r p h o l o g i c a l p r o c e s s e s will correlate with t h e construction of a

p-constituent; thus, all will u n d e r g o cyclic p h o n o l o g i c a l rules.

In this s t r o n g claim, P L P distinguishes itself from other theories of t h e mor-

p h o l o g y - p h o n o l o g y interface. F o r e x a m p l e , as w e shall s e e later, other m o d e l s
h a v e c l a i m e d that cyclicity o r noncyclicity is a stipulated property of each lexical
stratum (e.g., Halle a n d M o h a n a n , 1 9 8 5 ) — o r even that it is a diacritic property
specified directly for e a c h individual affix (e.g. Halle a n d K e n s t o w i c z , 1991). B u t
this kind of arbitrary variety is n o t possible within PLP. D u e t o the generic state-
m e n t of P C F , all levels of m - s t r u c t u r e will e x p e r i e n c e m a p p i n g t o p-structure in
Deriving Cyclicity 89

the s a m e fashion. A n d d u e to the e x i s t e n c e of p-frames in the u n d e r l y i n g repre-

sentation of every affix, p-structure will b e created a n e w with every instance of
affixation.
In the following sections, w e will e x p l o r e three specific places in the derivation
w h e r e P L P p e r m i t s — o r f o r c e s — r u l e s to apply noncyclically. T h e noncyclicity of
these three m o r p h o l o g i c a l l y defined e n v i r o n m e n t s follows from i n d e p e n d e n t l y
n e e d e d p r o p e r t i e s of the theory. I a r g u e , moreover, that they cover the data used
in the literature to m o t i v a t e less c o n s t r a i n e d m o d e l s of m o r p h o l o g y - p h o n o l o g y
interaction (discussed in Section 5). T h e s e three n o n c y c l i c p h e n o m e n a are: the
apparently universal n o n a p p l i c a t i o n of cyclic rules to b o u n d roots; the l a n g u a g e -
specific n o n a p p l i c a t i o n of cyclic rules to unaffixed s t e m s prior to initial affixation;
and the language-specific n o n c y c l i c application of p h o n o l o g i c a l rules to w o r d s
at the e n d of the lexical derivation. I turn first to the t r e a t m e n t of b o u n d roots
within PLP.

4.1. Bound Roots

B o u n d roots represent the o n e area in w h i c h past theories of the p h o n o l o g y -

m o r p h o l o g y interface h a v e o v e r g e n e r a t e d cyclicity. In particular, the version of
lexical p h o n o l o g y e x p o u n d e d in K i p a r s k y ( 1 9 8 2 ) predicts a p h o n o l o g i c a l cycle of
rules after e a c h m o r p h o l o g i c a l p r o c e s s , including the insertion of b o u n d roots. B u t
as K i p a r s k y o b s e r v e s (see also B r a m e , 1974; H a r r i s , 1983), these m o r p h e m e s d o
not constitute p h o n o l o g i c a l rule d o m a i n s o n their o w n , failing to u n d e r g o such
cyclic rules as stress a s s i g n m e n t — a n d even, in s o m e c a s e s , to c o n f o r m to m o r -
p h e m e structure constraints. T h e m i n i m a l m o r p h o l o g i c a l stature required for
cyclic p h o n o l o g i c a l rule application is that of free-standing stem, s u c h as w o u l d
b e created w h e n a b o u n d root is supplied with a m o r p h o l o g i c a l sister.
T h e free stem is precisely the input required in o r d e r to trigger P C F (9). Past
a c c o u n t s h a v e simply had to stipulate that b o u n d roots d o not qualify as p h o n o -
logical rule d o m a i n s . However, this generalization follows naturally from the
tenets of PLP, w h i c h is u n a b l e to assign such roots t h e requisite p-structure.
This point is illustrated in (28), w h i c h contrasts the 'derivation' of an unaffixed
b o u n d root (-ceive) with that of a free stem {stem).

(28) B O U N D ROOT STEM

UNDERLYING
REPRESENTATION: (() m a ceive) m a stem
MCF: (stem)
PCF: [stem] p a

PHONOLOGICAL RULES: [stem] p a

T h u s P L P derives the o n e instance of universally prohibited cyclic rule appli-

cation w i t h o u t any e x t r a stipulation.
90 Sharon Inkelas

T h e scenario predicted b y P L P is thus far o n e of absolute e x t r e m e s : rules never

apply to b o u n d roots, b u t a l w a y s apply in a cyclic fashion t o every (derived o r
n o n d e r i v e d ) free stem. H o w e v e r , this rigidity is s o m e w h a t illusory. P L P in fact
p o s s e s s e s flexibility at both e n d s of t h e lexical derivation. T h e flexibility derives
from t h e interaction b e t w e e n t h e t w o s o u r c e s of p-structure p r o v i d e d b y t h e
theory: affixation a n d P C F . A s w e will s e e , t h e o r d e r in w h i c h l a n g u a g e s c h o o s e
to t a p these t w o d e v i c e s allows for s o m e variation in t h e construction of p -
d o m a i n s at b o t h the very b e g i n n i n g a n d t h e very e n d of t h e lexicon. W e turn first
to t h e possibilities arising at the b e g i n n i n g .

4.2. Preaffixal S t e m C y c l e i n S t r a t u m 1

It is r e c o g n i z e d that b o u n d roots never constitute rule d o m a i n s . B u t is it t h e case

that n o n d e r i v e d s t e m s a l w a y s d o ? W h i l e P L P m a k e s this situation possible, it d o e s
not force it.
T h e t w o sources of p-structure in t h e lexicon m a y in p r i n c i p l e interact either in
a feeding m a n n e r o r in a b l e e d i n g m a n n e r . T h e f o r m e r is w h a t h a s b e e n implicitly
a s s u m e d thus far. T h a t is, w e h a v e a s s u m e d that before acquiring a n affix, a free
stem will u n d e r g o P C F a n d , as a n a u t o m a t i c c o n s e q u e n c e , a cycle of p h o n o l o g i c a l
rules.

(29) P C F FEEDS A F F I X A T I O N :
INPUT: stem
PCF: [stem]p
F I R S T CYCLE O F P H O N O L O G I C A L R U L E S : [stem] p

A F F I X A T I O N O F [[ ] s u f f i x ] :
p p [stem-suffix] p

S E C O N D CYCLE O F P H O N O L O G I C A L R U L E S : [stem-suffix] p

However, t h e o p p o s i t e o r d e r of application is also logically possible. U n d e r a

unification-based view o f s u b c a t e g o r i z a t i o n in w h i c h t h e c o m b i n a t i o n o f a s u b -
categorization frame with a n o t h e r m o r p h e m e is p e r m i t t e d a s long a s t h e output is
w e l l - f o r m e d , n o t h i n g will p r e v e n t a n affix from attaching directly t o a stem
14

w h i c h itself lacks p - s t r u c t u r e — t h e r e b y s u p p l y i n g t h e stem with that structure.

(30) AFFIXATION BLEEDS P C F :

INPUT: stem
A F F I X A T I O N OF [[ ] s u f f i x ] :
p p [stem-suffix] p

FIRST CYCLE OF PHONOLOGICAL RULES: [stem-suffix] p

PCF: —

D u e to the b r a c k e t erasure/locality c o n v e n t i o n s a s s u m e d here, there will b e n o

representational difference b e t w e e n f o r m s derived in t h e o r d e r s h o w n in (29) a n d
forms derived in t h e order s h o w n in ( 3 0 ) . B o t h p r o d u c e [stem-suffix] as output; p
Deriving Cyclicity 91

only (29), however, derives an i n t e r m e d i a t e stage w h e r e the input stem is an avail-

able constituent in p - s t r u c t u r e . O n l y in (29) is there a pre-affixal stem c y c l e .
W h a t I w o u l d like t o a r g u e h e r e is that b e c a u s e P C F a n d Affixation lack intrin-
sic o r d e r i n g , their o r d e r is p a r a m e t e r i z a b l e across l a n g u a g e s . In other w o r d s , there
o u g h t t o b e l a n g u a g e s in w h i c h Affixation b l e e d s P C F , predicting n o stem cycles
in c a s e an affix is attached; a n d l a n g u a g e s in w h i c h P C F a l w a y s applies first, in-
d u c i n g obligatory s t e m c y c l e s u n d e r all c i r c u m s t a n c e s . T h e t w o typological p o s -
sibilities are s c h e m a t i z e d below.

(31) TYPE A TYPE B

UNDERLYING UNDERLYING stem
REPRESENTATION: Stem REPRESENTATION:
PCF: [stem] p AFFIXATION: [stem-suffix] p

AFFIXATION: [stem-suffix] p PCF: —

(2 c y c l e s ) (1 cycle)

(Of c o u r s e , a stem w h i c h takes n o affixes will b e treated identically in both

scenarios, as will s t e m s w h i c h h a v e already u n d e r g o n e affixation.)
T h i s typological prediction is u n a v a i l a b l e to direct-access versions of lexical
p h o n o l o g y (Kiparsky, 1 9 8 2 ; M o h a n a n , 1 9 8 2 ; Pesetsky, 1979), w h i c h have n o
m e a n s of differentiating t h e t w o types of l a n g u a g e s s h o w n in (31). T h u s , a n y evi-
d e n c e for t h e d i c h o t o m y in (31) h a s dual i m p o r t a n c e : it n o t only supports o n e
particular p r o p o s a l within P L P , b u t it also s u p p o r t s that m o d e l in general o v e r
direct-access m o d e l s of lexical p h o n o l o g y .
W i t h this in m i n d , let us n o w turn to the e v i d e n c e . It h a s already b e e n s h o w n in
the literature that l a n g u a g e s of t y p e A exist. Diyari (Poser, 1989), for e x a m p l e ,
p r e s e n t s a clear c a s e in w h i c h t h e c o n s t r u c t i o n of m e t r i c a l structure o n m o n o m o r -
p h e m i c stems crucially p r e c e d e s a n y affixation. P u l l e y b l a n k ( 1 9 8 6 ) m a k e s similar
a r g u m e n t s for M a r g i a n d Tiv, w h e r e rules of t o n e a s s i g n m e n t ( a n d in the c a s e
of M a r g i , t o n e spreading) m u s t crucially apply to stems before the first suffix is
attached (see, e.g., p p . 6 8 , 7 4 ) . H e r e I p r e s e n t o n e e x a m p l e of a c a n d i d a t e for
T y p e B , n a m e l y C a r i b , a C a r i b a n l a n g u a g e s p o k e n in G u y a n a . T h e analysis is
a m e n d e d from I n k e l a s ( 1 9 8 9 ) ; t h e d a t a c o m e from Hoff ( 1 9 6 8 ) .

4.2.1. CARIB

Stress in n o n d e r i v e d s t e m s in C a r i b is perfectly alternating, subject to t h e

restriction that word-final syllables a r e n e v e r stressed. S o m e n o n d e r i v e d s t e m s
display stress o n o d d - n u m b e r e d syllables, w h i l e others display stress on even-
n u m b e r e d syllables. Phonetically, stress is correlated with vowel l e n g t h e n i n g in
the first t w o (nonfinal) stressed syllables of t h e w o r d , a n d with associated high
pitch with t h e s e c o n d stressed syllable from the left. O n l y vowel length a n d stress
will b e m a r k e d , as in ( 3 2 ) .
92 Sharon Inkelas

(32) O D D - N U M B E R E D STRESS:
N o . of syllables Stem Gloss
1 wo 'to b e a t '
2 e:ro 'this'
ku:pi 'bathe'
3 eimaka 'to c o m b a p a r t i n g '
tdikuwa 'polishing-stone'
4 d:rawd:ta 'howling monkey'
pd: yawd: ru 'cassava beer'
5 auwanoipono ' c a u s i n g laughter'
ko: kapo: take ' y o u will have m e bitten'

(33) E V E N - N U M B E R E D STRESS:
N o . of syllables Stem Gloss

3 akd: mi ' t r u m p e t e r bird'

tono: ro 'large b i r d '
4 ard: mari 'mythical snake'
kuriiyara 'canoe'
5 asd:pard:pi 'species of fish'
wotu: ropo: ro ' c a u s e to a s k '

A s s h o w n in ( 3 4 ) , m o r p h o l o g i c a l l y c o m p l e x forms d o not a d h e r e to the per-

fectly alternating pattern. In particular, b e c a u s e stress is assigned cyclically and is
structure-preserving, forms including prefixes or suffixes can exhibit nonalternat-
ing surface s t r e s s . 15

(34) aki: ma 'tease' ki'n-d: ki: ma: -no ' h e teases h e r '
awo: mi 'to get u p ' ay-a: wo: mi'-i ' y o u m u s t not get
up'
epd: nopi' 'help' ay-e:pd:nopi 'your being helped'
erne: pa 'to teach' ki'n-e:me:pa: -no ' h e teaches t h e m '
etdmboka 'untie' kl'n-e: tdmboka: -no ' s h e unties it'
eyd: to 'call' ki'n-e: yd: to-ya: -tor) 'they call h i m '
kurd: ma 'cure' si-ku: rd: ma-e T cure h i m '
kurd: ma ' l o o k after' ki'-ku: rd: ma-ko ' y o u m u s t look af-
ter m e '
ki'-ku: rd:ma-i 'you m u s t not look
after m e '
i-ku: rd: ma-ko ' y o u m u s t look af-
ter h i m '
Deriving Cyclicity 93

poro:pi 'stop' ni-pd: ro:pi'-i 'actually h e has

stopped'
ni-pd: ro: pi: -se ' so that he m a y
stop'
a-po: ro :pi-i ' y o u m u s t not s t o p '

B o t h the d e r i v e d a n d the n o n d e r i v e d w o r d s listed in the first and third c o l u m n s

of (34) exhibit stress on the s e c o n d syllable. H o w e v e r , n o t e that in e a c h c a s e the
s t e m s also exhibit stress on their o w n s e c o n d s y l l a b l e — t h a t is, the third syllable
of the derived w o r d , w h e r e n o postcyclic rule w o u l d p l a c e it. Of c o u r s e , this stem-
m e d i a l syllable is exactly w h e r e stress w o u l d b e assigned w e r e the prefix not p r e s -
ent. W e m a y thus c o n c l u d e that these s t e m s u n d e r g o stress a s s i g n m e n t both before
and after the prefix is a d d e d . 1 6

F o r m s c o n t a i n i n g s t e m s a n d suffixes only are consistent with the generalization

that stress a s s i g n m e n t is cyclic a n d structure-preserving.

(35) a. ko:roka 'scrub' ko: rokd: no 'he

scrubbed'
koro: mo 'recent' koro: mo-no ' a recent
thing'
b. bd: siya 'deputy-chief bd: siyd: -kof) 'deputy-
chiefs'
pako: to 'to cut w o o d ' pako: to-no 'fact of
wood-
cutting'
c. kuru: wese ' p a l m shell' kuru: wese-mho 'old p a l m
sheath'
d. eyd: to 'call' kin-e: yd: to-yd:-tog 'they call
him'

U n l i k e prefixation, however, suffixation n e v e r results in adjacent stresses.

T h e c o n t r a s t b e t w e e n o d d - a n d e v e n - n u m b e r e d stress is p h o n e m i c only w i t h i n
stems, t h o u g h p r e s e r v e d u n d e r s u b s e q u e n t derivation, as w e h a v e seen, by the
structure-filling stress rule. Hoff indicates that of the t w o p a t t e r n s , the o d d -
n u m b e r e d stress pattern is less frequent (p. 7 3 ) . T h i s suggests that stems with the
e v e n - n u m b e r e d stress pattern represent the u n m a r k e d c a s e .
G i v e n this, w e still are faced with t w o options for the basic stress rule. A s s u m -
ing that the rule o p e r a t e s from left to right, the basic stress foot could b e i a m b i c ,
or it c o u l d also b e t r o c h a i c , in w h i c h c a s e it w o u l d n e e d to follow a rule assigning
initial syllable extrametricality.

(36) IAMBIC A N A L Y S I S TROCHAIC ANALYSIS

C Y C L E 1: [kura.ma] ku [rama]
C Y C L E 2: [iku.ra.ma] i [ku.rama]
94 Sharon Inkelas

T h e i a m b i c analysis w o u l d require n e w l y inserted stress feet to overwrite w e a k

stress foot t e r m i n a l s a s s i g n e d on an earlier c y c l e (in order, for e x a m p l e , to p r o d u c e
the adjacent stresses in i-kurdma). B y contrast, the trochaic analysis w o u l d require
only that initial invisibility m u s t b e lost u p o n further prefixation, an a s s u m p t i o n
consistent with virtually all theories of extrametricality (including the o n e as-
s u m e d here).
I will thus tentatively a s s u m e that the correct analysis is trochaic feet and that
w e require the t w o rules in (37) a n d ( 3 8 ) . 1 7

(37) I N I T I A L S Y L L A B L E INVISIBILITY: [a . . .] p —> a[.. .] p

(38) STRESS ASSIGNMENT: F r o m the left e d g e , assign syllabic t r o c h e e s .

I turn n o w to the q u e s t i o n of h o w to c a p t u r e the lexical distinction b e t w e e n

stems falling into the o d d - n u m b e r e d pattern a n d those falling into the even-
n u m b e r e d pattern. O n e possibility is to a n n o t a t e o d d - n u m b e r e d stems with an
e x c e p t i o n feature b l o c k i n g the rule of extrametricality a s s i g n m e n t from affecting
t h e m . B u t a feature of this t y p e h a s n o theoretical status, a n d w e m a y thus reject it
as n o n e x p l a n a t o r y . A s e c o n d possibility is to assign these w o r d s an initial stress
foot in u n d e r l y i n g representation. Since the stress foot a s s i g n m e n t p r o c e s s is
structure-preserving, this initial foot will r e m a i n intact, c a u s i n g s u b s e q u e n t foot
a s s i g n m e n t to b e g i n on the third s y l l a b l e .
18

(39) REGULAR IRREGULAR

U N D E R L Y I N G REPRESENTATION: asaparapi pdyawaru
PCF: [asaparapi] [payawaru]
EXTRAMETRICALITY : a [saparapi] [payawaru]
TROCHEE ASSIGNMENT: a [saparapi] [payawaru]
(other rules) [ a s a : p a r a : pi] [ p a : y a w a : ru]
OUTPUT: ' s p e c i e s of fish' 'cassava beer'

W e are n o w p r e p a r e d to e x a m i n e the e v i d e n c e against a preaffixal stem c y c l e in

C a r i b . T h e first set of d a t a is disyllabic s t e m s .
D e s p i t e their a d h e r e n c e in citation form to the surface prohibition against final
stress, disyllabic s t e m s n o n e t h e l e s s separate into the s a m e t w o g r o u p s into w h i c h
larger stems are divided [see ( 3 2 ) - ( 3 3 ) ] . S o m e surface with initial, or o d d -
n u m b e r e d , stress in every e n v i r o n m e n t (40); w e m a y p l a c e these firmly in the o d d -
n u m b e r e d category. H o w e v e r , the other set exhibits alternating stress (41). W h e n
uttered alone, w i t h o u t suffixes, these s t e m s exhibit initial stress. B u t in the c o m -
p a n y of any suffix, they exhibit second-syllable, or e v e n - n u m b e r e d stress. W e thus
place t h e m in the e v e n - n u m b e r e d c a t e g o r y and require an explanation for their
b e h a v i o r in isolation.

(40) " O D D - N U M B E R E D STRESS" :

e:ro 'this' e:ro-me now
kd: mi 'flame' kd: mi-ro ' c a u s e to b e c o m e pale r e d '
kd:wo 'high' kd:wo-nd:ka 'from a b o v e '
kd: wo-nd: -kor) 'high ones'
Deriving Cyclicity 95

(41) " E V E N - N U M B E R E D STRESS" :

a: pi 'red, r i p e ' a}
api: -ro ' c a u s e to ripen'
e:ta 'hear' eta:
ei -topo ' m e a n s of h e a r i n g '
kdirai 'blackness' kardi-ma
kc 'blacken'
u:wa 'dance' uwd:
m -no 'being dancing'

T h e analysis d e v e l o p e d thus far w o u l d predict that the o d d - n u m b e r e d pattern is

the m a r k e d case, r e q u i r i n g a lexically specified foot (42a). This is consistent with
the n o n a l t e r n a t i n g nature of the stress of these disyllabic s t e m s . B y contrast, di-
syllabic s t e m s e x h i b i t i n g the e v e n - n u m b e r e d pattern are lexically unspecified
( 4 2 b ) — c o r r e l a t i n g with the context-sensitivity of their stress patterns.

(42) UNDERLYING REPRESENTATION:

a. karai 'blackness' Unmarked "even-numbered"

b. kdwo 'high' Marked "odd-numbered"

G i v e n these u n d e r l y i n g f o r m s , w e c a n n o w p r o v e that disyllabic s t e m s m u s t not

u n d e r g o a s t e m c y c l e prior to affixation (if a n y ) . To see w h y this m u s t b e so, let
us start b y a s s u m i n g the o p p o s i t e , n a m e l y that there is a preaffixal stem cycle.
F r o m the fact that all disyllabic s t e m s (including t h o s e in the " e v e n - n u m b e r e d "
c a t e g o r y ) p o s s e s s initial stress in isolation, w e k n o w that the extrametricality rule
m u s t fail to a p p l y on the c y c l e at w h i c h stress is assigned to t h e m . This c o u l d
plausibly b e attributed to a m i n i m a l i t y condition of the kind o b s e r v e d earlier in
A m e l e . H o w e v e r , even a s s u m i n g that w e m a n a g e to b l o c k invisibility from apply-
ing to disyllabic s t e m s , the s u b s e q u e n t a s s i g n m e n t of initial stress p o s e s a p r o b -
lem. O n c e stress is a s s i g n e d to the first syllable of disyllabic stems, then it should
thereafter exhibit the s a m e n o n a l t e r n a t i n g b e h a v i o r exhibited b y stems with un-
derlying stress. B u t this is not the c a s e : c o n t r a r y to the predictions of a stem cycle
account, the contrast b e t w e e n o d d - a n d e v e n - n u m b e r e d disyllabic stems is not
neutralized.

(43) " S T E M CYCLE" ACCOUNT:

REGULAR IRREGULAR
UNDERLYING
REPRESENTA- karai kdmi
TION:
PCF: [karai] [kami]
INVISIBILITY: [blocked b y m i n i m a l i t y ]
STRESS: [karai]
AFFIXATION: [karaima] [kamiro]
INVISIBILITY: ka [raima] ka [miro]
STRESS:
(late rules) *[ka:raima] [ka: miro]
'blacken' ' c a u s e to b e c o m e
pale r e d '
96 Sharon Inkelas

W e m a y thus c o n c l u d e that there c a n n o t b e a preaffixal stem cycle for disyllabic

s t e m s . Instead, Affixation m u s t b l e e d P C F , resulting in a stem cycle only on un-
affixed s t e m s .

(44) "No STEM C Y C L E " A C C O U N T :

REGULAR IRREGULAR
U N D E R L Y I N G REPRESENTATION: karai kdmi
AFFIXATION: [karaima] [kamiro]
INVISIBILITY: k a [raima] k a [miro]
STRESS: k a [raima] —
PCF: — —
(late rules) [karaima] [ k a : miro]
'blacken' ' c a u s e to b e c o m e
pale r e d '

T h e s e c o n d set of data s u p p o r t i n g the a b s e n c e of a stem cycle in C a r i b c o m e s

from the b e h a v i o r of a small set of " s t r o n g " suffixes w h i c h a p p e a r to alter the
stress p a t t e r n s of t h e b a s e s to w h i c h they attach. A s s h o w n in (45), they c a u s e
s t e m s o t h e r w i s e exhibiting e v e n - n u m b e r e d stress (a) to switch to an odd-
n u m b e r e d stress pattern. T h e contrast b e t w e e n s t e m s in the e v e n - n u m b e r e d cate-
g o r y (a) and those in the o d d - n u m b e r e d c a t e g o r y (b) is neutralized in the e n v i r o n -
m e n t of these suffixes.

(45) ISOLATION BEFORE "STRONG'

SUFFIX

a. kuri: yara ku: riyd: ra-ri' canoe

yamd: tu ydimatu: -ru 'basket'
b. su: rabarj su: rabd: -ni' ' b e a m of r o o f
kd: rawd: si kd: rawd: si-ri' 'rattle'

Strikingly, this neutralization o c c u r s only w h e n the b a s e is consonant-initial.

T h e stress p a t t e r n s of c o m p a r a b l e vowel-initial w o r d s in the e v e n - n u m b e r e d cate-
g o r y r e m a i n unaltered in the e n v i r o n m e n t of " s t r o n g " suffixes, as seen in (46).

(46) ISOLATION BEFORE "STRONG"

SUFFIX
aki: nu aki:nu-ru 'laziness' (*d: kind: -ru)
ineiku ine:ku-ru 'liana' (*i':neku: -ru)
i'rd:pa ird:pa-ri 'bow' : rapd: ri)
oko: mo okoimo-ri 'wasp' (*6:komo: -ri)

T h e r e are t w o p u z z l e s here. First, w h y d o e s the stress pattern shift at all in the

forms in (45)? W e k n o w C a r i b p o s s e s s n o general cyclic stress erasure p h e n o m e -
n o n . S e c o n d , w h y is the stress shift restricted to consonant-initial w o r d s ? O n e
possibility is to a s s u m e that the " s t r o n g " suffixes trigger a special rule of stress
foot deletion, w i p i n g the slate clean for the s u b s e q u e n t a s s i g n m e n t of o d d -
n u m b e r e d stress. H o w e v e r , this analysis is suspect for t w o r e a s o n s . First, odd-
Deriving Cyclicity 97

n u m b e r e d stress is s u p p o s e d to result only from lexical prespecification, not from

tabula rasa e n v i r o n m e n t s . S e c o n d , there is n o e x p l a n a t i o n for the correlation with
initial c o n s o n a n t s .
T h e solution I a d o p t instead is as follows. First, I c o n c l u d e from the contrasting
b e h a v i o r of c o n s o n a n t - a n d vowel-initial s t e m s that the rule of initial syllable in-
visibility p r o p o s e d earlier m u s t actually b e b r o k e n d o w n into t w o c o m p o n e n t s ,
initial c o n s o n a n t invisibility a n d initial vowel i n v i s i b i l i t y . W h e n b o t h apply, the
19

c o n s o n a n t invisibility rule feeds the vowel invisibility rule, giving the a p p e a r a n c e

of initial syllable invisibility. H o w e v e r , should the c o n s o n a n t invisibility rule b e
b l o c k e d from a p p l y i n g , the vowel invisibility rule will also fail, a n d the initial
syllable of the w o r d will b e visible for stress p u r p o s e s .
To a c c o u n t for the contrast b e t w e e n c o n s o n a n t - and vowel-initial stems u n d e r
" s t r o n g " suffixation, I further p r o p o s e that c o n s o n a n t invisibility is actually p a r t
a n d parcel of C a r i b ' s version of the P C F a l g o r i t h m .

(47) Initial Vowel Invisibility: [v . . . ] —» v [.. .]

(48) P C F (Carib): ( c . . . ) m -» <c . . . ) m c [.. . ] p

" S t r o n g " suffixes b l o c k the i m p o s i t i o n of initial c o n s o n a n t invisibility b y

b l o c k i n g the application of P - C o n s t i t u e n t F o r m a t i o n . T h a t is, in C a r i b , Affixation
p r e c e d e s — a n d thereby p r e e m p t s — P C F .

(49) C a r i b : Affixation < ( M C F < ) P C F

D e r i v a t i o n s of unaffixed (50) and suffixed (51) consonant-initial s t e m s are s h o w n

below:

(50) REGULAR IRREGULAR

U N D E R L Y I N G REPRESENTATION: kuriyara kdrawasi
MCF: (kuriyara) m (karawasi)m

PCF: k [uriyara] p k [arawasi] p

I N I T I A L V INVISIBILITY: ku [ r i y a r a ] p ka [rawasi] p

STRESS ASSIGNMENT: ku [rfyara]p ka [rawasi] p

(late rules) [kun:yara] p [ka:rawa:si] p

'canoe' 'rattle'

(51) REGULAR IRREGULAR

U N D E R L Y I N G REPRESENTATION: kuriyara kdrawasi
MCF <kuriyara) m (karawasi) m

AFFIXATION: (kuriyarari) m (karawasiri) m

[kuriyarari'lp [karawasirilp
I N I T I A L V INVISIBILITY: — —
STRESS ASSIGNMENT: [kuriyarari] p [karawasirilp
PCF: — —
(late rules) [kuriyarari'lp [karawasirilp
'canoe' 'rattle'
98 Sharon Inkelas

S i n c e the effect of b l o c k i n g P C F by Affixation is manifested only in c o n s o n a n t -

initial w o r d s , w e correctly predict n o alternations in the suffixed and unsuffixed
forms of vowel-initial s t e m s . T h e contrast b e t w e e n the b e h a v i o r of regular
consonant-initial a n d vowel-initial s t e m s u n d e r " s t r o n g " affixation is g e n e r a t e d
as follows.

(52) VOWEL-INITIAL CONSONANT-INITIAL

U N D E R L Y I N G REPRESENTATION: okomo yamatu
MCF: (okomo) m (yamatu) m

AFFIXATION: (okomori) m (yamaturi) m

[okomori] p [yamaturi]p
V O W E L INVISIBILITY: o [komori] p —
STRESS: o [k6mori] p [yamaturijp
PCF: — —
(late rules) [oko:mori] p [ya:matu:ri]p
'wasp' 'basket'

Vowel-initial stems will u n d e r g o Initial Vowel Invisibility regardless of w h e t h e r

Affixation or P C F applies first a n d will exhibit the s a m e stress patterns with or
w i t h o u t a strong suffix:

(53) SUFFIXED UNSUFFIXED

UNDERLYING
REPRESENTATION: okomo okomo
MCF: (okomo) m (okomo) m

AFFIXATION: (okomori) m

[okomori]p
V O W E L INVISIBILITY: o [komori'lp
STRESS RULE: o [komorilp
PCF: [okomo] p

V O W E L INVISIBILITY: o [komo] p

STRESS RULE: o [komo]p

(late rules) [oko:mori] p [oko:mo] p

'wasp(poss.)' 'wasp'

Finally, irregular, that is, o d d - n u m b e r e d , stress pattern stems will b e i m m u n e to

the effects of the strong suffixes, b e c a u s e their u n d e r l y i n g initial stress foot p r e -
vails regardless of the visibility or invisibility of their initial s y l l a b l e . 20

In s u m m a r y , w h i l e stress a s s i g n e d cyclically in C a r i b is n o r m a l l y preserved, the

alternations exhibited b e t w e e n suffixed a n d unsuffixed stems s h o w s that stress
m u s t not b e a s s i g n e d to s t e m s p r i o r to the insertion of the first suffix. T h i s moti-
vates the o r d e r i n g of P C F following Affixation in C a r i b .
Deriving Cyclicity 99

4.3. Noncyclic = Postcyclic

W e noted earlier that the source from w h i c h P L P derives cyclic rule application
is the p r e s e n c e of p - s u b c a t e g o r i z a t i o n frames in t h e lexical entries of affixes. A n y
stratum associated with affixation will necessarily b e associated with cyclic p h o -
nological rule application.
H o w e v e r , this leaves o p e n the possibility that a particular stratum m i g h t h a v e
n o affixes associated with it. In fact, this is the c a s e with the (postcyclic) lexical
w o r d stratum, as d e s c r i b e d b y K i p a r s k y ( 1 9 8 5 ) a n d Booij and R u b a c h ( 1 9 8 7 ) . T h e
d o m a i n of word-level lexical rules, this s t r a t u m d o e s n o t allow any m o r p h o l o g y
to take p l a c e .
In t e r m s of PLP, w e m a y c a p t u r e the nonaffixational nature of a stratum with a
constraint in u n d e r l y i n g representation against affixes s u b c a t e g o r i z i n g for c o n -
stituents of the a p p r o p r i a t e category. In other w o r d s , Affixation is t u r n e d off at a
particular level of the hierarchy. In t h e c o r r e s p o n d i n g constraint in (54), only the
m - f r a m e is m e n t i o n e d ; the a s s u m p t i o n is that since affixes will a l w a y s h a v e
m - f r a m e s , r e p e a t i n g the constraint o n p-frames w o u l d at best b e r e d u n d a n t . 21

(54) *(x, ( ) m ) m , w h e r e i > j , j a c a t e g o r y in the p - h i e r a r c h y

T h i s constraint interacts with a n o t h e r constraint, that g o v e r n i n g the level of the

hierarchy u p to w h i c h P C F o p e r a t e s in the lexicon. W e k n o w that constituents
b e l o w the level of the p h o n o l o g i c a l w o r d are strictly lexical; those at the p h r a s e
level a n d a b o v e are strictly postlexical. T h i s can b e c a p t u r e d formally with a lexi-
cal constraint prohibiting the o c c u r r e n c e of constituents above s o m e cutoff level
k in the lexicon.

(55) *[] pic w h e r e k < n; n a c a t e g o r y in the p - h i e r a r c h y

In fact, the constraint in (55) m a y b e v i e w e d as j u s t a special c a s e of structure

p r e s e r v a t i o n (Kiparsky, 1985).
W h e t h e r or n o t a system incorporating the constraint in (54) will p o s s e s s a
n o n c y c l i c level in the lexicon d e p e n d s entirely on the instantiation of ( 5 5 ) — t h a t
is, on w h e t h e r or not P C F is still o p e r a t i n g at level j . If P C F persists for a level
b e y o n d the Affixation cutoff, then a n o n c y c l i c level will result. If, however, P C F
and Affixation h a v e the s a m e cutoff point, then n o n o n c y c l i c lexical level is
expected.

(56) a. j < n: there will b e a n o n c y c l i c w o r d level

b . j > n: n o effect, n o n o n c y c l i c w o r d level e x p e c t e d

It follows as a corollary of this prediction that only o n e stratum, in particular

the last o n e , can b e m a d e n o n c y c l i c (57).

(57) At m o s t o n e n o n c y c l i c lexical level will occur; if so, it will b e the final

level.
100 Sharon Inkelas

(57) c o r r e s p o n d s to the c l a i m s about n o n c y c l i c lexical levels m a d e by Kiparsky

( 1 9 8 5 ) a n d Booij a n d R u b a c h ( 1 9 8 7 ) . Insofar as it follows from w h a t w e h a v e
p r o p o s e d , the theory is supported.
Let us first e x a m i n e the c a s e in w h i c h a n o n c y c l i c level is g e n e r a t e d . T h i s will
occur, I h a v e said, w h e n j in (54) is identified with a wow-final stratum in the
l e x i c o n — t h a t is, w h e n j in (54) is l o w e r in the h i e r a r c h y than the cutoff p o i n t n in
(55). In such a case, P C F will c o n t i n u e to g e n e r a t e constituents a b o v e the cutoff
level for affixation. B e c a u s e of the lack of m o r p h o l o g y , this level (or levels) will
be n o n c y c l i c .
It is clearly necessary, given this source, for n o n c y c l i c levels to b e at the e n d of
the lexicon. B u t h o w d o w e e n s u r e that there is only o n e ? T h i s , I argue, follows
not from inherent constraints in the theory but rather from learnability. T h e evi-
d e n c e that s o m e l a n g u a g e h a s t w o or m o r e c o n t i g u o u s levels at w h i c h affixation
is i m p o s s i b l e , as in (58), w o u l d b e slim at best.

(58) ...
affixational s t r a t u m k (cyclic)
non-affixational s t r a t u m k+1 ( n o n c y c l i c )
non-affixational stratum (noncyclic)

E m p i r i c a l e v i d e n c e for such a situation w o u l d b e identical to e m p i r i c a l evi-

d e n c e for s t r a t u m o r d e r i n g in a l a n g u a g e lacking affixes altogether. T h e l e a r n e r
w o u l d h a v e n o reason to d i v i d e u p into levels a set of o r d e r e d rules (k+1 rules,
k+2 rules) that w o u l d h a v e the s a m e effect if o r g a n i z e d into a single level.
If, on the other h a n d , the c a t e g o r y n in (55), at and above w h i c h n o m o r p h e m e
m a y attach, is a b o v e the l e x i c a l - p o s t l e x i c a l cutoff j in (54), then the statement
about subcategorization will b e v a c u o u s . S i n c e constituents of level j c a n n o t exist
in the lexicon anyway, the b a n on affixation at that level will h a v e n o noticeable
effect. In particular, n o n o n c y c l i c level j will occur.
T h u s , P C F p e r m i t s a single p o s t c y c l i c lexical level, but it d o e s not force o n e .
T h i s m a k e s the prediction that there will b e l a n g u a g e s with such a stratum a n d
l a n g u a g e s without, a prediction I will a t t e m p t to support in this section (see also
Z e e , this v o l u m e ) .
C a s e s for a lexical n o n c y c l i c s t r a t u m h a v e already b e e n m a d e in several p l a c e s
in the literature; see, for e x a m p l e , K i p a r s k y ( 1 9 8 5 ) , Booij and R u b a c h (1987) for
an analysis of Polish, a n d Inkelas and Z e e ( 1 9 8 8 ) , w h e r e a similar a c c o u n t for
S e r b o - C r o a t i a n is p r o v i d e d . B o t h w o r k s discuss rules w h i c h m u s t follow all affix-
ation but display distinctively lexical properties. I will not p r e s e n t these cases
h e r e , as they are available e l s e w h e r e , but c o n c e n t r a t e instead on the less familiar
c a t e g o r y of l a n g u a g e s w h i c h positively lack a postcyclic " w o r d - l e v e l " stratum.
O n e s o u r c e of positive e v i d e n c e that a l a n g u a g e lacked a postcyclic stratum
w o u l d b e a c a s e in w h i c h lexically a s s i g n e d invisibility persisted into the postlex-
ical c o m p o n e n t . T h i s w o u l d s h o w that n o word-level cycle of rules could h a v e
Deriving Cyclicity 101

applied lexically, as the a s s i g n m e n t by P C F of the word-level p-constituent w o u l d

obliterate the trace of invisibility in e m b e d d e d p-structure.
S u c h cases d o occur. Z e e (this v o l u m e ) d e m o n s t r a t e s that S t o k a v i a n dialects of
S e r b o - C r o a t i a n vary with respect to the p a r a m e t e r of w h e t h e r or not the p h o n o -
logical w o r d is available lexically. In m y t e r m s this a m o u n t s to a difference in
w h e t h e r or not the lexicon p e r m i t s P C F to construct a co constituent. Halle and
K e n s t o w i c z ( 1 9 9 1 ) a n a l y z e data from M a n a m and Latin w h i c h m o t i v a t e the a b -
s e n c e of a n o n c y c l i c lexical co level in both l a n g u a g e s . I p r e s e n t the M a n a m c a s e
h e r e as an illustration and then s h o w h o w P L P g e n e r a t e s facts of this kind.

4.3.1. MANAM

M a n a m displays a basic stress pattern in w h i c h stress resides on the word-final

syllable if it is heavy, a n d o t h e r w i s e , on the p e n u l t i m a t e syllable of the w o r d (Lich-
tenberk, 1983). A w a y of stating the generalization w i t h o u t disjunctions is that
stress surfaces on the syllable c o n t a i n i n g the p e n u l t i m a t e m o r a of the w o r d . T h i s
generalization is true of n o n d e r i v e d or prefixed w o r d s , and also of w o r d s e n d i n g
in any of a particular set of suffixes.

(59) UNSUFFIXED SUFFIXED

i-pandna ' h e ran' ?u-lele-?i?o T followed y o u '
wabubu 'night' tamd-da ' o u r (inc.) father'
u-zem T chewed (them)' tamd-T) ' y o u r (sg.) father'
malaborj 'flying fox' tama-mirj ' y o u r (pi.) father'

H o w e v e r , there is a n o t h e r set of suffixes w h i c h L i c h t e n b e r k t e r m s A P (ante-

p e n u l t ) suffixes. T h e s e c a u s e stress to o c c u r on the syllable c o n t a i n i n g the ante-
p e n u l t i m a t e m o r a of w o r d s that they e n d ; consequently, H a l l e a n d K e n s t o w i c z
a n a l y z e t h e s e as t r i g g e r i n g a rule of final syllable extrametricality (p. 4 6 7 ) .

(60) tdma-(ma) ' o u r (excl.) father' ?a-malipi-lipi-(la) ' y o u only

work'
tdma-(da) 'their father' u-rapun-(di) T waited for
them'
di-te-(a) 'they saw m e ' mogdru(r)a-0) 'his n o s e '
sirjdba-(lo) 'in the b u s h ' u-do(?-i) T t o o k it'

S o far, t h e s e facts are c o m p a t i b l e with t w o t y p e s of account. O n e possibility is

that stress is a s s i g n e d cyclically and there is n o postcyclic word-level stratum;
a n o t h e r possibility is that there is such a stratum, b u t n o stress is a s s i g n e d there.
H o w e v e r , e v i d e n c e from cliticization s h o w s that only the f o r m e r alternative is
possible.
M a n a m h a s a n u m b e r of enclitics. C o n v i n c i n g l y a r g u e d b y L i c h t e n b e r k a n d b y
Halle a n d K e n s t o w i c z to b e separate syntactic t e r m i n a l s , they are incorporated into
the p r e c e d i n g w o r d for p h o n o l o g i c a l p u r p o s e s , as s h o w n b y the fact that stress
102 Sharon Inkelas

rules apply to the h o s t - c l i t i c c o m b i n a t i o n . Stress shifts to a host-final unstressed

m o r a in c a s e a m o n o s y l l a b i c clitic follows.
(61) HOST: ?u-do?-i ' y o u take it'
HOST + CLITIC: ?u-do?-i=?i ' y o u take it o r '
W e m a y a s s u m e that M a n a m clitics f o r m p h o n o l o g i c a l w o r d s with their hosts
and that stress is a s s i g n e d postlexically on the d o m a i n of the p h o n o l o g i c a l w o r d .
W h a t is crucial for o u r p u r p o s e s is that this postlexical stress a s s i g n m e n t is
sensitive to a difference o b t a i n i n g a m o n g host-final u n s t r e s s e d m o r a s . T h e rele-
vant difference is b e t w e e n t h o s e final m o r a s w h i c h w e r e e x t r a m e t r i c a l at the t i m e
the lexical stress rule applied, a n d those w h i c h w e r e visible. O n l y the f o r m e r (61),
( 6 2 a ) — n e v e r the latter ( 6 2 b ) — a r e eligible for stress a s s i g n m e n t w h e n a m o n o -
syllabic clitic is a d d e d .

(62) a. ?u-dd?-i ' y o u take t h e m ' ?u-do?-i=?i ' y o u take t h e m o r '

b. ?u-do?-(i) ' y o u take it' ?u-do?-i= ?i ' y o u take it o r '
H a l l e and K e n s t o w i c z c o n c l u d e that it is only b e c a u s e the e x t r a m e t r i c a l suffixes
w e r e never metrified b y the earlier, cyclic rules that they are subject to stress as-
s i g n m e n t at the n o n c y c l i c clitic level. H a l l e a n d K e n s t o w i c z define t w o b l o c k s of
rule a s s i g n m e n t ; cyclic and n o n c y c l i c , and m a r k clitics diacritically as n o n c y c l i c .
B u t the noncyclicity of clitics should n o t h a v e to b e stipulated; it o u g h t to follow
naturally from their postlexical nature.
T h e relevance of d a t a of this k i n d to this p a p e r is that there C A N N O T b e a post-
cyclic stratum in the lexical p h o n o l o g y of M a n a m . W e k n o w from the fact that it
is only the o u t e r m o s t suffix w h i c h affects w o r d stress that stress m u s t b e assigned
cyclically in M a n a m , as H a l l e a n d K e n s t o w i c z h a v e argued. A n d w e k n o w that
the s a m e stress rule w h i c h applies lexically also applies postlexically (though
postlexically it a p p e a r s to respect existing metrical structure). If there w e r e a post-
cyclic lexical level, w e w o u l d incorrectly predict, given that invisibility d i s a p p e a r s
on the cycle (Inkelas, 1989), that all suffixed w o r d s w o u l d b e h a v e alike. A post-
cyclic level w o u l d neutralize the contrast p r o v i d e d b y the difference b e t w e e n vis-
ible and invisible suffixes, a contrast w h i c h m u s t persist until the postlexical level.
In (63) a partial derivation s h o w s h o w the contrast b e t w e e n forms with and with-
out final invisible m o r a s persists until the p o i n t of cliticization.
(63) I N V I S I B L E SUFFIX V I S I B L E SUFFIX
AFFIXATION: [ ? u d o ] . ?i
p [?udo?i] . p Lexical
phonology
STRESS RULE: [?udo] ,?i p
[?udo?i] P (

CLITICIZATION [?udo?i?i] . p
[?ud6?i?i] . p
Postlexical
phonology
STRESS RULE: [?udo?i?i] . p

OUTPUT: ?udo?i?i ?udd?i?i

' y o u t o o k it o r ' 'you took them or'
Deriving Cyclicity 103

T h e partial derivation in (64) s h o w s the b a d results of a neutralizing postcyclic

level of lexical rules.

(64) I N V I S I B L E SUFFIX V I S I B L E SUFFIX

AFFIXATION: [?udo] ?i P ( [?udo?i] P ( Lexical
phonology
STRESS RULE: [?udo] .?i Pi
[?udo?i] p t

PCF: [?udo?i] p m [?udo?i] D i + ,

[?udo?i] P ( + i
STRESS RULE:

[?udo?i?i] [?udo?i?i] Postlexical

CLITICIZATION:
P ( + l
pi+i
phonology
STRESS RULE:
OUTPUT: *?udo?i?i ?udo?i?i
' y o u t o o k it o r ' 'you took them or'

4.4. S u m m a r y of Predictions

B e l o w is a s u m m a r y of the predictions m a d e b y PLP, w h i c h forces cyclicity to

b e the u n m a r k e d c a s e but d o e s predict n o n c y c l i c rule application in a restricted
n u m b e r of e n v i r o n m e n t s .

(65) a. C y c l i c rules d o not apply to b o u n d roots.

b. In s o m e l a n g u a g e s , n o n d e r i v e d s t e m s a l w a y s constitute cyclic rule d o -
m a i n s , regardless of future affixation. In other l a n g u a g e s , the applica-
tion of rules to n o n d e r i v e d s t e m s in p r e e m p t e d b y the a t t a c h m e n t of the
first affix, if any.
c. T h e only n o n c y c l i c lexical rules will apply at the word-level stratum
(i.e., postcyclically), t h o u g h not all l a n g u a g e s will h a v e such a stratum
in the lexicon.

In severely limiting the role of n o n c y c l i c rule application, P L P differs from

m o d e l s that h a v e b e e n p r o p o s e d in the past. T h e version of standard lexical p h o -
n o l o g y d e v e l o p e d b y K i p a r s k y h o l d s that all p h o n o l o g i c a l rule application is cy-
clic, with stipulated e x c e p t i o n s for b o u n d roots (Kiparsky, 1982) and word-level
rules (Kiparsky, 1985). T a k i n g the o p p o s i t e v i e w are H a l l e a n d M o h a n a n ( 1 9 8 5 )
a n d M o h a n a n ( 1 9 8 6 ) , w h o s e version of lexical p h o n o l o g y takes all levels to b e
n o n c y c l i c . A third p e r s p e c t i v e is that of H a l l e and K e n s t o w i c z ( 1 9 9 1 ) , w h o p r o -
p o s e that it is an arbitrary p r o p e r t y of individual affixes w h e t h e r they trigger a
p a s s by cyclic p h o n o l o g i c a l rules. All of these v i e w s are possible within a frame-
w o r k of the p h o n o l o g y - m o r p h o l o g y interface in w h i c h the t i m i n g a n d a m o u n t of
access b y p h o n o l o g y to m o r p h o l o g y is negotiable. L i n g u i s t s have differed, given
this flexibility, as to w h e t h e r they take cyclicity to b e the m a r k e d or the un-
m a r k e d case.
Unfortunately, there is not space here to evaluate all the data s p a w n i n g these
104 Sharon Inkelas

diverse v i e w s . W h a t I h a v e a t t e m p t e d to d o is rather different. A s s u m i n g , as ar-

g u e d c o m p e l l i n g l y b y H a r g u s (this v o l u m e ) , the n e e d for cyclic rule application
in at least s o m e cases, I h a v e p r o p o s e d a restrictive m o d e l of the p h o n o l o g y -
m o r p h o l o g y interface w h i c h derives cyclicity as the u n m a r k e d option.

5. CONCLUSION

N o t every r e a d e r m a y agree that this p a p e r h a s simplified the theory of w h e n

p h o n o l o g i c a l rules m a y apply cyclically in the lexicon and w h e n they m a y not.
A l t h o u g h I h a v e a t t e m p t e d to s t r e a m l i n e the predictions a b o u t cyclicity triggered
by affixation, I h a v e i n t r o d u c e d n e w c o m p l i c a t i o n s r e g a r d i n g stem cycles and
w o r d cycles. T h e d a t a are still quite c o m p l e x a n d not perfectly understood, and
the p r o p o s a l s m a d e h e r e m a y well turn out to b e i n a d e q u a t e . B u t w h a t I h o p e will
prevail is the basic a p p r o a c h of relating cyclicity not to arbitrary diacritic proper-
ties of affixes or levels, but rather to i n d e p e n d e n t l y n e e d e d representations, w h o s e
properties w e u n d e r s t a n d .

ACKNOWLEDGMENTS

During the preparation of this paper I have benefitted from discussions with Sharon
Hargus, Larry Hyman, Richard Sproat, Draga Zee, an anonymous reviewer, and partici-
pants in the Lexical Phonology Workshop at the University of Washington.

NOTES

For arguments against treating the clitic group as a distinct p-constituent, see Buckley
1

(1991), Inkelas (1989), Kanerva (1990), Zee (1988), and Zee and Inkelas (1991). Additions
to this hierarchy have been proposed as well: Condoravdi (1990) has demonstrated that
Modern Greek requires a constituent (the "minimal phrase") between the phonological
phrase and the phonological word; Kanerva (1990) argues from Chichewa for a focal
phrase between the intonational phrase and the phonological phrase.
This move may also be seen as a meeting of prosodic hierarchy theory with Sproat's
2

(1985) proposal that phonological structure and morphological structure are distinct, dif-
fering in the geometry of their bracketing.
In (5) and subsequent examples, angle brackets indicate m-constituency, and square
3

brackets indicate p-constituency.

It may be necessary to relax this convention to permit access one constituent in, to
4
Deriving Cyclicity 105

accommodate the apparent need for juncture rules. Another use to which access to internal
structure has been put involves "strict cycle" rules, which putatively apply only across
(internal) morpheme boundaries (Kiparsky, 1982; Mascaro, 1976); but see Kiparsky (this
volume) for a reanalysis not requiring morphological boundary information.
5
Notably excluded from this set are the bracketing paradoxes discussed by, among oth-
ers, Sproat (1985). Certain apparent paradoxes can be handled elegantly in terms of p-
structure; for example, Nespor and Vogel (1986) and Booij and Rubach (1987) have ana-
lyzed certain affixes that do not behave as though they are phonologically part of the word
as exceptionally constituting phonological words on their own. However, I do not at present
see a way to extend this approach to all of the cases discussed by Sproat (see also Harris,
1989), in particular to cases in which a level 2 affix appears to be inside a level 1 affix.
6
It also fails to apply to prefix-final consonants, which Nespor and Vogel account for by
assigning phonological word status to the prefixes in question (see previous footnote).
7
Of course, not all invisibility effects result from the application of a general phono-
logical rule such as that in (17). English, among other languages (e.g., Japanese), possesses
suffixes which are inherently invisible. Especially in a theory which restricts access by
phonological rules to morphological structure, these phenomena pose a challenge; I return
to it in the next section.
8
A problem for postulating a bimoraic minimum in Amele is the large number of mono-
moraic words [e.g., be 'neck (front)', co 'lips', du 'neck (back)', su 'breast (woman's)',
ca 'add', jo 'house']. However, Ito and Hankamer (1989) have argued for Turkish that all
the forms which appear to violate minimal size conditions in that language are underived;
Ito (1990) has made a similar observation for Japanese. If minimality conditions are en-
forced only on (morphologically) complex words, then the Amele examples just cited
would not constitute counterexamples to a bimoraic minimal word condition. Unfortu-
nately, the derived environment hypothesis is difficult to falsify. For example, Amele has a
process of possessive formation which has a zero allomorph in the third person. Whether
we consider the monomoraic possessive co-0 'his/her lips' to be derived or underived
determines the success or failure of a derived environment minimal size condition in the
language. I leave the issue open here, since there is another option: if necessary, we could
always claim that minimal size constraints block phonological rules (such as invisibility
assignment), but not morphological rules. Similar claims about other phonological well-
formedness constraints are found elsewhere in constraint-based phonological theory (Hy-
man, 1991;Paradis, 1988).
9
In this the lexical PCF algorithm conforms to a precedent already set by a number of
postlexical phrasing algorithms in the literature. For example, phrasing algorithms devel-
oped for Greek (Condoravdi, 1990), Hausa (Inkelas, 1988), Korean (Cho, 1990), and
Shanghai (Selkirk and Shen, 1990) possess explicit elsewhere clauses invoked whenever
no more specific source is available to phrase the relevant material.
1 am assuming a general morpheme-based theory of morphology a la Lieber (1980)
1 0

or Kiparsky (1983), in which all affixes, as well as stems and roots, are lexically listed
items as opposed to rules (Anderson, 1986).
Proposals for a notion of phonological subcategorization exist in the literature, but
1 1

none involves the particular kind of phonological constituency at issue here. For example,
Klavans's (1985) phonological subcategorization frames refer to syntactic constituents on
which clitics phonologically lean; Booij and Lieber's prosodic subcategorization frames
106 Sharon Inkelas

(this volume) express the sensitivity of particular affixes to certain metrical properties of
the base. Sproat's (1985) proposal is the closest relative. Separating out issues of linear
precedence from issues of dominance, Sproat proposes to encode the former in the phono-
logical entry and the latter in the morphological entry of an affix. Although Sproat did not
invoke our notion of p-structure per se, his insight of separate phonological and morpho-
logical entries is also central to the proposal made here.
1 2
If roots are inserted as the first step in a derivation, and if morphological sensitivity is
always inward, then root sensitivity to phonological properties of adjacent morphemes
would be ruled out on independent grounds, except perhaps in root-root compounding, as
occurs, for example, in Japanese. Although the compounding process itself imposes pho-
nological properties in that language, I am not aware of any root-specific allomorphy based
on the phonological characteristics of the sister root there, or elsewhere.
1 3
The morphological process not discussed here is compounding; Inkelas (1989) ac-
counts for the two types of compounds mentioned earlier by positing two compounding
rules. One (Mcompounding) refers only to m-structure, and its output feeds right into PCF.
Such compounds correspond to a single p-constituent. The other compounding rule (Pcom-
pounding) refers both to m- and to p-structure, imposing p-structure on its output such that
the two elements of the compound correspond to distinct p-constituents. A more specific
source of p-structure for the compound, Pcompounding overrides PCF exactly as affix-
ation does. Both kinds of compounds, however, are associated with a cycle of phonological
rules, either by virtue of triggering PCF by default, or by virtue of construction-specific
p-structure assignment.
For arguments in favor of such a view, see Inkelas (1989). Evidence that it is necessary
1 4

comes from the ability of bound roots, which lack p-structure, to combine with affixes,
which possess p-frames. The p-frame of the affix interprets the root as its p-sister.
Note that the suffixes -no, -tor}, -se, and -ma always impose length on the preceding
1 5

vowel; the penultimate long vowels of words ending in these suffixes is thus a local effect
and can be overlooked for present purposes.
In possessing a cyclic, structure-preserving stress rule, Carib thus patterns with Diyari
1 6

and Warlpiri, as analyzed in Poser (1989), and with Greek, as analyzed in Steriade (1988).
A reviewer aptly points out that correlation between vowel quantity and stress is a
1 7

property of iambic, rather than trochaic, systems (Hayes, 1987, 1991). To the extent that
Carib violates this generalization, the analysis is suspect. However, in defense of the tro-
chaic account proposed here, I note that vowel length is noncontrastive in Carib, assigned
by a late rule that applies after stress foot construction. Moreover, note that vowel length-
ening affects only the first two stress feet. On the assumption that words are exhaustively
parsed by the footing algorithm, then vowel lengthening does not target all stressed syl-
lables. Finally, although vowel lengthening occurs only in open syllables, suggesting that
coda consonants contribute to weight, closed syllables do not interfere with the assignment
of binary feet (moxkd:ro 'they' vs. tuxkusi 'arrow'; ma?md: takd.ra 'species offish' vs.
oxkoto:poti'the cutting of you into pieces'). The conclusion is that we are not dealing with
prototypical quantity-sensitive feet.
In Inkelas (1989) I instead assumed that odd-numbered stems possess an underlyingly
1 8

long vowel and that the extrametricality rule only targeted an initial mora—thus failing to
render the entire first syllable of such forms invisible. But this analysis is problematic in
that vowel length is not otherwise phonemic in the language, and admitting it lexically
Deriving Cyclicity 107

would violate structure preservation (Kiparsky, 1982). By contrast, stress feet are clearly
needed in the lexical phonology. Admitting them in underlying representation is only a
small extension of existing apparatus. For the use of underlying feet in the literature to
capture exceptional stress patterns, see, e.g., Hayes (1981), Hammond (1989), and many
others.
1 9
The syllable canon in Carib is CV(C), such that removing the first CV from a word
will eliminate the stress-bearing capacity of the initial syllable.
W h a t still remains to be explained is why "weak" suffixes, such as those illustrated
2 0

in the examples involving disyllabic stems, do not display the sensitivity of "strong" suf-
fixes to the initial consonant-vowel status of the stem. One possibility, explored in Inkelas
(1989), is to assign to all "weak" suffixes the property of imposing invisibility on initial
consonants. Though it may seem complex at first glance, this account has the advantage of
making the more common, unmarked suffixes duplicate the effects of PCF, which is the
case in, for example, English.
2 1
At worst, it would incorrectly prohibit the representation of clitics, which subcatego-
rize for word-level or higher constituents in the p-hierarchy (Inkelas, 1989; Zee and Inkelas,
1991).

REFERENCES

Anderson, S. R. (1986). Disjunctive ordering in inflectional morphology. Natural Lan-

guage and Linguistic Theory 4, 1-32.
Anderson, S. R. (1988). Inflection. In Theoretical Morphology: Approaches in Modern
Linguistics (M. Hammond and M. Noonan, eds.), pp. 2 3 - 4 3 . Academic, New York.
Bickmore, L. (1990). Branching nodes and prosodic categories. In The Phonology-Syntax
Connection (S. Inkelas and D. Zee, eds.), pp. 1-17. CSLI Publications and University
of Chicago Press, Chicago.
Booij, G. (1985). Coordination reduction in complex words: A case for prosodic pho-
nology. In Advances in Non-Linear Phonology (H. van der Hulst and N. Smith, eds.),
pp. 143-160. Dordrecht: Foris.
Booij, G., and Rubach, J. (1984). Morphological and prosodic domains in lexical pho-
nology. Phonology Yearbook 1, 1-27.
Booij, G., and Rubach, J. (1987). Postcyclic versus postlexical rules in lexical phonology.
Linguistic Inquiry 18, 1 - 4 4 .
Borowsky, T. (1986). Topics in the Lexical Phonology of English. Doctoral dissertation,
University of Massachusetts, Amherst.
Brame, M. (1974). The cycle in phonology: Stress in Palestinian, Maltese, and Spanish.
Linguistic Inquiry 5, 3 9 - 4 0 .
Buckley, E. (1991). Second-position Clitics in Alsea. Unpublished manuscript, University
of California, Berkeley.
Buckley, E. (1992). Theoretical Aspects ofKashaya Phonology and Morphology. Doctoral
dissertation, University of California, Berkeley.
Cho, Y. Y. (1990). Syntax and phrasing in Korean. In The Phonology-Syntax Connection
108 Sharon Inkelas

(S. Inkelas and D. Zee, eds.), pp. 4 7 - 6 1 . CSLI Publications and University of Chicago
Press, Chicago.
Cohn, A. (1989). Stress in Indonesian and bracketing paradoxes. Natural Language and
Linguistic Theory 7, 167 - 216.
Condoravdi, C. (1990). Sandhi rules of Greek and prosodic theory. In The Phonology-
Syntax Connection (S. Inkelas and D. Zee, eds.), pp. 6 3 - 8 3 . CSLI Publications and
University of Chicago Press, Chicago.
Halle, M., and Kenstowicz, M. (1991). The free element condition and cyclic versus non-
cyclic stress. Linguistic Inquiry 2 2 , 4 5 7 - 5 0 1 .
Halle, M., and Mohanan, K. P. (1985). Segmental phonology of Modern English. Linguis-
tic Inquiry 16, 57 - 1 1 6 .
Hammond, M. (1989). Lexical stresses in Macedonian and Polish. Phonology 6, 1 9 - 3 8 .
Hargus, S. (1988). The Lexical Phonology ofSekani. Garland, New York.
Harris, J. (1983). Syllable Structure and Stress in Spanish. MIT Press, Cambridge, Mass.
Harris, J. (1989). The stress erasure convention and cliticization in Spanish. Linguistic
Inquiry 20, 3 3 9 - 3 6 3 .
Hayes, B. (1981). A Metrical Theory of Stress Rules. Doctoral dissertation, Massachusetts
Institute of Technology, Cambridge.
Hayes, B. (1987). A revised parametric metrical theory. Proceedings of the North Eastern
Linguistics Society 17, 2 7 4 - 2 8 9 .
Hayes, B. (1989). The prosodic hierarchy and meter. In Rhythm and Meter (P. Kiparsky
and G. Youmans, eds.), pp. 2 0 1 - 2 6 0 . Academic Press, Orlando.
Hayes, B. (1991). Metrical Stress Theory: Principles and Case Studies. Unpublished
manuscript, University of California, Los Angeles.
Hoff, B. (1968). The Carib Language. Nijhoff, The Hague.
Hyman, L. (1991). Imbrication in Cibemba. Unpublished manuscript, University of Cali-
fornia, Berkeley.
Inkelas, S. (1988). Prosodic effects on syntax: Hausa 'fa'. Proceedings of the West Coast
Conference on Formal Linguistics 7, 3 7 5 - 3 8 9 .
Inkelas, S. (1989). Prosodic Constituency in the Lexicon. Doctoral dissertation, Stanford
University, Stanford, Calif.
Inkelas, S., and Zee, D. (1988). Serbo-Croatian pitch accent: The interaction of tone, stress
and intonation. Language 64, 2 2 7 - 2 4 8 .
Ito, J. (1986). Syllable Theory in Prosodic Phonology. Doctoral dissertation, University of
Massachusetts, Amherst.
Ito, J. (1990). Prosodic minimality in Japanese. In CLS 26-11: Papers from the parasession
on the syllable in phonetics and phonology (K. Deaton, M. Noske, and M. Ziolkowski,
eds.), pp. 213-239.
Ito, J., and Hankamer, J. (1989). Notes on monosyllabism in Turkish. In Phonology at
Santa Cruz 1 (J. Ito and J. Runner, eds.), pp. 6 1 - 7 0 . Syntax Research Center, Univer-
sity of California, Santa Cruz.
Kaisse, E. (1985). Connected Speech. Academic Press, New York.
Kanerva, J. (1990). Focusing on phonological phrases in Chichewa. In The Syntax-Pho-
nology Connection (S. Inkelas and D. Zee, eds.), pp. 1 4 5 - 1 6 1 . CSLI Publications and
University of Chicago Press, Chicago.
Kiparsky, P. (1982). Lexical morphology and phonology. In Linguistics in the Morning
Calm (I.-S. Yang, ed.), pp. 3 - 9 1 . Hanshin Publishing Co., Seoul.
Deriving Cyclicity 109

Kiparsky, P. (1983). Word formation in the lexicon. In Proceedings of the 1982 Mid-
America Linguistics Conference (F. Ingemann, ed.), pp. 3 - 2 9 . University of Kansas,
Lawrence.
Kiparsky, P. (1985). Some consequences of lexical phonology. Phonology Yearbook 2,
85-138.
Klavans, J. (1985). The independence of syntax and phonology in cliticization. Language
61, 8 5 - 1 2 0 .
Lichtenberk, F. (1983). A Grammar of Manam. University of Hawaii Press, Honolulu.
Lieber, R. (1980). On the Organization of the Lexicon. Doctoral dissertation, Massachu-
setts Institute of Technology, Cambridge.
Mascaro, J. (1976). Catalan Phonology and the Phonological Cycle. Doctoral dissertation,
Massachusetts Institute of Technology. Distributed by the Indiana University Linguis-
tics Club.
McCarthy, J., and Prince, A. (1986). Prosodic Morphology. Unpublished manuscript, Uni-
versity of Massachusetts, Amherst, and Brandeis University, Waltham, Mass.
McCarthy, J., and Prince, A. (1990). Foot and word in prosodic morphology: The Arabic
broken plural. Natural Language and Linguistic Theory 8, 2 0 9 - 2 8 4 .
Mohanan, K. P. (1982). Lexical Phonology. Doctoral dissertation, Massachusetts Institute
of Technology, Cambridge.
Mohanan, K. P. (1986). The Theory of Lexical Phonology. Reidel, Dordrecht.
Nespor, M. (1984). ' T h e phonological word in Italian." In Advances in Non-linear Pho-
nology (H. van der Hulst and N. Smith, eds.), pp. 193-204. Foris, Dordrecht.
Nespor, M., and Vogel, I. (1982). Prosodic domains of external sandhi rules. In The Struc-
ture of Phonological Representations, part 1 (H. van der Hulst and N. Smith, eds.),
pp. 2 2 5 - 2 5 5 . Foris, Dordrecht.
Nespor, M., and Vogel, I. (1986). Prosodic Phonology. Foris, Dordrecht.
Paradis, C. (1988). On constraints and repair strategies. The Linguistic Review 6, 7 1 - 9 7 .
Pesetsky, D. (1979). Russian Morphology and Lexical Theory. Unpublished manuscript,
Massachusetts Institute of Technology, Cambridge.
Poser, W. (1984). The Phonetics and Phonology of Tone and Intonation in Japanese. Doc-
toral dissertation, Massachusetts Institute of Technology, Cambridge.
Poser, W. (1989). The metrical foot in Diyari. Phonology 6, 117 - 1 4 8 .
Pulleyblank, D. (1986). Tone in Lexical Phonology. Reidel, Dordrecht.
Roberts, J. (1987). Amele. Croom Helm, London.
Selkirk, E. (1978). On prosodic structure and its relation to syntactic structure. In Nordic
Prosody II (T. Fretheim, ed.), pp. 11 - 4 0 . TAPIR, Trondheim.
Selkirk, E. (1980). Prosodic domains in phonology: Sanskrit revisited. In Juncture (M.
Aronoff, ed.), pp. 1 0 7 - 1 2 9 . Anma Libri, Saratoga, Calif.
Selkirk, E. (1982). The Syntax of Words. MIT Press, Cambridge, Mass.
Selkirk, E. (1986). On derived domains in sentence phonology. Phonology Yearbook 3,
371-405.
Selkirk, E., and Shen, T. (1990). Prosodic Domains in Shanghai Chinese. In The Pho-
nology-Syntax Connection (S. Inkelas and D. Zee, eds.), pp. 3 1 3 - 3 3 7 . CSLI Publi-
cations and University of Chicago Press, Chicago.
Sproat, R. (1985). On Deriving The Lexicon. Doctoral dissertation, MIT, Cambridge, Mass.
Sproat, R. (1986). Malayalam Compounding: a Non-Stratum Ordered Account. Proceed-
ings of the West Coast Conference on Formal Linguistics 5, 2 6 8 - 2 8 8 .
110 Sharon Inkelas

Steriade, D. (1988). Greek accent: A case for preserving structure. Linguistic Inquiry 19,
271-314.
Vogel, I. (1989). The clitic group in prosodic phonology. To appear in Grammar in Prog-
ress: GLOW Studies for Henk van Riemsdijk. (J. Mascaro and M. Nespor, eds.). Foris,
Dordrecht.
Zee, D. (1988). Sonority Constraints on Prosodic Structure. Doctoral dissertation, Stan-
ford University, Stanford, Calif.
Zee, D., and Inkelas, S. (1990). Prosodically constrained syntax. In The Phonology-Syntax
Connection (S. Inkelas and D. Zee, eds.), CSLI Publications and University of Chi-
cago Press, Chicago.
Zee, D., and Inkelas, S. (1991). The place of Clitics in the Prosodic Hierarchy. Proceedings
of the West Coast Conference on Formal Linguistics 10, 5 0 5 - 5 2 0 .
INTERACTION BETWEEN MODULES IN LEXICAL
PHONOLOGY

DAVID ODDEN
Department of Linguistics
Ohio State University
Columbus, Ohio 43210

1. I N T R O D U C T I O N

T h i s article a d d r e s s e s q u e s t i o n s in the t h e o r y of c o m p o n e n t interaction in lexi-

cal p h o n o l o g y ( L P ) , q u e s t i o n s w h i c h arise in g i v i n g a formal interpretation to the
standard g r a p h i c m e t a p h o r in (1), a d a p t e d m i n i m a l l y from K i p a r s k y ( 1 9 8 2 ) .

(1) ,
LEXICON

u n d e r i v e d lexical
items

level 1 m o r p h o l o g y level 1 p h o n o l o g y

level 2 m o r p h o l o g y level 2 p h o n o l o g y

level n m o r p h o l o g y level n p h o n o l o g y

syntax postlexical p h o n o l o g y

ill
Phonetics and Phonology, Volume 4 Copyright © 1993 by Academic Press, Inc.
Studies in Lexical Phonology All rights of reproduction in any form reserved.
112 David Odden

T h i s display e m b o d i e s a n u m b e r of h y p o t h e s e s , the m o s t i m p o r t a n t b e i n g
spelled out in (2).

(2) a. T h e r e is a c o n s t r u c t " l e v e l " c o m m o n to p h o n o l o g y and m o r p h o l o g y .

b. T h e r e are significant formal differences b e t w e e n lexical and postlexical
phonology.
c. Levels in p h o n o l o g y are the s a m e as levels in m o r p h o l o g y .
d. C y c l i c application of p h o n o l o g i c a l rules derives from the interaction
b e t w e e n lexical p h o n o l o g y a n d m o r p h o l o g y as characterized above.
e. L e x i c a l p h o n o l o g y h a s n o access to the output of the syntax.
f. L e x i c a l p h o n o l o g y a n d m o r p h o l o g y interact so that m o r p h o l o g y has
access to p h o n o l o g i c a l p r o p e r t i e s derived by applying p h o n o l o g i c a l rules
on s o m e earlier level.

C l a i m s (2a) and (2b) are the level-ordering c l a i m s , and are simply a s s u m e d here
to b e true. T h e related claim (2c) that p h o n o l o g i c a l levels are the s a m e as m o r p h o -
logical levels has less support, in light of certain b r a c k e t i n g p a r a d o x e s . I also
maintain w i t h o u t c o m m e n t (or c o m m i t m e n t ) the standard a s s u m p t i o n of lexical
p h o n o l o g y that p h o n o l o g i c a l levels are i s o m o r p h i c with m o r p h o l o g i c a l levels,
since the identity of m o r p h o l o g i c a l a n d p h o n o l o g i c a l levels is entirely o r t h o g o n a l
to the question of interaction b e t w e e n p h o n o l o g y and m o r p h o l o g y . C l a i m (2d),
that cyclic b e h a v i o r C A N b e derived from a p h o n o l o g y - m o r p h o l o g y interaction,
is a purely theory-internal c l a i m . A s w e will see, other m o d e l s can derive cyclic
b e h a v i o r w i t h o u t this interaction.

1.1. N o n i n t e r a c t i v e L e x i c a l P h o n o l o g y

T h e p u r p o s e of this article is the presentation of a theory of p h o n o l o g y and

m o r p h o l o g y w h i c h c a p t u r e s the relevant generalizations r e g a r d i n g level o r d e r i n g
and cyclicity, w i t h o u t the L P theory of c o m p o n e n t interaction. Therefore, c l a i m s
(2e) and (2f) are the focus; they are the m a i n h y p o t h e s e s about the organization
of c o m p o n e n t s into g r a m m a r s a n d about the information available to e a c h c o m -
p o n e n t . I c o n s i d e r the alternative organization of c o m p o n e n t s in (3), a theory to
b e referred to as noninteractive lexical p h o n o l o g y ( N L P ) .
Interaction in Lexical Phonology 113

(3) SYNTAX

MORPHOLOGY

L e x i c o n of roots

level 1

level 2

level n

PHONOLOGY

Lexical Phonology

level 1

level 2

level n

i
Postlexical P h o n o l o g y

In effect, the p r o p o s a l is to retain the m o r e traditional o r d e r i n g of c o m p o n e n t s but

also a d o p t the notion of level and the l e x i c a l - p o s t l e x i c a l split. H a l l e and V e r g n a u d
(1987) similarly a d o p t this o r d e r i n g of m o r p h o l o g y before p h o n o l o g y .
In calling into q u e s t i o n certain a s s u m p t i o n s of LP, it is i m p o r t a n t to distinguish
the c l a i m s u n d e r investigation from interesting but tangential issues. T h e c l a i m
w h i c h is m o s t i m p o r t a n t to this theory, the o n e w h i c h m o s t o b v i o u s l y differenti-
114 David Odden

ates the m o d e l s , is the relative o r d e r i n g a n d noninteraction of c o m p o n e n t s . To

facilitate c o m p a r i s o n with LP, the level-ordering c l a i m s (2a,b,c) are retained from
LP, t h o u g h t h e formal m e c h a n i s m s w h i c h g e n e r a t e these types of b e h a v i o r are not
the s a m e in the t w o theories: it m a y well turn out that the distinction b e t w e e n
lexical a n d postlexical p h o n o l o g y is s p u r i o u s , or that " l e v e l " is a purely p h o n o -
logical notion, b u t these q u e s t i o n s are n o t investigated here.
N L P a s s u m e s , as d o e s LP, that there is a primitive notion of level, L to L . T h ex n

putative identity of m o r p h o l o g i c a l a n d p h o n o l o g i c a l levels w o u l d b e in N L P a

c o n s e q u e n c e of a s s u m p t i o n s a b o u t h o w levels in p h o n o l o g y are defined. A s in LP,
e a c h m o r p h o l o g i c a l rule is e n c o d e d for a specification of the level w h e r e it applies.
W o r d construction starts b y selecting a root, a n d m o r p h o l o g i c a l rules apply to this
structure, c o n c a t e n a t i n g a p p r o p r i a t e material with it. In N L P each m o r p h o l o g i c a l
rule p r o v i d e s a labeled b r a c k e t i n g as well as the s e g m e n t a l content of its affix, so
rules take the form of t h o s e in (4).

(4) [ Y]->[ [ Y]W]

X Z X
[ Y]->[ W[ Y]]
X Z X

T h e labels attached to these structures indicate the level o n w h i c h the operation

applies; it is this labeled b r a c k e t i n g w h i c h p h o n o l o g y reacts to, in applying rules
to m o r p h o l o g i c a l l y defined d o m a i n s .
In (5) w e see t h e m o r p h o l o g i c a l derivation of the M a l t e s e A r a b i c form hatfit-
kums ' s h e didn't snatch y o u ' w h i c h h a s t w o Level 1 m o r p h e m e s , the root hataf
a n d the subject m a r k e r it, a n d t w o Level 2 m o r p h e m e s , the object suffix hum and
the n e g a t i v e s.

(5) LX INSERTION OF ROOT

hataf

SUFFIXATION OF SUBJECT MARKER

hataf it

SUFFIXATION OF OBJECT MARKER

hataf it hum
Interaction in Lexical Phonology 115

SUFFIXATION OF NEGATIVE

' s h e d i d n ' t snatch y o u '

T h e only difference b e t w e e n this a n d standard lexical p h o n o l o g y is t h e inclusion

of labels o n the b r a c k e t s w h i c h indicate distinctions of level: Inkelas ( 1 9 8 9 ) , w h o
a s s u m e s t h e o r t h o d o x t h e o r y of LP, adopts a m o d e l quite similar to this, employ-
ing t h e labels a , p , 7 .
In the p h o n o l o g i c a l c o m p o n e n t , rules are e n c o d e d for d o m a i n of application,
specifically t h e lowest- a n d h i g h e s t - n u m b e r e d levels w h e r e the rule m a y apply.
T h i s is t h e s a m e as specifying, for instance, that a rule is " i n " Level 1 p h o n o l o g y .
In N L P , cyclicity is achieved in a m a n n e r a n a l o g o u s to that of LP. A s is well
k n o w n , cyclicity in L P c a n b e e x p r e s s e d in t w o w a y s . First, if a rule resides in t w o
c o n s e c u t i v e levels, cyclicity arises w h e n the rule applies at level L _ a n d at level
n x

L (interstratal cyclicity). S e c o n d , if L is d e e m e d cyclic, cyclicity result w h e n

n n

s o m e rule applies to e a c h substring resulting from m o r p h o l o g i c a l o p e r a t i o n s

within L (stratum-internal cyclicity).
n

N o w c o n s i d e r h o w N L P e x p r e s s e s cyclicity. I c o n s i d e r first interstratal cyclicity,

using t h e M a l t e s e A r a b i c form hatfitkums as an illustration. T h e highest constitu-
ent d o m i n a t e d b y L b e c o m e s the initial input to the p h o n o l o g y , a n d p h o n o l o g i c a l
x

rules e n c o d e d for application at L apply to this string. T h u s the b o x e d constituent

{

in the first step of (6) is the d o m a i n w h e r e Level 1 p h o n o l o g i c a l rules apply, b e -

g i n n i n g in this c a s e with stress a s s i g n m e n t . After the last rule defined at L applies
x

( A p o c o p e ) , the h i g h e s t constituent d o m i n a t e d b y L b e c o m e s the string subject to

2

p h o n o l o g i c a l rules, a n d rules e n c o d e d for L apply to this string. T h i s c o n t i n u e s

2

to the last lexical level a n d into postlexical p h o n o l o g y .

(6) L 2 STRESS

L, phonology

hdtaf it hum s
116 David Odden

APOCOPE

STRESS

N L P h a n d l e s stratum-internal cyclicity b y a p p l y i n g p h o n o l o g i c a l rules of level

L to the lowest u n p r o c e s s e d constituent d o m i n a t e d by L . S u p p o s e that L in
n n 2

M a l t e s e A r a b i c h a d b e e n cyclic. T h e n in the derivation (6), the lowest constituent

d o m i n a t e d b y L w o u l d h a v e defined the first d o m a i n of rule application; then the
2

next lowest constituent d o m i n a t e d b y L w o u l d define the next d o m a i n of rule

2

application, and so on. In this way, w e k e e p cyclicity a n d level ordering, w i t h o u t

interaction b e t w e e n p h o n o l o g y and m o r p h o l o g y : the issue of cyclicity in n o w a y
distinguishes N L P from LP.

1.2. T h e I n t e r p r e t a t i o n of B o x e s

T h e literature of L P has been virtually silent r e g a r d i n g the formal interpretation

of displays like (1). C o n s i d e r the fact that in (1), the b o x e s called "level 1 p h o -
n o l o g y " t h r o u g h "level n p h o n o l o g y , " and the b o x e s called "level 1 m o r p h o l -
o g y " through "level n m o r p h o l o g y " are c o n t a i n e d in a larger b o x called " l e x i -
c o n . " A c o n c e i v a b l e interpretation of such structures is that things sharing a b o x
have similar properties w h i c h things outside the b o x d o not have. B u t it is quite
unclear w h a t the shared formal properties of m o r p h o l o g y and lexical p h o n o l o g y
are, e x c e p t that in LP, these c o m p o n e n t s define the lexicon. In fact, taking displays
like (1) to b e a claim about similarity in formal properties, w e w o u l d c o n c l u d e
Interaction in Lexical Phonology 117

that lexical p h o n o l o g y a n d m o r p h o l o g y are m o r e similar than lexical and postlex-

ical p h o n o l o g y . W e m i g h t even c o n c l u d e that lexical and postlexical p h o n o l o g y
h a v e n o shared p r o p e r t i e s , since they share n o b o x . T h i s is clearly absurd, and
other a r r a n g e m e n t s of the b o x e s h a v e b e e n p r o p o s e d , for instance Kaisse and
S h a w ( 1 9 8 5 ) , w h e r e b o x e s overlap, or M o h a n a n ( 1 9 8 6 ) , w h e r e p h o n o l o g y and
m o r p h o l o g y d o not even share a b o x . If display (1) h a s a m e a n i n g , it c a n n o t b e a
c l a i m about similarity of c o m p o n e n t s .
M o r e plausibly, these structures c o u l d b e taken as g r a p h i c m e t a p h o r s for c l a i m s
a b o u t the o r d e r i n g of p r o c e s s e s ; t h u s , lexical p h o n o l o g y p r e c e d e s syntax, and
postlexical p h o n o l o g y follows syntax. S i n c e w e have n o substantive e v i d e n c e for
the real-time interpretation of c o m p o n e n t o r d e r i n g , a c l a i m about o r d e r i n g r e d u c e s
to a c l a i m about the information available to a given m o d u l e . If m o d u l e M "fol-
l o w s " m o d u l e L and " p r e c e d e s " m o d u l e N, then M h a s access to the results of
o p e r a t i o n s defined in L, but not to the results of o p e r a t i o n s defined in N. In this
sense, w h e n w e say that lexical p h o n o l o g y p r e c e d e s syntax, w e m e a n that infor-
m a t i o n p r o v i d e d by syntax is not available to lexical p h o n o l o g y . T h i s h a s b e e n
taken to entail that n o rule of lexical p h o n o l o g y h a s access to p r o p e r t i e s of other
w o r d s in the sentence, or to any fact a b o u t the syntactic structure of the sentence
w h i c h the w o r d a p p e a r s in. S u p p o s i n g that w e h a d criteria for d e c i d i n g w h e t h e r a
rule is lexical, then this s e e m s to m a k e the prediction that certain types of lan-
g u a g e s will not b e found.
T h e issues separating N L P and L P are the following t w o . First, N L P c l a i m s that
rules of lexical p h o n o l o g y h a v e a c c e s s to information from syntax. S e c o n d , N L P
disallows m o r p h o l o g y a c c e s s to information c o m i n g from p h o n o l o g i c a l rules.
T h e s e t w o issues are s e p a r a b l e — o n e c o u l d h a v e an interactive m o d e l of p h o -
n o l o g y and m o r p h o l o g y , a n d r e o r d e r syntax relative to lexical p h o n o l o g y . O r o n e
c o u l d retain the o r d e r i n g of p h o n o l o g y a n d m o r p h o l o g y relative to syntax b u t
reject the interactive aspect of LP. I thus c o n s i d e r these c l a i m s separately.
T w o c o n c l u s i o n s r e g a r d i n g w h a t l a n g u a g e s d o will e m e r g e in the c o u r s e of this
paper. First, w e will see that s y s t e m s exist with exactly the p r o p e r t i e s w h i c h L P
predicts should not exist, n a m e l y lexical rules a c c e s s i n g the output of syntax. S e c -
ond, it will b e a r g u e d that there are n o c o m p e l l i n g cases of rules of m o r p h o l o g y
a p p l y i n g after rules of p h o n o l o g y . T a k e n together, this should a r g u e for noninter-
active m o d e l (3) over m o d e l (1).
H o w e v e r , an e x t e n s i o n to LP, specifically H a y e s ' s theory of p r e c o m p i l e d p h o -
nology, can b e called on to h a n d l e a n y c o u n t e r e x a m p l e s w h e r e a p h o n o l o g i c a l rule
acts as t h o u g h it sees o u t s i d e its d o m a i n , such as a lexical rule seeing b e t w e e n
w o r d s , or for that matter, a Level 1 rule seeing material only available on Level 2.
T h i s e x t e n d e d version of L P ( E L P ) h a s access to all the information available to
the n o n i n t e r a c t i v e m o d e l , plus it a l l o w s for rules of m o r p h o l o g y w h i c h are sensi-
tive to derived p h o n o l o g i c a l information, a situation w h i c h is disallowed in the
noninteractive m o d e l .
118 David Odden

2. S Y N T A X A N D P H O N O L O G Y

T h e first p r o b l e m I look into briefly is the o r d e r i n g b e t w e e n lexical p h o n o l -

o g y and syntax. M o r e e x t e n s i v e d i s c u s s i o n of the d a t a and issues involved here
c a n b e found in O d d e n ( 1 9 9 0 a ) a n d O d d e n ( 1 9 8 7 ) , as well as H a y e s (1990). A
general description of K i m a t u u m b i p h o n o l o g y and m o r p h o l o g y can b e found in
Odden (1992).

2 . 1 . O r d e r i n g S a n d h i R u l e s in t h e L e x i c o n

T h e p r o b l e m I c o n s i d e r h e r e is that s o m e rules of K i m a t u u m b i p h o n o l o g y m u s t
b e lexical, b u t they also h a v e access to syntactic structure and p h o n o l o g i c a l p r o p -
erties of s u r r o u n d i n g w o r d s . T h e first rule is the phrase-level rule S h o r t e n i n g ,
w h i c h shortens a long vowel if it is the h e a d of a p h r a s e and is followed b y m a t e -
rial within its p h r a s e . This vowel length alternation is seen in (7).

(7) kjkoloombe 'cleaning shell'

k\l6lombe chadngu ' m y c l e a n i n g shell'
kjtuumbili 'monkey'
k\tumbili ywdaw\\le 'monkey who died'
naak\-tweet\ T t o o k it'
naaki-tweti k\koloombe T t o o k a c l e a n i n g shell'

S i n c e this rule involves m u l t i p l e w o r d s a n d syntactic structures, L P requires the

rule to b e a postlexical rule.
(8) SHORTENING

A (Y contains phonetic material)

T h e s e c o n d rule is G l i d e F o r m a t i o n ( G F ) , a strictly word-internal lexical rule.

T h a t rule desyllabifies a prevocalic h i g h vowel a n d c o m p e n s a t o r i l y l e n g t h e n s the
following vowel. E x a m p l e s of the v o w e l - g l i d e alternation a n d c o m p e n s a t o r y
l e n g t h e n i n g are seen in (9).

(9) /kj-kalaango/ kj-kdlaango 'frying pan'

CI. 7-frying p a n
/kj-yla/ ky-u^ld 'frog' (cf. kayld 'little frog')
CI. 7-frog
/j-kalaango/ l-kdlaango 'frying p a n s '
CI. 8-frying p a n
/j-yla/ y-uyld 'frogs'
CI. 8-frog
Interaction in Lexical Phonology 119

W e m u s t restrict G F so that a prevocalic long v o w e l d o e s not desyllabify, to ac-

c o u n t for the f o r m s in ( 1 0 ) : in the first e x a m p l e , the u n d e r l y i n g short v o w e l of the
locative prefix mu is l e n g t h e n e d before d i m o r a i c s t e m s by an i n d e p e n d e n t l y m o -
tivated rule. Vowel length w h i c h b l o c k s G F m a y d e r i v e from the c o m p e n s a t o r y
l e n g t h e n i n g effect of G F itself, as the form mw\\ute s h o w s , indicating that the rule
iterates from left to right.

(10) muu-ate 'in the h a n d s of b a n a n a s '

L o c . - h a n d of b a n a n a s
/mu-aanjy/ mwaanjy 'in the firewood'
Loc-fire wood
/my-j-ute/ mwij-ute ' y o u s h o u l d pull t h e m (CI. 9 ) '
2 p l . - t h e m (CI. 9)-pull
/ba-j-ute/ bayuute 'they pulled t h e m '
3pl.-them (CI. 9)-pull

T h r e e levels can b e m o t i v a t e d in K i m a t u u m b i . L e v e l 1 m o r p h o l o g y c o n s t r u c t s
the stem from the root, derivational affixes s u c h as the causative a n d benefactive,
a n d the stem-final tense inflection. Level 2 verbal m o r p h o l o g y includes addition
of object prefixes, tense-aspect prefixes, a n d subject prefixes. Level 2 n o m i n a l
m o r p h o l o g y is addition of the n o u n class prefixes. At L e v e l 3 , the locative n o m i n a l
prefixes ku-, pa-, a n d mu- are added, a n d in verbs the relative c l a u s e h e a d a g r e e -
m e n t prefixes such as pa- and cha- are a d d e d .
T h e G l i d e F o r m a t i o n rule applies cyclically, in particular interstratal-cyclically
b u t not s t r a t u m - i n t e r n a l - c y c l i c a l l y . T h i s is seen in the contrasting derivations of
mwjjute in (11a), w h i c h h a s the Level 2 prefixes mu a n d /, versus muyuyld in
( l i b ) , w h i c h h a s the level 3 prefix mu a n d the level 2 prefix /. T h e p r o b l e m is that
w e h a v e the s a m e b a s i c configuration of p h o n e m e s in b o t h c a s e s , b u t d e t e r m i n i n g
w h i c h of the v o w e l s b e c o m e s a glide requires k n o w i n g the level at w h i c h e a c h
m o r p h e m e is m a d e available to the p h o n o l o g y .

(11) a. [my-j-ute]
mu-j-ute Input to L 2

mwjjute Glide Formation

' y o u s h o u l d pull it'
b. [mu [i-yld]]
i-yld Input to L 2

yuyla Glide Formation

muyuyld Input to L 3

NA Glide Formation
in the f r o g s '

In the c a s e of mwiiute, the vowel s e q u e n c e u+j a p p e a r s entirely within Level 2,

so it is the first of the p r e v o c a l i c high v o w e l s w h i c h glides, g i v i n g the p h o n e t i c
120 David Odden

form. In the c a s e of muyyyld, the l o c a t i v e prefix my is only available at Level 3,

but / is available at Level 2, so / u n d e r g o e s G F since it is the only vowel e n c o u n -
tered at this stage of the derivation. T h e r e f o r e G F m u s t b e lexical.
N o w c o n s i d e r the o r d e r i n g of S h o r t e n i n g and GF. S h o r t e n i n g p r e c e d e s G F , as
s h o w n by the fact that S h o r t e n i n g d o e s not apply to the long vowel w h i c h arises
as a result of a p p l y i n g G F , at least at Level 2. T h i s is s h o w n by the f o r m s in (12).

(12) / m u - a k e \\l mwaake If ' y o u should not h u n t '

2 s S u b - h u n t not
(*mwake If)
/kj-ylachaangu/ kyyyla chadngu ' m y frog'
CI. 7-frog m y
i^kyyla chadngu)

T h e u n d e r l y i n g prefix plus stem vowel c o m b i n a t i o n u n d e r g o e s G F , w h i c h length-

e n s the stem vowel, but the derived l o n g vowel is not shortened in the p r e s e n c e of
a modifier. T h i s requires o r d e r i n g the lexical rule G F after the supposedly post-
lexical rule S h o r t e n i n g .
F u r t h e r data s h o w that S h o r t e n i n g is a Level 1 rule and actually follows G F on
that level. G F also applies at Level 1, the s t e m level, as w e see in (13), w h e r e the
short vowel of the reciprocal suffix -an- l e n g t h e n s as a result of desyllabifying the
suffix -/-.

(13) dk-a 'to net-hunt'

hunt-tense
dk-an-a ' t o n e t - h u n t e a c h other
hunt-recip.-tense
/ a k - j - a n - a / dk-y-aan-a ' t o net-hunt for e a c h o t h e r '
hunt-benefact.-recip.-tense

Interestingly, long vowels w h i c h arise b y a p p l y i n g G F at Level 1 d o u n d e r g o

S h o r t e n i n g , in contrast to long v o w e l s w h i c h derive by applying G F at Levels 2
a n d 3 . A s can b e seen in (14), G F is a p p l i c a b l e on t w o levels in u n d e r l y i n g ty-ak-
l~an-a ituumbili, n a m e l y on Level 1 to i+a, and on Level 2 to y+a. A s s h o w n by
the surface form twaakyana ituumbili, the long vowel derived by G F at Level 1
shortens, but the long vowel derived at Level 2 d o e s not.

(14) [tu-[ak-i-an-a]] ituumbili —>

we-hunt-benefact.-recip.-tense m o n k e y s
twaakyana ituumbili
' w e net-hunt m o n k e y s for e a c h o t h e r '

T h u s , S h o r t e n i n g is o r d e r e d after G F , and only applies at Level 1.

(15) ak-fan-a ituumbili Input to Level 1

akyaana ituumbili Glide Formation
Interaction in Lexical Phonology 121

akyana jtuumbili Shortening

tu-akyana jtuumbili Input to Level 2
twaakyana Ituumbili Glide Formation

F o r this derivation to b e p o s s i b l e , S h o r t e n i n g m u s t b e lexical, w h i c h m e a n s that it

is n e c e s s a r y to allow rules of the lexical c o m p o n e n t access to the output of syntax.
T h i s is i m p o s s i b l e in L P but is allowed in N L P .
O t h e r rules exhibit similar p r o p e r t i e s . O n e of these, Initial T o n e Insertion (ITI),
is also relevant in Section 3. T h i s rule, illustrated in (16), assigns an H t o n e to the
initial vowel of a lexical class of m o r p h e m e s , as long as the m o r p h e m e is p r e c e d e d
by a w o r d b e a r i n g n o stem H.

(16) kjndolo cha Mambdondo ' s w e e t p o t a t o of M a m b o o n d o '

kjwikilyo chd Mambdondo ' c o v e r of M a m b o o n d o '
mabigq ga-bili 'two beer brewing areas'
matanga gd-bili 'two cucumbers'
aatfbelekwd ku-Kjpooi ' h e w a s BORN in K i p o o i '
abelekjlxve k^,-K\pod\ ' h e w a s b o r n in K i p o o i '

T h e full list of m o r p h e m e s w h i c h u n d e r g o ITI is given in (17).

(17) a- d e m o n s t r a t i v e prefix
na 'with, and'
n(cheche) initial syllabic nasal of 'four'
n(tupu) initial syllabic nasal of 'is n o t '
ka- subord. verb prefix
maladu 'tomorrow'
kjtiwj 'how'
namanj 'what'
rial 'who'
gani 'which'
bulj 'how'
kill 'what'
gaaku 'what kind'
gu- et al. d e m o n s t r a t i v e prefix
wa- et al. associative prefix
gu- et al. n u m e r a l prefix
mu-, pa-, ku- locative prefixes

A n i m p o r t a n t condition on the t r i g g e r i n g e l e m e n t seen in (18) is that, w h i l e a

stem H in the p r e c e d i n g n o u n b l o c k s the rule, a prefixal H d o e s not.

(18) kj-wikilyo gdnj 'what type cover?'

kj-tumbili gani w h a t type m o n k e y ? '
4

k\-n 'ombe gdnf 'what type cows?'

122 David Odden

T h u s the stem H in ki-tuumbili b l o c k s a s s i g n m e n t of H to ganj, b u t the prefix H

in k(-ng'ombe d o e s not.
T h e fact that the rule only applies to a lexically specified set of m o r p h e m e s and
is sensitive to the s t e m - p r e f i x distinction a r g u e s that the rule should be lexical.
A n o r d e r i n g a r g u m e n t c i n c h e s this c o n c l u s i o n . T h e r e is a further restriction on
G l i d e F o r m a t i o n in K i m a t u u m b i , w h i c h is that an H - t o n e vowel c a n n o t u n d e r g o
G F before a long vowel, a l t h o u g h an L - t o n e vowel can. Word-internal e x a m p l e s
of this condition are seen in (19), w h e r e the prevocalic vowels / a n d y c a n n o t glide
before a long vowel b e c a u s e of this constraint.

(19) chatyoondite 'what w e peeled'

pan\adnd\\ke 'when I wrote'

N o w c o n s i d e r the d a t a in (20), with the prefix ky- before a long v o w e l .

(20) utijlf kuaanjy —> ut\l\ kwaanjy ' y o u s h o u l d run to the firewood'
util\kyaan]ii —> iit\l\ kyaanjy ' y o u r a n to the firewood'

A n H can b e assigned to ku- b y ITI, a n d this derived tone affects w h e t h e r G F can

apply before a long v o w e l — i f the prefix h a s a derived H , as in the s e c o n d ex-
a m p l e , then it c a n n o t u n d e r g o G F . T h i s s h o w s that ITI applies before G F . Since
G F is lexical, ITI m u s t also b e lexical. T h i s creates a n o t h e r p a r a d o x for the stan-
d a r d m o d e l of LP. T h e s a n d h i rule I T I h a s to a p p l y before a lexical rule, so it MUST
b e lexical, but the rule refers to p h o n o l o g i c a l and m o r p h o l o g i c a l properties of the
p r e c e d i n g word, so the rule C A N N O T b e lexical.

2.2. Precompilation

N L P has n o p r o b l e m s with this state of affairs, since in that theory the output
of the syntax is fully available to lexical p h o n o l o g y , and therefore these sandhi
rules can b e lexical. T h e q u e s t i o n r e m a i n s w h e t h e r o n e can h a n d l e these data,
retaining the s u p p o s e d o r d e r i n g b e t w e e n syntax and lexical phonology, by m o d i -
fying L P in s o m e way. H a y e s ( 1 9 9 0 ) , discussing similar p r o b l e m s , including the
S h o r t e n i n g - G l i d e F o r m a t i o n p a r a d o x of K i m a t u u m b i , p r o p o s e s a different m o d i -
fication of p h o n o l o g y , n a m e l y p r e c o m p i l a t i o n theory. In p r e c o m p i l a t i o n theory a
w o r d m a y have m u l t i p l e lexical derivations; h e n c e multiple outputs e m e r g e from
the lexicon for e a c h w o r d . E a c h of t h e s e derivations is t a g g e d for a property re-
ferred to as a lexical instantiation frame. L a n g u a g e s m a y define sets of instantia-
tion frames w h i c h serve as c o n t e x t for lexical p h o n o l o g i c a l rules. U p o n entering
the postlexical p h o n o l o g y , the frame definitions of the l a n g u a g e are consulted, and
the syntactic, m o r p h o l o g i c a l , and p h o n o l o g i c a l properties of the w o r d in its sen-
tence are c h e c k e d ; out of the various derivations g e n e r a t e d in the lexicon, the
correct form is then inserted into the sentence, and the string is submitted to the
Interaction in Lexical Phonology 123

postlexical p h o n o l o g y . T h u s , the p r e c o m p i l e d version of K i m a t u u m b i S h o r t e n i n g

is f o r m u l a t e d as in ( 2 1 ) .

(21) VV-->V/[ 1 [Frame 1]

T h e definition of F r a m e 1 is given as ( 2 2 ) .

(22) F R A M E 1: [ , .
X .. [ X ] Y] Y * 0

T h e t w o d e r i v a t i o n s in ( 2 3 ) are then g e n e r a t e d . T h e derivation w h e r e F r a m e 1

rules apply g e n e r a t e s the F r a m e 1 a l l o m o r p h , and t h e other, w h e r e t h e s e rules d o
not apply, g e n e r a t e s the e l s e w h e r e form.

(23) [kyaandangyo chadngu) 'my forest farm'

/andaangyo/

andangyo NA Shortening
k\-andangyo k\andaangyo Level 2 prefixing
kyaandangyo kyaandaangyo Output of lexical phonology

Later, the a p p r o p r i a t e a l l o m o r p h is selected. If the w o r d in q u e s t i o n a p p e a r s in the

syntactic c o n t e x t defined in ( 2 2 ) , the F r a m e 1 a l l o m o r p h is inserted; o t h e r w i s e the
e l s e w h e r e a l l o m o r p h is inserted.
F r a m e definitions m a y include p h o n o l o g i c a l information from n e i g h b o r i n g
w o r d s . K i m a t u u m b i Initial T o n e Insertion c o u l d b e stated to apply in the c o n t e x t
of F r a m e 2 .

(24) INITIAL TONE INSERTION-.

V V/ [ [ F r a m e 2]
[+D]

F r a m e 2 is defined as in ( 2 5 ) . 1

(25) FRAME 2 : [ s t e m ~ H] X ( X d o e s not contain )

s

P r e c o m p i l a t i o n thus allows lexical rules to indirectly see into s u r r o u n d i n g

w o r d s w i t h o u t directly l o o k i n g at t h e m . T h a t is, the rule s y s t e m freely g e n e r a t e s
all sorts of possibilities, then filters out the incorrect o n e s at a later stage. A n
o b s e r v a t i o n can t h u s b e m a d e a b o u t the p r e d i c t i o n s of LP. A l t h o u g h standard L P
d o e s not include p r e c o m p i l a t i o n , this m a c h i n e r y is basically consistent with the
architecture of the theory. S i n c e L P a l o n e c a n n o t h a n d l e K i m a t u u m b i , it m u s t
a d o p t the s u b t h e o r y of p r e c o m p i l a t i o n to a c h i e v e o b s e r v a t i o n a l adequacy. H o w -
ever, since b o t h E x t e n d e d L P and N L P can g e n e r a t e the correct forms, and since
the d e v i c e of p r e c o m p i l a t i o n allows p r e c o m p i l i n g any information about the syn-
124 David Odden

tax, m o r p h o l o g y , a n d p h o n o l o g y of s u r r o u n d i n g w o r d s , then E L P m a k e s n o p r e -
diction a b o u t information relations b e t w e e n syntax and p h o n o l o g y ; rather, it d e -
scribes these relations b y different m e a n s from NLP, but the forms g e n e r a b l e in
both theories are the s a m e .
If the theories are w e a k l y equivalent, then w e m u s t turn to s e c o n d a r y consid-
erations such as e l e g a n c e a n d c o m p u t a t i o n a l c o m p l e x i t y to evaluate the theories.
T h e r e is a c o m p u t a t i o n a l a d v a n t a g e to a l l o w i n g lexical rules to directly see the
o u t p u t of syntax. In p r e c o m p i l a t i o n theory, there can b e m u l t i p l e frames, and
frames can overlap, so e a c h frame definition d o u b l e s the n u m b e r of derivations
n e c e s s a r y for a form. F o r instance, if there are t w o frames defined in a l a n g u a g e ,
then four derivations are required for a w o r d , o n e for F r a m e 1, o n e for F r a m e 2,
o n e for F r a m e s 1 a n d 2, a n d o n e for the e l s e w h e r e form. In general, w h e n there
are n frames, w e n e e d 2 derivations. In the c a s e of K i m a t u u m b i (as discussed in
n

O d d e n , 1992), there are eight external sandhi rules w h i c h are p a r t of the lexical
p h o n o l o g y , so 2 5 6 parallel derivations are n e e d e d for e a c h w o r d . O n the other
h a n d , in the theory o r d e r i n g lexical p h o n o l o g y after syntax, only o n e derivation is
required, since the rules simply inspect the w o r d - e x t e r n a l c o n t e x t to d e t e r m i n e
w h e t h e r their c o n d i t i o n s for application are satisfied. P r e c o m p i l a t i o n theory thus
entails m o r e c o m p l e x c o m p u t a t i o n s .

3. M O R P H O L O G Y A N D P H O N O L O G Y

I n o w turn to the s e c o n d question of information a c c e s s , n a m e l y the f u n d a m e n -

tal question in m o r p h o l o g y - p h o n o l o g y interaction, w h i c h is w h e t h e r m o r p h o l o g y
c a n b e sensitive to the output of p h o n o l o g y . S i n c e m y thesis is that m o r p h o l o g y
exclusively PRECEDES p h o n o l o g y , I start with an investigation of a n e w c a s e w h e r e
m o r p h o l o g y s e e m s to FOLLOW p h o n o l o g y .

3.1. 1-SinguIar A l l o m o r p h y in A r a b i c

T h e p r o b l e m discussed h e r e involves the p h o n o l o g y of glides in Classical Ara-

bic, and an a l l o m o r p h of the 1-singular p o s s e s s i v e suffix. A n excellent analysis of
A r a b i c glides is available in B r a m e ( 1 9 7 0 ) . A basic fact about glides in Classical
A r a b i c is that they are p h o n o l o g i c a l l y " w e a k " a n d often elide. T h e i m p o r t a n t
generalization for o u r p u r p o s e s is that intervocalic glides delete, and the resulting
vowel cluster fuses into a single long v o w e l . S i n c e Classical A r a b i c d o e s not allow
long v o w e l s in closed syllables, this long v o w e l m a y then shorten. T h e s e rules are
seen in the p a r a d i g m of the perfective verb in (26). T h e left c o l u m n s h o w s the
p a r a d i g m of a v e r b w h i c h suffers n o alternations. T h e other t w o c o l u m n s illustrate
s t e m s with final y and w w h i c h delete intervocalically.
Interaction in Lexical Phonology 125

(26) S O U N D VERB FINAL y FINAL W

qatal-tu ramay-tu da ?aw-tu 1 sg.
qatal-nd ramay-nd da Taw-na lpl.
qatal-a ram-d da?-a 3 sg. m a s c .
qatal-at ram-at da T-at 3 sg. fem.
'kill' 'throw' 'call'

T h r e e rules are r e s p o n s i b l e for these alternation, namely, G l i d e Elision, w h i c h

deletes intervocalic glides; Vowel F u s i o n , w h i c h fuses vowel clusters into o n e
long v o w e l ; a n d C l o s e d Syllable S h o r t e n i n g .

(27) G L I D E ELISION: c
—con • 0/V- —V
+ son

V O W E L FUSION: O O

C L O S E D SYLLABLE SHORTENING:

D e r i v a t i o n s of illustrative f o r m s are given in (28).

(28) ramay-a ramay-at underlying

rama-a rama-at G l i d e Elision
rama ramat Vowel F u s i o n
NA ramat Shortening

T h e rules G l i d e Elision a n d Vowel F u s i o n are the t w o crucial rules involved in

this c a s e of p h o n o l o g y s u p p o s e d l y p r e c e d i n g m o r p h o l o g y .
T h e a l l o m o r p h y w e are c o n c e r n e d with is that of the 1-singular p r o n o m i n a l
suffix on n o u n s . A s w e see in (29), n o u n s are c o m p o s e d of a stem p l u s an obliga-
tory c a s e e n d i n g , u ( n o m . ) , i (gen.), or a ( a c c ) .

(29) ?al-kitab-u 'the book (nom.)'

def-book-nom.
?al-kitab-i 'the book (gen.)'
def-book-gen.
126 David Odden

?al-kitab-a 'the book (ace.)'

def-book-acc.

A p r o n o m i n a l suffix such as ha, ka, or na m a y b e added, and it stands after the

c a s e e n d i n g , as (30) s h o w s .

(30) kitab-u-ha 'her book (nom.)'

kitab-u-ka ' y o u r ( m a s c . sg.) b o o k ( n o m . ) '
kitab-i-ka ' y o u r ( m a s c . sg.) b o o k ( g e n . ) '
kitab-a-na 'our book (acc.)'

In (31) w e find e x a m p l e s of n o u n s w i t h the 1-singular suffix -T-. T h i s suffix is

vowel-initial a n d c o m b i n e s with the c a s e m a r k e r in such a w a y that the case
m a r k e r is c o m p l e t e l y lost.

(31) /kitab-u-T/ kitab-T 'my book (nom.)'

/kitab-i-T/ kitab-T 'my book (gen.)'
/kitab-a-T/ kitab-T 'my book (acc.)'

Vowel F u s i o n (27) c a n n o t apply w h e n the s e c o n d vowel in a s e q u e n c e of vowels

is long, b u t A r a b i c d o e s n o t tolerate v o w e l - v o w e l s e q u e n c e s , so the first vowel of
the u n d e r l y i n g v o w e l s e q u e n c e is therefore deleted; this can b e h a n d l e d b y m a k i n g
Vowel F u s i o n a m i r r o r i m a g e rule.
T h e 1-singular suffix is subject to p h o n o l o g i c a l l y c o n d i t i o n e d allomorphy, il-
lustrated in (32); if it stands i m m e d i a t e l y after a long vowel or d i p h t h o n g , it takes
the form ya. B y a r e g u l a r m o r p h o l o g i c a l principle, the indefinite suffix na or its
variant ni is not used w h e n followed b y a p r o n o u n suffix.

(32) NOUN NOUN + I S G . POSS.

yulam-a-ni 'slaves ( n o m . d u a l ) ' yulam-a-ya
s l a v e - n o m . dual-indef slave-nom. d u a l - m y
yulam-ay-ni 'slaves (oblique d u a l ) ' yulam-ay-ya
slave-obl. dual-indef slave-obl. d u a l - m y
muTallim-u-na 'teachers (nom. pi.)' mu Tallim-u-ya
t e a c h e r - n o m . pl.-indef teacher-nom. pi.-my
mufallim-T-na ' t e a c h e r s (oblique p i . ) ' muSallim-T-ya
teacher-obl. pl.-indef teacher-obl. p i . - m y

In these e x a m p l e s , the long vowel or d i p h t h o n g w h i c h conditions the ya variant is

present in u n d e r l y i n g representation. O t h e r e x a m p l e s , seen in (33), s h o w that long
v o w e l s w h i c h derive b y p h o n o l o g i c a l rule, in particular b y G l i d e Elision and
Vowel F u s i o n , also trigger the ya a l l o m o r p h .

(33) ?al-qahw-at-u 'the coffee ( n o m . ) '

/ma-qhaw-un/ ma-qha-n ' a coffee h o u s e ( n o m . ) '
/?al-ma-qhaw-u/ ?al-ma-qha ' t h e coffee h o u s e ( n o m . ) '
/ma-qhaw-u-1 sg./ ma-qha-ya ' m y coffee h o u s e '
Interaction in Lexical Phonology 127

?al-hawdy-at-u 'the hobby (nom.)'

/?al-haway-u/ ?al-hawd ' t h e desire ( n o m . ) '
/haway-u-lsg./ hawd-ya 'my love'

qaday-tu T settled'
/qadiy-un/ qadin 'a judge (nom.)'
/?al-qadiy-un/ ?al-qddi 'the judge (nom.)'
qadiy-u-1 s g . / qddi-ya 'my judge'

T h u s the stem for 'coffee' e n d s in the glide w, w h i c h surfaces p o s t c o n s o n a n t -

ally in ?alqdhwatu. Intervocalically, in ?almdqhd, the glide deletes before the
c a s e e n d i n g a n d the vowel cluster fuses into a long vowel. T h i s derived l o n g
v o w e l then c o n d i t i o n s the ya a l l o m o r p h . A s (34) indicates, this suggests that the
1-singular a l l o m o r p h is d e t e r m i n e d after G l i d e Elision a n d Syllable F u s i o n .

(34) /maqhaw-u/ underlying

maqhau G l i d e Elision
maqhd Vowel F u s i o n
maqhdya affixation of 1-singular

T h i s is w h a t L P predicts c o u l d h a p p e n , b u t it w o u l d s e e m to b e a p r o b l e m for N L P ,
w h i c h requires all m o r p h e m e s to b e c o n c a t e n a t e d before any p h o n o l o g i c a l rules
apply. H o w e v e r , it is s h o w n b e l o w that this c a s e C A N b e h a n d l e d in the noninter-
active theory.

3.2. Morphological and Phonological Operations

T h i s section p r e s e n t s a reanalysis of this a n d similar cases w h e r e m o r p h o l o g y

h a s b e e n c l a i m e d to follows p h o n o l o g y . A search of the literature reveals a n u m b e r
of putative c a s e s of p h o n o l o g i c a l rules a p p l y i n g before m o r p h o l o g y ; H a r g u s (this
v o l u m e ) p r o v i d e s an e x t e n s i v e list of such c a s e s . T h e largest class is typified b y
the e x a m p l e s in (35).

(35) O v e r a p p l i c a t i o n of p h o n o l o g i c a l rules u n d e r reduplication ( K i h e h e , Taga-

log, J a v a n e s e ) . In K i h e h e , the stem, e x c l u d i n g prefixes, reduplicates (kute-
lekateleka). P h o n o l o g i c a l rules of syllable fusion d r a w prefix material into
the stem, c a u s i n g t h e m to b e r e d u p l i c a t e d (kwiitakwiitd). ( O d d e n and
O d d e n , 1986)

T h e i m p e r a t i v e in D a n i s h is f o r m e d b y deleting the infinitive -Q suffix. D e -

letion follows a vowel l e n g t h e n i n g rule, so / b a e d o / b e c o m e s bee: eta ( w h i c h
is the infinitive), then the infinitive e n d i n g is deleted in the imperative, giv-
ing [ b a e : d ] . S i m i l a r rules deleting the a g r e e m e n t m o r p h e m e y in A b k h a z ,
and the verb suffix a in Icelandic, h a v e b e e n found. ( A n d e r s o n , 1975)

O n e large class i n c l u d e s overapplication of p h o n o l o g i c a l rules u n d e r reduplica-

tion, as in K i h e h e . A n o t h e r large class is typified by i m p e r a t i v e formation in D a n -
128 David Odden

ish, w h i c h deletes the -Q suffix of the infinitive. I m p e r a t i v e D e l e t i o n h a s to be

o r d e r e d after a p h o n o l o g i c a l rule of o p e n - s y l l a b l e vowel l e n g t h e n i n g .
T h e r e s e e m s to b e little d o u b t that the p h o n o l o g i c a l rules w h i c h supposedly
p r e c e d e m o r p h o l o g y in these cases are i n d e e d p h o n o l o g i c a l rules. W h a t is not at
all clear is that the s u p p o s e d l y m o r p h o l o g i c a l o p e r a t i o n s are part of m o r p h o l o g y .
C o n s i d e r the class of p o s t - p h o n o l o g i c a l subtractions, such as the D a n i s h i m p e r a -
tive: other c a s e s like this are c o n s i d e r e d in Section 3.2.2. In light of the rule or-
d e r i n g fact that truncation follows a p h o n o l o g i c a l rule, the only possible analysis
in N L P is to treat the deletion as p a r t of p h o n o l o g y , as ( 3 6 ) .

(36) D A N I S H IMPERATIVE T R U N C A T I O N
9-^0/ J
[Imper]

I thus a d o p t the analysis p r o p o s e d b y A n d e r s o n ( 1 9 7 5 ) that the imperative is b a s e d

o n a form identical to the infinitive.
T h e r e is n o t h i n g in t h e generally a c c e p t e d and m o t i v a t e d theory of p h o n o l o g y ,
b e it L P or nonlexical theories, w h i c h p r e c l u d e s h a v i n g a rule like (36) in the
p h o n o l o g y , since n o p r i n c i p l e of p h o n o l o g i c a l theory prohibits rules from refer-
ring to m o r p h o l o g i c a l p r o p e r t i e s . T h e r e f o r e the analysis (36) is p o s s i b l e in all
theories. In LP, it c o u l d also b e treated as a m o r p h e m e - d e l e t i o n rule and b e part of
the m o r p h o l o g i c a l c o m p o n e n t , as in (37).

(37) [Infin]-*0/ J
[Imper]

F o r that matter, if o n e a l l o w s deletion of p h o n o l o g i c a l units in t h e m o r p h o l o g y ,

not j u s t deletion of m o r p h e m e s , then o n e c o u l d a s s u m e a rule with the formal
statement (36) but p l a c e that rule in the m o r p h o l o g i c a l c o m p o n e n t . In other w o r d s ,
with n o further conditions on p h o n o l o g y or m o r p h o l o g y , three analyses are p o s -
sible and c a n n o t b e d i s t i n g u i s h e d e m p i r i c a l l y or on the basis of rule e l e g a n c e . If
o n e adopts the position that rules such as D a n i s h I m p e r a t i v e T r u n c a t i o n are p h o -
nological rules with m o r p h o l o g i c a l c o n d i t i o n s and are not rules of the m o r p h o -
logical c o m p o n e n t , then they clearly d o not s h o w that the output of p h o n o l o g y
can serve as the input to m o r p h o l o g y .
O n e of the basic s t u m b l i n g b l o c k s in r e s o l v i n g issues r e g a r d i n g the relation
b e t w e e n p h o n o l o g y a n d m o r p h o l o g y is this analytic ambiguity, a n d in particular
the fact that m a n y theories allow o n e to c o n s i g n p h o n o l o g i c a l p r o c e s s e s either to
the p h o n o l o g y or to the m o r p h o l o g y rather freely. W e m u s t m a k e clear w h a t w e
m e a n by " r u l e of m o r p h o l o g y " ; the c l a i m m a d e here, a n d the claim w h i c h m u s t
b e m a d e in L P if there is c o n t e n t to the c l a i m for the interleaving of p h o n o l o g y
and m o r p h o l o g y , is that a " r u l e of m o r p h o l o g y " is a rule in the m o r p h o l o g i c a l
c o m p o n e n t . Putting D a n i s h I m p e r a t i v e T r u n c a t i o n in the p h o n o l o g y but calling it
a " r u l e of m o r p h o l o g y " simply b e c a u s e it applies in a m o r p h o l o g i c a l l y defined
c o n t e x t trivializes the difference b e t w e e n p h o n o l o g y and m o r p h o l o g y . B y a n a l o -
Interaction in Lexical Phonology 129

g o u s r e a s o n i n g , w e s h o u l d call t h e K i m a t u u m b i rule S h o r t e n i n g a rule of syntax

b e c a u s e it a p p l i e s in a syntactically defined context.
T h e s y s t e m a t i c u n c e r t a i n t y a b o u t w h a t constitutes a p o s s i b l e m o r p h o l o g i c a l
rule surely n e e d s a p r i n c i p l e d resolution, so t o attack t h e p r o b l e m from t h e side of
m o r p h o l o g y , (38) is a d o p t e d t o restrict m o r p h o l o g y to c o n c a t e n a t i o n .

(38) PRINCIPLE OF M O R P H O L O G Y - P H O N O L O G Y SEGREGATION: T h e only opera-

tion a l l o w e d in m o r p h o l o g y is c o n c a t e n a t i o n .

O f c o u r s e , in t h i n k i n g a b o u t this restriction t o c o n c a t e n a t i o n , o n e s h o u l d k e e p in
m i n d that n o n l i n e a r r e p r e s e n t a t i o n s m a y m a k e it LOOK like a m o r p h o l o g i c a l p r o -
c e s s is d o i n g m e t a t h e s i s o r infixing, o r g e m i n a t i n g c o n s o n a n t s . A s M c C a r t h y
( 1 9 7 9 ) s h o w s , this is j u s t a n illusion.
T h e c o m p l e t e s e g r e g a t i o n of m o r p h o l o g i c a l a n d p h o n o l o g i c a l o p e r a t i o n s m a k e s
strong c l a i m s a b o u t m o r p h o l o g y . A c o n s e q u e n c e of this p r i n c i p l e is that w e rule
o u t o n theoretical g r o u n d s s u p p o s e d c a s e s of " P r o c e s s M o r p h o l o g y " ( M a t t h e w s ,
1974, inter alia), w h e r e m o r p h o l o g i c a l rules p e r f o r m p h o n o l o g i c a l c h a n g e s . E x -
a m p l e s of p r o c e s s m o r p h o l o g y m u s t b e r e a n a l y z e d as t w o p r o c e s s e s , n a m e l y
p u r e l y m o r p h o l o g i c a l c o n c a t e n a t i o n , p l u s a p h o n o l o g i c a l o p e r a t i o n . F o r instance,
the m o r p h o l o g i c a l p a r t of G e r m a n U m l a u t is s i m p l y stating t h e c o n d i t i o n s for
a d d i n g a n affix o r set of affixes. It h a p p e n s that t h e p h o n o l o g i c a l c o n t e n t of this
suffix is o r c o n t a i n s a floating v o w e l feature [ — b a c k ] . T h e p h o n o l o g y is then
r e s p o n s i b l e for l i n k i n g that feature t o t h e a p p r o p r i a t e v o w e l .

3.2.1. REDUPLICATION

S u c h a division of l a b o r r e m o v e s t h e K i h e h e k i n d of r e d u p l i c a t i o n from t h e
p o o l of s u p p o r t for the interactive m o d e l . T h e p r o b l e m of K i h e h e is that t h e entity
w h i c h r e d u p l i c a t e s is t h e stem, w h i c h is t h e o u t p u t of L e v e l 1 m o r p h o l o g y a n d
thus g e n e r a l l y e x c l u d e s t h e object prefix o r t h e infinitive prefix, w h i c h a r e at
L e v e l 2 . H o w e v e r , in c a s e s o m e prefixal e l e m e n t h a s fused syllabically w i t h t h e
initial s t e m syllable, t h e prefix s e g m e n t s g e t c o p i e d as well.

(39) ku-teleka 'to cook'

ku-teleka-teleka 'to cook a bit'
ku-lu-teleka ' t o c o o k it'
ku-lu-teleka-teleka ' t o c o o k it a b i t '
kw-iita ' t o spill'
kw-iita-kw-iita (/ku-ita/) ' t o spill a b i t '
ku-lw-iita ' t o spill it'
ku-lw-iit-lw-iita (/ku-lu-ita/) ' t o spill it a b i t '

It thus s e e m s that G l i d e F o r m a t i o n m u s t p r e c e d e r e d u p l i c a t i o n , in o r d e r t o explain

w h y c o p y i n g i n c l u d e s m a t e r i a l from a prefix.
A s n u m e r o u s r e s e a r c h e r s (e.g., M a r a n t z , 1982), h a v e o b s e r v e d , r e d u p l i c a t i o n is
not a single c o m p l e x o p e r a t i o n b u t is a set of interrelated p h e n o m e n a i m p l e m e n t e d
130 David Odden

by a n u m b e r of restricted rules. U n d e r such v i e w s of reduplication, the purely

m o r p h o l o g i c a l operation is the affixation p r o c e s s — a d e g e n e r a t e e l e m e n t such as
a syllable t e m p l a t e is a d d e d . In the c a s e of K i h e h e it is not trivially o b v i o u s w h a t
e l e m e n t is affixed, since multiple syllables a n d m o r p h e m e s are copied. I a s s u m e
that the root plus following suffixes form a p r o s o d i c w o r d OJ, and that the e m p t y
affix is OJ. R e d u p l i c a t i o n is the prefixing of an e m p t y OJ on Level 2 to the Level 1
constituent, the stem. T h e p h o n o l o g i c a l c o p y operation involves m a p p i n g the
e m p t y OJ o n t o the OJ created at Level 1 — t h e r e are n u m e r o u s m o d e l s of h o w this
p r o c e d u r e w o r k s , b u t for the sake of exposition I a d o p t the parafixation m o d e l of
C l e m e n t s ( 1 9 8 6 ) . T h e e m p t y OJ then receives a c o p y of the p r o s o d i c and s e g m e n t a l
material subordinate to the original co and is s e q u e n c e d to the left of that OJ. T h e
derivation of the r e d u p l i c a t e d form kutelekateleka is p r o v i d e d below.

(40) L 2 to underlying

telek a
ku 'a bit' cook tense
infinitive (redup.) (stem)

L 9 co L, to Lj phonology

A
a a a

ku
UA
telek a
mapping, copy

Li telek a

to

0)

A
a a a

ku
UA
telek a

L2
L ? to Lo to sequencing

A A
a a a a a a
li k A H A A
UA UA
ku telek a telek a

T h e difference b e t w e e n kutelekateleka and kwiitakwiita is simply that in the

latter case, w h e r e the infinitive prefix is c o p i e d contrary to the general pattern, the
Interaction in Lexical Phonology 131

rule G l i d e F o r m a t i o n applies before the c o p y i n g rule, so s e g m e n t a l material w h i c h

h a s accidentally b e c o m e d o m i n a t e d b y the b a s e co—in general, the o u t p u t of any
syllable r e o r g a n i z a t i o n — g e t c o p i e d a l o n g with the s e g m e n t s w h o s e m e m b e r s h i p
in the b a s e co is m o t i v a t e d on purely m o r p h o l o g i c a l g r o u n d s .

(41) Glide Formation (L ) 2

mapping, copy

L, co sequencing
^9 <JU
A A
a a

A\t\
• a a

M
kwit a
A d d i n g the p r o s o d i c affix co constitutes the entire m o r p h o l o g y of reduplication;
the ability to a d d this affix or the selection of a specific reduplication m o r p h e m e
as t h e realization of a m o r p h o l o g i c a l c a t e g o r y is not in any w a y affected b y p r i o r
application of p h o n o l o g i c a l rules in K i h e h e , n o r is it in any other l a n g u a g e . T h e
interesting a n d characteristic w o r k involved in reduplication is largely d o n e b y the
p h o n o l o g y , w h i c h r e c e i v e s this d e g e n e r a t e representation, n a m e l y a s e q u e n c e of
real s e g m e n t s p l u s s o m e s e g m e n t a l l y e m p t y p r o s o d i c t e m p l a t e , a n d the p h o n o l o g y
h a s the responsibility for filling in that t e m p l a t e . In K i h e h e a n d cases like it, this
takes p l a c e after certain p h o n o l o g i c a l r u l e s . 2

3.2.2. TRUNCATION

R e t u r n i n g to cases like D a n i s h , other cases of so-called subtractive m o r p h o l o g y

h a v e b e e n b r o u g h t out in the literature. M a r t i n ( 1 9 8 8 ) s h o w s that pluralization in
Koasati, seen in (42), m a y delete the stem-final r h y m e .
132 David Odden

(42) SINGULAR PLURAL GLOSS

pitdf-fi-n pit-li-n ' t o slice u p the m i d d l e '
tiwdp-li-n tiw-wi-n 'to open something'
atakd:-li-n atdk-li-n 'to hang something'
misip-li-n mis-li-n 'to w i n k '
koydf-fi-n koy-li-n ' t o cut s o m e t h i n g '

In N L P , this p r o c e s s m u s t result from a m o r p h o l o g i c a l l y c o n d i t i o n e d p h o n o l o g i c a l

rule, since all deletions m u s t b e p a r t of p h o n o l o g y .

(43) KOASATI RIME DELETION:

R ^ 0 / stem
[+Plural]

This c a s e is similar t o w h a t w e find in D a n i s h , e x c e p t that deletion affects a

h i g h e r level p r o s o d i c unit, n a m e l y a r i m e , and therefore indirectly affects multiple
segments.
Koasati constitutes a neutral territory, w h e r e both the L P and N L P a c c o u n t s are
plausible a n d consistent with t h e internal structure of t h e respective theories.
T h e r e is a similar deletion p r o c e s s in K i m a t u u m b i w h i c h m u s t b e analysed as
resulting from a deletion r u l e — a n d since this deletion is postlexical, the deletion
m u s t b e p h o n o l o g i c a l . T h i s rule deletes t h e s e g m e n t a l material of o n e of t h e
3

locative prefixes, ku-, after a v o w e l ; this rule, illustrated in ( 4 4 ) , is optional, so

there are t w o variants for e a c h s e n t e n c e .

(44) njyenda kyKlpodj T a m going to Kipooi.'

nyyendad Kipooi id.
eenddbutukd kuKjpdtlmu ' H e is r u n n i n g t o K i p a t i m u . '
eenddbutukda Klpdtlmu id.

It is a p p a r e n t that the locative prefix ku- is present in u n d e r l y i n g representations,

but is in part d e l e t e d — t h e e v i d e n c e for its u n d e r l y i n g p r e s e n c e even w h e n deleted
is that its tone a n d m o r a are p r e s e r v e d . T h e rule deletes the s e g m e n t a l material of
the syllable ku- but p r e s e r v e s t o n e a n d m o r a i c s t r u c t u r e . 4

(45) LOCATIVE TRUNCATION (optional):

o" Lo" —> 0 (prosodic structure p r e s e r v e d )
[ku]

T h e syllable p r e c e d i n g ku- takes t h e t o n e a n d m o r a originally part of the syl-

lable of ku-, so for this reason, the locative truncation rule is stated t o affect only
the s e g m e n t a l representation. A s seen in t h e derivation ( 4 6 ) , t h e H tone o n ku-
c o m e s from applying Initial T o n e Insertion.

(46) nyyenda ku-kppool underlying

nyyenda ky-kypool ITI
Interaction in Lexical Phonology 133

nyyenda /i kjpodj Locative Truncation

n\yendad k\po6\ reaffiliation of s t r a n d e d m o r a

Finally, ky- deletion is possible only if the prefix syllable is m o n o m o r a i c ; vari-

ous regular syllable fusions can m a k e the prefix b i m o r a i c , thus b l o c k i n g deletion,
as in (47).

(47) nyyenda kw\\s\wd T a m g o i n g to the i s l a n d s '

nyyenda kyynkoongo T a m g o i n g to M k o n g o '
nyyenda ky-my-koongo underlying
njyenda ky-m-koongo U-deletion
njyenda ky-m-koongo ITI
njyenda kyy-nkoongo N a s a l desyllabification
NA Locative Truncation

T h e s e c o n d e x a m p l e involves l e n g t h e n i n g of the locative prefix syllable as a c o n -

s e q u e n c e of postlexical desyllabification of the derived syllabic nasal in nkdongo;
this then a r g u e s that L o c a t i v e T r u n c a t i o n is postlexical. L o c a t i v e T r u n c a t i o n a p -
plies after p h o n o l o g i c a l rules, so it m u s t itself b e a p h o n o l o g i c a l rule, specifically
the dissociation of the s e g m e n t s of this m o r p h e m e , with retention of p r o s o d i c
structure.

3.2.3. TRUNCATION A N D DEFAULTS

T h e s e deletions a n d p r o s o d i c r e s t r u c t u r i n g s b r i n g us closer to the a p p a r e n t c a s e

of p o s t p h o n o l o g i c a l a l l o m o r p h y in A r a b i c w h i c h I started with. Before getting to
that case, I c o n s i d e r a n o t h e r e x a m p l e w h i c h is often treated as p h o n o l o g i c a l l y
c o n d i t i o n e d allomorphy, but w h i c h h a s a n o t h e r interpretation, as a m o r p h o l o g i -
cally c o n d i t i o n e d p h o n o l o g i c a l rule, o n e with a similarity to K i m a t u u m b i L o c a t i v e
T r u n c a t i o n . T h i s is the c a s e of K o r e a n i ~ ka allomorphy.
T h e relevant facts are as follows. C a s e m a r k e r s suffer certain variations, deter-
m i n e d by w h e t h e r they are a d d e d to a consonant-final b a s e or a vowel-final b a s e .
T h u s the accusative is ril after a vowel but il after a c o n s o n a n t , the topic m a r k e r
is nin after a vowel but in after a c o n s o n a n t , a n d the n o m i n a t i v e is i after a c o n -
sonant a n d ka after a vowel.

(48) citation: param pori

nominative: param-i pori-ka
accusative: param-il pori-ril
topic: param-in pori-nin
'wind' 'barley'

W r i t i n g rules to delete / or n after a c o n s o n a n t is u n p r o b l e m a t i c , and in fact

treating the il ~ ril alternation with suppletion-style a l l o m o r p h y fails to c a p t u r e
134 David Odden

the p h o n o l o g i c a l similarity b e t w e e n the a l l o m o r p h s . T h e p r o b l e m really is in the

n o m i n a t i v e ; can w e h a n d l e this alternation b y a p h o n o l o g i c a l rule? It turns out that
there is a very s i m p l e w a y to a c c o u n t for this a l l o m o r p h y b y an operation entirely
a n a l o g o u s to the K i m a t u u m b i L o c a t i v e T r u n c a t i o n and Koasati R i m e Deletion
rules. Specifically, w e a s s u m e the u n d e r l y i n g affix ka a n d i n v o k e a rule to delete
the s e g m e n t a l c o n t e n t of this syllable after a c o n s o n a n t .

(49) KOREAN NOMINATIVE DESTRUCTURING:

a a —> 0 (prosodic structure p r e s e r v e d )

| [+Nom]
C

T h i s will leave b e h i n d a s e g m e n t l e s s m o r a . Default rules then assign the n e c e s s a r y

features, and w e e n d u p with i. 5

(50) p a r a m - k a underlying

C V C V C C V

a a a

nominative destructuring

C V C V C

a a a

defaults, resyllabification

C V C V C V

\ \ If
a a a

A s it h a p p e n s , n o p h o n o l o g i c a l rules feed into (49), so K o r e a n is not crucial for

distinguishing the theories. N e v e r t h e l e s s , it suggests a direction for reanalysis of
other s u p p o s e d c a s e s of p h o n o l o g i c a l l y c o n d i t i o n e d a l l o m o r p h y : such rules m i g h t
b e slightly bizarre m o r p h o l o g i c a l l y c o n d i t i o n e d rules in the p h o n o l o g y . T h i s then
b r i n g s us to the Classical A r a b i c p o s s e s s i v e a l l o m o r p h y rule, w h e r e w e find ya
selected after long v o w e l s . T h i s will b e h a n d l e d as diphthongization arising from
Interaction in Lexical Phonology 135

p r o s o d i c destruction. T h e syllable structure of u n d e r l y i n g /: is r e d u c e d to a simple

C V c o r e syllable, with / a s s i g n e d to the o n s e t b y rule (51).

(51) CLASSICAL A R A B I C I-SINGULAR DIPHTHONGIZATION

a a a

V V C V / V

T h e syllable p e a k n o w h a s n o s e g m e n t a l m a t e r i a l , so default rules fill in the values

for a.

(52) maqha i output of Glide Elision, Vowel Fusion

V V V V

maqha i diphthongization

V V C V

maqha i a defaults

V V C V

T h e s e e x a m p l e s s h o w that deletion is n e e d e d in p h o n o l o g y , w h i c h is hardly

surprising, a n d that deletion can affect m u l t i p l e s e g m e n t s , as l o n g as they f o r m a
h i g h e r level constituent. In this enterprise of r e a n a l y z i n g putative feeding from
p h o n o l o g y into m o r p h o l o g y , it is i m p o r t a n t to k n o w w h a t limits are to b e i m p o s e d .
T h e limits are, of c o u r s e , the limits i m p o s e d on p h o n o l o g i c a l analysis: w e d o not
p r o p o s e that A N Y w e l l - m o t i v a t e d c o n s t r a i n t s on p h o n o l o g y b e r e l a x e d for t h e s a k e
of r e a n a l y z i n g s u p p o s e d m o r p h o l o g i c a l o p e r a t i o n s . C o n s i s t e n t with the constraint
that p h o n o l o g i c a l o p e r a t i o n s apply to p r o s o d i c or ( s u b ) s e g m e n t a l constituents, w e
predict that there is n o so-called subtractive m o r p h o l o g y affecting m o r p h e m e s
w h i c h are not p h o n o l o g i c a l constituents. A specific case of this constraint on
analysis can b e seen in C h i m w i i n i .

3.2.4. T H E CHIMWIINI PASSIVE PERFECTIVE

Kisseberth a n d A b a s h e i k h ( 1 9 7 4 ) n o t e that the p a s s i v e suffix in C h i m w i i n i is

generally o: w, as w e see in (53).
136 David Odden

(53) ku-lum-o:w-a 'to b e bitten

na-kimb-il-o: w-a ' s h e is b e i n g sung t o '
ku-dar-o: w-a 'to be touched'

Before the p a s s i v e suffix, dental / b e c o m e s alveolar, as the forms of (54) show.

(54) ku-ya:l-a 'to sow' ku-yal-o: w-a 'to b e s o w n '

k-i:ngil-a 'to enter' k-ingil-o: w-a 'to be entered'
x-fu: ngul-a 'to o p e n ' x-fungul-o: w-a 'to b e o p e n e d '

Kisseberth and A b a s h e i k h suggest that this c o u l d b e d o n e b y a p h o n o l o g i c a l rule,

as in (55).

(55) /->// + o:w

T h e perfective tense p a s s i v e is s o m e w h a t p r o b l e m a t i c , since the p a s s i v e suffix

o: w is not found o n the surface. F u r t h e r m o r e , the perfective passive irregularly
selects the final vowel a rather than e, as (56) s h o w s .

(56) lum-i :l-e 'he bit' lum-i: l-a ' h e w a s bitten'

som-e:l-e 'he read' chi-som-e: l-a 'it w a s r e a d '
fungi :l-e 'he opened' i-fungi:l-a 'it w a s o p e n e d '

T h e s e e x a m p l e s also s h o w that the rule c h a n g i n g dental / to alveolar still applies,

even t h o u g h the s u p p o s e d c o n d i t i o n i n g factor, the passive m o r p h e m e , is not p h o -
netically present.
In a m o d e l c o u n t e n a n c i n g m o r p h e m e deletions and p h o n o l o g y feeding into
m o r p h o l o g y , w e c o u l d h a n d l e this b y a p p l y i n g the dental-to-alveolar rule first,
then deleting -o:w.

(57) lum-i :l-o: w-a underlying

lum-i: l-o: w-a /-to- /
lum-i: l-a o: w - D e l e t i o n

T h i s is not allowed in the m o d e l p r o p o s e d here, since m o r p h e m e deletion is

barred. In this case, the rule c a n n o t b e p h o n o l o g i c a l deletion, since the s e q u e n c e
-o:w is not any kind of constituent; if it is unsyllabified, it is certainly not a r i m e .
If it is syllabified, it straddles syllables.
Therefore, the only solution is to directly restrict the insertion of the passive
affix, so that it is not insertable in [4-PERFECTIVE] verbs. T h e n h o w about the
dental-to-alveolar c h a n g e ? A s K i s s e b e r t h a n d A b a s h e i k h point out, this c h a n g e
need not b e triggered by the p h o n e m e s e q u e n c e -o:w but could equally refer to
the feature [ + P A S S I V E ] . C e r t a i n l y the c h a n g e from dental to alveolar d o e s not in-
volve feature spreading from o n e of the suffix s e g m e n t s . (I a s s u m e that the rule
deletes the feature [ + d i s t r i b u t e d ] , with later default a s s i g n m e n t of the feature
[ — distributed].)

(58) / -> / / + [PASSIVE]

Interaction in Lexical Phonology 137

T h e r e f o r e , N L P is forced to a d o p t the solution s u g g e s t e d b y K i s s e b e r t h a n d

A b a s h e i k h that t h e m u t a t i o n of / is t r i g g e r e d n o t by the p a s s i v e suffix itself, b u t
the m o r p h o l o g i c a l feature [PASSIVE].
L a c k of s p a c e p r e v e n t s actually r e a n a l y z i n g o t h e r c a s e s w h i c h c o u l d b e c i t e d —
s u p p o s e d e x a m p l e s from J a v a n e s e ( H a r g u s , this v o l u m e ) , S h i , a n d L u g a n d a (Hy-
m a n , 1990) a r e r e a n a l y z e d in O d d e n ( 1 9 9 0 b ) — b u t it should b e clear that it will
take m u c h s t r a n g e r a l l o m o r p h y than is currently available to give u n a m b i g u o u s
s u p p o r t to t h e c l a i m that p h o n o l o g y c a n feed into m o r p h o l o g y . A l l o m o r p h y like
E n g l i s h go ~ went o r be ~ am ~ is ~ were certainly w o u l d qualify, b u t n o n e of
this variation is p h o n o l o g i c a l l y c o n d i t i o n e d , m u c h less c o n d i t i o n e d b y d e r i v e d
phonological information.

3.3. Apparent Problems

T h e r e a r e still u n t o u c h e d c a s e s w h e r e p h o n o l o g i c a l r e a n a l y s i s is n o t p o s s i b l e .
T h e best k n o w n c a s e is the v e r b - t o - n o u n derivational suffix -al in E n g l i s h , w h i c h ,
it is often said, c a n attach o n l y to final-stressed stems.

(59) arrival disposal acquittal refusal reversal

survival bestowal withdrawal betrothal avowal
renewal revival approval transferral betrayal
appraisal deferral referral perusal upheaval
burial denial

T h i s stands as o n e of the s t r o n g e r a r g u m e n t s available for p h o n o l o g y p r e c e d i n g

m o r p h o l o g y , since the putatively d e r i v e d p h o n o l o g i c a l c o n d i t i o n is o n e of the fac-
tors d e t e r m i n i n g w h e t h e r the affix c a n b e u s e d at all; that is, w e are n o t d e a l i n g
w i t h v a r i a t i o n s in t h e s h a p e of a m o r p h e m e o r w i t h p h o n o l o g i c a l deletion. T h i s
c a s e is n e v e r t h e l e s s n o t strong e n o u g h to force a d o p t i o n of L P ; given t h e n o n p r o -
ductivity of this affix a n d the small n u m b e r of f o r m s available, it is i m p o s s i b l e to
really test a n y h y p o t h e s i s r e g a r d i n g this affix. S u p p o s i n g that position of stress
w e r e the c o r r e c t g e n e r a l i z a t i o n , it is p o s s i b l e that stress in these w o r d s is p r e s e n t
in u n d e r l y i n g r e p r e s e n t a t i o n , so this w o u l d n o t b e a c a s e of m o r p h o l o g y b e i n g
sensitive to d e r i v e d p h o n o l o g i c a l i n f o r m a t i o n . F u r t h e r m o r e , there is a different
g e n e r a l i z a t i o n w h i c h c o v e r s the data, n a m e l y that -al only c o m b i n e s with L a t i n a t e
b o u n d prefix p l u s m o n o s y l l a b i c r o o t . 6

Booij ( 1 9 8 1 ) a n d Booij a n d R u b a c h ( 1 9 8 4 , 1987) p r o p o s e t w o other c a s e s from

D u t c h ; these yield to r e a n a l y s i s . T h e first c a s e i n v o l v e s the s u p p o s e d interaction
b e t w e e n s c h w a deletion a n d a rule of a l l o m o r p h y . S u p p o s e d l y , the final s c h w a of
ambassade ' e m b a s s y ' deletes before eur, g i v i n g ambassadeur.

(60) 9->0f V

T h e n an a l l o m o r p h y rule r e p l a c e s eur with ris in t h e f e m i n i n e , giving

ambassadrice.
138 David Odden

(61) ambasadd-or affixation

ambasad-dr s c h w a deletion
ambasad-or-isQ affixation
ambasad-r-iso nce-allomorphy

T h e interleaving of p h o n o l o g y a n d m o r p h o l o g y is m o t i v a t e d by the fact that the

feminine a l l o m o r p h y is consonant-initial, h e n c e could not trigger S c h w a Deletion,
so if the feminine a l l o m o r p h w e r e a d d e d directly to ambassade, w e w o u l d gener-
ate incorrect *ambassaderice.
T h e r e are a n u m b e r of possibilities for reanalysis, all of w h i c h are consistent
with the data cited b y Booij a n d R u b a c h . O n e possibility is that s c h w a deletes
before rice. Or, the a l l o m o r p h y rule c o u l d b e a m o r p h o l o g i c a l l y c o n d i t i o n e d p h o -
nological rule, like the cases w e h a v e seen earlier. A n o t h e r possibility is to derive
ambassadrice from the root ambassad, not the n o u n ambassade. A fourth p o s -
sibility is simply to n o t derive ambassadrice from a n y t h i n g , e x c e p t F r e n c h .
T h e s e c o n d e x a m p l e , in m a n y w a y s similar to E n g l i s h -al, is the case of the
suffix -ief w h i c h only p r o d u c t i v e l y attaches to n o u n s e n d i n g in unstressed /.

(62) psychologie 'psychology' psychologisch 'psychological'

hysterie 'hysteria' hysterisch 'hysterical'
agressie 'aggression' agressief 'agressive'
dctie 'action' actief 'active'

D u t c h stress is certainly not transparently p r e d i c t a b l e like Latin or A r a b i c stress,

so an o b v i o u s direction to l o o k for reanalysis is to a s s u m e that stress is present in
u n d e r l y i n g representations, in w h i c h c a s e w e are n o longer dealing with derived
p h o n o l o g i c a l information. Booij (personal c o m m u n i c a t i o n ) informs m e that the
default position for stress a s s i g n m e n t is o n the penult, and van der H u l s t ( 1 9 8 4 :
235) confirms this generalization: this fits w i t h o u r h y p o t h e s i s . W o r d s with irregu-
lar stress, especially final stress, will b e entered in the lexicon with stress p r e -
assigned, w h e r e a s w o r d s with p e n u l t i m a t e stress will h a v e n o u n d e r l y i n g stress.
T h e condition for affixation of -ief is then simply, as Booij and R u b a c h a s s u m e ,
that it only attaches to w o r d s e n d i n g with u n s t r e s s e d i. W o r d s such as psychologie
with final lexical stress d o not satisfy this condition, so c a n n o t take the affix -ief.

3.4. Precyclicity o r P r e c o m p i l a t i o n ?

To close this investigation, I l o o k at a p r o b l e m w h e r e L P and N L P seem to

m a k e different p r e d i c t i o n s . T h i s is a c a s e in M a l t e s e A r a b i c w h e r e Level 1 p h o -
n o l o g y n e e d s to a c c e s s Level 2 information. T h i s w o u l d s e e m to refute the inter-
active theory, since Level 2 m o r p h o l o g y h a s not even b e e n d o n e at the stage of
Level 1 p h o n o l o g y . H o w e v e r , there is a w a y out for LP.
T h e cyclic Stress rule of M a l t e s e is involved, w h i c h B r a m e (1974) formulates
as in (63) (additional data c o m e from Aquilina, 1965; A q u i l i n a and Isserlin, 1981).
Interaction in Lexical Phonology 139

(63) STRESS:
V -> [+stress] / C ((VC) VC )]
0
1

A p p l y i n g after stress is an A p o c o p e rule w h i c h deletes an u n s t r e s s e d vowel in an

o p e n syllable.

(64) APOCOPE:
V->0/ CV

T h e s e rules interact to a c c o u n t for the p a r a d i g m of hataf in ( 6 5 ) .

(65) /hataf-t/ htdft T snatched'

/hataf-na/ htafna 'we snatched'
/hataf/ hataf 'he snatched'
/hataf-u/ hdtfu 'she snatched'
/hataf-it/ hdtfet 'she snatched'

Stress a n d A p o c o p e apply to u n d e r l y i n g hataf+u to yield hdtfu, a n d apply to

hataf+na to yield htdfna.

(66) hataf hataf-u hataf-na underlying

hataf hdtaf-u hatdf-na Stress
NA hdtf-u htdf-na Apocope

O t h e r m o r p h e m e s w h i c h are Level 2 suffixes, seen in (67), c a n b e a d d e d to the

verb after affixation of subject a g r e e m e n t , i n c l u d i n g the object suffixes -kum ' y o u
(pi.)' and -ik ' y o u ( s g . ) ' , and the n e g a t i v e suffix -s.

(67) hatdf-s ' h e didn't s n a t c h '

hatf-it-kom ' s h e snatched you (pi.)'
hdtf-ek ' h e snatched you ( s g . ) '

T h e a r g u m e n t that t h e s e affixes are at a different level is the cyclic pattern of

stress a s s i g n m e n t . O n e fact w h i c h cyclic stress e x p l a i n s is t h e surface contrast in
(68) b e t w e e n bistratal [hatafna] ' h e s n a t c h e d u s ' a n d m o n o s t r a t a l [htafna] ' w e
s n a t c h e d ' . O n Level 1, Stress and A p o c o p e apply to the form [hataf] ' h e s n a t c h e d '
and to [hataf + n a ] ' w e s n a t c h e d ' . T h e s e c o n d form is directly m a p p e d o n t o the
p h o n e t i c o u t p u t [htafna]. T h e first form hataf c o n t a i n s a Level 2 suffix, so Stress
and A p o c o p e reapply on the L c y c l e . T h e stress is r e a s s i g n e d to the penult, but
2

d u e to the previously a s s i g n e d stress on the first syllable, A p o c o p e is b l o c k e d and

the p h o n e t i c form is [hatafna].

(68) [hatafna] [htafna]

'he snatched us' 'we snatched'
hataf hataf-na input to L j
hataf hatdf-na Stress
NA htdf-na Apocope
140 David Odden

hdtaf-na input to L 2

hatdf-na Stress
NA Apocope

T h e r e f o r e the object suffixes a n d n e g a t i v e s m u s t only b e available on Level 2,

a n d Stress a n d A p o c o p e are cyclic rules.
N o w w e turn to the other rule. T h e Level 2 affixes have a further peculiarity
seen in ( 6 9 ) ; namely, they lengthen a p r e c e d i n g vowel. T h u s htdftu+na becomes
htaftuuna a n d htdfna+kom b e c o m e s htafniekom. T h e n e g a t i v e suffix -s also in-
d u c e s B o u n d a r y L e n g t h e n i n g , s o ma hdtfu+s b e c o m e ma hatfuus.

(69) htaftuuna ' y o u (pi.) snatched u s ' (htdftu ' y o u (pi.) s n a t c h e d ' )
htafniekom ' w e s n a t c h e d y o u (pi.)' (htdfna ' w e s n a t c h e d ' )
ma hatfuus 'they didn't s n a t c h ' (hdtfu 'they s n a t c h e d ' )

A straightforward formulation of this l e n g t h e n i n g is possible within a n y theory:

any Level 2 suffix i n d u c e s L e n g t h e n i n g .

(70) B O U N D A R Y L E N G T H E N I N G (Level 2):

V - > V V / ] X

N o t e that this derived length attracts stress, so w e get hatfuus, not * hatfuus. There-
fore, B o u n d a r y L e n g t h e n i n g p r e c e d e s Stress o n Level 2.
N o w w e arrive at the p a r a d o x in ( 7 1 ) . T h e p r o b l e m is that verb stems w h i c h e n d
with a vowel, such as ?dra, m u s t l e n g t h e n their final vowel before a Level 2 suffix
as predicted, a n d this l e n g t h e n i n g m u s t take p l a c e o n Level 1 before L stress x

is assigned so that t h e initial vowel r e m a i n s unstressed a n d therefore u n d e r g o e s

Apocope.

(71) /?ara-0/ ?dra 'he read'

/?ara-0 + na/ ?rdana ' h e read u s '
/jara-0/ jdra 'it h a p p e n e d '
/jara-0 + 1-i/ jrda-li 'it h a p p e n e d to m e '

In LP, L e n g t h e n i n g w o u l d h a v e to b e a s s i g n e d to Level 2, since it is triggered only

b y Level 2 suffixes. T h e r e f o r e a s s i g n m e n t of Stress o n Level 1 m u s t p r e c e d e
L e n g t h e n i n g a n d should not b e sensitive to the output of L e n g t h e n i n g . B u t this
prediction is incorrect, as seen in t h e derivation (72).

(72) ?ara+& subject affixing

?dra+0 Stress
NA Apocope
[L ]
2

?dra+na object affixing

?draa+na Lengthening
Interaction in Lexical Phonology 141

?arda+na Stress
NA Apocope
*[?araana]

O n t h e Level 1 c y c l e , stress should b e a s s i g n e d t o t h e first vowel, j u s t as it is in

the unsuffixed form; o n Level 2 , t h e final vowel is l e n g t h e n e d a n d that vowel then
gets t h e stress, b u t t h e s u b o r d i n a t e d stress o n t h e initial vowel w o u l d incorrectly
b l o c k A p o c o p e . T h e correct derivation requires that L e n g t h e n i n g apply prior to
Level 1 Stress, g i v i n g ?araa+na a s t h e input to Stress, so that t h e p e n u l t i m a t e
v o w e l is stressed a n d t h e initial vowel is n e v e r stressed.

(73) ?ara-na output of m o r p h o l o g y

?araa-na (precyclic) B o u n d a r y L e n g t h e n i n g
?arda- Level 1 Stress
?rda- Level 1 A p o c o p e
?rdana Level 2 (nothing applies)

T h e s e facts c a n b e a c c o m m o d a t e d in a n y t h e o r y w h e r e all m o r p h e m e s a r e c o n -
c a t e n a t e d before a n y p h o n o l o g i c a l rules apply, p r o v i d i n g that w e allow s o m e
rules, a n d in particular, B o u n d a r y L e n g t h e n i n g , to a p p l y precyclically, so that
Level 2 suffixes c a n b e seen a n d t h u s trigger l e n g t h e n i n g before Level 1 p h o n o -
logical rules apply. In fact, w e c a n also a c c o u n t for this p r o b l e m if w e treat
L e n g t h e n i n g as a p r e c o m p i l e d rule, w h i c h w o u l d b e written as (74).

(74) LENGTHENING (precompiled);

V —> V : / ] [ F R A M E i]

T h e Level 1 p h o n o l o g y will g e n e r a t e b o t h ?ara a n d ?araa. T h e f o r m e r leaves

Level 1 as ?dra a n d t h e latter leaves Level 1 as ?rda.

(75) ?ara

[+F,] [-F,] Lengthening

?araa NA Stress
?arda ?dra Apocope
?rda NA

F r a m e 1 is defined a s in ( 7 6 ) .

(76) {Object 1
F r a m e 1: [VERB J [Negative J

S o w h e n a negative o r object affix is e n c o u n t e r e d in Level 2 , t h e variant ?rda is

selected.
142 David Odden

4. SUMMARY

T o c o n c l u d e , I h a v e investigated t w o manifestations of the q u e s t i o n of w h a t

i n f o r m a t i o n is available to e a c h c o m p o n e n t in the g r a m m a r and c o n s i d e r e d h o w
L P and N L P e x p l a i n the b e h a v i o r of linguistic s y s t e m s . It w a s s h o w n that plain
L P c a n n o t e x p l a i n the b e h a v i o r of s a n d h i rules in K i m a t u u m b i or precyclic
B o u n d a r y L e n g t h e n i n g in M a l t e s e , b u t e x t e n d i n g the t h e o r y w i t h p r e c o m p i l a t i o n
puts the theories o n an e q u a l footing descriptively. A r g u m e n t s h a v e b e e n given
against the c l a i m that m o r p h o l o g y a n d p h o n o l o g y interact in the w a y i m p l i e d by
the m o d e l (1). T h i s certainly d o e s n o t refute LP, since the m o d e l (1) m i g h t still b e
right, e v e n if there is n o e v i d e n c e for it in the r e a l m of i n f o r m a t i o n a c c e s s . A t this
stage, t h o u g h , it is n o t o b v i o u s w h e r e else w e c o u l d d e r i v e s u p p o r t for the inter-
active m o d e l .

NOTES

Sharon Hargus has suggested the possibility of eliminating the negative condition "not
1

preceded by H " in this rule by giving the morphemes in question an underlying initial H,
which deletes after a stem H. This possibility can be ruled out on two grounds. First, the
citation form of the relevant morphemes lacks the H, cf. mu-kjkdlaango 'in the frying pan':
in the citation form, there is no preceding H, so no reason to delete the putative underlying
H of /my-kjkalaango/. Second, ITI is subject to a syntactic condition that the toneless stem
which conditions ITI and the morpheme which undergoes the rule cannot be separated by
a righthand S-bracket. So, despite the fact that muundu 'person' has no H tones and the
following morpheme mu is one of the morphemes undergoing ITI, ITI does not apply in
the sentence naabik\tee nama \yaydapim\lwe na muyndu] my-kjkdlaango 'I put the meat
which was bought by the person in the frying pan' because the determinant and focus are
in different clauses. If we construe ITI as deletion of an underlying H, we must further
expand the environment for that rule, so that if the H is separated from the preceding word
by ] , then the initial H tone must also be deleted, even if there is no preceding H.
s

An analysis of reduplication in LP, viewed as an operation in the morphology, would

2

most likely have the same form as the NLP analysis. Since reduplication may (under resyl-
labification, as in reduplication of kwiita or kulwiitd) copy an object prefix or the infinitive
prefix ku, reduplication follows prefixation of these morphemes. The problem is identifying
the substring subject to reduplication. Given inputs such as REDUP•+ku[teleka] and RE-
DUP+ku[lu[teleka]] which reduplicate as kutelekateleka and kulutelekateleka, the morpho-
logical constituent which copies is the stem. But by the assumptions of LP, the internal
morphological structure of the verb is not recoverable. Morphological structure simply
does not suffice to identify the correct substring which reduplicates in the case of kulwii-
talwiita, where the structure which is copied includes nonstem material (Iw, the object
prefix). Therefore, in the LP account, some prosodic structure must form the basis for
Interaction in Lexical Phonology 143

identifying the structures subject to reduplication, and this structure must include the stem
syllables but may not include prefix syllables except when prefix syllables fuse with stem
syllables by phonological rules.
3
The reason that postlexical deletion, especially any deletion applying after phonologi-
cal rules, should be phonological is that the morphological component is, in the theory of
LP, part of the lexicon. Of course, one could expand LP in such a way that "postlexical
morphology" is not a theoretical anomaly, but such a move would seriously undermine the
motivation for distinguishing between the lexical and postlexical components.
4
It is beyond the scope of this article to present a complete theory of rules of this type,
but it is important to know something about how such rules are constrained. It is assumed
here that rules may either delete the segmental and prosodic material under a specified
prosodic constituent, or may delete the segmental material under the constituent leaving
prosodic structure intact. Rules of the former type are written as simply deleting the rele-
vant prosodic constituent. Rules of the latter type are written as deletion of the prosodic
element, with the additional annotation that "prosodic structure is preserved."
5
It is often assumed that i, not /, is the vowel which arises from default rules in Korean.
However, i has a restricted distribution in Korean: no morphemes ends in i except for the
demonstratives ki and ni. The relevant generalization is that i cannot be prepausal—the
demonstratives can never be prepausal. A similar constraint on i appears in Tigrinya; epen-
thesis inserts i in word-final position, but the vowel is realized phonetically as i in that
position. Chung (1991) provides other arguments that word-finally, / and not / arises by
default.
6
The forms denial and trial do not conform to this generalization; note, however, that
the stress-final generalization is falsified by burial.

REFERENCES

Aquilina, J. (1965). Teach Yourself Maltese. English Universities Press, London.

Aquilina, J., and Isserlin, B. (1981). A Survey of Contemporary Dialectal Maltese. B. S. J.
Isserlin, Leeds.
Anderson, S. (1975). On the interaction of phonological rules of various types. Journal of
Linguistics 11, 3 9 - 6 2 .
Booij, G. (1981). Rule ordering, rule application, and the organization of grammars. In
Phonologica 1980 (W. U. Dressier, ed.), pp. 4 5 - 5 6 . Institut fur Sprachwissenschaft,
Innsbruck.
Booij, G., and Rubach, J. (1984). Morphological and prosodic domains in lexical pho-
nology. Phonology Yearbook 1, 1 8 1 - 2 0 7 .
Booij, G., and Rubach, J. (1987). Postcyclic versus postlexical rules in lexical phonology.
Linguistic Inquiry 18, 1 - 4 4 .
Brame, M. (1970). Arabic Phonology. Doctoral dissertation, Massachusetts Institute of
Technology, Cambridge.
Brame, M. (1974). The cycle in phonology: Stress in Palestinian, Maltese, and Spanish.
Linguistic Inquiry 5, 3 9 - 6 0 .
144 David Odden

Chung, Y. H. (1991). The Lexical Tone System of North Kyungsang Korean. Doctoral dis-
sertation, Ohio State University, Columbus.
Clements, G. N. (1986). The problem of transfer in non-linear phonology. Cornell Working
Papers in Linguistics 7, 1 - 3 6 .
Halle, M., and Vergnaud, J.-R. (1987). An Essay on Stress. MIT Press, Cambridge, Mass.
Hayes, B. (1990). Precompiled phrasal phonology. In The Phonology-Syntax Connection
(S. Inkelas and D. Zee, eds.), pp. 8 5 - 1 0 8 . CSLI Publications and University of Chi-
cago Press, Chicago.
Hulst, H. van der (1984). Syllable Structure and Stress in Dutch. Foris, Dordrecht.
Hyman, L. (1990). Conceptual Issues in the Comparative Study of the Bantu Verb Stem.
Paper presented at the 21st Conference on African Linguistics, University of Georgia,
Athens.
Inkelas, S. (1989). Prosodic Constituency in the Lexicon. Doctoral dissertation, Stanford
University, Stanford, Calif.
Kaisse, E., and Shaw, P. (1985). On the theory of lexical phonology. Phonology Yearbook
2, 1-30.
Kiparsky, P. (1982). Lexical phonology and morphology. In Linguistics in the Morning
Calm (I. S. Yang, ed.), pp. 3 - 9 1 . Hanshin, Seoul.
Kisseberth, C , and Abasheikh, M. (1976). On the interaction of phonology and mor-
phology: a Chi-mwi: ni example. Studies in African Linguistics 7 , 3 1 — 110.
Marantz, A. (1982). Re reduplication. Linguistic Inquiry 1 3 , 4 3 5 - 4 8 2 .
Martin, J. (1988). Subtractive morphology as dissociation. Proceedings of the West Coast
Conference on Formal Linguistics 8, 2 2 9 - 2 4 0 .
Matthews, P. (1974). Morphology: An Introduction to the Theory of Word Formation.
Cambridge University Press, Cambridge.
McCarthy, J. (1979). Formal Problems in Semitic Phonology and Morphology. Doctoral
dissertation, Massachusetts Institute of Technology, Cambridge.
Mohanan, K. P. (1986). The Theory of Lexical Phonology. Reidel, Dordrecht.
Odden, D. (1987). Kimatuumbi phrasal phonology. Phonology Yearbook 4, 1 3 - 3 6 .
Odden, D. (1990a). Syntax, lexical rules and postlexical rules in Kimatuumbi. In The
Phonology-syntax Connection (S. Inkelas and D. Zee, eds.), pp. 2 5 9 - 2 7 7 . CSLI Pub-
lications and University of Chicago Press, Chicago.
Odden, D. (1990b). Phonology and its interaction with syntax and morphology. Studies in
the Linguistic Sciences 20, 6 9 - 1 0 8 .
Odden, D. (1992). Kimatuumbi Phonology and Morphology. Unpublished manuscript,
Ohio State University, Columbus.
Odden, D., and Odden, M. (1986). Ordered reduplication in Kihehe. Linguistic Inquiry 16,
497-503.
THE STRUCTURE OF THE SLAVE
(NORTHERN ATHABASKAN) VERB

KEREN D.RICE
Department of Linguistics
University of Toronto
Toronto, Ontario, Canada M5S 1A1

1. INTRODUCTION

T h e A t h a b a s k a n verb is c o m m o n l y d e s c r i b e d as consisting of a stem and a

n u m b e r of prefixes, both inflectional and derivational in nature, w h o s e o r d e r i n g
is u n p r e d i c t a b l e and requires a slot-and-filler, or t e m p l a t e , analysis. T h e verb in
A t h a b a s k a n l a n g u a g e s is u n u s u a l in several w a y s , p o s i n g p r o b l e m s for universal
theories of w o r d formation. First, it is generally o b s e r v e d that inflection stands
outside of derivation rather than inside of or interspersed with derivation (e.g.
A n d e r s o n , 1982, 1988; W i l l i a m s , 1981). In A t h a b a s k a n l a n g u a g e s , inflectional
and derivational e l e m e n t s are i n t e r m i n g l e d within the w o r d . S e c o n d , l a n g u a g e s
requiring slot-and-filler m o r p h o l o g y a p p e a r to b e u n u s u a l (e.g., M y e r s , 1987;
S p e a s , 1984), creating a third t y p e of m o r p h o l o g y distinct from c o n c a t e n a t i v e and
n o n c o n c a t e n a t i v e m o r p h o l o g y . A t h a b a s k a n l a n g u a g e s a p p e a r to require such m o r -
p h o l o g y to d e s c r i b e the order of verbal m o r p h e m e s . Third, m o r p h o l o g i c a l c o -
o c c u r r e n c e restrictions in l a n g u a g e s h a v e been a r g u e d to operate on a principle of
adjacency w h i c h requires that m o r p h o l o g i c a l s u b c a t e g o r i z a t i o n frames refer only
to adjacent e l e m e n t s (e.g., Allen, 1978; Lieber, 1 9 8 1 ; Siegel, 1978). A t h a b a s k a n
l a n g u a g e s s h o w c o o c c u r r e n c e restrictions b e t w e e n nonadjacent m o r p h e m e s that
are not e x p l a i n e d given a theory of slot-and-filler m o r p h o l o g y (e.g., Randoja,
1989; S p e a s , 1986, 1989; T h o m a s - F l i n d e r s , 1983). A n d fourth, d o m a i n s defined
by the m o r p h o l o g y and t h o s e required by the p h o n o l o g y h a v e b e e n argued to b e

145
Phonetics and Phonology, Volume 4 Copyright © 1993 by Academic Press, Inc.
Studies in Lexical Phonology All rights of reproduction in any form reserved.
146 Keren D. Rice

i s o m o r p h i c (e.g., Kiparsky, 1982; M o h a n a n , 1982, 1986). In A t h a b a s k a n lan-

g u a g e s the d o m a i n s r e q u i r e d for the p h o n o l o g y are n o n i s o m o r p h i c with those re-
quired b y the m o r p h o l o g y . In m o s t a c c o u n t s of the A t h a b a s k a n verb (e.g., Har-
1

g u s , 1988; Kari, 1976; R a n d o j a , 1989; R i c e , 1989; S p e a s , 1986) it is a s s u m e d that

affixes are diacritically m a r k e d for p h o n o l o g i c a l d o m a i n in o r d e r to a c c o u n t for
this discrepancy.
In this article, I w o u l d like to e x a m i n e the A t h a b a s k a n verb afresh. T h e m a i n
goal is to set forth a n e w p r o p o s a l r e g a r d i n g the verb that will a c c o u n t for the
properties noted a b o v e . I first give s o m e b a c k g r o u n d on the structure of w h a t is
traditionally called the verb w o r d a n d on the t r e a t m e n t of the verb in the literature.
I then turn to three q u e s t i o n s a b o u t the verb: Is it a m o r p h o l o g i c a l or a syntactic
construct; Is the o r d e r of m o r p h e m e s stipulated or predictable; and A r e the p h o -
nological rule d o m a i n s stipulated or d e r i v e d ? I e n d with a c o m p a r i s o n of the
theory that I p r o p o s e with a m o d e l of lexical p h o n o l o g y a n d m o r p h o l o g y , con-
c l u d i n g that w h i l e I h a v e a r g u e d against lexical p h o n o l o g y and m o r p h o l o g y for
the v e r b structure, the v e r b nevertheless p r o v i d e s s u p p o r t for a n u m b e r of prin-
ciples of the theory.

2. T H E S T R U C T U R E O F T H E S L A V E V E R B

T h e verb in an A t h a b a s k a n l a n g u a g e is generally characterized as a template,

c o n s i s t i n g of a string of fixed-order position classes a n d m o r p h e m e s that are
m a r k e d lexically for the position that they fill in the t e m p l a t e . In addition,
b o u n d a r y types are associated with the different m o r p h e m e positions to a c c o u n t
for their p h o n o l o g i c a l p r o p e r t i e s . T h e t e m p l a t e p r o p o s e d for Slave ([slevi]), an
A t h a b a s k a n l a n g u a g e of C a n a d a , in (1) is slightly adapted from that p r o p o s e d b y
Rice (1989).
(1) p r e v e r b # distributive # iterative # incorporate # direct object %
D D D D I
deictic subject % t h e m e - d e r i v a t i o n - a s p e c t 4- c o n j u g a t i o n / m o d e + subject
I D/I D I I
= voice + s t e m
Several o b s e r v a t i o n s a b o u t the v e r b are in o r d e r c o n c e r n i n g the content of the
slots, the identification of m o r p h e m e s as inflectional (I) or derivational (D), the
c o o c c u r r e n c e restrictions b e t w e e n positions in the verb, and the use of p h o n o l o g i -
cal b o u n d a r i e s .

2 . 1 . T h e C o n t e n t of t h e T e m p l a t i c P o s i t i o n s

In this section I identify briefly e a c h of the m o r p h e m e positions, starting at the

left of the verb, with the p r e v e r b s .
The Structure of the Slave Verb 147

PREVERBS, traditionally called incorporated postpositions and a d v e r b s , repre-

sent o b l i q u e relations or m a n n e r . S e e Kari ( 1 9 8 9 , 1990) and R i c e ( 1 9 9 1 b ) for
details. T y p i c a l m e a n i n g s of p r e v e r b s include ' a r o u n d , away, u p onto, out of,
across, to a point, into fire, into air, in half, to pieces, e x c e s s ' . W h i l e the m e a n i n g s
of m a n y of the m o r p h e m e s in this position are transparent, with s o m e the m e a n i n g
is defined only in c o m b i n a t i o n with the stem.
T h e DISTRIBUTIVE m o r p h e m e is aspectual, m a r k i n g distributivity: e a c h one
separately. It can h a v e scope over the agent, t h e m e , action, or location.
T h e ITERATIVE m o r p h e m e , a n o t h e r aspectual marker, indicates that an action is
habitual or repeated, d e p e n d i n g on other m o r p h e m e s p r e s e n t within the verb.
INCORPORATES are of t w o t y p e s , n o u n s that are internal a r g u m e n t s (Rice,
1991c) and s t e m s with an adverbial function. T h e m e a n i n g of the verb with an
incorporate differs in systematic w a y s from the m e a n i n g of the verb w i t h o u t the
incorporate. S e e R i c e (1989) and A x e l r o d ( 1 9 9 0 ) for discussion.
D I R E C T OBJECT m o r p h e m e s m a r k the p e r s o n and n u m b e r of the direct object.
DEICTIC SUBJECT, or third p e r s o n subject, position includes t w o m o r p h e m e s , k-
'third p e r s o n plural h u m a n subject' a n d ts'- 'unspecified s u b j e c t ' . It h a s b e e n ar-
g u e d for Navajo, a n o t h e r A t h a b a s k a n l a n g u a g e , that the c o g n a t e m o r p h e m e s rep-
resent a t y p e of object m a r k i n g a n d can b e c o l l a p s e d with the object m a r k e r s
( S p e a s , 1991). I a s s u m e that such an analysis is possible for Slave; this question
requires further close attention.
T H E M E - D E R I V A T I O N - A S P E C T is a g r a b - b a g position in w h i c h three t y p e s of ele-
m e n t s can be identified. First, p r o d u c t i v e aspectual m o r p h e m e s are found. T h e s e
include d ' i n c e p t i v e ' , n ' t e r m i n a t i v e , c o m p l e t i v e ' , i ' t r a n s i t i o n a l ' , and u ' c o n a -
t i v e ' . I will call these m o r p h e m e s SECONDARY ASPECT. S e c o n d , there are GENDER
2

m o r p h e m e s , generally called derivational prefixes. T h e s e include d 'fire', d ' b e n e -

f a c t i v e ' , d ' b y m o u t h ' , n ' m i n d , f e e l i n g ' , n ' w a t e r ' , y 'dual subject'. T h e s e m o r -
p h e m e s certainly m a r k g e n d e r historically. S o m e are e x t r e m e l y p r o d u c t i v e ; for
instance, the m o r p h e m e d ' b y m o u t h ' o c c u r s in a w i d e r a n g e of verbs h a v i n g to
d o with n o i s e ; e x a m p l e s include ' w h i s t l e ' , ' s n o r e ' , ' b u r p ' , ' s i t ' , ' b a r k ' , ' c o u g h ' ,
' s q u e a k ' , ' a s k ' , ' h o n k ' ( g o o s e call), ' w h i n e , f u s s ' , ' a r g u e ' , 'defend (help with
w o r d s ) ' , ' w a l k l a u g h i n g , crying, e t c . ' , 'joke (tease with w o r d s ) ' , ' w i n with w o r d s ' .
O t h e r s o c c u r only in restricted c a s e s . For e x a m p l e , the m o r p h e m e n o c c u r s in
verbs m e a n i n g ' h a n d l e unspecified object (water) on object, or w a s h ' and ' h a n d l e
in w a t e r ' . W i t h o u t n, the m e a n i n g of the verb w o u l d not include the c o n c e p t of
water. T h e prefix y m u s t o c c u r in certain verbs with a dual subject (e.g. 'dual
a r r i v e ' ) ; however, it is not generally found even w h e n the stem requires a dual
subject. Third, there are m o r p h e m e s w h i c h always o c c u r with a particular verb
stem; these m o r p h e m e s are generally t e r m e d THEMATIC in the A t h a b a s k a n litera-
ture. T h e y are part of the u n d e r l y i n g representation of the lexical item. For in-
stance, the basic lexical e n t r y for the verb ' h a n d l e singular object ( u n c o n t r o l l e d ) '
includes the prefix y, with every derivative b a s e d on this lexical e n t r y requiring
this m o r p h e m e . I call both the g e n d e r and the t h e m a t i c m o r p h e m e s GENDER in this
148 Keren D. Rice

article. W h i l e in general g e n d e r p r e c e d e s s e c o n d a r y aspect, the o r d e r i n g m a y b e

overridden by p h o n o l o g i c a l constraints. In Slave, the o r d e r i n g of these mor-
p h e m e s is: u, y gender, d, n, y s e c o n d a r y aspect, i. S e e R i c e ( 1 9 8 9 ) o n Slave;
H a r g u s (1988) on S e k a n i ; Kari ( 1 9 8 9 ) on A h t n a ; and S p e a s ( 1 9 8 6 ) , W r i g h t
( 1 9 8 6 ) , and M c D o n o u g h (1990) o n Navajo.
C O N J U G A T I O N and MODE will b e c o n s i d e r e d together. M o d e is a m i s n o m e r
w h i c h is used for c o n v e n i e n c e in A t h a b a s k a n literature. T h r e e m o r p h e m e s are
found in this position in Slave, 0 imperfective, n perfective, and ghu o p t a t i v e . 3

T h e first t w o m a r k aspect and the third indicates m o d e , or in s o m e cases a r e m o t e

future tense. A m o r p h e m e in this position m u s t o c c u r in every verb. I call these
m o r p h e m e s PRIMARY ASPECT. T h e m o r p h e m e s called conjugation h a v e the forms
0 , n w, and y in S l a v e . E a c h verb requires a particular conjugation pattern or set
y
4

of conjugation m a r k e r s for the imperfective, perfective, and optative. T h e conju-

gation pattern is d e t e r m i n e d in t w o different w a y s . First, an u n d e r l y i n g lexical
entry has a conjugation pattern associated with it, largely d e t e r m i n a b l e by s e m a n -
tics. F o r instance, verbs of m o t i o n require the conjugation pattern n imperfective,
n perfective, n optative, w h i l e those i n v o l v i n g sustained actions over time require
0 imperfective, y perfective, 0 optative. S e c o n d , p r e v e r b a n d aspectual m o r -
p h e m e s are conjugation c h o o s e r s a n d d e t e r m i n e the c h o i c e of conjugation pattern
for a particular verb. F o r e x a m p l e , the a d v e r b dah ' u p o n t o a horizontal surface'
requires w imperfective, w perfective, w optative conjugation m a r k i n g ; the adverb
kd 'out from i n s i d e ' requires 0 imperfective, y perfective, 0 optative conjugation
m a r k i n g ; and the s e c o n d a r y aspectual m a r k e r d ' i n c e p t i v e ' requires 0 imperfec-
tive, w perfective, 0 optative conjugation m a r k i n g . T h e c u s t o m a r y and distribu-
tive also select conjugation p a t t e r n s .
T h e SUBJECT m a r k e r s indicate the p e r s o n and n u m b e r of the subject. T h e s e
include first p e r s o n singular, s e c o n d p e r s o n singular, first p e r s o n plural, s e c o n d
person plural, and third p e r s o n ( n u m b e r is not d e t e r m i n e d in this position).
T h e c o n j u g a t i o n - p r i m a r y a s p e c t - s u b j e c t p o r t i o n s of the verb c o m b i n e in w a y s
that are not a l w a y s predictable. F o r instance, the first p e r s o n singular subject h a s
the form h e x c e p t in the perfective of 0 a n d h voice e l e m e n t verbs, w h e r e it is /.
W h i l e it is p o s s i b l e to assign the m o r p h e m e h the m e a n i n g 'first p e r s o n singular
subject', the m o r p h e m e / i n c l u d e s m o r e than o n e m e a n i n g , n a m e l y first person
singular subject and perfective p r i m a r y aspect. T h e optative is predictably u or
wo- e x c e p t w h e n the conjugation m a r k e r is n or w, w h e n it is 'wo or wo (the acute
accent indicates that a high tone falls on the vowel of the p r e c e d i n g syllable).
O t h e r m o r p h e m e s s h o w similar patterns. W h i l e the s e c o n d p e r s o n singular is regu-
larly nasalization in certain e n v i r o n m e n t s , in 0 and h voice e l e m e n t perfectives it
has the form ne in these e n v i r o n m e n t s . T h e third p e r s o n exhibits similar allomor-
phy, with an unusual form in the perfective of 0 and h voice e l e m e n t verbs, n and
w conjugation optatives also display u n e x p e c t e d patterns. T h e n o n s y s t e m a t i c
c o m b i n a t i o n s of c o n j u g a t i o n - p r i m a r y a s p e c t - s u b j e c t suggest that in at least s o m e
cases this stretch of the verb should b e treated as a single unit, or p o r t m a n t e a u
The Structure of the Slave Verb 149

m o r p h , with c o m p l e x m e a n i n g , as p r o p o s e d , for instance, b y A n d e r s o n ( 1 9 8 2 ) for

G e o r g i a n a n d b y W i l l i a m s ( 1 9 8 1 ) for Latin.
T h e so-called classifiers, h e r e labeled ' v o i c e ' , m a r k v a l e n c e / v o i c e a n d transi-
tivity, c h a n g i n g the a r g u m e n t structure required by a verb. F o r instance, the h (* t)
voice e l e m e n t a d d s an agent a r g u m e n t , creating a transitive/causative, a n d the d
voice e l e m e n t c a u s e s the loss of the agent a r g u m e n t , creating a passive. In addi-
tion, a voice e l e m e n t m a y b e part of the lexical entry of a verb as the form of the
voice e l e m e n t is not always predictable from the a r g u m e n t structure of the verb.
Finally, the stem itself is actually c o m p l e x , consisting of a root a n d an aspectual
suffix.
In (2), the verb t e m p l a t e , as redefined b a s e d on the a b o v e discussion, is s h o w n .

(2) p r e v e r b # distributive # iterative # i n c o r p o r a t e # object a g r e e m e n t % third

p e r s o n subject % g e n d e r + s e c o n d a r y aspect + p r i m a r y aspect + subject
a g r e e m e n t = voice + verb root + aspect

2.2. Inflection o r D e r i v a t i o n ?

I l o o k at the inflection/derivation distinction briefly h e r e a n d return to it in m o r e

d e p t h later. I h a v e labeled m o s t m o r p h e m e s in (1) as inflectional (I) or derivational
(D) following traditional A t h a b a s k a n practice (e.g., Kari, 1979, 1990; Sapir and
Hoijer, 1969). T h e p r o n o m i n a l a g r e e m e n t m a r k e r s (object of postposition, direct
object, deictic subject, subject) are treated as inflectional in the literature. T h e
status of aspectual a n d g e n d e r material is not clear; I return to this q u e s t i o n below.
O n the surface at least, the A t h a b a s k a n v e r b is highly m a r k e d , with inflectional
affixation a p p e a r i n g linearly inside of derivational affixation. 5

2.3. Cooccurrence Restrictions

C o o c c u r r e n c e restrictions b e t w e e n positions within the A t h a b a s k a n verb are

frequently found. I m e n t i o n s o m e briefly; see H a r g u s ( 1 9 8 8 ) on S e k a n i , R i c e
( 1 9 8 9 ) on Slave, S p e a s (1986) on Navajo, and R a n d o j a ( 1 9 8 9 ) o n B e a v e r for m o r e
e x t e n s i v e discussion.
A s d i s c u s s e d earlier, conjugation m o r p h e m e s d o not o c c u r freely but are d e p e n -
dent on other information. T h e y can be d e t e r m i n e d by the u n d e r l y i n g s e m a n t i c s
of the verb or by preverbs, s e c o n d a r y aspect m o r p h e m e s , the distributive, and the
iterative. S o m e c o o c c u r r e n c e restrictions hold b e t w e e n the voice e l e m e n t s and
a n a p h o r i c p r o n o u n s , with the reflexive a n d reciprocal generally r e q u i r i n g the d
voice e l e m e n t . In addition, the iterative usually m u s t o c c u r with the d voice ele-
m e n t , at least in intransitive verbs.
O t h e r m o r p h e m e s are d i s c o n t i n u o u s ; for instance, certain preverbs require a
g e n d e r m o r p h e m e . T h e p r e v e r b di 'fire' o c c u r s only with the g e n d e r m o r p h e m e
d 'fire'.
150 Keren D. Rice

2.4. Phonological Boundaries

P h o n o l o g i c a l rule d o m a i n s are indicated by b o u n d a r y s y m b o l s in (1) and (2).

T h e verb divides into four major d o m a i n s with respect to p h o n o l o g i c a l rules (e.g.,
H a r g u s , 1988; Kari, 1975, 1976; Li, 1946; Randoja, 1989; R i c e , 1989). T h e sym-
bols #, %, + , and = delineate these d o m a i n s . # separates the so-called disjunct
m o r p h e m e s , w h i c h interact little with e a c h other or with the rest of the verb. It is
generally r e g a r d e d as the strongest of the w o r d - i n t e r n a l b o u n d a r i e s . + separates
the m o r p h e m e s called conjunct prefixes, w h i c h enter closely into p h o n o l o g i c a l
c o m b i n a t i o n . T h e % b o u n d a r y is u s e d for the direct objects and deictic subjects.
W h i l e these m o r p h e m e s often pattern with the conjunct m o r p h e m e s , in s o m e w a y s
they pattern with the disjunct m o r p h e m e s , t h u s b e i n g i n t e r m e d i a t e b e t w e e n the
two. Finally, the s y m b o l = separates the voice e l e m e n t a n d stem from the conjunct
m o r p h e m e s ( H a r g u s , 1988; Kari, 1976; M c D o n o u g h , 1990; R i c e , 1989). T h e
b o u n d a r i e s thus define p h o n o l o g i c a l rule d o m a i n s , with the conjunct m o r p h e m e s
m o r e closely b o u n d to the verb stem than the disjunct m o r p h e m e s .

2.5. T h e D e r i v a t i o n of t h e V e r b W o r d

W h i l e (1) represents the linear o r d e r i n g of m o r p h e m e s , the verb has often b e e n

treated as c o m p o s e d of several distinct levels m o r p h o l o g i c a l l y (e.g., H a r g u s ,
1988; Kari, 1976, 1990; Li, 1946; R a n d o j a , 1989; R i c e , 1989; Sapir a n d Hoijer,
1969). T h e basic lexical entry, k n o w n as the VERB T H E M E , is the structure that is
entered in the lexicon. It obligatorily includes a root a n d a voice e l e m e n t , w h i c h
m a y b e null. In addition, the t h e m e includes any m o r p h e m e s that m u s t b e present
in all forms of the verb. T h e s e i n c l u d e p r e v e r b s and t h e m a t i c (gender) m o r p h e m e s .
S o m e e x a m p l e s of t h e m e s are given in (3).

(3) a. d-dg 'drink'

voice-stem
b. ya-h-ti 'preach, bark'
pre verb-voice-stem
c. d-l-we ' s g . fall'
gender-voice-stem

In the first stage of w o r d formation, a level called the verb b a s e is formed. At

this level, derivational affixes are a d d e d to the verb t h e m e . T h e s e include prever-
bal, gender, a n d s e c o n d a r y aspectual i t e m s . S o m e s a m p l e bases f o r m e d on the
t h e m e s in (3) are given in (4). Verb w o r d s are also s h o w n ; see the discussion
below.

(4) a. verb t h e m e : d-dg ' d r i n k (object)'

verb w o r d : he do 's/he drinks
(object)'
The Structure of the Slave Verb 151

verb base: te-d-d-do ' d r i n k to e x c e s s '

preverb-sec asp-voice-stem
verb w o r d : tedehdo 's/he d r a n k to
excess'
b. verb t h e m e : ya-h-ti 'preach, bark'
preverb-voice-stem
verb w o r d : yahti 's/he preaches,
barks'
verb b a s e : xa-ya-d-d-h-ti 'pray'
preverb-preverb-gender-
voice-voice-stem
verb word: xayadeti 's/he prays'
verb b a s e : k' a-ya-?e-h-ti 'interpret'
preverb-preverb-DO- voice-stem
verb w o r d : k'aya?ehti ' s / h e interprets'
c. verb t h e m e : d-l-we 'fall'
gender-voice-stem
verb base: kd-d-d-l-we 'fall o u t '
preverb-gender-asp-voice-stem
verb w o r d : kddedehwe ' s h e / h e / i t fell o u t '
verb b a s e : teh-d-l-we 'fall into w a t e r '
preverb-gender-voice-stem
verb w o r d : tedgwe ' s / h e fell into
water'
A t the final stage of w o r d formation, the verb w o r d is p r o d u c e d . A t this level,
inflectional affixes are a d d e d (e.g., subject, object m a r k e r s , p e r h a p s conjugation
a n d p r i m a r y aspect) and the formation of the verb w o r d is c o m p l e t e . ( S e e Kari,
1979, 1990, 1992, for a far m o r e highly articulated m o d e l of w o r d formation
in A h t n a . )
T h i s m o d e l of w o r d f o r m a t i o n i n c l u d e s three levels: v e r b t h e m e (basic lexical
entry), verb b a s e (verb m i n u s inflection), a n d verb w o r d (inflected verb). S u c h a
m o d e l of w o r d formation is p r o p o s e d in o r d e r to a c c o u n t for p a r a d i g m a t i c p r o p -
erties of the A t h a b a s k a n verb. It results in m a k i n g the A t h a b a s k a n v e r b far m o r e
like verbs in o t h e r l a n g u a g e s as well, w i t h derivational m o r p h o l o g y p r e c e d i n g
inflectional m o r p h o l o g y derivationally.
G i v e n this m o d e l of w o r d f o r m a t i o n a n d the p h o n o l o g i c a l b o u n d a r i e s d i s c u s s e d
in Section 2.4, it is o b v i o u s that w o r d formation a n d p h o n o l o g y d o not take p l a c e
in t a n d e m , as t h e p h o n o l o g i c a l d o m a i n s are n o t defined until the v e r b w o r d is
f o r m e d (see H a r g u s , 1986, for s o m e c o m m e n t s ) . B e c a u s e of this lack of i s o m o r -
p h i s m b e t w e e n w o r d f o r m a t i o n a n d p h o n o l o g y , the A t h a b a s k a n literature r e c o g -
nizes t w o m o d e l s of the verb. O n e (verb t h e m e , v e r b b a s e , verb w o r d ) a c c o u n t s
for the m o r p h o l o g i c a l structure of the verb, a l l o w i n g for derivation to p r e c e d e
152 Keren D. Rice

inflection. T h e s e c o n d (disjunct, conjunct prefixes, stem) a c c o u n t s for the p h o n o -

logical structure of the verb. T h i s s e c o n d t y p e of structure is c o d e d as b o u n d a r y
s y m b o l s (or s o m e other diacritic) on the lexical entry of the affixes.

3. P R E V I O U S T R E A T M E N T S O F W O R D F O R M A T I O N

M a n y a c c o u n t s of the A t h a b a s k a n verb h a v e b e e n p r o p o s e d , falling into t w o

basic c a t e g o r i e s . O n e , r e p r e s e n t e d best by the w o r k of H a r g u s ( 1 9 8 8 , etc.), as-
s u m e s the lexical p h o n o l o g y framework, a r g u i n g that a templatic structure such
as that in (1) represents the structure of the verb m o r p h o s y n t a c t i c a l l y as well as
phonologically. T h e level at w h i c h a particular m o r p h e m e is a d d e d to the verb and
the o r d e r in w h i c h the m o r p h e m e s are a d d e d is stipulated in the g r a m m a r of the
l a n g u a g e . C o o c c u r r e n c e restrictions are r e p r e s e n t e d t h r o u g h c o m p l e x lexical en-
tries. In a m o d e l w h e r e p h o n o l o g y a n d m o r p h o l o g y p r o c e e d in t a n d e m , the fact
that inflection is n o r m a l l y o u t s i d e of derivation is accidental. N o n a d j a c e n t c o o c -
c u r r e n c e restrictions within the verb are m e t b y diacritic m a r k i n g s on affixes. M o r -
phosyntactically, then, the A t h a b a s k a n verb m u s t b e r e g a r d e d as u n u s u a l . T h i s
m o d e l leaves u n e x p l a i n e d certain facts about the p h o n o l o g y ; for instance, the p h o -
n o l o g y of s t e m s and of disjunct m o r p h e m e s is similar w h i l e that of the conjunct
m o r p h e m e s is different. S i n c e s t e m s a n d disjunct m o r p h e m e s a p p e a r on n o n -
adjacent levels of the lexical p h o n o l o g y , this similarity in their p h o n o l o g y is
surprising.
T h e s e c o n d t y p e of analysis is r e p r e s e n t e d by w o r k of authors such as Kari,
S p e a s , W r i g h t , and R a n d o j a . T h e s e authors p r o p o s e m o d e l s in m a n y w a y s similar
to that d e s c r i b e d in S e c t i o n 2 . 5 , a r g u i n g that the m o r p h o s y n t a c t i c structure of the
verb is not i s o m o r p h i c to the linear o r d e r i n g of affixes. S p e a s ( 1 9 8 4 , 1986, 1990),
e c h o e d b y R a n d o j a ( 1 9 8 9 ) , p o i n t s out that in the interest of m a i n t a i n i n g a restric-
tive theory of m o r p h o l o g y a n d w o r d formation, n o alternative solution to the m o r -
p h o l o g y p r o b l e m is available.
S p e a s , Kari, W r i g h t , a n d R a n d o j a i m p l e m e n t this basic m o d e l in very different
w a y s . W h i l e S p e a s ( 1 9 9 0 ) a n d K a r i (1979) a c c o u n t elegantly for the m o r p h o s y n -
tactic side of verb structure, they offer n o systematic a c c o u n t of the p h o n o l o g y
(but see Kari, 1990, for discussion of A h t n a p h o n o l o g y ) .
W r i g h t ( 1 9 8 3 , 1986) and, particularly, R a n d o j a ( 1 9 8 9 ) attempt to integrate the
p h o n o l o g y and the m o r p h o l o g y sides into a c o h e r e n t m o d e l . Randoja, in an ex-
a m i n a t i o n of verb structure in Beaver, p r o p o s e s that the basic structure of the
verb is e n c o d e d in a t h e m a t i c t e m p l a t e w h i c h represents both the lexical entry of
the verb and the division of the verb into p h o n o l o g i c a l rule d o m a i n s . S h e argues
that derivational affixation p r e c e d e s inflectional affixation, with m o r p h e m e s
a c h i e v i n g their surface position b y m a p p i n g to the thematic t e m p l a t e . T h e order
The Structure of the Slave Verb 153

of m a p p i n g follows from universal principles (inflection outside derivation, adja-

c e n c y ) , with the slot in the t e m p l a t e to w h i c h the affix m a p s being stipulated in
the lexical entry of the affix. T h u s , e a c h affix is m a r k e d for the d o m a i n to w h i c h
it is m a p p e d . It is this basic type of m o d e l that I w o u l d like to p u r s u e in this article.
M y m o d e l differs from that p r o p o s e d by R a n d o j a in that I argue, following S p e a s
( 1 9 9 0 , 1991), that the basic verb structure is phrasal in nature a n d the w o r d l i k e
nature of the verb is a p h o n o l o g i c a l characteristic only. I p r o p o s e to eliminate the
l a n g u a g e - p a r t i c u l a r t e m p l a t e in favor of a structure w h e r e the o r d e r i n g of m o r -
p h e m e s within the verb w o r d follows from p r o p e r t i e s of universal g r a m m a r . Fi-
nally, I p r o p o s e to e l i m i n a t e diacritic m a r k i n g of p h o n o l o g i c a l levels, p r o p o s i n g
that the p h o n o l o g i c a l levels are derived by a l g o r i t h m .
B e f o r e t u r n i n g to m y p r o p o s a l , I s u m m a r i z e s o m e of t h e p r o b l e m s that m u s t b e
dealt with in a c c o u n t i n g for the structure of the Slave verb.

1. W h i l e inflection generally o c c u r s o u t s i d e of derivation, in Slave inflection is

linearly inside of derivation.
2. W h i l e the o r d e r of verbal affixes generally reflects scopal properties or rele-
v a n c e to the verb ( B y b e e , 1985), in the Slave verb the order of m o r p h e m e s
a p p e a r s to b e r a n d o m .
3. W h i l e m o r p h o l o g i c a l s u b c a t e g o r i z a t i o n frames generally d o not state d e p e n -
d e n c i e s b e t w e e n nonadjacent items, in Slave d e p e n d e n c i e s exist b e t w e e n
nonadjacent items.
4 . W h i l e lexical p h o n o l o g y p r o p o s e s that w o r d formation and p h o n o l o g y p r o -
ceed in t a n d e m , in Slave the linear o r d e r i n g of affixes represents p h o n o l o g i -
cal rule d o m a i n s b u t d o e s not readily a c c o u n t for w o r d formation.

T h e m o d e l that I p r o p o s e a c c o u n t s for these properties as follows.

1. T h e conjunct m o r p h e m e s are functional items w h i c h head phrasal projec-

tions. F o l l o w i n g w o r k by Pollock ( 1 9 8 9 ) , C h o m s k y ( 1 9 8 8 ) , S p e a s ( 1 9 9 0 ,
1991), and others, they o c c u r as syntactic objects outside of the verb p h r a s e ,
a n d thus outside of any e l e m e n t s that m i g h t be c o n s t r u e d to b e derivational.
Inflection thus is outside of d e r i v a t i o n . 6

2. T h e o r d e r i n g of m o r p h e m e s within the conjunct portion of the verb reflects

the o r d e r i n g e x p e c t e d b a s e d on scopal properties or relevance to the verb.
Position classes are u n n e c e s s a r y as o r d e r i n g is b a s e d on general principles
related to s c o p e .
3. N o n a d j a c e n t d e p e n d e n c i e s are a c c o u n t e d for by treating d i s c o n t i n u o u s con-
stituents as syntactic objects, in the sense of D i S c i u l l o a n d W i l l i a m s ( 1 9 8 8 ) :
they are phrasal units that are listed in the lexicon. M o r p h o l o g i c a l violations
of adjacency thus are only superficial.
4 . P h o n o l o g i c a l d o m a i n s are d e t e r m i n e d by m a p p i n g of syntactic structure
o n t o p r o s o d i c structure. P h o n o l o g i c a l structure is thus directly related to
syntactic structure.
154 Keren D. Rice

4. T H E S Y N T A X O F T H E S L A V E V E R B

C o n v e n t i o n has it that in p o l y s y n t h e t i e l a n g u a g e s the word takes over the func-

tion of the sentence. In generative g r a m m a r , o n e a p p r o a c h to the p r o b l e m of poly-
synthetic l a n g u a g e s is to c l a i m that a w o r d in such l a n g u a g e s actually is a sentence
(e.g., A n d e r s o n , 1982, 1988; Baker, 1988; C h o m s k y , 1988; D i S c i u l l o and W i l -
liams, 1988; S p e a s , 1990). In this section, I e x p l o r e the possibility that the Atha-
b a s k a n verb is phrasally rather than lexically formed. I a r g u e that the traditionally
t e r m e d disjunct m o r p h e m e s are lexical items and the conjunct m o r p h e m e s func-
tional items. F o l l o w i n g Pollock ( 1 9 8 9 ) and C h o m s k y (1988) a m o n g others, I sug-
gest that the functional i t e m s project i n d e p e n d e n t positions in the syntax, h e a d i n g
phrasal p r o j e c t i o n s .
7

4.1. T h e F u n c t i o n a l N a t u r e of t h e C o n j u n c t M o r p h e m e s

A s discussed in Section 2.2, d e t e r m i n i n g w h i c h m o r p h e m e s are inflectional or

functional in Slave is not entirely straightforward. In this section, I a r g u e that the
traditional disjunct m o r p h e m e s are lexical items and the conjunct m o r p h e m e s
functional i t e m s . In m a k i n g this a r g u m e n t , I a s s u m e that the lexical/functional
contrast can b e established in several w a y s . First, functional items are syntactically
active w h i l e lexical items are n o t . In d e t e r m i n i n g w h i c h items are syntactically
8

active, I follow A n d e r s o n ( 1 9 8 2 , 1988), w h o argues that syntactically active items

s h o w configurational, a g r e e m e n t , inherent (e.g., g e n d e r ) , a n d p h r a s a l p r o p e r t i e s .
Second, functional items are obligatory, b e i n g m a r k e d e a c h t i m e a c a t e g o r y to
w h i c h they apply a p p e a r s ( A n d e r s o n , 1982; B y b e e , 1 9 8 5 : 2 7 ) . L e x i c a l items, on
the other h a n d , are not obligatory in this sense. Third, functional items can c o m -
b i n e to form p o r t m a n t e a u m o r p h e m e s with m o r e than o n e e l e m e n t of m e a n i n g in
a single entry. L e x i c a l i t e m s d o not c o m b i n e with e a c h other or with functional
items. S e e A n d e r s o n ( 1 9 8 8 ) for discussion. Finally, functional classes are gener-
ally closed classes while lexical classes tend to b e open.
G i v e n these criteria, the following m o r p h e m e s in Slave can b e c o n s i d e r e d to b e
functional.

1. P R O N O M I N A L AGREEMENT. S a x o n (1986) argues that the p r o n o m i n a l ele-

m e n t s of D o g r i b , a l a n g u a g e closely related to Slave, represent a g r e e m e n t b e t w e e n
a n o u n p h r a s e and the clausal e l e m e n t on w h i c h it d e p e n d s syntactically. S h e ad-
ditionally argues that p r o n o m i n a l inflection is obligatory, even w h e n a specified
n o u n is present, and that n o u n s are a r g u m e n t s rather than adjuncts. G i v e n these
characteristics, she c o n c l u d e s that the p r o n o m i n a l e l e m e n t s represent a g r e e m e n t
and thus are functional. A similar a r g u m e n t is m a d e by S p e a s (1989) for m o s t
p r o n o m i n a l e l e m e n t s in Navajo. G i v e n the similarities b e t w e e n D o g r i b and Slave,
I believe that Saxon's a r g u m e n t s b a s e d on relevance to the syntax and obligatori-
The Structure of the Slave Verb 155

ness can b e e x t e n d e d to Slave as well. In addition, as discussed in Section 2 . 1 , the

subject m a r k e r s at least conflate with aspect m o r p h e m e s to yield p o r t m a n t e a u
forms, a n o t h e r d i a g n o s t i c of their functional nature.
2 . PRIMARY ASPECT. A n d e r s o n ( 1 9 8 2 ) points out that t e n s e / a s p e c t play an im-
portant role syntactically, so o n e m i g h t e x p e c t these m o r p h e m e s to b e of syntactic
relevance. T h e r e are positive r e a s o n s in Slave to c o n s i d e r the p r i m a r y aspect m o r -
p h e m e s as inflectional. First, c o o c c u r r e n c e restrictions are found b e t w e e n p r i m a r y
aspect a n d aspectual c a t e g o r y - a s s i g n i n g m o r p h e m e s w h i c h follow the v e r b root:
if the verb stem is optative, then the m o r p h e m e optative m u s t b e p r e s e n t in pri-
m a r y aspect position, a n d so on. S e c o n d , c o o c c u r r e n c e restrictions are found with
other postverbal m a t e r i a l . For instance, the imperfective v e r b c o m b i n e s w i t h the
postverbal particle gha to yield a future. T h e optative c o m b i n e s with the postver-
bal sand to g i v e a p r o h i b i t i v e m e a n i n g . If the postverbal particles are r e g a r d e d as
higher p r e d i c a t e s (see R i c e , 1 9 8 9 ) , it is p o s s i b l e to view this as selection of pri-
m a r y aspect b y a h i g h e r verb, a configurational property. Third, p r i m a r y aspect is
an obligatory p a r t of the verb, again an indication that it is functional. Finally, t h e
subject and p r i m a r y aspect m o r p h e m e s c o m b i n e to form p o r t m a n t e a u m o r p h s ,
suggesting that e a c h of the c o m p o n e n t s is functional.
3 . C O N J U G A T I O N . T W O facts suggest the functional nature of the conjugation
m o r p h e m e s . First, t h e s e m o r p h e m e s are obligatory. S e c o n d , they c o m b i n e with
p r i m a r y aspect a n d subject in u n p r e d i c t a b l e w a y s , suggesting that a single m o r -
p h e m e m a y i n c l u d e the m e a n i n g conjugation, p r i m a r y aspect, a n d subject. A g a i n ,
since functional m o r p h e m e s form p o r t m a n t e a u s only with other functional items,
this suggests that these m o r p h e m e s m u s t b e functional.
4 . S E C O N D A R Y ASPECT. T h e s e m o r p h e m e s s h o w c o o c c u r r e n c e restrictions with
t e m p o r a l a d v e r b s that are clearly outside of the verb, and they are required in o r d e r
to yield the particular m e a n i n g . F o r instance, t h e inceptive, w h i c h m a r k s a p o i n t
in t i m e , d o e s n o t o c c u r with an a d v e r b indicating a span of t i m e . In addition, s o m e
of these m o r p h e m e s c o m b i n e in u n e x p e c t e d w a y s p h o n o l o g i c a l l y with the conju-
gation m a r k e r s a n d subject p r o n o u n s . T h e s e c o m b i n a t i o n s can b e treated as port-
m a n t e a u m o r p h e m e s , p r o v i d i n g e v i d e n c e for their inflectional nature.
5 . G E N D E R . T h e m o r p h e m e s that I h a v e labeled g e n d e r (those n o r m a l l y t e r m e d
derivational) h a v e s o m e n o n l o c a l correlates, as d i s c u s s e d in Section 2 . 1 . G i v e n
this, these m o r p h e m e s a p p e a r to b e functional. In addition, they s h o w the s a m e
patterns of c o m b i n a t i o n with conjugation and p r i m a r y aspect as the s e c o n d a r y
aspect m o r p h e m e s . W h i l e these m o r p h e m e s h a v e n o n l o c a l p r o p e r t i e s , they are n o t
found with every verb that h a s a particular item as an a r g u m e n t (e.g., ' g o into
w a t e r ' d o e s not o c c u r with the n). T h e g e n d e r m o r p h e m e s a p p e a r to have b e e n
productively inflectional historically, s h o w i n g regular a g r e e m e n t with an argu-
m e n t of the v e r b ; however, it is not clear that this is the case synchronically. Fi-
nally, the g e n d e r m o r p h e m e s , like the s e c o n d a r y aspect m o r p h e m e s , can c o m b i n e
in u n p r e d i c t a b l e w a y s with the conjugation m a r k e r s a n d subject p r o n o u n s , f o r m -
156 Keren D. Rice

ing p o r t m a n t e a u m o r p h s . I c o n s i d e r g e n d e r m o r p h e m e s to b e functional, under-

standing that p r o b l e m s exist with this definition. See also Section 5 for discussion.

T h e other p r i m a r y c a n d i d a t e s for functional status are the distributive and itera-

tive m o r p h e m e s . T h e s e m o r p h e m e s are aspectual in m e a n i n g and thus m i g h t be
e x p e c t e d to b e inflectional. H o w e v e r , unlike the p r i m a r y and s e c o n d a r y aspect
m o r p h e m e s , these m o r p h e m e s are n o t obligatory with a particular m e a n i n g , a cri-
terion identified b y A n d e r s o n and B y b e e as i m p o r t a n t for inflection. F o r instance,
the iterative m e a n i n g c a n b e e x p r e s s e d simply b y stem c h o i c e , and the iterative
m o r p h e m e n e e d not b e p r e s e n t in the verb w o r d . T h i s c a n b e seen in (5).

(5) nididhah ' s / h e p i c k e d u p plural o b j e c t s '

ninadfdheh, nididheh ' s / h e p i c k e d u p plural objects r e p e a t e d l y '

T h e first line is not an iterative form. In the s e c o n d line, the first w o r d includes the
iterative m o r p h e m e na, w h i l e this m o r p h e m e is absent in the second word, with
iterative m e a n i n g c o n t r i b u t e d solely by the v e r b stem. In the c a s e of imperfective,
perfective, a n d optative, or p r i m a r y aspect m o r p h e m e s , these m o r p h e m e s are
obligatory a n d aspect is not d e t e r m i n a b l e b y stem form alone. B y the criterion of
obligatoriness, then, the iterative d o e s n o t qualify as functional.
T h e distributive m o r p h e m e is like the iterative m o r p h e m e in that the m e a n i n g
of distributivity can b e given solely b y stem form or b y c h o i c e of conjugation.

(6) ndneyihkwa T w h i p p e d it' ( y conjugation)

ndydnehkwa, ndnehkwa T w h i p p e d it r e p e a t e d l y ' (w conjugation)

In the distributive e x a m p l e s on the s e c o n d line, the distributive m e a n i n g can b e

e x p r e s s e d only t h r o u g h the c h o i c e of the w conjugation marker, as in the s e c o n d
form; the n o n d i s t r i b u t i v e r e a d i n g is assigned to the form with the y conjugation
m a r k e r given on the first line. T h u s , w h i l e the iterative and distributive m o r -
p h e m e s are aspectual semantically, the fact that they are optional suggests that
they are not functional.
In order to c o m p l e t e this survey of the verb, I will c o m m e n t on the status of the
voice e l e m e n t s . T h e s e m o r p h e m e s d o not a p p e a r to b e functional. Voice and tran-
sitivity m a r k i n g s are not obligatory in that, w h i l e every verb m u s t contain a voice
element, this e l e m e n t d o e s not a l w a y s h a v e a g r a m m a t i c a l function. W h i l e the h
voice e l e m e n t m a r k s transitives a n d the d passives, h n e e d not b e p r e s e n t in
transitives and d n e e d not b e p r e s e n t in passives. F u r t h e r m o r e , h can b e present
in intransitives a n d d in n o n p a s s i v e s . F o r instance, the transitive verb ' k i c k '
(stem td) h a s a 0 voice e l e m e n t w h i l e the intransitive verb ' d u a l g o ' (stem the)
has an h voice e l e m e n t . T h e s e m o r p h e m e s s h o w the types of idiosyncrasies a s s o -
ciated with derivational m o r p h o l o g y rather than with inflection. I thus a s s u m e 9

that the voice e l e m e n t c o m b i n e s with the verb stem lexically, with voice and tran-
sitivity alternations d e t e r m i n e d in the lexicon. S e e also M c D o n o u g h ( 1 9 9 0 ) ,
Speas (1991).
The Structure of the Slave Verb 157

S o far, I h a v e p r o p o s e d that the p r o n o m i n a l a g r e e m e n t m a r k e r s , p r i m a r y and

s e c o n d a r y aspect, conjugation, a n d g e n d e r are functional. W h e n the verb structure
as a w h o l e is e x a m i n e d , a striking p r o p e r t y is evident: these are exactly the m o r -
p h e m e s that fall in t h e conjunct p o r t i o n of t h e v e r b w o r d . T h e disjunct/conjunct
distinction thus can b e seen as r e d u c i n g to a distinction b e t w e e n lexical c a t e g o r i e s
(disjunct m o r p h e m e s , s t e m s ) and functional c a t e g o r i e s (conjunct m o r p h e m e s ) .

4.2. T h e O r d e r i n g of t h e F u n c t i o n a l I t e m s

S o far I h a v e a r g u e d that the conjunct m o r p h e m e s can b e viewed as functional.

I n o w a d d r e s s the o r d e r i n g of these e l e m e n t s , e x a m i n i n g w h e t h e r their o r d e r i n g is
a l a n g u a g e - p a r t i c u l a r p r o p e r t y or follows from m o r e general principles.
A n a s s u m p t i o n m a y b e m a d e that the o r d e r of m o r p h e m e s at the level of logical
form is a reflection of scopal p r o p e r t i e s (e.g., Baker, 1988; S p e a s , 1991). T h e o r d e r
of m o r p h e m e s m u s t directly reflect scopal relations or m u s t b e m a p p a b l e to the
scopal o r d e r i n g . If a l a n g u a g e d o e s not have m o v e m e n t at logical form, o n e m i g h t
e x p e c t the syntactic o r d e r i n g of m o r p h e m e s to b e a direct reflection of scope. W i t h
this h y p o t h e s i s in m i n d , I turn to an e x a m i n a t i o n of the o r d e r i n g of the functional
e l e m e n t s in Slave. I u s e the t e r m SCOPE in the following d i s c u s s i o n ; b y this I m e a n
s o m e t h i n g similar in n a t u r e to B y b e e ' s ( 1 9 8 5 ) t e r m R E L E V A N C E T O T H E VERB.
W h e n the position of the verb stem is ignored, the following o r d e r of m o r -
p h e m e s is found.

(7) DO-gender-secondary aspect-conjugation/primary aspect-subject

T h e subject m o r p h e m e , w h i c h o c c u r s on the right e d g e of the inflectional c o m -

plex in Slave, can b e v i e w e d as b e i n g relevant to an entire sentence (e.g., S p e a s ,
1991); if o r d e r i n g is a c o n s e q u e n c e of s c o p e , o n e m i g h t e x p e c t to find it a p p e a r i n g
on an e d g e . A s p e c t m a y b e seen as h a v i n g s c o p e over the verb and the direct
object. In Slave, p r i m a r y aspect is required in every verb w h i l e s e c o n d a r y aspect
is not, and s o m e s e c o n d a r y aspects o c c u r with a restricted r a n g e of p r i m a r y as-
p e c t s ; it thus a p p e a r s that p r i m a r y aspect m a y h a v e s c o p e over s e c o n d a r y aspect.
G e n d e r in Slave g e n e r a l l y represents c o n c o r d w i t h n o n a g e n t i v e t h e m a t i c roles, or
n o n s u b j e c t s , so it is not u n r e a s o n a b l e to think of this m o r p h e m e as h a v i n g s c o p e
over the direct object, but not o v e r other functional material. Finally, the direct
object h a s s c o p e only over the v e r b itself.
B a s e d on these criteria, the o r d e r i n g of the Slave functional m o r p h e m e s a p p e a r s
to b e a c o n s e q u e n c e of their scopal p r o p e r t i e s . Strikingly, the o r d e r found in Slave
d o e s not a p p e a r to b e u n i q u e to Slave. In recent w o r k on the o r d e r i n g of functional
e l e m e n t s , S p e a s ( 1 9 9 1 ) e x a m i n e s six l a n g u a g e s (English, F r e n c h , M o d e r n G r e e k ,
Finnish, B a s q u e , Navajo), finding the m o r p h e m e o r d e r in (8) to b e constant across
languages. 1 0

(8) subject a g r e e m e n t - t e n s e - a s p e c t - o b j e c t a g r e e m e n t - v o i c e - v e r b
158 Keren D. Rice

T h e l a n g u a g e s d i s c u s s e d by S p e a s d o not h a v e g e n d e r within the verb, so the

m o d e l s are not directly c o m p a r a b l e . However, it is notable that the o r d e r of func-
tional e l e m e n t s in Slave m a y not b e u n i q u e to this l a n g u a g e b u t m a y b e found
cross-linguistically. If this is true, a l a n g u a g e - p a r t i c u l a r statement of scopal rela-
tionships is u n n e c e s s a r y a n d the o r d e r of functional i t e m s in Slave can b e seen to
follow from a theory of scopal o r d e r i n g that is p a r t of universal g r a m m a r .

4 . 3 . T h e S t r u c t u r e of t h e F u n c t i o n a l C o m p l e x

A s s u m i n g that the functional m o r p h e m e s project phrasally, I p r o p o s e the struc-

ture in (9) for the Slave verb.

(9)

Certain aspects of this structure require c o m m e n t . First, the functional c a t e g o -

ries are hierarchically arrayed a b o v e the verb p h r a s e , with m o r p h e m e s of greater
The Structure of the Slave Verb 159

s c o p e b e i n g h i g h e r than m o r p h e m e s within their scope. S c o p e relationships are

thus structurally e n c o d e d . T h e o r d e r i n g of Subject A g r e e m e n t at the top of the
tree follows from this a s s u m p t i o n . T h e verb p h r a s e , as required by the scopal
o r d e r i n g p r i n c i p l e , is s u b o r d i n a t e to the functional c a t e g o r i e s , a p p e a r i n g at the
b o t t o m of the tree. S e c o n d , within the verb p h r a s e , the h e a d of the verb p h r a s e is
indicated as a root, w i t h o u t category. E v i d e n c e for this c o m e s from the fact that
m a n y n o u n s a n d verbs share a root, with c a t e g o r y status achieved t h r o u g h suffix-
ation. T h i s root h a s syntactic p r o p e r t i e s in that it specifies an a r g u m e n t structure,
but is also a m o r p h o l o g i c a l entity in that it lacks c a t e g o r y features. T h e c a t e g o r y
features are p r e s e n t as the head of C P . T h i s e l e m e n t is a syntactic h e a d but is
1 1

m o r p h o l o g i c a l l y s u b c a t e g o r i z e d for b y the root. In o r d e r for the root to receive its

c a t e g o r y features, it m u s t raise into h e a d position, or to CP, by X ° m o v e m e n t .
F o l l o w i n g B a k e r and H a l e ( 1 9 9 0 ) , I a s s u m e that functional h e a d s d o not serve as
barriers for m o v e m e n t , a n d the root stops w h e n it r e a c h e s an e l e m e n t m a r k e d
[ + V ] ; in other w o r d s , it stops w h e n it receives c a t e g o r y features, satisfying its
m o r p h o l o g i c a l s u b c a t e g o r i z a t i o n frame. In this way, the d i s c r e p a n c y b e t w e e n (1)
a n d (9) is a c c o u n t e d for: the root m o v e s from its syntactic position in the verb
p h r a s e to its surface position in C P in o r d e r to receive a category. T h e output of
X ° m o v e m e n t is s h o w n in (10). T r a c e s of m o v e d i t e m s and material in specifier
positions are o m i t t e d .

(10) CP

SubAgr V

PrimAsp SubAgr

SecAsp PrimAsp

Gen SecAsp

/X
DOAgr Gen

/ \
VP DOAgr

/\
VP NP
A s
V Adv

Prev Root,

A c o n s e q u e n c e of raising is that the functional h e a d s and the verb achieve single

w o r d status, f o r m i n g a c o m p l e x w o r d (notice the a b s e n c e of phrasal levels in the
160 Keren D. Rice

derived structure for the functional c a t e g o r i e s ) ; the linear string r e s e m b l e s that in

the t e m p l a t e in (1).

4.4. T h e S t r u c t u r e of t h e VP

S o far I h a v e largely ignored the internal structure of the verb p h r a s e . T h e low-

est functional c a t e g o r y takes as its c o m p l e m e n t the verb p h r a s e . T h e m o r p h e m e s
that are within the verb p h r a s e in Slave are the root and the disjunct m o r p h e m e s
(preverbs, distributive a n d iterative a d v e r b , incorporates). T h e structure of this
part of the verb is i n c l u d e d in (9); see R i c e ( 1 9 9 1 b ) for details. I a s s u m e that each
of the e l e m e n t s of the V P is itself a word, that is, these e l e m e n t s d o not form a
single lexical item.
T h e n o u n p h r a s e that is d a u g h t e r of V P is the direct object. I a s s u m e that this
n o u n p h r a s e generally m o v e s into Specifier of Object A g r e e m e n t position. In cer-
tain cases, w h i c h I will not attempt to a c c o u n t for here, it r e m a i n s within the verb
p h r a s e , giving the effect of n o u n incorporation. I posit a single V P - i n t e r n a l adverb
position for the iterative and distributive m o r p h e m e s ; see R i c e ( 1 9 9 1 b ) for details.
I treat these a d v e r b s as d a u g h t e r s of a phrasal rather than an X ° projection: they
are involved in the selection of functional items w h i c h are not strictly adjacent. If
they were part of the verb root (i.e., X ° ) , they w o u l d b e e x p e c t e d to select only
strictly locally within the w o r d . T h e p r e v e r b s also a p p e a r to b e lexical categories
w h i c h are not p a r t of the root. T h e y d o n o t h a v e syntactic correlates in the way the
functional items d o . S o m e preverbs m a y o c c u r as separate lexical items indepen-
dent of the verb word. F o r instance, the p r e v e r b dah ' a b o v e , located on t o p ' can
o c c u r as an i n d e p e n d e n t postposition, as in the p h r a s e shih dage ' o n top of the
m o u n t a i n , located on t o p of the m o u n t a i n ' . T h e p r e v e r b ?6ne ' a w a y f r o m ' occurs
as a postposition, as in the postpositional p h r a s e s shih ?6ne ' b e y o n d the m o u n t a i n '
or yah?6ne ' o v e r t h e r e ' . I treat the preverbs as d a u g h t e r s of V .
T h e ordering of m o r p h e m e s within the V P c a n p e r h a p s b e seen as following
from universal properties of s c o p e , j u s t as the o r d e r i n g of the functional items can
b e . See R i c e ( 1 9 9 1 b ) for discussion. T h u s templatic o r d e r i n g b e c o m e s u n n e c e s -
sary for the verb overall, as an o v e r a r c h i n g principle exists that d e t e r m i n e s the
order in w h i c h the m o r p h e m e s c o m e .
Restrictions exist on the content of incorporates within the verb p h r a s e . S t e m s
can b e incorporated (i.e., n e e d not m o v e into Specifier of Direct Object A g r e e m e n t
position), even p o s s e s s e d n o u n s with an a g r e e m e n t p r o n o u n as possessor; h o w -
ever, heavier n o u n p h r a s e s are not possible, and m u s t o c c u r in S p e c of D O A g r . A
similar restriction holds of preverbs: a modified p r e v e r b is not allowed within the
verb phrase. T h e r e a s o n s for such restrictions are b e y o n d the scope of this article.

4.5. S u m m a r y

In this section, I h a v e d e v e l o p e d a proposal of S p e a s (1990) for the structure of

the verb. I h a v e suggested that the o r d e r i n g of inflection inside of derivation is a
The Structure of the Slave Verb 161

surface p h e n o m e n o n o n l y ; Slave is n o t highly m a r k e d in r e q u i r i n g that inflection

not b e syntactically accessible. T h i s result is achieved b y treating t h e verb c o m -
plex as syntactic, as follows from t h e a s s u m p t i o n that m o r p h e m e s with syntactic
properties are syntactically accessible. T h e A t h a b a s k a n verb n e e d n o t b e t e m -
p l a t e , as t h e o r d e r i n g of m o r p h e m e s results from s c o p e relationships. T h u s , this
t r e a t m e n t allows for the elimination of a n u m b e r of p r o b l e m a t i c areas that m a k e
the verb highly m a r k e d a m o n g l a n g u a g e s .
A major p r o b l e m that r e m a i n s t o b e a c c o u n t e d for is that of p h o n o l o g i c a l d o -
m a i n s . Before t u r n i n g to this p r o b l e m , I w o u l d like to r e m a r k briefly o n t h e u n -
d e r l y i n g representation of verbs.

5. T H E U N D E R L Y I N G R E P R E S E N T A T I O N O F T H E V E R B

A s d i s c u s s e d in Section 2 . 5 , t h e basic lexical entry of a verb obligatorily in-

c l u d e s a voice e l e m e n t a n d a root. I a s s u m e that t h e v o i c e e l e m e n t is c o m b i n e d
with t h e root lexically, with voice a n d transitivity alternations d e t e r m i n e d in t h e
lexicon; s e e Section 4.1 for discussion. T h e m i n i m a l lexical entry of a verb is thus
as s h o w n in (11).

(11) voice-root] R o o t

M o r e c o m p l e x lexical entries exist, as illustrated in Section 2 . 5 . Pre verbs a n d

other m o r p h e m e s c a n o c c u r within t h e verb t h e m e . In such cases, t h e m e a n i n g is
defined o n t h e entry as a w h o l e , n o t o n individual m o r p h e m e s . F o r instance, in the
verb t h e m e n-h-ji ' s c a r e ' it is n o t p o s s i b l e to assign m e a n i n g s to t h e individual
e l e m e n t s of t h e t h e m e . In s u c h structures, t h e a s s u m p t i o n m a d e in t h e A t h a b a s k a n
literature h a s b e e n that these are single w o r d s ( s e e , e.g., Kari, R a n d o j a , R i c e ,
S p e a s , W r i g h t , a n d m a n y others). H o w e v e r , an alternative solution is available.
D i S c i u l l o a n d W i l l i a m s ( 1 9 8 8 ) , in a study of E n g l i s h phrasal i d i o m s , suggest that
these i d i o m s are syntactic objects that a r e listed in t h e lexicon. T h e i d i o m s are like
w o r d s in that their m e a n i n g s a r e n o n c o m p o s i t i o n a l b u t differ from w o r d s in b e i n g
phrasal. I p r o p o s e that t h e d i s c o n t i n u o u s verb t h e m e s in Slave entries should b e
c o n s i d e r e d as c o m p a r a b l e to E n g l i s h phrasal i d i o m s . ( 1 2 ) gives an e x a m p l e .

(12) [h]
voice M R o o t [jtfJpreverb 'preach, bark'
E a c h m o r p h e m e is labeled for category. W h e n this phrasal unit is inserted into t h e
larger syntactic structure, the m o r p h e m e s are correctly placed. N o further stipu-
lation of position is required, a s it is a direct c o n s e q u e n c e of the phrasal structure,
w h i c h in turn is p r e d i c t e d from scopal p r o p e r t i e s .
Lexical entries c a n also include g e n d e r material a n d direct objects. F o r instance,
the verb ' s c a r e ' h a s a g e n d e r m o r p h e m e a n d 'tell a l i e ' a direct object that m u s t
o c c u r with t h e verb s t e m . 1 2
162 Keren D. Rice

(13) [h] 'scare'

voice Iji] Root L"Jgender '(tell a) l i e '
[ ^ ' I R O O J ^ D O
B y treating d i s c o n t i n u o u s verb t h e m e s as idiomatic, the benefits achieved by
the analysis p r o p o s e d here c a n b e m a i n t a i n e d . Slave m a y b e unusual in the n u m b e r
of phrasal i d i o m s it h a s , b u t the c o n s t r u c t is not in and of itself u n e x p e c t e d .

6. T H E P H O N O L O G I C A L S T R U C T U R E O F T H E S L A V E V E R B

S o far, I h a v e s u g g e s t e d that the Slave verb is phrasal in origin a n d is not

a lexical construct. However, the verb is a single unit phonologically, as h a s
long b e e n r e c o g n i z e d in the A t h a b a s k a n literature. T h i s status is achieved partly
t h r o u g h raising, w h i c h creates an X ° level of the functional c o m p l e x a n d stem. I
suggest in this section that w o r d status is also achieved t h r o u g h m a p p i n g to p r o -
sodic d o m a i n s .

6 . 1 . E v i d e n c e for D o m a i n s o f R u l e A p p l i c a t i o n

A s d i s c u s s e d in Section 2.4, the verb in Slave (and in A t h a b a s k a n l a n g u a g e s in

general) is generally c o n c e i v e d as b e i n g divided into a n u m b e r of p h o n o l o g i c a l
d o m a i n s (e.g., H a r g u s , 1 9 8 8 ; Kari, 1976; Randoja, 1 9 8 9 ; R i c e , 1989). T h e major
d o m a i n s are the disjunct (lexical) m o r p h e m e s , the conjunct (functional) m o r -
p h e m e s , a n d the s t e m s ; the functional items t h e m s e l v e s divide into t w o d o m a i n s .
In this section I briefly outline t h e types of p h o n o l o g i c a l e v i d e n c e that distinguish
these d o m a i n s ; m o r e extensive discussion of e v i d e n c e for d o m a i n s in Slave is
found in R i c e ( 1 9 9 2 ) .
T h e functional a n d lexical d o m a i n s differ in several w a y s , with lexical items
sharing properties with stems a n d functional i t e m s h a v i n g u n i q u e p r o p e r t i e s . O n e
difference b e t w e e n these m o r p h e m e s is in c a n o n i c a l p h o n o l o g i c a l form. T h e lexi-
cal m o r p h e m e s and stems h a v e the following properties.

(14) a. T h e y c a n b e g i n with a n y c o n s o n a n t in the u n d e r l y i n g inventory.

b. T h e y c a n contain a n y v o w e l .
c. T h e s e m o r p h e m e s c a n have the form C V ( C ) a n d C V C V ( C ) . G e n e r a l -
izing, they each constitute a p r o s o d i c foot.

T h e functional items, on the other h a n d , s h o w a m o r e limited r a n g e of properties.

(15) a. T h e y c a n b e g i n with only a subset of the u n d e r l y i n g c o n s o n a n t s .

b. T h e y generally contain the vowel [e], although [i], [a], [u] are possible,
[e] c a n b e treated as epenthetic, and thus m o s t conjunct m o r p h e m e s can
be viewed as h a v i n g the c a n o n i c a l s h a p e C. ( S e e Randoja, 1989, for
detailed discussion.)
The Structure of the Slave Verb 163

A s e c o n d m a j o r difference b e t w e e n lexical and functional m o r p h e m e s is in their

c o n d i t i o n i n g of prefixal alternants. Functional i t e m s c o m b i n e freely with e a c h
other b u t s h o w little c o m b i n a t i o n with lexical m o r p h e m e s . L e x i c a l items, on the
other h a n d , generally pattern as i n d e p e n d e n t w o r d s , only rarely c o m b i n i n g with
functional i t e m s or with e a c h other. To give but o n e e x a m p l e , the s e c o n d p e r s o n
singular subject can b e v i e w e d as h a v i n g the u n d e r l y i n g representation Inf. W h e n
it follows a functional e l e m e n t , it surfaces as nasalization o n the p r e c e d i n g v o w e l ;
w h e n it is word-initial or follows a lexical e l e m e n t , it surfaces in the form / n e / .
T h e p r o c e s s a c c o u n t i n g for this surface a l l o m o r p h y is generally d e s c r i b e d as e p e n -
thesis if n o syllable is p r e s e n t on the functional d o m a i n . N u m e r o u s other p h o -
1 3

nological p r o c e s s e s define these d o m a i n s . A third t y p e of e v i d e n c e for p h o n o l o g i -

cal d o m a i n s c o m e s from e d g e s . F o r instance, c l o s e d syllables are found only in
certain locations in the Slave verb: at the e d g e of a lexical item, at the e d g e of an
a g r e e m e n t m o r p h e m e , or w o r d finally, defining the s a m e d o m a i n s as defined b y
p h o n o l o g i c a l rules. I a s s u m e the c o r r e c t n e s s of these d o m a i n s (see R i c e , 1992 for
details) a n d c o n s i d e r the question of h o w they are d e t e r m i n e d .

6.2. D e r i v i n g t h e D o m a i n s

R u l e d o m a i n s in the verb h a v e b e e n a c c o u n t e d for with b o u n d a r i e s (1) or dia-

critics m a r k i n g the level at w h i c h a m o r p h e m e is attached (e.g., H a r g u s , 1 9 8 8 ;
Randoja, 1989; R i c e , 1989). In all cases, the entry of e a c h n o n s t e m m o r p h e m e
includes, in addition to p h o n o l o g i c a l a n d s e m a n t i c information, a statement of its
level of affixation. In this section, I p r o p o s e a rather different a c c o u n t of the p h o -
nological d o m a i n s , o n e in w h i c h the d o m a i n s are derived by inspection of the
syntactic structure. T h e d o m a i n s thus are not primitive, as in the other theories,
b u t are derived.
Different t y p e s of p h o n o l o g i c a l d o m a i n s m u s t b e d e r i v e d from t h e structure in
(10). First, the traditional verb w o r d is a single d o m a i n phonologically. S e c o n d ,
lexical m o r p h e m e s e a c h pattern p h o n o l o g i c a l l y as separate i t e m s a n d m u s t b e
defined as such within the larger verb w o r d . Third, functional m o r p h e m e s and the
stem pattern as a single item, and finally a g r e e m e n t m o r p h e m e s , both subject a n d
direct object, form the e d g e s of d o m a i n s .
In order to derive the p h o n o l o g i c a l d o m a i n s , I appeal to the literature o n the
derivation of p h o n o l o g i c a l d o m a i n s from syntactic structures. It h a s b e e n a r g u e d
that p h o n o l o g i c a l rules o p e r a t e in p r o s o d i c a l l y rather than m o r p h o l o g i c a l l y or syn-
tactically defined d o m a i n s . S e e , for instance, Selkirk ( 1 9 8 6 ) , N e s p o r a n d Vogel
( 1 9 8 6 ) , H a y e s ( 1 9 8 9 ) , a n d m a n y of the p a p e r s in Inkelas and Z e e ( 1 9 9 0 ) . Selkirk
(1986) p r o p o s e s that syntactic d o m a i n s are m a p p e d o n t o p h o n o l o g i c a l d o m a i n s
by an a l g o r i t h m that refers to e d g e s of syntactic constituents. She argues that a
p h o n o l o g i c a l d o m a i n m a y b e d e t e r m i n e d by seeking out an e d g e , right or left, of
a category of the X - b a r hierarchy, X or X m a x
. H a l e and Selkirk ( 1 9 8 6 ) add a
 164 Keren D. Rice

further parameter, a r g u i n g that l a n g u a g e s m a y refer to functional and lexical cate-

gories as well as to e d g e a n d level.
I p r o p o s e that p h o n o l o g i c a l d o m a i n s in Slave are read off the derived syntactic
structures as follows. T h e verb word, w h i c h I refer to as the PHONOLOGICAL
PHRASE, is defined b y m a r k i n g t h e right e d g e of X , w h e r e X is a functional
m a x

category. This defines the right e d g e of C P as a single word. I t e m s p r e c e d i n g the

verb w o r d are in their o w n functional projections (e.g., D P ) and form their o w n
p h o n o l o g i c a l p h r a s e s . S i n c e the surface structure of the verb contains n o other
m a x i m a l functional projections, the entire verb w o r d is i n c l u d e d in the p h o n o l o g i -
cal p h r a s e . W i t h i n the p h o n o l o g i c a l p h r a s e , w o r d s are d e t e r m i n e d by m a r k i n g the
right e d g e of X ° , w h e r e X is lexical. L e x i c a l categories i n c l u d e preverb, adverb,
n o u n , v e r b , and postposition. N o u n , verb, and p r e v e r b are likely lexical categories
universally; for the others it m a y b e n e c e s s a r y to list as p a r t of the basic lexical
entry that they are lexical. T h i s information is n e e d e d for the m o r p h o s y n t a x a n d
is available to the p h o n o l o g y . M a r k i n g the right e d g e of lexical X ° defines e a c h
disjunct item as a word, as a c c o r d s with traditional t r e a t m e n t s . T h e functional
items p l u s the verb s t e m form a w o r d on their o w n . Finally, w h a t I call SMALL
WORDS are defined by m a r k i n g the right e d g e of a g r e e m e n t . T h u s the direct objects
and the span of m o r p h e m e s from g e n d e r t h r o u g h subject form small w o r d s within
a w o r d . T h e isolation of a g r e e m e n t is an u n u s u a l feature of this analysis. T h e
t r e a t m e n t of a g r e e m e n t as distinct from other inflection is not a characteristic of
j u s t Slave; D r e s h e r (this v o l u m e ) argues that Old E n g l i s h a g r e e m e n t is distinct
from inflection such as n u m b e r , gender, and p e r s o n m a r k i n g , and h e defines a level
w h i c h includes all inflection e x c e p t for a g r e e m e n t .
T h e settings for the derivation of p h o n o l o g i c a l d o m a i n s are s u m m a r i z e d in (16).

(16) a. right e d g e of X m a x
, X is f u n c t i o n a l — p h o n o l o g i c a l p h r a s e
b. right e d g e of X ° , X is lexical (noun, verb, adverb, p o s t p o s i t i o n ) — w o r d
c. right e d g e of a g r e e m e n t — s m a l l w o r d

In (17), a s a m p l e b r a c k e t e d S-structure is given, with the derived p h o n o l o g i c a l

d o m a i n s indicated b e n e a t h it.

(17) [[V, Prev A d v N ] V P D O G e n d e r S e c A s p P r i m A s p Subj V ] C P

( ) phrase, X P
( )( )() ( ) w o r d , X ° , lex
[ ][ ][] ( )( )[ ] small word,
X § a r

A s s u m i n g the strict layer h y p o t h e s i s (e.g., H a y e s , 1989; N e s p o r and Vogel, 1986;

Selkirk, 1980), e a c h w o r d constitutes a small w o r d ( m a r k e d with square brackets).
T h i s h y p o t h e s i s requires that (1) a given n o n t e r m i n a l unit of the p r o s o d i c hierar-
chy, X^, is c o m p o s e d of o n e or m o r e units of the i m m e d i a t e l y lower category,
X ^ , and (2) a unit of a given level of the p r o s o d i c hierarchy is exhaustively
1
The Structure of the Slave Verb 165

c o n t a i n e d in the s u p e r o r d i n a t e unit of w h i c h it is a part ( N e s p o r and Vogel, 1 9 8 6 :

7). If the w o r d s d o n o t contain small w o r d s , then t h e first condition of the strict
layer h y p o t h e s i s is n o t met. S m a l l w o r d s c a n n o t span w o r d s b y the s e c o n d c o n d i -
tion of h y p o t h e s i s .
M a n y p r o c e s s e s reveal t h e n e e d for these d o m a i n s ; I m e n t i o n only a few. T h e
small w o r d is m o t i v a t e d b y restrictions o n t h e distribution of closed s y l l a b l e s . 14

Preverbs a n d i n c o r p o r a t e d stems c a n e n d in a c o n s o n a n t ; t h e reciprocal m o r -

p h e m e , a direct object, is consonant-final, ?eieh\ a n d in subject a g r e e m e n t p o s i -
tion t h e first p e r s o n singular h a n d t h e s e c o n d p e r s o n plural ah a r e c o n s o n a n t -
final. S t e m s c a n also b e consonant-final. Consonant-final m o r p h e m e s thus o c c u r
only at the e d g e of a small w o r d . T h e small w o r d is also the d o m a i n of syllabifi-
cation. It is also n e c e s s a r y t o a c c o u n t for t h e fact that certain p h o n o l o g i c a l p r o -
cesses o c c u r only w i t h i n the small w o r d and not b e t w e e n small w o r d s ; specifically,
direct objects p a t t e r n with t h e lexical m o r p h e m e s in certain w a y s . T h e w o r d is
required t o a c c o u n t for n u m e r o u s p h o n o l o g i c a l p r o c e s s e s that h a p p e n within t h e
conjunct d o m a i n . T h e w o r d forms t h e d o m a i n of foot formation (the m i n i m u m
p r o s o d i c w o r d ) and of the n u m e r o u s rules of Slave that m a k e reference t o the foot,
i n c l u d i n g e x t r a - h i g h tone formation, conjugation t o n e m a p p i n g , voicing a s s i m i -
lation, v o w e l assimilation, a n d others (Rice, 1991a). Finally, various p r o c e s s e s
o c c u r w i t h i n t h e p h o n o l o g i c a l p h r a s e . For instance, t h e e d g e of the p h o n o l o g i c a l
p h r a s e defines t h e e n v i r o n m e n t for insertion of a glottal stop after a h i g h - t o n e
v o w e l ; t o n e d i s p l a c e m e n t in H a r e (Rice, 1991a) h a s the p h o n o l o g i c a l p h r a s e as its
d o m a i n . S o m e of these p r o c e s s e s are s u m m a r i z e d in (18).

(18) PHONOLOGICAL PHRASE: d o m a i n of syllabification, m i n i m a l w o r d c o n -

straints, glottal epenthesis, assimilation t o quality of a vowel
W O R D : / a / raising, foot-based p r o c e s s e s , assimilation to nasality of a vowel
SMALL WORD: rhymal constraints

T h e Slave p h o n o l o g i c a l d o m a i n s follow from i n d e p e n d e n t l y required structural

p r o p e r t i e s of m o r p h e m e s (category) c o u p l e d with the e n d - b a s e d theory and n e e d
not b e e n c o d e d i n d e p e n d e n t l y in the representation of e a c h m o r p h e m e .

7. SUMMARY

I h a v e a r g u e d , c o n t r a r y to the usual h y p o t h e s i s that the verb in Slave is a lexical

construct, that t h e verb " w o r d " in Slave is p h r a s a l , with its single w o r d status
b e i n g a c o n s e q u e n c e of raising and m a p p i n g t o p r o s o d i c structure. A lexical e n t r y
can b e a single w o r d o r a phrasal i d i o m , consisting of m o r e than o n e w o r d . T h e
o r d e r i n g of m o r p h e m e s is predictable b a s e d o n s c o p e relations, o r relevance t o the
verb, with t h e actual o r d e r i n g differing from t h e u n d e r l y i n g o r d e r largely in t h e
166 Keren D. Rice

position of the verb root. In o r d e r to derive the surface position of the root, I
s u g g e s t e d that raising m o v e s it from its position within the verb p h r a s e in order to
assign it c a t e g o r y status.
T h e traditional single ' w o r d ' status of the verb in Slave is in this a c c o u n t a
derived property. T h e e n d - b a s e d a l g o r i t h m defines the verb as a single unit by
seeking out the right e d g e of phrasal projections of functional categories. T h e
p h o n o l o g i c a l d o m a i n s within the w o r d are d e t e r m i n e d b y first m a r k i n g the right
e d g e of major c a t e g o r y lexical items a n d s e c o n d m a r k i n g the right e d g e of a g r e e -
m e n t m o r p h e m e s . T h e verb " w o r d " is thus not a lexical c o n s t r u c t but a p h o n o -
logical o n e , and verb affixes are syntactically w o r d s .

8. C O M P A R I S O N W I T H L E X I C A L P H O N O L O G Y

H o w d o e s this m o d e l c o m p a r e with lexical p h o n o l o g y ? Different a s s u m p t i o n s

are m a d e in the theories, so in s o m e sense they are not c o m p a r a b l e . F o r instance,
in m a n y versions of the theory of lexical p h o n o l o g y , w o r d formation takes p l a c e
entirely in the lexicon, so given the status of the verb as a single w o r d p h o n o l o g i -
cally, the h y p o t h e s i s in lexical p h o n o l o g y is that it is f o r m e d in the lexicon (Har-
g u s , 1988). S e c o n d , s o m e of the characteristics of the verb that I h a v e identified
as u n u s u a l are not issues in lexical p h o n o l o g y ; for instance, it is not c l a i m e d that
there is a relationship b e t w e e n the o r d e r i n g of levels and t h e position of deriva-
tional a n d inflectional m o r p h o l o g y . T h e n e e d for a t e m p l a t e is not necessarily
m a r k e d within lexical p h o n o l o g y as the t e m p l a t e c a n s i m p l y b e v i e w e d as an ex-
tension of the notion of o r d e r i n g of levels to o r d e r i n g within levels. D e s p i t e the
fact that the t w o theories are in m a n y w a y s i n c o m m e n s u r a t e , I believe that the
m o d e l p r o p o s e d in this article enjoys s o m e a d v a n t a g e s . First, it allows for an ac-
c o u n t of the o r d e r i n g of m o r p h e m e s , with functional i t e m s o u t s i d e of lexical items
a n d p e r h a p s with o r d e r i n g within these categories itself b e i n g a c o n s e q u e n c e of
s c o p e o r relevance to the verb. In the lexical p h o n o l o g y m o d e l , the o r d e r i n g of
m o r p h e m e s both on a n d within levels m u s t b e stated in the g r a m m a r . Second, it
m a k e s Slave l o o k m o r e like other l a n g u a g e s w h e r e inflection is syntactically a c -
cessible in an outer layer of the w o r d . W h i l e this is not particularly a c o n c e r n of
lexical p h o n o l o g y , it still m i g h t b e v i e w e d as an a d v a n t a g e that such an account is
possible. Third, the rule d o m a i n s are derived a n d n e e d not b e stipulated as part of
the lexical entry of e a c h m o r p h e m e . T h e m a r k i n g of m o r p h e m e s as b e l o n g i n g to
a particular level has b e e n an issue of controversy in lexical p h o n o l o g y (see G o l d -
smith, 1990, for an o v e r v i e w ) , and the issue d i s a p p e a r s if rule d o m a i n s are d e -
rived. F o u r t h , s o m e t h i n g I h a v e not dealt with in depth, the p h o n o l o g i c a l rules
apply on the derived p r o s o d i c d o m a i n s ; the s a m e rules are available e v e r y w h e r e
a n d it is the s e g m e n t a l and metrical m a k e u p of the d o m a i n that m a k e s the results
The Structure of the Slave Verb 167

different in the different d o m a i n s . T h u s the fact that the stem and the disjunct
prefixes share similar rules is not surprising [as it is in lexical p h o n o l o g y , as this
a p p e a r s to b e a violation of either the strong d o m a i n h y p o t h e s i s (Kiparsky, 1984)
or the c o n t i n u o u s stratum h y p o t h e s i s ( M o h a n a n , 1 9 8 6 ) ] — t h e y share p r o p e r t i e s
b e c a u s e their s e g m e n t a l a n d metrical p r o p e r t i e s are similar, a n d t h e s e p r o p e r t i e s
are similar b e c a u s e they are m a j o r - c a t e g o r y lexical i t e m s as o p p o s e d to functional
i t e m s . In a lexical p h o n o l o g y m o d e l , s o m e of the basic tenets of lexical p h o n o l o g y
m u s t b e rejected since rules apply to d i s c o n t i n u o u s d o m a i n s .
W h a t general c o n c l u s i o n s c a n b e d r a w n with respect to lexical p h o n o l o g y ? T h i s
is a difficult q u e s t i o n to answer, as I have s u g g e s t e d that a l a n g u a g e that a p p e a r s
to h a v e c o m p l e x m o r p h o l o g y , a n d t h u s a p p e a r s to b e a g o o d test c a s e for lexical
p h o n o l o g y , d o e s not in fact have c o m p l e x m o r p h o l o g y . T h e surface c o m p l e x i t i e s
in the m o r p h o l o g y result from derived rather than u n d e r l y i n g p r o p e r t i e s . T h u s ,
any c o n c l u s i o n s r e g a r d i n g the interaction of p h o n o l o g y a n d m o r p h o l o g y in w o r d
formation say n o t h i n g at all about lexical p h o n o l o g y . H o w e v e r , the Slave findings
p e r h a p s force a reevaluation of other l a n g u a g e s that l o o k particularly t r o u b l e s o m e
for lexical p h o n o l o g y . If they receive similar r e a n a l y s e s , it m a y b e that l a n g u a g e s
in w h i c h w o r d formation is truly m o r p h o l o g i c a l rather than resulting from p h o -
nological d o m a i n a s s i g n m e n t are in fact well a c c o u n t e d for b y the m o d e l .
T h e p h o n o l o g i c a l m o d e l that is briefly d e s c r i b e d in this p a p e r also c a n b e
v i e w e d as p r o v i d i n g s u p p o r t for s o m e of the tenets of lexical p h o n o l o g y . K i p a r s k y
( 1 9 8 4 ) p r o p o s e s the strong d o m a i n h y p o t h e s i s , w h i c h allows the g r a m m a r of a
l a n g u a g e to stipulate w h e r e a rule c e a s e s to apply, b u t n o t to turn a rule on. All
rules are thus potentially a p p l i c a b l e at the first level and apply there if p e r m i t t e d
by other p r i n c i p l e s such as the strict cycle condition a n d structure preservation. In
the m o d e l that I h a v e p r o p o s e d , this is precisely the c a s e : all rules are a p p l i c a b l e
at the small w o r d and fail to apply there if their structural description is not m e t
(i.e., the foot is not p r e s e n t yet) or if they violate structure preservation (e.g.,
insertion of glottal stop at the e d g e of a p h o n o l o g i c a l p h r a s e ) . T h u s , in a sense this
m o d e l s u p p o r t s m a n y of the principles of lexical p h o n o l o g y .

9. CONCLUSION

T h e m o d e l of v e r b structure that I h a v e d e v e l o p e d in this article is m e a n t to set

out a research p r o g r a m , a n d as such it inevitably raises m o r e q u e s t i o n s than it
a n s w e r s . First, certain p r o b l e m s particular to Slave (and to the A t h a b a s k a n fam-
ily) h a v e not b e e n discussed. F o r instance, I h a v e i g n o r e d the deictic subjects,
a n d restrictions o n i n c o r p o r a t e s h a v e n o t b e e n a c c o u n t e d for. T h e p h o n o l o g i c a l
m e c h a n i s m required for o r d e r i n g s o m e of the m o r p h e m e s within the small w o r d
is not discussed. T h e m o d e l forces n o u n p h r a s e s to b e a n a l y z e d as d e t e r m i n e r
168 Keren D. Rice

p h r a s e s ; the e v i d e n c e for this m u s t b e explored. Details of d o m a i n p h o n o l o g y need

to b e w o r k e d out: w h i l e m u c h of the p h o n o l o g y o p e r a t e s quite straightforwardly
off metrical structure built within the w o r d s defined by the p r o s o d i c algorithm,
s o m e p r o b l e m s are found. In addition to these k i n d s of l a n g u a g e - p a r t i c u l a r p r o b -
l e m s , m a n y other p r o b l e m s with the a c c o u n t r e m a i n . M u c h of w h a t I have p r o -
p o s e d about the syntactic structure is highly speculative, and far m o r e detailed
a r g u m e n t s m u s t b e d e v e l o p e d for the position outlined here. I h a v e m a d e use of
s c o p e for the o r d e r i n g of m o r p h e m e s ; j u s t w h a t is m e a n t by s c o p e r e m a i n s to be
w o r k e d out. T h e claim that scopal o r d e r i n g is a universal p r o p e r t y also requires
c o n s i d e r a b l e investigation. I h a v e a r g u e d for a syntactic m o d e l ; however, m u c h of
w h a t I h a v e said about the p h o n o l o g y c o u l d also follow from a lexical treatment
of w o r d formation w h e r e the verb h a d a lexical structure similar to the syntactic
structure that I h a v e p r o p o s e d , so the a c c o u n t of verb formation is quite indepen-
dent of the a c c o u n t offered of the p h o n o l o g y . T h e t r e a t m e n t of a g r e e m e n t as spe-
cial by the p h o n o l o g y requires e x p l a n a t i o n ; at this point it serves only to describe
the o b s e r v a b l e facts.
D e s p i t e these p r o b l e m s , the a p p r o a c h taken in this article s e e m s highly p r o m -
ising. It m a k e s the A t h a b a s k a n l a n g u a g e s less bizarre in their formal properties
and m o r e like better u n d e r s t o o d l a n g u a g e s . It also lends s o m e strength to m a n y
tenets of the theory of lexical p h o n o l o g y . T h i s c o n c l u s i o n is p r o m i s i n g a n d is
definitely w o r t h y of further exploration.

ACKNOWLEDGMENTS

I have several people to thank for their helpful discussion of this article: Leslie Saxon,
Peggy Speas, Eloise Jelinek, Sharon Hargus, Elizabeth Cowper, Diane Massam, Aryeh
Faltz, the participants in the lexical phonology conference, and the reviewer of an earlier
version of the article.

NOTES

Inkelas (1989) proposes a version of lexical phonology in which cases of lack of iso-
1

morphism receive an account.

2
In Rice (1989) I identified a number of different aspectual morphemes of the form i-
and two of the form n-. The i- morphemes include semelfactive (action performed a single
time), sedative (segmented action), and transitional; the n- morphemes are completive and
inchoative. Kari (1989) argues that the particular aspectual meaning carried by these mor-
phemes is a property of the semantics of the verb as a whole and is not a result of the
existence of numerous homophonous affixes. I adopt this position here.
The Structure of the Slave Verb 169

3
1 use orthography in most cases. The following symbols should be noted, gh is voiced
velar fricative, dh a voiced dental fricative, th a voiceless dental fricative, an acute accent
represents a high tone, a hook under a vowel represents nasalization. The symbol n, a
palatal nasal, is an abstract representation for a morpheme which may surface as a high
front vowel, as nasalization on a vowel, or as voicing on a continuant, depending on pho-
nological and morphological context.
4
w- is the reflex in Hare, a Slave dialect, of *s; and y- is the reflex of *gh.
5
It is interesting that inflectional and derivational affixes also appear in a marked order
in the Slave noun. In particular, the inflectional morpheme indicating possessive agreement
is phonologically closer to the stem than the derivational augmentative and diminutive
morphemes. See Rice (1991a) for some discussion.
6
T h e inflection/derivation question disappears in some ways since I treat the verb as
phrasal rather than lexical in nature. The so-called derivational items are treated as lexical
categories and the inflectional items as functional categories.
7
This position thus represents an abandonment of the strong version of the lexicalist
hypothesis, which requires that all word formation take place in the lexicon. Instead, in-
flectional morphology is part of syntax proper and lexical operations are restricted to deri-
vational morphology. This assumption alone rules out the type of lexical phonology model
proposed by Hargus (1988), as in that model the verb word is formed in its entirety in the
lexicon.
8
1 use the term F U N C T I O N A L where Anderson (1982, 1988) uses the term I N F L E C T I O N A L .
9
In addition, the classifiers have unique phonological properties.
10
T h i s ordering is similar to that found by Bybee (1985) in her survey of morpheme
ordering in fifty languages; however, Bybee's survey is based on surface morpheme order
and Speas's on a more abstract underlying order, so they are not directly comparable.
1 1
The rudiments of this analysis come from work by Palma dos Santos (1991).
1 2
It is useful to summarize the types of word formation found in the Slave verb. First,
voice elements can be added to the root in the lexicon. Otherwise no lexical word formation
is found. Syntactic "word formation" arises from the operation of Raising. Finally, as dis-
cussed in Section 6, phonological "word formation" results from the imposition of pro-
sodic structure on the syntactic structure.
13
T h i s process is also often thought of as deletion on the functional domain (e.g., Rice,
1989). Whether epenthesis or deletion is the preferred analysis does not affect the claim
that the morpheme patterns differently depending on its position in the word.
1 4
Similar restrictions are found within nouns and postpositions, which also constitute
small words by the definition given.

REFERENCES

Allen, M. (1978). Morphological Investigations. Ph.D. dissertation, University of Con-

necticut, Storrs.
Anderson, S. R. (1982). Where's morphology. Linguistic Inquiry 13, 5 7 1 - 6 1 2 .
Anderson, S. R. (1988). Morphological theory. In Linguistics: The Cambridge Survey,
vol. 1, Linguistic Theory: Foundations (F. J. Newmeyer, ed.), 1 4 6 - 1 9 1 .
170 Keren D. Rice

Axelrod, M. (1990). Incorporation in Koyukon Athapaskan. International Journal of

American Linguistics 56, 1 7 9 - 1 9 5 .
Baker, M. (1988). Incorporation: A Theory of Grammatical Function Changing. Univer-
sity of Chicago Press, Chicago.
Baker, M., and Hale, K. (1990). Relativized minimality and pronoun incorporation. Lin-
guistic Inquiry 21, 2 8 9 - 2 9 7 .
Bybee, J. L. (1985). Morphology, A Study of the Relation between Meaning and Form.
Benjamins, Philadelphia.
Chomsky, N. (1988). Some notes on economy of derivation and representation. MIT Work-
ing Papers in Linguistics 10, 4 3 - 7 4 .
DiSciullo, A.-M., and Williams, E. (1988). On the Definition of Word (Linguistic Inquiry
Monograph 14). MIT Press, Cambridge, Mass.
Goldsmith, J. 1990. Autosegmental and Metrical Phonology. Blackwell, Oxford.
Hale, K., and Selkirk, E. O. (1987). Government and tonal phrasing in Papago. Phonology
Yearbook*, 1 5 1 - 1 8 3 .
Hargus, S. (1986). Phonological evidence for prefixation in Navajo verbal morphology.
Proceedings of West Coast Conference on Formal Linguistics 5, 5 3 - 6 7 .
Hargus, S. (1988). The Lexical Phonology ofSekani. Garland, New York.
Hayes, B. (1989). The prosodic hierarchy in meter. In Rhythm and Meter (P. Kiparsky and
G. Youmans, eds.), pp. 201 - 2 6 0 . Academic Press, Orlando.
Inkelas, S. (1989). Prosodic Constituency in the Lexicon. Doctoral dissertation, Stanford
University, Stanford, Calif.
Inkelas, S., and Zee, D. (1990). The Phonology-Syntax Connection. CSLI Publications
and University of Chicago Press, Chicago.
Kari, J. (1975). The disjunct boundary in the Navajo and Tanaina verb prefix complexes.
International Journal of American Linguistics 41, 3 3 0 - 3 4 5 .
Kari, J. (1976). Navajo Verb Prefix Phonology. Garland, New York.
Kari, J. (1979). Athabaskan Verb Theme Categories: Ahtna (Alaska Native Language Cen-
ter Research Papers 2). Alaska Native Language Center, Fairbanks.
Kari, J. (1989). Affix positions and zones in the Athapaskan verb complex: Ahtna and
Navajo. International Journal of American Linguistics 55, 4 2 4 - 4 5 4 .
Kari, J. (1990). Ahtna Dictionary. University of Alaska Press, Fairbanks.
Kari, J. (1992). Some concepts in Ahtna Athabaskan word formation. In Morphology Now
(M. Aronoff, ed.), pp. 1 0 7 - 1 3 1 . State University of New York Press, Albany.
Kiparsky, P. (1982). Lexical morphology and phonology, In Linguistics in the Morning
Calm (I.-S. Yang, ed.), pp. 3 - 9 1 . Hanshin, Seoul.
Kiparsky, P. (1984). On the lexical phonology of Icelandic. In Nordic Prosody HI (C.-C.
Elert et al., eds.), pp. 135-162. University of Umea, Umea, Sweden.
Li, F.-K. (1946). Chipewyan. In Linguistic Structures of Native America (H. Hoijer, ed.),
pp. 3 9 8 - 4 2 3 . Viking Fund Publications in Anthropology, New York.
Lieber, R. (1981). On the Organization of the Lexicon. Doctoral dissertation, Massachu-
setts Institute of Technology, Cambridge.
McDonough, J. (1990). Topics in the Phonology and Morphology of Navajo Verbs. Doc-
toral dissertation, University of Massachusetts, Amherst.
Mohanan, K. P. (1982). Lexical Phonology. Doctoral dissertation, Massachusetts Institute
of Technology, Cambridge.
The Structure of the Slave Verb 171

Mohanan, K. P. (1986). The Theory of Lexical Phonology Reidel, Dordrecht.

Myers, S. (1987). Tone and the Structure of Words in Shona. Doctoral dissertation, Uni-
versity of Massachusetts, Amherst.
Nespor, M., and Vogel, I. (1986). Prosodic Phonology. Foris, Dordrecht.
Palma dos Santos, A. (1991). Negative Inflection in the Athapaskan Verb. Unpublished
manuscript, University of Toronto.
Pollock, J-Y. (1989). Verb movement, universal grammar, and the structure of IP. Linguis-
tic Inquiry 20, 3 6 5 - 4 2 4 .
Randoja, T. (1989). The Phonology and Morphology of Halfway River Beaver. Doctoral
dissertation, University of Ottawa.
Rice, K. (1989). A Grammar of Slave. Mouton de Gruyter, Berlin.
Rice, K. (1991a). Prosodic constituency in Hare (Athapaskan): Evidence for the foot. Lin-
gua 82, 2 0 1 - 2 4 5 .
Rice, K. (1991b). Predicting the order of the disjunct morphemes in the Athapaskan lan-
guages. Toronto Working Papers in Linguistics 10, 99 - 1 2 1 .
Rice, K. (1991c). Intransitives in Slave (Northern Athapaskan): Arguments for unaccusa-
tives. International Journal of American Linguistics 57, 51 - 6 9 .
Rice, K. (1992). On deriving rule domains: The Athapaskan case. Proceedings of West
Coast Conference on Formal Linguistics 10, 4 1 7 - 4 3 0 .
Sapir, E., and Hoijer, H. (1969). The Phonology and Morphology of the Navaho Language
(University of California Publications in Linguistics 50). University of California
Press, Berkeley and Los Angeles.
Saxon, L. (1986). The Syntax of Pronouns in Dogrib (Athapaskan): Some Theoretical Con-
sequences. Doctoral dissertation, University of California, San Diego.
Selkirk, E. O. (1980). Prosodic domains in phonology: Sanskrit revisited. In Juncture (M.
Aronoff and M.-L. Kean, eds.), pp. 1 0 7 - 1 2 9 . Anma Libri, Saratoga.
Selkirk, E. O. (1986). On derived domains in sentence phonology. Phonology Yearbook 3,
371-405.
Siegel, D. (1978). The adjacency constraint and the theory of morphology. Proceedings of
the North Eastern Linguistics Society 8, 189-197.
Speas, M. (1984). Navajo prefixes and word structure typology. MIT Working Papers in
Linguistics 7, 8 6 - 1 0 9 .
Speas, M. (1986). Adjunctions and Projections in Syntax. Doctoral dissertation, MIT, Cam-
bridge, Mass.
Speas, M. (1990). Phrase Structure in Natural Language. Kluwer, Dordrecht.
Speas, M. (1991). Functional heads and the Mirror Principle. Lingua 84, 181-214.
Thomas-Flinders, T. (1983). Morphological Structures. Doctoral dissertation, University
of California, Los Angeles.
Williams, E. (1981). On the notions "lexically related" and "head of a word." Linguistic
Inquiry 12, 2 4 5 - 2 7 4 .
Wright, M. (1983). The CV skeleton and verb prefix phonology in Navajo. Proceedings of
the North Eastern Linguistics Society 1 4 , 4 6 1 - 4 7 7 .
Wright, M. (1986). Mapping and movement of partial matrices in Navajo. Proceedings of
the North Eastern Linguistics Society 17, 6 8 5 - 6 9 9 .
LOOKING INTO WORDS

RICHARD SPROAT
Linguistics Research Department
AT&T Bell Laboratories
Murray Hill, New Jersey 07974

1. INTRODUCTION

Lexicalist theories of m o r p h o l o g y h a v e generally a s s u m e d s o m e version of the

notion of LEXICAL INTEGRITY, originating in C h o m s k y ( 1 9 7 0 ) . W h i l e there h a s
b e e n a substantial a m o u n t of d i s a g r e e m e n t a b o u t w h a t precise r a n g e of d a t a the
p r i n c i p l e is i n t e n d e d to c o v e r and h o w best to h a n d l e d a t a w h i c h a p p e a r to involve
flagrant violations of it, there h a s b e e n a general a s s u m p t i o n that postlexical p r o -
c e s s e s are b a r r e d from m a k i n g u s e of information a b o u t the internal properties of
w o r d s . A m o n g the m o r e specific v i e w s on lexical integrity is the i m p l e m e n t a t i o n
of the principle, within the theory of lexical p h o n o l o g y and m o r p h o l o g y , as a
special c a s e of the m o r e general p r i n c i p l e of BRACKETING ERASURE ( B E ) (Kipar-
sky, 1982a; M o h a n a n , 1986; Pesetsky, 1979) or OPACITY ( M o h a n a n , 1982). W h i l e
various v i e w s h a v e b e e n e x p r e s s e d on h o w B E applies within the m o r p h o l o g i c a l
c o m p o n e n t (see M o h a n a n , 1982, 1986, for descriptions of t w o of the differing
v i e w s , and see Kiparsky, 1982b, and H a r g u s , 1985, for s o m e v i e w s on h o w B E
m a y fail to apply within the lexicon in s o m e c a s e s ) , it is generally a g r e e d that it at
least applies at the o u t p u t of the m o r p h o l o g i c a l c o m p o n e n t , erasing internal brack-
ets a n d thus obliterating information a b o u t the internal structure of w o r d s . L e x i c a l
integrity thus follows from B E . F o r e x a m p l e , M o h a n a n ( 1 9 8 6 : 2 4 - 2 5 ) a r g u e s :

Another consequence of BE, as pointed out in Pesetsky (1979), is that it derives the prin-
ciple of Lexical Integrity, first proposed in Chomsky (1970). The Lexical Integrity Hy-
pothesis says that syntactic rules cannot have access to the internal structure of words. It

173
Phonetics and Phonology, Volume 4 Copyright © 1993 by Academic Press, Inc.
Studies in Lexical Phonology All rights of reproduction in any form reserved.
174 Richard Sproat

prevents, for example, a pronoun taking father in fatherless as its antecedent, exempts the
self'm self-destruction from the syntactic conditions governing anaphor binding, and pre-
vents -al and -ion from being attached to refuse and destroy in the syntax. Given that
morphological operations take place in the lexicon, and that internal brackets are erased
by BE in the output of the lexicon, it follows that syntactic operations will be blind to the
internal structure created by morphological operations.

In fact, given B E , not only are syntactic operations s u p p o s e d l y blind to m o r p h o -

logical structure, but all postlexical o p e r a t i o n s are s u p p o s e d l y blind to such struc-
ture ( M o h a n a n , 1 9 8 6 : 2 4 ) . I t e r m this view the STRONG INTERPRETATION of brack-
e t i n g erasure.
M y p u r p o s e in this article is to tie t o g e t h e r t w o rather different recent p i e c e s of
research on w h i c h I h a v e b e e n collaborating a n d to b r i n g the data and c o n c l u s i o n s
of e a c h to b e a r on the q u e s t i o n of the status of lexical integrity and its i m p l e m e n -
tation via B E . T h e t w o p i e c e s of r e s e a r c h c o n c e r n on the o n e h a n d the sensitivity
of rules of p h o n e t i c i m p l e m e n t a t i o n to s o m e aspects of m o r p h o l o g i c a l structure,
and on the other the relationship b e t w e e n m o r p h o l o g y and p r a g m a t i c s . In both
cases it will b e a r g u e d that there m u s t b e s o m e sensitivity of INTERPRETIVE COM-
P O N E N T S , such as p h o n e t i c s a n d p r a g m a t i c s , to the internal structure of w o r d s . I
take w h a t I believe to b e the uncontroversial view that p h o n e t i c i m p l e m e n t a t i o n
reads off (postlexical) p h o n o l o g i c a l structure, and that p r a g m a t i c s — a t least the
part of p r a g m a t i c s that deals with a n a p h o r a r e s o l u t i o n — r e a d s directly off s e m a n -
tic structure (and ultimately off syntactic structure). In particular, n o " b a c k d o o r s "
into lexical structure are a s s u m e d to exist for these c o m p o n e n t s . U n d e r that view,
the sensitivity to w o r d structure of the interpretive c o m p o n e n t s in question is
therefore potentially p r o b l e m a t i c for the strong interpretation of B E . I c o n c l u d e
the article b y p o i n t i n g out that t h e d a t a p r e s e n t e d are n o t i n c o m p a t i b l e with s o m e
recent v i e w s of the nature of B E , but that they are i n c o m p a t i b l e at least with the
strong interpretation.
O n e point w h i c h will b e clear is that the m o s t extensive and h e n c e p e r s u a s i v e
e v i d e n c e that interpretive c o m p o n e n t s m u s t b e able to " s e e " inside w o r d s c o m e s
from c o m p o u n d i n g . S o m e a u t h o r s , such as F a b b ( 1 9 8 4 ) , h a v e argued that (at least
s o m e ) c o m p o u n d i n g should b e h a n d l e d in the syntax. Indeed, the even stronger
position that there is n o separate m o r p h o l o g i c a l c o m p o n e n t , and that all m o r -
p h o l o g y m u s t therefore b e d o n e " i n the s y n t a x , " has b e e n taken in Sproat (1985)
a n d L i e b e r ( 1 9 9 1 ) . N e e d l e s s to say, u n d e r either of these a p p r o a c h e s the m o s t
p e r s u a s i v e of the e v i d e n c e d i s c u s s e d h e r e is u n p r o b l e m a t i c . Indeed, it is consistent
with any of the following p o s s i b l e theories.

(1) a. C o m p o u n d i n g is d o n e in the syntax.

b. All m o r p h o l o g y is d o n e in the syntax.
c. M o r p h o l o g y is d o n e in the lexicon, but there is n o B E , at least on the
strong interpretation.

I turn n o w to a description of the data.

Looking into Words 175

2. S E N S I T I V I T Y O F P H O N E T I C I M P L E M E N T A T I O N R U L E S
TO LEXICAL BOUNDARIES

A s part of a study of the articulatory p h o n e t i c s of a l l o p h o n i c variation of the

p h o n e m e l\l in E n g l i s h , Sproat and F u j i m u r a ( 1 9 8 9 , 1993) investigated the effects
of b o u n d a r y strength on intervocalic p r e b o u n d a r y /1/s. Specifically, w e elicited a
n u m b e r of utterances from five speakers (four A m e r i c a n M i d w e s t e r n , o n e " B r i t -
i s h " ) in w h i c h /1/s w e r e p r o d u c e d in the e n v i r o n m e n t / i l ] i / , w h e r e the first vowel
w a s a l w a y s a stressed / i / , the s e c o n d vowel w a s a l w a y s an u n s t r e s s e d /U ( h e n c e
the /1/s w e r e c o n t a i n e d w i t h i n trochaic stress c o n t e x t ) , a n d the only variable w a s
the linguistic b o u n d a r y d e n o t e d b y the left bracket. T h i s b o u n d a r y w a s varied over
the following possibilities, w h e r e e a c h b o u n d a r y c a s e is followed b y the sen-
t e n c e ^ ) u s e d as stimuli.

(2) a. 0 (no b o u n d a r y ) Mr. Beelik wants actors/Mr. Beelik's from Madison. (2

speakers only)
b. + (stratum I b o u n d a r y ) The beelic men are actors.
c. # ( s t r a t u m II b o u n d a r y ) The heeling men are actors.
d. C ( c o m p o u n d , stratum III b o u n d a r y ) The heel equator's amazing.
e. P (phrasal b o u n d a r y separating V P internal constituents): / gave Bee I
equated actors.
f. V (phrasal b o u n d a r y separating subject from V P ) : Beel equates the
actors.
g. | (Major intonation b r e a k b e t w e e n an utterance-initial vocative a n d the
r e m a i n d e r of the utterance): Beel, equate the actors.

T h e utterances w e r e p e r f o r m e d at a m o d e r a t e r e a d i n g rate, and all the utter-

a n c e s investigated w e r e fluent p r o d u c t i o n s of this type. T h e major c o n c l u s i o n
of the s t u d y — w h i c h i n c l u d e d syllable ( m o r p h e m e ) - i n i t i a l as well as the syllable
(morpheme)-final c o n t e x t s m e n t i o n e d a b o v e — w a s that the a l l o p h o n i c variation
b e t w e e n syllable-initial (light) /1/s a n d syllable-final (dark) /1/s is not categorical.
R a t h e r the variation is c o n t i n u o u s a n d is p r o d u c e d by a p h o n e t i c i m p l e m e n t a t i o n
f u n c t i o n w h o s e p a r a m e t e r s are b o t h discrete v a r i a b l e s — r e f e r r i n g directly to
1

w h e t h e r the l\l is syllable-initial or s y l l a b l e - f i n a l — a n d c o n t i n u o u s variables. O n e

such c o n t i n u o u s variable, and o n e w h i c h e x p l a i n s a great deal of the variation in
p r e b o u n d a r y /1/s, is the duration of the p r e b o u n d a r y r i m e ; /1/s in shorter r i m e s are
lighter a l o n g a n u m b e r of different acoustic, articulatory, and physiological di-
m e n s i o n s than /1/s in l o n g e r r i m e s .
O n e of the findings of this investigation w a s that there w a s a g o o d correlation
b e t w e e n the m e a s u r e d acoustic duration of the p r e b o u n d a r y r i m e — w h e r e this w a s
defined as the s e q u e n c e /il/, the tacit a s s u m p t i o n b e i n g m a d e that there w a s n o
^ s y l l a b i f i c a t i o n of the l\l across the w e a k e r b o u n d a r i e s — a n d an a priori linguistic
notion of b o u n d a r y strength; in particular, p r e b o u n d a r y r i m e s before w e a k e r
 176 Richard Sproat
Duration of preboundary rime (msec)

0 1 2 3 4 5 6

Boundary strength

Figure 1. Sensitivity of phonetic implementation rules to boundary strength, (a), speaker CS;
(b), speaker CC.

b o u n d a r i e s w e r e shorter. A typical e x a m p l e of the variation is plotted in Figure l a

for o n e of the A m e r i c a n s p e a k e r s . In this figure, the x axis gives ad h o c n u m e r i c a l
values for the b o u n d a r i e s a c c o r d i n g to an a priori sense of their relative strength.
T h e y axis gives the duration of t h e p r e b o u n d a r y r i m e as defined above, in
milliseconds.
Looking into Words 177

Lighter/I/

#
Retraction of tongue dorsum for l\l (mm)

C
\ C
-69
I
-70

V
I
-71
>

Darker W

i r
3 4

Boundary strength

Figure 1. Continued.

All s p e a k e r s s h o w e d similar t r e n d s , the only n o t a b l e difference b e i n g in the

i m p l e m e n t a t i o n of the V a n d P b o u n d a r y c o n t e x t s , w h e r e s o m e speakers reversed
the trend s h o w n for s p e a k e r C S , p r e s u m a b l y b e c a u s e of differing intonational im-
p l e m e n t a t i o n strategies. I have, however, not d o n e a systematic study of the into-
national i m p l e m e n t a t i o n of the phrasal cases: m y p r i m a r y interest for the current
178 Richard Sproat

p u r p o s e s is to discuss the lexical cases involving the b o u n d a r y contexts labeled

+ , #, and C. T h e r e should, of c o u r s e , b e n o t h i n g surprising in the c o n c l u s i o n that
p h o n e t i c i m p l e m e n t a t i o n rules, such as those r e s p o n s i b l e for assigning duration,
should b e sensitive to different k i n d s of b o u n d a r i e s ; see, inter alia, L e h i s t e (1980),
P i e r r e h u m b e r t and B e c k m a n (1988), and S i l v e r m a n (1988), w h o discuss the sen-
sitivity of p h o n o l o g i c a l i m p l e m e n t a t i o n rules to various k i n d s of phrasal b o u n d a -
ries. However, the e v i d e n c e that there is s o m e sensitivity to w o r d - i n t e r n a l contexts
is p e r h a p s s o m e w h a t novel; it is also p r o b l e m a t i c if the view is taken that w o r d -
internal b o u n d a r i e s are invisible to postlexical p r o c e s s e s , a s s u m i n g that p h o n e t i c
i m p l e m e n t a t i o n rules are c o n s i d e r e d to b e postlexical p r o c e s s e s (see Kiparsky,
1985, for discussion of that issue, as well as M o h a n a n , 1986, etc.). N o n e t h e l e s s , it
is clear from an e x a m i n a t i o n of F i g u r e l a that w h i l e on the o n e h a n d r i m e s p r e -
c e d i n g the s u p p o s e d lexical c o m p o u n d (C) b o u n d a r y are shorter in duration than
r i m e s p r e c e d i n g phrasal b o u n d a r i e s , they are also longer than r i m e s p r e c e d i n g
other lexical b o u n d a r i e s . N o w , for n o s p e a k e r is it the c a s e that the + or #
b o u n d a r y c o n t e x t s s h o w a significant difference from e a c h other or from the n o -
b o u n d a r y c o n t e x t (0). O n the other h a n d , for every s p e a k e r (except speaker R S ,
w h o s e trend is n o n e t h e l e s s in the required direction) the difference b e t w e e n the
c o m p o u n d b o u n d a r y cases and the other lexical cases is significant, as the results
of a t test confirm, for the distinction b e t w e e n + and C b o u n d a r y c o n t e x t s . 2

(3) Speaker t P
CS t -8.33
6 = <0.0005
cc t = -6.19
6 <0.001
AD t = -2.87
6 <0.05
DB f = -5.27
6
<0.005
RS t = -1.25
6
0.26

All speakers show a 2 7 - 6 9 m s e c difference in the m e a n p r e b o u n d a r y r i m e dura-

tions of the + and C c o n t e x t s . O n e interpretation of these data, then, is that w h i l e
there is n o e v i d e n c e of sensitivity to w e a k e r lexical b o u n d a r i e s ( + and # ) , duration
rules are sensitive to the e d g e s of other d o m a i n s , including c o m p o u n d b o u n d a r i e s .
N o t e that o n e could also relate the a b o v e data to the o b s e r v a t i o n s discussed in
L e h i s t e ( 1 9 7 2 , 1980, etc.) that there is an inverse relationship b e t w e e n a syllable's
duration and the n u m b e r of syllables in the w o r d c o n t a i n i n g that syllable. Lehiste
( 1 9 8 0 : 5 ) notes, for e x a m p l e , that the length of the syllable speed as an isolated
w o r d w a s systematically l o n g e r than the s a m e syllable in the w o r d speedy. It
should b e b o r n e in m i n d that in the w o r k reported in L e h i s t e (1972) m a n y of the
test e x a m p l e s w e r e w o r d s p r o d u c e d in isolation; it is not clear that, with w o r d s
p r o d u c e d in a phrasal context, the effect that Lehiste reports is nearly so strong
(J. van Santen, p e r s o n a l c o m m u n i c a t i o n , 1989). N o n e t h e l e s s , o n e m i g h t s u p p o s e
that the r i m e in heel in the c o m p o u n d case is longer than the c o r r e s p o n d i n g r i m e
Looking into Words 179

in the # and + c a s e s simply b e c a u s e of the effect of L e h i s t e ' s o b s e r v e d t e n d e n c i e s .

W h i l e this m i g h t a c c o u n t for the difference, n o t e that in o r d e r for the expla-
nation to work, the w o r d b o u n d a r y after heel in heel equator m u s t b e visible
q u a w o r d b o u n d a r y . T h i s m a y s e e m painfully o b v i o u s a n d u n p r o b l e m a t i c a n d
3

p r o b a b l y w o u l d s e e m so to the majority of p h o n e t i c i a n s w h o h a v e investigated

duration. B u t again, given the strong interpretation of B E , it is not at all o b v i o u s
that the required w o r d b o u n d a r y information w o u l d b e available to p h o n e t i c
implementation.
It should also b e p o i n t e d out that duration is n o t the only p h o n e t i c p r o p e r t y
w h i c h s h o w s itself sensitive to at least s o m e lexical b o u n d a r i e s in o u r w o r k on
T h e a l l o p h o n i c variation in l\l itself is also sensitive to b o u n d a r y strength. F o r
e x a m p l e , the d e g r e e of t o n g u e d o r s u m retraction also s h o w s a significant differ-
e n c e in b e h a v i o r b e t w e e n c o m p o u n d a n d other lexical b o u n d a r i e s , as s u g g e s t e d
b y F i g u r e l b , a n d also the following results of a t test for the c o m p a r i s o n b e t w e e n
t o n g u e d o r s u m retraction in the + a n d C e n v i r o n m e n t s .

(4) Speaker t P
CS h = 6.67 <0.001
cc - 4.00 <0.01
AD *6 = 2.52 <0.05
DB *6 = 1.74 0.13
RS *6 - 2.14 0.076

T o n g u e d o r s u m retraction is o n e d e t e r m i n a n t of lightness of /If: a m o r e retracted

t o n g u e d o r s u m — t h a t is, a m o r e strongly n e g a t i v e y value in F i g u r e l b — y i e l d s a
d a r k e r /If. N o w , as n o t e d above, a major d e t e r m i n a n t of lightness of p r e b o u n d a r y
/1/s w a s found to b e the duration of the p r e b o u n d a r y r i m e , a n d so the data in F i g u r e
l b c a n at the very least b e v i e w e d as an i n d e p e n d e n t c h e c k o n t h e s a m e sensitivity
to b o u n d a r y strength as discussed a b o v e . O n the other h a n d , S p r o a t a n d F u j i m u r a
( 1 9 8 9 ) n o t e that d u r a t i o n c a n n o t explain all of the o b s e r v e d variation, a n d it is at
least p o s s i b l e that s e g m e n t s adjacent to b o u n d a r i e s m a y b e m o r e directly sensitive
to the b o u n d a r y in q u e s t i o n ( m u c h as tones are sensitive, following P i e r r e h u m b e r t
a n d B e c k m a n , 1988, etc.). Therefore, the sensitivity of the quality of l\l to the
strength of the b o u n d a r y s h o w n in F i g u r e l b m a y b e partly indicative of an i n d e -
p e n d e n t sensitivity of p h o n e t i c i m p l e m e n t a t i o n to w i t h i n - w o r d b o u n d a r y strength.
I turn n o w to the q u e s t i o n of h o w to refer to b o u n d a r i e s in p h o n e t i c i m p l e m e n -
tation. O n e of the positive c o n t r i b u t i o n s of the w o r k on level-ordered m o r p h o l o g y
dating b a c k to Siegel ( 1 9 7 4 ) and c u l m i n a t i n g in lexical p h o n o l o g y and m o r -
p h o l o g y is the o b s e r v a t i o n that b o u n d a r y s y m b o l s in the s e n s e of C h o m s k y a n d
H a l l e ( 1 9 6 8 ) are both u n d e s i r a b l e a n d unnecessary. T h e w o r k d o n e by b o u n d a r i e s
in the older t h e o r y is t a k e n over b y lexical strata in t h e n e w e r a p p r o a c h . T h e p r o b -
lem for the c u r r e n t d i s c u s s i o n is that information a b o u t the strata at w h i c h lexical
180 Richard Sproat

c o n s t r u c t i o n s , including c o m p o u n d s , are derived is s u p p o s e d to b e invisible to

postlexical p r o c e s s e s . S o , unlike the b o u n d a r y s y m b o l s of C h o m s k y and Halle
( 1 9 6 8 ) , w h i c h m i g h t c o n c e i v a b l y h a v e struck a r o u n d until p h o n e t i c i m p l e m e n t a -
tion, n o such information is available u n d e r the later theories.
T h e r e are t w o w a y s out of this p r o b l e m , both of w h i c h a p p e a r to c o m p r o m i s e
the strictest a s s u m p t i o n s about B E . Let us c o n c e n t r a t e on the c a s e of c o m p o u n d s ,
w h e r e there is clear e v i d e n c e that the w o r d - i n t e r n a l b o u n d a r y m u s t b e visible. T h e
first solution w o u l d m a k e use of the idea that, from a metrical p h o n o l o g i c a l point
of view, E n g l i s h c o m p o u n d s consist of t w o p h o n o l o g i c a l w o r d s co, in the sense of
Selkirk (1980) and s u b s e q u e n t research, including Booij ( 1 9 8 3 ) and Booij and
L i e b e r (this v o l u m e ) . T h e representation of beel equator w o u l d thus b e metrically
as in (5a). If w e a s s u m e that the m e t r i c a l structure is available to p h o n e t i c i m p l e -
m e n t a t i o n , then w e c o u l d a c c o u n t at least for the facts of p r e b o u n d a r y lengthen-
ing b y allowing p h o n e t i c i m p l e m e n t a t i o n rules to refer to the metrical p h o n o l o g i -
cal w o r d .

(5) a. (o b. y

03 Q)

beel equator

A slightly different interpretation is available u n d e r the theory p r o p o s e d in In-

kelas ( 1 9 8 9 , this v o l u m e ) . Inkelas s u g g e s t s that the lexical strata of lexical p h o -
n o l o g y should b e reinterpreted as b e i n g p r o s o d i c d o m a i n s o n a p a r with the p r o -
sodic d o m a i n s discussed for p h r a s a l p h o n o l o g y in various w o r k including N e s p o r
and Vogel (1986) a n d P i e r r e h u m b e r t a n d B e c k m a n ( 1 9 8 8 ) . U n d e r h e r analysis
( 1 9 8 9 : 9 4 - 9 6 ) , a c o m p o u n d in E n g l i s h w o u l d b e considered, phonologically, to
consist of a c o m b i n a t i o n of t w o p ( = s t r a t u m II) d o m a i n s into a 7 d o m a i n , as
indicated in (5b). T h i s view is particularly interesting since it is k n o w n (Pierre-
h u m b e r t and B e c k m a n , 1988) that p h o n e t i c i m p l e m e n t a t i o n rules are sensitive to
phrasal p r o s o d i c d o m a i n s , a n d it s e e m s a natural e x t e n s i o n to a s s u m e s o m e sen-
sitivity to lexical p r o s o d i c d o m a i n s also.
O n either of these v i e w s , however, the strong interpretation of B E is c o m p r o -
m i s e d . In the s e c o n d view, the c o m p r o m i s e is rather direct, since in Inkelas's
theory, b r a c k e t i n g erasure is i m p l e m e n t e d as erasure of information about w o r d -
internal p r o s o d i c b o u n d a r i e s . T h e r e is a c o m p l i c a t i o n h e r e : Inkelas's m o d e l of
p r o s o d i c b r a c k e t erasure ( 1 9 8 9 : 5 7 ) d o e s not erase internal b r a c k e t s if the t w o
d a u g h t e r s of a n o d e are of a different p r o s o d i c t y p e from the m o t h e r n o d e ; in (5b),
then, the t w o P n o d e s w o u l d r e m a i n visible, so o n e m i g h t explain the above data
o n that basis. However, if the w h o l e c o m p o u n d beel equator w e r e e m b e d d e d in a
larger c o m p o u n d , the t w o (3 n o d e s w o u l d b e erased u n d e r Inkelas's theory, and so
o n e w o u l d e x p e c t the r i m e of beel in this larger c o m p o u n d to h a v e a categorically
different duration from that in beel equator. T h i s s e e m s highly unlikely, so it
Looking into Words 181

w o u l d still s e e m n e c e s s a r y to c o m p r o m i s e B E . Inkelas d o e s , however, suggest an

alternative interpretation of b r a c k e t i n g e r a s u r e w h i c h d o e s not involve actual era-
sure of p r o s o d i c information, and I return to that alternative below.
In the first view, the c o m p r o m i s e is m o r e indirect, in that internal m o r p h o l o g i c a l
structure b e c o m e s visible to postlexical p r o c e s s e s b y e n c o d i n g at least part of that
structure in m e t r i c a l representation, w h i c h is not generally subject to e r a s u r e of its
internal s t r u c t u r e . At the very least this w e a k e n s the predictive p o w e r of B E ,
4

since s o m e of m o r p h o l o g i c a l structure b e c o m e s visible via s m u g g l i n g t h r o u g h

metrical structure. 5

I n o w turn to the other set of data w h i c h suggest that word-internal structure

m a y b e visible to interpretive c o m p o n e n t s .

3. P R A G M A T I C S A N D W O R D - I N T E R N A L S T R U C T U R E

In S p r o a t and W a r d ( 1 9 8 7 ) ; Ward, Sproat, and M c K o o n ( 1 9 9 1 ) ; M c K o o n ,

Ratcliff, Ward, and Sproat ( 1 9 9 0 ) ; and M c K o o n , Ward, Sproat, and Ratcliff
( 1 9 9 3 ) , w e h a v e investigated the status of so-called a n a p h o r i c islands. It will b e
recalled that Postal's ( 1 9 6 9 ) original study c o n c l u d e d that there w a s a g r a m m a t i c a l
(i.e., m o r p h o s y n t a c t i c ) constraint p r o h i b i t i n g a n a p h o r i c reference into and out of
w o r d s . W h i l e various researchers, i n c l u d i n g Lakoff and R o s s ( 1 9 7 2 ) , C o r u m
6

( 1 9 7 3 ) , a n d Watt ( 1 9 7 5 ) , h a v e n o t e d the fact that s o m e c o n s t r u c t i o n s w h i c h o u g h t

to b e infelicitous a c c o r d i n g to the a n a p h o r i c island condition are in fact quite
felicitous, it h a s n o n e t h e l e s s generally b e e n a s s u m e d that p r o n o u n s m a y not take
as their a n t e c e d e n t s e l e m e n t s w h i c h are inside w o r d s . T h o s e cases w h e r e e x -
a m p l e s s e e m fully felicitous h a v e b e e n s u g g e s t e d to b e derived b y " p r a g m a t i c
i n f e r e n c e " (cf. Shibatani a n d K a g e y a m a , 1988), t h o u g h w h a t form this "infer-
e n c e " takes and h o w such " i n f e r e n c e " differs from " r e a l " a n a p h o r a h a s n e v e r
b e e n defined. T h e s u p p o s e d categorical prohibition o n reference by p r o n o u n s into
w o r d s h a s b e e n t a k e n to b e derivable from lexical integrity a n d therefore from B E
b y S i m p s o n ( 1 9 8 3 ) and M o h a n a n ( 1 9 8 6 ) — s e e , e.g., the q u o t e from M o h a n a n in
Section 1.
T h e r e is a p r o b l e m with this v i e w : it a p p e a r s to b e the case, as a r g u e d by Sproat
and W a r d ( 1 9 8 7 ) , W a r d et al. ( 1 9 9 1 ) , M c K o o n et al. ( 1 9 9 0 , 1993), that there is n o
specifically m o r p h o l o g i c a l constraint w h i c h prohibits p r o n o u n s from taking their
a n t e c e d e n t s within w o r d s . A s is a r g u e d in detail in t h o s e references, as well as in
the w o r k of several authors cited a b o v e , fully felicitous e x a m p l e s of a p p a r e n t vio-
lations of the so-called a n a p h o r i c island condition can readily b e found. T h e fol-
lowing naturally o c c u r r i n g e x a m p l e s should suffice to m a k e the point (in these
e x a m p l e s , boldface m a r k s i n t e n d e d coreferents a n d b r a c k e t s delimit the lexical
item c o n t a i n i n g the a n t e c e d e n t ) .
182 Richard Sproat

(6) a. Patty is a definite [Kal Kan cat]. Every day she waits for it.
Television a d v e r t i s e m e n t for Kal K a n , January 2 8 , 1987)
b. There's a [Thurber story] about his maid.. .
H e a r d in c o n v e r s a t i o n , S e p t e m b e r 7, 1988
c. We went up to [Constable country]', we stayed in the village he was
born in.
H e a r d in c o n v e r s a t i o n , O c t o b e r 1 1 , 1988
d. / refer you to the [Schachter paper]', he's very proud of it. . .
S p e a k e r in r e s p o n s e to a question at N o r t h E a s t e r n Linguistics Society,
N o v e m b e r 12, 1988
e. Well, action is still needed. If we 're to finish the job, Reagan's Regiments
will have to become the [Bush Brigades]. Soon he'll be the chief, and
he '11 need you every bit as much as I did.
R. R e a g a n , farewell speech, J a n u a r y 1 1 , 1989, reported in A s s o c i a t e d
Press N e w s w i r e
f. Millions of [Oprah Winfrey fans] were thoroughly confused last week
when, during her show, she emotionally denied and denounced a vile
rumor about herself.
M i k e R o y k o , M a y 2 2 , 1989, cited b y M c C a w l e y , 1989, as an e x a m p l e of
reflexive u s a g e — i . e . , not as an e x a m p l e of an a n a p h o r i c island
g. / had a [paperroute] once but my boss said I took too long deliv-
erin' 'em.
" L . A . L a w , " 1987
h. I'm a [mystery-story buff] and read (and watch on PBS) a lot of them.
N o r t h w e s t e r n University electronic bulletin board, January, 1989
i. We asked [Saab 9000-CD owners] about its road-handling. ..
Television ad for S a a b , M a r c h 12, 1989
j. For a [SYNTAX slot], I'd rather see someone with more extensive
coursework in it.
H e a r d d u r i n g a discussion of various subdisciplines of linguistics, w h e r e
the s p e a k e r w a s c o n t r a s t i n g syntax with other subdisciplines, J a n u a r y
18, 1987
k. At the same time as coffee beans were introduced, the Arabs made
changes in [coffee preparation] that greatly improved its flavor.
Schapira, J., Schapira, D . , a n d Schapira, K., The book of coffee and tea,
1982, p . 7
1. A : Are we ciderless? B : Yes we're [ciderless]. You should have told
me—I would have brought some.
In c o n v e r s a t i o n , J a n u a r y 2 3 , 1987
m. Do [parental] reactions affect their children?
H e a r d in c o n v e r s a t i o n by Jill Burstein, M a r c h 15, 1990
Looking into Words 183

n. "I heard someone say," he began, "that you are a [New Zealander]. I
was out there as a small boy."
M a r s h , N., Night at the Vulcan, 1 9 5 1 , p . 2 0 7
o. Our neighbors, who are sort of [New York City-ites], they have jobs
there . . .
H e a r d in conversation, D e c e m b e r 30, 1990

In W a r d et al. ( 1 9 9 1 ) , w e suggest that it is w r o n g to claim that apparently fe-

licitous e x a m p l e s like the a b o v e are really u n g r a m m a t i c a l and are s o m e h o w a m -
nestied b y p r a g m a t i c factors. N o t e that e x a m p l e s of the other half of the a n a p h o r i c
island constraint, n a m e l y the prohibition in E n g l i s h on p r o n o u n s o c c u r r i n g within
w o r d s , are n e v e r a m n e s t i e d by p r a g m a t i c factors.

(7) * / don't eat bananas because I'm a them -hater.

(7) is not a c c e p t a b l e u n d e r any c o n d i t i o n s . T h u s , if the e x a m p l e s in (6) are c o n -

7

sidered u n g r a m m a t i c a l , w e w o u l d h a v e to explain w h y their u n g r a m m a t i c a l i t y can

b e a m n e s t i e d , w h e r e a s that of (7) c a n n o t .
Still, w e m u s t a c c o u n t for the p h e n o m e n o n that Postal ( 1 9 6 9 ) originally d e -
scribed: clearly n o t all c a s e s w h e r e an a n a p h o r finds its a n t e c e d e n t within a w o r d
are felicitous. A n d w e m u s t also p r o v i d e an a c c o u n t of w h i c h factors r e n d e r s o m e
e x a m p l e s felicitous.

3 . 1 . W h y R e f e r e n c e i n t o W o r d s Is O f t e n Infelicitous

Various c o n s i d e r a t i o n s lead o n e to e x p e c t that reference into w o r d s will often

b e infelicitous. F o r o n e thing, w o r d s often h a v e idiosyncratic s e m a n t i c s a n d it is a
p r e c o n d i t i o n for felicity of reference into w o r d s that the w o r d in w h i c h the in-
tended a n t e c e d e n t is found b e sufficiently t r a n s p a r e n t semantically. C o n s i d e r the
following e x a m p l e s .

(8) a. Fred is a cowboy. #He says they can be difficult to look after.
b. #John wants to be a fireman because he likes putting them out.

W o r d s like cowboy and fireman are lexicalized (institutionalized, o p a q u e ) c o m -

p o u n d s . W h a t this m e a n s is that a l t h o u g h the w o r d s m o r p h o l o g i c a l l y contain cow
and fire respectively, and a l t h o u g h (in the real w o r l d ) c o w b o y s are s o m e h o w re-
lated to c o w s and firemen to fires, it is very clear that o n e d o e s not get the s e m a n t i c
referent of either w o r d by any d e c o m p o s i t i o n a l strategy. Since interpreting the
m e a n i n g of cowboy a n d fireman d o e s not require the s e m a n t i c interpretation of
the m o r p h o l o g i c a l parts, w e w o u l d not e x p e c t reference to those parts ( m o r e cor-
rectly, to the d i s c o u r s e entities they m i g h t e v o k e ) to b e felicitous.
In cases w h e r e the s e m a n t i c s of a w o r d are sufficiently transparent, reference
184 Richard Sproat

into w o r d s is often quite felicitous. C o n s i d e r s o m e of the e x a m p l e s of naturally

o c c u r r i n g d a t a from above. In synthetic c o m p o u n d s of the t y p e found in (6i), for
e x a m p l e , there is n o q u e s t i o n that b y a n d large such constructions are interpreted
compositionally, since it is d e r i v a b l e from the m e a n i n g of own and -er that a Saab
9000-CD owner is g l o s s e d ' s o m e o n e w h o o w n s a S a a b 9 0 0 0 - C D ' . In c o m p l e t e l y
n o n c e f o r m a t i o n s such as Kal Kan cat in (6a), w e can appeal to the fact (cf. D o w n -
ing, 1977) that hearers are able to c o m p u t e an interpretation (appropriate to the
discourse context) for such cases b y c o n s i d e r i n g a p p r o p r i a t e relations b e t w e e n the
parts of the c o m p o u n d ; the hearer in this c a s e w o u l d n e e d to figure out w h a t rela-
tion b e t w e e n the s u b s t a n c e Kal K a n a n d cats is b e i n g c o m m u n i c a t e d . T h i s cru-
cially involves a c c e s s i n g the referents of the t e r m s Kal Kan a n d cat, unlike the
situation with cowboy a n d fireman.
A s e c o n d c o n s i d e r a t i o n revolves a r o u n d the syntactic function, position, or t y p e
of i t e m s w h i c h are c o n t a i n e d within w o r d s . H e r e I j u s t c o n s i d e r the c a s e of the
left-hand m e m b e r of a c o m p o u n d . It is generally a s s u m e d that the left-hand m e m -
b e r of a c o m p o u n d functions as a modifier of the r i g h t - h a n d m e m b e r ; see Levi
( 1 9 7 8 ) , inter alia. S o , if it turns out that there is r e a s o n to a s s u m e that p r e n o m i n a l
modifiers are generally less a c c e s s i b l e — i n a sense to b e clarified b e l o w — t h a n
other syntactic positions, w e h a v e a n o t h e r part of the e x p l a n a t i o n for w h y refer-
e n c e into c o m p o u n d s is often infelicitous. W i t h this in m i n d , c o n s i d e r the follow-
ing d i s c o u r s e , w h e r e either (9a) or (9b) can b e the s e c o n d sentence.

(9) John doesn't like to visit his relatives very much.

a. His intolerable aunt is hostile.
b . His hostile aunt is intolerable.
He never has a very good time.

In an e x p e r i m e n t r e p o r t e d in M c K o o n et al. ( 1 9 9 0 ) a n d W a r d et al. ( 1 9 9 1 ) , s u b -
j e c t s read b l o c k s of e x a m p l e d i s c o u r s e s such as the t w o r e p r e s e n t e d in (9) a n d
w e r e s u b s e q u e n t l y p r e s e n t e d w i t h lists of test w o r d s w h e r e they w e r e required to
indicate as rapidly a n d accurately as possible w h e t h e r or not the test w o r d w a s in
o n e of t h e p r e c e d i n g d i s c o u r s e s . In cases w h e r e a w o r d w a s a modifier, such as
hostile in (9b), subjects r e s p o n d e d significantly m o r e slowly to the test item hos-
tile than w h e n that w o r d h a d a p p e a r e d in a predicate as in (9a); in the t w o dis-
c o u r s e s in (9), the c o m p l e m e n t a r y results w e r e found for intolerable. S o there is
g o o d psycholinguistic e v i d e n c e that modifier position is i n d e e d less accessible
than at least o n e other syntactic position, a n d w e suggest in W a r d et al. (1991) that
this fact is a n o t h e r factor in d e t e r m i n i n g the relative infelicity of reference into
words.
N o t e that the a b o v e f a c t o r s — s e m a n t i c opacity and the effect of the t y p e of
syntactic function (or position) of s o m e word-internal e l e m e n t s — d o h i n g e on
properties of the m o r p h o l o g i c a l c o n s t r u c t i o n s involved. B u t there is n o m o r p h o -
syntactic PROHIBITION against reference into w o r d s . Rather, various properties of
Looking into Words 185

w o r d s c o n s p i r e to r e n d e r such reference infelicitous (or not fully felicitous) in

many cases.

3.2. I n c r e a s i n g t h e Felicity of R e f e r e n c e i n t o W o r d s

Various s e m a n t i c a n d p r a g m a t i c factors m a y m a k e reference into w o r d s felici-

tous. O n e p r e c o n d i t i o n I h a v e already d i s c u s s e d is that the w o r d w h i c h c o n t a i n s
the i n t e n d e d a n t e c e d e n t b e semantically transparent. O t h e r factors that are relevant
are contrast and topicality. F o r e x a m p l e , in (6j), contrast b e t w e e n syntax and other
subdisciplines of linguistics helps r e n d e r the d i s c o u r s e entity e v o k e d b y syntax in
the c o m p o u n d syntax slot m o r e accessible for s u b s e q u e n t reference b y t h e p r o -
n o u n it. A l s o , p s y c h o l i n g u i s t i c studies r e p o r t e d in M c K o o n et al. ( 1 9 9 3 ) a n d W a r d
et al. ( 1 9 9 3 ) s h o w that d i s c o u r s e s in w h i c h the i n t e n d e d w o r d - i n t e r n a l a n t e c e d e n t
is topical felicitate s u b s e q u e n t reference to the d i s c o u r s e entity e v o k e d b y that
antecedent. For example, a discourse concerning animals and nature but with no
p r i o r explicit m e n t i o n of the w o r d deer will r e n d e r reference to deer, as e v o k e d b y
the a p p e a r a n c e of the m o r p h e m e deer within a c o m p o u n d such as deer hunter,
m o r e accessible for s u b s e q u e n t reference than a d i s c o u r s e in w h i c h d e e r are not
topical.

(10) a. TOPICAL: Sam likes the outdoor life. Having grown up in rural Ken-
tucky, he knows a lot about nature and is an expert at fishing and shoot-
ing. He goes on hunting trips as often as he can. He used to hunt just
small game, like rabbit and quail. However, lately he's taken up deer
hunting. He thinks that they are really exciting to track.
b. N O N T O P I C A L : Sam has many interests in the outdoors. He's an avid
skier, and each winter he takes about a month off from work to ski in
Colorado. In the summertime, he visits his parents in Montana where
he has a chance to do some mountain climbing. Lately, he's taken up
deer hunting. He thinks that they are really exciting to track.

T h i s accessibility-raising effect can b e verified b y m e a s u r i n g s u b j e c t s ' r e a d i n g

t i m e for the (final) s e n t e n c e c o n t a i n i n g an a n a p h o r {they) referring to deer; r e a d i n g
t i m e is significantly faster in the topical c o n t e x t than in the n o n t o p i c a l context. W e
a r g u e that this is b e c a u s e in the n o n t o p i c a l context, it is h a r d e r for the subject to
resolve the reference of the p r o n o u n (they) than in the topical c o n t e x t . 8

3.3. Summary

So, u n d e r a p p r o p r i a t e d i s c o u r s e c o n d i t i o n s reference into w o r d s is p o s s i b l e ,

apparently c o n t r a d i c t i n g c l a i m s that such reference is m o r p h o s y n t a c t i c a l l y ruled
out. Indeed, M o h a n a n ' s o w n e x a m p l e s , fatherless a n d self-destruction, have no
186 Richard Sproat

special properties in this r e g a r d . F o r the first case, the following s e e m s perfectly

9

felicitous [cf. (6,1)].

(11) Mary has been fatherless for years; he died when she was five.

S o reference to father in fatherless is n o t in general ruled out. W i t h regard to the

s e c o n d e x a m p l e , self-destruction, t h e issue arises as to w h e t h e r self- should b e
c o n s i d e r e d to b e an a n a p h o r o r m e r e l y an operator w h i c h b i n d s t w o a r g u m e n t s in
the a r g u m e n t structure of its b a s e , as s u g g e s t e d b y D i Sciullo and W i l l i a m s ( 1 9 8 7 ) .
If self- is j u s t a lexical o p e r a t o r a n d n o t an anaphor, then n o special statement is
n e e d e d to e x e m p t self- from a n a p h o r b i n d i n g principles, w h i c h w o u l d not apply
to it anyway. F o r those w h o h a v e a r g u e d that self- is an a n a p h o r (e.g., Lieber,
1984, 1 9 9 1 ; Sproat, 1985), it is n o t difficult to find e x a m p l e s w h i c h s e e m to sug-
gest that self- c a n b e b o u n d b y an e l e m e n t outside its c o n t a i n i n g w o r d .

(12) John's gradual self-destruction by drug use was heartrending to his family.

W h a t is the relevance of these o b s e r v a t i o n s to the issue of B E ? Clearly w e want

to say that t h e r e a s o n t h e e x a m p l e s in ( 3 ) a r e felicitous is that t h e a n t e c e d e n t s in
question e v o k e the relevant d i s c o u r s e entities b y b e i n g MORPHOLOGICALLY p r e s -
ent. T o b e s u r e — a n d t o allay a n y p o s s i b l e m i s u n d e r s t a n d i n g s — t h e r e are cases
w h i c h s e e m to require a different analysis.

(13) A : You're not English, are you? B : No, I was brought up there . . .
(Gaslight)

H e r e , it s e e m s unlikely that England—the a n t e c e d e n t of there—is morphologi-

cally present in English. Rather, there s e e m s to b e s o m e sort of lexical inference
g o i n g on, in t h e sense that English functions a s t h e P R O V E N A N C E adjective
related to England, a n d it is this well-instantiated LEXICAL r e l a t i o n s h i p w h i c h 10

renders t h e e x a m p l e felicitous. T h e analysis of this particular c a s e is r e m i n i s c e n t

of t h e analysis of t h e RELATEDNESS PARADOX theoretical linguist in M o h a n a n
( 1 9 8 6 : 2 5 ) ; recall that in that c a s e theoretical is n o r m a l l y interpreted as modifying
linguistics, n o t linguist, since the m e a n i n g is ' o n e w h o practices theoretical lin-
g u i s t i c s ' . M o h a n a n suggests, correctly I believe, that theoretical c a n modify lin-
guistics, n o t b e c a u s e t h e latter is m o r p h o l o g i c a l l y c o n t a i n e d within linguist—it
patently i s n ' t — b u t rather b e c a u s e linguist is k n o w n to m e a n ' o n e w h o practices
linguistics'. T h e modification of linguistics is thus at s o m e s e m a n t i c level a n d not
at the m o r p h o l o g i c a l level; see also S p e n c e r ( 1 9 8 8 ) for discussion of this kind of
relatedness p a r a d o x . It s e e m s to b e true, in fact, that cases like linguist w h i c h
allow for these k i n d s of relatedness p a r a d o x e s also typically allow for the kind of
reference exhibited in ( 1 3 ) . C o m p a r e t h e relatedness p a r a d o x e s in (14a,c) with
( 1 4 b , d ) , w h e r e t h e p r o n o u n corefers with part of the m e a n i n g of the contain-
ing w o r d .
Looking into Words 187

(14) a. John's an East German, (cf. East Germany)

b. John's a German though he hasn't lived there for many years.
c. John's a transformational linguist.
d. John's a linguist; he says it's an exciting field.

Plausibly, similar m e c h a n i s m s for extracting Germany from German and linguis-

tics from linguist are involved in all c a s e s .
O n e m i g h t a t t e m p t an analysis a l o n g the lines of that required for (13) or theo-
retical linguist b y s u g g e s t i n g that the c o m p o u n d Kal Kan cat is lexically related
to Kal Kan by s o m e r e l a t i o n — s a y the C A T - W H I C H - L I K E S - X r e l a t i o n — a n d
that it is b e c a u s e of this s e m a n t i c relationship that reference to Kal K a n is felici-
tous in (6a). T h e p r o b l e m with this view is that, as W a r d et al. (1991) a r g u e for
e x a m p l e s like (13), and as S p e n c e r ( 1 9 8 8 ) a r g u e s for e x a m p l e s like theoretical
linguist, such s e m a n t i c inference s e e m s to require m i n i m a l l y that the c o n t a i n i n g
w o r d s — t h a t is, English c o n t a i n s England, and linguist c o n t a i n s linguistics—be
listed lexical i t e m s or that there at least b e a well-instantiated c o n v e n t i o n a l rela-
tionship b e t w e e n the c o n t a i n i n g w o r d a n d the w o r d w h i c h it notionally c o n t a i n s .
P r e s u m a b l y Kal Kan cat h a s neither of these characteristics; it is a c o m p l e t e l y
n o n c e f o r m a t i o n . Clearly then, t h e m o s t straightforward analysis of an e x a m p l e
like (6a) is that reference to Kal K a n is p o s s i b l e b e c a u s e of the m o r p h o l o g i c a l
p r e s e n c e of Kal Kan within Kal Kan cat. Precisely put, Kal Kan e v o k e s a dis-
c o u r s e entity w h i c h c o r r e s p o n d s to the s u b s t a n c e Kal K a n , and w h i c h is suffi-
ciently accessible in the d i s c o u r s e for s u b s e q u e n t a n a p h o r i c reference. O n l y a
prejudice that c o m p o n e n t s of w o r d s c a n n o t b e visible at postlexical levels w o u l d
prevent o n e from a d o p t i n g this m o s t straightforward analysis.
If, as I h a v e suggested, there is n o m o r p h o s y n t a c t i c restriction p r e v e n t i n g ana-
p h o r s from referring into w o r d s , then t w o things w o u l d a p p e a r to follow. First of
all, e v i d e n c e for lexical integrity, or its i m p l e m e n t a t i o n via B E , c a n n o t b e sought
in that d o m a i n , c o n t r a S i m p s o n ( 1 9 8 3 ) a n d M o h a n a n ( 1 9 8 6 ) ; the data on w h i c h
S i m p s o n ' s a n d M o h a n a n ' s a r g u m e n t s rest are simply incorrect. S e c o n d , the c o n -
clusion w o u l d actually s e e m to be p r o b l e m a t i c for B E , at least on its strongest
interpretation. It is true that the g r a m m a r itself d o e s not n e e d to b e able to see
inside Kal Kan cat in o r d e r to allow reference to the d i s c o u r s e entity e v o k e d by
Kal Kan to serve as a felicitous a n t e c e d e n t for an a n a p h o r . Indeed, in W a r d et
11

al. ( 1 9 9 1 ) , w e followed R e i n h a r t ( 1 9 8 3 ) in s u g g e s t i n g that the g r a m m a r n e e d m a k e

n o statement a b o u t the referential possibilities of p r o n o u n s ; u n d e r this a p p r o a c h ,
word-internal a n t e c e d e n t s such as Kal Kan are in n o way special and are g o v e r n e d
by the s a m e k i n d s of p r a g m a t i c constraints as g o v e r n other instances of p r o n o m i -
nal a n a p h o r a . B u t the p r a g m a t i c s m u s t still be able to see Kal Kan, and if we take
seriously M o h a n a n ' s suggestion that all postlexical p r o c e s s e s are b a r r e d from hav-
ing access to the internals of w o r d s , then o n e w o u l d think that such visibility
would not be available.
188 Richard Sproat

4. S O M E C O N C L U S I O N S

T h e data p r e s e n t e d in the p r e c e d i n g t w o sections is at worst p r o b l e m a t i c for the

strong interpretation of B E , a n d at the very l e a s t — i f word-internal structure is
allowed to b e s m u g g l e d t h r o u g h in w a y s other than m o r p h o l o g i c a l bracket-
i n g — s u g g e s t s that the strong interpretation of B E m i g h t not b e d o i n g as m u c h
w o r k as initially a p p e a r s .
T h e strong interpretation of B E e x p o u n d e d in M o h a n a n ( 1 9 8 6 ) w a s cast in
t e r m s of a theory of m o r p h o l o g y w h e r e w o r d s h a v e a single structure. Similarly,
M o h a n a n ' s view of lexical insertion into syntactic structure a p p e a r s to b e that p h o -
netic strings are inserted into syntactic frames (see p p . 145ff., for e x a m p l e ) ; o n e
c o n s e q u e n c e of this v i e w is that M o h a n a n e x t e n d s the d o m a i n of B E into the
postlexical c o m p o n e n t , w h e r e it erases information about syntactic constituency
with the result that the later p h o n e t i c i m p l e m e n t a t i o n rules h a v e n o access to such
structure. O n e r e a s o n a b l e interpretation of r e c e n t w o r k on phrasal p h o n o l o g y is
that syntactic structures on the o n e h a n d and p r o s o d i c or p h o n o l o g i c a l structures
o n the other exist in parallel a n d that p h o n e t i c i m p l e m e n t a t i o n accesses only p r o -
s o d i c / p h o n o l o g i c a l s t r u c t u r e . If p h o n e t i c i m p l e m e n t a t i o n accesses only p r o -
12

sodic structure, however, it follows (trivially) that it c a n n o t access syntactic struc-

ture, and thus B E s e e m s to b e u n n e c e s s a r y at the postlexical level.
Related to this view of the relationship b e t w e e n syntax a n d phrasal p h o n o l o g y
are theories of w o r d structure w h e r e real m o r p h o l o g i c a l (or m o r p h o s y n t a c t i c )
structure is parallel to b u t separate from p h o n o l o g i c a l (or m o r p h o p h o n o l o g i c a l )
structure. Sproat (1985) a r g u e d for such a view, and m o r e recently Inkelas (1989)
h a s a d o p t e d the s a m e position, recasting m o r p h o p h o n o l o g i c a l structure in t e r m s
of p r o s o d i c structure in the lexicon, as discussed a b o v e . A n o b v i o u s n e x t step is
to a s s u m e that p h o n e t i c i m p l e m e n t a t i o n rules only h a v e access to lexical p r o s o d i c
information and not to m o r p h o s y n t a c t i c structure. O n the other h a n d , the evoca-
tion of d i s c o u r s e entities in the p r a g m a t i c s w o u l d h a v e access to m o r p h o s y n t a c t i c
structure. G i v e n that m o r p h o s y n t a c t i c a n d m o r p h o p h o n o l o g i c a l (lexical p r o s o d i c )
structures are separate, o n e could, of c o u r s e , i m a g i n e a scenario w h e r e B E applies,
say, in p r o s o d i c structure b u t not in m o r p h o s y n t a c t i c structure, a n d u n d e r such a
s c e n a r i o o n e or the other set of data p r e s e n t e d in this p a p e r w o u l d then b e u n p r o b -
lematic. W i t h that in m i n d , let us c o n s i d e r the status of B E at e a c h of the levels
separately.
First of all, theories of B E in p h o n o l o g i c a l structure h a v e b e e n p r o p o s e d w h i c h
are u n p r o b l e m a t i c from the p o i n t of view of the p h o n e t i c i m p l e m e n t a t i o n data
discussed in this paper. O n e such idea is the suggestion of Inkelas ( 1 9 8 9 : 5 8 ) that
B E m a y b e reinterpreted " a s a constraint o n w h a t p h o n o l o g i c a l rules m a y refer
t o " and that " i n s t e a d of forcing internal b r a c k e t s to b e erased, w e n e e d only i m -
p o s e the r e q u i r e m e n t on rules that they look only at the highest n o d e [i.e., highest
relative to the current analysis] in p r o s o d i c structure." T h u s , B E is simply a P H O -
N O L O G I C A L locality condition. O b v i o u s l y this view is c o m p l e t e l y c o m p a t i b l e with
Looking into Words 189

the data d e s c r i b e d in this paper. If w e a s s u m e that p h o n e t i c i m p l e m e n t a t i o n ap-

plies to a n n o t a t e d p r o s o d i c structures such as t h o s e in (5), then it s e e m s r e a s o n a b l e
to a s s u m e that p h o n e t i c i m p l e m e n t a t i o n will h a v e a c c e s s to all d o m a i n s repre-
sented in such structures. Of c o u r s e , it will b e u p to p h o n e t i c i m p l e m e n t a t i o n to
d e c i d e w h e t h e r a n d h o w it m a k e s use of such information. F o r e x a m p l e , it w o u l d
b e consistent with the d a t a in this p a p e r to a s s u m e that p r e b o u n d a r y l e n g t h e n i n g
m a y in p r i n c i p l e h a v e a c c e s s to all levels of p r o s o d i c b o u n d a r y b u t in fact is only
sensitive to b o u n d a r i e s at or a b o v e the c o m p o u n d level. W e d o not currently un-
d e r s t a n d e n o u g h a b o u t p h o n e t i c p h e n o m e n a such as p r e b o u n d a r y l e n g t h e n i n g to
say w h a t r a n g e of sublexical p r o s o d i c b o u n d a r i e s m a y b e visible to these p h e -
n o m e n a cross-linguistically, t h o u g h it is clear that at least s o m e sensitivity to such
b o u n d a r i e s is required. It m a y in any event b e a s s u m e d that p h o n e t i c i m p l e m e n -
tation is not c o n s t r a i n e d b y p h o n o l o g i c a l l o c a l i t y .
13

T u r n i n g n o w to B E in the m o r p h o s y n t a c t i c representation, n o t e again that the

original c o n t e n t of lexical integrity w a s specifically c o n c e r n e d with the relation of
m o r p h o l o g y to syntax, so in the parallel structures view of m o r p h o l o g y , it is B E
at the m o r p h o s y n t a c t i c level of representation w h i c h is relevant for lexical integ-
rity as originally c o n s t r u e d . Lexical integrity h a s a l w a y s b e e n a p r o b l e m a t i c prin-
ciple. For o n e thing, as a r g u e d in Sproat ( 1 9 8 8 ) , it is not clear to w h a t extent it is
an e m p i r i c a l c l a i m that syntax c a n n o t affect the internal structure of w o r d s ; t h e o -
ries such as that of B a k e r ( 1 9 8 8 ) w o u l d a p p e a r to violate it m a s s i v e l y on at least
the m o s t simplistic v i e w . In any event, it s e e m s clear that, if m y interpretation
1 4

of the d a t a from so-called a n a p h o r i c islands is correct, p r a g m a t i c s at least m u s t

b e able to l o o k i n s i d e w o r d s , s u g g e s t i n g that real m o r p h o l o g i c a l c o n t e n t m a y b e
visible.
Of c o u r s e , m o r p h o s y n t a c t i c B E has b e e n a p p e a l e d to to p e r f o r m other services
b e s i d e s d e r i v i n g lexical integrity. S o , as P e s e t s k y ( 1 9 7 9 ) p o i n t e d o u t , B E s u b -
1 5

s u m e s the adjacency condition of Allen ( 1 9 7 8 ) . Recall that u n d e r the adjacency

condition, a m o r p h o l o g i c a l rule attaching X to [[Y]Z] m a y m a k e reference to
p r o p e r t i e s of the m o r p h e m e Z , but it m a y m a k e n o reference to any p r o p e r t i e s of
Y. A s P e s e t s k y argued, B E is even stronger than the adjacency c o n d i t i o n and thus
properly s u b s u m e s it: since all b r a c k e t s e x c e p t the o u t e r m o s t pair in [ Y Z ] h a v e
b e e n e r a s e d , the rule attaching X can only refer to the o u t e r m o s t b r a c k e t s and
1 6

their labels and is b l o c k e d from referring even to Z , a n d a fortiori to Y. T h u s , an

affix like -merit can attach to a w o r d like enjoy b e c a u s e the b r a c k e t s s u r r o u n d i n g
enjoy are labeled V a n d -ment is s u b c a t e g o r i z e d to attach to V. It is irrelevant that
the m o r p h e m e adjacent to -ment, n a m e l y joy, is labeled N ; indeed, p r o p e r t i e s of
m o r e d e e p l y e m b e d d e d m o r p h e m e s w h i c h are not t h e m s e l v e s p e r c o l a t e d to h i g h e r
n o d e s are systematically irrelevant to the a t t a c h m e n t of m o r p h e m e s later on in the
derivation. B E predicts this systematic m y o p i a in that it requires that m o r p h o l o g i -
cal derivations only l o o k at features of the t o p m o s t n o d e .
T h e p r o b l e m with a d d u c i n g these c o n s i d e r a t i o n s as e v i d e n c e for the necessity
of B E in m o r p h o l o g y is that there are all sorts of p l a c e s in g r a m m a r w h e r e only
190 Richard Sproat

the t o p m o s t n o d e is relevant. In syntax, for e x a m p l e , if o n e is c h e c k i n g to see that

a verb's subcategorization r e q u i r e m e n t s are satisfied, o n e only n e e d s to look at the
t o p m o s t n o d e of the verb's sister(s) to establish that those n o d e s m e e t the sub-
categorization r e q u i r e m e n t s ; alternatively, o n e m a y w a n t to say that o n e looks at
the h e a d of the sister(s), but that can apparently a l w a y s be recast as an instance of
l o o k i n g at the t o p m o s t n o d e u n d e r the a s s u m p t i o n that the h e a d ' s features are
inherited by all p h r a s e s w h i c h it h e a d s . Verbs m a y specify a subcategorization for
an NP, but they apparently m a y not specify a subcategorization for an N P w h i c h
is modified by at least o n e adjective, or an N P w h i c h is followed by a relative
clause. S u c h information is p r e s u m a b l y a v a i l a b l e — n o o n e to m y k n o w l e d g e h a s
p r o p o s e d the application of B E in the syntax to render it u n a v a i l a b l e — a n d yet 1 7

information about a n y t h i n g other than the t o p m o s t n o d e s of sisters, and w h a t e v e r

information is inherited b y t h e m from their h e a d s , is systematically irrelevant to
subcategorization c o n s i d e r a t i o n s . P r e s u m a b l y this is b e c a u s e subcategorization
frames are restricted to specifying a set of slots and the features of, but n o struc-
tural information about, the fillers of t h o s e slots. H o w e v e r this is to b e stated ex-
actly, it w o u l d s e e m to s u b s u m e the adjacency condition every bit as well as B E .
A l t h o u g h I h a v e called into q u e s t i o n the usefulness a n d c o r r e c t n e s s of w h a t I
h a v e t e r m e d the strong interpretation of B E , the ultimate c o n c l u s i o n of this p a p e r
is, I think, a positive o n e . T h e data p r e s e n t e d h e r e give a better sense of the kinds
of w a y s in w h i c h postlexical c o m p o n e n t s m a y m a k e reference to w o r d structure,
and they suggest h o w the correct versions of B E — a n d lexical integrity m o r e gen-
e r a l l y — s h o u l d look. I w o u l d like to e n d b y suggesting that a fruitful area of future
research on m o r p h o l o g y w o u l d b e a systematic investigation of the relevance of
m o r p h o l o g i c a l structure to p h o n e t i c i m p l e m e n t a t i o n a n d p r a g m a t i c s . A s far as I
know, a l t h o u g h p h o n e t i c i m p l e m e n t a t i o n is at least d i s c u s s e d in the lexical p h o -
n o l o g y literature, w h a t d i s c u s s i o n there h a s b e e n h a s usually b e e n at a rather high
level, and even w h e n it d o e s treat the issue in s o m e detail, it has focused on the
p h o n e t i c i m p l e m e n t a t i o n of purely p h o n o l o g i c a l information, with n o attention to
the possible relevance of w o r d structure itself to p h o n e t i c s . T h e r e are few if any
studies from the lexical m o r p h o l o g y tradition that a d d r e s s the relationship b e -
t w e e n m o r p h o l o g y and p r a g m a t i c s . W o r d s d o not exist in a v a c u u m , yet research-
ers h a v e on the w h o l e t e n d e d to take a rather n a r r o w view of the k i n d s of data that
are relevant to theories of m o r p h o l o g y , usually limiting t h e m s e l v e s to at m o s t
discussing the relation b e t w e e n m o r p h o l o g y and p h o n o l o g y or m o r p h o l o g y and
syntax. T h e field is ready to b e c o m e m o r e c o s m o p o l i t a n .

ACKNOWLEDGMENTS

I thank Sharon Hargus and an anonymous reviewer for helpful comments on this article.
Looking into Words 191

NOTES

1
Not a low-level phonological rule, contra the analysis of l\l variation in Halle and
Mohanan (1985) and Mohanan (1986). We make the common assumption (cf. Liberman
and Pierrehumbert, 1984) that phonetic implementation rules introduce continuous varia-
tion whereas phonological rules of any kind, including low-level ones, introduce categori-
cal variation.
2
N o t e that all speakers pronounced beel equator as a compound—i.e., deaccenting
equator—except speaker AD, who did place some prominence on equator in some of the
examples; this accenting difference did not appear to show up in the form of a different
trend for speaker AD.
3
If it is not visible, then the whole compound beel equator should count as a tetrasylla-
ble word, and the rime of beel should be shorter in this case than in the disyllabic beeling,
by Lehiste's argument.
4
On the interpretation of Booij and Lieber (this volume) and apparently also Cohn
(1989), metrical structure—prosodic structure, in their terms (see Inkelas, 1989, for a dis-
cussion of the terms prosodic and metrical)—is the only structure to which lexical pho-
nological rules can refer. On that view, then, the only representation of morphological
structure that the phonology sees is the metrical structure, so if BE were to apply so as to
render morphological information invisible to phonological structure, it would have to
erase metrical structure. Of course, metrically weak morphemes such as affixes might be
incorporated into whatever metrical structure they attach to, so that the resulting structure
is indistinguishable from a metrical structure derived from a monomorphemic input. So
BE might in such cases follow from metrical considerations alone. On the other hand, there
is no reason to assume on purely metrical grounds that the phonological wordhood of the
components of a compound should be erased.
5
1 note in passing that Booij (1983) uses metrical structure to account for some Dutch
data which appear to compromise lexical integrity. Also Mohanan (1986:60 n. 10) inti-
mates that such a compromise might be possible since "postlexical phonological opera-
tions have access to the phonological structures exiting from the lexicon (e.g. stress trees,
syllable trees)."
6
1 continue to use Postal's original terminology and talk of reference into words. How-
ever, in doing so I am being rather terminologically loose: the correct way to describe the
situation is to say that an anaphor is used to refer to a discourse entity evoked by a mor-
pheme which is morphologically contained within another word.
7
There are a number of possible reasons for this: as Paul Kiparsky has noted (personal
communication) pronouns are closed-class items and as such are generally barred from
undergoing morphological derivation. Of course, as an anonymous reviewer points out,
there are derived forms of prepositions such as thereto, thereabout. However, it is fair to
say that such formations are not productive in present-day English.
A somewhat related possibility is the more general statement that functional categories
and their projections cannot serve as the basis of morphological derivation in English. This
would link the unavailability of *r/z<?ra-hater with the unacceptability of *the The Bronx
hater (Fabb, 1984); note that on recent views of phrase structure, in particular Abney
(1987), a fully specified NP is considered to be a projection of a functional category (in
this case the determiner the), and therefore the full DP The Bronx could not be morpho-
192 Richard Sproat

logically contained within the compound by the proposed prohibition. See Sproat (1985,
1988) for a third possible account.
Finally, note that if a language lacks whatever grammatical constraint of English pre-
vents pronouns from occurring within words, then we would expect pronouns which do
occur, say as the lefthand member of a compound, to be able to corefer with antecedents
outside the word. This is exactly what we find in Sanskrit (Gillon, 1990).

(i) rdgasya-an-upayoge katham [tat-saktih] upayujyate

passion-non-use how it-potentiality use
If there is no causal role for passion, in what way does its potentiality have a
causal role?'

Of course, it is necessary to check that the subject is resolving the reference correctly;
8

see McKoon et al. (1993) and Ward et al. (1991) for details of the methods used for veri-
fying that reference was resolved. Note also that topicality was shown in the cited work to
have a significant (though reduced) effect on reference to full noun phrases in discourses
similar to (10) (e.g., lately he's taken up hunting deer).
T h e fact that BE prevents -al and -ion from attaching to refuse and destroy in the
9

syntax—as noted by Mohanan in the above-quoted passage—has, of course, nothing to do

with anaphoric islandhood, and is therefore largely irrelevant to the points discussed in this
article. Still, it is worth pointing out that as far as one can tell, the prohibition on attachment
of derivational affixes in the syntax has essentially zero empirical content. What would go
wrong if such affixes W E R E attached in the syntax?
I . e . , a lexical relationship for which one can find many instances: for any country, one
10

can find (or construct) a provenance term related to that country.

Thanks to an anonymous reviewer for pointing out some unclarities in an earlier ver-
1 1

sion of this discussion.

N o t e that data discussed in Pierrehumbert and Beckman (1988), among others, sug-
12

gest that phonetic implementation M U S T access prosodic structure—apparently contra Mo-

hanan (1986:175) who suggests, "Thus phonological structures at all levels exist only in
the lexical and syntactic modules, and are dissolved in the implementational module."
Another view possibly compatible with the phonetic data presented in this paper is the
1 3

view intimated in Booij and Rubach (1987) that bracketings constructed at stratum II are
not erased by BE, since BE only applies at cyclic strata and stratum II is a noncyclic stra-
tum. A further compatible view is that of Hammond (1984). Hammond argues that BE
does not apply to compounds, although it applies to other morphological constructions. In
general, any version of lexical phonology and morphology which assumes a sufficiently
weak interpretation of BE will be compatible with more or fewer of the data presented here.
For a particularly lucid discussion of the various arguments for and against lexical
1 4

integrity, see LeRoux (1988).

Thanks to Sharon Hargus for reminding me of this point.
1 5

I n Pesetsky's theory, BE applied at the end of each cycle.

1 6

Again, Mohanan's proposal that BE apply in the syntax was intended to rule out pho-
1 7

netic implementation rules from accessing syntactic information.

Looking into Words 193

REFERENCES

Abney, S. (1987). The English Noun Phrase in its Sentential Aspect. Doctoral dissertation,
Massachusetts Institute of Technology, Cambridge.
Allen, M. (1978). Morphological Investigations. Doctoral dissertation, University of Con-
necticut, Storrs.
Baker, M. (1988). Incorporation: A Theory of Grammatical Function Changing. Univer-
sity of Chicago Press, Chicago.
Booij, G. (1983). Coordination and reduction in complex words: A case for prosodic
phonology. Vrije Universiteit Working Papers in Linguistics 3, Vrije Universiteit,
Amsterdam.
Booij, G. and Rubach, J. (1987). Postcyclic versus postlexical rules in lexical phonology.
Linguistic Inquiry 18, 1 - 4 4 .
Chomsky, N. (1970). Remarks on nominalization. In Readings in English Transforma-
tional Grammar (R. Jacobs and P. Rosenbaum, eds.), pp. 1 8 4 - 2 2 1 . Ginn, Waltham,
Mass.
Chomsky, N., and Halle, M. (1968). The Sound Pattern of English. Harper and Row, New
York.
Cohn, A. (1989). Stress in Indonesian and bracketing paradoxes. Natural Language and
Linguistic Theory 7, 167-216.
Corum, C. (1973). Anaphoric peninsulas. In Papers from the Regional Meeting of the Chi-
cago Linguistic Society 9, 8 9 - 9 7 .
Di Sciullo, A. M., and Williams, E. (1987). On the Definition of Word. MIT Press, Cam-
bridge, Mass.
Downing, P. (1977). On the creation and use of English compound nouns. Language 53,
810-842.
Fabb, N. (1984). Syntactic Affixation. Doctoral dissertation, Massachusetts Institute of
Technology, Cambridge.
Gillon, B. (1990). Sanskrit word formation and context free rules. Toronto Working Papers
in Linguistics 11(2).
Halle, M., and Mohanan, K. P. (1985). Segmental phonology of Modern English. Linguis-
tic Inquiry 16, 57 - 1 1 6 .
Hammond, M. (1984). Level ordering, inflection, and the righthand head rule. MIT Work-
ing Papers in Linguistics 7, 3 3 - 5 2 .
Hargus, S. (1985). The Lexical Phonology of Sekani. Doctoral dissertation, University of
California, Berkeley.
Inkelas, S. (1989). Prosodic Constituency in the Lexicon. Doctoral dissertation, Stanford
University, Stanford, Calif.
Kiparsky, P. (1982a). Lexical morphology and phonology. In Linguistics in the Morning
Calm, (I.-S. Yang, ed.), pp. 3 - 9 1 . Hanshin, Seoul.
Kiparsky, P. (1982b). Word formation and the lexicon. Mid-America Linguistics Confer-
ence 3 - 2 9 .
Kiparsky, P. (1985). Some consequences of lexical phonology. Phonology Yearbook 2,
85-138.
194 Richard Sproat

Lakoff, G., and Ross, J. (1972). A note on anaphoric islands and causatives. Linguistic
Inquiry 3, 1 2 1 - 1 2 5 .
Le Roux, C. (1988). On the Interface of Morphology and Syntax: Evidence from Verb-
Particle Combinations in Afrikaans. Master's thesis, University of Stellenbosch, Stel-
lenbosch, South Africa.
Lehiste, I. (1972). The timing of utterances and linguistic boundaries. Journal of the Acous-
tical Society of America 51, 2 0 1 8 - 2 0 2 4 .
Lehiste, I. (1980). Phonetic manifestation of syntactic structure in English. Annual Bulletin
of the Research Institute of Logopaedics and Phoniatrics 14, 1 - 2 7 .
Levi, J. (1978). The Syntax and Semantics of Complex Nominals. Academic Press, New
York.
Liberman, M., and Pierrehumbert, J. (1984). Intonational invariants under changes in pitch
range and length. In Language Sound Structure. (M. Aronoff and R. Oehrle, eds.),
pp. 1 5 7 - 2 3 3 . MIT Press, Cambridge, Mass.
Lieber, R. (1984). Grammatical rules and sublexical elements. In Papers from the Para-
session on Lexical Semantics, pp. 187-199. Chicago Linguistic Society, Chicago.
Lieber, R. (1991). Deconstructing Morphology: Word Formation in a Government-Binding
Syntax. The University of Chicago Press, Chicago.
McCawley, J. (1989). 1989 Linguistic Flea Circus. Unpublished manuscript, University of
Chicago.
McKoon, G , Ratcliff, R., Ward, G , and Sproat, R. (1990). Structural and Discourse Ma-
nipulations of Salience in the Interpretations on Anaphora. Paper presented at the
Third Annual CUNY Conference on Human Sentence Processing, New York, N.Y.
McKoon, G , Ward, G , Sproat, R., and Ratcliff, R. (1993). Morphosyntactic and pragmatic
factors affecting the accessibility of discourse entities. Journal of Memory and Lan-
guage 32, 1 - 2 0 .
Mohanan, K. P. (1982). Lexical Phonology. Doctoral dissertation, Massachusetts Institute
of Technology, Cambridge.
Mohanan, K. P. (1986). The Theory of Lexical Phonology. Reidel, Dordrecht.
Nespor, M., and Vogel, I. (1986). Prosodic Phonology. Foris, Dordrecht.
Pesetsky, D. (1979). Russian Morphology and Lexical Theory. Unpublished manuscript,
Massachusetts Institute of Technology, Cambridge.
Pierrehumbert, J., and Beckman, M. (1988). Japanese Tone Structure. MIT Press, Cam-
bridge, Mass.
Postal, P. (1969). Anaphoric islands. Papers from the Regional Meeting of the Chicago
Linguistic Society 5, 2 0 5 - 2 3 9 .
Reinhart, T. (1983). Anaphora and Semantic Interpretation. University of Chicago Press,
Chicago.
Selkirk, E. (1980). The role of prosodic categories in English word stress. Linguistic In-
quiry 11, 5 6 3 - 6 0 6 .
Shibatani, M., and Kageyama, T. (1988). Word formation in a modular theory of grammar.
Language 64, 451 - 4 8 4 .
Siegel, D. (1974). Topics in English Morphology. Doctoral dissertation, Massachusetts In-
stitute of Technology, Cambridge.
Silverman, K. (1988). Utterance-internal prosodic boundaries. Proceedings of the Austra-
lian International Conference on Speech Science and Technology 2, 8 6 - 9 1 .
Looking into Words 195

Simpson, J. (1983). Aspects of Warlpiri Morphology and Syntax. Doctoral dissertation,

Massachusetts Institute of Technology, Cambridge.
Spencer, A. (1988). Bracketing paradoxes and the English lexicon. Language 64, 6 6 3 -
682.
Sproat, R. (1985). On Deriving the Lexicon. Doctoral dissertation. Massachusetts Institute
of Technology, Cambridge.
Sproat, R. (1988). On anaphoric islandhood. In Theoretical Morphology (M. Hammond
and M. Noonan, eds.), pp. 291 - 3 0 1 . Academic Press, San Diego.
Sproat, R., and Fujimura, O. (1989). Articulatory Evidence for the Non-categoricalness of
English III allophones. Paper presented at the winter meeting of the Linguistic Society
of America, Washington, D.C.
Sproat, R., and Fujimura, O. (1993). Allophonic Variation in English III and its implications
for phonetic implementation. Journal of Phonetics 21.
Sproat, R., and Ward, G. (1987). Pragmatic considerations in anaphoric island phenomena.
Papers from the Regional Meeting of the Chicago Linguistic Society 23, 3 2 1 - 3 3 5 .
van Santen, J. (1989). Modeling Contextual Effects on Vowel Duration. I. Description of
Individual Factors. Unpublished manuscript, AT&T Bell Laboratories.
Ward, G., Sproat, R., and McKoon, G. (1991). A pragmatic analysis of so-called anaphoric
islands. Language 6 7 , 4 3 9 - 4 7 4 .
Watt, W. (1975). The indiscreetness with which impenetrables are penetrated. Lingua 37,
95-128.
ON THE WORD LEVEL

TONI BOROWSKY
Department of Linguistics
University of Sydney
Sydney, New South Wales 2006, Australia

1. INTRODUCTION

In this article I p u t forward a m o d e l of lexical p h o n o l o g y in w h i c h the p r o p e r t i e s

of the t w o major levels follow from the fact that at the W o r d level all the p h o n o -
logical p r o c e s s e s p r e c e d e , rather than follow, all m o r p h o l o g i c a l o p e r a t i o n s . A t the
S t e m level, the m o r p h o l o g y p r e c e d e s a n d feeds p h o n o l o g y in a cyclic fashion, as
is usual in any standard lexical p h o n o l o g y .
T h e theory I p r o p o s e c h a l l e n g e s t w o b a s i c a s s u m p t i o n s in lexical p h o n o l o g y :
First, m o s t w o r k in lexical p h o n o l o g y is c h a r a c t e r i z e d — i f not explicitly, then im-
p l i c i t l y — b y the idea that w h e r e there is interaction b e t w e e n the m o r p h o l o g i c a l
and t h e p h o n o l o g i c a l m o d u l e s , the p h o n o l o g y follows the m o r p h o l o g y . I offer
1

instead a m o d e l in w h i c h this o r d e r i n g is reversed o n the W o r d level. S e c o n d , I

c o n s i d e r the idea that the t w o classes of affixes c a n b e d i a g n o s e d on p u r e l y p h o -
nological g r o u n d s . Affixes at the W o r d level are generally defined negatively, in
t e r m s of the p h o n o l o g i c a l rules (usually t h o s e of t h e first level) w h i c h d o not apply
at this l e v e l . T h i s a l o n e is offered as an e x p l a n a t i o n for t h e fact that these affixes
2

display w h a t is k n o w n as " o p e n j u n c t u r e " w h i l e the affixes of the first level dis-

play " c l o s e j u n c t u r e . " Yet in m o s t lexical p h o n o l o g i e s , even t h o u g h it is, i m p l i c -
itly at least, r e c o g n i z e d that the t w o levels h a v e quite distinct p r o p e r t i e s , they are
a s s u m e d to b e the s a m e . In this article I s h o w that the differences in the p h o n o l o g y
of the t w o lexical levels are m u c h m o r e than the differences in the sets of rules
w h i c h apply or d o not apply at e a c h level. I claim that the t w o levels are c h a r a c -
terized b y distinct interfacing b e t w e e n the m o r p h o l o g y a n d the p h o n o l o g y . T h e

199
Phonetics and Phonology, Volume 4 Copyright © 1993 by Academic Press, Inc.
Studies in Lexical Phonology All rights of reproduction in any form reserved.
 200 Toni Borowsky

p h o n o l o g i c a l p r o p e r t i e s w h i c h c h a r a c t e r i s e e a c h level are d e r i v e d from the inter-

action b e t w e e n the t w o m o d u l e s .
In (1) is s h o w n a d i a g r a m of the p r o p o s e d m o d e l . In relevant m o r p h o l o g i c a l
respects, it can b e c o n s t r u e d as identical w i t h the m o d e l p r o p o s e d in Selkirk
( 1 9 8 2 ) . T h e r e are t w o lexical d o m a i n s , a n d only t w o , c o r r e s p o n d i n g to Stem level
( = Level 1) and W o r d level ( = Level 2). (See also Aronoff and Sridhar, 1983;
Sproat, 1985).

(1) - LEXICON -
- PHONOLOGY-
Morphological Lexical
operations, phonology,
Stem/Level 1 Level 1

Morphological
Lexical
operations,
phonology
Word/Level 2

Postlexical
phonology

T h i s m o d e l is d r a m a t i c a l l y different from the available alternate m o d e l s in

w h i c h it is a s s u m e d that the W o r d level is either cyclic like the S t e m level (e.g.,
B o r o w s k y , 1986; Kiparsky, 1982) or n o n - or p o s t c y c l i c (e.g., Booij and R u b a c h ,
1987; H a l l e and M o h a n a n , 1985). I p r e s e n t a s c h e m a t i z e d version of these m o d e l s
in (2). In the c o u r s e of this article w e shall see m a n y r e a s o n s to reject this sort of
model.

(2) - LEXICON
PHONOLOGY-
Morphological
Lexical
operations,
phonology
Level /

Morphological
Lexical
operations,
phonology
Level n

Postlexical
phonology

T h e t w o lexical d o m a i n s exhibit m a n y differences w h i c h are, in the theory in

(1), a direct result of the different t y p e of interaction b e t w e e n the p h o n o l o g i c a l
On the Word Level 201

and m o r p h o l o g i c a l s y s t e m s of e a c h d o m a i n . O n the first d o m a i n , the S t e m level,

the m o r p h o l o g y and p h o n o l o g y interact as usual in standard lexical p h o n o l o g y .
E v e r y m o r p h o l o g i c a l operation feeds into the p h o n o l o g i c a l system, and the result
is r e t u r n e d to the m o r p h o l o g i c a l system. T h e p h o n o l o g y of this level exhibits the
collection of p r o p e r t i e s c o n s i d e r e d to b e diagnostic of lexical p h o n o l o g y : that is,
the rules are structure-preserving, are cyclic, a n d o b e y the strict c y c l e condition.
After all Level 1 o p e r a t i o n s are c o m p l e t e d , the resulting f o r m s c o m p l e t e a n o t h e r
circuit t h r o u g h the p h o n o l o g i c a l system before any Level 2 affixation takes p l a c e .
T h i s , the w o r d c y c l e , is the last p h o n o l o g i c a l d o m a i n in the lexicon and constitutes
w h a t I call the W o r d level. R u l e s of this level m a y b e n o n - s t r u c t u r e - p r e s e r v i n g .
3

T h e W o r d level is not cyclic in the usual sense: it d o e s not s h o w e v i d e n c e of strict-

cyclic effects. S i n c e I will s h o w that, at least for English, there is n o e v i d e n c e that
there is any lexical p h o n o l o g y b e y o n d this o n e cycle at the W o r d level, there is
also n o e v i d e n c e to d e t e r m i n e w h e t h e r or not it is cyclic. Word-level affixes m a y
be attached, b u t n o further lexical p h o n o l o g y takes place. It thus a p p e a r s that on
this level p h o n o l o g y p r e c e d e s b u t d o e s not follow m o r p h o l o g y . T h e n e x t cycle
t h r o u g h the p h o n o l o g y , after Word-level m o r p h o l o g y , is the first d o m a i n of the
postlexical p h o n o l o g y — t h e p h o n o l o g i c a l w o r d .
T h e article is structured as follows: I b e g i n by s h o w i n g , on the basis of data
from E n g l i s h a n d G e r m a n , that there exists a single cycle at w h i c h the rules w h i c h
apply at the W o r d level m u s t apply. T h i s c y c l e m u s t p r e c e d e any affixation at this
level. I then g o on to suggest that there is n o other p h o n o l o g i c a l d o m a i n at the
W o r d level on the basis of a c o n s i d e r a t i o n of the nature of Level 2 p h o n o l o g i c a l
p r o c e s s e s as well as the structure of forms derived at this level.
M y p r o p o s a l , therefore, h a s t w o parts. I m a k e an e m p i r i c a l point: that the W o r d
is a p h o n o l o g i c a l d o m a i n . T h e r e m u s t b e a W o r d cycle at Level 2 before any
m o r p h o l o g y , a n d this c y c l e a p p e a r s to b e the only p h o n o l o g i c a l cycle for w h i c h
there is any e v i d e n c e at this level. B a s e d on this observation, I m a k e a theoretical
proposal about the interplay b e t w e e n the t w o m o d u l e s that c o m p r i s e the lexicon
w h i c h bears crucially on h o w the lexical p h o n o l o g y is to b e structured and inter-
preted. In the c o u r s e of the discussion I s h o w that noncyclicity (as s u g g e s t e d by
Booij and R u b a c h , 1985; Halle a n d M o h a n a n , 1985; Kiparsky, 1985; and others)
is not an a d e q u a t e e x p l a n a t i o n of the p r o p e r t i e s of the W o r d level and is i n c o m -
patible with the facts d i s c u s s e d in this paper. I then c o n s i d e r an alternative analysis
of the facts w h i c h has b e e n s u g g e s t e d to m e and show it to b e i n a d e q u a t e . T h e
A p p e n d i x p r o v i d e s further e x a m p l e s of rules of the type discussed in the b o d y of
the article.

2. T H E W O R D C Y C L E

In B o r o w s k y ( 1 9 8 6 ) I argue, on the basis of the m o r p h o p h o n e m i c alternations

exemplified in (3), that there m u s t be a single cycle through the p h o n o l o g y after
202 Toni Borowsky

all Level 1 p r o c e s s e s , but before any L e v e l 2 m o r p h o l o g y . T h i s cycle c o r r e s p o n d s

to the m o r p h o l o g i c a l o p e r a t i o n in Selkirk (1982) w h i c h c h a n g e s a S t e m form to a
W o r d form. It is on this c y c l e that the p r o c e s s w h i c h a c c o u n t s for these alterna-
tions take p l a c e .

(3) a. [ + son]-»[ + syll]/C ]

] ]aff2 ]aff 1
wonder wondering wondrous
cycle cycling cyclic
theater theatergoer theatrical
meter metering metric, metrical
rhythm rhythm-and-blues rhythmic
anger angry
center centering central centrality
[saykl] [sayklirj] 4
[siklik]
[sayklirj] *[siklik]
b. big - > 0 / N ] a ( w h e r e N = h o m o r g a n i c nasal)
] ]aff2 ]aff 1
long longing elongate, longitude,
strong strongly strongest
bomb bombing bombard
crumb crumby crumble
thumb thumbing thimble
[brjirj] [ibrjget]
c. n —> 0 / m ]
] ]aff2 ]aff 1
condemn condemning condemnation
autumn autumning autumnal
hymn hymning hymnal
[him] [himirj] [himnal]
*[himnirj] *[himal]
d. £-»0/ [+nasal]]
] ]aff 2 ]aff 1
resign resigning resignation
sign signer signature
paradigm paradigmatic
[sayn] [saynar] [signacur]
c
[saygn] *[saygner] *[sinocur]

N o t e that t h o u g h the details of the c h a n g e s induced in e a c h set of alternations

are different, the patterns of c h a n g e s are identical; all are found in the word-final
e n v i r o n m e n t , as s h o w n in the first c o l u m n ; they a p p e a r too before all Level 2
suffixes, second c o l u m n . Before Level 1 suffixes, as s h o w n by the e x a m p l e s in the
third c o l u m n , n o n e of these p h e n o m e n a take place.
On the Word Level 203

T h e s e facts are m o s t a d e q u a t e l y e x p l a i n e d as d u e to p r o s o d i c licensing (Ito,

1986). E a c h c h a n g e is a syllabification-induced c h a n g e w h i c h o c c u r s to license,
or eliminate, material left u n l i c e n s e d b y t h e Level 1 p h o n o l o g y . T h a t is, at Level 2
w h e n t h e final c o n s o n a n t s of Level 1, p r e v i o u s l y licensed b y extrametricality, b e -
c o m e visible t o t h e p h o n o l o g y , these c h a n g e s are i n d u c e d . T h u s t h o u g h t h e facts
in (3) are set o u t as if illustrating different rules, I a s s u m e that they a r e actually
the result of a single p r o c e s s : p r o s o d i c licensing. B e a r i n g this in m i n d , I c o n c e n -
trate m y d i s c u s s i o n o n t h e first of t h e s e e x a m p l e s , S o n o r a n t Syllabification, d r a w -
ing attention t h r o u g h o u t to t h e similarities with t h e other alternations.
Final s o n o r a n t s b e c o m e syllabic at t h e w o r d e d g e . In t h e e x a m p l e s given in (3a)
w e see that t h o u g h these sonorants are consistently nonsyllabic w h e r e followed by
Level 1 suffixes, as in metric, they are a l w a y s syllabic w o r d finally, as in meter,
and they m a y optionally b e syllabic before all Level 2 affixes, as in metering.
N o t i c e that this alternation is n o t pandialectal; s o m e speakers d o n o t p r o n o u n c e
these f o r m s with a syllabic sonorant if followed b y a vowel-initial L e v e l 2 affix.
T h i s is t h e result of a postlexical desyllabification rule w h i c h is optional in t h e
dialects u n d e r discussion. T h e crucial thing to explain h e r e is t h e fact that there is
never a syllabic s o n o r a n t before a Level 1 affix o r a nonsyllabic s o n o r a n t w o r d
finally. T h e only p l a c e optionality arises is before W o r d level affixes. T h a t is,
t h o u g h both alternatives are available in s o m e dialects before Level 2 affixes, thus
[miytrirj] o r miytrirj], n o dialect ever h a s a syllabic sonorant in a n y Level 1 d e -
rivative *[met9rik]. ( N o t e also that t h e optionality of syllabicity or nonsyllabicity
of s o n o r a n t s d o e s n o t e x t e n d to a n y of t h e other p r o c e s s e s : t h e forms hymning or
signing a r e n e v e r available as *[himnirj] o r *[signirj].) T h i s a r r a n g e m e n t of facts
suggests that t h e p r o c e s s b y w h i c h s o n o r a n t s b e c o m e syllabic d o e s n o t take p l a c e
at t h e first level of t h e lexicon. A t the S t e m level, final sonorants b e c o m e onsets
of vowel initial suffixes, as in (4).

(4)

center sen (tr)

<j cr

centr + al
AA
sen tr dl

In the s a m e w a y t h e c o n s o n a n t s participating in t h e other alternations are licensed

at Level 1 b y regular syllabification: the c o n s o n a n t s , s h o w n in boldface, in elong-
ate, solemnity, a n d autumnal turn u p as onsets, w h i l e the g in signature is li-
c e n s e d in the c o d a b y t h e syllabification of t h e following n into t h e suffix onset.
At Level 2 t h e final sonorant is syllabic even before vowel-initial suffixes
attached at this level. T h e p r o c e s s w h i c h m a k e s final sonorants syllabic m a k e s
n o reference to t h e following suffix; it a p p e a r s to see only t h e w o r d as it
e m e r g e s from t h e Level 1 p h o n o l o g y . T h e final c o n s o n a n t is n o t licensed as the
onset of t h e suffix syllable. T h e other c o n s o n a n t s delete in precisely t h e s a m e
circumstances.
204 Toni Borowsky

All these alternations are easily a c c o u n t e d for if p r o s o d i c licensing is enforced

before the Level 2 affixes are a d j o i n e d — t h a t is, on the W o r d cycle. T h e n the t w o
different surface e n v i r o n m e n t s , word-final and p r e c e d i n g Level 2 affixes, are in
fact the s a m e w h e n the p r o c e s s e s occur, as s h o w n in (5).

(5) o-

A
center sen (tr) Level 1

a o~
Level 2: word cycle
sen tr Phonology: Sonorant Syllabification

(7 <J

sen t r + ing Morphology

cr a o~ <x o~

A it
[sen t r] [sen
An
t r + irj]

(or [sen trirj] Postlexical: optional desyllabification ) 5

CT O"

Since p r o s o d i c licensing takes p l a c e before any Level 2 affixation, the option of

incorporating final stray c o n s o n a n t s as o n s e t s of following suffixal syllables is
simply not available.
Before i n t r o d u c i n g additional e v i d e n c e found in E n g l i s h for the w o r d cycle, let
m e turn to s o m e similar facts from G e r m a n .

2.1. G e r m a n Morphophonemics

C o u n t e r p a r t s of t w o of the E n g l i s h p r o c e s s e s s h o w n a b o v e exist as well in Ger-

m a n . T h e G e r m a n facts present the s a m e p r o b l e m s as the E n g l i s h o n e s d o and
m u s t b e analysed in the s a m e way.
In G e r m a n , syllabic and nonsyllabic s o n o r a n t s alternate, and g deletes after [rj]
in p r o c e s s e s similar to the E n g l i s h o n e s . Before Level 1 affixes, sonorants are
always nonsyllabic. Word-final s o n o r a n t s are obligatorily syllabic w h e n absolute
final and optionally syllabic w h e n followed b y Level 2 affixes. C o m p a r e the t w o 6

patterns in (6).

(6) L E V E L 1: obligatorily nonsyllabic

filtr + ieren ' t o filter'
registr 4- ieren 'to register'
On the Word Level 205

metr + isch 'metrical'

(Gelt)wechsl + er 'changer'
Verdunklung 'blackout'

L E V E L 2: syllabic finally, else optional

silber + ig (silbr + ig) 'silvery'
zucker + ig (zuckr + ig) 'sugary'
schmuddle + ig (etc.) 'grimy'
Verwechselung 'mixup'
Verdunkelung 'darkening'

Similarly, in word-final position, a n d before Level 2 affixes, [g] deletes after [rj],
as well as w h e n it c a n n o t b e syllabified (Hall, 1989b).

(7) L E V E L 1:
diphthong + ier + en 'to d i p h t h o n g i z e '
tang + ier + en 'to affect'
laryng + al 'laryngeal'

versus

L E V E L 2:
Spreng + ung 'explosion'
sing 4- e '(I) s i n g '
Jung + en 'boys (nom.)'
Jung + ling '(a) y o u t h '

T h e analysis of b o t h sets of facts is identical to the o n e I h a v e given for the

E n g l i s h alternations. P r o s o d i c licensing takes p l a c e before affixation at Level 2.
T h u s the relevant c o n s o n a n t s are not syllabified into the onset of o n e of these
affixes a n d are licensed b y S o n o r a n t Syllabification for final sonorants, or deleted,
as for final g s after [rj]. T h e optional p r o n u n c i a t i o n with desyllabified s o n o r a n t s
arise by a postlexical reduction p r o c e s s . T h u s , I c o n c l u d e that G e r m a n h a s a w o r d
cycle at Level 2 j u s t as d o e s E n g l i s h .
A n o t h e r e x t r e m e l y interesting e x a m p l e of a w o r d - c y c l e rule in G e r m a n is the
[c, x] alternation. Briefly, the s e g m e n t s [ 9 ] a n d [x] are in c o m p l e m e n t a r y distri-
bution. T h e velar fricative o c c u r s after [ + b a c k ] v o w e l s w h i l e the palatal fricative
o c c u r s after [ — b a c k ] v o w e l s a n d e l s e w h e r e . E x a m p l e s are given in (8).

(8) [9] [x]

ich T Buch 'books'
hochlich 'highly' hoch 'highly'
Kuche 'kitchen' Koch 'cook'
riechen 'to smell' rauchen 'to s m o k e '
sicher 'sure' Sprache 'language'
206 Toni Borowsky

T h i s distribution can b e stated informally as in ( 9 ) . T h e analysis I present b e l o w

7

is taken from Hall ( 1 9 8 9 a ) after Bloomfield ( 1 9 3 0 ) .

(9) f-»jc/[+back]

T h e m i n i m a l pairs in (10) a p p e a r to b e c o u n t e r e x a m p l e s to this rule. In (10a) the

palatal fricative o c c u r s after [ + b a c k ] v o w e l s w h e r e the [x] is e x p e c t e d . In (10b)
the [x] o c c u r s as it should.

(10) a.
Kuhchen [ku:90n] 'little c o w '
Pfauchen [ p ao£on]
f
'little p e a c o c k '
Tauchen [taocon] 'little r o p e '
b. [x]
Kuchen [ku:xsn] 'cake'
pfauchen [ p aoxon]
f
'to hiss'
tauchen [taoxan] 'to dive'

T h e distinguishing factor in these e x a m p l e s is that the f o r m s in w h i c h g appears

after a [ + b a c k ] vowel are m o r p h o l o g i c a l l y c o m p l e x forms derived at Level 2. (I
follow h e r e H a l l ' s , 1989a, analysis a n d refer the r e a d e r to that article for a r g u m e n -
tation s u p p o r t i n g the a s s i g n m e n t of this rule to Level 2.) T h u s a l t h o u g h the p h o -
nological e n v i r o n m e n t s a p p e a r to b e identical, the e x a m p l e s in (10a) h a v e a W o r d -
level m o r p h e m e b o u n d a r y b e t w e e n the v o w e l and the fricative, w h i l e those in
(10b) are m o n o m o r p h e m i c : [[Kuh]chen] vs. [kuchen].
In o r d e r to a c c o u n t for this, Hall ( 1 9 8 9 a ) builds into his rule (11) the stipulation
that the relevant s e g m e n t s m u s t b e t a u t o m o r p h e m i c .

(11) -son
+cont
-voice

V ,C

u- 1
[back] [+high]

B a c k n e s s spreads from a v o w e l o n t o a following c o n s o n a n t so long as both seg-

m e n t s are inside the s a m e m o r p h e m e — a r e t a u t o m o r p h e m i c . ( N o t e that this is the
8

equivalent of saying that the rule is b l o c k e d across the W o r d / L e v e l 2 boundary.)

Clearly this rule is a n o t h e r c a n d i d a t e for application at w o r d cycle. If rule (9)
applies on the W o r d cycle, before affixation of the d i m i n u t i v e m o r p h e m e , the fact
that this [c] is not affected by Fricative A s s i m i l a t i o n is unsurprising.
On the Word Level 207

W h a t is p e r h a p s surprising a n d m a k e s this rule of particular interest to us is the

fact that it d o e s not apply after the affix h a s b e e n attached. T h i s is certainly w h a t
w e w o u l d n o r m a l l y e x p e c t if p h o n o l o g y follows m o r p h o l o g y — w h e t h e r cyclically
or n o n c y c l i c a l l y — s i n c e the target a n d the trigger c o o c c u r only after the m o r -
phology. H o w e v e r , the rule m u s t b e b l o c k e d from applying in precisely this e n v i -
r o n m e n t . Hall a c c o m p l i s h e s this with his t a u t o m o r p h e m i c condition.
(Hall's t e r m " t a u t o m o r p h e m i c " is slightly m i s l e a d i n g since it m u s t e x c l u d e
all m o r p h o l o g i c a l c o m p l e x i t y arising before the W o r d level. Let us instead use
the w o r d TAUTOLEXICAL with the specific interpretation, in the s a m e W o r d .
Word-level rules are limited to a p p l y i n g within the d o m a i n of a W o r d , or m o r -
p h e m e , as it is available at Level 2. T h a t is, any form at Level 2, even if derived
at Level 1, is a W o r d a n d is treated in the s a m e w a y as an u n d e r i v e d w o r d . All
internal m o r p h o l o g i c a l information from Level 1 is g o n e at Level 2. T h u s it is
only W o r d - l e v e l / L e v e l 2 m o r p h e m e b o u n d a r i e s w h i c h m a k e t w o m o r p h e m e s not
tautolexical.)
Restricting a rule to tautolexical application is, at first g l a n c e , an u n u s u a l and
p r o b a b l y h i g h l y u n d e s i r a b l e c o n d i t i o n o n a rule. In the p r o p o s e d m o d e l this a p -
parently b a r o q u e c o n d i t i o n on Hall's rule is derived. In fact, tautolexicality is a
characteristic p r o p e r t y of rule application at the W o r d level a n d not at all b a r o q u e .
All W o r d level rules, i n c l u d i n g the G e r m a n Fricative A s s i m i l a t i o n rule, apply on
the cycle before m o r p h o l o g y . T h e r e is therefore n o n e e d for a stipulation to e n s u r e
that a rule applies tautolexically. O n the W o r d cycle, only o n e m o r p h e m e can b e
seen by the p h o n o l o g y ; all internal m o r p h o l o g i c a l structure from Level 1 h a s b e e n
erased, a n d n o additional m o r p h e m e s h a v e yet b e e n a d d e d . T h e next c y c l e t h r o u g h
the p h o n o l o g i c a l s y s t e m will b e postlexical, w h e r e the rule is not a v a i l a b l e . C o n -
9

sider the following derivation.

(12)
S t e m / L e v e l 1: hoch kuh kuchen
hoch hoch Umlaut
(Lieber,
1989)
W o r d / L e v e l 2: hoch hoch kuh kuchen
[x] [x] Fricative
Assimilation
Morphology: hoch+lich kuh+chen
[hox] [hoclic] [ku:can] [ku: xan]

At level 2, if the u m l a u t e d a l l o m o r p h is selected, the fricative is a palatal; if not,

the fricative is velar. O n the w o r d c y c l e of Kuh, the rule d o e s not apply. Thereafter
the rule d o e s not get a n o t h e r o p p o r t u n i t y to apply. A s a Word-level rule, its t i m e
208 Toni Borowsky

h a s c o m e and g o n e — b e f o r e the m o r p h o l o g y creates the s e q u e n c e m e e t i n g its

structural description.

2.2. S u m m a r y

To c o n c l u d e this section, I recapitulate: w e h a v e seen further e v i d e n c e of the

need for the W o r d c y c l e in G e r m a n . In addition, a n d m o r e importantly, w e have
seen a property of Word-level rules w h i c h I h a v e argued is characteristic: they
apply only tautolexically. T h e G e r m a n facts s h o w that a W o r d - c y c l e rule, in this
case Fricative A s s i m i l a t i o n , d o e s not apply after the W o r d cycle even if its struc-
tural description is met. T h e picture of the W o r d level w e see e m e r g i n g from these
facts is o n e in w h i c h there is a single d o m a i n of p h o n o l o g i c a l rule application
p r i o r to any m o r p h o l o g y . N o e v i d e n c e h a s b e e n a d d u c e d so far for a n o t h e r cycle
t h r o u g h p h o n o l o g y after m o r p h o l o g y . H o w e v e r , w e c o u n t as e v i d e n c e against the
c l a i m that there m a y be a n o t h e r c y c l e the fact that W o r d - c y c l e rules m u s t b e stipu-
lated to apply only tautolexically.

3. E N G L I S H A L L O P H O N I C R U L E S

I n o w i n t r o d u c e a series of a l l o p h o n i c rules d r a w n from various dialects of

English. T h e s e rules exhibit the s a m e p r o p e r t i e s as the p r o c e s s e s discussed in the
p r e v i o u s sections. A l t h o u g h these rules are a l l o p h o n i c and p e r h a p s the sort of
rules w e m i g h t intuitively e x p e c t to b e postlexical rather than lexical, w e will see
that they h a v e clear lexical p r o p e r t i e s . All of t h e m are sensitive to m o r p h o l o g i c a l
structure in a w a y that is quite uncharacteristic of postlexical rules. S o m e of t h e m
can b e s h o w n not to b e postlexical quite directly. W h i l e s o m e of the rules h a v e
general postlexical c o u n t e r p a r t s in other dialects, in the dialects c o n c e r n e d they
are clearly contextually d e t e r m i n e d a n d h a v e taken on the characteristics of W o r d -
level rules. N o t e again that all of these rules violate structure preservation b e c a u s e
they create n e w s e g m e n t s — a l l o p h o n e s . S i n c e t h e s e rules b e h a v e in precisely the
s a m e w a y as the rules d i s c u s s e d so far, they p r o v i d e strong support for the d o m a i n
and m o d e of rule application p r o p o s e d herein.
All data a n d s o m e a n a l y s e s are taken from W e l l s ( 1 9 8 2 ) a n d H a r r i s ( 1 9 8 9 ,
1991). I h a v e confined m y discussion in the text to four of these rules. F u r t h e r
e x a m p l e s of this kind of rule, with brief d e s c r i p t i o n s of each, c a n b e found in the
A p p e n d i x . In e a c h case, in the alternations discussed, the a l l o p h o n e found in
the unaffixed w o r d o c c u r s also if followed b y a Level 2 affix. In m o n o m o r p h e m i c
f o r m s , or those derived at Level 1, a different a l l o p h o n e is found. In other w o r d s ,
these alternations s h o w the s a m e patterns of distribution as the m o r p h o p h o n e m i c
rules discussed a b o v e .
On the Word Level 209

3.1. Belfast D e n t a l i z a t i o n

T h e r e is a rule of dentalization found in the dialect of E n g l i s h s p o k e n in Belfast

w h i c h c a u s e s alveolar c o n s o n a n t s to b e c o m e dental w h e n followed by tautosyl-
labic r.

(13) t, d, n, /->/, d, n,ll (z)r

( N o t e that s o m e dialects of E n g l i s h h a v e a general and unrestricted dentalization

p r o c e s s ; h o w e v e r the dialect w e are c o n c e r n e d with is not o n e of these.)
In Belfast the dental realization of alveolars before r occurs in m o n o m o r p h e m i c
f o r m s as well as in f o r m s d e r i v e d at Level 1. E x a m p l e s are given illustrating the
alternation in t h e s e e n v i r o n m e n t s in (14).

(14) Dental Nondental

a. train, drain bedroom, hard rain
b. elementary, sanitary™ element row

C o m p a r e the t w o sets of e x a m p l e s in (15). In these w e see the dentalized f o r m s

in the m o n o m o r p h e m i c w o r d s in the first c o l u m n w h i l e the w o r d s in the s e c o n d
c o l u m n s h o w n o dentalization even t h o u g h they a p p e a r to b e in a l m o s t identical
environments.

(15) Dental Nondental

spider wider
ladder louder
spanner finer
pillar filler
matter fatter

T h e w o r d spider with a dental c o n t r a s t s with wider with an alveolar d. T h e differ-

e n c e b e t w e e n the f o r m s in (15) is the fact that the f o r m s in the first c o l u m n are
m o n o m o r p h e m i c , w h i l e those in the s e c o n d are derived at Level 2. T h e rule d o e s
not apply in any of the forms in w h i c h the c o n d i t i o n i n g (Q) r is in a Level 2 affix
and therefore not in the s a m e m o r p h e m e as the dentalizing c o n s o n a n t . T h u s w e
see the s a m e pattern e m e r g i n g as w e found with the G e r m a n Fricative A s s i m i l a -
tion rule: if the trigger of the rule is in a Word-level suffix, w h i l e the target is in
the p r e c e d i n g m o r p h e m e , the rule d o e s not apply. W h e r e the rule d o e s apply, it
applies w h e n b o t h trigger a n d target are tautolexical: that is, O N T H E W O R D CYCLE.
N o t e that the rule c a n n o t b e a postlexical rule in Belfast E n g l i s h b e c a u s e p o s t l e x -
ically there is n o w a y to distinguish a m o n o m o r p h e m i c w o r d like spider [dor]
from a b i m o r p h e m i c w o r d like wider [dor].
In o r d e r to a c c o u n t for these facts, H a r r i s ( 1 9 8 9 ) a n a l y s e s this rule as restricted
to Level 1 in spite of the fact that it violates both structure preservation and the
strict c y c l e c o n s t r a i n t ( S C C ) . A s pointed out a b o v e , it is clearly not postlexical.
1 1
210 Toni Borowsky

O n the basis of these properties, as well as the fact that its pattern is exactly that
of the clearcut Word-level rules, I assign it to the W o r d level instead, p a c e Hall.
Let us then c o m p a r e the derivations of the forms in (16). At Level 1, the rule
d o e s not apply at a l l — i t is b l o c k e d by structure p r e s e r v a t i o n as well as the S C C .
At Level 2 it applies on the w o r d cycle to g e n e r a t e the correct f o r m s in the first
and last c a s e . S i n c e m o r p h o l o g y follows p h o n o l o g y , the rule d o e s not get an o p -
portunity to apply on the W o r d cycle in the m i d d l e c a s e . After the affix -er is
attached to wide, the rule d o e s not apply. T h u s wider h a s a n o n d e n t a l d.

(16) spider wide element + ary

S t e m / L e v e l 1: — — — Dentalization
W o r d / L e v e l 2: wide
W o r d cycle spidor — elementry Dentalization
Morphology: wide + r
[spaydr] [waydr] [elamentri]

O n c e again, this c a s e s h o w s that a Word-level rule m u s t apply on the w o r d cycle,

and further, m u s t not get a n o t h e r o p p o r t u n i t y to apply at all thereafter. T h e rule
only applies tautolexically. In m y t e r m s , the d o m a i n of this rule is the W o r d cycle.
After the m o r p h o l o g y h a s t a k e n p l a c e , the rule is n o l o n g e r active. T h u s even w h e n
a suffix is a d d e d w h i c h results in a string w h i c h m e e t s the structural description
of the rule, the rule d o e s not apply.
N o t e that if w e w e r e to a s s u m e , following H a l l e ( 1 9 8 8 ) , H a l l e and M o h a n a n
( 1 9 8 5 ) , or Booij and R u b a c h (1987) or others, that the rule is a p p l y i n g on a n o n -
cyclic d o m a i n , w e w o u l d b e forced to stipulate that the rule can apply only tauto-
lexically. T h e rule w o u l d h a v e to b e b l o c k e d b y the intervening bracket. Yet it is
not b l o c k e d b y the Level 1 b r a c k e t in [[element]ary] b e c a u s e b y the t i m e it gets
to a p p l y — t h a t is, at the W o r d l e v e l — a l l Level 1 b r a c k e t s h a v e b e e n erased. S o
only the Level 2 b r a c k e t is a blocker. H o w e v e r , if the rule applies before any
affixation at the W o r d level, n o t h i n g special n e e d s to b e said. T h e correct result is
derived quite simply.

3.2. A i t k e n ' s L a w

In various Scottish a n d Irish E n g l i s h dialects, r e p o r t e d in Wells (1982) and

H a r r i s ( 1 9 8 7 , 1989, 1990), a m o n g o t h e r s , there is a rule w h i c h l e n g t h e n s stressed
v o w e l s w h e n they are followed by voiced fricatives or r, or w h e n w o r d final. In
these dialects, unlike m o s t other varieties of E n g l i s h , there is n o p h o n e m i c length
distinction in v o w e l s . L e n g t h is d e t e r m i n e d b y this rule, k n o w n as Aitken's L a w
(Wells, 1982). T h e rule is therefore clearly not structure-preserving in these dia-
lects a n d thus c a n n o t apply at the stem level. H a r r i s ( 1 9 9 1 : 1 0 1 ) gives an informal
s t a t e m e n t of the rule w h i c h I adapt as follows.
On the Word Level 211

(17) AITKEN'S LAW: Stressed v o w e l s are long before a [ + v o i c e , + c o n t i n u -

1 2

ant] s e g m e n t or before ]. E l s e w h e r e they are short.

Accordingly, as p o i n t e d out by Wells (p. 4 0 0 ) , a w o r d such as mood [ m u d ] r h y m e s

with good, and bead h a s a v o w e l w h i c h is t h e s a m e length as the v o w e l in bid.
L o n g v o w e l s are found only in the Aitken's law e n v i r o n m e n t s : agree, seize, beer,
breathe, Seery, versus short v o w e l s in feet, feed, cease, feel, Sheedy, a n d so on. In
these e x a m p l e s w e see the rule violating the S C C as well.
O n c e again w e m i g h t a s s u m e that this rule is a postlexical rule, w e r e it not for
the fact that length derived b y Aitken's L a w in the word-final e n v i r o n m e n t is re-
tained e v e n w h e n the syllable is closed by a word-level suffixal c o n s o n a n t . S o , o n e
finds m i n i m a l pairs like those b e l o w w h e r e a morpheme-final v o w e l is followed
b y a suffixal -d. T h e m o r p h o l o g i c a l l y c o m p l e x forms are vowel final for the pur-
p o s e s of Aitken's law.

(18) ]d d]
agreed [9gri:d] greed [grid]
kneed [i:] need [nid]
brewed [u:] brood [brud]
stayed [e:] staid [sted]
toed [o:] toad [tod]
gnawed [o:] node [nod]

T h e sole difference b e t w e e n a form like [ b r u : d] brewed and [brud] brood is the

m o r p h e m e b o u n d a r y w h i c h intervenes b e t w e e n t h e v o w e l and the c o n s o n a n t . T h e
l e n g t h e n i n g found on word-final vowels is carried over to the suffixed f o r m s . S o
the l e n g t h e n i n g of the v o w e l s m u s t take p l a c e before the syllable is closed b y the
past-tense affix, w h i c h should o t h e r w i s e b l e e d the application of the rule. In the
m o n o m o r p h e m i c w o r d s there is n o l e n g t h e n i n g b e c a u s e the vowel is never final.
Aitken's law m u s t apply at the word-level (see also H a r r i s , 1990). It c a n n o t b e
postlexical b e c a u s e the b o u n d a r y b e t w e e n the t w o m o r p h e m e s w o u l d b e erased
b y the postlexical p h o n o l o g y and there w o u l d b e n o w a y to distinguish the m o n -
o m o r p h e m i c w o r d s from the b i m o r p h e m i c w o r d s . T h i s pattern is u n m i s t a k a b l y
the Word-level pattern w e have o b s e r v e d so far.

(19) Level 2: agreed vs. greed

Word cycle: gri grid
gri: DNA Aitken's law
Morphology: gri: + d

In s o m e Scottish E n g l i s h dialects, there is a related d i p h t h o n g i z a t i o n . T h e w o r d -

final c o n t e x t of Aitken's law generalizes to b e c o m e a syllable-final e n v i r o n m e n t .
T h e t w o variants are distinguished in quality: thus [Ai] alternates with [ a ' e ] . fire
[fa'e.ir], high, alive, prize, all with [ a e ] , versus wipe, tribe, wide with [Ai] (Wells,
f
212 Toni Borowsky

1 9 8 2 : 4 0 5 ) . T h e long variant [a e] a p p e a r s in word-final position even w h e n fol-

T

l o w e d by an apparently tautosyllabic c o n s o n a n t , as s h o w n in (20). A g a i n the

word-final e n v i r o n m e n t is u n c h a n g e d for the p u r p o s e s of this rule by the following
suffix. T h u s the rule a c c o u n t i n g for this a l l o p h o n y m u s t apply before the suffixes
are incorporated. M o n o m o r p h e m i c viper contrasts with b i m o r p h e m i c wipe+r ex-
actly as predicted if the rule applies on the w o r d cycle and m o r p h o l o g y only takes
place afterwards.

(20) [a-e] [Ai]

tied tide
sighed side
shyness shining
viper wiper
spider wider

3 . 3 . L o n d o n [AU] ~ [DU]

In L o n d o n Vernacular E n g l i s h , syllable-final /1/s are s o m e t i m e s vocalized. Har-

ris ( 1 9 8 9 ) s h o w s , b a s e d again on the description of these facts in Wells ( 1 9 8 2 :
3 1 2 ) , that in this e n v i r o n m e n t v o w e l s p r e c e d i n g this / s h o w certain coarticulatory
effects. F o r e x a m p l e , the length distinction in high vowels is neutralized so that
pull and pool are both p r o n o u n c e d [ p u u ] and fill, feel are [fiu]. I m p o r t a n t to m y
discussion is the fact that the vowel of cola, Roland, [AU], b e c o m e s [DU] w h e n
followed b y tautosyllabic /. T h e rule is given informally in (21).

(21) [AU]->[DU]/ 1L 1 3

T h e vowel of an unaffixed w o r d is m a i n t a i n e d in w o r d s derived b y affixation at

Level 2 even t h o u g h the / h a s apparently b e c o m e the o n s e t of the affix syllable.
T h u s o n c e again w e find contrasts such as those s h o w n in (22).

(22) [AU] [DU]

Roland rolling
slowly goalie
polar roller

O n the other h a n d , f o r m s d e r i v e d at Level 1 b e h a v e as if they are m o n o m o r p h e m i c

at Level 2 ; thus polar, from [[pole]ar], h a s the vowel [AU], s h o w i n g that the rule
did not apply to this form on the first Stem-level cycle. T h u s yet again, w e h a v e a
n o n - s t r u c t u r e - p r e s e r v i n g rule a p p l y i n g before m o r p h o l o g y at Level 2. C o m p a r e
the derivation of polar with that of roller. T h e rule m u s t apply to roll, but not to
pole, before the final / is syllabified into the o n s e t of the suffix syllable. T h a t is, it
m u s t apply on the w o r d c y c l e — a t the W o r d level.
On the Word Level 213

(23) polar roller

Level 1/stem: po(le)
ro(ll) Rule(21)DNA
Morphology: pol + ar
Phonology:
Rule ( 2 1 ) D N A
Level 2/word:
Word cycle: polar roll
NA [rou] R u l e 21
roll + er
Morphology:
[roulor]
Postlexical: [pAulor]

3.4. N o r t h e r n Irish daze v e r s u s days

In N o r t h e r n Irish E n g l i s h (Wells, 1982; H a r r i s , 1990), there is an allophonic

alternation b e t w e e n the vowel found w o r d finally in w o r d s like stay, day, lay, with
[ e : ] , and the s a m e vowel w h e n nonfinal: fade, fate, station, cater, fail, with [io].
N o t e that the rule d o e s not apply in the syllable-final e n v i r o n m e n t — c o n s i d e r sta-
tion as c o m p a r e d with stay.

(24) [io]->[e:]/ #

W e c a n see from the a b o v e e x a m p l e s that the rule is n o n - s t r u c t u r e - p r e s e r v i n g a n d

violates the S C C and therefore c a n n o t b e a Stem-level rule. Yet nevertheless w e
again find that a contrast arises b e t w e e n the t w o a l l o p h o n e s w h e n w e look at forms
derived at the W o r d level, so the rule c a n n o t b e postlexical, either.

(25) [13] [e:]

Daly daily
Reagan ray gun
daze days
staid stayed
grains greyness

T h u s w e fix this rule on the W o r d level. In o r d e r to derive the forms in (25), the
rule m u s t apply o n t h e W o r d c y c l e before any Word-level affixation, as follows.

(26) Level 2: Daly versus daily

Word cycle: — day
[e:] R u l e (24)
Affixation: — day + ly
Output: [dioliy] [ d e : liy]

S o far w e h a v e seen a very distinctive set of rules, all s u p p o r t i n g the c l a i m that

there is a single cycle t h r o u g h the p h o n o l o g y before any m o r p h o l o g y takes p l a c e
214 Toni Borowsky

at Level 2, that is, at the W o r d level. ( M o r e such rules a p p e a r in the A p p e n d i x . )

E a c h of the rules w e h a v e seen is oblivious to material i n t r o d u c e d by the m o r -
p h o l o g y of Level 2; this is e v i d e n c e d by t w o things: e a c h rule a p p e a r s to apply
only inside the m o r p h e m e c o n c e r n e d — t h e rules apply only tautolexically; and at
least t w o of t h e m are clearly b l o c k e d from applying w h e n their structural d e s c r i p -
tion is satisfied b y a string w h i c h s p a n s t w o Word-level m o r p h e m e s .

4. T H E W O R D L E V E L

I c o n c l u d e d the p r e v i o u s section with the c l a i m that there exists a c y c l e — t h e

W o r d c y c l e — w h i c h takes p l a c e after Level 1 but before any Level 2 m o r p h o l o g y .
It should b e clear at this p o i n t that unless w e a s s u m e the e x i s t e n c e of such a cycle,
w e c a n n o t a d e q u a t e l y explain the array of facts w e h a v e seen. In this section I take
this a r g u m e n t o n e step further a n d a r g u e that this cycle is the only p h o n o l o g i c a l
c y c l e at Level 2 . 1 suggest also that this cycle is e x t e n d e d to all the formatives at
this level. T h a t is, I a r g u e that all Word-level affixes also g o t h r o u g h the p h o n o l -
o g y on the w o r d cycle in E n g l i s h , and only thereafter are they j o i n e d t o g e t h e r b y
m o r p h o l o g i c a l p r o c e s s e s . T h e r e is n o further lexical p h o n o l o g y after Level 2
morphology.
T h i s study of the p h o n o l o g i c a l p r o c e s s e s associated with the W o r d level in E n -
glish h a s revealed a surprising fact given standard a s s u m p t i o n s about lexical p h o -
nology. It turns out that every p h o n o l o g i c a l rule at this level is a w o r d - c y c l e rule.
I h a v e found n o Level 2 / W o r d - l e v e l p h o n o l o g i c a l rules w h i c h are not W o r d - c y c l e
rules. Q u i t e simply, a l t h o u g h w e m a y for theoretical r e a s o n s w a n t to a r g u e that the
results of m o r p h o l o g i c a l o p e r a t i o n s at this level m u s t cycle again t h r o u g h the lexi-
cal p h o n o l o g y ( w h e t h e r j u s t o n c e , as in a n o n c y c l i c theory, or m o r e than o n c e , as
for a cyclic m o d e l ) , there is n o e v i d e n c e for this. G i v e n this, there is n o empirical
r e a s o n to a s s u m e any further c y c l e t h r o u g h the p h o n o l o g y at this level. All the
k n o w n facts are c o m p a t i b l e with the v i e w that the next d o m a i n of p h o n o l o g i c a l
rule application is the d o m a i n of the p h o n o l o g i c a l w o r d — i n the postlexical
phonology. 14

N o w n o t e that it d o e s not follow that, b e c a u s e there is n o e v i d e n c e of p h o n o l o g y

after the w o r d - c y c l e , that all the AFFIXES should h a v e u n d e r g o n e p h o n o l o g y at
15

the w o r d c y c l e p r i o r to m o r p h o l o g i c a l c o n c a t e n a t i o n . It is entirely p o s s i b l e that

the fact that p h o n o l o g y p r e c e d e s m o r p h o l o g y is only apparent. T h a t is, b e c a u s e
all the rules are w o r d - c y c l e rules, they p r e c e d e m o r p h o l o g y . T h e n there simply is
n o m o r e lexical p h o n o l o g y in E n g l i s h — t h o u g h it is entirely possible that there
might have been.
In the following sections I c o n s i d e r a n u m b e r of p h e n o m e n a in s u p p o r t of the
c l a i m that there is n o further lexical p h o n o l o g y after the W o r d c y c l e and that
On the Word Level 215

therefore the affixes m u s t h a v e u n d e r g o n e their p h o n o l o g y on this c y c l e . First I

discuss the affix / i n g / , w h i c h is an affix that u n d o u b t e d l y u n d e r g o e s a W o r d - c y c l e
p h o n o l o g i c a l rule. T h e n I c o n s i d e r s o m e lexical rules w h i c h m y s t e r i o u s l y s e e m to
skip over Level 2. T h e s e t w o sections t o g e t h e r p r o v i d e e v i d e n c e that s o m e W o r d -
level affixes are i n d e p e n d e n t p h o n o l o g i c a l d o m a i n s on the W o r d level and that
there is n o p h o n o l o g y b e y o n d the W o r d c y c l e . L a s t I l o o k at the general p r o p e r t i e s
of f o r m s derived by affixation at the W o r d level a n d s h o w that these properties are
only e x p l a i n e d if the a b o v e c o n c l u s i o n s are d r a w n .

4 . 1 . R u l e s W h i c h A p p l y t o Affixes: / - i n g /

T h e r e is o n e L e v e l 2 affix in E n g l i s h w h i c h s h o w s u n e q u i v o c a l l y the applica-

tion of a p h o n o l o g i c a l rule w h i c h is itself a w o r d - c y c l e rule. T h a t is, the affix
/ - i n g / requires the operation of the rule w h i c h deletes the final g in the s a m e w a y
as a w o r d like long d o e s . T h u s w e k n o w that this affix u n d e r g o e s word-level
1 6

phonology.
T h e usual a s s u m p t i o n a b o u t the p h o n o l o g i c a l b e h a v i o r of b o u n d m o r p h e m e s
like / - i n g / is that p h o n o l o g y should take p l a c e only after they are attached to the
w o r d . G i v e n this, w e h a v e n o r e a s o n to e x p e c t that the g should not b e h a v e like
any final g at Level 1 — t h a t is, it should b e e x t r a m e t r i c a l o n its o w n cycle a n d
r e s c u e d b y syllabification into the onset of the following syllable [[[swrn^] iNg]
est] on the next. O f c o u r s e it is not. T h e form ^[swirjirjgost] is u n g r a m m a t i c a l .
G i v e n this, let us s u p p o s e instead that the g is not e x t r a m e t r i c a l . T h e n w e can
derive the correct result if w e e n s u r e the derivation is cyclic, as s u g g e s t e d in B o -
r o w s k y ( 1 9 8 6 ) — t h a t is, the g will d i s a p p e a r on the Iswiq+ingl cycle. T h i s is
directly c o u n t e r to all o t h e r c l a i m s a b o u t t h e W o r d level in t h e literature (Booij
and R u b a c h , 1987; H a l l e a n d K e n s t o w i c z , 1989; H a l l e and M o h a n a n , 1985; Halle
a n d Vergnaud, 1987), w h e r e it is a r g u e s that the W o r d level is n o n c y c l i c .
H o w e v e r , if the derivation is cyclic, there are still p r o b l e m s . W e h a v e seen that
o n e of the p r o p e r t i e s of t h e s e w o r d - c y c l e rules is t h e fact that if their structural
description is m e t on a c y c l e created after affixation at Level 2, the rule d o e s not
apply. T h a t is, in t h o s e c a s e s , G e r m a n Fricative A s s i m i l a t i o n and Belfast D e n t a l -
ization, in w h i c h the trigger a n d the target of the rule a p p e a r e d in different m o r -
p h e m e s , the rule never applied. T h e rules only applied tautolexically. If the deri-
vation w e r e cyclic, then the rule should get a n o t h e r c h a n c e to apply o n the c y c l e
at w h i c h the trigger and target b o t h occur. B u t this is not the c a s e ; neither of these
t w o rules applies after the w o r d c y c l e . N o t e that the s a m e situation arises if the
1 7

derivation is n o n c y c l i c a n d all m o r p h o l o g y p r e c e d e s all p h o n o l o g y . T h e r e is still

a difficulty e x p l a i n i n g the b e h a v i o r of these rules w i t h o u t resorting to the use of
b r a c k e t s as a b l o c k i n g d e v i c e . (Interestingly, there d o not s e e m to b e a n y parallel
e x a m p l e s , that I k n o w of, of rules w h i c h are crucially b l o c k e d b y the Stem-level
boundary.) In the c a s e of / i n g / , the g w o u l d h a v e to delete sensitive to the W o r d
216 Toni Borowsky

b o u n d a r y ]. Yet the p r o p e r ch ar acter izatio n of this deletion is in t e r m s of syllable

structure: the g deletes if n o t syllabifiable, at the W o r d level. It m u s t b e syllabifi-
able in the s a m e W o r d — t h a t is, tautolexically.
T h e tautolexical p r o p e r t y of these rules is u n p r o b l e m a t i c if all Word-level p h o -
n o l o g y applies to all f o r m a t i v e s on the W o r d c y c l e , before any m o r p h o l o g y . In-
d e e d , it is p r e d i c t e d . T h e n the final g of / i n g / is p r o s o d i c a l l y licensed at the s a m e
t i m e that the final g of long i s — o n the w o r d c y c l e . S i n c e neither g can b e licensed,
they delete.
T h u s w h i l e G e r m a n Fricative A s s i m i l a t i o n and Belfast Dentalization gave evi-
d e n c e that Word-level rules d o not apply across the W o r d affix b o u n d a r y , the affix
/ - i n g / p r o v i d e s e v i d e n c e that there is at least o n e w o r d - c y c l e affix w h i c h under-
g o e s a W o r d - c y c l e rule, i n d e p e n d e n t of a n y t h i n g it m a y b e adjoined to. A n affix
u n d e r g o e s a tautolexical W o r d - c y c l e rule.

4.2. R u l e s W h i c h S e e m N o t t o A p p l y at L e v e l 2 : N a s a l A s s i m i l a t i o n
and Stress

T h e r e are rules w h i c h are a p p a r e n t l y available at d o m a i n s b o t h p r e c e d i n g and

following the W o r d level b u t d o not apply at the W o r d d o m a i n itself. S t a n d a r d
lexical p h o n o l o g i c a l solutions m u s t stipulate that the rule is b l o c k e d at this level.
Consider, for e x a m p l e , the rules r e s p o n s i b l e for N a s a l A s s i m i l a t i o n and Stress
A s s i g n m e n t in E n g l i s h .
N a s a l A s s i m i l a t i o n a p p l i e s regularly at Level 1 as well as postlexically, but it
a p p e a r s to skip Level 2. At Level 1, N a s a l A s s i m i l a t i o n applies to derive f o r m s
like impress, compel from / i N + p r e s s / and / c o N + p e l / , and so on. It also clearly
applies postlexically: pu[r\]kin 'pumpkin', Vancouver, i[m] Bolivia, a n d so on.
It d o e s not apply at L e v e l 2 b e t w e e n the prefix un- a n d a following w o r d : there is
n o obligatory nasal a s s i m i l a t i o n — u n b e l i e v a b l e is [Anbiliivabl]. H o w e v e r , post-
lexical application of this rule will give us the optional [Am]believable, showing
that the rule h a s not t u r n e d off.
H a l l e a n d M o h a n a n ( 1 9 8 5 ) as well as B o r o w s k y ( 1 9 8 6 ) h a v e a r g u e d i n d e p e n -
dently that in fact N a s a l A s s i m i l a t i o n d o e s apply at Level 2 to g e n e r a t e f o r ms like
long before the g deletes. W h y , then, d o e s it n o t apply in unbelievable and others?
It has b e e n p r o p o s e d (Booij a n d R u b a c h , 1987) that s o m e prefixes m u s t r e m a i n
separate from the rest of the form in o r d e r to b l o c k certain rules from a p p l y i n g to
t h e m . T h a t is, they m a i n t a i n s o m e sort of p h o n o l o g i c a l integrity even after affix-
ation. In Booij a n d R u b a c h ' s s y s t e m , t h e s e affixes form a PROSODIC W O R D or MOT
a n d as such are treated b y the p h o n o l o g y as a separate d o m a i n for p h o n o l o g i c a l
rules. A c c o r d i n g to Booij a n d R u b a c h , the a b o v e m e n t i o n e d facts a b o u t un- fall
out b e c a u s e the prefix is a p r o s o d i c constituent on its o w n . un- is a m o t w h i c h is
separate from the s e c o n d m o t : -able. First p r o s o d i c structure is a s s i g n e d and then
o t h e r p h o n o l o g i c a l rules t a k e p l a c e .
On the Word Level 217

T h e notion m o t or p r o s o d i c w o r d is, in principle, similar to m y claim that at the

w o r d level, affixes c y c l e i n d e p e n d e n t l y t h r o u g h the p h o n o l o g y . Booij and R u b a c h
are u n c l e a r a b o u t how, or i n d e e d w h e n , their m o t is derived. F o r t h e m it m u s t b e
stipulated that certain f o r m s h a v e this p r o s o d i c structure so that their p h o n o l o g i c a l
b e h a v i o r follows. In the m o d e l of the p h o n o l o g y p r o p o s e d herein, the m o t is d e -
rived on the W o r d c y c l e . M o t is w o r d . T h e theory predicts that all affixes at the
W o r d level should retain their p h o n o l o g i c a l integrity in this way, with respect to
lexical rules. T h e y d o this b e c a u s e p h o n o l o g y p r e c e d e s m o r p h o l o g y a n d d o e s not
interact w i t h it at this level. S i n c e e a c h affix is o r g a n i z e d b y the p h o n o l o g y as an
i n d e p e n d e n t f o r m on the W o r d cycle, the affix h a s a separate identity.
N o assimilation takes p l a c e b e t w e e n / u n - / a n d a following c o n s o n a n t in the
adjacent m o r p h e m e at the W o r d level b e c a u s e the m o r p h e m e s are not adjacent
w h e n they c y c l e t h r o u g h the p h o n o l o g i c a l s y s t e m . T h e prefix g o e s t h r o u g h the
p h o n o l o g y , a n d the stem to w h i c h it attaches g o e s t h r o u g h the p h o n o l o g y , a n d the
rules w h i c h are a p p l i c a b l e a p p l y to both f o r m s independently. N a s a l A s s i m i l a t i o n
is not applicable, since the n is n o t adjacent to an assimilator. T h i s is different
from the c l a i m of Booij and R u b a c h . F o r t h e m , the affixes a n d stem are m o r p h o -
logically adjacent b u t in distinct p r o s o d i c / p h o n o l o g i c a l constituents, w h i l e in m y
s y s t e m they are not m o r p h o l o g i c a l l y adjacent w h i l e the Word-level p h o n o l o g i c a l
rules are a p p l y i n g .
T h u s the state of b e i n g a m o t is derived from the o r d e r of events in m y system.
T h e p r o p e r t y of b e i n g a p r o s o d i c unit is derived quite simply via the p h o n o l o g y ;
every form that c y c l e s t h r o u g h the p h o n o l o g y at s o m e p o i n t in the derivation is a
p r o s o d i c unit in that d o m a i n . A t the S t e m level, the m o r p h o l o g y creates c o m p l e x
forms w h i c h c y c l e t h r o u g h the p h o n o l o g y . E a c h such unit is a p r o s o d i c unit. At
the W o r d level, n o c o m p l e x forms arise until after the p h o n o l o g y has t a k e n p l a c e
on all f o r m a t i v e s . All the affixes attached at the W o r d level, as well as all f o r m s
w h i c h h a v e c o m e from Level 1 into this level, are p r o s o d i c units. N o L e v e l 1 af-
fix c o u l d ever b e a m o t b e c a u s e at Level 1 n o affix can constitute, b y itself, a
d o m a i n for p h o n o l o g i c a l rule application. O n the other h a n d , all m o r p h e m e s are
p r o s o d i c units at the W o r d level. In other w o r d s , it is not the c a s e that N a s a l
A s s i m i l a t i o n skips Level 2. T h e rule m a y apply at every l e v e l — c y c l i c a l l y at the
first level, noncyclically w h e n postlexical, a n d tautolexically at the W o r d level.
Its tautolexical application at the W o r d l e v e l — ( i t applies in long b u t not in
un+believable)—is entirely consistent with the architecture of the system p r o -
p o s e d herein.
F u r t h e r e v i d e n c e that f o r m s at the W o r d level constitute i n d e p e n d e n t p r o s o d i c
d o m a i n s c o m e s from the stress rules. M o s t w o r k in lexical p h o n o l o g y m a k e s the
a s s u m p t i o n that the general stress rules d o n o t apply at the W o r d level in the w a y
they d o at t h e first level. T h e W o r d affixes n e v e r affect previously a s s i g n e d
stresses. To a c c o u n t for this it is said either that e a c h of the affix classes is m a r k e d
as stress-neutral or stress-sensitive; or that the b l o c k of stress-neutral affixes is
218 Toni Borowsky

attached after the stress rules a p p l y . T h a t is, the failure of certain stress rules to
18

apply at the W o r d level is stipulated.

It c a n n o t b e the c a s e that the regular stress rules h a v e turned off altogether after
Level 1, since they apply to c o m p o u n d s , and s o m e of the so-called stress-neutral
affixes are t h e m s e l v e s s t r e s s e d — s o m e of the prefixes for e x a m p l e , as well as
/ - h o o d / . T h e g r a m m a r m u s t surely d o m o r e than m e r e l y stipulate, to a c c o u n t for
1 9

the fact that the o n e class of affixes never h a s any effect on previously assigned
stresses w h i l e the other d o e s . It d o e s not follow from cyclicity. It is j u s t as p o s s i b l e
that earlier stresses could b e w i p e d out o n a n o n c y c l i c d o m a i n as on a cyclic
d o m a i n . N o t e too that the fact that it is a l w a y s the second-level affixes w h i c h are
stress neutral and the first-level affixes stress sensitive is u n e x p l a i n e d so long as
s o m e stress rules c a n a p p l y at the s e c o n d level. It c o u l d in p r i n c i p l e b e the other
w a y r o u n d given any other lexical p h o n o l o g y m o d e l — b u t it is not.
If w e a s s u m e that the stress rules apply to e a c h m o r p h e m e individually on the
w o r d c y c l e , it follows that these affixes will a l w a y s b e stress-neutral with respect
2 0

to the rest of the m o r p h o l o g i c a l structure derived at the s a m e level. O n the other

h a n d , Level 1 affixes attach p r i o r to p h o n o l o g y — t h e output of every m o r p h o l o g i -
cal operation creates a w o r d w h i c h m u s t b e i n p u t to a p h o n o l o g i c a l cycle. T h u s it
follows that these affixes will add to the syllable c o u n t of the w h o l e form and thus
affect the overall stress pattern. In other w o r d s , the p h o n o l o g i c a l b e h a v i o r of the
classes of affixes s t e m s from the architecture of the m o d e l , and it is not merely a
stipulation a b o u t e a c h class of affixes o r a b o u t the stress rules.

4.3. The Characteristics of Stem and W o r d

T h e r e is a clear difference b e t w e e n the w a y derived forms at the S t e m and W o r d

levels l o o k on t h e surface. T h e s e differences are visibly manifest in the surface
s h a p e s of the f o r m s associated with e a c h level. R u l e s of the S t e m level typically
knit the m o r p h e m e s together, e r a s i n g the internal e d g e s and r e d o i n g the p r o s o d i c
organization every t i m e a m o r p h e m e is a d d e d . T h e output is a form w h i c h looks
like an u n d e r i v e d f o r m — a form s h o w i n g close j u n c t u r e . For e x a m p l e , there are
n o u n d e r l y i n g g e m i n a t e c o n s o n a n t s in E n g l i s h . T h e r e are n o g e m i n a t e s in any
u n d e r i v e d w o r d s in English. If t w o identical c o n s o n a n t s arise t h r o u g h the c o n c a t e -
nation of t w o m o r p h e m e s , the resulting s e q u e n c e is d e g e m i n a t e d , m a k i n g the re-
sulting w o r d l o o k like an u n d e r i v e d w o r d ; for e x a m p l e , /in + n u m e r a b l e / —>
[inumarobl].
O n the other h a n d , forms d e r i v e d at the W o r d level h a v e sharp internal e d g e s
w h i c h exhibit all sorts of w o r d - e d g e c h a r a c t e r i s t i c s . For e x a m p l e , the syllable
21

a p p e n d i x , w h i c h a p p e a r s only in absolute final position of m o n o m o r p h e m i c

w o r d s , a p p e a r s regularly followed b y Word-level suffixes: the s e q u e n c e of con-
s o n a n t s e n d i n g the first syllable of the w o r d worldly is a s e q u e n c e not p e r m i t t e d
internally in m o n o m o r p h e m i c f o r m s or in w o r d s derived at Level 1 (see B o -
rowsky, 1989). Word-level rules d o not knit e d g e s together b u t are instead char-
On the Word Level 219

acteristically sensitive to the e d g e s in a m a n n e r never found with rules of the first

level. Word-level forms look like c o n c a t e n a t i o n s of m o r p h e m e s and not like un-
d e r i v e d w o r d s . T h u s there is n o obligatory d e g e m i n a t i o n if the adjunction of t w o
m o r p h e m e s creates a g e m i n a t e . T h e g e m i n a t e m a y r e m a i n , d e p e n d i n g o n s p e e c h
rate; for instance, unnatural, pine needle, rat trap, bus stop. At the W o r d level the
previously a s s i g n e d p r o s o d i c o r g a n i z a t i o n is m a i n t a i n e d n o m a t t e r h o w m a n y
Word-level affixes are attached; consciencelessness.
M o s t w o r k in lexical p h o n o l o g y m a k e s little of these o b s e r v a t i o n s , m e r e l y
22

attributing the differences b e t w e e n the t w o levels to the fact that the levels are
c h a r a c t e r i z e d b y distinct sets of rules and, m o r e recently, to the c l a i m that the
W o r d level is n o n c y c l i c . H o w e v e r , neither of these ideas gives a satisfactory e x -
planation of the o b s e r v e d differences in S t e m forms and W o r d forms. T h e hy-
p o t h e s i s that certain rules are restricted to certain levels d o e s n o t e x p l a i n w h y t h e
forms should b e superficially so different. To say that the d e g e m i n a t i o n rule, or
the stress rule, or constraints on the structure of a w e l l - f o r m e d syllable turn off
after the first level is to explain n o n e of t h e s e things. It leaves aside as well the
fact that those rules w h i c h d o apply at the s e c o n d level are still not a c c o m p l i s h i n g
the s a m e sort of e d g e - e r a s i n g function as the rules of the first level. T h e output of
the Word-level p h o n o l o g y a n d m o r p h o l o g y n e v e r l o o k s like an u n d e r i v e d w o r d .
N e i t h e r can the c l a i m that the w o r d level is n o n c y c l i c , as h a s b e e n s u g g e s t e d by
Halle a n d M o h a n a n ( 1 9 8 5 ) , Booij a n d R u b a c h ( 1 9 8 7 ) , or K i p a r s k y (1985) p r o v i d e
an a d e q u a t e e x p l a n a t i o n . In fact, as w e will see below, the c l a i m is descriptively
i n a d e q u a t e . If the W o r d level is n o n c y c l i c , rules should take p l a c e across the b o a r d
within the p r o s c r i b e d d o m a i n , as in the postlexical p h o n o l o g y , and not b e re-
stricted to smaller d o m a i n s internal to the p r o s c r i b e d d o m a i n , as it s e e m s to b e at
the W o r d level.
A l t h o u g h it m a y apply cyclically inside nested p r o s o d i c d o m a i n s , postlexical
p h o n o l o g y is r e p u t e d to h a v e the p r o p e r t y of a p p l y i n g the rules across the b o a r d
within e a c h of these d o m a i n s , i g n o r i n g c o n s t i t u e n c y information internal to that
p r o s o d i c d o m a i n . T h u s , for e x a m p l e , the F l a p p i n g rule c h a n g e s / t , d / to [D] in foot-
m e d i a l position w h e r e v e r its structural description is m e t within s o m e larger p r o -
sodic d o m a i n . T h u s it applies inside w o r d s , as well as inside s o m e phrasal struc-
tures: repe[D]i[D]iveness, mor[D]ality, go [D]omorrow, ge[D] Anne. Word-level
p h o n o l o g y s h o w s n o t h i n g like this taking place. W h a t w e have seen instead is that
m o r p h o l o g i c a l b o u n d a r i e s at the W o r d level create i n d e p e n d e n t p h o n o l o g i c a l d o -
m a i n s internal to the d o m a i n of the w h o l e w o r d and that rules respect t h o s e inter-
nal d o m a i n s — t h e y apply tautolexically.
Let us a s s u m e that the first or stem level of the lexicon contains, inter alia, a list
of stems and a list of affixes, as well as the rules w h i c h relate e a c h stem to the
others in the list (or, alternately v i e w e d , the rules d e r i v e e a c h s t e m from another).
W h a t this m e a n s is that e a c h form derived at Level 1 exists as an i n d e p e n d e n t
form in the l i s t . A s s u m e as well that the W o r d level is distinguished by the fact
23

that every possible derived form is actively derived, and these lexical entries are
220 Toni Borowsky

not i n d e p e n d e n t l y listed. S o , in this lexicon, all the stems, as well as both classes
of affixes, are listed, but n o d e r i v e d Word-level f o r m s o c c u r in the list. If w e as-
s u m e all this to b e the case, the set of affixes w h i c h select for S t e m s (i.e., the
Level 1 affixes) can b e identified w i t h o u t explicit m a r k i n g . T h e s e affixes are the
o n e s w h i c h o c c u r in the listed f o r m s . T h e Word-level affixes o c c u r in the lexical
lists only as separate b o u n d m o r p h e m e s . T h e y d o not also a p p e a r in the listed
w o r d s . T h e t w o classes of affixes c a n b e differentiated by structure preservation,
as it w e r e . T h e set of affixes w h i c h also o c c u r as c o m p o n e n t s of listed w o r d s will
b e the Level 1/Stem-level affixes.
T h e fact that the S t e m level is cyclic is d u e to the fact that the existing lexical
items are derived from o n e another. T h u s presidential is derived from /president
+ ial/, a n d presidentiality is d e r i v e d from /presidential + ity/. Similarly, the fact
that the level is s t r u c t u r e - p r e s e r v i n g follows: relating f o r m s in existing i t e m s al-
lows n o r o o m for innovation in the d e r i v a t i o n s . Clearly, n o n e w s e g m e n t s or struc-
tures c o u l d b e i n t r o d u c e d . (I refer the r e a d e r to Sproat, 1985, w h o m a k e s the s a m e
observations.) N o t e , lastly, that the S C C , b y w h i c h structure-changing rules are
b l o c k e d in n o n d e r i v e d e n v i r o n m e n t s , is also derived: listed entries will b l o c k any
internal c h a n g e s d u e to t h e e l s e w h e r e condition (see Kiparsky, 1982), and thus
the only p e r m i s s i b l e c h a n g e s are o n e s at e d g e s w h e r e n e w material is b e i n g
incorporated.
N o n e of these characteristics carry over to the W o r d level. W e d o not readily
a s s u m e that m o r p h o l o g i c a l l y c o m p l e x w o r d s at this level m u s t b e l i s t e d — a t least
not in the w a y that conjoin, conjunctive, conjunction, for e x a m p l e , are listed. T h e
first level of the lexicon is not c o n s i d e r e d to b e " p r o d u c t i v e " b e c a u s e it is
2 4

m a i n l y c o n c e r n e d with existing w o r d s . (For e x a m p l e , a w o r d like enjunctive is a

possible, but n o n e x i s t i n g , w o r d b e c a u s e it is not in the list even t h o u g h it c a n b e
derived b y the rules.) Word-level f o r m s like swingingest, unchocolatyest, or
neighbourhoodlessness are not i n d e p e n d e n t l y listed; they are actively derived
by the m o r p h o l o g y — t h i s level is p r o d u c t i v e . T h e r e is n o cyclic or structure-
p r e s e r v i n g p h o n o l o g y b e c a u s e existing f o r m s are not b e i n g related to e a c h other.
S o , truly p r o d u c t i v e affixes s e e m to attach to fully p r o c e s s e d w o r d s . 2 5

In c o n c l u s i o n , I h a v e a r g u e d that the interaction of p h o n o l o g y and m o r p h o l o g y

in the lexicon differs on the t w o levels of the lexicon. I h a v e s h o w n that the p h o -
nological p r o p e r t i e s of each level are derived not from cyclicity/noncyclicity, but
from the fact that there is n o lexical p h o n o l o g y after affixation at the W o r d level
since the p h o n o l o g y p r e c e d e s the m o r p h o l o g y on this level.
T h e a r g u m e n t can b e s u m m a r i z e d as follows: the attested rules w h i c h can b e
fixed to apply at the W o r d level m u s t all apply on a p h o n o l o g i c a l cycle w h i c h
takes place before any m o r p h o l o g y — t h e word cycle. T h e s e rules d o not apply
after the w o r d c y c l e . T h e r e is n o p h o n o l o g i c a l e v i d e n c e of interaction with the
m o r p h o l o g y at the W o r d level. In fact, the p h o n o l o g i c a l characteristics of the
W o r d level suggest that there is n o n e . F r o m this I c o n c l u d e that the lexicon has
t w o d o m a i n s at w h i c h p h o n o l o g y takes place. T h i s is structured so that at the S t e m
On the Word Level 221

level the p h o n o l o g y and m o r p h o l o g y interfeed in the w a y usually a s s u m e d in m o s t

standard lexical p h o n o l o g i e s . O n exiting the S t e m level, all f o r m s cycle through
the p h o n o l o g y on their w a y to the m o r p h o l o g y of the W o r d level. T h i s is the W o r d -
c y c l e and constitutes the s e c o n d and final lexical p h o n o l o g i c a l d o m a i n . It is the
only p h o n o l o g i c a l c y c l e at this level a n d links the t w o d o m a i n s . T h e output of the
Word-level m o r p h o l o g y enters the p h o n o l o g y again o n its w a y into the syntax;
this is the first postlexical d o m a i n , the p h o n o l o g i c a l w o r d . It is this d o m a i n w h i c h
c o r r e s p o n d s with the d o m a i n identified as the w o r d level in Booij and R u b a c h
( 1 9 8 7 ) , K i p a r s k y ( 1 9 8 5 ) , a n d so on. T h i s is the so-called " p o s t c y c l i c " or " n o n -
c y c l i c " d o m a i n , not the w o r d level, a n d it is not a lexical d o m a i n .

5. A N A L T E R N A T I V E P R O P O S A L

A n alternative w a y of d e s c r i b i n g s o m e of the facts I h a v e c o n s i d e r e d h a s b e e n

s u g g e s t e d to m e b y Paul K i p a r s k y . A l t h o u g h the analysis is, I believe, incorrect,
26

I take it seriously b e c a u s e it reflects a very interesting o b s e r v a t i o n about the nature

of m a n y of t h e w o r d - l e v e l rules. S i n c e m y c l a i m is s o m e w h a t innovative, I d o n o t
w a n t to d i s c o u n t any alternatives w h i c h w o u l d allow t h e s e facts to b e a c c o m m o -
dated within a traditional view of lexical p h o n o l o g y . I believe that the view I
p r e s e n t a c c o u n t s for the peculiar p r o p e r t i e s of word-level p h o n o l o g y , b u t it is en-
tirely p o s s i b l e that further r e s e a r c h will p r o v i d e the crucial e v i d e n c e to s u p p o r t
o n e or the other of these t w o v i e w s .
T h e c r u x of the alternative idea is the o b s e r v a t i o n that m a n y of the w o r d - c y c l e
rules w e h a v e seen are sensitive to syllable structure as it is defined at the w o r d
cycle. If w e m a k e a single a s s u m p t i o n a b o u t syllabification p r o c e s s e s in the lexi-
con, w e can a c c o u n t for m a n y of the facts w e h a v e seen within a theory in w h i c h
the p h o n o l o g y , after the w o r d c y c l e , is cyclic. A s s u m e that all lexical syllabifica-
tion is p u r e l y structure-building. Postlexically, syllabification m a y b e structure-
c h a n g i n g . T h u s syllabification affects only unsyllabified s e g m e n t s — a t both levels
of the lexicon. At the first level, final c o n s o n a n t e x t r a m e t r i c a l i t y ( H a y e s , 1982)
b l o c k s the syllabification of final c o n s o n a n t s until the following c y c l e ( B o r o w s k y ,
1986; ltd, 1986), w h e n the C loses its e x t r a m e t r i c a l status a n d is incorporated into
the suffix syllable. A s s u m e , in addition, that there is n o such final c o n s o n a n t ex-
trametricality at the W o r d level. T h e n , by general p r i n c i p l e s of p r o s o d i c licensing,
e a c h m o r p h o l o g i c a l constituent will b e an i n d e p e n d e n t syllabic d o m a i n if the m o r -
p h e m e s are attached cyclically.
It follows that there will b e n o ^ s y l l a b i f i c a t i o n of any previously o r g a n i z e d
s e g m e n t s w h e n suffixes are attached, since such an operation w o u l d b e structure-
c h a n g i n g . A n y c o n s o n a n t still left unsyllabified by the t i m e a form r e a c h e s Level 2
will b e dealt with o n the w o r d cycle, since it c a n n o t r e m a i n e x t r a m e t r i c a l any
longer. O n the w o r d cycle, all stray material is syllabified (or deleted). W h e n an
222 Toni Borowsky

affix is attached, the n e w material is syllabified, b u t the p r e v i o u s cycle's syllabifi-

cation is left intact b e c a u s e there is n o ^ s y l l a b i f i c a t i o n of material o r g a n i z e d ear-
lier. A s an e x a m p l e illustrating the difference b e t w e e n syllabification strategies at
the t w o levels, c o m p a r e the derivations of longer and longish.

(27) Level 1: a

k
lorj
k (g) lor) (g)
Syllabification
Final C extrametricality

kA Morphology—extrametricality lost
lor) g + er
a a

k A
lor) g er
Syllabification

Level 2:
Word cycle
lor) g 8 —• 0
0
a
A
lor] + ir/g Morphology
a a

AA
Iov iv (g) Syllabification, etc.
H o w d o e s this idea explain the facts I h a v e b e e n d i s c u s s i n g ? If syllabification
from the w o r d cycle is m a i n t a i n e d t h r o u g h o u t the W o r d level, then any rule w h i c h
is sensitive to syllabification will apply in syllables w h i c h h a v e not b e e n restruc-
tured. F o r e x a m p l e , a rule w h i c h I h a v e c l a i m e d m u s t apply on the w o r d c y c l e
could in fact apply at any t i m e at the w o r d level b e c a u s e the syllable structure at
this level will not b e c h a n g e d even after suffixation. T h u s , so long as w e allow
syllabification to take p l a c e either before or at the w o r d cycle, the other rules will
apply w i t h o u t p r o b l e m .

(28)

Level 1: po (/)
hA h
po I + ar ro (11) Syllabification
cr

Level 2:
A
pole polar
A
roll Syllabification

AA
roll + er
Syllabification
Rule (21)

Recall that the m o s t serious a r g u m e n t against the idea that the W o r d level w a s
cyclic after the w o r d cycle w e r e t h o s e cases in w h i c h the addition of an affix
On the Word Level 223

p r o d u c e d a s e q u e n c e w h i c h m e t the structural description of a w o r d - c y c l e rule. In

these cases the rule w a s b l o c k e d in a way w h i c h runs c o u n t e r to o u r usual e x p e c -
tations a b o u t cyclic rule application. T h e r e w e r e t w o e x a m p l e s of t h i s — o n e w a s
Belfast D e n t a l i z a t i o n , and the other w a s G e r m a n Fricative A s s i m i l a t i o n . Belfast
D e n t a l i z a t i o n applied only w h e n the target a n d the trigger of the rule are tauto-
m o r p h e m i c . If syllabification of the w o r d is left u n d i s t u r b e d after the w o r d cycle,
the structural description of a syllable-sensitive rule like this will not b e m e t after
an affix is adjoined. T h u s , even if the rule d o e s apply cyclically, its structural
d e s c r i p t i o n w o u l d n o t b e m e t at the W o r d level. C o n s i d e r as an e x a m p l e the deri-
vation of wider.
(29) Level 1: a

A
way (d) Syllabification

Level 2:
Word cycle:
/IN
way d Syllabification

/IN I I
Morphology: way d + dr Syllabification

NA Belfast Dentalization

(d and r are not tautosyllabic)

In this derivation w e see that at Level 1 the final c o n s o n a n t is e x t r a m e t r i c a l . S i n c e

n o affix is adjoined, the f o r m g o e s t h r o u g h to the W o r d level, w h e r e at the w o r d
c y c l e the final c o n s o n a n t s are l i c e n s e d — i n this c a s e b y syllabification. M o r -
p h o l o g y t h e n adjoins an affix. Belfast D e n t a l i z a t i o n is b l o c k e d at this p o i n t b e -
c a u s e the t w o relevant s e g m e n t s are not tautosyllabic. A s long as this syllabifica-
tion is m a i n t a i n e d t h r o u g h this level, the rule will n o t apply h e r e . T h u s this a l o n e
can a c c o u n t for the fact that e v e n w h e n the t w o s e g m e n t s o c c u r a c r o s s a b o u n d a r y
the relevant rule d o e s not apply, and it is therefore entirely p o s s i b l e that the W o r d
level is, after all, cyclic.
Unfortunately, h o w e v e r attractive this p r o p o s a l m a y s e e m , g i v e n the p r e s e n t
state of o u r k n o w l e d g e it r e m a i n s i n c o m p a t i b l e w i t h m a n y of the facts. In addition
it d o e s not suffice as an e x p l a n a t i o n of the o b s e r v e d properties of the W o r d level.
T h e analysis p r o p o s e d a b o v e e x p l a i n s the cases in w h i c h the desired effect is to
b l o c k ^ s y l l a b i f i c a t i o n of word-final c o n s o n a n t s into suffixal o n s e t s .
(30) * o- (7

N/1
V C ] V
w i d e r Belfast Dentalization
r o 11 i ng London [^u] / [vu]
224 Toni Borowsky

However, it is m o r e difficult to explain those cases in w h i c h the syllabification of

suffixal material is to b e i n c o r p o r a t e d into t h e word-final syllable, or in the case
of prefixes, the word-initial syllable.

(31) * o-

K
v ] c
agree d Aitken's law
paw s London pause/paws (Appendix)

(32) *

A
VC f V

II I
u n e thical

In (31) the suffix c o n s o n a n t is stray a n d therefore free to b e syllabified a n d incor-

porated into t h e p r e v i o u s syllable. T h e s a m e s h o u l d b e t h e c a s e in (32). T h i s m e a n s
that there is a potential difference b e t w e e n these t w o e n v i r o n m e n t s a n d the o n e in
(30), for t h e p u r p o s e s of p h o n o l o g i c a l rule application. I d o n o t k n o w w h e t h e r this
difference is ever exploited. T h e attested cases d o n o t distinguish b e t w e e n the t w o
types of e n v i r o n m e n t — a l l t h e given data m a k e n o distinction b e t w e e n t h e various
types of suffixes attached at Level 2 ; t h e f o r m day]s h a s an o p e n syllable equiva-
lent in every w a y to that in dai]ly, for t h e p u r p o s e s of t h e rule a c c o u n t i n g for the
distribution of t h e [ e : ] ~ [ i s ] a l l o p h o n e s in N o r t h e r n Irish.
N o n e of t h e a s s u m p t i o n s a b o u t syllabification, w h i c h w e h a v e m a d e in order to
m a k e this p r o p o s a l work, c a n b l o c k t h e incorporation of unsyllabified material
into an already structured s y l l a b l e — t h i s is certainly n o t covered b y t h e usual in-
terpretation of w h a t it m e a n s to b e structure-changing. B e s i d e s , w e h a v e already
allowed for the syllabification p r o c e d u r e s o n t h e W o r d cycle: stray c o n s o n a n t s on
the w o r d cycle m a y b e i n c o r p o r a t e d — f i n a l syllabic s o n o r a n t s — o r deleted, if u n -
syllabified b y t h e first level syllabification p r o c e s s e s .
A s J o h n M c C a r t h y (personal c o m m u n i c a t i o n ) h a s p o i n t e d out to m e , t h e syllab-
ification e x p l a n a t i o n m a k e s a n o t h e r prediction. In t h e k n o w n dialects, on this
view, the onsetless syllables w h i c h arise at Level 2 w o u l d persist till t h e postlexi-
cal p h o n o l o g y , w h e r e they w o u l d b e given o n s e t s b y resyllabification. F o r e x -
a m p l e , in an r-less dialect like m y o w n , the w o r d wondering is p r o n o u n c e d with
2 7

an r as the onset of the final syllable: [wAndarirj], a n d not *[wAnd9irj]. This s h o w s

clearly that resyllabification took p l a c e , since syllable-final r s a r e never pro-
n o u n c e d ; c o m p a r e wonder, w h i c h I p r o n o u n c e [ w A n d s ] . O n e m i g h t therefore ex-
pect a dialect s o m e w h e r e in w h i c h the e m p t y onsets of vowel-initial Word-level
affixes are filled b y s o m e sort of c o n s o n a n t insertion, say. Yet there are n o such
cases that I k n o w of. N o E n g l i s h dialect h a s forms like [[roll]?er], [[roll]?ing]
with a glottal stop filling t h e onset of the s e c o n d syllable.
On the Word Level 225

In addition to m a k i n g incorrect predictions a b o u t the k i n d s of p h e n o m e n a w e

m a y find, the p r o p o s a l is observationally w r o n g . Consider, for e x a m p l e , the fact
that G e r m a n Fricative A s s i m i l a t i o n is not, as far as I can tell, sensitive to syllable
structure. M a n i p u l a t i n g the rule in different w a y s d o e s not m a k e any difference.
E v e n if w e a s s u m e that G e r m a n h a s a p h o n e m i c distinction b e t w e e n fx/ and / c / , it
is not clear h o w to d e s c r i b e the generalization a b o u t their distribution in syllabic
t e r m s . A l t h o u g h neither the r e v i e w e r n o r Paul K i p a r s k y actually m a d e any attempt
to explain this particular p h e n o m e n o n in t e r m s of syllable structure, a fairly plau-
sible analysis m i g h t b e given along this line: the feature [ + b a c k ] is spread from
v o w e l s to tautosyllabic fricatives; the other alternant, [c], o c c u r s e l s e w h e r e . It is
difficult to reconcile such an analysis with m o n o m o r p h e m i c forms like: Buche
' b e e c h t r e e ' , Knochen ' b o n e ' , Kuchen ' c a k e ' , Sprache ' l a n g u a g e ' , and so on (all
with [x]). A m b i s y l l a b i c i t y m u s t b e ruled out as a p o s s i b l e e x p l a n a t o r y factor b e -
c a u s e not only are there m i n i m a l p a i r s — c o m p a r e sicher with [ 9 ] versus kachel
[x], but w e will then b e stuck again with n o e x p l a n a t i o n for the crucial -chen c a s e s .
B e s i d e s , ambisyllabicity, aside from b e i n g a postlexical p h e n o m e n o n , m u s t b e the
result of a s t r u c t u r e - c h a n g i n g syllabification p r o c e d u r e and thus is not within the
spirit of the p r o p o s a l I a m c o n s i d e r i n g h e r e .
Similarly, Aitken's law is also n o t clearly syllable-sensitive. F r o m the d i s c u s -
sion in Wells ( 1 9 8 2 ) , o n e m i g h t get the idea that Aitken's law is s o m e kind of
open-syllable l e n g t h e n i n g . J o h n H a r r i s and N o r v a l S m i t h h a v e p o i n t e d out to m e
(personal c o m m u n i c a t i o n ) that this is not the c a s e . T h e r e are m i n i m a l pairs such
as Seery [si: ri] (long vowel before r) versus Sheedy [sidi] (short vowel) w h i c h
illustrate that the rule applies e v e n w h e n the c o n d i t i o n i n g c o n s o n a n t is not tauto-
syllabic; a n d it d o e s not apply if the vowel is m e r e l y in an o p e n syllable (see also
ready a n d booty w h i c h h a v e short v o w e l s a n d are not * [ r e : d i ] or * [ b u : t i ] ) . T h e
stress r e q u i r e m e n t on the rule m a k e s it an interesting c a n d i d a t e for a prosodically
c o n d i t i o n e d r u l e — w h i c h m a y r e i n t r o d u c e s o m e syllabic c o n d i t i o n i n g ; however,
further research w o u l d h a v e to b e d o n e to take this r e m a r k o u t of the b o u n d s of
p u r e speculation. I therefore a s s u m e that the rule is not sensitive to syllable struc-
ture in any crucial w a y ; N o r is the N o r t h e r n Irish a l t e r n a t i o n : [ e : ] is word-final,
28

as in stay, w h i l e the other a l l o p h o n e is found syllable finally, as in state, station

with [10].
G i v e n that m a n y of the Word-level rules are n o t syllable-sensitive, the p r o p o s a l
outlined in this section can be rejected, but it b r i n g s u p a n u m b e r of related q u e s -
tions w h i c h p r o p o n e n t s of this idea have not a t t e m p t e d to answer. W h y should it
be the c a s e that syllabification at the t w o lexical levels is different? W h y d o so
m a n y rules a p p e a r to b e sensitive to the syllable structure assigned at the w o r d
c y c l e ? W h y d o e s it a p p e a r that syllable structure is retained at this level? T h e
analysis outlined in this section simply states that it is, without offering any expla-
nation. This view predicts as well that there should b e Level 2 rules w h i c h apply
across the b o u n d a r i e s b e t w e e n m o r p h e m e s . I have not yet u n c o v e r e d any such
rules.
226 Toni Borowsky

However, if all p h o n o l o g y takes p l a c e before m o r p h o l o g y in t h e w o r d d o m a i n ,

these o b s e r v a t i o n s follow. It c o u l d only b e t h e syllable structure a s s i g n e d o n the
w o r d cycle that will b e relevant for rules at t h e W o r d level. T h e system predicts
that all Word-level rules w h i c h are sensitive t o syllable structure will refer to the
syllable structure a s s i g n e d o n this cycle. S e c o n d , since n o rules apply after t h e
m o r p h o l o g y of t h e W o r d level, n o p h o n o l o g y across the m o r p h e m e e d g e s will b e
found at this level. All e d g e - e r a s i n g p h o n o l o g y other than a u t o m a t i c resyllabifi-
cation will b e postlexical. T h u s , u n i f o r m c o n t i n u o u s syllabification p r o c e s s e s m a y
b e m a i n t a i n e d t h r o u g h all levels if t h e distinction b e t w e e n levels is seen to b e a
c o n s e q u e n c e of t h e w a y t h e m o r p h o l o g y a n d p h o n o l o g y feed e a c h other at e a c h
level rather than a c o n s e q u e n c e of stipulations a b o u t h o w individual rules apply
at e a c h level.

6. A P P E N D I X : F U R T H E R E X A M P L E S O F W O R D - L E V E L R U L E S

In this A p p e n d i x I list, with brief descriptions, s o m e further e x a m p l e s of rules

of E n g l i s h 2 9
w h i c h I c o n s i d e r t o b e Word-level rules.

6.1. ^ - T e n s i n g

T h e rule of ^ - T e n s i n g is found in m a n y dialects of English. In m o s t of these

it applies g e n e r a l l y a n d u n c o n d i t i o n a l l y b u t in o t h e r s — N e w York City, Phila-
delphia, Belfast (Irish E n g l i s h ) — t h e rule is contextually d e t e r m i n e d . T h e vowel
ce tenses w h e n followed b y a tautosyllabic nasal fricative o r voiced obstruent
( t h o u g h n o t d in P h i l a d e l p h i a ) .
T h e data a n d analysis p r e s e n t e d h e r e are taken from D u n l a p (1988) o n Phila-
d e l p h i a E n g l i s h a n d H a r r i s ( 1 9 8 9 ) o n Belfast English.

(33) ce T e n s i n g :
nasals
ce - > E / _ fricatives
v o i c e d stops

[E]
graph graphic
psychopath psychopathic
mass, massive massive
class, classy classical, classify
classing

T h e f o r m graphic s h o w s that t h e rule c a n n o t apply cyclically at Level 1 b e c a u s e

that w o u l d g e n e r a t e gr[[E]phic. T h e rule m u s t apply at t h e Word-level b e c a u s e
On the Word Level 227

w e get t e n s e [E] in b o t h affixes a n d unaffixed f o r m s . S i n c e the rule is sensitive to

a tautosyllabic c o n s o n a n t , it m u s t apply before affixation, a n d s u b s e q u e n t syllabi-
fication at L e v e l 2 t a k e s the final c o n s o n a n t out of the syllable to m a k e the onset
of the suffix.

(34) Level 1: Cycle 1 Cycle 2

class class + ify
NA NA ^-Tensing

Level 2 :
Word cycle cl[E]ss r^-Tensing
Cycle 2 cl[E]ss + y

T h e unaffixed class u n d e r g o e s the rule, b e c o m i n g cl[E]ss. T h e n the affix is incor-

porated, rendering the correct output cl[E]ssy.

6.2. V o w e l R o u n d i n g i n A d e l a i d e E n g l i s h

In the A d e l a i d e dialect of A u s t r a l i a n E n g l i s h as d e s c r i b e d b y S i m p s o n ( 1 9 8 0 ) ,
there is a rule similar to the L o n d o n rule in (28). C e r t a i n v o w e l s are r o u n d e d in
the e n v i r o n m e n t of a tautosyllabic /.

(35) [+back, ahigh] -» [+round] / ]

T h u s there is a c o n t r a s t b e t w e e n u n r o u n d e d V in holy, lowly, Julie versus r o u n d e d

V in goal, fool, pole. A g a i n this r o u n d i n g is c a r r i e d over from the unaffixed f o r m s
to t h o s e f o r m s affixed at Level 2, even t h o u g h the c o n d i t i o n i n g / has b e e n m o v e d
to the o n s e t of the affix syllable. T h a t is, the syllable-final / in goal conditions
r o u n d i n g b e f o r e affixation to create a c o n t r a s t b e t w e e n m o n o m o r p h e m i c holy and
bimorphemic goalie.

(36) [ - round] [+round]

holy goal
bowler (hat) bowler 'one who bowls'
Julie fooling

6.3. L o n d o n pause—paws

In L o n d o n E n g l i s h (data from Wells, 1 9 8 2 : 3 1 0 - 3 1 3 a n d H a r r i s , 1991), the

vowel o h a s t w o a l l o p h o n e s : [ o s ] and [ o u ] . T h e [ o s ] variant o c c u r s in final o p e n
syllables, the o t h e r variant [ou] o c c u r s e l s e w h e r e .

(37) [09] [ou]

saw, soar3° sauce
law, lore lord
paw, pore thought
228 Toni Borowsky

O n c e again, the c o m p l e m e n t a r y distribution seen in the m o n o m o r p h e m i c e x a m -

ples d o e s not carry over to f o r m s d e r i v e d at Level 2. W h i c h e v e r variant is found
in unaffixed f o r m s is found as well before suffixes.

(38) [oo] [ou]

bored board
paws pause
poorly Crawley
law term Law ton

T h e affixes d o not close the syllables for the rule. Instead the rule applies as if the
final vowel of the stem w a s in an o p e n syllable, itself final. T h e rule thus applies
on the w o r d cycle before any Level 2 m o r p h o l o g y .
T h e r e are a few other rules w h i c h a p p e a r to m e to b e e x a m p l e s of w o r d - c y c l e
rules; however, I d o not h a v e e n o u g h information a b o u t t h e m to be absolutely
sure. T h e s e rules exhibit the s a m e sort of patterning s h o w n by all the previous
examples.
P r e - / B r e a k i n g (Wells, 1982) m a k e s a m o n o s y l l a b i c w o r d disyllabic before /.
A s usual, the effects of this rule carry o v e r to affixed forms in w h i c h the / appears
n o w to b e in the onset of the affix syllable.

(39) feel [fiol/fiyol] rule [ruol/ruwol]

feeling [fiolirj/fiyolirj] ruling [ruolirj/ruwolirj]

Sledd ( 1 9 5 8 ) m e n t i o n s an a l l o p h o n i c alternation found in dialects of S o u t h e r n

A m e r i c a n w h i c h again s h o w s similar distribution patterns. T h e r e are t w o allo-
p h o n e s of the III vowel. O n e is found followed by tautosyllabic r as well as w h e n
the r is followed b y Level 2 affixes. T h e other is found e l s e w h e r e .

(40) [IO] [i:]

beer
beery Erie
jeer
jeering hero

ACKNOWLEDGMENTS

I thank Geert Booij, Allan James, John Harris, Harry van der Hulst, John McCarthy,
Norval Smith, and three anonymous reviewers for their comments on earlier versions of
this paper. John Harris, Paul Kiparsky, and John McCarthy also provided me with addi-
tional examples of word-level rules. I am also grateful to students in my seminars at the
University of Delaware and Rijksuniversiteit te Leiden, in particular Barbara Bullock and
On the Word Level 229

Mark Verhijde, as well as to audiences at the Lexical Phonology Workshop, Seattle; the
University of Texas, Austin; and the University of Massachusetts, Amherst. The article was
completed while I had the honor of being a fellow at the Netherlands Institute for Advanced
Study (NIAS), and I am grateful indeed to the Institute for providing me with such a pleas-
ant and wonderful environment.

NOTES

1
Three general types are distinguishable:

1. The interactive model: Each morphological operation is input to a phonological cycle

(e.g., Kiparsky, 1982; Mohanan, 1986; Pesetsky, 1979).
2. The noninteractive model: All morphology precedes all phonology (e.g., Halle and
Kenstowicz, 1989; Halle and Vergnaud, 1987; Odden, this volume; Sproat, 1985).
3. The combination model: Morphology feeds phonology cyclically at some levels and
applies in a block before phonological rules at another (Booij and Rubach, 1985;
Halle and Mohanan, 1985; Kiparsky, 1985).
2
There has been very little in-depth study of the phonology associated with the stress-
neutral affixes, as far as I know. The processes discussed in Section 2, which have been
widely discussed, and more recently the phenomena discussed by Dunlap (1987), Hall
(1989a), and Harris (1989, 1991) are the only ones to have received much attention.
3
In Borowsky (1986, 1989) I showed that the pattern of so-called lexical properties only
holds at the first level. Halle and Mohanan (1985) argue that the second level violates the
strict cycle condition; the cyclicity of this level has been called into question by Booij and
Rubach (1987), Halle and Mohanan (1985), and Halle and Vergnaud (1987), for example.
4
See note 5.
5
A reviewer has suggested that the optional desyllabification I propose may not be an
optional postlexical rule. The reviewer cites examples like: whistling, wrestling, hustling,
crippling, enabling, and troubling, all of which s/he finds acceptable with nonsyllabic /l/,
while examples like Sicily, Esalen, oscillatory, insolent, and sibilant are rather better with
syllabic /If. I suggest that the latter cases are different in that they all have underlying
vowels (either full vowels or schwa). These vowels, even when they are destressed, are
different from the syllabic sonorant cases where there is never a vowel. The two cases
merge very late—possibly in the phonetic implementation—and we can assume that any
desyllabification that takes place must take place before this. No doubt word frequency also
has some effect, as suggested by the same reviewer.
6
1 have taken most of the data from Rubach (1990). The interpretation of the facts is
mine, however. An anonymous reviewer argues that the proper treatment of sonorants is to
desyllabify them in certain environments, as suggested by Rubach. This solution misses
the generalization. The crucial fact here is that sonorants are never syllabic before Level 1
affixes even though they are optionally syllabic before those that are found at Level 2.
Desyllabification leaves this unexplained. Note that this is the same pattern found with the
English cases. The same reviewer goes on to suggest that the array of facts found for g
230 Toni Borowsky

deletion is unconvincing because it is so small: that the relevant words should simply be
marked as exceptions to g-deletion. Yet the observation that the facts pattern in one direc-
tion and not in the opposite direction is a valid one. g does not delete before the Level 1
affixes. This is significant. Marking forms as exceptions when their pattern is predictable,
even if there are only a few such forms, misses the point. True exceptions would not fall
into a pattern and we would get a random set of variations.
7
The formulation of the rule is not crucial and does not affect my point, since it is the
pattern of distribution I am concerned with. The correct analysis seems to me to be one
similar to Hall's, in which the fricative is unspecified and gets [back] by rule and is other-
wise nonback.
8
If you think about it, all the rules seen so far—English as well as German—have a
similar property. The crucial factor for all these rules is the fact that they are blind to any
material outside their immediate domain. Most of the rules I have discussed so far have the
property of applying only within the Word built in, by dint of the ] in their conditioning
environment. (Note it may equally well be [.) This seems to be similar to saying these
processes only apply tautomorphemically. The sole difference is that in these rules the word
edge is the conditioning environment.
9
In some dialects of German the rule is postlexical and in fact the forms in (10) are all
pronounced with [x] as expected.
10
N o t e that in these dialects these words are pronounced without a vowel in the suffix:
[elamentriy], [saenitriy].
1 1
Harris formulates the rule as a blank-filling rule, which gets him around the SCC
violation. However, on his analysis it remains an odd rule since it applies O N L Y at Level 1,
which is peculiar by virtue of the fact that it is clearly not structure-preserving, as he him-
self notes. I think this is dubious since it is quite unlike any of the other rules of English
which are restricted to Level 1. These are a small set of highly morphologized rules, e.g.,
Trisyllabic Laxing, Velar Softening, and Spirantization. Belfast Dentalization is quite un-
like any of these rules.
I t is not entirely clear what the effect of stress on Aitken's law is. The stress condition
1 2

is perhaps more complicated than is presented. As John McCarthy has pointed out to me,
the only unstressed vowels in English are schwas, which would not undergo the rule. The
only cases of nonschwa unstressed vowels occur word finally. He notes as well that there is
no interaction of cyclic stress assignment with the rule which supports the claim that it is a
Word-level rule.
1 3
See Harris (1990) for arguments leading to his assigning this rule to Level 2.
1 4
Booij and Rubach (1987) discuss Dutch Syllable Final devoicing ( [ — son] —>
[ — voice] / $) as an example of what they refer to as a word-level rule which applies
non(post)cyclically. That this rule is not cyclic is illustrated by alternations like held [t]
'hero' vs. heldin [d] 'heroine'. They argue further that the rule is lexical on the grounds
that it precedes two other clearly postlexical processes, Voicing Assimilation and Resyllab-
ification. This argument, based on ordering of rules, does not conclusively place Dutch
Final Devoicing in the lexicon, however, and I assume that this rule and others like it (e.g.,
English r-deletion) are very early postlexical rules which apply in the first postlexical do-
main of phonological word.
A study of the few rules associated with the rest of the Level 2 affixes reveals only
1 5
On the Word Level 231

one possible candidate which could show some phonological interaction between affix and
stem or affix and affix. That is the cluster of rules, Voicing Assimilation and epenthesis/
deletion, found with the inflectional affixes -ed and -es. I suggest, however, that these pro-
cesses are postlexical and do not take place at Level 2 at all. We know that they must apply
postlexically in certain syntactic structures: the dog's bone, the cat's milk, the horse's
mouth; as well as after postlexical reduction processes: the cat's gonna bite, the dog's
gonna scratch, the horse's gonna snort, etc. There is no evidence to force us to apply the
rules at Level 2 when the affixes are attached.
1 6
T h e same observation holds for the l-ungl affix in German. Bloomfield (1930) sug-
gested that the German affix /-chen/ should be treated as an independent form for the pur-
poses of the rule which determines the distribution of [x] and [c] exactly as I suggest in
this system. The affix must cycle independently through the phonology in order to derive
the [c] from some underlying underspecified fricative.
1 7
Another possibility which has been suggested to me is the one to be discussed in
Section 5.
1 8
Note that I exclude the compound stress rule or the nuclear stress rule. When I refer
to the regular stress rules I mean the word-stress rules. For an account of the different
compound stress phenomena compatible with my view, see, e.g., Sproat (1985), Sel-
kirk (1982).
1 9
Perhaps also /-ly/, l-yl and /-ish/—all of which retain their full vowels. It could be
that these affixes have tense vowels but are not stressed, -ize is another affix which seems
stressed to me. This affix is generally problematic, and it is sometimes claimed that it has
a dual class membership (Kiparsky, 1982; Selkirk, 1982). I personally believe this affix is
really a Level 1 affix and not stress-neutral at all—as evidenced by words like democratize,
automatize, catholicize, and others. I suggest that the affix is itself stressed and thus shifts
stress off adjacent stressed syllables. It does not have any effect on stress in a word like
standardization because there is no stress adjacent to its own stress. Stress cannot shift to
the last syllable of standard because that would cause a clash.
2 0
Note that it does not follow that all Level 2 affixes should be stressed just because
each of them has been cycled on. For example, suffixes like -ness, -less, and even -dom
with internal [a]s do not get stressed (e.g., stresslessness, mercilessness have only one
stress per word). On the other hand, affixes like un-, re-, and -hood are stressed even though
their stresses do not appear to affect the rest of the stresses of the word—e.g., reroute,
unfortunate, boyhood. Similarly, affixes consisting of one consonant, say, either would be
repaired by syllabification by the insertion of a [a], or they could escape syllabification till
later. Repairing them by means of epenthesis would have the effect of making the English
inflectional affixes l-dl and l-zl into /-ad/ and /-dz/. This could be why we cannot decide
which is the correct underlying representation for these affixes.
2 1
The only potential edge-erasing phonology we find is CV syllabification, which we
assume takes place automatically and continuously throughout the derivation (ltd, 1986).
However, even if syllabification were to be suspended, there is no evidence which would
force the syllabification of stem final Cs to apply at the word level. It could be done on the
first postlexical phonological domain, as the Dutch final devoicing facts mentioned in
note 14 suggest.
2 2
With the exception of Sproat (1985), who makes most of these observations.
232 Toni Borowsky

2 3
Following Kiparsky (1982) we could assume that the output of every cycle is a "lexi-
cal entry." This could be interpreted to mean that each form derived at Level 1 is "listed"
in a strange sense (not exactly that of Jackendoff): the lexical identity rule for each form.
2 4
Of course this is not true. We C A N produce new words by means of the morphological
and phonological rules at the first level. The lexicon is not merely a list of existing words.
But it does seem that by some measure of productivity the Word level is more so than the
Stem level.
2 5
Suppose that to be productive in the sense of Word-level productive means that the
affix is used on line in the actual creation of complex words. That is, Stem-level forms may
be analyzable but when you use them in production you just do a lexical look-up of the
whole word. It is lexically listed. On the other hand, when you use a Word-level affixed
form, you actually put it together during production. It is not available as a prepackaged
form. Psycholinguistic data, as far as I know, have shown that Level 1 phonological and
morphological processes are never affected in speech errors and so on, while Level 2 pro-
cesses are. We could suppose therefore that the base of affixation in these forms is not an
underlying representation but an actually occurring word. This would support my picture
of the lexicon. I am grateful to Donca Steriade for making these points to me.
2 6
A similar suggestion was made by an anonymous reviewer.
2 7
Conservative South African English.
2 8
It may be possible to redefine these rules as foot rules, similar to English Flapping.
Even on this assumption, though, we cannot maintain the idea that lexical syllabification is
limited to structure-building processes only. However, until more is known about the rules
in question, I leave these speculations for future research.
2 9
English and German are not the only languages which have a word cycle. For some
word-level rules in Arabic dialects see Dunlap (1987), who discusses two rules: Palestinian
Backing (from Younes, 1984) and Bedouin Hijazi Liquid Emphaticization (from Al-
Mozaini, 1982). Booij and Rubach (1987) refer to Rochet's (1973) discussion of Loi de
position and other examples from French dialects. Danish Grave Assimilation appears to
me to be a similar type of rule (see Borowsky, Ito, and Mester, 1984, and references cited
therein). Most of the well-known word-final cluster simplification rules, especially in Indo-
European languages (Catalan: see, e.g., Kiparsky, 1985; Danish: Ito, 1984; Icelandic: e.g.,
Kiparsky, 1985; Ito, 1986; etc.), appear to be examples of the same Word-level prosodic
licensing phenomena as the English cases in (3).
30
R e c a l l that London English is a nonrhotic dialect: lore and law are homophonous.

REFERENCES

Al-Mozainy, (1981). Vowel Alternations in a Bedouin Hijazi Arabic Dialect: Abstractness

and Stress. Doctoral dissertation, University of Texas, Austin.
Aronoff, M., and Sridhar, S. N. (1983). Morphological Levels in English and Kannada; or
Atarizing Reagan. Papers from the Parasession on the Interplay of Phonology, Mor-
phology and Syntax, pp. 3 - 1 6 . Chicago Linguistic Society, Chicago.
On the Word Level 233

Bloomfield, L. (1930). German c and x. Le Maitre Phonetique 3 20, 2 7 - 2 8 .

Booij, G., and Rubach, J. (1987). Postcyclic versus postlexical rules in lexical phonology.
Linguistic Inquiry 18, 1 - 4 4 .
Borowsky, T. (1986). Topics in the Lexical Phonology of English. Doctoral dissertation,
University of Massachusetts, Amherst.
Borowsky, T. (1989). Syllable Codas in English and Structure-Preservation. Natural Lan-
guage and Linguistic Theory 7, 146-166.
Borowsky, T., Ito, J., and Mester, A. (1984). The formal representation of ambisyllabicity:
Evidence from Danish. Proceedings of the North Eastern Linguistic Society 14,
38-48.
Dunlap, E. (1987). Tensing in New York English. Unpublished manuscript, University of
Massachusetts, Amherst.
Hall, T.-A. (1989a). Lexical Phonology and the distribution of German [c] and [x]. Pho-
nology 6, 1-17.
Hall, T.-A. (1989b). German syllabification, the velar nasal, and the representation of
schwa. Linguistics 27, 8 0 7 - 8 4 2 .
Halle, M. (1988). Why Phonological Strata Should Not Include Affixation. Unpublished
manuscript, Massachusetts Institute of Technology, Cambridge.
Halle, M., and Kenstowicz, M. (1989). On Cyclic and Noncyclic Stress. Unpublished
manuscript, Massachusetts Institute of Technology, Cambridge.
Halle, M., and Mohanan, K. P (1985). Segmental Phonology of Modern English. Linguis-
tic Inquiry 16, 1 5 - 1 1 6 .
Halle, M., and Vergnaud, J.-R. (1987). An Essay on Stress. MIT Press, Cambridge, Mass.
Harris, J. (1989). Toward a lexical analysis of sound change in progress. Journal of Lin-
guistics 25, 3 5 - 5 6 .
Harris, J. (1991). Derived phonological contrasts. In Studies in the Pronunciation of En-
glish; A Commemorative Volume in Honour of A. C. Gimson (S. Ramsavan, ed.),
pp. 8 7 - 1 0 5 . Croom Helm, London.
Hayes, B. (1982). Extrametricality and English stress. Linguistic Inquiry 13, 2 2 7 - 2 7 6 .
Ito, J. (1984). Consonant Loss in Danish and Phonological Theory. Paper presented at ICU
Summer Institute in Linguistics, Tokyo.
Ito, J. (1986). Syllable Theory in Prosodic Phonology. Doctoral dissertation, University of
Massachusetts, Amherst.
Ito, J. (1989). A prosodic theory of epenthesis. Natural Language and Linguistic Theory 7,
217-259.
Kahn, D. (1976). Syllable Based Generalizations in English Phonology. Doctoral disser-
tation, Massachusetts Institute of Technology, Cambridge.
Kiparsky, P. (1982). From cyclic phonology to lexical phonology. In The Structure of Pho-
nological Representations. Part I (H. van der Hulst and N. Smith, eds.), pp. 131 - 1 7 5 .
Foris, Dordrecht.
Kiparsky, P. (1985). Some consequences of lexical phonology. Phonology Yearbook 2,
85-138.
Lieber, R. (1979). On the Organization of the Lexicon. Doctoral dissertation, Massachu-
setts Institute of Technology, Cambridge.
Mohanan, K. P. (1986). The Theory of Lexical Phonology. Reidel, Dordrecht.
234 Toni Borowsky

Rochet, B. (1973). On the Status of the Word in French Phonology. International Review
of Applied Linguistics in Language Teaching 25, 187-196.
Rubach, J. (1990). Final devoicing and cyclic syllabification in German. Linguistic Inquiry
21, 7 9 - 9 4 .
Selkirk, E. O. (1982). The Syntax of Words. MIT Press, Cambridge, Mass.
Simpson, J. (1980 Cyclic Syllabification and a First Cycle Rule of Vowel-rounding in Some
Dialects of Australian English. Unpublished manuscript, Massachusetts Institute of
Technology, Cambridge.
Sledd, J. (1958). Some questions of English phonology. Language 34, 2 5 2 - 2 5 8 .
Sproat, R. (1985). On Deriving the Lexicon. Doctoral dissertation, Massachusetts Institute
of Technology, Cambridge.
Wells, J. (1982). Accents of English, 3 vols. Cambridge University Press, Cambridge.
Younes, M. (1984). Emphasis and the low vowels in Palestinian Arabic. In Working Papers
in Cognitive Science (J. McCarthy and A. Woodbury, eds.).
STRUCTURE PRESERVATION AND
POSTLEXICAL TONOLOGY IN DAGBANI

LARRY M. HYMAN
Department of Linguistics
University of California
Berkeley, California 94720

1. I N T R O D U C T I O N

A m o n g t h e m o s t attractive ideas m o t i v a t i n g t h e f r a m e w o r k of lexical p h o -

n o l o g y is the v i e w that a single rule m a y apply in different p l a c e s within the p h o -
n o l o g y b u t s h o w different effects, d e p e n d i n g o n w h e r e it applies. F o r i n s t a n c e , it
h a s b e e n h y p o t h e s i z e d that a s t r u c t u r e - c h a n g i n g rule a p p l y i n g lexically can apply
only to d e r i v e d d o m a i n s , m u s t b e structure-preserving, and m a y h a v e lexical ex-
c e p t i o n s . O n the other h a n d , the s a m e rule a p p l y i n g postlexically m a y h a v e j u s t
the o p p o s i t e p r o p e r t i e s . T h a t is, it m a y apply across the b o a r d (i.e., also to n o n -
d e r i v e d d o m a i n s ) , it m a y i n t r o d u c e n e w s e g m e n t s , and it is n o t e x p e c t e d to h a v e
lexical e x c e p t i o n s . W h i l e these p u r p o r t e d differences are still b e i n g tested (and
h a v e s o m e t i m e s b e e n c h a l l e n g e d in the literature), there c a n b e n o d o u b t about
1

the g e n e r a l t e n d e n c y for lexical rule application to b e of a different c h a r a c t e r from

its postlexical counterpart.
In this article I e x a m i n e the issue of structure p r e s e r v a t i o n (SP) as it p e r t a i n s to
the t o n o l o g y of D a g b a n i , a G u r l a n g u a g e s p o k e n in G h a n a . T h e issue of interest
is that D a g b a n i h a s a p e r v a s i v e constraint forbidding c o n t o u r tones t h r o u g h o u t
m o s t of the tonology. B y SP, a rule w h e r e b y a h i g h t o n e (H) spreads o n t o a follow-
ing l o w - t o n e (L) syllable a u t o m a t i c a l l y delinks that L, so as to avoid the u n a c -
c e p t a b l e contour. A t a later stage in the derivation, however, the s a m e H - s p r e a d i n g
rule reapplies, this t i m e W I T H O U T L - d e l i n k i n g . A s a result, H L c o n t o u r tones

235
Phonetics and Phonology, Volume 4 Copyright © 1993 by Academic Press, Inc.
Studies in Lexical Phonology All rights of reproduction in any form reserved.
236 Larry M. Hyman

ultimately surface in violation of SP. T h e key issue here c o n c e r n s the place in

D a g b a n i p h o n o l o g y at w h i c h S P h a s apparently turned off. A s w e shall see, this
level d o e s not c o r r e s p o n d to the lexical/postlexical distinction postulated by the
theory. Finally, since multiple application of t o n e - s p r e a d i n g h a s a c u m u l a t i v e ef-
fect, the strong d o m a i n h y p o t h e s i s ( S D H ) of K i p a r s k y (1984) predicts that a H
tone c o u l d potentially spread as m a n y syllables a w a y from its u n d e r l y i n g position
as there are strata or levels within the l a n g u a g e - p a r t i c u l a r p h o n o l o g y . T h u s , a sec-
o n d issue in this study c o n c e r n s the a p p r o p r i a t e n e s s of the S D H in predicting the
o b s e r v e d tonal facts in D a g b a n i .
T h e p a p e r is o r g a n i z e d as follows. In Section 2 I p r e s e n t the m o r p h o l o g i c a l and
p h o n o l o g i c a l structure of n o u n s in D a g b a n i , including their lexical tonal represen-
tation. In Section 3 I establish that the rule of H - t o n e S p r e a d i n g alluded to a b o v e
applies at the postlexical level but h a s t w o different effects: o n e s u p p o r t i n g versus
o n e violating SP. In Section 4 I c o n s i d e r different interpretations of these facts,
a r g u i n g finally in the c o n c l u s i o n in Section 5 for a modification b o t h of S P and of
the S D H .

2. L E X I C A L T O N O L O G Y

For e x p o s i t o r y r e a s o n s , m o s t of t h e d i s c u s s i o n in this article c o n c e r n s the real-

ization of t o n e on n o u n s or on w o r d s that follow n o u n s . Before addressing the
2

specific tonal properties of n o u n s in D a g b a n i , a few general o b s e r v a t i o n s c o n c e r n -

ing the overall system are in order. First, the t o n e - b e a r i n g unit ( T B U ) in this lan-
g u a g e is the syllable (or, p e r h a p s m o r e appropriately, the h e a d m o r a of the syl-
l a b l e ) . S e c o n d , a surface T B U m a y b e realized on either a H or a L pitch level,
3

or it can b e realized with a H L falling contour. T h e c o r r e s p o n d i n g L H rising c o n -

tour d o e s not exist in D a g b a n i . Finally, as will b e seen in m a n y of the e x a m p l e s ,
the l a n g u a g e p o s s e s s e s a surface p h o n e m i c d o w n s t e p , contrasting the s e q u e n c e s
H - H and H - H . ! 4

A s a p r e l i m i n a r y to u n d e r s t a n d i n g the n o m i n a l t o n o l o g y of D a g b a n i , the e x -
a m p l e s in (1) are d e s i g n e d to illustrate that the vast majority of n o u n s e n d with an
overt ( n o u n class) suffix. 5

(1) a. bi-d 'child' pi. bi-hi

b . pdg-d 'woman' pi. pdg-bd
c. tib-li 'ear' pi. tlb-d
d. wdb-gu 'elephant' pi. wdb-ri
e. bih-im 'milk'

O n the other h a n d , as seen in (2), a limited of n o u n s d o not show an overt suffix. 6

(2) a. bd 'father' ma 'mother'

zd 'starch' zd 'friend' (~zdri)
Structure Preservation in Dagbani 237

b. kddu 'banana' kurcu 'pig'

kurwa 'pot' bdaji 'bag'
c. dbdbdi 'plantain' dkdrmd 'drummer'
W h i l e those in (2a) are native, the n o u n s in (2b) and (2c) are all b o r r o w i n g s (the
latter set s h o w i n g an initial [a], the only vowel that can b e g i n a D a g b a n i n o u n —
a n d only in b o r r o w i n g s ) . W h i l e t h e majority of n o u n s form their plural b y substi-
tuting a different n o u n class suffix, as seen a b o v e in ( l a - d ) , the n o u n s in (2) all
form their plural by a d d i n g the self-standing form mmd, as in bd mmd ' f a t h e r s ' ,
kddu mmd ' b a n a n a s ' .
I n o w s h o w that bisyllabic n o u n s consisting of a m o n o s y l l a b i c stem p l u s an
overt suffix fall into o n e of four general t o n e classes, as distinguished in ( 3 a - d ) .
(3) CVC.CV CV.CV CV.V
a. jieb-gd 'crocodile' pdg-d 'woman' dd-6 'man'
b. wdb-gu 'elephant' sdn-d 'stranger' no-6 'chicken'
c. ddb-li 'slave' wa-hu 'horse' nd-d 'chief
d. zdb-gu 'hair' kpdrj-a 'guineafowl'

T h e three c o l u m n s in (3) r e p r e s e n t the three different bisyllabic structures that are

attested: C V C . C V , CV.CV, and CV.V. W h i l e the n o u n s in (3a) and (3b) h a v e the
s a m e surface realization in isolation, it h a p p e n s that their tonal p r o p e r t i e s are quite
different in context. Finally, only three t o n e patterns w e r e n o t e d on CV.V n o u n s ,
w h o s e vocalic suffix is u n d e r l y i n g /-a/.
I n o w a r g u e that n o u n class suffixes are all u n d e r l y i n g l y H, and that the tonal
classes in ( 3 a - d ) can b e d i s t i n g u i s h e d b y r e p r e s e n t i n g the u n d e r l y i n g s t e m tones
as follows ( c f . G o a d , 1 9 8 8 ) . 7

(4) a. /pag/ 'woman' c. /war/ 'horse'

L
b. /san/ 'stranger' d. /kparj/ 'guineafowl'
H HL

A s indicated in (4), it is p r o p o s e d that stems such as in (3a) are u n d e r l y i n g l y

toneless, those in (3b) are H , t h o s e in (3c) are L, and t h o s e in (3d) are H L .
To justify this position, c o n s i d e r first the n o u n + adjective s e q u e n c e s in (5).
(5) a. / p a g + tita-li/ —» pag titd-li 'big woman'
H H
b. /san + tita-li/ —» sd:n titd-li ' b i g stranger'
H H H
c. / w a r + tita-li/ —» war titd-li 'big horse'
L H H
d. /kparj + tita-li/ -> kpd: n titd-li
1
'big g u i n e a f o w l '
HL H H

A s seen, w h e n followed by an adjectival n o u n such as titd-li ' b i g ' ( w h i c h consists

of a H stem + H n o u n class suffix), a modified n o u n a p p e a r s w i t h o u t its suffix. 8
238 Larry M. Hyman

T h e surface tones of (5b,c) follow directly from the u n d e r l y i n g representations

w i t h o u t modification. T h e H of the u n d e r l y i n g H L of /kparj/ ' g u i n e a f o w l ' is real-
ized on the n o u n stem in (5d), w h i l e the L c a u s e s the following H ' s of the adjective
to b e l o w e r e d to a d o w n s t e p . In (6a) the toneless n o u n stem / p a g / ' w o m a n ' re-
ceives a L t o n e b y default.
T u r n i n g to a m o r e c o m p l e x set of alternations, c o n s i d e r the realization of these
s a m e n o u n s w h e n followed b y a H tone n o u n in (6).

(6) a. pdg-d yi-li 'woman's house' [pay yfli]

b. sdn-d yi-li
f
'stranger's house' [san yfli]
!

c. wd-hu yi-li 'horse's house' [wah yfli]

d. kpdrj-'d yi-li 'guineafowl's house' [kparj yiTi]!

In (6) these n o u n s a p p e a r within the genitive construction, w h i c h in D a g b a n i c o n -

sists of a s i m p l e j u x t a p o s i t i o n of p o s s e s s o r + p o s s e s s e d . A s seen in the p h o n e t i c
9

transcriptions to the right, the final v o w e l of the p o s s e s s o r is generally elided w h e n

not p h r a s e final, a p h e n o m e n o n that is quite w i d e s p r e a d in G u r l a n g u a g e s (see,
e.g., Rialland, 1980). A s seen in (6a,b), the t w o H - H tone patterns in (3a,b) again
s h o w a difference in context: n o u n s such as sdn-d ' s t r a n g e r ' condition a following
d o w n s t e p , w h i l e n o u n s such as pdg-d ' w o m a n ' d o not. In (6c), the L - H of wd-hu
has b e c o m e L - L , w h i l e in (4d), the H - L of kpdrj-d ' g u i n e a f o w l ' is realized
as H - H .
!

T h e realizations in (6) are those found quite generally w h e n full n o u n s are

followed b y any w o r d b e g i n n i n g with a H tone. B y "full n o u n s " I m e a n n o u n s
that consist of a stem 4- suffix, as w e h a v e seen. In (6c) the u n d e r l y i n g L - H
s e q u e n c e of / w a r - h u / ' h o r s e ' b e c o m e s L - L b y the postlexical L - t o n e S p r e a d i n g
rule in ( 7 a ) . 1 0

(7) a. a a a b. a a a

L H H H © H //

T h e H - H s e q u e n c e on kpdrj- d in (6d) is directly attributable to the fact that the

! 1

n o u n s t e m / k p a r j / is u n d e r l y i n g l y H L a n d the following suffix -a is u n d e r l y i n g l y

H. In isolation such n o u n s are realized H - L as the result of the late rule in (7b),
w h i c h lowers a single H to L before p a u s e .
!

T h i s leaves the p r o b l e m of a c c o u n t i n g for the difference b e t w e e n (6a) and (6b).

T h o u g h b o t h are realized H - H in isolation, n o u n s such as sdn-d ' s t r a n g e r ' con-
dition a d o w n s t e p o n a following H, w h i l e n o u n s such as pdg-d ' w o m a n ' d o not.
In this p r o p o s a l , the u n d e r l y i n g representations are as in (8).

(8) a. [[pag] a] 'woman' c. [[war] hu] 'horse'

H L H
b. [[san] a] 'stranger' d. [[kparj] a] 'guineafowl'
H H HL H
Structure Preservation in Dagbani 239

T h e lexical t o n e rules that are n e e d e d to derive t h e surface f o r m s of t h e s e n o u n s

are f o r m a l i z e d in (9).

(9) a. MEEUSSEN'S RULE: b. H ANTICIPATION: C. DEFAULT L .

o" (a) @ o- @

H H H L

i
L

T h e rule in (9a) dissimilates a H to L w h e n i m m e d i a t e l y p r e c e d e d b y a H (as

h a p p e n s w i t h " M e e u s s e n ' s R u l e " in B a n t u ) . B y a rule of H - t o n e A n t i c i p a t i o n , (9b)
spreads the H of the suffix o n t o a p r e c e d i n g toneless n o u n stem [thereby a c c o u n t -
ing for the H t o n e that a p p e a r s o n (a) n o u n s w h e n s u f f i x e d ] . Finally, (9c) s h o w s
11

that after (9a) a n d (9b) h a v e applied, any T B U that is still toneless r e c e i v e s a L

t o n e b y default. T h e o u t p u t of the three rules in (9) p r o v i d e s the lexical r e p r e s e n -
tation of e a c h n o u n . Finally, in the c a s e of H - L n o u n s derived b y (9a), a postlexi-
cal rule of r i g h t w a r d H - t o n e S p r e a d i n g ( H T S ) applies, w h i c h also delinks t h e L,
x

as f o r m a l i z e d in (10).

(10) (<x)

H L

T h e t w o H - H n o u n p a t t e r n s can b e n o w derived as follows.

(11) [[pag]a] [[san]a] Underlying representations

I I I
H H H

[san-a] Meeussen's Rule (9a)

I I
H L
[pag-a] H-tone Anticipation (9a)

N
H
H-tone Spreading (10)

V
H L

[pag-a] [san-a( )]
!
Surface representation

A s seen in the left c o l u m n , the H - H of pdg-d ' w o m a n ' is derived by anticipat-

ing the H of the suffix -d o n t o the p r e c e d i n g toneless stem / p a g / . T h e H - H of 1 2

sdn-d ' s t r a n g e r ' is d e r i v e d in quite a different way. First, M e e u s s e n ' s R u l e con-

240 Larry M. Hyman

verts the u n d e r l y i n g H - H s e q u e n c e to H - L . T h e n H - t o n e S p r e a d i n g (a postlexical

rule) spreads the H, delinking the L. T h e configuration in (12a) of an u n l i n k e d L
tone w e d g e d b e t w e e n H ' s is realized as a H followed by a d o w n s t e p p e d H . T h u s ,
!

(6b) w o u l d b e r e p r e s e n t e d as in (12b).
(12) a CT cr t>. !
yi-li 'stranger's house'

I I v v
H L H H L H
T h e floating L following such n o u n s as sdn-d ' s t r a n g e r ' h a s n o effect before
p a u s e . However, as p o i n t e d out b y W i l s o n ( 1 9 7 0 : 4 0 9 ) , n o u n s such as pdg-d and
sdn-d h a v e different surface realizations w h e n a c c o m p a n i e d b y the vocative into-
nation w h i c h involves a l e n g t h e n e d final v o w e l .

(13) a. pdg-dd 'Woman!' b . sdn-dd 'Stranger!'

A s seen in (13b), the floating L is, in this c a s e , allowed to link o n t o the l e n g t h e n e d

final vowel, suggesting that the e x t r a length constitutes an additional T B U (syl-
lable?). W i t h these derivations and additional o b s e r v a t i o n s established, w e can
n o w turn to the theoretical p r o b l e m s p o s e d b y the rule(s) of H - t o n e S p r e a d i n g .

3. P O S T L E X I C A L T O N O L O G Y

In Section 2 I p r e s e n t e d full n o u n s followed b y a p o s s e s s e d n o u n with initial H

tone [cf. (4)] and n o u n s t e m s followed b y adjectives with b o t h initial H [cf. (5)]
a n d initial L [cf. ( 7 ) ] . W h a t w a s not s h o w n w a s the realization of the four full n o u n
forms followed by a p o s s e s s e d n o u n with initial L. T h i s g a p is n o w filled by the
data in (14).

(14) a. pag-a kodu —• pdg(-d) kodu 'woman's banana'

V II
H L H
b. san-a kodu —• sdn(-d) kodu 'stranger's banana'

I I II
H L LH
C. wa-hu kodu —• wd-h(u) kodu 'horse's banana'

I I II
L H L H
d. kparj-a kodu • kpdrj-'(d) kodu 'guineafowl's banana'

HL H LH

T h e input tones indicated on pdg-d and sdn-d in (14a,b) are t h o s e lexically derived
in (11). In the outputs in (14), the final v o w e l of the p o s s e s s o r n o u n is s h o w n in
p a r e n t h e s e s since, as w a s already said, it is subject to reduction, even deletion.
Structure Preservation in Dagbani 241

T h e derivations of (14c,d) are straightforward, i n v o l v i n g the s p r e a d i n g of the H

of the first suffix o n t o the L syllable of k o d u ' b a n a n a ' . T h e postlexical application
of H T S [rule (10)] c a u s e s this L to delink, w h i c h d e r i v e s i n t e r m e d i a t e ko du.
t
!

H o w e v e r , as already p o i n t e d out, since there is only o n e H T B U before p a u s e ,

!

'du b e c o m e s L. Finally, the H on the -hit suffix in (14c) is lost, either as a result
of the vowel r e d u c t i o n p r o c e s s or by t h e rule of L - t o n e S p r e a d i n g in (7a).
T h e intrigue of the a b o v e c o m e s from the c o m p a r i s o n of (14a) and (14b).
W h e r e a s H T S j will derive the correct o u t p u t in (15a), it will not p r o d u c e the
correct output in ( 1 5 b ) .

(15) a. pag(-a) kodu [pa7 kodu] (<pdg ko'du)

H LH
b. san(-a) kodu *[san kodu] (cf. [san kodu])

V\ I I
H L L H
H T S j applies b e t w e e n the t w o n o u n s in (15a) to derive the i n t e r m e d i a t e form
s h o w n in the right c o l u m n , w h i c h t h e n u n d e r g o e s final l o w e r i n g of H t o L. In
!

(15b), H T S j s p r e a d s the stem H of first n o u n o n t o the following suffix L, w h i c h

then d e l i n k s . A s s h o w n , this should p r o d u c e a H - L - H s e q u e n c e in the output.
W h a t is o b s e r v e d in ( 1 4 b ) , however, is a H L falling t o n e on [ k o d u ] . B u t w h e r e
d o e s the H of this fall c o m e f r o m ? A n d crucially, r e g a r d i n g the issue of SP, w h y
1 3

d o e s n ' t the L of the falling tone delink in a c c o r d a n c e with the p e r v a s i v e constraint

against H L t o n e s ?
T h e d a t a in (14a,b) s h o w that w h e n a H t o n e spreads o n t o a L - H s e q u e n c e
postlexically, t w o c o n t r a s t i n g realizations are obtained: H - H (simplified to H - L )
!

a n d H L - H . In (16) w e o b s e r v e a similar o p p o s i t i o n w h e n the L t o n e adjective

/ k a r - h 7 ' b i g ' follows these n o u n s .

(16) a. pag kar-li [pag kar-li] 'big woman'

L L H
b. sa: rj kar-li [sa:T| kar-li] 'big stranger'

H L H
C. war kar-li [war kar-li] 'big horse'

L L H
d. kpa:rj kar-li [kpa.T| kar-li] 'big guineafowl'

HL L H

A g a i n , the tonal inputs are those o b t a i n e d at the e n d of the lexical p h o n o l o g y .

T h e s e tones c o m b i n e to p r o d u c e the postlexical outputs in (16a) and (16c) with-
242 Larry M. Hyman

out further modification. In ( 1 6 b ) , a postlexical application of H T S correctly d e - {

rives the output ( t h r o u g h an i n t e r m e d i a t e kdr- li w h i c h simplifies to kdr-li before

!
y

p a u s e ) . In (16d), however, w e obtain [kpa:rj k a r - l i ] , with a H L falling t o n e on

[kar], apparently derived b y a H - t o n e S p r e a d i n g rule of s o m e sort.
F r o m the f o r m s in (14a,b) a n d (16b,d), w e tentatively arrive at the following
descriptive generalizations in (17).

(17) a. (a) a gives rise to (d)

H L H L
b. a a gives rise to (d) d

H L L H L

In (17a) a H - L s e q u e n c e b e c o m e s H - H followed by an u n l i n k e d L ( w h i c h can

condition d o w n s t e p on a following H ) , w h i l e in (17b), w h e r e there is an unlinked
L b e t w e e n the linked H a n d L tones, the surface instead is a H - H L s e q u e n c e . A s
final e v i d e n c e for these " g e n e r a l i z a t i o n s , " c o n s i d e r the realization of the n o u n s
in (18), first seen in (2a), w h i c h d o not h a v e an overt (i.e., s e g m e n t a l ) n o u n class
suffix.

(18) a. ma 'mother' ma kodu 'mother's banana'

A
L H L H L H
b . bd —• bd 'father' bd kodu

i i n I
H H H L H L L H
W h i l e these t w o n o u n s s h o w a single t o n e in isolation, L versus H respectively, it
is clear from the genitive f o r m s to the right that they actually h a v e a H - t o n e suffix.
In (18a) this H is a n c h o r e d o n t o the L of kodu ' b a n a n a ' , w h i c h is then delinked,
creating i n t e r m e d i a t e ko'du, w h o s e H then is lowered to kodu before p a u s e . In
!

(18b), the H suffix first dissimilates b y (9a) to b e c o m e L. T h e n , as s h o w n , s o m e -

h o w the H of bd s p r e a d s over the u n l i n k e d L o n t o the linked L of kodu to create
the H L falling tone. T h e question is h o w to b r i n g this result about.
Let us refer to the H - t o n e S p r e a d i n g rule that results in a H L falling tone as
H T S . T h e r e are at least t w o p r o b l e m s associated with H T S . T h e first h a s already
2 2

b e e n m e n t i o n e d , namely, W h y doesn't it c a u s e delinking of the L ? T h e s e c o n d

p r o b l e m is p e r h a p s even m o r e basic: H o w can any rule of H T S apply across an
u n l i n k e d L ? T h e n o r m a l e x p e c t a t i o n is that a free L in a s e q u e n c e such as w e h a v e
to the right in (18b) w o u l d BLOCK H T S , as the earliest studies on floating tones
s h o w e d . If, as I believe, floating L's should universally b l o c k H T S , then o n e p o s -
sibility is to delete t h e m , as n e e d e d . T h u s , a " n a t u r a l " solution to this p r o b l e m
w o u l d b e to r e c o g n i z e the following three rules.
Structure Preservation in Dagbani 243

(19) a. HTS, b. ©-DELETION c. HTS ?

© — - 0 / L
, A
H L

First, H T S j applies in (19a), automatically d e l i n k i n g the L (by S P ) . T h i s free L

tone (indicated b y © ) is then deleted w h e n followed by a n o t h e r L t o n e . 1 4
A t this
p o i n t H T S 2 c a n apply, as in (19c). A s seen, this rule of H - t o n e S p r e a d i n g d o e s not
delink t h e L a n d t h u s d o e s not respect SP.
O n the other h a n d , it m i g h t b e s u g g e s t e d that © - d e l e t i o n is n o t n e e d e d , a n d in
fact, that H T S 2 REQUIRES a free L tone in o r d e r to apply, as in (20a).

(20) a. b.

H © L H (©)
1 L

T h e derivation in ( 2 1 a ) , however, s h o w s that H T S d o e s n o t require a free L .

2

(21) lexical representation a. pag-a akarma b. san-a akarma

v v I I i v I
H L H H L L H
HTS, san-a akarma
pag-a akarma

M / II I i v I
H L L H
H L H
HTS. san-a akarma
pag-a akarma

H L H H © L H
Contour Simplification san-a akarma

H © L H
phonetic representation [pa^ akarma] [san akarma]

In this e x a m p l e , the n o u n [ a k a r m a ] ' d r u m m e r ' h a s a single L t o n e lexically

linked to its first t w o syllables. H T S first applies, d e l i n k i n g this L from t h e first
x

syllable of the n o u n , followed b y H T S , w h i c h f o r m s a H L c o n t o u r o n t h e s e c o n d

2

syllable. T h u s , if there is n o © - d e l e t i o n , H T S w o u l d h a v e t o b e f o r m u l a t e d as in
2

(20b), that is, with an optional free L. Finally, a rule of L absorption a p p l i e s , as

in (22), r e m o v i n g the L of a H L c o n t o u r w h e n followed directly b y a n o t h e r L or
by p a u s e .1 5

(22) CONTOUR SIMPLIFICATION:

A L, //
244 Larry M. Hyman

T h e need for this rule is seen in the last stage of the derivation in (21b), since the
output is not *[san a k a r m a ] .
T h i s , then, c o m p l e t e s the presentation of the t w o postlexical p h e n o m e n a
d u b b e d H T S and H T S . In the n e x t section I c o n s i d e r different interpretations of
] 2

these facts.

4. D I S C U S S I O N

In o r d e r to a c c o u n t for the t w o H - t o n e S p r e a d i n g p h e n o m e n a , H T S j and H T S ,

2

w e m u s t first raise the q u e s t i o n of w h e t h e r they should b e v i e w e d as o n e or t w o

rules. A s a first a r g u m e n t in favor of v i e w i n g both H T S p r o c e s s e s as a single rule,
there is the o b v i o u s formal similarity b e t w e e n t h e m : both spread a H o n t o a fol-
l o w i n g L - t o n e syllable. A s a s e c o n d a r g u m e n t , there is the fact that only the pu-
tative rule of © - D e l e t i o n (19b) is possibly o r d e r e d b e t w e e n t h e m . If there is n o
rule of © - D e l e t i o n , the t w o " r u l e s " can b e o r d e r e d c o n s e c u t i v e l y in the p h o -
nology. A s w e shall see shortly, H T S j a n d H T S 2 apply in exactly the s a m e d o -
m a i n . T h u s , a c c o r d i n g to the practices of standard generative p h o n o l o g y , H T S {

and H T S should b e collapsed, if a f o r m a l i s m can b e found.

2

However, standing in the w a y of this single-rule analysis is the fact that H T S ]

and H T S 2 h a v e slightly different p r o p e r t i e s . First, H T S j c a n n o t apply across a
free L, w h i l e H T S 2 c a n — a l t e r n a t i v e l y , as seen in (19b), © - D e l e t i o n can in-
tervene b e t w e e n H T S t and H T S , in w h i c h c a s e the t w o p r o c e s s e s c a n n o t in
2

any case be c o l l a p s e d . 16
A s a s e c o n d difference, H T S t c a u s e s the L to delink,
w h i l e H T S d o e s not, thereby f o r m i n g a H L c o n t o u r o t h e r w i s e unattested in the
2

language.
T o slightly restate these findings, there are three logical analyses, as s u m m a -
rized b e l o w in (23).

(23) a. H T S ] a n d H T S are t w o separate rules with an accidental r e s e m b l a n c e ;

2

the fact that the properties of these t w o rules differ slightly m u s t simply
b e stipulated.
b. H T S j a n d H T S constitute a single rule w h i c h h a s t w o separate ap-
2

plications; the p r o p e r t i e s of the t w o applications differ b e c a u s e of the

p l a c e in the p h o n o l o g y w h e r e e a c h application takes place.
c. H T S j and H T S constitute a single ( c o m p l e x ) rule, w h i c h applies only
2

o n c e in the postlexical tonology.

W e h a v e already c o n s i d e r e d (23a) in c o n n e c t i o n with the discussion of the rules

in (19) a n d (20). It clearly is not a p p e a l i n g to set u p t w o such similar rules, and I
thus view the solution in (23a) as a last resort. T h e solution in (23b) is the o n e I
Structure Preservation in Dagbani 245

w o u l d like to support. In o r d e r for it to g o t h r o u g h , however, it is n e c e s s a r y to

s h o w that the first application of H T S (i.e., H T S j ) o c c u r s in o n e p l a c e in the
p h o n o l o g y , w h i l e the s e c o n d application ( H T S ) o c c u r s later in a n o t h e r p l a c e .
2

Unfortunately for this analysis, HTS and H T S apply in exactly the SAME place
{ 2

in the p h o n o l o g y .
A s a first d e m o n s t r a t i o n of this fact, c o n s i d e r the derivations in (24).

v
(24) a. kddu 'banana' versus kddu titd-li 'big banana'

II v
L H
LH
b. pdg(-d) kodu titd-li 'woman's big banana'

H L H
C. sdn(-d) kddu titd-li 11
'stranger's big banana'

H L L H

In (24a) w e see that w h e n a L - H n o u n such as kddu ' b a n a n a ' is followed b y a H

t o n e , it flattens to kddu, w h i c h consists of a single L t o n e linked to b o t h syl-
l a b l e s . In (24b) the H of pdg-d ' w o m a n ' spreads twice o n t o this s a m e w o r d
1 8

kddu, first yielding kddu, then kddu. In (24c), the lexical H of the n o u n stem san
' s t r a n g e r ' first s p r e a d s o n t o the suffixal -a, delinking its L, a n d then spreads
again o n t o the first syllable of kddu ( p r o d u c i n g kddu, w h i c h then simplifies b y
L - a b s o r p t i o n to kddu, as seen).
T h e derivation in (24b) s h o w s that the s a m e w o r d m a y b e affected b y b o t h
H T S j a n d H T S . N o w c o m p a r e ( 2 5 a ) , w h e r e H T S and H T S b o t h c r o s s a w o r d
2 x 2

b o u n d a r y , thus involving the t h r e e - w o r d s e q u e n c e in the derivation.

(25) a. pdg(-d) war kar-li 'woman's big horse'

V ^ - ? " 1 I
H L L H
b. sdn(-d) war kar-li 'stranger's big horse'

W 1 II
H L L L H
T h e derivation in (25b) s h o w s the lexical H of the n o u n stem san s p r e a d i n g o n t o
the suffix -a, followed b y a s e c o n d s p r e a d i n g of this H o n t o the following n o u n
stem war ' h o r s e ' . C o m b i n i n g these o b s e r v a t i o n s , w e arrive at the c o n c l u s i o n that
both H T S j and H T S are free to apply either within or across w o r d s .
2

P e r h a p s w e m i g h t still m a i n t a i n that the t w o applications of H T S o c c u r at dif-

ferent levels in the postlexical p h o n o l o g y . A s s u m i n g the distinction m a d e by
Kaisse ( 1 9 8 5 ) , H T S j m i g h t apply at P j and H T S at P . In this c a s e w e could,
2 2

following K a i s s e ( 1 9 9 0 ) and R i c e ( 1 9 9 0 ) , say that S P is respected at P (and h e n c e x

H T S , requires L d e l i n k i n g ) , w h i l e S P is not respected at P (and h e n c e H T S is

2 2
246 Larry M. Hyman

a l l o w e d to create a H L c o n t o u r tone). Unfortunately, it h a s b e e n i m p o s s i b l e to find

a n y c o n f i r m i n g e v i d e n c e since, for instance, b o t h H T S a n d H T S are obligatory x 2

a n d n o t subject to t e m p o , stylistic, or other variables affecting their application

differentially. In addition, p r o s o d i c d o m a i n s are totally irrelevant: H T S j and
H T S will apply a n y w h e r e in an utterance u n l e s s the respective H and L tones are
2

s e p a r a t e d b y p a u s e . T h e data in ( 2 6 ) - ( 3 0 ) s h o w that both rules apply within a

n u m b e r of c o n s t r u c t i o n s w h i c h in other l a n g u a g e s w o u l d h a v e divided u p into t w o
or m o r e p h o n o l o g i c a l p h r a s e s .

(26) ( N o n b r a n c h i n g ) subject + verb:

a. d zdg-si bd 'he refused them'

i
L L
v I
H
b. pdg(-d) zag-si bd 'a woman refused them'

H L H
C. sdn(-d) zdg-si bd 'a stranger refused them'

V W I
H L L H

(27) ( B r a n c h i n g ) subject + v e r b :
a. doo rjwun jie pdg(-d) ] Su zdg-si bd 'a man who saw a woman refused them'

H L H
b. doo rjwun jie sdn(-d) ]Su zdg-si bd 'a man who saw a stranger refused them'

H L L H

(28) Object! + object : 2

a. d ti pdg(-d) kodu 19
'he gave a woman a banana'

I I \L^I
L L H LH
b. d ti sdn(-d) kodu 'he gave a stranger a banana'

I I VV1I
L L H L L H
(29) Object + a d v e r b (sdhld 'yesterday'):
a. d jie pdg(-d) sohld 'he saw a woman yesterday'

L L H L H
b. d jie sdn(-d) sohld 'he saw a stranger yesterday'

I I VV1 I
L L H L L H
Structure Preservation in Dagbani 247

(30) Object + b r a n c h i n g object [a kddu ' y o u r (sg) b a n a n a ' ] :

a. d ti pdg(-d) d kddu 'he gave a woman your banana'

I I
L L H L LH

W 1 II
b. d ti sdn(-d) d kddu 'he gave a woman your banana'

I I
L L H L L LH

S i n c e b o t h H T S a n d H T S apply a n y w h e r e within the utterance d o m a i n , it is

x 2

clear that w e c a n n o t assign t h e m to different postlexical strata or p r o s o d i c d o -

m a i n s — a n d h e n c e that w e c a n n o t a c c o u n t for their different b e h a v i o r vis-a-vis S P
in this way. H o w e v e r , there is a n o t h e r possibility that n e e d s to b e c o n s i d e r e d :
p e r h a p s H T S is a n o r m a l postlexical rule (applying a n y w h e r e within the utter-
x

a n c e ) , a n d H T S is a so-called p h o n e t i c i m p l e m e n t a t i o n (PI) rule ( L i b e r m a n and

2

P i e r r e h u m b e r t , 1984). In this case, w e w o u l d say that the w h o l e of D a g b a n i p h o -

n o l o g y r e s p e c t s the o n e - t o n e - p e r - T B U constraint, w h e r e a s the p h o n e t i c s d o e s not.
A s L i b e r m a n a n d P i e r r e h u m b e r t point out, the d i v i d i n g line b e t w e e n d o m a i n - f r e e
postlexical p h o n o l o g y a n d PI rules is not a clear o n e . B o t h are allowed to h a v e
optional a n d gradual effects (cf. Kiparsky, 1985), a n d b o t h of c o u r s e o c c u r late in
the derivation. In D a g b a n i , w e n o t e that if H T S is a PI rule, then at least three 2

rules m u s t follow it. T h e first is C o n t o u r Simplification, f o r m u l a t e d in (22), w h i c h ,

a p p l y i n g after H T S , w o u l d also h a v e to b e a PI rule. T h e s e c o n d rule, interroga-
2

tive H - L o w e r i n g , is s u m m a r i z e d in (31a).

(31) Interrogative intonation:

a. H-LOWERING: H"] C G ] Y b. FINAL LENGTHENING: V]TJ

i i
Ln
V:

A c c o r d i n g to (31a), the final H or s e q u e n c e of H ' s o c c u r r i n g within the last clitic

g r o u p ( C G ) of an interrogative utterance b e c o m e s L . 2 0
In addition, the final v o w e l
is l e n g t h e n e d , as seen in a c o m p a r i s o n of the declarative and interrogative sen-
tences in (32).

(32) a. d jie sdn-d -> d jie sdn-d: 'did h e see a s t r a n g e r ? '

'did h e see this
b. d jie sdn-d=rjd? —» d fte sdn-d—yd: 2
stranger?'
'did h e see their
c. d jie bi= sdn-d —> d fie bi=sdn-d: stranger?'
'did h e see a big
d. d jie sd:n titd-li —> d jie sd:n titd-li: stranger?'

(32b) s h o w s that the d e m o n s t r a t i v e =rjd? ' t h i s ' is an enclitic, w h i l e (32c) s h o w s

that p o s s e s s i v e p r o n o u n s such as bi— 'their' are proclitics. O n the other h a n d , a
248 Larry M. Hyman

n o u n (stem) + adjective s e q u e n c e constitutes T W O C G S , as seen in (32d). T h a t

interrogative H - L o w e r i n g is a l o w e r i n g rule, rather than s o m e kind of spreading
rule, is seen from the tonal m i n i m a l pair in (33).

(33) a. d jie wdb t-ttd-li —• o jie wdb ttid-li: 'did he see a big elephant?'
| | | \j/ I I I \|/ & -8* 'elephant')
db

L L H H L L H L
b. d jie zdb !
t4td-li —• d jie zdb !
tkd-li: 'did he see a big hair?'

L L H L H L L H L L

In e a c h c a s e it is the last H t o n e that has lowered to L, p r o d u c i n g a surface pitch

height contrast b e t w e e n L a n d L in j u s t this o n e c o n s t r u c t i o n .
! 22

T h e r e l e v a n c e of interrogative H - L o w e r i n g is seen in (34).

(34) a. d jie sdn(-d) kodu —• d jie sdn(-d) kodu: 'did he see the
| | \ | | | | | \ ' \ " ' ] | stranger's banana?'
L L H L L H L L H L L L

After l o w e r i n g h a s o c c u r r e d , w e obtain a H L - L s e q u e n c e on kodu:. H e n c e it m u s t

b e the c a s e that H - L o w e r i n g follows L-absorption. T h u s , if H T S is a P I rule, then 2

so is C o n t o u r Specification a n d so is H - L o w e r i n g ! Finally, the derivation in (35a)

s h o w s t h a t H - L o w e r i n g , w h i c h affects a single T B U before p a u s e , is a THIRD rule
!

that m u s t follow H T S . 2

(35) a. /zab-gu/ 'hair' —• zdb-'gu —• zdb-gu

I I
HL H
b. /zab-gu/ 'hair' —• zdb-gu —• *zdb-gu

I I
HL H
In ( 3 5 b ) , if H l o w e r i n g applies first to this form, then H T S will incorrectly derive
!
2

a final H L falling t o n e . W h i l e it is c o n c e i v a b l e that PI rules m i g h t b e o r d e r e d like

true p h o n o l o g i c a l rules, the result j u s t o b t a i n e d further blurs the distinction. I thus
c o n c l u d e that there is n o r e a s o n to attribute H T S to p h o n e t i c i m p l e m e n t a t i o n .
2

R e t u r n i n g to the three logical analyses listed in (23), it has b e e n hard to s u b -

stantiate the view in ( 2 3 b ) , since w e h a v e found n o c o m p e l l i n g reason to assign
H T S j a n d H T S to t w o different strata or c o m p o n e n t s . W e have yet to c o n s i d e r
2

the possibility in (23c), that b o t h p r o c e s s e s b e stated as a single ( c o m p l e x ) rule

a p p l y i n g o n c e in the p h o n o l o g y . Certainly this is consistent with the facts: w h e r -
ever both H T S and H T S can apply, they b o t h d o apply. W h a t w e n e e d is a
t 2

c o m p l e x rule that says: S p r e a d a H o n t o a following L T B U , delinking this L,

a n d if there is a s e c o n d L T B U , the H spreads also onto it, forming a H L contour.
Structure Preservation in Dagbani 249

If tones link directly to T B U s (here, syllables), then the rule is f o r m u l a t e d as

in (36a).

(36) a. a (a) a b. O O O

U4"1 U4"l
H L L H L L
A s seen, the m i d d l e T B U m u s t b e optional, b e c a u s e H T S will also apply across
2

a free L tone, as w e h a v e seen. T h e rule can be slightly simplified if w e instead

a s s u m e that t o n e s link directly to s o m e k i n d of tonal n o d e , as in (36b), in w h i c h
c a s e the m i d d l e n o d e can b e either linked or f r e e . In b o t h f o r m u l a t i o n s , I i g n o r e
23

the p r o b l e m of distinguishing b e t w e e n t w o L t o n e features, as indicated, versus

o n e L linked to t w o syllables or tonal n o d e s . A n o t h e r p r o b l e m not a d d r e s s e d is
w h a t to d o a b o u t H - L - H s e q u e n c e s , w h e r e H T S ! obligatorily applies b u t H T S 2

c a n n o t apply (since there is n o s e c o n d L t o n e ) , since o n e interpretation of the rules

in (36) is that either b o t h H - s p r e a d i n g p r o c e s s e s a p p l y — o r else neither. E v e n
a s s u m i n g that t h e s e difficulties can b e o v e r c o m e , there is r e a s o n to reject this
a p p r o a c h . O n e o b v i o u s objection is that a p h o n o l o g i c a l rule should in general b e
restricted to affecting only o n e target (though p e r h a p s a p p l y i n g iteratively to a
string of targets). If c o n t r a r y f o r m u l a t i o n s such as in (36) are allowed, then there
is n o r e a s o n w h y w e c o u l d not add o n e m o r e syllable (or tonal n o d e ) with L t o n e
a n d h a v e H T S potentially apply THREE t i m e s , and so on. In other w o r d s , despite
a n y t e m p t a t i o n o n e m i g h t h a v e to say that there is o n e postlexical p r o c e s s of H T S
that affects u p to t w o L - t o n e syllables, it is formally u n d e s i r a b l e to c o n c e p t u a l i z e
the c h a n g e of H - L - L to H - H - H L as a single rule application. In the following
section I therefore return to the idea that there is a single H T S p r o c e s s that applies
m o r e than o n c e in the p h o n o l o g y .

5. CONCLUSION

T h e position arrived at in the p r e c e d i n g section is that H T S applies at the }

postlexical level, followed b y the application of H T S at the very s a m e level.

2

S i n c e postlexical H T S j respects the " o n e - t o n e - p e r - T B U " lexical constraint, I a m

in a g r e e m e n t with K a i s s e ( 1 9 9 0 ) a n d R i c e ( 1 9 9 0 ) that S P m a y persist into t h e
postlexical p h o n o l o g y . W h a t is c u r i o u s , however, is that the constraint against H L
c o n t o u r s is arbitrarily a b a n d o n e d , since H T S a p p l i e s i m m e d i a t e l y following
2

H T S i. If this r e p r e s e n t s , as I h a v e suggested, the p o i n t at w h i c h S P b e c o m e s in-

operative, then it a p p e a r s that S P not only n e e d not turn off at the e n d of the lexi-
cal p h o n o l o g y , but also m a y turn off at any d e s i g n a t e d p o i n t in the postlexical
derivation. 24
250 Larry M. Hyman

W i t h S P s o m e w h a t in doubt, I w o u l d like to c o n c l u d e with a brief consideration

of K i p a r s k y ' s ( 1 9 8 4 ) S D H . A c c o r d i n g to the S D H , all p h o n o l o g i c a l rules poten-
tially apply from the b e g i n n i n g of the p h o n o l o g y and c o n t i n u e to apply at each
stratum until/unless they are specifically t u r n e d off. W h i l e a rule can cease to
apply at a given level (e.g., b e restricted to lexical stratum 1), a rule c a n n o t b e
turned on in the m i d d l e of the p h o n o l o g y . A n exclusively postlexical application
w o u l d b e achieved in cases w h e r e a lexical application w o u l d violate a specific
p r i n c i p l e such as SP. In the c a s e of H T S , there is n o a priori reason w h y it could
not first apply lexically. W i t h this in m i n d , w e can reinterpret the derivation of
pdg-d ' w o m a n ' of (11) a l o n g the lines of K e n s t o w i c z et al.'s (1988) analysis of
M o o r e . Instead of v i e w i n g the H of the suffix as prelinked, as in (37a), w h e r e the
H spreads leftwards o n t o the toneless n o u n stem, in (37b) w e start with the suffixal
H as unlinked.

(37) a. /pag-a/ —• pag-a b. /pag-a/ — • pag-a —• pag-a

I V I V
H H H H H
After this u n l i n k e d H associates to the leftmost T B U , H T S applies lexically to
derive the H - H s e q u e n c e . W i t h a lexical application w e see that a H m a y m i g r a t e
three syllables to the right of its initial link, s p r e a d i n g o n c e lexically, a s e c o n d
t i m e by H T S , and a third time by H T S [recall ( 2 4 b ) ] . T h e lexical application is
x 2

n e e d e d only in cases such as (37b), w h e r e I h a v e s u p p o s e d that an initial H is

followed b y a toneless T B U . In c a s e the H is followed by a L, such as in the
derivation in (38), it is n o t p o s s i b l e to tell w h e t h e r the word-initial H - t o n e spread-
ing is from putative lexical H T S or from H T S . x

(38) /san-a/ — • san-a —• sdn-d 'stranger'

II II v
H H H L H L
W h a t is clear is that if it is from lexical H T S , then this form c a n n o t then trigger
H T S ! postlexically, b e c a u s e of the floating L t o n e that b l o c k s H T S . Instead, it x

can only trigger H T S , w h i c h is not b l o c k e d b y floating L.

2

I will not resolve this a m b i g u i t y e x c e p t to n o t e that a s i m p l e r (and n o n a m b i g u -

ous) solution w o u l d b e r e a c h e d if lexical H T S w e r e restricted to a p p l y i n g to cases
w h e r e the H is followed b y a toneless T B U — u n l i k e either H T S or H T S . W h a t x 2

is interesting from the p o i n t of view of the S D H is the prediction that if H T S

applies at every stratum (lexically a n d postlexically), then o n e should o b s e r v e the
H spreading as m a n y T B U s to the right as there are strata. In a l a n g u a g e w h e r e ,
say, H T S applies at lexical stratum 1, lexical stratum 2, a n d postlexical stratum 1,
the H should spread a m a x i m u m of three T B U s to the right. R a t h e r than c o u n t i n g
to the n u m b e r 3 (or s o m e h i g h e r n u m b e r if a l a n g u a g e h a s m o r e than three strata),
the lexical m o d e l predicts that the rule c o u n t s only o n e T B U , but three t i m e s ! Is
Structure Preservation in Dagbani 251

there then such a l a n g u a g e w h i c h confirms this p r e d i c t i o n ? It w a s h o p e d that D a g -

bani w o u l d b e this l a n g u a g e . However, as I h a v e s h o w n , the multiple versions of
H T S s h o w differences (e.g., with respect to S P ) that c a n n o t b e neatly attributed to
strata.

ACKNOWLEDGMENTS

Materials for this article were gathered in the field methods course given in the Spring
of 1988 at the University of Southern California. I am indebted to Mr. Abdul Saedu, who
served as informant, and to the members of that course for their participation in the discov-
ery of the tone system of Dagbani. I would like to thank Drs. W. A. A. Wilson and A.
Naden for sending me their work on Dagbani and related Gur languages and for their com-
ments and general help during our initial analysis of these materials. For previous work on
Dagbani tonology, see especially Wilson (1970). Finally, thanks to Ellen Kaisse and an
anonymous Academic Press referee for comments.

NOTES

JFor example, Harris (1987) argues that structure preservation is not respected in the
lexical phonology in Southeastern Bantu, while Kaisse (1990) and Rice (1990) give ex-
amples where structure preservation persists into the postlexical phonology. Given the con-
clusion I reach in this study, it may be that some of the claimed lexical/postlexical distinc-
tions are simply "tendencies" that are reflected in many, but not all, languages.
2
The verbal tonology has also been studied and analyzed and supports the conclusions
reached on the basis of the nominal tonology.
3
The following tone marks are written over the first (or only) V of each syllable: ' =
H(igh) tone,' = L(ow) tone, = HL (falling) tone; marks downstep tone.
A !

4
1 might note that the drop from H to H is considerably greater in interval than in other
!

languages I have studied. As we shall see, if only one TBU occurs with the H tone before !

pause, it is actually realized L.

5
Of 148 "basic nouns," 139 (or 94%) occur with a noun class suffix.
6
We shall see below that these nouns may actually have a floating H suffix, however;
see (18).
7
This analysis was greatly facilitated by a talk given by Michael Kenstowicz in April
1988 at the University of Southern California, which has since appeared as Kenstowicz,
Nikiema, and Ourso (1988).
8
These forms show a number of segmental alternations which are irrelevant to the tonal
phonology, e.g., the length of the vowel in sd:n in (5b) and kpd:n 'guineafowl' in (5d)
and the surface [r] in war 'horse' in (5c), etc. The forms that are shown should not be
confused with the similar but not identical phonetic transcriptions in (4), where the suffix
252 Larry M. Hyman

is present but is subject to final vowel reduction. To underscore this point, consider the
corresponding genitive and pre-adjective realizations of jieb-gd 'crocodile' and wdb-gu
'elephant', whose isolation forms were originally cited in (3a,b):

W jieb-g1
yi-li 'crocodile's house'
wdb-gi •'yi-li 'elephant's house'

(ii) jieb titd-li 'big crocodile'

wdb titd-li 'big elephant'

In (i) the vowel of the suffixes -gd and -gu is replaced by a very short, perhaps epenthetic
centralized vowel following the Igl of the suffix. In (ii), on the other hand, both this vowel
and the Igl are missing, again illustrating that a noun appears bare (without suffix) when
modified by an adjective.
9
This contrasts with the situation in other Gur languages such as Moore and Lama,
where Kenstowicz et al. (1988) have shown the need for a L tone genitive particle.
10
T h o u g h irrelevant to the analysis, it is possible that at this postlexical level the two
H's would have already conflated as a single H linked to the two syllables.
1 1
In Section 5 I shall consider an alternative whereby the H of the suffix comes in
unlinked, associates first to the noun stem, and then spreads rightward.
1 2
According to the common practice of filling a gap in the pattern, one might propose
that nouns such as /pag/ are actually underlyingly LH. Though not an insurmountable prob-
lem, there are at least two reasons for not doing this. First, there is the question of why a
H-tone suffix should cause a LH contour to become H, rather than simplifying to L [as in
fact happens elsewhere in the language by the rule of contour simplification in (22)]. Sec-
ond, there is the problem that this H does not surface when followed directly by a L
tone—cf. (16a).
13
I t cannot literally be from H T S ] , as conceived up to this point, since the L of kddu
would then be expected to delink, ultimately yielding *sdn kodu.
( 19b) could also apply to delete the free L tone before pause, since in that position it
1 4

also has no surface effect. On the other hand, free L's are not deleted before H tone, since
they are needed to condition downstep.
Cf. Wilson (1970:414): "A sequence HL other than V(C)CV becomes HH before L."
15

The effect of this rule is to guarantee that a HL falling tone will surface only if it is followed
by a H tone, a constraint that is found in a number of other tone languages, e.g., Kinande,
Luganda, etc. There is an (intonational) exception to this, however; see (34).
I f HTS] did apply across the free L tone in (15b) and (16d), for instance, we would
1 6

obtain the incorrect outputs *[san kodu] (from intermediate *sdn ko'du) and * [kparj kar-
li] (from intermediate *kpdjj kdr-'li), respectively.
1 7
After H T S applies to this form and others cited below, a H - H L sequence is obtained,
2

e.g., intermediate kddu in (24c). As transcribed, this fall is simplified by contour simplifi-
cation (22).
18
T h i s rule of L-tone Spreading (LTS) spreads rightwards onto a H TBU if the latter is
in turn also followed by a H TBU, i.e., L - H - H becomes L - L - H . The H - H sequence either
may be a single H that has spread by HTS j , or it may be a separate H feature from the next
morpheme. In either case the rule applies. (For discussion of the relevance of such ex-
Structure Preservation in Dagbani 253

amples to the feature geometry of tone within a parametric framework, see Hyman and
Pulleyblank, 1988.)
1 9
I n this example [and also in (29a)], only HTS j can apply, since after the initial spread-
ing of H, the intermediate output of 'banana' (ko'du) does not have a following L TBU For
H T S to apply. (Recall that this intermediate representation becomes kodu before pause.)
2

2 0
T h i s differs slightly from Wilson's (1970:414) statement of the rule ["Before the ?
[interrogative] marker a final H tone or H tone sequence is lowered to L back as far as the
last Downstep of the utterance (if any)"] because of the data in (32d).
2 1
(32b) also shows that when a form would normally take a glottal stop prepausally in
a declarative utterance (cf. Hyman, 1989), the glottal stop is not observed within the cor-
responding interrogative.
22
T h e r e is an alternative to H-Lowering whereby a boundary interrogative L delinks the
H('s) on the appropriate TBU(s). Assuming that the affected sequences of H's have all been
fused into a single H-tone feature at the CG level, the rule of H-Lowering simply targets
the last H of the domain. In this alternative, this H would have to link to single tonal node,
or else it is not clear how a boundary L can dislodge it from a sequence of TBUs. For this
reason, and also for the reason that we know we need intonational rules that modify lexical
tones (see, e.g., Hyman, 1990), I stick with the analysis given above.
2 3
This seems to be what Wilson's (1970:413) rule is intended to do, which changes a
[ + h ] [ - h] [ - h] sequence into [ + h ] [ + h ] [ + h ] , independent of whether the second [ - h]
is linked or not.
2 4
Since the discussion of SP centers around a single tonal constraint, what is needed is
a demonstration that more than one such constraint turns off at the same arbitrary point in
the postlexical phonology. Unfortunately, this is hard to demonstrate. It can be pointed out
that the constraint against LH contours is never violated in Dagbani. However, none of the
three rules that were said to follow H T S (L-absorption, interrogative H-Lowering, H -
2
!

Lowering) would in any case produce a LH contour—which COULD conceivably surface if

the above analysis is correct.

REFERENCES

Goad, H. (1988). Tone in the Dagbani Noun Phrase. Unpublished manuscript, University
of Southern California, Los Angeles.
Harris, J. (1987). Non-structure-preserving rules in lexical phonology: Southeastern Bantu
harmony. Lingua 73, 2 5 5 - 2 9 2 .
Hyman, L. M. (1989). The phonology of final glottal stops. In Proceedings of the West
Coast Conference on Linguistics 18, 113 - 1 3 0 .
Hyman, L. M. (1990). Boundary tonology and the prosodic hierarchy. In The Phonology-
Syntax Connection (S. Inkelas and D. Zee, eds.), pp. 1 0 9 - 1 2 5 . CSLI Publications and
University of Chicago Press, Chicago.
Hyman, L. M., and Pulleyblank, D. (1988). On feature copying: parameters of tone rules.
In Language, Speech and Mind (L. M. Hyman and C. N. Li, eds.), pp. 3 8 - 4 8 . Rout-
ledge, London.
254 Larry M. Hyman

Kaisse, E. M. (1985). Connected Speech: The Interaction of Syntax and Phonology. Aca-
demic Press, Orlando.
Kaisse, E. M. (1990). Toward a typology of postlexical rules. In The Phono logy-Syntax
Connection (S. Inkelas and D. Zee, eds.), pp. 1 2 7 - 1 4 3 . CSLI Publications and Uni-
versity of Chicago Press, Chicago.
Kenstowicz, M., Nikiema, E., and Ourso, M. (1988). Tonal polarity in two Gur languages.
Studies in the Linguistic Sciences 18, 7 7 - 1 0 3 .
Kiparsky, P. (1984). On the lexical phonology of Icelandic. In Nordic Prosody III (C.-C.
Elert et al., eds.), pp. 135-162. University of Umea, Umea, Sweden.
Kiparsky, P. (1985). Some consequences of lexical phonology. Phonology Yearbook 2,
85-138.
Lieberman, M., and Pierrehumbert, J. (1984). Intonational invariance under changes in
pitch range and length. In Language Sound Structure (M. Aronoff and R. T. Oehrle,
eds.), pp. 1 5 7 - 2 3 3 . MIT Press, Cambridge, Mass.
Rialland, A. (1980). Marques de ponctuation et d'integration dans l'enonce en gurma. Bul-
letin de la Societe de Linguistique de Paris 7 5 , 4 1 5 - 4 3 2 .
Rice, K. D. (1990). Predicting rule domains in the phrasal phonology. In The Phonology-
Syntax Connection (S. Inkelas and D. Zee, eds.), pp. 2 8 9 - 3 1 2 .
Wilson, W. A. A. 1970. External tonal sandhi in Dagbani. African Language Studies 11,
405-416.
(POST) LEXICAL RULE APPLICATION

GREGORY K. IVERSON
Department of Linguistics
University of Wisconsin—Milwaukee
Milwaukee, Wisconsin 53201

1. INTRODUCTION

T h e t h e o r y of lexical p h o n o l o g y d e v e l o p e d b y K i p a r s k y ( 1 9 8 2 ) p r o p o s e s a fun-
d a m e n t a l distinction b e t w e e n LEXICAL and POSTLEXICAL r u l e s — t h e latter b u t n o t
the f o r m e r apply across the b o a r d , that is, w i t h o u t r e g a r d for derivational h i s t o r y
or m o r p h o l o g i c a l c o m p o s i t i o n ; are typically e x c e p t i o n l e s s ; a n d c o m m o n l y p r e s e n t
only n o n n e u t r a l i z i n g or " a l l o p h o n i c " effects. A major descriptive feature of the
theory as originally c o n c e i v e d is that a given p h o n o l o g i c a l rule h a s m e m b e r s h i p
in o n e or the other of t h e s e t w o c a t e g o r i e s , b u t n o t both. K i p a r s k y ( 1 9 8 5 ) a r g u e s
that in certain c a s e s , however, specifically nasal c o n s o n a n t assimilation in C a t a l a n
a n d o b s t r u e n t voice assimilation in R u s s i a n , a single rule m u s t b e a c c o r d e d post-
lexical as well as lexical status. T h i s relaxation of the t h e o r y is n e c e s s i t a t e d for
Catalan, h e m a i n t a i n s , by the a p p a r e n t o r d e r i n g p a r a d o x that in s o m e d e r i v a t i o n s
nasal assimilation m u s t b o t h p r e c e d e and follow another, p r e s u m a b l y lexical, rule.
In the p r e s e n t article it is s h o w n that the C a t a l a n o r d e r i n g p a r a d o x falls a s i d e u n d e r
m o r e g e n e r a l c o n c e p t i o n s of rule interaction a n d feature representation, a n d that
the t w o rules actually b o t h apply only within the postlexical d o m a i n . T h i s result
in turn h a s i m p o r t a n t i m p l i c a t i o n s in lexical p h o n o l o g y for i m p l e m e n t a t i o n of the
derived e n v i r o n m e n t constraint on p h o n o l o g i c a l p r o c e s s e s s u c h as palatalization
in K o r e a n , w h o s e effects are s o m e t i m e s derivationally restricted, s o m e t i m e s not.
D e s p i t e a p p e a r a n c e s to the contrary, rules like t h e s e m a y also apply within a single
c o m p o n e n t of the p h o n o l o g y if it is only structure-preserving applications of rules

255
Phonetics and Phonology, Volume 4 Copyright © 1993 by Academic Press, Inc.
Studies in Lexical Phonology All rights of reproduction in any form reserved.
256 Gregory K. Iverson

w h i c h m u s t o b s e r v e the d e r i v e d e n v i r o n m e n t constraint, and not, as c o n v e n t i o n a l

lexical p h o n o l o g y c o n v e r s e l y h a s it, that any rule restricted to derived e n v i r o n -
m e n t s m u s t also b e s t r u c t u r e - p r e s e r v i n g . A further c o n s e q u e n c e of this association
b e t w e e n structure p r e s e r v a t i o n a n d the d e r i v e d e n v i r o n m e n t constraint is that p h o -
n o l o g i c a l feature underspecification will b e d e t e r m i n e d within the lexicon, inde-
p e n d e n t of the c o n t e x t u a l r e q u i r e m e n t s of p a r t i c u l a r p h o n o l o g i c a l rules.

2. CATALAN

A s d e s c r i b e d b y M a s c a r o ( 1 9 7 6 ) , the nasal / n / in C a t a l a n regularly a s s i m i -

lates in p l a c e of articulation to an i m m e d i a t e l y following c o n s o n a n t . K i p a r s k y
( 1 9 8 5 : 9 5 ) illustrates w i t h the following alternations in the p r o n u n c i a t i o n of son
'(they) a r e ' .

(1) unassimilated: so[n] amies 'they are friends'

labial: so[m] pocs 'they are f e w '
labiodental: so[m] feligos 'they are h a p p y '
dental: so[n] dos 'they are t w o '
alveolar: so[n] sincers 'they are s i n c e r e '
postalveolar: so[n] rics 'they are r i c h '
laminopalatal: SO[q] [Hermans'they are b r o t h e r s '
palatal: 1
SO[q] [X]iures 'they are free'
velar: so[t)] grans 'they are b i g '

T h e other p h o n e m i c n a s a l s in the system, bilabial / m / and palatal / j i / , d o not

assimilate at all, e x c e p t that Iml b e c o m e s l a b i o d e n t a l before a n o t h e r labiodental:
c o m p a r e so[m] dos ' w e are t w o ' with so[rq\ feligos ' w e are h a p p y ' . B u t the pala-
tal r e m a i n s palatal in a[ j i ] felig ' h a p p y y e a r ' , and so on, and the velar nasal, w h i c h
derives as in o t h e r I n d o - E u r o p e a n l a n g u a g e s from a nasal p l u s stop cluster, always
r e m a i n s velar: ft'frj] pa T have b r e a d ' , and so on.
T h e o r d e r i n g p a r a d o x in C a t a l a n c o n c e r n s the interaction of N a s a l A s s i m i l a t i o n
with a n o t h e r g e n e r a l rule that deletes a stop h o m o r g a n i c with a p r e c e d i n g c o n s o -
nant w h e n the cluster is tautosyllabic, that is, not i m m e d i a t e l y followed by a vowel
in the s a m e w o r d . H e n c e , C l u s t e r Simplification r e d u c e s / m p / to Iml in / k a m p /
camp 'field' w o r d finally ( [ k a m ] ' t h e field', [ k a m es] ' t h e field i s ' ) and in c o n s o -
nantally suffixed c o n s t r u c t i o n s ( [ k a m s ] 'fields'), b u t it d o e s n o t a p p l y in forms
w h e r e the suffix is vowel-initial: [ k a m p £t] 'little field'. S i n c e the nasal's p l a c e of
articulation is p r e d i c t a b l e from the following u n d e r l y i n g stop before it deletes, the
least r e d u n d a n t lexical representation of [ k a m ] w o u l d be as / k a N p / , with the nasal
unspecified for p l a c e of articulation. Similarly, the velar nasal in [ben,] venc T
sell' w o u l d d e r i v e from an underspecified r e p r e s e n t a t i o n / b e N k / , as w o u l d the
alveolar in [bin] vint ' t w e n t y ' , p r e s u m a b l y from / b i N t / .
(Post) Lexical Rule Application 257

A s K i p a r s k y illustrates in his derivation of M a s c a r a ' s e x a m p l e venc vint pans

[b£rj b i m p a n s ] T sell t w e n t y loaves of b r e a d ' , accordingly, t w o results m u s t b e
assured. First, o n c e underspecified / N / b e c o m e s either labial or velar in assimila-
tion to its following c o n s o n a n t , it m u s t n o t u n d e r g o assimilation again after cluster
reduction has applied, for [b£rj b i m ] d o e s n o t b e c o m e * [ b £ m b i m ] . But, second,
the c o r o n a l realization of underspecified / N / is subject to assimilation a s e c o n d
time, b e c a u s e i n t e r m e d i a t e [bin p a n s ] ( < /biNt p a n + s / ) b e c o m e s [bim p a n s ] .

(2) /beNk biNt pan+s/

bcrjk bint Nasal Assimilation
bcrj bin C l u s t e r Simplification

bim Nasal Assimilation

[b£rj bim pans]

T h e o r d e r i n g of the rules in this f a s h i o n — p u t t i n g aside for the m o m e n t the

q u e s t i o n of h o w to prevent reapplication of N a s a l Assimilation in the s e q u e n c e
[b£rj b i m ] — v i o l a t e s the irreflexivity a s s u m p t i o n of standard generative practice
(a rule m a y not p r e c e d e itself) b e c a u s e N a s a l A s s i m i l a t i o n is applied both before
and after C l u s t e r Simplification. K i p a r s k y r e m o v e s this part of the p r o b l e m for the
theory first by o r d e r i n g N a s a l A s s i m i l a t i o n before Cluster Simplification in the
lexical p h o n o l o g y p e r se, then b y a s s i g n i n g N a s a l A s s i m i l a t i o n to t h e postlexical
p h o n o l o g y as well. A s N a s a l A s s i m i l a t i o n thus a p p e a r s in t w o separate, i n d e p e n -
dently s e q u e n c e d c o m p o n e n t s of the p h o n o l o g y , its application both before a n d
after Cluster Simplification in the derivation of the [m] in [bim p a n s ] turns out to
be simply a derivative property of the rule's m u l t i p l e c o m p o n e n t i a l a s s i g n m e n t
rather than a stipulation contradicting the p r e c e p t s of extrinsic o r d e r i n g .
T h e other part of the p r o b l e m in a derivation like (2), however, is that postlexi-
cal application of nasal assimilation m u s t b e restricted j u s t to / n / , b e c a u s e labials,
palatals, and velars d o not assimilate in p l a c e of a r t i c u l a t i o n . K i p a r s k y h e r e p r o -
2

poses to c h a r a c t e r i z e the nasal w h i c h d o e s assimilate as underlyingly c o m p l e t e l y

unspecified for p l a c e of articulation features, as / N / . O n this i m p l e m e n t a t i o n of
radical underspecification, the assimilating nasal will b e r e p r e s e n t e d w i t h o u t p l a c e
features not j u s t w h e n it a p p e a r s before a t a u t o m o r p h e m i c c o n s o n a n t , as in vint,
but even w h e n it d o e s not, as in son. In fact, as he later points out, this idea can b e
e x t e n d e d (and a r g u a b l y should b e — c f . , e.g., Y i p , 1 9 9 1 , for English; Iverson,
1989, for K o r e a n ; Paradis and Prunet, 1989, for n u m e r o u s other l a n g u a g e s ) to all
the anterior c o r o n a l s b e c a u s e p l a c e of articulation in general need be specified
only for labials, palatals, and velars; that is, place features should be u n m a r k e d for
the dental or alveolar o b s t r u e n t s ft d s/, too. For this class of c o n s o n a n t s n o w with
no inherent place of articulation specifications at all, a p p r o p r i a t e coronal place
features will be filled in later in the derivation by default rules w h e n e v e r a s s i m i -
lation (in the case of nasals) has not applied instead.
C h a r a c t e r i z i n g all t h e anterior c o r o n a l s as unspecified for p l a c e of articulation
features m e a n s that the p r o b l e m a t i c derivation in (2), w h i c h m o t i v a t e d a s s i g n m e n t
258 Gregory K. Iverson

of N a s a l A s s i m i l a t i o n to b o t h the lexical and postlexical c o m p o n e n t s in the first

p l a c e , takes on a rather different character.

(3) /beNk biNT paN+S/

berjk Nasal Assimilation
bcrj biN Cluster Simplification
bim Nasal Assimilation

pan+s Default P l a c e F e a t u r e s
[b£rj bim pans]

In (3), even coronal o b s t r u e n t s are systematically underspecified for p l a c e of

articulation features. B u t w h i l e N a s a l A s s i m i l a t i o n still applies before C l u s t e r
Simplification in the derivation of [b£rj] a n d after it in the derivation of [ b i m ] ,
there is n o violation of irreflexivity b e c a u s e it is not the c a s e that Nasal A s s i m i l a -
tion m u s t a p p l y t w i c e in the SAME derivation. H e r e , rather, the rules interact freely
( o r d e r e d locally in the sense of A n d e r s o n , 1974, or not o r d e r e d at all, p e r Kout-
s o u d a s , S a n d e r s , a n d N o l l , 1974). N a s a l A s s i m i l a t i o n simply applies w h e n e v e r its
structural description is m e t : before Cluster Simplification in / b e N k / , w h e r e velar
p l a c e features are p r e s e n t ; but not at all in / b i N T / , w h e r e p l a c e features are not
available until the u n d e r s p e c i f i e d / N / c o m e s into j u x t a p o s i t i o n with labially speci-
fied / p / in the following w o r d via the application of Cluster Simplification. C o n -
sistent radical underspecification of p l a c e of articulation features thus naturally
r e m o v e s the a p p a r e n t o r d e r i n g p a r a d o x in Catalan, p r o v i d e d that the m o d e of rule
interaction p e r m i t s N a s a l A s s i m i l a t i o n to apply to any representations that satisfy
its structural description.
W i t h n o o r d e r i n g p a r a d o x , Cluster Simplification a n d N a s a l A s s i m i l a t i o n m a y
b o t h apply exclusively within the s a m e c o m p o n e n t of the p h o n o l o g y . K i p a r s k y
m a i n t a i n s that C l u s t e r Simplification m u s t b e a lexical rule b e c a u s e it a p p e a r s not
to interact with, that is, n o t b e b l e d by, the postlexical resyllabification that takes
p l a c e b e t w e e n w o r d s ( / k a N p e s / —> / k a m p e s / —> [ k a . m e s ] , not * [ k a m . p e s ] ; cf.
/ k a N p + £ t / —> / k a m p + £ t / —> [kam.p£t]). B u t there are g o o d r e a s o n s to s u p p o s e
that the rule is postlexical instead. First, if r e m o v a l of the constituent-final stop is
a c c o m p l i s h e d t h r o u g h stray e r a s u r e of a c o n s o n a n t not h a v i n g satisfied the t e m -
plate for syllabification, as deletion p h e n o m e n a in general are characterized u n d e r
the p r o s o d i c theory of the syllable (Ito, 1986, 1989), then C a t a l a n e m e r g e s as j u s t
the r e v e r s e of o t h e r l a n g u a g e s with p l a c e of articulation constraints on the syllabic
i n c o r p o r a t i o n of c o n s o n a n t clusters. For e x a m p l e , in D i o l a F o g n y (Sapir, 1965),
clusters m a y o c c u r at the e n d of the w o r d j u s t in c a s e they are h o m o r g a n i c , but in
h e t e r o r g a n i c clusters the s e c o n d c o n s o n a n t is deleted. C a t a l a n is quite the o p p o s i t e
of this since it freely tolerates h e t e r o r g a n i c clusters word-finally (serp ' s n a k e ' ,
pore ' p i g ' , bosc ' f o r e s t ' ) , r e d u c i n g only the h o m o r g a n i c o n e s . If ltd (cf. also G o l d -
smith, 1990) is right that the relevant place of articulation constraint o n c o n s o n a n t
clusters h a s to d o with the licensing of appropriate p h o n o l o g i c a l features in the
(Post) Lexical Rule Application 259

syllable c o d a , then any l a n g u a g e w h i c h p e r m i t s word-final h e t e r o r g a n i c clusters

(two p l a c e specifications) should also allow h o m o r g a n i c clusters (one p l a c e speci-
fication). T h e fact that C a t a l a n contradicts this implication suggests that other fac-
tors are at play in the d e t e r m i n a t i o n of its surface syllabification.
T h o s e factors, G o n z a l e z ( 1 9 8 9 ) clarifies, c o n c e r n h o w tightly b o u n d the w o r d
c o n t a i n i n g the final c o n s o n a n t cluster is with the following vowel-initial w o r d . At
a major syntactic constituent break, as b e t w e e n N P and V P in el camp ( [ k a m ] ) es
. . . ' t h e field is . . .', cluster simplification i n d e e d is essentially obligatory. B u t it
is optional w h e n the association b e t w e e n constituents is tighter, as within N P s like
el camp ( [ k a m p ] / [ k a m ] ) espanyol ' t h e S p a n i s h field', el pont ( [ p 6 n t ] / [ p 6 n ] ) alt
'the tall b r i d g e ' , a n d so on. Word-final p r e v o c a l i c clusters syllabify entirely within
the c o d a w h e n at the e n d of a major syntactic p h r a s e , in other w o r d s , but optionally
split b e t w e e n the c o d a of o n e syllable a n d the onset of a n o t h e r w h e n they are
internal to the p h r a s e . T h i s syllabification m u s t b e postlexical b e c a u s e it is sensi-
tive to constituent c a t e g o r i e s larger than the w o r d , the m a x i m u m d o m a i n of the
purely lexical c o m p o n e n t . A s it is defined on a s e q u e n c e of h o m o r g a n i c c o n s o -
nants at the e n d of the syllable, w h i c h s o m e t i m e s straddles the division b e t w e e n
w o r d s , application of C l u s t e r Simplification t h u s d e p e n d s o n postlexical syllabi-
fication after all, w h i c h is the reason it is m o r e likely to apply at the e n d of an N P
than at the e n d simply of a N o u n . 3

(4) c c 1

I , Jo-
Root Root

Place

L i k e Cluster Simplification, N a s a l A s s i m i l a t i o n m u s t also b e a postlexical rule,

since it t o o applies b e t w e e n w o r d s (cf. [bim p a n s ] ) , and so the rules n o w interact
as illustrated in ( 3 ) . 4

B u t a n o t h e r part of K i p a r s k y ' s r e a s o n i n g that N a s a l A s s i m i l a t i o n m u s t apply

postlexically (besides, in his view, lexically) is that the underlyingly unspecified
nasal takes on even the nondistinctive place features of a following c o n s o n a n t , as
exemplified in (1). T h e anterior coronal o b s t r u e n t stops are dental rather than al-
veolar, for e x a m p l e , so before t h e m IN/ a s s u m e s dental articulation too; before
labiodental fricatives, similarly, IN/ (as well as / m / ) is labiodental, and it is post-
alveolar before p o s t a l v e o l a r /r/, and so on. B y the p r i n c i p l e of structure p r e s e r v a -
tion ( " l e x i c a l rules c a n n o t i n t r o d u c e or refer to r e d u n d a n t f e a t u r e s " ) , crucially,
assimilation to n o n c o n t r a s t i v e place of articulation features like [ ± d i s t r i b u t e d ] —
w h i c h distinguishes bilabials from labiodentals, dentals from alveolars, and post-
alveolars from l a m i n o p a l a t a l s — i s prohibited e x c e p t in the postlexical phonology,
w h e r e " a l l o p h o n i c " features are first m a d e reference to. If the rule of Nasal A s -
similation is to a c c o u n t for the o c c u r r e n c e of labiodental, dental, postalveolar, and
260 Gregory K. Iverson

laminopalatal nasals in Catalan u n d e r lexical p h o n o l o g y ' s a s s u m p t i o n of structure

preservation, therefore, it m u s t b e t h r o u g h the rule's postlexical application. T h i s
is consistent, of c o u r s e , with the status of Nasal A s s i m i l a t i o n (and Cluster S i m p l i -
fication) as strictly postlexical, but it turns out to b e quite b e s i d e the point w h e n
various p h o n e t i c detail and default rules are taken into a c c o u n t . 5

T h u s , as noted too b y G o l d s m i t h ( 1 9 9 0 ) in this c o n n e c t i o n , the a s s i g n m e n t of

particularly labiodental p l a c e of articulation to labial fricatives but bilabial place
to labial stops in C a t a l a n is itself a postlexical matter, b e c a u s e n o potential contrast
exists in this l a n g u a g e b e t w e e n the t w o refinements of labial place specification:
for t h e m the feature [distributed] is entirely r e d u n d a n t here. A postlexical rule
spelling out predictable labiodental articulation for labial fricatives, and bilabial
for all other labials, w o u l d b e as in (5).

(5) a. T—sonorant "1

Labial —> [ — distributed] / + continuant

b. Labial -> [+distributed]

F o l l o w i n g these r e q u i r e m e n t s , If I is filled out as labiodental, Ipl and Iml (by d e -

fault) as bilabial.
U n d e r the obligatory c o n t o u r principle ( O C P ) (e.g., M c C a r t h y , 1986), m o r e -
over, w h i c h h o l d s that adjacent identical features are prohibited, a s e q u e n c e of
Iml followed b y If I in the s a m e p h o n o l o g i c a l d o m a i n will fall u n d e r o n e Labial
articulator n o d e (in the g e o m e t r i c sense of Sagey, 1986) rather than t w o separate
o n e s . W h e n rule (5a) applies to If I, therefore, it p r o v i d e s the feature [-distributed]
to the Labial n o d e w h i c h If I shares with the p r e c e d i n g nasal in an /mf/ s e q u e n c e ,
and so both s e g m e n t s s i m u l t a n e o u s l y b e c o m e l a b i o d e n t a l . W h e n followed b y Ipl
6

(or by n o c o n s o n a n t at all), on the other h a n d , the inapplicability of (5a) allows

(5b) to c o m p l e t e the representation of Iml as bilabial.

— sonorant
(6) a. [ + nasal] +continuant b . [+nasal] [ — sonorant]

Place Place

Labial Labial

[—distributed] [-(-distributed]
([mf]) ([mp])
(Post) Lexical Rule Application 261

A n a n a l o g o u s a c c o u n t a c c o m m o d a t e s the other s u b p h o n e m i c places of ar-

ticulation. Since dental III ([t]) and Idl ([d]) are phonetically [ + d i s t r i b u t e d ] , / n /
(< INI) in /nt/ ([nt]) and / n d / ([nd]) clusters a c q u i r e s that feature t o o b y virtue of
the s a m e r e d u n d a n c y rule w h i c h assigns [ + d i s t r i b u t e d ] to an anterior c o r o n a l stop
sharing its P l a c e n o d e with the nasal. Before the n o n a n t e r i o r c o r o n a l s , w h i c h c o m -
prise the [ + d i s t r i b u t e d ] laminopalatal fricatives / s , z7 and the palatal sonorants
A , j i / , the P l a c e - s h a r i n g nasal also b e c o m e s [ + d i s t r i b u t e d ] ; but e l s e w h e r e , such
as w h e n sharing P l a c e with the r e m a i n i n g c o r o n a l s / s , z, 1, r/ or w h e n i n d e p e n -
dent of other s e g m e n t s (phrase finally, before a v o w e l ) , it is default-specified as
[ - d i s t r i b u t e d ] . Finally, a s s u m i n g with K i p a r s k y that p o s t a l v e o l a r / r / and the pala-
tals are [ + h i g h ] , this feature too automatically accrues to the nasal w h e n it shares
its P l a c e n o d e with a following / r / ; o t h e r w i s e , t h e nasal s h a r i n g P l a c e with a
7

coronal is given the default value of [ - h i g h ] .

Before these postlexical r e d u n d a n c i e s c o m e into play, however, the unspecified
nasal in a cluster will h a v e a c q u i r e d its basic p l a c e of articulation from a following
labial, palatal, or velar via the p h o n o l o g i c a l rule of Nasal A s s i m i l a t i o n .

(7) [+nasal] [+consonantal]

Place

" P l a c e " d o m i n a t e s o n e of the articulators Labial, C o r o n a l , or Dorsal. T h e only

nasal to acquire o n e of these specifications b y the s p r e a d i n g effects of rule (7), as
K i p a r s k y ' s a c c o u n t also h a s it, is the unspecified o n e , b e c a u s e that is the only nasal
w h i c h d o e s not already h a v e an o c c u p i e d P l a c e n o d e . T h u s , the palatal (Coronal)
nasal d o e s not assimilate to a following dental (initially unspecified, later C o r o -
nal), n o r the bilabial o n e (Labial) to a following velar (Dorsal), b e c a u s e these
nasals already are specified for Place articulators; inherently unspecified / N / ,
t h o u g h , d o e s assimilate to the Labial, C o r o n a l , or D o r s a l Place of a following
c o n s o n a n t b e c a u s e this fills the void in its o w n representation. T h e r e a s o n w h y
Iml assimilates only to a following labiodental is that Iml shares its Place specifi-
cation with following labials, but not with other kinds of s e g m e n t s , a n d p h o n e t i c
detail rules like (5a) spelling out the labiodental quality of If I also affect P l a c e -
shared Iml.
U n d e r this g e o m e t r i c system of feature representation, to recapitulate, the labial
nasal is identified by p r e s e n c e of the Labial articulator, the palatal nasal b y p r e s -
e n c e of the C o r o n a l articulator d o m i n a t i n g the feature [-anterior], and the velar
nasal by p r e s e n c e of the D o r s a l articulator [which it only acquires via (7) from a
following stop]. W h e n (7) d o e s not c o m e into play to p r o v i d e features to the lexi-
cally unspecified nasal INI, its e m p t y Place n o d e is filled out by default, and the
s e g m e n t e m e r g e s as alveolar [n].
262 Gregory K. Iverson

(8) Coronal Place

[+anterior]

R u l e s like (5) defined o n either anterior or n o n a n t e r i o r c o r o n a l s w o u l d not affect

palatals adjacent to dentals since, a l t h o u g h they h a v e C o r o n a l in c o m m o n , their
r e m a i n i n g specifications u n d e r P l a c e are not identical and so c o u l d not b e shared,
j u s t as D o r s a l and L a b i a l are i n d e p e n d e n t w i t h respect to a velar p r e c e d i n g a bi-
labial. B u t in view of the operation of postlexical p h o n e t i c detail rules like (5) on
linked as well as u n l i n k e d m a t r i c e s , the precise h o m o r g a n i c i t y of clusters c o m -
p o s e d of nasals sharing basic p l a c e of articulation with a following c o n s o n a n t
is d u e not to a rule of assimilation p e r se, but rather to the unifying effects of
the O C R
T h i s c i r c u m s t a n c e limits the postlexical application of Catalan Nasal A s s i m i -
lation following Cluster Simplification to s e q u e n c e s b e t w e e n w o r d s of the un-
specified nasal followed b y a c o n s o n a n t with an inherent articulator n o d e , as in
vint pans [bim p a n s ] . A p p l i c a t i o n of N a s a l A s s i m i l a t i o n p r i o r to Cluster Simpli-
fication also affects only s e q u e n c e s consisting of the unspecified nasal followed
by a c o n s o n a n t with an inherent P l a c e specification. All other values of nasals are
either u n d e r l y i n g or filled in b y p h o n e t i c detail c u m default rules as exemplified
in (5). A s a c o n s e q u e n c e , there is n o r e q u i r e m e n t for either of the Catalan rules
d i s c u s s e d h e r e to b e a c c o r d e d lexical as well as postlexical status; that is, both are
interpretable as strictly postlexical.
T h e celebrated n e u t r a l i z i n g / a l l o p h o n i c rule of R u s s i a n voice assimilation dis-
c u s s e d at length b y K i p a r s k y ( 1 9 8 5 ) will also apply within a single c o m p o n e n t of
the p h o n o l o g y if an idea p u t forward b y M a c f a r l a n d and P i e r r e h u m b e r t ( 1 9 9 1 ) is
a d o p t e d (cf. P l a p p , 1990), w h o c o n s i d e r structure preservation to b e subject to the
a u t o s e g m e n t a l linking c o n d i t i o n that " a s s o c i a t i o n lines are interpreted as e x h a u s -
t i v e " ( H a y e s , 1986). O n this r e a d i n g of structure preservation, novel s e g m e n t s
m a y b e i n t r o d u c e d even into the lexical p h o n o l o g y p r o p e r if they are the p r o d u c t s
of assimilation, or feature spread, b e c a u s e the resulting shared feature configura-
tions d o not stand in violation of m o n o s e g m e n t a l l y defined statements w h i c h ex-
c l u d e a l l o p h o n e s from the p h o n e m e inventory. F o r e x a m p l e , the voiced fricative
[7] could b e i n t r o d u c e d into the lexical c o m p o n e n t of R u s s i a n p h o n o l o g y , despite
the fact that it is strictly an a l l o p h o n e of / x / , b e c a u s e its [ 4 - voice] feature is a l w a y s
shared with a n e i g h b o r i n g voiced sound, never i n d e p e n d e n t l y ascribed to [ 7 ] . U n -
d e r this (rather extensive) relaxation of structure preservation, both neutralizing
and n o n n e u t r a l i z i n g applications of R u s s i a n voice assimilation can b e directly ac-
c o m m o d a t e d within the lexical p h o n o l o g y p e r s e — i n d e e d , for that matter, within
the postlexical c o m p o n e n t instead, since neutralizable inputs to the rule apparently
(Post) Lexical Rule Application 263

always h a p p e n to b e crucially derived. B u t there are other clear c a s e s for w h i c h ,

in view of the standard t h e o r y ' s restriction of j u s t lexical rules to derived e n v i r o n -
m e n t s , it w o u l d s e e m that dual lexical/postlexical status n o n e t h e l e s s is necessary.
T h e following sections e x p l o r e t h e significance of this sort of rule, w h i c h , w h e n
neutralizing, exhibits familiar derivationally restricted effects, but w h e n not neu-
tralizing applies w i t h o u t restriction. It will b e a r g u e d that rules like these limit
structure-preserving applications to d e r i v e d e n v i r o n m e n t s as a m a t t e r of general
principle, rather than of c o m p o n e n t i a l a s s i g n m e n t , a n d as a result that the appli-
cations w h i c h fail to respect structure preservation are i n d e p e n d e n t of c o n s i d e r a -
tions of a u t o s e g m e n t a l linking.

3. KOREAN

T h e palatalization of c o r o n a l s in K o r e a n is o n e such case, for in s o m e forms it

takes p l a c e only in d e r i v e d e n v i r o n m e n t s ; in others it applies across t h e b o a r d . 8

T h e l a n g u a g e ' s o b s t r u e n t s y s t e m consists of the t h r e e - w a y stop contrasts, includ-

ing o n e affricate series, a n d the t w o - w a y fricative contrasts illustrated in (9).

(9) p t c k s (Lax)
ph t h gh k h (Aspirated)
p' t' c' k' s' (Tense)
Of these, the plosives /t t / c o n v e r t to t h e affricates / c c / w h e n in position before
h h

[i] or [y] in t h e n e x t m o r p h e m e , in w h i c h position t h e fricatives /s s 7 also t a k e o n

p o s t a l v e o l a r articulation ([s s ' ] ) . T h e palatalization p r o c e s s thus gives rise to al-
9

ternations such a s in (10) (with r e d u n d a n t voicing of t h e lax n o n c o n t i n u a n t s in

intervocalic c o n t e x t s ) .

(10) /tat-/ 'close' [taft'a] (indie.) [taji] (noun)

/tot-/ 'rise' [toft'a] (indie.) [toji] (noun)
/pat -/
h
'field' [pat ttl]
h
(obj.) [pac i] h
(subj.)
/os-/ 'cloth' [osttl] (obj.) [osi] (subj.)

U n d e r the o b s t r u e n t system of (9), the effect of palatalization in the first three

of these forms is neutralizing, b u t in the fourth it is not. T h e r e are m a n y appar-
ent e x c e p t i o n s to t h e rule, however, b e c a u s e n o t only d o t h e affricates a p p e a r
m o r p h e m e - i n t e r n a l l y before [i], as in ( l i b ) , b u t so d o the coronal stops, as e x e m -
plified in (11a).

(11) a. [madil 'knot' ( < /mati/) b. [cip ]n

'house'
[p9t i] h
'endure' [c'ija] ' t e a r ' (imp.)
[t i] h
'dust' [cijs] 'bark'(imp.) 1 0
264 Gregory K. Iverson

U n l i k e the stops, the t w o fricatives u n d e r g o palatalization even in m o r p h e m e -

internal e n v i r o n m e n t s . 11

(12) [si] 'poem' (*[si])

[sikan] 'time' (*[sikan])
[s'i] 'seed' (*[s'i]). 12

T h e r e w o u l d clearly s e e m to b e a single generalization here, n a m e l y that coron-

als palatalize before [i] (and [y]), e x c e p t that, if it w o u l d m e r g e an underlying
contrast, the p r o c e s s is b l o c k e d in i n t r a m o r p h e m i c c o n t e x t s . This classic derived
e n v i r o n m e n t restriction m e a n s that K o r e a n palatalization u n d e r p r e s e n t theoretical
a s s u m p t i o n s m u s t b e a lexical rule, for it is not free to apply across the b o a r d w h e n
it affects the stops. Yet the rule w o u l d also s e e m to b e postlexical with respect to
the fricatives, b e c a u s e these u n d e r g o palatalization irrespective of any m o r p h o -
logical c o n s i d e r a t i o n s ; that is, for t h e m the derived e n v i r o n m e n t restriction plays
n o role and the rule d o e s apply across the board. If palatalization is postlexical as
well as lexical, however, then w h a t prevents it from a p p l y i n g , incorrectly, to o b -
struent stops in m o r p h e m e - i n t e r n a l e n v i r o n m e n t s too?
Short of resorting to the null h y p o t h e s i s that fricative palatalization is a different
rule from stop palatalization ( A h n , 1988; C h o and Sells, 1 9 9 1 ) , the alternative 1 3

suggested by Iverson ( 1 9 8 7 ; also Iverson and W h e e l e r , 1988) lies in reaffirmation

of K i p a r s k y ' s (1973) revised alternation condition, w h i c h restricts j u s t neutraliz-
ing rules to crucially d e r i v e d f o r m s . K a i s s e ( 1 9 8 6 ) r e a c h e s fundamentally the
s a m e c o n c l u s i o n in h e r analysis of c o n s o n a n t d e v o i c i n g p h e n o m e n a in Turkish,
and H u a l d e ( 1 9 8 9 ) , in consideration of an apparently n o n c y c l i c lexical rule oper-
ating in the p h o n o l o g y of B a s q u e , similarly advocates factoring the derived envi-
r o n m e n t constraint out of the strict cycle condition as an i n d e p e n d e n t r e s t r i c t i o n . 14

K i p a r s k y ' s ( 1 9 7 3 ) p r o p o s a l , to recapitulate here, h o l d s that p h o n o l o g i c a l rules o b -

serve the derived e n v i r o n m e n t constraint only if they effect neutralizations, or
" p r e s e r v e s t r u c t u r e . " C o n s t r u e d as a constraint on rule applications rather than
1 5

on rules p e r se, this principle p e r m i t s the various manifestations of K o r e a n pala-

talization to fall directly u n d e r a single generalization, b e c a u s e palatalization of
/ t i / to [ci] is neutralizing (/ . . . ci . . ./ s e q u e n c e s o c c u r a m o n g the rule's inputs)
and thus restricted to d e r i v e d e n v i r o n m e n t s , b u t palatalization of / s i / to [si] is not
(the exclusive s o u r c e of [s] is palatalization itself). U n d e r the interpretation that
the derived e n v i r o n m e n t constraint is valid j u s t for neutralizing rule applications,
therefore, all the o c c u r r e n c e s of K o r e a n palatalization will b e i m p l e m e n t e d by a
single lexical rule w h o s e a p p a r e n t postlexical effects are d u e to the general invali-
dation of the derived e n v i r o n m e n t constraint w h e n e v e r the rule is not neutralizing.
T h e effect of limiting the derived e n v i r o n m e n t constraint j u s t to neutralizing
applications of rules is to incorporate into that constraint the o t h e r w i s e indepen-
d e n t r e q u i r e m e n t of structure preservation. T h e function of this principle is to
b l o c k lexical rules from creating or referring to any s e g m e n t s not found in the
u n d e r l y i n g inventory, that is, to force lexical rules to b e strictly neutralizing. Since
(Post) Lexical Rule Application 265

lexical rules are defined to b e j u s t those w h i c h m u s t o b e y the derived e n v i r o n m e n t

constraint, t h e c o n v e n t i o n a l t h e o r y ' s further attribution to t h e m of structure p r e s -
ervation e n c o d e s their limitation to neutralization as a distinctly separate restric-
tion. U n d e r the a s s u m p t i o n s here, t h o u g h , w h e r e the derived e n v i r o n m e n t c o n -
straint itself is sensitive to neutralization, lexical rules p r e s e r v e structure b e c a u s e
their application typically is restricted to derived e n v i r o n m e n t s ; a t y p i c a l l y — p r e -
cisely w h e n it w o u l d not be s t r u c t u r e - p r e s e r v i n g — e v e n a lexical rule m a y apply
across the b o a r d . In this way, structure preservation integrates with a major prin-
ciple of the theory, the d e r i v e d e n v i r o n m e n t constraint, a n d so m a k e s p o s s i b l e the
unified description of derivationally sensitive neutralizations with across-the-
b o a r d rule applications that fail to p r e s e r v e structure or a p p e a r to be postlexical.
T h e fact that lexically restricted rules usually d o p r e s e r v e structure derives from
the basic lexical/postlexical d i c h o t o m y inherent in the theory, w h e r e i n j u s t those
rules are defined as lexical w h i c h m u s t o b s e r v e the derived e n v i r o n m e n t c o n -
straint; these, in turn, constitute the class of rules limited b y the p r o p e r t y of struc-
ture preservation. T h e innovation s u g g e s t e d h e r e is to reverse this implication b e -
t w e e n structure preservation and the derived e n v i r o n m e n t constraint, from the
c o n v e n t i o n a l relationship e x p r e s s e d in (13a) to the alternative in (13b).

(13) a. If a rule is lexical (observes the derived e n v i r o n m e n t constraint), then

it is also structure-preserving (neutralizing),
b. If a rule application is neutralizing (structure-preserving), then it also
o b s e r v e s the derived e n v i r o n m e n t constraint.

In d i s t i n g u i s h i n g b e t w e e n a rule per se a n d its individual applications, (13b)

retains structure p r e s e r v a t i o n , or neutralization, as definitive for i m p l e m e n t a t i o n
of the d e r i v e d e n v i r o n m e n t constraint, but it also a c c o m m o d a t e s within the lexical
c o m p o n e n t certain rule applications w h i c h are not structure-preserving a n d h e n c e
not restricted to derived e n v i r o n m e n t s . Structure preservation on this p r o p o s a l
16

thus r e m a i n s a p r o p e r t y of traditionally lexical rules, t h o u g h not necessarily of all

lexical rule applications. B y reversing the association b e t w e e n structure preser-
vation and the d e r i v e d e n v i r o n m e n t constraint, however, (13b) raises the m o r e
fundamental q u e s t i o n of w h e t h e r a strict functional b o u n d a r y really should be
d r a w n after all b e t w e e n lexical and postlexical rules in the p h o n o l o g y , and sug-
gests as an a n s w e r that, as e l s e w h e r e in g r a m m a r , it is general c o n d i t i o n s on rules
w h i c h are d e t e r m i n a t i v e of their application rather than the architecture of m o d u -
lar c o m p o n e n t r y .

4. C O N T E X T - S E N S I T I V I T Y I N U N D E R S P E C I F I C A T I O N

B e s i d e s c o n d i t i o n s , however, the p h o n o l o g i c a l representations t h e m s e l v e s can

be m a d e to constrain the applicability of rules, a possibility w h i c h in principle
266 Gregory K. Iverson

offers an alternative a c c o u n t of h o w derivationally restricted rules m a y have un-

restricted postlexical effects. Radical feature underspecification, w h e n d e t e r m i n e d
a c c o r d i n g to p h o n o l o g i c a l context, plays a critical role in this c o n n e c t i o n , as it
allows m a x i m i z a t i o n of the feature fill-in function of p h o n o l o g i c a l rules by m a k -
ing both values of a feature available in lexical representation. T h o u g h it will b e
s u g g e s t e d b e l o w that this is m i s g u i d e d , a n d that feature underspecification should
instead b e context-free, first the derivational c o n s e q u e n c e s of underspecification
defined on p h o n o l o g i c a l c o n t e x t n e e d to b e explored.
K i p a r s k y (1984) illustrates the context-sensitive i m p l e m e n t a t i o n of feature u n -
derspecification with E n g l i s h quantity alternations, w h e r e v o w e l s before c o n s o -
n a n t clusters in the s a m e m o r p h e m e m a y b e either long (paint) or short (tent), but
only short before Level 1 suffixes w h o s e a m a l g a m a t i o n with stems results in con-
sonant clusters. T h e vowel in mean is therefore predictably short before the ir-
regular inflectional suffix -t (meant), t h o u g h it is long e v e r y w h e r e else (meaning,
means, etc.); this e l s e w h e r e [ + l o n g ] value can b e supplied b y default if only ba-
sically short vowels outside the d o m a i n of the precluster s h o r t e n i n g rule are un-
derlyingly specified, so that lexically the vowel in pen is [ — l o n g ] , w h i l e that in
mean is unspecified. In apparently e x c e p t i o n a l paint, however, w h e r e the vowel
is long despite its precluster a p p e a r a n c e , specification as [ + l o n g ] b l o c k s applica-
tion of the s h o r t e n i n g rule b e c a u s e the relevant e n v i r o n m e n t is not crucially d e -
rived. A n d since the invariably short vowel in a w o r d like tent always appears
before a cluster of c o n s o n a n t s , its length value, like that of mean, m a y be left
lexically unspecified as well, to b e later filled in as [ — long] b y the s a m e rule of
derivation w h i c h p r o v i d e s that feature to the vowel in meant. T h e result is that
s o m e v o w e l s are specified as [ + l o n g ] (paint), s o m e as [ — long] (pen), and s o m e
not at all (tent, mean); yet this m o d e of underspecification is still radical since the
contexts w h e r e plus values o c c u r (before clusters) never overlap with those w h e r e
m i n u s values o c c u r (not before clusters); that is, only o n e feature value is specified
in any given c o n t e x t found a m o n g u n d e r l y i n g r e p r e s e n t a t i o n s .
17

(14) LONG V : [ + l o n g ] before clusters, o t h e r w i s e unspecified

SHORT V : unspecified before clusters, o t h e r w i s e [ — long]

A similar organization of the lexicon is p o s s i b l e with r e g a r d to the variably

restricted rule of palatalization in K o r e a n . T h e required derivational results w o u l d
obtain if the s e g m e n t s for w h i c h palatalization is never neutralizing are lexically
unspecified for the features that define palatal articulation, say, [ + h i g h ] subordi-
nated to a C o r o n a l n o d e (with certain other features to b e filled in b y segmental
r e d u n d a n c y rules), but t h o s e w h i c h contrast in t e r m s of the palatalization features
are underlyingly specified either [ + h i g h ] or [ - h i g h ] , d e p e n d i n g on context. T h e
basic rule of Palatalization can b e given as in (15a), spreading the t o n g u e b o d y
features of a high front vowel or glide (here characterized as c o r o n a l , following
H u m e , 1989, and G o r e c k a , 1989) to a p r e c e d i n g underspecified c o n s o n a n t ; (15b)
(Post) Lexical Rule Application 267

gives the default configuration for c o n s o n a n t s w i t h o u t inherent place of articula-

tion s p e c i f i c a t i o n s . 18

(15) a. [+consonantal] [-consonantal] b. [+consonantal]

Place Place Place

j j
Coronal Coronal

[+high] [-high]

A s s u m i n g with K i p a r s k y that the derived e n v i r o n m e n t constraint p e r the strict

c y c l e condition restricts only s t r u c t u r e - c h a n g i n g applications of lexical rules,
the apparently postlexical effects of lexical palatalization with respect to the frica-
tives /s s 7 are m e r e l y the reflexes of s t r u c t u r e - b u i l d i n g — t h o u g h not structure-
p r e s e r v i n g — r u l e application, b e c a u s e lexically t h e s e s e g m e n t s are a l w a y s u n d e r -
specified for p l a c e of articulation features. Parallel to the t r e a t m e n t of E n g l i s h long
and short v o w e l s before c o n s o n a n t clusters, then, the lexical representation of the
s e g m e n t s for w h i c h palatalization is neutralizing will vary a c c o r d i n g to w h e t h e r
they are followed b y HI. T h e m e d i a l o b s t r u e n t in m o n o m o r p h e m i c [ k a c i ] ' v a l u e ' ,
n

for e x a m p l e , is consistent with rule (15a), w h i c h m e a n s the rule itself c o u l d supply

the a p p r o p r i a t e C o r o n a l configuration in the c o u r s e of ordinary derivation to an
u n d e r l y i n g representation like / k a T i / , w h e r e T represents an o b s t r u e n t stop un-
h

specified for p l a c e of articulation. A l t e r n a t i n g m o r p h e m e s like [ p a t - ] / [ p a c - ]

h h

'field' could b e r e p r e s e n t e d similarly, / p a T - / , so that in s u p p l y i n g the C o r o n a l

n

articulator n o d e rule (15) w o u l d also fill in [ + h i g h ] w h e n [i] follows ( [ p a c i ] , b y n

(15a)), [ - h i g h ] o t h e r w i s e ( [ p a t t t l ] , b y (15b)). B u t affricates in other k i n d s of

h

m o r p h e m e s w o u l d h a v e to b e specified for at least part of this configuration, b e -

c a u s e [ + h i g h ] articulation is not p r e d i c t a b l e in m o s t e n v i r o n m e n t s ( m o r p h e m e
finally, before any vowel other than [i]); the m a n y f o r m s of this sort w o u l d h a v e
to b e r e p r e s e n t e d w i t h specified Ic c c 7 ( [ + h i g h ] ) rather than with underspecified
h

/ T T T 7 . Finally, in o r d e r to prevent application of palatalization in m o n o m o r -

h

p h e m i c structures c o n t a i n i n g any of [t t t ' ] before [i], e.g., [madi] < / m a t i / ' k n o t ' ,
h

c o r o n a l stops in this e n v i r o n m e n t w o u l d n e e d to b e regularly specified as [ — h i g h ] ,

w h e r e a s in all other e n v i r o n m e n t s — a g a i n , m o r p h e m e finally and before v o w e l s
other than [ i ] — [ t t t' ] could still b e unspecified, later to b e given the default value
n

[ - high] by rule (15b). W h a t e m e r g e s from this is the following s c h e m e of lexical

specification for the feature [high] a m o n g c o r o n a l o b s t r u e n t s . 19

(16) C O N T E X T - S E N S I T I V E UNDERSPECIFICATION :
It t t 7
h
[ - h i g h ] before / i / , o t h e r w i s e unspecified (IT T T 7 )
n

Ic c c 7 :
h
unspecified (IT T T 7 ) before / i / , o t h e r w i s e [ + h i g h ]
h

/ss7: unspecified ( / S S 7 )
268 Gregory K. Iverson

T h e reason that palatalization w o u l d have n o effect on m o n o m o r p h e m i c [. . .

ti . . .] s e q u e n c e s u n d e r this system of representation is that, as a lexical rule,
its feature-changing applications w o u l d be confined to derived representations.
Structure-building applications of the rule predictably are not restricted in this
way, t h o u g h with respect to the fricatives the rule d o e s create s e g m e n t s not found
in the p h o n e m e inventory a n d h e n c e stands in violation of structure preservation.
T h i s aside, m o r p h e m e - i n t e r n a l b l o c k i n g of j u s t the neutralizing applications of
Korean palatalization in t e r m s of context-sensitive feature underspecification is a
direct c o n s e q u e n c e of [ — high] lexical m a r k i n g for any [t t t ' ] w h i c h h a p p e n to
h

be followed by [i].
It is clear, however, that the only reason [t t t ' ] in these e n v i r o n m e n t s are
h

specified for [high] in the first place is to prevent t h e m from u n d e r g o i n g palatal-

ization; that is, [t t t ' ] before [i] in the s a m e m o r p h e m e are m a r k e d [ - h i g h ]
h

b e c a u s e they fail to u n d e r g o palatalization, and they fail to u n d e r g o palatalization

b e c a u s e they are m a r k e d [ — h i g h ] . At the s a m e time, any gains in t e r m s of sim-
2 0

plicity relative to the lexical representation of [c c c ' ] before [i] are only illusory,
h

as these are offset by the c o m p l e m e n t a r y listing of [ — high] for the plain stops in
the palatalizing e n v i r o n m e n t . If [t t t ' ] w e r e unspecified irrespective of context,
h

on the other hand, as will b e p r o p o s e d presently, and [c c c ' ] w e r e as a c o n s e -

h

q u e n c e specified [ + h i g h ] in all of their o c c u r r e n c e s , then the derived e n v i r o n m e n t

restriction on rule (15a) could not b e i m p o s e d by the strict cycle condition. If
radically underspecified for the feature [high] e v e r y w h e r e , in o t h e r w o r d s , the
coronal stops w o u l d incorrectly palatalize in m o r p h e m e - i n t e r n a l e n v i r o n m e n t s
w h e n it is a s s u m e d that j u s t feature-changing rule applications o b s e r v e the derived
e n v i r o n m e n t constraint.
Of c o u r s e , palatalization of stops in Korean always w o u l d be feature-changing
u n d e r the alternative of contrastive feature specification, w h i c h , t h o u g h not a d v o -
cated here, is exemplified for c o m p a r i s o n in (17). B u t the derived e n v i r o n m e n t
constraint will properly affect even context-free radically underspecified represen-
tations, as in (18), if, following (13b), it is structure-preserving rather than simply
structure-changing rule applications that are derivationally restricted.

(17) C O N T R A S T I V E SPECIFICATION:
Ii t t 7
h
[-high]
/c c c ' /
h
[+high]
/ss7 unspecified

(18) C O N T E X T - F R E E UNDERSPECIFICATION :
/tt t7
h
unspecified
/cc c7 h
[+high]
Is s 7 unspecified

W i t h respect to i m p l e m e n t a t i o n of (13b), context-free radical underspecifica-

tion p r o v i d e s sufficient information to establish that palatalization of stops w o u l d
(Post) Lexical Rule Application 269

b e neutralizing, or structure-preserving, but that of fricatives w o u l d not, b e c a u s e

underspecification theory stipulates that at m o s t o n e feature value m a y b e present
in u n d e r l y i n g representation (in a given e n v i r o n m e n t , a n y w a y ) , the other b e i n g
regularly supplied b y rule. A n y alternation b e t w e e n the sets in (18) of under-
specified It t t'I (= IT T T 7 ) and specified Ic c c 7 is thus neutralizing b e c a u s e
h h h

the value [ + h i g h ] is p r e s e n t u n d e r l y i n g l y a m o n g n o n c o n t i n u a n t o b s t r u e n t s , the

class of stops and affricates; but a m o n g fricatives, neither value of [high] occurs
in u n d e r l y i n g representations, so for t h e m rule (15a) could h a v e only a n o n n e u -
tralizing e f f e c t . For any binary feature subject to context-free underspecifica-
21

tion, in fact, the p a r a d i g m a t i c e n v i r o n m e n t in w h i c h overt values m a y b e found in

u n d e r l y i n g representation constitutes a potential neutralization site; other e n v i r o n -
m e n t s d o not. T h i s p o i n t is i m p o r t a n t in m e e t i n g K i p a r s k y ' s ( 1 9 8 2 : 4 0 ) objection
that the p r o p e r t y of neutralization c a n n o t b e d e t e r m i n e d from an inspection of the
g r a m m a r alone, b u t rather requires c h e c k i n g all of the derivations. O n the c o n -
trary, j u s t as the p r o p e r t y of structure p r e s e r v a t i o n is d e t e r m i n a b l e from c o m p a r i -
son with the inventory of u n d e r l y i n g r e p r e s e n t a t i o n s , so too is the p r o p e r t y of
neutralization with respect to u n d e r l y i n g contrasts.
T h e context-free a p p r o a c h to radical underspecification in (18) is confirmed
over the context-sensitive characterization in (16) by t w o further aspects of K o -
rean p h o n o l o g y , namely, syllable-final o b s t r u e n t neutralization a n d aspiration c o -
a l e s c e n c e , b o t h of w h i c h are optimally d e s c r i b e d only if Itl is the least specified
c o n s o n a n t in this l a n g u a g e , as it h a s also b e e n c h a r a c t e r i z e d for E n g l i s h (e.g.,
G o l d s m i t h , 1990; Y i p , 1991), M a o r i (de C h e n e , 1988; S a n d e r s , 1990), and several
other l a n g u a g e s . T h u s the obstruents in (9) all r e d u c e to o n e of three u n r e l e a s e d
stops in syllable-final position: the labials to [ p ] , the velars to [ k ] , and all the
n n

others to [ f ] . Iverson a n d K i m ( 1 9 8 7 ) c h a r a c t e r i z e this p r o c e s s in g e o m e t r i c t e r m s

1

as the d e l i n k i n g of terminal features from o b s t r u e n t s in syllable-final position,

r e m o v i n g all specified laryngeal features, the m a n n e r feature [continuant], a n d any
s e c o n d a r y p l a c e of articulation features s u b o r d i n a t e to the p r i m a r y p l a c e n o d e s .
R e d u n d a n c y rules c o m e into play to identify the u n m a r k e d laryngeal articulation
as neither aspirated n o r tense, a n d the u n m a r k e d m a n n e r of articulation as n o n c o n -
tinuant and nonaffricate, w h i c h a c c o u n t s for w h y the ultimate output of the rule is
a plain lax stop out of the set of [ p f k ] , with p r i m a r y p l a c e of articulation
n n

r e m a i n i n g intact. Surprisingly, however, syllable-final Ihl also u n d e r g o e s this p r o -

cess, as s h o w n particularly in careful p r o n u n c i a t i o n s in w h i c h the Ihl also aspi-
rates a following lax stop, as in / c o h + k o / —» [ c o t k o ] ' g o o d a n d ' , / t a h + c i / —>
n h

[ t a f c i ] ' t o u c h ( s u s p . ) ' . T h e fact that Ihl, a s e g m e n t with n o inherent p l a c e fea-

h

tures of its o w n , regularly e m e r g e s as [ t ] t o o u n d e r syllable-final c o n d i t i o n s lends

1

c o n s i d e r a b l e s u p p o r t to the lexical representation of Itl with n o specified features

at all, b e c a u s e the s a m e default r e d u n d a n c y rules that fill in the p h o n e t i c values
for Itl e v e r y w h e r e a n d for neutralized coronal o b s t r u e n t s syllable finally will also
p r o v i d e those values to Ihl in the syllable-final e n v i r o n m e n t , p r o d u c i n g [ f ] . 1 2 2

T h e identification of Itl as c o m p l e t e l y unspecified in K o r e a n p h o n o l o g y is at

270 Gregory K. Iverson

o b v i o u s o d d s with the diacritic b l o c k i n g of palatalization in n o n d e r i v e d / . . . t i . . . /

s e q u e n c e s t h r o u g h lexical specification of j u s t these / t / s as [ — high] on the m o d e l
of (16). To assert that HI is in general the unspecified K o r e a n obstruent, but that
in the m a n y m o r p h e m e s in w h i c h l\l i m m e d i a t e l y follows it is not, is to cloud the
generalization that all t a u t o m o r p h e m i c /. . . ti . . ./ structures e s c a p e palataliza-
tion for the very s a m e r e a s o n . T h e b l o c k i n g of m o r p h e m e - i n t e r n a l palatalization
t h r o u g h lexical specification of [ — high] w o u l d instead characterize every under-
lying /. . . ti . . ./ s e q u e n c e in the l a n g u a g e as idiosyncratic, o b s c u r i n g the gener-
alization that it is all / t / s — r a t h e r than j u s t certain m o r p h e m e s w h i c h contain
t h e m — w h i c h fail to u n d e r g o palatalization w h e n in a n o n d e r i v e d e n v i r o n m e n t . 23

T h e restriction of j u s t structure-preserving rule applications to derived forms,

however, allows for consistent, context-free radical underspecification of un-
m a r k e d K o r e a n Itl as in (18) w h i l e still a c c o m m o d a t i n g a p p r o p r i a t e applicational
distinctions in the rule of palatalization, a n d r e s e r v i n g for g e n u i n e l y exceptional,
m a r k e d cases the b l o c k i n g p o w e r of m o r e c o m p l e t e lexical representation 2 4

5. CONCLUSION

To s u m m a r i z e , it w o u l d s e e m that t h e characterization of certain rules as both

lexical and postlexical is not the o p t i m a l w a y to c a p t u r e p h o n o l o g i c a l generaliza-
tions w h i c h s o m e t i m e s o b e y the derived e n v i r o n m e n t constraint, s o m e t i m e s not.
Rather, the relevant restriction a p p e a r s to b e that all a n d only structure-preserving,
or neutralizing, rule applications are c o n s t r a i n e d to derived e n v i r o n m e n t s , with
the c o n s e q u e n c e that structure-building applications of lexical rules n e e d not
(though m a y ) b e structure-preserving. T h i s r e n d e r i n g of the correlation, in turn,
calls into question the utility of d r a w i n g a basic functional distinction b e t w e e n
lexical and postlexical rule m o d u l e s in the p h o n o l o g y , suggesting instead that the
properties of structure preservation and derived e n v i r o n m e n t restriction affiliate
in p r e d i c t a b l e w a y s with t h e rules t h e m s e l v e s rather than with the c o m p o n e n t s in
w h i c h they apply. W i t h respect to radical feature underspecification, moreover,
the limitation of structure-preserving rule applications to derived e n v i r o n m e n t s
implies that the allocation of features to p h o n e m e s is consistently context-free,
thus p r o v i d i n g for a m o r e direct association b e t w e e n the relative c o m p l e x i t y of
representation and any idiosyncrasies of derivation.

NOTES

That the nasal assimilates only as far as laminopalatal place when before (palatal) [A]
1

is puzzling and presumably requires a special restriction. Catalan does have phonemic
(Post) Lexical Rule Application 271

/ n / , but simple avoidance of merger with this segment cannot be the whole story since the
same nasal does merge with phonemic Iml in so[m] pocs 'they are few'.
2
Except that labial nasals do assimilate before labiodentals; see below.
3
With no Root node, the melodically empty C will be stray-erased at the end of the
derivation.
4
The syntactic conditions on syllabification in Catalan would seem to be fully express-
able in terms of parametric variation per the theory of syntax-phonology interaction de-
veloped in Kaisse (1985:186ff.).
5
The postlexical derivation of velar nasals makes unnecessary the special lexical treat-
ment of them Kiparsky ( 1 9 8 5 : 1 0 1 - 1 0 3 ) proposes, because the velar nasal under present
assumptions does not exist except at the postlexical level. Variation in the retention of
postnasal [k] (as with [p] and [t]) is then due generally to the syntactic conditions on syl-
labification, though under the same stylistic conditions the velar stop is somewhat more
likely to be retained than the labial or dental (Gonzales, 1989).
6
The blocking of rule (5b) in this case will follow from the elsewhere condition. This
predictable precedence of the more specific rule (5a) over (5b) assures that the labiodental
articulation specified for the fricative IfI percolates onto the nasal Iml in an Imfl cluster,
rather than the (default) bilabial articulation of Iml onto IfI.
7
Though not with a palatal (cf. note 1).
8
Korean palatalization has been described from various perspectives by Kim-Renaud
(1974), S.-G. Kim (1976), C.-W. Kim and Ahn (1983), Ahn (1985, 1988), Sohn (1987),
Iverson (1987), Iverson and Wheeler (1988), and Cho and Sells (1991).
9
Presumably It' I would undergo this process too, except that the only member of the
tense stop series which occurs finally in stems is the velar IV!I (Chung, 1980).
10
Palatalization occasionally results in homophony, e.g., [maji] derives from either
/ m a t + i / 'the eldest' or /maci/ 'hempen paper', [kac i] from either / k a t + i / 'together' or
h h

/ k a c i / 'value'.
h

1 1
Besides Is s 7 , the alveolar sonorants also evince allophonic palatalization between
([mujii] < /mun-f-i/ 'door (subj.)') as well as within morphemes ([huAAun] < /hullyun/
'magnificent'). Cho and Sells (1991) suggest that morpheme-internal instances of It t t 7 h

"palatalize," too—but without affricating—so that the alveolar stops in (1 la) would also
reflect a degree of superficial coarticulation with HI. But the effect here seems to be no
more than is universally instantiated in such contexts, and in any case is not nearly as
categorically salient as among the coronal sonorants and fricatives.
1 2
As pointed out to me by Sang-Cheol Ahn, there is some variation among speakers as
to the extent of palatalization with respect to tense / s 7 , i.e., /s'i/ seems to vary individually
between essentially unpalatalized ([s'i]) and sharply palatalized ([s'i]) articulations. Since
the other potentially palatalizable tense obstruent, / t 7 , happens not to occur in relevant
derived environments (cf. note 9), palatalization for some speakers appears to exclude the
tense obstruents in general.
1 3
A variant of the Cho and Sells (1991) two-rule approach is outlined by Kiparsky (this
volume) in which palatalization per se would always be allophonic, applying first in the
derived environments of the lexical phonology, then without morphological restriction in
the postlexical phonology. But before exiting the lexicon, the intermediate postalveolar
stops (It l \'I) produced by the rule in derived environments would merge with the pho-
h

nemic affricates (Ic c c 7 ) by a separate rule applying at the word level. The violations of
h
272 Gregory K. Iverson

structure preservation entailed by this version of palatalization's application in the lexicon

could then be sanctioned as per Macfarland and Pierrehumbert (1991), who invoke the
autosegmental linking condition (Hayes, 1986) in order to grant lexical status to feature-
spreading, allophonic assimilation rules. Irrespective of whether relaxation of structure
preservation to this extent is ultimately well founded, however, Korean palatalization thus
described still remains split between two rules, the neutralizing one of which applies
context-free. Further, the attribution of postalveolar ([+anterior]) or palatalized ([+high])
articulation to nonaffricated coronal stops before [i] in forms like those in (11a) would
represent a considerably greater degree of coarticulation than the phonetics warrants (cf.
also note 11).
1 4
The rule raises word-final /a/ to [e] after high vowels, as in /mutil+a/ —» [mutiAe] 'the
boy'. It must be noncyclic since the only constituent boundary before which it applies is at
the level of the word; hence /a/ does not raise in / m u t i l + a + k / —> [mutiAak] 'the boy (erg.)';
yet the rule must be restricted to derived environments, since /a/ does not raise morpheme-
internally, either: /ikas+i/ [ikasi] 'to learn', /muga/ —> [nurva] 'limit'. Cf. also Booij
and Rubach (1987) for other instances of noncyclic lexical rules.
1 5
As defined by Kiparsky (1973), a rule of the form A —> B / C D is neutralizing
just in case there exist instances of CBD in the input to the rule.
1 6
The property of structure preservation among lexical rules has been used to explain
the neutrality of specific vowels in harmony systems, e.g., lil and lei in Finnish (Kiparsky,
1985). Precisely those vowels are neutral for which the effect of harmony would be allo-
phonic (the simple backing of HI or Id in Finnish would create novel back unrounded
vowels, but the backing of lyl or 161 results in merger with underlying lul and /o/). If all
applications of vowel harmony must be structure-preserving, which characterization of the
rule as strictly lexical under the conventional theory would entail, then the neutrality of lil
and lei would seem to fall out, although it begs the question of why neutral vowels then do
not simply undergo rounding along with backing so as to merge with extant vowels in the
system (the harmonic rounding of lil in Yawelmani Yokuts, for example, results in merger
with underlying lul rather than in either nonapplication of the rule or in creation of the
novel segment [y]). Under (13b), harmonic neutrality derives as per Steriade (1987), from
the representational distinction between contrastive and redundant feature values.
1 7
Of course, vowel length today would be represented not as a binary feature on the
segmental or melodic tier, but as a timing unit on the skeletal or moraic tier (Clements and
Keyser, 1983; Hayes, 1989), so that the structure of long vowels is VV, that of short vowels
just V. Shortening then consists in the disassociation of the vowel's melodic features from
one of its two timing elements, with the default interpretation that short segments in general
are unmarked (if not strictly underspecified) relative to long or geminate ones. (Cf. also
Myers, 1987; Yip, 1987.) Kiparsky's analysis varying the underlying specifications of a
segmental feature [long] in English is nonetheless illustrative of the context-sensitive im-
plementation of radical underspecification theory.
1 8
Nothing crucial hinges here on spreading the vowel's presumed Coronal articulator to
the underspecified consonant's Place node (as opposed to just spreading [+high] and al-
ways providing Coronal by default). The important point is that the output of Palatalization
for stops is the same representation as underlies that of the affricates.
1 9
When listed with either of its binary values, the feature [high] is ultimately dependent
on there being an articulator node under Place, which, when not otherwise specified, is
Coronal.
T h i s circularity is endemic as well to the class of traditional sequence redundancy
2 0
(Post) Lexical Rule Application 273

rules, which also predict features based on properties of segments found elsewhere in the
morpheme (e.g., all the features of I si are redundant in initial triconsonantal clusters of
English even though Is/ would otherwise be specified [+continuant]). Here, however, there
is no interaction with derivation, since classical sequence redundancy rules apply strictly
within morphemes; and neither does the representational savings they provide have to be
undone by marking of the feature for blocking effect purposes in complementary environ-
ments (Itl need not be specified [-continuant] in any environment even if Isl in clusters is
also unspecified for that feature because there are no lexical exceptions to the cluster gen-
eralization). Cf. Ao (1991) for a possible case in which a sequence redundancy rule none-
theless does appear to apply after a rule of phonological derivation, though the effect there
is still inert, i.e., neither feeding nor bleeding.
2 1
In terms of the definition in note 15, supplying the feature [+high] to underspecified
ITI is neutralizing because [+high] coronals (Ic c c 7 , also underspecified for [continuant])
h

exist in the input to the rule. Supplying [+high] to Is s'i (specified as [+continuant]),
however, is not neutralizing because no [+high, +continuant] consonants exist in the input
to the rule. A similar characterization of feature redundancy is outlined in Steriade (1987).
T h i s account is based on Iverson (1989), in which laryngeal segments are configured
2 2

with specifications for laryngeal features and for [continuant], but without distinctive place
of articulation features.
2 3
The representation of Itl as unspecified in all environments has the consequence that
Korean morphemes with /. . . t i . . . / are no more complex (in fact, one feature less so) than
those with /. . . c i . . ./. Similarly, invariant radical underspecification of vowel quantity in
English does not distinguish the long vowels in paint and pain (both with /VV/), whereas
the context-sensitive approach outlined in (14) would specify the vowel as long in paint
but not in pain. In the absence of evidence showing Korean /. . . t i . . ./ to be more marked
than /. . . ci . . ./, or of any suggesting the vowel in paint is in some way more remarkable
than that in pain, this consequence seems quite correct.
2 4
This property also holds of underspecified segments for which rule application
is blocked by virtue of exceptional marking elsewhere in the word. For example, the
loanword in Hungarian biiro [byro:] 'bureau' contravenes the regular front/back vowel
harmony pattern of the language (Ringen, 1988); its exceptionality can be encoded by
assigning the specification [ — back] just to the word's leftmost vowel in underlying repre-
sentation. In regular cases, [ — back] is a property of either all the vowels in the morpheme
(/OrOm, [ — back]/ - » [orom] orom 'joy') or none of them (/vArOs/ —> [varos] varos 'city',
with [+back] throughout by default). In irregular biiro, however, the lexical attachment of
[ — back] just to the first vowel (or to the second in similarly disharmonic sofdr [sof0: r]
'chauffeur') causes the other vowel's backness value to be supplied by default rather than
harmony since, in consequence of the morpheme's exceptional lexical marking, the spread-
ing within it of [ — back] would be neutralizing and so inapplicable in this nonderived
context.

REFERENCES

Ahn, S.-C. (1985). The Interplay of Phonology and Morphology in Korean. Doctoral dis-
sertation, University of Illinois, Urbana.
Ahn, S.-C. (1988). Lexicality vs. postlexicality in Korean palatalization. In Linguistics in
274 Gregory K. Iverson

the Morning Calm 2 (The Linguistic Society of Korea, ed.), pp. 2 4 9 - 2 6 3 . Hanshin,
Seoul.
Anderson, S. R. (1974). The Organization of Phonology. Academic Press, New York.
Ao, B. (1991). Kikongo nasal harmony and context-sensitive underspecification. Linguistic
Inquiry 22, 193-196.
Booij, G., and Rubach, J. (1987). Postcyclic versus postlexical rules in lexical phonology.
Linguistic Inquiry 18, 1 - 4 4 .
Cho, Y.-M., and Sells, P. (1991). A Lexical Account of Phrasal Suffixes in Korean. Unpub-
lished manuscript, Stanford University, Stanford, Calif.
Chung, K. (1980). Neutralization in Korean: A Functional View. Doctoral dissertation,
University of Texas, Austin.
Clements, G. N., and Keyser, S. J. (1983). CV Phonology: A Generative Theory of the
Syllable. MIT Press, Cambridge, Mass.
de Chene, B. (1988). Japanese x-epenthesis and Inflectional Boundary Buffer-Segments.
Paper presented at the winter meeting of the Linguistic Society of America, New
Orleans.
Goldsmith, J. (1990). Autosegmental and Metrical Phonology. Blackwell, Oxford.
Gonzalez, M. (1989). Catalan Nasal Assimilation: Lexical or Postlexical? Unpublished
manuscript, University of Massachusetts, Amherst.
Gorecka, A. (1989). Are Front Vowels Coronal? Paper presented at the winter meeting of
the Linguistic Society of America, Washington, D.C.
Hayes, B. (1986). Inalterability in CV phonology. Language 62, 3 2 1 - 3 5 1 .
Hayes, B. (1989). Compensatory lengthening in moraic phonology. Linguistic Inquiry 20,
253-306.
Hualde, J. (1989). The strict cycle condition and noncyclic rules. Linguistic Inquiry 20,
675-680.
Hume, E. (1989). Blocking Effects in Korean Umlaut: Evidence for the Coronality of Front
Vowels. Paper presented at the winter meeting of the Linguistic Society of America,
Washington, D.C.
Ito, J. (1986). Syllable Theory in Prosodic Phonology. Doctoral dissertation, University of
Massachusetts, Amherst,
ltd, J. (1989). A prosodic theory of epenthesis. Natural Language & Linguistic Theory 7,
217-259.
Iverson, G. (1987). The Revised Alternation Condition in lexical phonology. Nordic Jour-
nal of Linguistics 10, 151 - 1 6 4 .
Iverson, G. (1989). On the category Supralaryngeal. Phonology 6, 2 8 5 - 3 0 3 .
Iverson, G , and Kim, K.-H. (1987). Underspecification and hierarchical feature represen-
tation in Korean consonantal phonology. In Papers From the Parasession on Autoseg-
mental and Metrical Phonology, pp. 1 8 2 - 1 9 8 . Chicago Linguistic Society, Chicago.
Iverson, G , and Wheeler, D. (1988). Blocking and the elsewhere condition. In Theoretical
Morphology: Approaches in Modern Linguistics (M. Hammond and M. Noonan,
eds.), pp. 3 2 5 - 3 3 8 . Academic Press, San Diego.
Kaisse, E. (1985). Connected Speech: The Interaction of Syntax and Phonology. Academic
Press, Orlando.
Kaisse, E. (1986). Locating Turkish devoicing. Proceedings of the West Coast Conference
on Formal Linguistics 5, 119-128.
(Post) Lexical Rule Application 275

Kim, C.-W, and S.-C. Ahn. (1983). Palatalization in Korean Revisited. Paper presented at
the winter meeting of the Linguistic Society of America, Minneapolis.
Kim, S.-G. (1976). Palatalization in Korean. Doctoral dissertation, University of Texas,
Austin.
Kim-Renaud, Y.-K. (1974). Korean Consonantal Phonology. Doctoral dissertation, Uni-
versity of Hawaii, Manoa.
Kiparsky, P. (1973). Abstractness, opacity, and global rules. In Three Dimensions of Lin-
guistic Theory (O. Fujimura, ed.), pp. 5 7 - 8 6 . TEC, Tokyo.
Kiparsky, P. (1982). Lexical morphology and phonology. In Linguistics in the Morning
Calm (I.-S. Yang, ed.), pp. 3 - 9 1 . Hanshin, Seoul.
Kiparsky, P. (1984). On the lexical phonology of Icelandic. Nordic Prosody III (C.-C. Elert,
ed.), pp. 1 3 5 - 1 6 4 . Almqvist and Wiksell, Stockholm.
Kiparsky, P. (1985). Some consequences of lexical phonology. Phonology Yearbook 2,
85-138.
Koutsoudas, A., Sanders, G , and Noll, C. (1974). The application of phonological rules.
Language 50, 1-28.
Macfarland, T., and Pierrehumbert, J. (1991). On ich-Laut, ach-Laut and structure preser-
vation. Phonology 8, 171-180.
Mascaro, J. (1976). Catalan Phonology and the Phonological Cycle. Doctoral dissertation,
Massachusetts Institute of Technology, Cambridge.
McCarthy, J. (1986). OCP effects: gemination and antigemination. Linguistic Inquiry 17,
207-264.
Myers, S. (1987). Vowel shortening in English. Natural Language & Linguistic Theory 5,
485-518.
Paradis, C , and Prunet, J.-F. (1989). On coronal transparency. Phonology 6, 3 1 7 - 3 4 8 .
Plapp, R. (1990). The Geometry of Russian Voice Assimilation. Paper presented at the Min-
nesota Conference on Language and Linguistics, Minneapolis.
Ringen, C. (1988). Transparency in Hungarian vowel harmony. Phonology 5, 3 2 7 - 3 4 2 .
Sagey, E. (1986). The Representation of Features and Relations in Non-linear Phonology.
Doctoral dissertation, Massachusetts Institute of Technology, Cambridge.
Sanders, G. (1990). On the analysis and implications of Maori verb alternations. Lingua
80, 149-196.
Sapir, J. D. (1965). A Grammar of Diola-Fogny (West African Language Monographs 3).
Cambridge University Press, London.
Sohn, H.-S. (1987). Underspecification in Korean Phonology. Doctoral dissertation, Uni-
versity of Illinois, Urbana.
Steriade, D. (1987). Redundant values. In Papers from the Parasession on Autosegmental
and Metrical Phonology, pp. 3 3 9 - 3 6 2 . Chicago Linguistics Society, Chicago.
Yip, M. (1987). English vowel epenthesis. Natural Language & Linguistic Theory 5,
463-484.
Yip, M. (1991). Coronals, consonant clusters and the coda condition. In The Special Status
of Coronals: Internal and External Evidence (C. Paradis and J.-F. Prunet, eds.),
pp. 6 1 - 7 8 . Academic Press, San Diego.
BLOCKING IN NONDERIVED ENVIRONMENTS

PAUL KIPARSKY
Department of Linguistics
Stanford University
Stanford, California 94305

1. T H E P R O B L E M

S o m e p h o n o l o g i c a l rules apply freely across m o r p h e m e b o u n d a r i e s , a n d m o r -

p h e m e internally w h e r e fed b y s o m e earlier p h o n o l o g i c a l rule, b u t a r e b l o c k e d
e l s e w h e r e , in w h a t a r e referred t o as " n o n d e r i v e d e n v i r o n m e n t s . " I will call this
s y n d r o m e N O N D E R I V E D E N V I R O N M E N T BLOCKING ( N D E B ) . T h e t h e o r e t i c a l b a s i s
for N D E B w a s sought first in constraints o n u n d e r l y i n g representations (the alter-
nation c o n d i t i o n ) a n d later in constraints o n rule application (e.g., t h e revised al-
ternation condition, t h e strict c y c l e condition, t h e e l s e w h e r e c o n d i t i o n ) . In this
article, I p r o p o s e to derive it from m o r e general a s s u m p t i o n s , n a m e l y (1) that
lexical r e p r e s e n t a t i o n s c a n b e underspecified, (2) that p h o n o l o g i c a l rules c a n a p -
ply in structure-building ( " f e a t u r e - f i l l i n g " ) m o d e , a n d (3) that learners construct
the simplest g r a m m a r . It will also b e s h o w n that these a s s u m p t i o n s lead to p r e d i c -
tions a b o u t t h e n a t u r e a n d s c o p e of derived e n v i r o n m e n t effects w h i c h a r e e m p i r i -
cally m o r e accurate than those w h i c h flow from a n y of t h e a b o v e m e n t i o n e d
approaches.
W h a t rules a r e subject to N D E B ? T h e best a p p r o x i m a t i o n so far s e e m s to b e
that it is t h e class of obligatory neutralization rules. T h e revised alternation c o n -
dition ( R A C , Kiparsky, 1 9 7 3 ; Iverson a n d W h e e l e r , 1988) e x p r e s s e s this directly
by simply prohibiting obligatory neutralization rules from a p p l y i n g in n o n d e r i v e d
e n v i r o n m e n t s . B u t t h e R A C is really n o m o r e than a descriptive generalization
dressed u p as a principle a n d is unstatable as a formal condition o n p h o n o l o g i c a l

277
Phonetics and Phonology, Volume 4 Copyright © 1993 by Academic Press, Inc.
Studies in Lexical Phonology All rights of reproduction in any form reserved.
278 Paul Kiparsky

rules. Notoriously, the class of neutralization rules c a n n o t b e formally specified,

for w h e t h e r a rule is neutralizing d e p e n d s o n the derivations of the g r a m m a r . T h e
restriction to obligatory rules is likewise suspect, b e c a u s e e v e r y w h e r e else the
principles g o v e r n i n g the application of p h o n o l o g i c a l rules apply equally to o p -
tional and obligatory rules. ( T h e r e is n o reason to believe that, w h e n optional rules
b e c o m e obligatory in the c o u r s e of l a n g u a g e c h a n g e , they suddenly b e c o m e s u b -
j e c t to different p r i n c i p l e s of rule application.) A n d the c o n c e p t of " d e r i v e d envi-
r o n m e n t " itself plays n o role e l s e w h e r e in the theory of g r a m m a r and indeed re-
quires a peculiar disjunctive definition. M o r e o v e r , the R A C is p r o b a b l y i n a d e q u a t e
on factual g r o u n d s as well, since it effectively b a n s ALL absolute neutralization,
enforcing a d e g r e e of c o n c r e t e n e s s in u n d e r l y i n g representations w h i c h t h r o w s out
m a n y w e l l - m o t i v a t e d a n a l y s e s (Dresher, 1981). R e m e d y i n g this latter flaw b y lim-
iting the R A C to n o n a u t o m a t i c rules (Kiparsky, 1973) w o u l d m a k e it theoretically
even m o r e objectionable, for d e t e r m i n i n g w h e t h e r a rule is a u t o m a t i c or not again
involves l o o k i n g at all the d e r i v a t i o n s .
T h e search for a m o r e a d e q u a t e a c c o u n t of N D E B effects led to t w o m a i n alter-
natives. T h e m o r e w i d e l y a c c e p t e d of these i n v o k e s s o m e version of the strict
cycle condition ( S C C ) first p r o p o s e d for p h o n o l o g y by K e a n ( 1 9 7 4 ) . T h e restric-
tion to obligatory neutralization rules w a s specified stipulatively in the version of
M a s c a r o (1976) and r e c e i v e d little attention in later w o r k (Halle, 1978; R u b a c h ,
1984). In K i p a r s k y ( 1 9 8 2 ) it w a s further p r o p o s e d to r e d u c e the S C C itself to the
e l s e w h e r e c o n d i t i o n . T h i s a l l o w e d the r e m a i n i n g p r o b l e m a t i c predicate " d e -
1

r i v e d " to b e r e p l a c e d b y " d i s t i n c t from a lexical i t e m , " a further i m p r o v e m e n t

b e c a u s e distinctness is a basic p r e d i c a t e of the theory of g r a m m a r , and b e c a u s e it
unifies the p h o n o l o g i c a l a n d m o r p h o l o g i c a l k i n d of " d e r i v e d n e s s . " T h e S C C ,
then, m a d e it possible to avoid all references to obligatoriness, to neutralization,
to " d e r i v e d e n v i r o n m e n t s , " and to a rule's " a u t o m a t i c " status, and in addition h a d
the virtue of p e r m i t t i n g absolute neutralization, as long as this is effected by n o n -
cyclic (or postlexical) rules. T h e s e gains w e r e , however, offset by a major degra-
dation in empirical adequacy. T h e S C C defines the nature and s c o p e of N D E B in
ways w h i c h are not at all b o r n e out b y the e v i d e n c e . In particular, it h a s b e c o m e
rather clear that N D E B is correlated with a rule's cyclicity a n d / o r lexicality only
a p p r o x i m a t e l y if at all. T h e R A C h a d struck m u c h nearer the m a r k in c o n n e c t i n g
N D E B to a rule's o b l i g a t o r i n e s s and to its neutralizing status.
T h e s e c o n d t y p e of alternative a c c o u n t for the N D E B takes the derivational
character of such n o t i o n s as " o b l i g a t o r y neutralization r u l e " and " d e r i v e d e n v i -
r o n m e n t " as indications that the p h e n o m e n o n should not be attributed to any con-
straints on g r a m m a r s at all, but rather to the l a n g u a g e acquisition p r o c e s s . In this
vein, A n d e r s o n (1981) suggests that N D E B arises from the learner's lack of evi-
d e n c e for a rule's applicability in n o n d e r i v e d e n v i r o n m e n t s , and that nothing m o r e
than this need b e s a i d . I c o n c u r with A n d e r s o n on the first point but not on the
2

second. N D E B arises in situations w h e r e only derived e n v i r o n m e n t s p r o v i d e posi-

tive positive e v i d e n c e for the application of a rule, but it represents a generaliza-
Blocking in Nonderived Environments 279

tion w h i c h g r a m m a r s m u s t b e c a p a b l e of e x p r e s s i n g in a p r i n c i p l e d way. T h e al-

ternatives are all u n a c c e p t a b l e . W r i t i n g the " d e r i v e d e n v i r o n m e n t s " expressly
into the structural d e s c r i p t i o n s of rules as s o m e c o m b i n a t i o n of lexical and m o r -
p h o l o g i c a l c o n d i t i o n s , aside from b e i n g ad h o c , w o u l d in s o m e c a s e s require
global c o n d i t i o n s o t h e r w i s e u n k n o w n in p h o n o l o g i c a l rules (Kiparsky, 1973).
Listing the n o n d e r i v e d cases, or m a r k i n g t h e m as lexical e x c e p t i o n s , w o u l d m i s s
the generalization that precisely n o n d e r i v e d e n v i r o n m e n t s are e x e m p t — p a r t i c u -
larly unfortunate in the case of rules w h i c h o t h e r w i s e h a v e n o lexical e x c e p t i o n s .
W o r s e , it w o u l d b e entirely out of the q u e s t i o n for m o r p h e m e s with t w o potential
alternation sites, of w h i c h o n l y the o n e in t h e d e r i v e d e n v i r o n m e n t u n d e r g o e s the
rule. F o r e x a m p l e , F i n n i s h C o n s o n a n t G r a d a t i o n , w h i c h simplifies a g e m i n a t e in
the o n s e t of a c l o s e d syllable, affects only t h e s e c o n d g e m i n a t e in s u c h w o r d s as
/ a t t e n t a a t t i + n / —» attentaati+n ' a s s a s s i n a t i o n ' . A s s o c i a t i n g an e x c e p t i o n feature
3

w i t h the first tt b u t n o t w i t h the s e c o n d w o u l d (beside m i s s i n g the generalization

that only g e m i n a t e s in n o n d e r i v e d e n v i r o n m e n t s are " e x c e p t i o n s " to the rule)
c o n t r a v e n e the robust generalization that e x c e p t i o n features are associated with
lexical i t e m s or at m o s t with m o r p h e m e s , but at any rate not with individual seg-
m e n t s or syllables.
In this article I a r g u e that in o r d e r for learners to arrive at the correct projections
from the e v i d e n c e accessible to t h e m , the t h e o r y of g r a m m a r m u s t m a k e available
rules a n d r e p r e s e n t a t i o n s of a particular kind, n a m e l y structure-building rules a n d
underspecified representations. G i v e n such rules a n d representations, it will fol-
l o w that t h e simplest g r a m m a r s c o n s t r u c t e d b y learners will project precisely t h e
o b s e r v e d r a n g e of N D E B effects. In s u p p o r t of this position I s h o w that N D E B
effects are in m a n y w a y s sharply c o n s t r a i n e d b y structural p r i n c i p l e s . I p r o v i d e a
partial vindication for the claim b e h i n d the original R A C that obligatoriness a n d
neutralization are a n e c e s s a r y c o n d i t i o n for N D E B ( t h o u g h , I a r g u e , n o t a suffi-
cient c o n d i t i o n ) , and that cyclicity a n d lexical status are only indirectly c o n n e c t e d
to it. A m o n g the other c o n s e q u e n c e s of m y p r o p o s a l for w h i c h I also give e v i d e n c e
are that N D E B is restricted to structure-building (feature-filling) rules and that the
" d e r i v e d n e s s " of an e n v i r o n m e n t m u s t b e u n d e r s t o o d in a specific sense w h i c h
differs from w h a t h a s hitherto b e e n s u p p o s e d .
T h e next section of this article presents p h o n o l o g i c a l e v i d e n c e in s u p p o r t of the
position that N D E B is i n d e p e n d e n t of cyclicity a n d lexicality, a n d that it is a s s o -
ciated only with obligatory neutralization rules. I then i n t r o d u c e m y n e w a c c o u n t
of N D E B a n d s h o w h o w it w o r k s in the s t a n d a r d r a n g e of e x a m p l e s . S u b s e q u e n t
sections p r e s e n t s o m e additional k i n d s of e v i d e n c e s u p p o r t i n g it and s h o w h o w
s o m e a p p a r e n t l y recalcitrant c a s e s c a n b e dealt with in a w a y that is consistent
with it. I c o n c l u d e by d i s c u s s i n g N D E B in p r o s o d i c p h o n o l o g y , taking u p first
the possibility of unifying it with the free e l e m e n t condition p r o p o s e d for rules
of metrical structure a s s i g n m e n t , then its role in syllable structure and quantity,
and finally its relation to b l o c k i n g of h a r m o n y and tonal s p r e a d i n g by o p a q u e
elements.
280 Paul Kiparsky

2. N D E B IS N O T S P E C I F I C T O C Y C L I C O R L E X I C A L R U L E S

A s its n a m e indicates, the S C C w a s originally a condition on the application of

cyclic rules. T h e reduction to the e l s e w h e r e condition w o u l d m a k e it a condition
on lexical rules, i n d e p e n d e n t l y of their cyclic status. In this section I argue, con-
trary to both these v i e w s , that the N D E B p h e n o m e n o n is i n d e p e n d e n t both of
cyclicity and of the lexical status of a rule.

2.1. There Are Cyclic Lexical Rules with N o N D E B

First I d e m o n s t r a t e this i n d e p e n d e n c e in o n e direction b y exhibiting a cyclic

lexical rule w h i c h applies freely in n o n d e r i v e d e n v i r o n m e n t s . T h i s is the vowel
c o a l e s c e n c e p r o c e s s in F i n n i s h stated in (1).

(1) COALESCENCE:
ea, ed —> ee
da —> 66
oa —> oo

C o a l e s c e n c e is found in several regional dialects and applies as an optional rule in

colloquial Helsinki s p e e c h ( w h i c h the cited data represent). It is applicable to any
unstressed vowel s e q u e n c e s of the form (1), w h e t h e r u n d e r l y i n g , as in (2a), or
derived, as in (2b).

(2) a. / p i m e a / pimea ~ pimee 'dark' (nom.sg.)

b . / n i m e + t a 7 nime+a~ nimee 'name' (part.sg.)

T h e rule w h i c h feeds C o a l e s c e n c e in (2b) deletes t in an onset after a short vowel

w h i c h d o e s not b e a r p r i m a r y stress ( w h i c h is a l w a y s o n the initial syllable), as in
/ h a t t u + t a / hattua ' h a t ' , / j o + t a / jota ' w h a t ' , / v e e s e e + t a 7 veeseetd ' t o i l e t ' .
4

(3) ^-DELETION: a a a

t—• 0

T h e r e is n o f-Deletion after the long vowels created b y C o a l e s c e n c e . For e x -

a m p l e , only the u n c o a l e s c e d variant of (2a) allows r-Deletion after it. B e c a u s e
C o a l e s c e n c e is optional, this results in the variation seen in (4).

(4) / p i m e a + t a 7 pimedd ~ pimee+ta (*pimee+d) ' d a r k ' (part.sg.)

Blocking in Nonderived Environments 281

So, since /-Deletion feeds C o a l e s c e n c e a n d C o a l e s c e n c e b l e e d s /-Deletion, C o -

a l e s c e n c e m u s t apply both before and after /-Deletion. Specifically, the o r d e r
is: (1) C o a l e s c e n c e within s t e m s , (2) /-Deletion in s t e m + s u f f i x c o m b i n a t i o n s ,
(3) C o a l e s c e n c e in s t e m + s u f f i x c o m b i n a t i o n s . T h i s o r d e r of application requires
that /-Deletion and C o a l e s c e n c e apply cyclically, in that order, with " a n y w h e r e "
^syllabification:

(5) /pimea/ /nime/

FIRST /-Deletion — —
CYCLE: Coalescence pimee ~ pimea —
(optional)
SECOND Morphology pimee+ta ~ pimea+ta nime+ta
CYCLE: /-Deletion pimee-\~ta ~ pimea+a nime+a
Coalescence — nimee ~ nime+a
(optional)

A s e c o n d a r g u m e n t for the cyclic lexical status of b o t h rules c o m e s from their

interaction with the m o r p h o l o g y . B o t h rules m u s t apply to s t e m s before the allo-
m o r p h s of lexical suffixes a d d e d to t h e m are selected b y the p h o n o l o g i c a l l y c o n -
ditioned a l l o m o r p h y r u l e s in (6).

(6) a. ILLATIVE SINGULAR: T h e e n d i n g is -seen after a " c o n t r a c t e d " (underly-

ingly disyllabic) long v o w e l , and -(h)Wn e l s e w h e r e . E.g., altaa+seen
(stem / a l t a s V / ) 'into the p o o l ' , talo+(h)on (not *talo+seeri) 'into the
house'. 5

b. POSSESSIVE SUFFIX: After a c a s e suffix e n d i n g in short v o w e l , the 3-sg.

p o s s e s s i v e suffix h a s an optional a l l o m o r p h -Wn. E.g., talo+ssa+an ~
talo+ssa+nsa 'in h i s / h e r / i t s / t h e i r h o u s e ' , but only kala+t+nsa
[-» kala-\msa, by (8)], ' h i s / h e r / i t s / t h e i r fishes' (not *kala+t+an).

T h e e x a m p l e s in (7) s h o w h o w the right f o r m s are d e r i v e d only if the a l l o m o r p h s

are c h o s e n AFTER the application of C o a l e s c e n c e and /-Deletion to the stem.

(7) a. / p i m e a / ' d a r k ' : (ill.sg.) pimea + (h)an ~ pimee+seen (*pimea+seeri)

[(6a) fed b y C o a l e s c e n c e ]
b. / n i m e + t a / ' n a m e ' (part.sg.): 3 p o s s . nime+a+an ~ nimee+nsa
(~ *nimee+an) [(6b) bled by C o a l e s c e n c e ]
c. / k a l a + t a / 'fish' (part.sg.): 3 p o s s . *kala+a+an (kala+a+nsa) [(6b)
bled by /-Deletion; this derivation requires that ^ s y l l a b i f i c a t i o n is a
cyclic rule or an e v e r y w h e r e rule]

O n t h e a s s u m p t i o n that cyclic p h o n o l o g i c a l rules are interspersed with the m o r -

phology, the c o r r e c t distribution of a l l o m o r p h s follows directly from the cyclic
status of C o a l e s c e n c e a n d /-Deletion. T h e c h o i c e of a l l o m o r p h s is s i m p l y deter-
m i n e d by the s t e m ' s p h o n o l o g i c a l s h a p e at the point w h e n the suffixes are a d d e d .
282 Paul Kiparsky

F o r theories w h i c h d o not allow cyclic p h o n o l o g y to interact with m o r p h o l o g y

(e.g., O d d e n , this v o l u m e ) , it is n o t so c l e a r h o w to a p p r o a c h these facts. O n e m o v e
w o u l d b e to stipulate that -seen a n d -Yn a l l o m o r p h s are licensed by a filter, or
i n t r o d u c e d by a r e p l a c e m e n t p r o c e s s in place of the " b a s i c " a l l o m o r p h s , AFTER
the p h o n o l o g i c a l rules h a v e applied. T h i s fails b e c a u s e rule (8), w h i c h deletes a
c o n s o n a n t before a p o s s e s s i v e suffix, such as the suffix-final -n in (9), d o e s not
license the p o s s e s s i v e suffix a l l o m o r p h y stated in (6), as seen from the u n g r a m -
maticality of the -Yn a l l o m o r p h of the p o s s e s s i v e suffix in (10a).

(8) C->0/ +Poss

(9) / h u o n e e + s e e n + n i / —» huoneeseeni ' m y r o o m ' (ill.sg.)

/ h a t t u + h u n + s i / - > hattu(h)usi ' y o u r h a t ' (ill.sg.)

(10) a. * / h a t t u + h u n + V n / - > * h a t t u + ( h ) u + V n - > * h a t t u + ( h ) u + u n

b. / h a t t u + h u n + n s a / —» h a t t u + ( h ) u + n s a

Indeed, as the analysis of K a n e r v a (1987) m a k e s clear, the conditions on the - V n

a l l o m o r p h of the 3.p. p o s s e s s i v e suffix w o u l d on these a s s u m p t i o n s h a v e to b e
c h e c k e d by filters at two separate stages of the p h o n o l o g i c a l derivation. This
w o u l d o b v i o u s l y require a still m o r e drastic w e a k e n i n g of m o r p h o l o g i c a l theory.
In s u m , I h a v e p r e s e n t e d t w o a r g u m e n t s that /-Deletion and C o a l e s c e n c e are
lexical rules w h i c h apply cyclically in that order. T h e first a r g u m e n t is that this
resolves the o r d e r i n g p a r a d o x raised by the data in (2) and (4) and predicts the
correct interaction of C o a l e s c e n c e a n d /-Deletion in all cases. T h e s e c o n d argu-
m e n t is that it predicts the o t h e r w i s e p r o b l e m a t i c p h o n o l o g y / a l l o m o r p h y inter-
action in (7) a n d is consistent with the restrictive and empirically well sup-
p o r t e d position that " a l l o m o r p h y " is not r e p l a c e m e n t but selection (Lieber, 1987;
Z w i c k y , 1986), and that " g l o b a l " c o n d i t i o n s on the selection of a l l o m o r p h s are
not allowed.
B u t recall that, as (2) s h o w s , C o a l e s c e n c e applies in n o n d e r i v e d e n v i r o n m e n t s .
It is thus an e x a m p l e of a cyclic lexical rule (significantly, an optional, a l t h o u g h
neutralizing, one) w h i c h applies across the b o a r d . T h e u p s h o t is that the N D E B
effect c a n n o t b e a p r o p e r t y of all cyclic lexical rules.

2.2. N D E B in W o r d - L e v e l a n d Postlexical R u l e s

A fair a m o u n t of e v i d e n c e h a s already a c c u m u l a t e d w h i c h s h o w s the i n d e p e n -

d e n c e of N D E B a n d cyclic/lexical status in the other direction. N D E B effects h a v e
b e e n found b o t h in rules w h i c h apply noncyclically at the w o r d level ( H a r g u s ,
1985, 1989; H u a l d e , 1989; K a i s s e , 1986; Shaw, 1 9 8 5 : 1 9 9 ) , and in postlexical
rules (Clark, 1990:117ff.; H a r g u s , 1985; Iverson and W h e e l e r , 1988; R i c e , 1988).
M a n y of the rules p r e s e n t e d in the earlier literature as cases of " s t r i c t c y c l i c i t y "
h a v e turned out to b e in reality n o n c y c l i c . For e x a m p l e , the Finnish assibilation
Blocking in Nonderived Environments 283

rule ( 1 1 ) s h o w s classic d e r i v e d - e n v i r o n m e n t b e h a v i o r (Kiparsky, 1 9 7 3 ) . In a w o r d

like / t i l a t + i / —> tilasi ' o r d e r e d ' , Assibilation applies only to the s e c o n d /t/, the
o n e before a d e r i v e d i. H o w e v e r , Assibilation m u s t in fact apply at the w o r d level.
In derivations like ( 1 2 ) it is fed by a rule w h i c h raises -e to -i at the e n d of a w o r d .

(11) t-^sl—i

(12) a. / v e t e / —> veti —> vest ' w a t e r ' ( n o m . s g . )

b. / v e t e + n a / —> vetend (ess.sg.)

S i n c e Assibilation is fed b y a word-level rule, it m u s t itself apply at least at the

w o r d level or later ( t h o u g h it may, as far as these d a t a are c o n c e r n e d , also apply
cyclically).
A n o t h e r such rule in F i n n i s h is C o n s o n a n t G r a d a t i o n , w h i c h d e g e m i n a t e s
d o u b l e stops a n d w e a k e n s s i m p l e stops in the onset of a c l o s e d syllable. T h e e x -
a m p l e s in ( 1 3 ) illustrate the regular alternation b e t w e e n g e m i n a t e a n d s i m p l e stops
i n d u c e d by C o n s o n a n t G r a d a t i o n in the w o r d hattu ' h a t ' .

(13) a. hattu ( n o m . s g . ) , hattu+a (part.), hattu+na (ess.)

b. hatu+n (gen.), hatu+ssa (iness.), hatu+sta (elat.), hatu+ksi (transl.)

B o t h i n d i g e n o u s a n d b o r r o w e d v o c a b u l a r y testify to the rule's failure to apply in

nonderived environments.

(14) a. N A T I V E A N D N A T I V I Z E D MORPHEMES: sitten " t h e n " , kippis " c h e e r s ! " ,

hellanlettas (affectionate e x c l a m a t i o n ) , tattis " t h a n k s " , G e n . P l . -tten
(e.g., huone+i+tten "rooms")
b. L O A N S : appelsiini ' o r a n g e ' , pikkelsi ' p i c k l e s ' , attentaatti 'assassina-
tion ( a t t e m p t ) ' , okkultaatio ' o c c u l t a t i o n ' , kettinki ' c h a i n ' , rottinki 'rat-
tan', appellatiivi ' c o m m o n n o u n ' , opportunisti ' o p p o r t u n i s t ' , hokkus-
pokkus ' h o c u s p o c u s ' , bakkantti ' b a c c h a n t e ' , supportti ' s u p p o r t (of
a lathe)'

T h e following e x a m p l e illustrates all three cases in a single word, n a m e l y (taking

the tt's from left to right), ( 1 ) n o n a p p l i c a t i o n in a n o n d e r i v e d e n v i r o n m e n t ,
( 2 ) application in a m o r p h o l o g i c a l l y derived e n v i r o n m e n t , and ( 3 ) application in
a p h o n o l o g i c a l l y derived e n v i r o n m e n t .

(15) /hottentotti+ttoma+ta/ -» hottentotti+ttom+ta —> hottentoti+ton+ta

' H o t t e n t o t l e s s ' (part.sg.)

I return to the derivation of such cases in Section 5 . 1 .

I turn n o w to the e v i d e n c e that C o n s o n a n t G r a d a t i o n is a word-level rule. T h a t
it m u s t apply at w o r d level is s h o w n by the fact that it is fed by a rule w h i c h deletes
word-final -e in p o l y s y l l a b l e s . T h e closed syllable resulting from this word-level
6

deletion p r o c e s s triggers d e g e m i n a t i o n . 7

(16) / v a a t t e C e / —> vaatteC —> vaateC

284 Paul Kiparsky

T h a t C o n s o n a n t G r a d a t i o n applies only at w o r d level, not also in the cyclic

p h o n o l o g y , is suggested b y its interaction with the c o n s o n a n t deletion rule (8).
Deletion by rule (8) b l e e d s G r a d a t i o n in cases like (17).

(17) / h a t t u + n + s i / —> hattusi (*hatusi) 'your hat's' (hat+gen.sg.+2-sg.poss.)

/ h a t t u + t + n i / —> hattuni {*hatuni) 'my hats' (hat+nom.pl. +1-sg.poss.)

If C o n s o n a n t G r a d a t i o n w e r e cyclic, it w o u l d w r o n g l y apply at the stage hattu+n,

hattu+t, before the p o s s e s s i v e suffix is a d d e d o n the last cycle, triggering the
application of ( 8 ) . 8

Finally, I note that C o n s o n a n t G r a d a t i o n is lexical (this is actually inessential to

m y a r g u m e n t ; the crucial point is that it is a word-level rule). It applies only within
the m o r p h o l o g i c a l word, e x c l u d i n g clitics p h o n o l o g i c a l l y adjoined to it. This im-
plies lexical status if w e a d o p t Inkelas and Z e e ' s suggestion that the "clitic g r o u p "
is not a separate p r o s o d i c c a t e g o r y b u t the postlexical incarnation of the p h o n o -
logical w o r d (Inkelas, 1989; Z e e , this v o l u m e ) , in w h i c h c a s e application in (18)
c a n n o t b e b l o c k e d by restricting the rule to a p r o s o d i c d o m a i n . A g e m i n a t e arising
across clitic b o u n d a r y is not subject to C o n s o n a n t G r a d a t i o n . 9

(18) / m e n e C # p a s / —> mene[pp]as (-/» *mene[p]as) 'go!'

/ i t s e C # k i n / —> itse[kk]in (-/> *itse[k]in) 'self, t o o '

B u t if word-level rules are not c o n s t r a i n e d b y the S C C , and C o n s o n a n t G r a d a -

tion applies at w o r d level, w h a t b l o c k s it from a p p l y i n g in n o n d e r i v e d cases like
( 1 4 ) ? T h e c o n c l u s i o n is that word-level rules can b e restricted to derived e n v i r o n -
m e n t s exactly as cyclic rules can.
I turn n o w to the e v i d e n c e that N D E B m u s t also b e c o u n t e n a n c e d for post-
lexical rules.
E v i d e n c e w o u l d b e e x p e c t e d to b e scarce for rules w h i c h apply O N L Y postlexi-
cally, for such rules are n o r m a l l y n o n n e u t r a l i z i n g (in c o n s e q u e n c e of the strict
d o m a i n h y p o t h e s i s and structure p r e s e r v a t i o n ) , a n d n o n n e u t r a l i z i n g rules are not
subject to N D E B (for r e a s o n s to b e e x p l a i n e d later). T u r k i s h final stop devoicing
m a y b e a case, h o w e v e r (Kaisse, 1986). B u t the p o i n t can b e m a d e equally well
with rules w h i c h apply both lexically and postlexically. If N D E B w e r e turned off
postlexically, any b l o c k i n g in their lexical applications w o u l d b e overridden by
their postlexical applications, and so they w o u l d effectively apply across the
board. Yet m a n y rules with this dual status d o display N D E B .
T h e Sanskrit ruki rule has b e e n cited as a s h o w c a s e e x a m p l e of derived-
e n v i r o n m e n t b e h a v i o r (Kiparsky, 1982). In Classical Sanskrit it applies strictly
within the m o r p h o l o g i c a l / l e x i c a l w o r d . In Vedic, however, it also applies option-
ally across clitic b o u n d a r y ( M a c d o n e l l , 1 9 6 8 : 4 5 - 4 6 ; Selkirk, 1980; Wacker-
nagel, 1 8 9 5 : 2 3 7 ; Whitney, 1 8 8 7 : 6 4 ) and ( m o r e rarely) b e t w e e n w o r d s within a
phonological phrase.
Blocking in Nonderived Environments 285

(19) a. CLITIC B O U N D A R Y : rcchdntisma 'they w e n t ' (RV. 10.102.6)

b. W O R D B O U N D A R Y : nu sthirdm ' n o w the strong o n e ' (RV. 1.64.15)
A p p l i c a t i o n of the rule across w o r d b o u n d a r y in Vedic is certainly e v i d e n c e that it
applied postlexically in that d i a l e c t , t h o u g h not necessarily only postlexically.
10

B u t the ruki rule w a s restricted to " d e r i v e d e n v i r o n m e n t s " even in Vedic. T h e r e -

fore, in o r d e r to block postlexical applications of the ruki rule in such m o r p h o -
logically s i m p l e w o r d s as bisa ' s p r o u t ' , busa ' m i s t ' , it m u s t be possible to restrict
even postlexical rules to derived e n v i r o n m e n t s . W h a t e v e r principle is responsible
for N D E B m u s t stay in force postlexically here.
T h e proposal p r e s e n t e d in the next section is d e s i g n e d to a c c o m m o d a t e such
cases w i t h o u t losing the respective theoretical and empirical virtues of p r e v i o u s
a c c o u n t s of the N D E B .

3. N D E B E F F E C T S E X P L A I N E D A W A Y

3.1. The Underspecification Account

I a d o p t the following a s s u m p t i o n s .
(20) a. Underspecification (Kiparsky, 1982; A r c h a n g e l i and Pulleyblank,
1989).
b. Default rules m a y be o r d e r e d to apply either cyclically, at the w o r d
level, or postlexically (Pulleyblank, 1986; R i c e , 1988).
c. T h e optimal g r a m m a r is the simplest ( C h o m s k y and Halle, 1968).
d. S t r u c t u r e - c h a n g i n g rules are to b e d e c o m p o s e d into deletion (delink-
ing) plus structure-building (Poser, 1982; M a s c a r o 1987; C h o , 1990).
For p u r p o s e s of the a r g u m e n t , underspecification is u n d e r s t o o d in the strictest
sense, that is, w h a t Steriade ( 1 9 8 7 a ) calls RADICAL UNDERSPECIFICATION, includ-
ing the a s s u m p t i o n of strict binarity of feature specifications in u n d e r l y i n g lexical
representations. In e a c h e n v i r o n m e n t , w e can h a v e at m o s t [OFJ and [ a F ] , w h e r e
[ — a F ] is the value assigned by the m o s t specific rule ( l a n g u a g e - p a r t i c u l a r or uni-
versal) w h i c h is a p p l i c a b l e in that e n v i r o n m e n t . (That is, the relevant " e n v i r o n -
m e n t s " are defined by the rule system, including the m a r k e d n e s s rules of universal
g r a m m a r ) . T h e e s s e n c e of m y p r o p o s a l could be m a i n t a i n e d even u n d e r w e a k e r
versions of underspecification, however, as is s h o w n below.
A corollary of (20c) and (20d) is that rules will b e structure-building if possible
(i.e., unless positive e v i d e n c e requires positing a deletion rule). A corollary of
(20c) and (20a) is that u n d e r l y i n g representations are m i n i m a l l y specified.
M y principal thesis is:
(21) N D E B is the result of structure-building rules applying to underspecified
representations.
286 Paul Kiparsky

F o r e x a m p l e , the F i n n i s h Assibilation rule (11) w o u l d b e a structure-building

rule that assigns the feature specification [ + c o n t i n u a n t ] to a coronal obstruent
before i. E l s e w h e r e , c o r o n a l o b s t r u e n t s will b e assigned the default specification
[ — continuant].

(22) a.
I +coronal
—> [ + c o n t i n u a n t ] / i
+obstruent_

b. [ + o b s t r u e n t ] —> [ - c o n t i n u a n t ]

G i v e n these rules, the u n d e r l y i n g representations for non-alternating / t / a n d / s /

m u s t b e as follows.

(23) Itl Isl

Before /; [ - cont] [Ocont]
Elsewhere: [Ocont] [+cont]

In m o r p h e m e - i n t e r n a l / t i / and / s a / s e q u e n c e s , the c o n s o n a n t s are specified as re-

spectively [ - c o n t i n u a n t ] a n d [ + c o n t i n u a n t ] . T h e derivations in (24) s h o w h o w
this w o r k s (capitals d e n o t e s e g m e n t s unspecified for the feature [ ± continuant]).

(24) a. / t i l a T + i / -> [tilas+i] [by(22a,b)]

b. / s a T a / - > [sata] [by (22b)]

S o the difference b e t w e e n the application of Assibilation in derived and n o n -

derived e n v i r o n m e n t s follows directly from their different lexical representations
a n d from the structure-building status of the rule. B o t h of these are in turn deter-
m i n e d b y the simplicity p r i n c i p l e (20c). F o r lexical representations, simplicity
enforces the m i n i m a l u n d e r l y i n g feature specifications. G i v e n that the feature
[continuant] is distinctive and rules (22a,b) (the latter from universal g r a m m a r ) ,
the m i n i m a l system of r e p r e s e n t a t i o n s m u s t b e as in (23). A n d the learner's as-
s u m p t i o n that Assibilation is structure-building is also a c o n s e q u e n c e of the sim-
plicity principle, b e c a u s e (20d) g u a r a n t e e s that structure-building rules are the
m i n i m a l (simplest) rules, w h i c h therefore will b e c h o s e n in the a b s e n c e of c o n -
trary e v i d e n c e . C o n t r a Iverson a n d W h e e l e r , 1988, such underspecification d o e s
not involve the " a d h o c " " j u d i c i o u s m a n i p u l a t i o n " of the u n d e r l y i n g represen-
tations. It is dictated i n d e p e n d e n t l y of the N D E B effect by the r e q u i r e m e n t that
the m i n i m a l l y specified u n d e r l y i n g representations b e selected, in accord with
simplicity.
U n d e r these a s s u m p t i o n s , then, derived contexts are exactly those contexts in
w h i c h the learner will a s s u m e that a rule applies. In s u m , b y divorcing N D E B
from cyclicity/lexicality, a n d tying it instead to the structure-building versus
s t r u c t u r e - c h a n g i n g status of a rule (the latter b e i n g a n a l y z e d as the f o r m e r with
deletion), w e get the N D E B effect essentially for free. In addition, the properties
of N D E B i m m e d i a t e l y follow as well, as will n o w b e s h o w n .
Blocking in Nonderived Environments 287

I noted earlier that N D E B d o e s not a p p e a r in optional rules or in n o n n e u t r a l i -

zation rules. T h e f o r m e r generalization c a n b e d e r i v e d in the following way. M y
a s s u m p t i o n is that the simplest g r a m m a r results from m a x i m a l l y underspecified
lexical r e p r e s e n t a t i o n s and structure-building rules, and so m o r e fully specified
lexical representations and s t r u c t u r e - c h a n g i n g rules will be set u p only if positive
e v i d e n c e r e q u i r e s t h e m . A n a c r o s s - t h e - b o a r d optional rule can b e learned straight-
forwardly in c o n f o r m i t y with these constraints, for its very optionality p r o v i d e s
such positive e v i d e n c e for m o r e fully specified lexical representations, h e n c e for
the s t r u c t u r e - c h a n g i n g c h a r a c t e r of the rule. F o r e x a m p l e , in the c a s e of Finnish
C o a l e s c e n c e [rule (1)], the positive e v i d e n c e required for the learner to postulate
lexical r e p r e s e n t a t i o n s like leal in a w o r d like pimee is o b v i o u s l y the optional
u n c o a l e s c e d alternant pimea [see (2a)]. O n the o t h e r h a n d , a putative optional rule
restricted to a p p l y i n g in derived e n v i r o n m e n t s c a n only b e learned on the basis of
NEGATIVE e v i d e n c e , n a m e l y the n o n e x i s t e n c e of o n e of the variants in n o n d e r i v e d
cases. O n general learnability g r o u n d s , s u c h a system w o u l d b e at least unstable.
A s for w h y n o n n e u t r a l i z a t i o n rules are n o t b l o c k e d in n o n d e r i v e d e n v i r o n -
m e n t s , t h e reason is that in the c a s e of such rules the learner h a s distributional
e v i d e n c e for the " a b s t r a c t " u n d e r l y i n g form (with simplicity forcing the abstrac-
tion to b e m a d e ) . F o r e x a m p l e , in the simplest g r a m m a r of English, aspiration is
unspecified in u n d e r l y i n g representations a n d i n t r o d u c e d o n the basis of the p h o -
netic context.
T h a t rules should apply regularly in d e r i v e d e n v i r o n m e n t s requires a further
a s s u m p t i o n , however, n a m e l y the strict binarity a s s u m p t i o n . A s a s i m p l e illustra-
tion of w h a t is at stake in strict binarity, c o n s i d e r the interaction of the Assibilation
rule (11) w i t h the rule raising word-final e to /. B o t h are structure-building rules
a p p l y i n g to underspecified a r c h i p h o n e m e s . T h e i r interaction is illustrated in (25),
w h e r e ITI r e p r e s e n t s a c o r o n a l obstruent unspecified for c o n t i n u a n c y (t ~ s), and
IEI is a n o n l o w front v o w e l unspecified for h e i g h t (e ~ *')•

(25) a. /. . . T i / : lasi, lasi+na 'glass'

b. /. . . ti/: koti, koti+na 'home'
c. /. . . T E / : vesi, vete+na 'water'
d. /. . . s E / : kuusi, kuuse+na 'fir'

Binarity can b e m a i n t a i n e d if e a c h e n v i r o n m e n t s h o w s at m o s t a t w o - w a y o p -
position. T h i s is clearly the c a s e for continuancy, as s h o w n in (26), w h e r e the
c o l u m n s represent, respectively, s e g m e n t s alternating b e t w e e n t a n d s, fixed t a n d y

fixed s.

(26) /t-s/ III Isl

Before lil — [-cont] [Ocont]
Before IEI [Ocont] — [+cont]
Elsewhere — [Ocont] [+cont]
288 Paul Kiparsky

S u p p o s e that, in addition to the four alternation patterns in (25), Finnish h a d a fifth

pattern, exemplified by the m a d e - u p w o r d in (27).

(27) /. . . t E / : *mati, * materia

If this situation w e r e to b e r e p r e s e n t e d in a purely p h o n o l o g i c a l way, it w o u l d

require a three-way contrast / T E / : / s E / : / t E / for vesi, kuusi, *mati respectively.
B u t such a three-way contrast w o u l d violate the strict binarity principle. S o the
a b s e n c e of w o r d s like (27) w o u l d b e predicted if w e a s s u m e d strict binarity in
addition to underspecification.
In the case of height, F i n n i s h h a s a m a r g i n a l three-way contrast.

(28) /e~i/ Id l\l

Inenv ] [Ohigh] [-high] [+high]
Elsewhere — [Ohigh] [ - high]

T h e e to / rule fails to apply to a class of w o r d s c o m p r i s i n g nursery w o r d s such as

nukke ' d o l l ' , nolle ' t e d d y b e a r ' , h y p o c o r i s t i c n a m e s such as Kalle ' C h a r l e y ' , Ville
' W i l l i e ' , and abbreviations such as Yle ' P u b l i c R a d i o ' (from Yleisradio). H e r e the
final vowels w o u l d b e e x p e c t e d to raise as in (25c) and ( 2 5 d ) . T h e only way n

these w o r d s c o u l d b e treated as p h o n o l o g i c a l l y r e g u l a r on the p r e s e n t view is by

lexically specifying their final vowel as n o n h i g h , a n d this w o u l d entail a three-
w a y contrast as seen in the top line of (28). If they are specified as e x c e p t i o n s to
the e —> / rule, then all the right forms are derived in conformity with strict bi-
narity. T h e m a r g i n a l status of these stems m a k e s this alternative quite r e a s o n a b l e .
Similarly, the e x i s t e n c e of hypothetical w o r d s like (27) w o u l d not force the aban-
d o n m e n t of strict binarity, p r o v i d e d the correct w a y of a c c o u n t i n g for t h e m turned
out to b e as e x c e p t i o n s to rule (11).
N o t e that even if w e w e r e to c o m p r o m i s e on strict binarity, our analysis of the
N D E B effects in (25) c o u l d still b e m a i n t a i n e d . A l l o w i n g three-way distinctions
w o u l d yield a formally c o h e r e n t but w e a k e r theory. W e would, for e x a m p l e ,
lose the e x p l a n a t i o n for the n o n e x i s t e n c e of the pattern in (27). T h u s the viability
of strict binarity and the p r o p e r e x p l a n a t i o n of N D E B are s o m e w h a t separate
questions.
It is also clear w h y N D E B can persist in postlexical applications of rules w h i c h
first apply in the lexical m o d u l e , as in the previously cited cases. Postlexical ap-
plications are b l o c k e d in n o n d e r i v e d cases for exactly the s a m e r e a s o n s that the
lexical applications are b l o c k e d in t h e m , that is, b e c a u s e the l a n g u a g e d o e s not
have a s t r u c t u r e - c h a n g i n g rule. This d o e s not m e a n , however, that the b l o c k i n g
effect MUST always persist postlexically. It will not persist if the postlexical m o d -
ule contains, in addition to the structure-building rule w h i c h it shares with the
lexical m o d u l e , a s t r u c t u r e - c h a n g i n g rule of its o w n w h i c h deletes or delinks the
specifications of the s a m e feature. T h i s deletion rule will then create n e w inputs
to w h i c h the other rule, still in structure-building m o d e , can apply. E x a m p l e s for
w h i c h this m a y b e the a p p r o p r i a t e analysis are discussed in Kiparsky ( 1 9 8 5 ) .
Blocking in Nonderived Environments 289

Finally, cases of n o n f e e d i n g o r d e r in cyclic rules, w h i c h h a v e b e e n cited as

e v i d e n c e for t h e S C C ( R u b a c h , 1 9 8 4 : 1 3 ) , will n o w b e a c c o u n t e d for b y the cyclic
a s s i g n m e n t of default rules (on w h i c h m o r e in Kiparsky, 1992).
T h e w a y in w h i c h the p r o p o s e d t r e a t m e n t a c c o m m o d a t e s t h e standard e x a m p l e s
of strict cyclicity should n o w b e clear.

3.2. Korean Palatalization

A n o t h e r a p p a r e n t lexical-cum-postlexical rule with " d e r i v e d e n v i r o n m e n t " b e -

h a v i o r is K o r e a n Palatalization. A s p r e s e n t e d in Iverson a n d W h e e l e r ( 1 9 8 8 ) , t h e
K o r e a n situation w o u l d m a t c h t h e Sanskrit ruki rule very closely. However, the
additional facts b r o u g h t to light b y H u m e ( 1 9 9 0 ) a n d b y C h o a n d Sells (1991)
m o t i v a t e a s o m e w h a t different analysis. F o r this r e a s o n I discuss it separately here.
T h e p h o n e m i c inventory of K o r e a n i n c l u d e s t h e c o r o n a l c o n s o n a n t s , /t, t , c,
h

c , s, n, 1/. T h e p h o n e m e s /t, t / are n o r m a l l y realized as dental stops a n d / c , c /

h h h

are n o r m a l l y realized as postalveolar affricates. Before i, all c o r o n a l s a r e palatal-

ized; I will write the resulting prepalatal s e g m e n t s as [{, f , c, q , §, n, A.]. Palatal-
h h

ization applies uniformly t o all c o r o n a l s before / regardless of w h e t h e r t h e e n v i -

r o n m e n t is derived o r not.

(29) a. / s i / —> [§i] ' p o e m '

/k'ini/—> [k'ifii] ' m e a l '
/ p ' a l l i / - ^ [p'aXXi] 'fast'
b. / o s + i / —> [o§i] ' c l o t h e s - N o m . '
c. / p a t ilarj/ —»[pag irarj] ' r i d g e of a field'
h 1

In (29c), t h e stop is deaspirated in word-final position a n d voiced m e d i a l l y b y a

separate rule.
In addition to b e c o m i n g palatalized, /t, t / g e t affricated if t h e triggering i is
h

within t h e s a m e w o r d , thereby c o i n c i d i n g with t h e a l l o p h o n e s of u n d e r l y i n g Id

and lc l in that e n v i r o n m e n t .
h

(30) a. / p a t + i / - > [pachi] 'field ( n o m . ) '

h

Ihx t o t + i / —> [hae doji] ' s u n r i s e ( n o m . ) '

/ m a t + i t a / —> [majida] 'eldest ( c o p u l a ) '
b. / t a c + i / —» [tachi] ' a n c h o r ( n o m . ) '
h

/ n a c + i / —> [naji] ' d a y ( n o m . ) '

Affrication only applies before a derived /. Before t a u t o m o r p h e m i c /, It, t l are h

retained, as palatalized [{, | ] , contrasting with [q, c ] from / c , c / in that position.

h h h

(It is these palatalized but unaffricated stops that the analysis p r o p o s e d b y Iverson
and W h e e l e r neglects.)

(31) / m a t i / [maoU] ' k n o t ' , / c a n t i / [can^i] ' g r a s s '

290 Paul Kiparsky

Id and / c / a p p e a r in lexical r e p r e s e n t a t i o n s before all v o w e l s and m o r p h e m e

h

finally.

(32) curjkuk ' C h i n a ' , -coc a

h
' e v e n ' , nac ' d a y ' , kac- 'to h a v e ' , tac h
'anchor'

T h e prepalatals are not p h o n e m i c a n d arise only from palatalization before / i / ,

w h i c h is a u t o m a t i c and obligatory.

(33) a. * [ . . . § a . . . ] , * [ . . . nu . . . ] , * [ . . . fo . . . ] , * [ . . . 9a.. .]
b. * [ . . . s i . . . ] , * [ . . . n i . . . ] , * [ . . . t i . . . ] , * [ . . . c i . . .]

S u p p o s e that u n d e r l y i n g /t, t / are distinguished from / c , c / b y the feature

h h

[ ± anterior]. T h e m i n i m a l l y specified lexical representations of palatalized and

nonpalatalized c o r o n a l s are then:

(34) /t, t / h
/c, c /h
/ s , n, 1/
Before i [+ant] [Oant] [Oant]
Elsewhere [Pant] [ - ant] [Oant]

L e t us take the p h o n e t i c specifications of the palatalized a n d nonpalatalized c o r o -

nals to b e

(35) t \ c 5 s §_
High - + - + - +
Anterior + - - - + -
D e l a y e d release — — 4- + — —

and a s s u m e that Palatalization s p r e a d s the features [ - a n t e r i o r ] and [ + h i g h ] to

coronal c o n s o n a n t s from a following [ + h i g h ] front vowel. (If w e a d o p t the p r o -
posal of C l e m e n t s , 1989, a n d L a h i r i a n d E v e r s , 1989, that front v o w e l s are
[ + c o r o n a l ] , w e can s i m p l y take Palatalization to b e spread of the Place node.)
Palatalization applies freely to all c o r o n a l s , b o t h in the lexical p h o n o l o g y a n d the
postlexical p h o n o l o g y . In the latter, it applies in feature-changing fashion. A l -
t h o u g h [ — anterior, + h i g h ] c o r o n a l s are not in the lexical inventory, structure
preservation m u s t not b l o c k Palatalization from applying lexically, p e r h a p s b e -
c a u s e , in virtue of the linking condition ( H a y e s , 1986; Ito, 1986), the relevant
lexical m a r k i n g constraint is inapplicable to the multiply linked structures that
result from spreading ( M a c f a r l a n d a n d P i e r r e h u m b e r t , 1991). T h e different out-
c o m e of lexical and postlexical palatalization of /t, t / is d u e to a word-level rule
h

that specifies the [ — anterior, + h i g h ] o b s t r u e n t stops as [ + d e l a y e d r e l e a s e ] . 12

E l s e w h e r e , c o r o n a l s are n o n p a l a t a l b y the default rule (36).

(36) [ + c o r o n a l ] —> [ + a n t e r i o r ]

Moreover, s o m e dialects (e.g., that of South H a m k y e n g ) have a lexical rule of

vowel fronting (umlaut) triggered b y a high front vowel or glide. A c c o r d i n g to
R a m s e y ( 1 9 7 8 ) a n d H u m e ( 1 9 9 0 ) , vowel fronting is b l o c k e d by any intervening
coronal (hence palatalized) c o n s o n a n t . In t e r m s of the present proposal, such dia-
Blocking in Nonderived Environments 291

lects differ m i n i m a l l y from the n o n u m l a u t i n g dialects. In the latter, palatality

spreads to the nearest C; in the u m l a u t i n g dialects palatality spreads to the nearest
eligible p l a c e n o d e , w h e t h e r C or V. T h a t is, u m l a u t is simply a generalization of
palatalization. T h i s relationship b e t w e e n u m l a u t a n d palatalization c o u l d not b e
c a p t u r e d if the palatalization rule w e r e split into a lexical a n d a postlexical part,
as Iverson a n d W h e e l e r p r o p o s e .
A n a p p a r e n t p r o b l e m for the word-level status of Affrication arises in c a s e s like
(37), w h e r e the n o m i n a t i v e c a s e e n d i n g -i on the face of it a p p e a r s to b e syntacti-
cally attached to a p h r a s e and p h o n o l o g i c a l l y cliticized to its last w o r d . 1 3

(37) sup -kwa

h
pa[c ]-i h
p uli-ta h

forest-and field-nom. b e g r e e n - d e c l .

H o w e v e r , C h o and Sells ( 1 9 9 1 ) a r g u e o n i n d e p e n d e n t g r o u n d s , including m o r p h o -

logical e v i d e n c e , that the e n d i n g s in question really are lexical suffixes, even in
cases like (37).
M o s t d e r i v e d - e n v i r o n m e n t rules d i s c u s s e d so far can b e a n a l y z e d straightfor-
w a r d l y as structure-building rules a p p l y i n g to underspecified representations. In
Section 5 I e x a m i n e s o m e trickier c a s e s (including F i n n i s h C o n s o n a n t G r a d a t i o n )
and a r g u e that, c o n t r a r y to a p p e a r a n c e s , they are not s t r u c t u r e - c h a n g i n g rule
either.
H a v i n g s h o w n h o w N D E B can b e r e d u c e d to the application of structure-
b u i l d i n g rules to underspecified r e p r e s e n t a t i o n s a n d h o w its m a i n p r o p e r t i e s are
thereby e x p l a i n e d , I p r o c e e d to s o m e m o r e surprising c o n s e q u e n c e s of this w a y of
l o o k i n g at the matter.

4. V A C U O U S L Y D E R I V E D E N V I R O N M E N T S C O U N T
AS UNDERIVED

A clear-cut e m p i r i c a l difference b e t w e e n the S C C and the underspecification

a c c o u n t has to d o with v a c u o u s l y derived e n v i r o n m e n t s : e n v i r o n m e n t s not distinct
from the u n d e r l y i n g e n v i r o n m e n t w h i c h arise in the c o u r s e of the derivation.
O n the classical S C C story, v a c u o u s l y derived e n v i r o n m e n t s should trigger the
application of lexical rules. In fact, the a p p a r e n t confirmation of this prediction of
the S C C w a s o n e of t h e m o s t interesting results of M a s c a r o ( 1 9 7 6 ) .
T h e p r e s e n t p r o p o s a l , t h o u g h , m a k e s j u s t the o p p o s i t e prediction. A feature
specification w h i c h b l o c k s a rule from a p p l y i n g in the lexical representation
should b l o c k the rule equally w h e n t h e s a m e e n v i r o n m e n t arises in d e r i v e d f o r m s .

4.1. Finnish

T h e prediction that v a c u o u s l y derived e n v i r o n m e n t s b e h a v e like u n d e r i v e d en-

v i r o n m e n t s is confirmed by the Finnish assibilation rule (11). It can b e triggered
292 Paul Kiparsky

by the past tense suffix / (38a). Certain final vowels, including are truncated
before it ( 3 8 b ) . 1 4

(38) a. halut+i+vat —> halusivat 'they w a n t e d '

b. karsi+i+vat —> karsivat 'they p r u n e d '

T h e Itl in (38) is [ — cont] b e c a u s e it p r e c e d e s HI in the u n d e r l y i n g representations.

In the a c c o u n t p r o p o s e d here, that m a k e s it i m m u n e from assibilation before d e -
rived HI as well. O n the S C C account, however, c a s e 2 of the definition of " d e -
rived e n v i r o n m e n t " should m a k e it subject to assibilation. In fact, t is NEVER
c h a n g e d to s in such c a s e s . Past tense f o r m s of verbs in -ti thus furnish a case of a
vacuously derived e n v i r o n m e n t w h i c h s u p p o r t s m y p r o p o s a l over the S C C .

(39) a. vaati+vat —> vaativat (*vaasivat) 'they d e m a n d '

b. vaati+i+vat —> vaativat (*vaasivat) 'they d e m a n d e d '

Since the Itl in (39b) starts out p r e c e d i n g an III in the s a m e m o r p h e m e and e n d s

u p p r e c e d i n g an HI in a different m o r p h e m e , the structural condition for assibila-
tion is m e t in a derived e n v i r o n m e n t , a n d so, a c c o r d i n g to the older view of d e -
rived e n v i r o n m e n t rules, assibilation should apply. O n the p r e s e n t p r o p o s a l , the
Itl m u s t be lexically specified as [ — c o n t i n u a n t ] for the r e a s o n s previously stated
(otherwise it w o u l d b e / s / ) , a n d this lexical specification will b l o c k the structure-
b u i l d i n g assibilation rule from a p p l y i n g in derived e n v i r o n m e n t s as w e l l . 1 5

W h a t a b o u t the c a s e s w h e r e it h a s b e e n specifically a r g u e d that N D E B d o dis-

a p p e a r in v a c u o u s l y derived e n v i r o n m e n t s ? T h e t w o m o s t i m p r e s s i v e e x a m p l e s
are Icelandic and Catalan.

4.2. I c e l a n d i c

T h e Icelandic rule of w-Umlaut, w h i c h turns a to d if u follows in the next

syllable ( A n d e r s o n , 1969; Kiparsky, 1984; O r e s n i k , 1977) applies regularly in
m o r p h o l o g i c a l l y derived e n v i r o n m e n t s such as (40a) but is b l o c k e d m o r p h e m e
internally in cases such as (40b).

(40) a. / h a r d + u r n / —> hordum ' h a r d ' (dat.pl.)

/ s a g a + u r / - » sbgur ' s a g a s ' (nom.pl.)
/ k a l l a + u m / —» kollum ' c a l l ' (1-pl.)
b. / a k u r / —> akur 'field'

T h e vowel in the s e c o n d syllable of w o r d s like akur is s y n c o p a t e d before syllabic

e n d i n g s , and w h e n their a thereby c o m e s to stand before -u, it d o e s u n d e r g o u-
Umlaut.

(41) a. dat.sg. / a k u r + i / —» akr-\-i

b. dat.pl. / a k u r + u m / - > bkr+um

In K i p a r s k y ( 1 9 8 4 ) , the contrast b e t w e e n (40b) and (41b) w a s e x p l a i n e d on the

a s s u m p t i o n that v a c u o u s l y derived e n v i r o n m e n t s , unlike n o n d e r i v e d e n v i r o n -
Blocking in Nonderived Environments 293

merits, are e x e m p t from N D E B , w h i c h is c o n t r a r y to the line w e are d e v e l o p -

ing here.
W h a t is not e x p l a i n e d on this analysis is why, in all w o r d s w h i c h s h o w w-umlaut
in v a c u o u s l y derived e n v i r o n m e n t s , such as (41), the c o n s o n a n t after the n o n -
u m l a u t - t r i g g e r i n g u is r. N a t i v e stems e n d i n g in other final c o n s o n a n t s h a v e only
6 . . . u, not a . . . u (e.g., hofud ' h e a d ' ) , and the loans w h i c h h a v e a . . . u d o not
s e e m to u n d e r g o s y n c o p e , such as kaktus ' c a c t u s ' , Bakkus ' B a c c h u s ' . It is suspi-
c i o u s , then, that precisely -r is subject to a rule inserting u before it in e n v i r o n -
m e n t s w h e r e it c a n n o t b e syllabified. A n d this epenthetic u also fails to trigger
w-Umlaut, even t h o u g h the e n v i r o n m e n t is o b v i o u s l y derived.

(42) / d a g + r / —> dag+Vr —» dagur 'day'

In p r e v i o u s a n a l y s e s the failure of w-umlaut in (42) w a s h a n d l e d by o r d e r i n g E p e n -

thesis after w-Umlaut. In the c o n t e x t of current p h o n o l o g i c a l theory, e p e n t h e s i s is
m o r e plausibly a n a l y z e d as insertion of an e m p t y vowel slot, w h i c h fails to trigger
u m l a u t b e c a u s e it has n o feature content at the point w h e r e u m l a u t applies and is
later a s s i g n e d the features of u. W h i c h e v e r of these options is taken, the non-
U m l a u t i n g nature of epenthetic u i m m e d i a t e l y raises the possibility of represent-
ing akur as / a k r / , w i t h w-epenthesis, w h i c h w o u l d automatically yield the failure
of w-umlaut. In the earlier analyses m e n t i o n e d above, this possibility w a s rejected
on the g r o u n d s that there is a class of s t e m s w h i c h g e n u i n e l y h a v e this u n d e r l y i n g
shape, such as /lifr/ lifur, w h i c h differ systematically from stems like akur in their
behavior, as O r e s n i k s h o w s .
H o w e v e r , this c o n c l u s i o n m a y h a v e b e e n p r e m a t u r e . O t t o s s o n ( 1 9 8 8 ) has n o w
w o r k e d out a detailed analysis of Icelandic in the lexical p h o n o l o g y f r a m e w o r k in
w h i c h the case for v a c u o u s l y derived e n v i r o n m e n t s d i s a p p e a r s . In his p r o p o s a l ,
the u of akur is a n a l y z e d as e p e n t h e t i c , a n d the difference b e t w e e n the p a r a d i g m s
of the akur and lifur t y p e s is a r g u e d to b e m o r p h o l o g i c a l rather than p h o n o l o g i c a l .
If this is accepted, then Icelandic w-umlaut, while s h o w i n g standard N D E B ef-
fects, gives n o s u p p o r t to the position that v a c u o u s l y derived e n v i r o n m e n t s c o u n t
as derived for p u r p o s e s of the N D E B effect.

4.3. Catalan

P e r h a p s the m o s t interesting case of vacuously derived e n v i r o n m e n t s appar-

ently c o u n t i n g as derived is the Catalan Vowel L o w e r i n g rule, to w h i c h M a s c a r o
d e v o t e d an e x t e n s i v e a n d insightful discussion in this theoretical context. H e un-
covered the generalization that Vowel L o w e r i n g (o, e —» o, c) takes place before
any suffix that assigns stress to the p r e c e d i n g syllable, and on that basis p r o p o s e d
the rule that it applies to stressed syllables. In o r d e r to block Vowel L o w e r i n g in
unsuffixed s t e m s , he a s s u m e d that s t e m s bear u n d e r l y i n g stress and i n v o k e d the
S C C to block Vowel L o w e r i n g of syllables w h o s e stresses are not derived. Stress
v a c u o u s l y assigned to the stem by the prestressed suffixes m u s t thus c o u n t as
derived for p u r p o s e s of the S C C , on this analysis:
294 Paul Kiparsky

(43) sentrd ' c e n t e r ' , scntr+ik

direkto ' d i r e c t o r ' , direktor+i

T h e a s s u m p t i o n that stress in C a t a l a n is m a r k e d in the lexicon is therefore crucial

to M a s c a r a ' s analysis of Vowel L o w e r i n g . A recent study of Catalan stress, h o w -
ever, h a s c o n s i d e r a b l y u n d e r m i n e d this p r e m i s e . A l s i n a ( 1 9 8 7 ) a r g u e s that stress
is not m a r k e d in the lexicon, but p r e d i c t a b l e . T h e stress of all the w o r d s in (43)
16

is thus " d e r i v e d " in exactly the s a m e way, n a m e l y s i m p l y b y the application of

the stress rule. A l s i n a o b s e r v e s that l o w e r i n g can b e formulated as a rule w h i c h
a s s i g n s [-How] to v o w e l s before any suffix at Level 1 (in fact, the e n v i r o n m e n t
m i g h t b e simply before any syllable). T h i s rule applies in structure-building fash-
ion to v o w e l s not already specified for the feature [low].

(44) a. / s E n t r o / —> sEntrQ -» sentrd

b. / s E n t r a + i k / —» sEntr+ik —> sentr+ik

T h a reason Presuffixal Vowel L o w e r i n g can simply apply before any suffix is that,
unless the suffix is prestressed, the stem vowel will s i m p l y get r e d u c e d and the
effects of Vowel L o w e r i n g will b e u n d e t e c t a b l e .
M a s c a r o also a r g u e s for N D E B in a rule h e m o t i v a t e s for C a t a l a n w h i c h takes
o to u a n d e to Q } Let us refer to this rule as M i d Vowel R e d u c t i o n . Certain
1

w o r d s (unassimilated loans a n d learned w o r d s , it s e e m s ) d o not o b s e r v e M i d

Vowel R e d u c t i o n .

(45) [boston] ' B o s t o n '

[katedro] ' ( a c a d e m i c ) c h a i r '
[soprano] ' s o p r a n o '

M a s c a r o suggests that these u n s t r e s s e d m i d vowels fail to u n d e r g o M i d Vowel

R e d u c t i o n b y the S C C (again on the a s s u m p t i o n that they are underlyingly un-
stressed). Interestingly, stress-shifted derivatives of such exceptional w o r d s d o un-
d e r g o reduction.

(46) [bustun+ya] 'Bostonian'

[ k a t a d r a t i k ] ' h o l d e r of an ( a c a d e m i c ) c h a i r '

M a s c a r o a s s u m e s that application of the stress rule o n the derived w o r d cycle

c a u s e s the u n s t r e s s e d v o w e l s to get v a c u o u s l y respecified as unstressed, and that
s u c h v a c u o u s respecification creates d e r i v e d e n v i r o n m e n t s , so that t h e contrast
b e t w e e n (45) and (46) is a c a s e of the S C C at w o r k . T h i s c a s e has a rather different
character from the previously d i s c u s s e d Vowel L o w e r i n g case, however, for the
w o r d s in (45) are, as M a s c a r o notes, e x c e p t i o n a l . A reviewer suggests that w h a t
c o u l d b e g o i n g o n h e r e is the e l i m i n a t i o n of e x c e p t i o n features in derivatives ( K i -
parsky, 1973), as in F r e n c h Hitler (with /i-aspire) versus hitlerien ( r e g u l a r ) . 18
Blocking in Nonderived Environments 295

5. S T R U C T U R E - C H A N G I N G R U L E S A R E N O T S U B J E C T T O N D E B

R e d u c i n g N D E B to the operation of structure-building rules on underspecified

representations predicts that s t r u c t u r e - c h a n g i n g rules will n e v e r b e subject to it,
b e c a u s e they c a n delink a prespecified feature.
T h i s prediction is i n d e p e n d e n t of the c l a i m of strict binarity of feature specifi-
cations in lexical r e p r e s e n t a t i o n s . M y analysis of N D E B is consistent with the
e x i s t e n c e of t h r e e - w a y lexical feature distinctions. W h a t d e p e n d s on strict binarity
is the regularity of alternations in derived e n v i r o n m e n t s .
In the m a t e r i a l e x a m i n e d so far, r e p r e s e n t i n g the standard types of cases cited
in the S C C literature, it is fairly clear that w e are i n d e e d dealing with structure-
b u i l d i n g rules. F o r s o m e rules this is not so o b v i o u s , t h o u g h .

5.1. Finnish Consonant Gradation

T h e p r o b l e m raised b y C o n s o n a n t G r a d a t i o n is that, in a s e q u e n c e such as /. . .

V t t V t t V C C V . . ./, the result is /. . . V t V t V C C V . . ./ [see (15) for an e x a m p l e ] .
T h e m i d d l e I til acts as a g e m i n a t e since it triggers G r a d a t i o n on its left, but itself
u n d e r g o e s d e g e m i n a t i o n d u e to the c l o s e d syllable o n its right. S u c h instances of
m u l t i p l e application m a k e C o n s o n a n t G r a d a t i o n l o o k like a s t r u c t u r e - c h a n g i n g
rule, b e c a u s e the / t t / m u s t b e u n d e r l y i n g to trigger the rule. B u t this is inconsistent
with o u r c l a i m that N D E B is a p r o p e r t y only of structure-building rules.
A strictly structure-building a c c o u n t of d e g e m i n a t i o n is p o s s i b l e by separating
skeletal a n d m e l o d i c content. S u p p o s e that the lexical representation of a m o r -
p h e m e consists of a m e l o d y (subject to O C P , a n d appropriately underspecified)
a n d a skeleton (also p e r h a p s underspecified). T h e association b e t w e e n the m e l o d i c
e l e m e n t s a n d the skeletal slots is p r e d i c t a b l e b y the standard association c o n v e n -
tions and therefore n o t specified in the lexicon. (47) s h o w s , u s i n g C V notation,
h o w this w o u l d w o r k for a w o r d like aatto ' e v e ' .

(47) v v c c v v v c c v

- N N I
a t o a t o

C o n s o n a n t G r a d a t i o n ( d e g e m i n a t i o n ) is then s i m p l y the following Finnish-

specific restriction on the universal association p r o c e d u r e .

(48) Before a V C r h y m e , an o b s t r u e n t stop is associated only to the onset.

If a - C or - C C V e n d i n g is a d d e d to the stem, (48) forces o n e of the adjacent C

slots to r e m a i n vacant. A s s u m i n g that preference is given to o n s e t s , this yields, for
the genitive aaton,
296 Paul Kiparsky

(49) v v c c v +c v v c c v +c

- N I I I
a t o +n a t o +n

W e k n o w from other e v i d e n c e that u n a s s o c i a t e d C-slots trigger C o n s o n a n t Gra-

dation a n d that they get deleted only postlexically, if at all. B o t h these a s s u m p t i o n s
are i n d e p e n d e n t l y m o t i v a t e d by word-final " g h o s t c o n s o n a n t s " (actually " g h o s t
syllables," see K e y s e r and Kiparsky, 1984). F o r instance, in aate ' i d e a ' , the dele-
tion of the word-final vowel in the n o m i n a t i v e results in a syllable closed b y an
u n a s s o c i a t e d C-slot, c a u s i n g C o n s o n a n t G r a d a t i o n .
(50) V V C C V C V V V C C V C

- N i l
a t e a t e

T h e final e m p t y C-slot is not realized in isolation b e c a u s e it c a n n o t b e associated

with any c o n s o n a n t a l s e g m e n t , but its persistence into the postlexical p h o n o l o g y
is s h o w n by the fact that a c o p y of the next c o n s o n a n t spreads o n t o it in sandhi,
resulting in g e m i n a t i o n , as in aatev voitti ' t h e idea w o n ' . In genitive aatteen, the
condition (48) is inapplicable, so association is m a x i m a l .
(51) v v c c v c v +c v v c c v c v +c

- \l N I
a t e +n a t e +n

O n this view of C o n s o n a n t G r a d a t i o n , cases with multiple d e g e m i n a t i o n in

c o n s e c u t i v e syllables are not p r o b l e m a t i c any longer, as the derivations of
hatu+ttoma+na, hatu+ton ' h a t l e s s ' (ess. a n d n o m . sg.) in (52) show. In the
n o m i n a t i v e , the V slot of the -a of the privative suffix / t t o m a / is deleted, again
closing the syllable and c a u s i n g d e g e m i n a t i o n . 19

(52) a. c v c c v + c c v c v + c v

II I IN I I I I I
h a t u + t o m a +n a

b. c v c c v + c c v c

I I I I III
h a t u + t o m a
G e m i n a t e s followed by t a u t o m o r p h e m i c closed syllables [such as the first tt in
hottentotti; see (15)] m u s t then b e lexically p r e a s s o c i a t e d with the coda. C o n s e -
quently they will not u n d e r g o gradation. A s before, the possibility of such lexical
Blocking in Nonderived Environments 297

preassociation of g e m i n a t e s e l s e w h e r e than before t a u t o m o r p h e m i c closed syl-

lables w o u l d b e e x c l u d e d b y strict binarity (or m o r e precisely, by its a n a l o g for
nonfeatural i n f o r m a t i o n ) , w h i c h again raises the e m p i r i c a l issues d i s c u s s e d in
Section 3 . 1 .

5.2. C h u m a s h S i b i l a n t H a r m o n y

A n a p p a r e n t d e r i v e d - e n v i r o n m e n t effect in a s t r u c t u r e - c h a n g i n g rule of C h u -
m a s h , a l a n g u a g e formerly s p o k e n a l o n g the central California coast, h a s b e e n
p r e s e n t e d as a p r o b l e m for the present p r o p o s a l by P o s e r ( 1 9 9 0 ) . M y reanalysis
d r a w s on his fuller earlier t r e a t m e n t (Poser, 1982), and on his source, A p p l e -
gate ( 1 9 7 2 ) .
In C h u m a s h , a p r o c e s s of Sibilant H a r m o n y m a k e s sibilants and affricates agree
in laminality ([ ± anterior] in Poser, 1982, [ ± distributed] in Poser, 1990, and here)
with the r i g h t m o s t sibilant or affricate in the word. T h e rule is s t r u c t u r e - c h a n g i n g
and assimilates u n d e r l y i n g apicals to l a m i n a l s (53) as well as u n d e r l y i n g l a m i n a l s
to a p i c a l s .
20

(53) /ha+s+xintila/ hasxintila 'his Indian n a m e '

/ha+s+xintila+was/ hasxintiliwas 'his f o r m e r Indian n a m e '

(54) / p + i s + a l - f anan'/ pisannan' ' d o n ' t you t w o g o '

/s+is + sili+uluaqpey+us/ sis uleqpeyus
h
'they t w o w a n t to
follow it'

A n o t h e r rule of C h u m a s h , P r e - C o r o n a l L a m i n a l i z a t i o n ( P C L ) (56), m a k e s a
sibilant laminal i m m e d i a t e l y before a n o n s t r i d e n t c o r o n a l (t, /, n):

(55) /s+nan'/ man 'he goes'

/s+tepu?/ stepu? 'he gambles'
/s+loxit'/ sloxif ' h e surpasses m e '

(56) PCL: [+cor, +strid] -> [+distr] / [ + c o r , - strid]

Sibilants w h i c h b e c o m e laminal by P C L trigger Sibilant H a r m o n y , but they d o

not u n d e r g o it.

(57) / s + i s + lu + sisin/ sislusisin 'they t w o are g o n e a w r y '

/s+ti + yep+us/ stiyepus ' h e tells h i m '

T h e s e d a t a present an o r d e r i n g p a r a d o x w h i c h Poser ( 1 9 8 2 ) solves by d e -

c o m p o s i n g sibilant h a r m o n y into a s t r u c t u r e - c h a n g i n g delinking p r o c e s s and a
structure-building spread p r o c e s s , and o r d e r i n g P C L b e t w e e n t h e m .

(58) a. D i s t r i b u t e d D e l i n k i n g : delink [ ± distr] from all strident coronals e x c e p t

the last in a w o r d
b. P C L [rule (56)]
c. D i s t r i b u t e d S p r e a d i n g : spread [ ± distr] leftward
298 Paul Kiparsky

G i v e n this o r d e r i n g , l a m i n a l s from P C L e s c a p e D i s t r i b u t e d D e l i n k i n g , and so the

structure-building rule of D i s t r i b u t e d S p r e a d i n g c a n n o t apply to t h e m . U s i n g S to
s y m b o l i z e c o r o n a l s unspecified for laminality, w e then derive cases like (57)
as follows:

(59) / s + i s + l u + s i s i n / - > ( D i s t r i b u t e d D e l i n k i n g ) S+iS+lu+Sisin -> (PCL)

S+is+lu+Sisin —> (Distributed S p r e a d i n g ) sislusisin

N o w the p r o b l e m is that P C L exhibits N D E B effects. M o r p h e m e internally, a

contrast b e t w e e n s a n d s is p r e s e r v e d in the P C L e n v i r o n m e n t :

(60) a. stumukun 'mistletoe'

slow' ' e a g l e '
wastu? ' p l e a t '
b. wasti 'of a flow, of liquid in m o t i o n '

W h i l e s from P C L d o e s not u n d e r g o Sibilant H a r m o n y , s before t a u t o m o r p h e m i c

t in w o r d s like (60b) ( w h i c h o n P o s e r ' s analysis w o u l d b e u n d e r l y i n g s) d o e s
u n d e r g o Sibilant H a r m o n y :

(61) / s + w a s t i + l o k ' i n + u s / —> (Distributed D e l i n k i n g ) S+waSti+lok'in+us —>

( P C L ) N A —> s+wasti+lok'in+us ' t h e flow stops on h i m '

T h e contrast b e t w e e n (61) a n d (59) follows in Poser's analysis b e c a u s e P C L is

a s s u m e d not to apply in n o n d e r i v e d e n v i r o n m e n t s , a n d b e c a u s e the [ ± d i s t r ]
specifications in m o r p h e m e s like t h o s e in (60) ( w h i c h are fully m a r k e d in the
lexicon) are detected b y D i s t r i b u t e d D e l i n k i n g if a sibilant follows.
It m i g h t s e e m , then, that the a p p a r e n t derived e n v i r o n m e n t effect in the appli-
cation of P C L c a n n o t b e r e d u c e d to b l o c k i n g of a structure-building rule by speci-
fied representations. T h e r e a s o n is that, by the structure-changing operation of
Distributed D e l i n k i n g , all sibilants e x c e p t for the last in a w o r d get unspecified,
including both derived cases like (57), to w h i c h P C L applies, a n d m o r p h e m e -
internal cases like (60, 6 1 ) , to w h i c h it d o e s not apply. P C L therefore c a n n o t dis-
tinguish b e t w e e n t h e m on the basis of underspecification. B u t this contradicts o u r
proposal that the difference b e t w e e n derived and m o r p h e m e - i n t e r n a l cases is en-
c o d e d in the representation itself as a difference b e t w e e n the unspecified and the
specified value of the relevant feature.
M y idea can, however, b e sustained b y modifying Poser's analysis along the
following lines. S u p p o s e that a " s h a r e d feature c o n v e n t i o n " (Steriade, 1982)
m e r g e s adjacent p l a c e n o d e s if they are identical.

(62) c c c c

Place Place —• Place

Blocking in Nonderived Environments 299

S u p p o s e f u r t h e r m o r e that this m e r g e r is restricted to t a u t o m o r p h e m i c s e q u e n c e s ,

for r e a s o n s e x p l o r e d b y M c C a r t h y ( 1 9 8 6 ) . O n l y r e m a i n i n g s e q u e n c e s of c o r o n -
2 1

als with identical p l a c e n o d e s — i n practice, h e t e r o m o r p h e m i c o n e s — a r e subject

to P C L . 2 2

(63) c c

Place [+Distributed] /- Place

Coronal Coronal

[-Strident] [-Strident]

D i s t r i b u t e d S p r e a d i n g then applies iteratively leftward in structure-building m o d e .

(64) Coronal Coronal

[ + Strident] [ +Strident]

[ +Distributed]

In the illustrative derivation below, I s y m b o l i z e c o r o n a l s unspecified for [ ± dis-

tributed] b y S, T, a n d m a r k the result of a p p l y i n g the shared feature c o n v e n t i o n
(62) to a c o r o n a l cluster b y writing si.
(65) underlying wasT i S + wasTi+Lok 'iN+ uS S+iS+Lu+SiSiN
Distributed — S + waS T / + L o k 'iN+ uS
Delinking
shared feature — S + waSTi+Lok 'iN+ uS
convention

PCL —
— (blocked) S+is+Lu+SiSiN
Distributed — s+is+Lu + SiSiN
Spreading
Default wasti s+waTti+lok'in+us s+is+lu+sisin
O n this analysis, C h u m a s h is c o m p a t i b l e with m y c l a i m that N D E B occurs only
in n o n - s t r u c t u r e - c h a n g i n g r u l e s .
2 3

5.3. C h a m o r r o V o w e l L o w e r i n g

A n o t h e r " d e r i v e d - e n v i r o n m e n t " rule that apparently applies in structure-

c h a n g i n g fashion is the rule that g o v e r n s the distribution of high and m i d v o w e l s
in C h a m o r r o ( C h u n g , 1983).
300 Paul Kiparsky

(66) N o n l o w v o w e l s are realized as

a. e o in stressed closed syllables.
y

b. i, u e l s e w h e r e .

T h e following w o r d s illustrate the r e g u l a r distribution of h i g h and m i d v o w e l s .

(67) munddrjgu ' s t o m a c h of a c o w '

met gut ' s t r o n g '
pisaw 'fishing l i n e '

T h e shifting of stress b y the cyclic application of stress a s s i g n m e n t p r o d u c e s

systematic alternations in d e r i v e d f o r m s b e t w e e n high and m i d vowels, in accord
with ( 6 6 ) .2 4

(68) lapis lapessu ' ( y ) pencil'

m

hugdndu huganddnna '(his) p l a y i n g '

A l t h o u g h Vowel L o w e r i n g b y (66a) " a p p l i e s to the vast majority of n o n -

c o m p l e x w o r d s , including virtually all native C h a m o r r o w o r d s , " C h u n g records
" a handful of w o r d s " in w h i c h it u n a c c o u n t a b l y d o e s not apply. A l s o , " a rather
large class of l o a n s " h a v e low v o w e l s in stressed o p e n syllables, c o n t r a r y to (66b).
T h e s e t w o classes of cases w h e r e the n o r m a l distribution of vowel height is b r o k e n
in n o n d e r i v e d e n v i r o n m e n t s are illustrated in (69).

(69) a. hungan ' y e s ' , asut ' b l u e ' , listu ' q u i c k '
( [ + h i g h ] w h e r e [ — high] is e x p e c t e d by (66a))
b. neni ' b a b y ' , ispeyus ' m i r r o r ' , sitbesa ' b e e r '
([ — high] w h e r e [ + h i g h ] is e x p e c t e d b y (66b))

O n m y a s s u m p t i o n s , the lexical representations of these m o r p h e m e s w o u l d h a v e

[ + h i g h ] and [ — high] v o w e l s respectively, w h i l e t h o s e of the m o r p h e m e s in (67)
w o u l d h a v e [Ohigh] v o w e l s , w h i c h w o u l d b e specified as [ + h i g h ] a n d [ — high] in
the a p p r o p r i a t e c o n t e x t s b y (66).
Is (66) structure-building or s t r u c t u r e - c h a n g i n g ? F r o m w h a t w a s said in the
p r e c e d i n g p a r a g r a p h , it s e e m s that it m u s t b e structure-building, so it d o e s not
lower the specified high v o w e l s in (69a) or raise the specified m i d vowels in
(69b). B u t it also s e e m s that it m u s t b e structure-changing, for it d o e s raise the
specified o p e n syllable m i d v o w e l s in (69b) w h e n the stress shifts off t h e m in
suffixed forms.

(70) neni ' b a b y ' , ninina 'his b a b y '

W e s e e m to h a v e arrived at a contradiction.
T h e p r o b l e m is only apparent. T h e height alternation is the result of t w o sepa-
rate rules, a structure-building l o w e r i n g rule (with [ + h i g h ] default) conditioned
b y syllable structure, and a s t r u c t u r e - c h a n g i n g raising rule c o n d i t i o n e d by stress.
Blocking in Nonderived Environments 301

(H)) [ [ - h i g h ] in closed s y l l a b l e s ] , . ,
[ - l o w ] -> < \ " \ (non-structure-changing)
[ [ + h i g h ] (elsewhere) J

(72) [ — low] —> [ + h i g h ] in unstressed syllables (structure-changing)

If these rules apply to the p r o p o s e d lexical representations, b o t h the height alter-

nations a n d the distribution of u n p r e d i c t a b l e height are fully a c c o u n t e d for. If
cyclic, these rules i m p l y that [ ± high] is distinctive only in stressed syllables, a n d
consequently, that N D E B effects with respect to L o w e r i n g o c c u r only in stressed
o p e n syllables.

(73) l\l Id
Stressed c l o s e d syllable [+high] [Ohigh]
Stressed o p e n syllable [Ohigh] [ — high]
U n s t r e s s e d syllable [Ohigh] —

T h e prediction, w h i c h a p p e a r s to b e correct, is that there are n o w o r d s with alter-

nations like hugdndu ~ *hdgand6niia, as o p p o s e d to the t y p e in (70).
T h e d e c o m p o s i t i o n required by the underspecification a c c o u n t of the C h a m o r r o
vowel alternations is thus confirmed. N o t e also that strict binarity is again
preserved.

6. N D E B E F F E C T S I N P R O S O D I C R U L E S

6.1. Stress: Unifying F E C and N D E B

B y attributing the b l o c k i n g of rules in n o n d e r i v e d e n v i r o n m e n t s to their n o n -

s t r u c t u r e - c h a n g i n g status, w e r e o p e n the question of h o w to unify N D E B a n d
"strict c y c l e " effects with the free e l e m e n t c o n d i t i o n ( F E C ) (Prince, 1985).

(74) R u l e s of p r i m a r y metrical analysis d o not affect existing m e t r i c a l structure.

Let us c o n s i d e r the distinction b e t w e e n rules of p r i m a r y metrical analysis and

rules of metrical reanalysis to b e a special c a s e of the m o r e general d i c h o t o m y
b e t w e e n structure-building a n d s t r u c t u r e - c h a n g i n g rules. T h i s w o u l d m e a n that
F E C d e p e n d s on the s a m e p r o p e r t y that I h a v e j u s t a r g u e d to be relevant for
N D E B . Let us also a s s u m e as before that structure-building is the basic m o d e of
rule application a n d that s t r u c t u r e - c h a n g i n g rule application is the c o m b i n e d result
of deletion and structure-building ( w h e r e the deletion c o m p o n e n t of this c o m b i -
nation c o u l d be either a deletion rule or a filtering/projection m e c h a n i s m a la Halle
a n d Vergnaud, 1987). T h e fact that rules of m e t r i c a l reanalysis typically " r e a s -
s e r t " the l a n g u a g e ' s basic foot structure (Prince, 1985) i n d e p e n d e n t l y confirms
302 Paul Kiparsky

the principle that requires this d e c o m p o s i t i o n , since it m e a n s that they m u s t be

v i e w e d a n y w a y as c o n s i s t i n g of destressing plus reapplication of the rules of pri-
m a r y metrical analysis.
F E C effects c o m e in t w o sharply different t y p e s , strong a n d w e a k . S t r o n g F E C
effects, d o c u m e n t e d for G r e e k (Steriade, 1988), Diyari and Warlpiri (Poser, 1986,
1989), M a n a m and A r a b i c (Halle a n d K e n s t o w i c z , 1990), a n d C h o c t a w ( L o m -
bardi a n d M c C a r t h y , 1990), involve absolute inviolability of all previously as-
signed metrical constituency. W e a k F E C effects, seen in E n g l i s h (Kiparsky, 1982),
C h a m o r r o ( C h u n g , 1983), I n d o n e s i a n ( C o h n , 1989), a n d D a k o t a (Shaw, 1985),
involve the b l o c k i n g of m e t r i c a l structure r e a s s i g n m e n t in d o m a i n s w h i c h are al-
r e a d y footed in their entirety (as in standardization), w h i l e already footed syl-
lables d o get reassociated with as yet unfooted syllables (as in solidify, original).
It w a s the w e a k F E C effects w h i c h w e r e assimilated to the S C C in K i p a r s k y
( 1 9 8 2 ) , a n d m o r e recently K a g e r ( 1 9 8 9 ) h a s p r o p o s e d r e f o r m u l a t i n g the F E C itself
so as to a c c o u n t for t h e s e c a s e s . B u t K a g e r ' s w e a k e n e d version of the F E C has
2 5

n o t h i n g to say a b o u t strong F E C effects, w h e r e foot b o u n d a r i e s c o n s t r u c t e d on

previous cycles are entirely t a b o o for later foot construction.
T h e right m o v e , then, w o u l d b e to take the strong F E C as representing the basic
N D E B s y n d r o m e in the d o m a i n of stress. T h e w e a k F E C is the result of c o m p o s -
ing stress a s s i g n m e n t with partial deletion of metrical structure, specifically of
b r a c k e t i n g (i.e., deforestation with retention of p r o m i n e n c e m a r k e d by metrical
grids). P r o p o s a l s a l o n g t h e s e lines h a v e in fact b e e n m a d e b y H a l l e a n d V e r g n a u d
( 1 9 8 7 ) , P o s e r ( 1 9 8 9 ) , Steriade ( 1 9 8 8 ) , a n d H a l l e a n d K e n s t o w i c z (1990) (though
they differ substantially on h o w to d o this, especially on the conditions u n d e r
w h i c h metrical structure from p r e v i o u s cycles is retained a n d w h e n it is c h a n g e d ) .
Accordingly, in the E n g l i s h / C h a m o r r o / D a k o t a t y p e of stress a s s i g n m e n t there is
deforestation in addition to the a s s i g n m e n t of metrical structure. T h e strong F E C
prohibits rules from c h a n g i n g existing metrical c o n s t i t u e n c y or p r o m i n e n c e rela-
tions ( w h e t h e r t h e s e originate in the lexicon or t h r o u g h the application of earlier
stress rules). Stress a s s i g n m e n t is thus structure-building in the strictest sense. If
m y p r o p o s a l is right, this m e a n s that there is n o "strict c y c l i c i t y " in stress rules
other than w h a t results from the F E C itself, a n d that the F E C is a special case
of N D E B . 2 6

If w e a k F E C effects are the result of deforestation, then they w o u l d in effect

r e d u c e to a b a n on " p r o m i n e n c e - c h a n g i n g . " T h i s requires a definition of w h a t it
m e a n s to c h a n g e p r o m i n e n c e . W e m u s t define a stress a s s i g n m e n t as p r o m i n e n c e -
c h a n g i n g if and only if its input a n d output h a v e i n c o m p a t i b l e p r o m i n e n c e rela-
tions, w h e r e relative p r o m i n e n c e is defined b y grid height in the usual w a y . T h e 2 7

basic interpretation of grid h e i g h t tells u s , for e x a m p l e , that (75) asserts the p r o m i -

n e n c e relations (76).

(75) * *

(a){(bc)(def))
Blocking in Nonderived Environments 303

(76) 1. a, b, d, > c, e,f

2. c=e=f
3. a,b> d

W e m u s t further a s s u m e that relative p r o m i n e n c e a m o n g e l e m e n t s at the t o p

grid level is undefined. A s s i g n i n g s u p e r o r d i n a t e metrical structure to the top grid
level therefore d o e s not assign i n c o m p a t i b l e p r o m i n e n c e relations and is thereby
structure-building, as d e s i r e d . 28
F o r instance, (75) d o e s not assert (77), b u t simply
leaves the relative p r o m i n e n c e of a and b unspecified.

(77) a = b

A l s o , m e t r i c a l l y u n a s s o c i a t e d (unfooted) m a t e r i a l m u s t not h a v e any p r o m i n e n c e

relations defined to it, so that a d d i n g m e t r i c a l structure to such e l e m e n t s is not a
s t r u c t u r e - c h a n g i n g operation either.
G i v e n t h e s e c o n v e n t i o n s , foot c o n s t r u c t i o n across a previously built foot is
b l o c k e d , as in the third c y c l e of standardization. 29

{start dard) (iz) + a (tion) -h *(stan) (dar diz) (a) (tion)

T h e E n g l i s h Stress R u l e is correctly b l o c k e d here, since it w o u l d reverse the

p r o m i n e n c e relations in -dardize-. B u t it is free to apply in the s e c o n d c y c l e of
solidity a n d original, as desired.

(so) lid + i (ty) -> (so) (li di) (ty)

(ori) gin + (al) —> (o) (ri gin) (al)

In b o t h t h e s e cases, the stress rule is free to apply on the third cycle b e c a u s e it

d o e s not c h a n g e any of the p r o m i n e n c e relations defined on the first, in the sense
j u s t defined.
K a g e r ( 1 9 8 9 : 1 2 7 f f . ) h a s a r g u e d that t w o of the E n g l i s h destressing rules, S o -
norant D e s t r e s s i n g and the A r a b R u l e , actually b e l o n g in the s u b s y s t e m of p r i m a r y
metrical analysis rules, in that they are o r d e r e d a m o n g t h e m , a n d like t h e m , are
structure-building ( h e n c e subject to the F E C ) . F r o m this h e p r o p o s e s to derive
their restriction to n o n b r a n c h i n g feet. W e retain this result, for destressing of the
disyllabic foot in (80a) is b l o c k e d b e c a u s e p r o m i n e n c e relations are c h a n g e d on
-nonga-, b u t destressing of the m o n o s y l l a b i c foot in (80b) is p e r m i t t e d b e c a u s e n o
p r o m i n e n c e relations are c h a n g e d .

(80) a . * * * * *
(Mo) (non ga) (he) (la) -h (Monon ga) (he) (la)
* * * * *
(Ha) (cken) (sack) —> (Hacken) (sack)
304 Paul Kiparsky

6.2. Q u a n t i t y

Since N D E B effects with respect to quantity exist, a vowel or syllable m u s t b e

capable of b e i n g lexically m a r k e d as distinctively short in e n v i r o n m e n t s w h e r e
l e n g t h e n i n g w o u l d apply to it, a n d c o n v e r s e l y as distinctively long in environ-
m e n t s w h e r e shortening w o u l d apply to it. Intrinsically privative representations
are e x c l u d e d for length j u s t as for all other p h o n o l o g i c a l p r o p e r t i e s . Ito (1990)
reports that J a p a n e s e i m p o s e s a t w o - m o r a m i n i m a l foot r e q u i r e m e n t on derived
w o r d s such as (81).

(81) a. T r u n c a t i o n s : choko ' c h o c o l a t e '

b. R e d u p l i c a t e d f o r m a t i o n s (two feet): mi 'look' —> miimii 'while
looking'

But it also has n u m e r o u s m o n o m o r a i c s t e m s like (82), w h i c h , unless suffixed,

surface as w o r d s consisting of d e g e n e r a t e feet.

(82) su ' v i n e g a r ' , ta 'field'

A s Ito points out, such stems m u s t b e lexically specified as short. Lexically speci-
fied s u b m i n i m a l feet will not b e l e n g t h e n e d t h r o u g h the structure-building assign-
m e n t of the canonical b i m o r a i c foot. F o r instance, a s s u m i n g a m o r a i c representa-
tion of length, su w o u l d h a v e the u n d e r l y i n g representation (83),

(83) F

su
w h i c h e s c a p e s the m i n i m a l foot r e q u i r e m e n t in virtue of its lexically specified
metrical structure. A s t e m c o n f o r m i n g to t h e canonical pattern, such as suu ' n u m -
b e r ' , has n o metrical structure in the lexicon a n d gets assigned a full t w o - m o r a
foot b y the o r d i n a r y p r o s o d i c p a r s i n g p r o c e s s e s o p e r a t i n g o n / s u / . Finally, if
truncation and reduplication s u p e r i m p o s e their o w n foot patterns on the lexically
specified p r o s o d i c structure of the s t e m (on the principle that affix properties
s u p e r s e d e stem properties), they will effectively neutralize u n d e r l y i n g lexical
quantity.

6.3. D i s h a r m o n y

T h e S C C h a s b e e n used to a c c o u n t for idiosyncratically o p a q u e s e g m e n t s in

tone and h a r m o n y p r o c e s s e s . T h e idea is to h a v e the S C C b l o c k the m o r p h e m e -
Blocking in Nonderived Environments 305

internal spread of a h a r m o n i c feature lexically associated to a s e g m e n t (Arch- 3 0

angeli a n d P u l l e y b l a n k , 1989; L e v e r g o o d , 1984; Pulleyblank, 1982; Steriade,

1987a). For instance, A r c h a n g e l i and P u l l e y b l a n k p r o p o s e for Yoruba a right-to-
left s p r e a d i n g of [ — A T R ] . If p r e s e n t as a floating feature in the lexical r e p r e s e n -
tation, [ — A T R ] spreads m a x i m a l l y over all eligible v o w e l s , e.g., ekq ' p a p ' .

(84) e k o — + ~ e k o - + e k o

-A -A -A

A d i s h a r m o n i c w o r d such as telq 'tailor' is treated as h a v i n g its last vowel lexi-

cally p r e a s s o c i a t e d with [ — A T R ] .

(85) t e I o —>- t e I o tailor

-A +A-A

Leftward s p r e a d i n g of the p r e l i n k e d a u t o s e g m e n t in (85) is taken to b e b l o c k e d by

the S C C , a n d the u n a s s o c i a t e d v o w e l of the initial syllable is later a s s i g n e d the
default specification [ + A T R ] , b y the s a m e rule w h i c h assigns it to regular har-
m o n i c [ + A T R ] w o r d s . W e will w a n t to retain the idea that d i s h a r m o n y results
from there b e i n g m o r e than o n e h a r m o n i c span a n d that v o w e l s not r e a c h e d b y
h a r m o n y get a s s i g n e d their u n m a r k e d values of the h a r m o n i c feature b y default.
B u t clearly this i m p l e m e n t a t i o n of it will not w o r k w i t h o u t a specific principle
such as the S C C , so it r e p r e s e n t s a potential p r o b l e m for the proposal to explain
away N D E B effects.
T h e r e is, however, an alternative a p p r o a c h to d i s h a r m o n y , w h i c h p r e s e r v e s the
basic insight yet d o e s not require stipulating b l o c k i n g in n o n d e r i v e d e n v i r o n -
m e n t s . This is to m a r k off h a r m o n i c spans directly (see Hulst a n d S m i t h , 1986, for
o n e t r e a t m e n t of this k i n d ) . In the n o r m a l case, the h a r m o n i c span is p r e d i c t a b l e ;
it is typically the p h o n o l o g i c a l w o r d . A regular w o r d like ekq constitutes o n e
h a r m o n i c span, over w h i c h the h a r m o n i c a u t o s e g m e n t spreads m a x i m a l l y . T h e
exceptional nature of d i s h a r m o n i c w o r d s such as telq is that they are m a r k e d as
h a v i n g t w o h a r m o n i c spans. S p r e a d i n g then fails and [ + A T R ] is a s s i g n e d by
default.

(86) te ' lo te ~ lo

— |
-A +A -A

A n a r g u m e n t in favor of this a p p r o a c h is that even w o r d s w h i c h are not dishar-

m o n i c can fall into separate h a r m o n i c spans. F o r e x a m p l e , in Finnish, w h e r e /, e
306 Paul Kiparsky

are neutral, long w o r d s with b a c k v o w e l s followed b y several neutral v o w e l s , such

as karamelli, nevertheless often take front suffixes.

(87) a. vari+a ' c o l o r ' (part.sg.)

b. pari+a ' p a i r ' (part.sg.)
c. karamelli+a ~ karamelli+a ' p i e c e of c a n d y ' (part.sg.)

S i n c e /, e are neutral, these s t e m s are not intrinsically d i s h a r m o n i c , and so there

is n o lexically specified b a c k n e s s feature to p r e a s s o c i a t e . A d d i n g [ — b a c k ] to o n e
of the neutral v o w e l s for the sake of h a v i n g s o m e t h i n g to preassociate w o u l d entail
sacrificing a major generalization of F i n n i s h p h o n o l o g y , that the neutral v o w e l s
are unspecified for b a c k n e s s . B u t the variation in these w o r d s is readily captured
b y a l l o w i n g t h e m to consist of t w o h a r m o n i c spans. It s e e m s that such w o r d s
n o r m a l l y consist of t w o stress feet, so that the generalization m a y simply b e that
the h a r m o n i c span in long s t e m s is either the w o r d or the f o o t . 31

(88) a. [karamelli + a]
b. [kara] [melli + d\

Additional justification for this t r e a t m e n t of d i s h a r m o n y can b e found in a lan-

g u a g e like T u r k i s h , w h i c h h a s b o t h r o u n d i n g h a r m o n y and b a c k n e s s h a r m o n y .
T h e a s s u m p t i o n of h a r m o n y spans m a k e s sense out of the fact that r o u n d i n g and
b a c k n e s s d i s h a r m o n y never o c c u r i n d e p e n d e n t l y of e a c h other in T u r k i s h . T h e
solidarity of r o u n d i n g a n d b a c k n e s s w o u l d o t h e r w i s e b e p u z z l i n g since the t w o
features spread b y separate p r o c e s s e s , as w e can tell from the fact that n o n h i g h
v o w e l s b l o c k the spread of r o u n d i n g b u t not the spread of b a c k n e s s . If r o u n d i n g
and b a c k n e s s d i s h a r m o n y are d u e to lexical linking of the respective features
[round] a n d [back] to s o m e v o w e l , then a hypothetical w o r d of the n o n e x i s t i n g
type w h e r e j u s t o n e of t h e s e features fails to spread c o u l d b e easily r e p r e s e n t e d b y
lexically linking j u s t that o n e feature. T h e result w o u l d b e a r a n g e of d i s h a r m o n i c
vowel s e q u e n c e s w h i c h in fact d o not exist in the l a n g u a g e . F o r e x a m p l e , since
spreading in T u r k i s h is left to right, in a w o r d of the f o r m . . . u the
feature [back] w o u l d b e floating a n d the feature [round] lexically linked to the first
vowel, and in a w o r d of the f o r m . . . . . . . u . . . it w o u l d b e the other w a y r o u n d .
If d i s h a r m o n y is dealt with b y delimiting h a r m o n i c spans as h e r e advocated, then
the systematic a b s e n c e of such vowel c o m b i n a t i o n s in T u r k i s h is a c c o u n t e d for by
the straightforward a s s u m p t i o n that T u r k i s h spreads r o u n d i n g a n d b a c k n e s s in the
s a m e d o m a i n . T h i s w o r k s for all actually o c c u r r i n g d i s h a r m o n i c vowel c o m b i -
3 2

nations in the T u r k i s h v o c a b u l a r y listed b y C l e m e n t s and Sezer (1983). For ex-

a m p l e , a w o r d like muzip ' m i s c h i e v o u s ' consists of t w o h a r m o n i c spans, the first
associated with [ + r o u n d ] a n d [ + b a c k ] , the s e c o n d u n a s s o c i a t e d and assigned
[ - r o u n d ] , [ - b a c k ] by default.
T h e r e is yet a n o t h e r r e a s o n w h y prelinking with S C C - i n d u c e d b l o c k i n g of
Blocking in Nonderived Environments 307

spread is not the right a p p r o a c h to d i s h a r m o n y . T h i s is that it p r e c l u d e s the p h o -

n o l o g y from e x p r e s s i n g the generalization, w h i c h h o l d s for T u r k i s h j u s t as for any
other l a n g u a g e , that [ + r o u n d ] v o w e l s are [ + b a c k ] in the u n m a r k e d c a s e . T h e
v o w e l u in a m o n o s y l l a b i c w o r d w o u l d b e lexically m a r k e d as [ + r o u n d ] a n d as-
signed the feature specification [ + b a c k ] b y the universal default rule [ + r o u n d ] —>
[ + b a c k ] . T h e s a m e should b e true for the u in a d i s h a r m o n i c w o r d like muzip. B u t
u n d e r the S C C t r e a t m e n t of d i s h a r m o n y , this c a n n o t b e d o n e , b e c a u s e b o t h
[ + r o u n d ] a n d [ + b a c k ] w o u l d h a v e to b e linked lexically in o r d e r to prevent
t h e m from s p r e a d i n g b y h a r m o n y . In this way, the S C C forces us to a b a n d o n the
underspecification analysis that other c o n s i d e r a t i o n s d e m a n d . U n d e r the span-
delimitation solution, however, the underspecification analysis is available.
In addition, t h e latter solution affords a p r i n c i p l e d a c c o u n t of C l e m e n t s and
S e z e r ' s generalization that only the five v o w e l s /, e, a, o, u m a y b e o p a q u e in
T u r k i s h . T h i s is of c o u r s e j u s t the s u b s y s t e m of v o w e l s w h i c h is u n m a r k e d for
b a c k n e s s . F o r t h e other three v o w e l s , *, d, u, t h e v a l u e of t h e feature [back] is not
p r e d i c t a b l e from the features [high] a n d [ r o u n d ] . In o r d e r to derive this distribu-
tion of v o w e l s w e n e e d only a s s u m e , in addition, that the feature [back] is lexically
distinctive only in initial s p a n s . T h i s restricts t h e v o w e l s i, b\ u to the initial ( p o s -
sibly o n l y ) h a r m o n i c span of the w o r d .
I c o n c l u d e that opacity a n d d i s h a r m o n y d o not p r o v i d e e v i d e n c e for t h e S C C
or i n d e e d for any p r i n c i p l e w h i c h b l o c k s the application of rules in n o n d e r i v e d
environments.

7. CONCLUSIONS

P r e v i o u s t r e a t m e n t s of N D E B h a v e not s u c c e e d e d in r e c o n c i l i n g e m p i r i c a l c o v -
e r a g e with theoretical adequacy. T h e R A C p r o v i d e s a descriptively fairly a c c u r a t e
c i r c u m s c r i p t i o n of t h e p h e n o m e n o n , e x c e p t that formulating t h e constraint as a
categorical prohibition of absolute neutralization is p r o b a b l y too strong. T h e R A C
is, however, clearly unsatisfactory as a p r i n c i p l e of g r a m m a r , b e c a u s e of t h e for-
m a l indefinability of the class of " n e u t r a l i z a t i o n r u l e s , " the undesirability of hav-
ing different p r i n c i p l e s for obligatory a n d optional rules, a n d the d u b i o u s status of
the c o n c e p t of " d e r i v e d e n v i r o n m e n t . " T h e S C C , on the other h a n d , is preferable
o n g e n e r a l theoretical g r o u n d s b u t s i m p l y fails t o m a t c h t h e facts in m a n y specific
instances. In this article I h a v e p r e s e n t e d a n e w interpretation w h i c h resolves this
d i l e m m a . It m a k e s essential u s e of underspecification and of d e c o m p o s i t i o n of
s t r u c t u r e - c h a n g i n g rules. Its m a i n a d v a n t a g e s are that it r e d u c e s the b l o c k i n g ef-
fects to i n d e p e n d e n t p r i n c i p l e s of g r a m m a r , predicts their restriction to structure-
building rules a n d to n o n v a c u o u s rule applications, recaptures the e m p i r i c a l
generalizations b e h i n d the original restriction of the alternation condition to
308 Paul Kiparsky

obligatory neutralization rules, and correctly s u b s u m e s the F E C of p r o s o d i c

phonology.

ACKNOWLEDGMENTS

I would like to thank Young-mee Cho for her helpful advice at several stages in the
preparation of this article. It has also benefited from the pointed and constructive comments
of Ellen Kaisse, John McCarthy, and an anonymous reviewer. I take responsibility for any
remaining errors.

NOTES

1
See also Giegerich (1988) for a different proposal to this effect.
2
Recently, Hammond (1991) has also argued for an acquisition-based account of SCC
effects, based on a "morphological WYSIWYG" principle. I hope to return to Hammond's
proposal on another occasion.
3
Consonant Gradation, as well as other cases which establish the same general point, is
discussed at greater length below.
4
Unless otherwise stated, I assume the analysis of Finnish word phonology presented in
Keyser and Kiparsky (1984), complemented by the morphological analysis of Kanerva
(1987). The reader is referred to those sources for the evidence supporting the rules and
underlying forms which figure in the discussion below. I should add that I know of no
reason to believe that these rules might belong to different levels of the lexical phonology,
or indeed, that Finnish even has more than one level.
5
The -h- is obligatory after a primary stress and after a long vowel, e.g., maa+han 'into
the ground', mi+hin 'into what', veesee+hen 'into the toilet' (words of the latter type,
with noncontracted unstressed long vowels, are loans, and they are often treated as con-
tracted in nonstandard speech, e.g. veesee+seen). Otherwise, - h- is deleted, except in ar-
chaic, poetic, or dialectal usage.
6
T h e deletion cannot be cyclic because it it is blocked by any following suffix, e.g.,
/vaatteCe+n/ —» vaatteen.
7
Detailed motivation for this analysis is presented in Keyser and Kiparsky (1984).
8
The reader points out that these data could be reconciled with cyclic application of
Consonant Gradation on the assumption that final consonants are cyclically extrametrical,
and that rule (8) is ordered before it. Even on these assumptions, Consonant Gradation
would have to apply at the word level as well (and extrametricality would have to assumed
be turned off there), because of such cases as (16) and (49). Therefore, this alternative does
not call into question the status of Consonant Gradation as a word-level rule subject
to NDEB.
9
T h e data in (18) represent the actual pronunciation. Postlexical gemination is normally
Blocking in Nonderived Environments 309

not shown in the spelling (menepds, itsekin). For the evidence that clitics like -kin are
syntactic elements and not attached lexically see Kanerva (1987).
1 0
On the Zee/Inkelas assumptions mentioned above, so is application across clitic
boundary. Moreover, on those assumptions both the fact that the ruki rule stopped applying
across clitic boundary and the fact that it stopped applying across word boundary in clas-
sical Sanskrit can be seen as consequences of a single change, namely, that the rule became
confined to the lexical phonology. This development represents a characteristic trajectory
of phonological rules (Zee, this volume).
1 1
As shown by the failure of sandhi gemination, and by forms like gen.sg. / n u k k e + n /
nuke+n (not /nukkeCe+n/ *nukkee+n) and part.sg. /nukke+ta/ nukke+a (not /nukke-
C e + t a / *nuket+ta), there is no protective "ghost consonant" after them, as in (16) and in
(50) below.
1 2
T h e analysis is similar to that of English palatalization, which applies both lexically
and postlexically, but with the lexical palatals undergoing (word-level?) spirantization, e.g.,
/edition/ [s] vs. /hit you/ [c] (Borowsky, 1986).
1 3
Similar examples can be given for other endings beginning with -/.
1 4
It is specifically the first vowel which is truncated, as shown by sequences of dissimi-
lar vowels, e.g., /tunte+i + vat/ tunsivat 'they knew', /vaati+eCe/ vaade 'demand'.
1 5
John McCarthy points out that even simple cases like /koti/ koti should constitute
vacuously derived environments for t —> s, since raising would apply vacuously to the final
vowel.
1 6
It is immaterial for the point whether Stress Assignment is done along the lines pro-
posed by Harris (1983, 1987) for Spanish, or following the somewhat different analysis
developed by Alsina.
1 7
Actually, this rule is formulated more generally as applying to the low vowels D, £ as
well, but in Mascaro's analysis these get raised to mid vowels in unstressed position any-
way by the second Vowel Reduction rule, which brooks no exceptions whatever.
1 8
A similar case of vacuously derived environments apparently counting as derived for
the SCC was noted by Steriade (1987b note 23) for Yakan (data from Behrens, 1973:25).
Here unstressed a is raised to / in derived stems; on a Mascarovian account, the latter
condition could be eliminated.
1 9
Note also that if Consonant Gradation is not deletion but lack of association, it auto-
matically follows that gradated consonants cannot undergo /-Deletion [rule (3)], in such
words as hatu + ton. As John McCarthy has pointed out to me, ordering /-Deletion before
Consonant Gradation would be problematic because the structural description of t-Deletion
crucially refers to a combined prosodic and melodic context, which is not available until
after Consonant Gradation, on the analysis suggested here.
2 0
Geminate sibilants are realized as s , s .
h h

2 1
Some cases which at first blush seem to be exceptions to PCL could, if the proposed
analysis is on the right track, be viewed as exceptionally undergoing the shared feature
convention in the lexical phonology in spite of being heteromorphemic: s+lu+skumu 'it
branches into four', s+netus 'he does it to him' (Applegate 1972:120).
2 2
O u r formulation supposes that [distributed] is dependent on [strident].
2 3
Applegate has many other examples of s ~ s alternations in wasti-type, words (1972:
347, 351, 352, 367, 375), but none that clearly involve vmy/w?-type words. The derivation
/ha p+xoslo? + s/ —> apxoslos 'you blow your nose' is given on p. 522, but no */xoslo/ is
310 Paul Kiparsky

actually cited (is / h a + s + n o x s / asnoxs 'nose' somehow related?). The analysis I have pro-
posed would actually predict that underlying /st/ does not undergo Sibilant Harmony, since
the multiple link established by the shared feature convention would block spreading.
2 4
Chung describes a pattern of optionality in the application of the high-mid alternation
in vowels with secondary stress, which is of great theoretical interest but not directly
relevant here.
25
K a g e r ' s version of the FEC furthermore covers cases like original(-ity), which were
problematic for the SCC account.
2 6
From this perspective, the distinction between strong and weak FEC effects is akin to
that between stress-neutral and stress-dominant affixes. It is no accident, then, that clitic-
triggered stress-assignment always displays strong FEC effects (Arabic, Manam, Greek),
and that weak FEC effects are associated with stress subordination.
2 7
A similar proposal was made by Bruce Hayes in a talk presented at a phonology work-
shop at the University of Massachusetts, Amherst, in 1983, of which I have seen only the
handout.
28
C o n t r a Harris (1989:358).
2 9
Adopting the notation of Kager (1989), I delimit feet by parentheses and extrametrical
elements by angled brackets.
3 0
With the no-crossing constraint blocking any spread across it.
3 1
Finnish linguists sometimes characterize such long words as "quasi-compounds."
There is no morphological support for this, in cases like (88) at least, but the intuition
nevertheless has a real basis, to which my proposal also does justice, in that compounds
similarly have at least two stress feet.
3 2
This would be the unmarked situation, though different harmonic processes could
potentially be restricted to different spans as well, a possibility which is apparently realized
in some Turkic languages.

REFERENCES

Alsina, A. (1987). Stress Assignment in Catalan. Unpublished manuscript, Stanford

University.
Anderson, S. (1969). An outline of the phonology of Modern Icelandic vowels. Founda-
tions of Language 5, 5 3 - 7 2 .
Anderson, S. (1981). Why phonology isn't "natural." Linguistic Inquiry 1 2 , 4 9 3 - 5 3 9 .
Applegate, J. (1972). Ineseno Chumash Grammar. Doctoral dissertation, University of Cal-
ifornia, Berkeley.
Archangeli, D., and Pulleyblank, D. (1989). Yoruba vowel harmony. Linguistic Inquiry 20,
173-217.
Behrens, D. (1973). Yakan Phonemics and Morphophonemics (Papers in Philippine Lin-
guistics 7). Australian National University, Canberra.
Borowsky, T. (1986). Topics in English Phonology. Doctoral dissertation, University of
Massachusetts, Amherst.
Blocking in Nonderived Environments 311

Cho, Y. (1990). Parameters of Consonantal Assimilation. Doctoral dissertation, Stanford

University, Stanford, Calif.
Cho, Y., and Sells, P. (1991). A Lexical Account of Phrasal Suffixes in Korean. Unpublished
manuscript, Stanford University, Stanford, Calif.
Chomsky, N., and Halle, M. (1968). The Sound Pattern of English. Harper & Row, New
York.
Chung, S. (1983). Transderivational relationships in Chamorro phonology. Language 59,
35-66.
Clark, M. (1990). The Tonal System oflgbo. Foris, Dordrecht.
Clements, G. N. (1989). A Unified Set of Features for Consonants and Vowels. Unpub-
lished manuscript, Cornell University, Ithaca, N.Y.
Clements, G. N., and Sezer, E. (1983). Vowel and consonant disharmony in Turkish. In
The structure of Phonological Representations (H. van der Hulst and N. Smith, eds.)
vol. 2, pp. 2 1 3 - 2 5 5 . Foris, Dordrecht.
Cohn, A. (1989). Stress in Indonesian and bracketing paradoxes. Natural Language and
Linguistic Theory 7, 167-216.
Dresher, E. (1981). On the learnability of abstract phonology. In The Logical Problem of
Language Acquisition (C. L. Baker and J. McCarthy, eds.). MIT Press, Cambridge,
Mass.
Giegerich, H. J. (1988). Strict cyclicity and elsewhere. Lingua 75, 1 2 5 - 1 3 4 .
Halle, M. (1978). Formal vs. functional considerations in phonology. In Festschrift for O.
Szemerenyi (B. Brogyanyi, ed.). Benjamins, Amsterdam.
Halle, M., and Kenstowicz, M. (1990). The Free-Element Condition and Cyclic vs. Non-
cyclic Stress. Unpublished manuscript, Massachusetts Institute of Technology,
Cambridge.
Halle, M., and Vergnaud, J.-R. (1987). An Essay on Stress. MIT Press, Cambridge, Mass.
Hammond, M. (1991). Deriving the Strict Cycle Condition. Unpublished manuscript, Uni-
versity of Arizona, Tucson.
Hargus, S. (1985). The Lexical Phonology of Sekani. Doctoral dissertation, University of
California, Los Angeles.
Hargus, S. (1989). Underspecification and derived-only rules in Sekani phonology. In
Theoretical Perspectives on Native American Languages (D. B. Gerdts and K. Mi-
chelson, eds.), pp. 7 0 - 1 0 3 . State University of New York Press, Albany.
Harris, J. (1983). Syllable Structure and Stress in Spanish: a Nonlinear Analysis. MIT
Press, Cambridge, Mass.
Harris, J. (1987). The accentual pattern of verb paradigms in Spanish. Natural Language
and Linguistic Theory 5, 6 1 - 9 0 .
Harris, J. (1989). The stress erasure convention and cliticization in Spanish. Linguistic
Inquiry 20, 3 3 9 - 3 6 3 .
Hayes, B. (1986). Inalterability in CV phonology. Language 62, 3 2 1 - 3 5 2 .
Hualde, J. I. (1989). The strict cycle condition and noncyclic rules. Linguistic Inquiry 20,
675-680.
Hulst, H. van der, and Smith, N. (1986). On neutral vowels. In The Phonological Represen-
tation of Suprasegmentals (H. van der Hulst and M. Mous, eds.), pp. 2 3 3 - 2 7 9 . Foris,
Dordrecht.
312 Paul Kiparsky

Hume, E. (1990). Front vowels, palatal consonants, and the rule of Umlaut in Korean.
Proceedings of the North Eastern Linguistic Society 21.
Inkelas, S. (1989). Prosodic Constituency in the Lexicon. Doctoral dissertation, Stanford
University, Stanford, Calif.
Ito, J. (1986). Syllable Theory in Prosodic Phonology. Doctoral dissertation, University of
Massachusetts, Amherst.
Ito, J. (1990). Prosodic minimality in Japanese. Syntax Research Center, Cowell College,
University of California, Santa Cruz, Technical Report, SRC-90-04.
Iverson, G. K., and Wheeler, D. W. (1988). Blocking and the elsewhere condition. In Theo-
retical Morphology (M. Hammond and M. Noonan, eds.), pp. 3 2 5 - 3 3 8 . Academic
Press, San Diego.
Kager, R. (1989). A Metrical Theory of Stress and Destressing in English and Dutch. Foris,
Dordrecht.
Kaisse, E. (1986). Locating Turkish devoicing. Proceedings of the West Coast Conference
on Formal Linguistics 5, 119 - 1 2 8 .
Kanerva, J. (1987). Morphological integrity and syntax: The evidence from Finnish pos-
sessive suffixes. Language 6 3 , 4 9 8 - 5 2 1 .
Kean, M.-L. (1974). The strict cycle in phonology. Linguistic Inquiry 5, 1 7 9 - 2 0 3 .
Keyser, S. J., and Kiparsky, P. (1984). Syllable structure in Finnish phonology. In Language
Sound Structure (M. Aronoff and R. T. Oehrle, eds.), pp. 7 - 3 1 . MIT Press, Cam-
bridge, Mass.
Kiparsky, P. (1973). Phonological representations. In Three Dimensions of Linguistic
Theory (Osamu Fujimura, ed.). TEC, Tokyo.
Kiparsky, P. (1982). Lexical phonology and morphology. In Linguistics in the Morning
Calm (Linguistic Society of Korea, ed.), pp. 3 - 9 1 . Hanshin, Seoul.
Kiparsky, P. (1984). On the lexical phonology of Icelandic. In Nordic Prosody III(C. Elert,
I. Johansson, and E. Strangert, eds.), pp. 135-164. Almqvist & Wiksell, Stockholm.
Kiparsky, P. (1985). Some consequences of lexical phonology. Phonology Yearbook 2,
82-138.
Kiparsky, P. (1992). Underspecification and Harmony Systems. Unpublished manuscript,
Stanford University, Stanford, Calif.
Lahiri, A., and Evers, V. (1989). Palatalization and coronality. In The special status
of Coronals: Internal and External Evidence (C. Paradis and J.-F. Prunet, eds.),
pp. 7 9 - 1 0 0 . Academic Press, San Diego.
Levergood, B. (1984). Rule governed vowel harmony and the strict cycle. Proceedings of
the North Eastern Linguistic Society 14, 2 7 5 - 2 9 3 .
Lombardi, L., and McCarthy, J. (1991). Prosodic circumscription in Choctaw morphology.
Phonology 8, 3 7 - 7 2 .
Macdonell, A. (1968). Vedic Grammar. Indological Book House, Varanasi.
Macfarland, T , and Pierrehumbert, J. (1991). On ich-Laut, ach-Laut and structure preser-
vation. Phonology 8, 171 - 1 8 0 .
Mascaro, J. (1976). Catalan Phonology. Doctoral dissertation, Massachusetts Institute of
Technology, Cambridge.
Mascaro, J. (1987). A Reduction and Spreading Theory of Voicing and Other Sound Effects.
Unpublished manuscript, Barcelona.
Blocking in Nonderived Environments 313

McCarthy, J. (1986). OCP effects: Gemination and antigemination. Linguistic Inquiry 17,
207-263.
Oresnik, J. (1977). Modern Icelandic u-umlaut from the descriptive point of view. Gripla
2, 151-182. Reprinted in Studies in the Phonology and Morphology of Modem Ice-
landic (J. Oresnik, ed., 1985). Buske, Hamburg.
Ottosson, K. (1988). Fragments of the Lexical Morphology and Phonology of Icelandic.
Unpublished manuscript, University of Maryland, College Park.
Poser, W. (1982). Phonological representations and action-at-a-distance. In The Structure
of Phonological Representations, part 2 (H. van der Hulst and N. Smith, eds.),
pp. 1 2 1 - 1 5 8 . Foris, Dordrecht.
Poser, W. (1986). Diyari stress, metrical structure assignment, and the nature of metrical
representation. Proceedings of the West Coast Conference on Formal Linguistics 5,
178-191.
Poser, W. (1989). The metrical foot in Diyari. Phonology 6, 1 1 7 - 1 4 8 .
Poser, W. (1990). Are Strict Cycle Effects Derivable? Unpublished manuscript, Stanford
University, Stanford, Calif.
Prince, A. (1985). Improving tree theory. Proceedings of the Berkeley Linguistic Society
11,471-490.
Pulleyblank, D. (1986). Tone in Lexical Phonology. Reidel, Dordrecht.
Ramsey, S. R. (1978). Accent and Morphology in Korean Dialects. The Society of Korean
Linguistics, Seoul.
Rice, K. (1988). Continuant voicing in Slave (Northern Athapaskan): The cyclic applica-
tion of default rules. In Theoretical Morphology (M. Hammond and M. Noonan, eds.),
pp. 3 7 1 - 3 8 8 . Academic Press, San Diego.
Rubach, J. (1984). Cyclic and Lexical Phonology. Foris, Dordrecht.
Selkirk, E. (1980). Prosodic domains in phonology: Sanskrit revisited. In Juncture (M.
Aronoff and M.-L. Kean, eds.), pp. 1 0 7 - 1 2 9 . Anma Libri, Saratoga, Calif.
Shaw, P. (1985). Lexical phonology. Phonology Yearbook!, 1 7 1 - 2 0 0 .
Steriade, D. (1982). Greek Prosodies and the Nature of Syllabification. Doctoral disserta-
tion, Massachusetts Institute of Technology, Cambridge.
Steriade, D. (1987a). Redundant values. Papers from the Regional Meeting of the Chicago
Linguistic Society 24, 3 3 9 - 3 6 2 .
Steriade, D. (1987b). Vowel Tiers and Geminate Blockage. Unpublished manuscript, Mas-
sachusetts Institute of Technology, Cambridge.
Steriade, D. (1988). Greek accent: a case for preserving structure. Linguistic Inquiry 19,
271-314.
Wackernagel, J. (1895). Altindische Grammatik, vol. 1. Vandenhoeck & Ruprecht,
Gottingen.
Whitney, W. D. (1887). Sanskrit Grammar. Harvard University Press, Cambridge, Mass.
Zwicky, A. (1986). The general case: Basic form versus default form. Proceedings of the
Berkeley Linguistics Society 12, 305 - 1 4 .
ARE STRICT CYCLE EFFECTS DERIVABLE?

WILLIAM J. POSER
Department of Linguistics
Stanford University
Stanford, California 94305

1. D E R I V E D E N V I R O N M E N T E F F E C T S

D e r i v e d e n v i r o n m e n t effects, in w h i c h a rule fails to apply in n o n d e r i v e d e n v i -

r o n m e n t s that o t h e r w i s e satisfy its structural description, are generally attributed
to the strict c y c l e condition (Kean, 1974; M a s c a r o , 1976). K i p a r s k y (this v o l u m e )
p r o p o s e s to e l i m i n a t e the strict cycle condition, a r g u i n g that d e r i v e d e n v i r o n m e n t
effects result from the application of structure-building rules to underspecified
representations. T h e p u r p o s e of this note is to p r e s e n t a c a s e that a p p e a r s to p o s e
serious p r o b l e m s for K i p a r s k y ' s attempt to e l i m i n a t e the strict cycle condition.
T o illustrate K i p a r s k y ' s a p p r o a c h , c o n s i d e r his analysis of the w e l l - k n o w n
F i n n i s h rule that takes / t / to / s / before / i / , w h i c h applies only in d e r i v e d e n v i r o n -
m e n t s . Relevant f o r m s , along with the p r o p o s e d u n d e r l y i n g representations, are
given in (1).

(1) underlying
nominative essive representation gloss
a. last lasina laTi 'glass'
b. koti kotina koti 'home'
c. vesi vetend veTE 'water'
d. kuusi kuusena kuusE 'fir'

T h e derived e n v i r o n m e n t effect is illustrated by the contrast b e t w e e n ( l b ) and

( l c ) . In ( l b ) Itl p r e c e d e s I'll in the u n d e r l y i n g representation and is unaffected. In

315
Phonetics and Phonology, Volume 4 Copyright © 1993 by Academic Press, Inc.
Studies in Lexical Phonology All rights of reproduction in any form reserved.
316 William J. Poser

( l c ) Itl u n d e r l y i n g l y p r e c e d e s / e / , and it is only w h e n / e / raises to lil in word-final

position that the p r e c e d i n g Itl b e c o m e s Isl.
O n K i p a r s k y ' s analysis there are t w o u n d e r l y i n g n o n - l o w front v o w e l s , / i / ,
w h i c h is [ + h i g h ] , a n d IEI, w h i c h is [Ohigh]. T h e r e is n o distinctively [ - h i g h ]
n o n - l o w front vowel as this is forbidden by p r i n c i p l e s of radical underspecifica-
tion. T h e a r c h i s e g m e n t IEI m a y r e c e i v e a value for [high] either by the E - R a i s i n g
rule ( 2 ) or by the default rule ( 3 ) .

(2) F I N A L E-RAISING:
[]->[+high]/ #

(3) HIGH DEFAULT:

[]-»[-high]

F o r [t] and [s] w e have three u n d e r l y i n g possibilities, Isl ( [ + c n t ] ) , ITI ([Ocnt]),

and Itl ([ — cnt]). Isl is p e r m i t t e d only before u n d e r l y i n g HI, w h i l e Itl is p e r m i t t e d
only w h e n not p r e c e d i n g / i / , so that in n o single e n v i r o n m e n t is there an under-
lying distinction b e t w e e n [ 4 - c n t ] a n d [ — c n t ] . Before lil w e c a n have either Itl or
ITI; e l s e w h e r e w e can h a v e Isl or ITI. T h e default rule supplies the value [ — c n t ] ,
p r o d u c i n g [t]. T h e rule that t a k e s Itl to Isl before lil is a purely feature-filling rule
and so applies to ITI, filling in the value [ + c n t ] , but not to Itl, w h i c h is already
specified [ — c n t ] .
T h e d e r i v e d e n v i r o n m e n t effect results from the fact that d e r i v e d [i] (that is,
IEI) m a y be p r e c e d e d either by Isl, w h i c h is a l w a y s realized as [s], or by ITI,
w h i c h is realized as [t] w h e n IEI is realized as [e], and as [s] w h e n IEI is realized
as [i], w h i l e u n d e r i v e d [i] (HI) m a y b e p r e c e d e d either by Itl, w h i c h is a l w a y s
realized as [t], or by ITI, w h i c h before lil is a l w a y s realized as [s].
T h e p r o b l e m a t i c e x a m p l e involves t w o rules of C h u m a s h , a n o w extinct lan-
g u a g e of California. T h e data p r e s e n t e d h e r e are from the I n e s e n o dialect, d e -
scribed by A p p l e g a t e ( 1 9 7 2 ) on the basis of the field notes of J o h n P. H a r r i n g t o n .
T h e situation in the V e n t u r e n o dialect, d e s c r i b e d in the p o s t h u m o u s l y p u b l i s h e d
H a r r i n g t o n ( 1 9 7 4 ) , is very similar. T h e analysis given h e r e recapitulates that
of Poser ( 1 9 8 2 ) , w h i c h m a y b e c o n s u l t e d for additional e x a m p l e s and details
omitted here. 1

2. C H U M A S H S I B I L A N T H A R M O N Y

T h e first of the t w o relevant rules is Sibilant H a r m o n y , w h i c h c a u s e s all sibi-

lants (including affricates) to a g r e e in laminality with the last sibilant in the w o r d . 2

( 4 ) illustrates the fact that the third p e r s o n subject prefix surfaces as [s] w h e n n o
other sibilant follows. But w h e n the past tense suffix / w a s / is a d d e d as in ( 5 ) , Isl
b e c o m e s Isl.
Are Strict Cycle Effects Derivable? 317

(4) hasxintila / h a + s 4- xintila/ 'his g e n t i l e '

(5) hasxintilawas / h a + s + xintila + w a s / 'his f o r m e r g e n t i l e '

Sibilant H a r m o n y also c a u s e s u n d e r l y i n g I si to surface as I si. (6) s h o w s that the

dual subject prefix is u n d e r l y i n g l y / i s / since that is the form in w h i c h it a p p e a r s
w h e n n o other sibilant follows. In (7), w h e r e the last sibilant is / s / , the dual prefix
h a r m o n i z e s and surfaces as / s / .

(6) pisanan? / p + is + al 4- n a n ? / ' d o n ' t you t w o g o '

(7) sishuleqpeyus / s + is + sili + u l u a q p e y + u s / 'they t w o w a n t to

follow it'

T h e s e e x a m p l e s d e m o n s t r a t e that Sibilant H a r m o n y is feature-changing, since

it c a u s e s u n d e r l y i n g I si to surface as I si a n d u n d e r l y i n g III to surface as I si. For a
h a r m o n y s y s t e m to b e purely feature-filling, it m u s t b e the c a s e that h a r m o n i z i n g
s e g m e n t s are unspecified for the h a r m o n y feature. Therefore, s e g m e n t s subject to
h a r m o n y will all take o n the s a m e value, w h a t e v e r value is supplied b y the relevant
default rule, w h e n they lie outside the d o m a i n of a h a r m o n y trigger. In a c a s e like
the o n e at h a n d , in w h i c h the value of the h a r m o n y feature is u n p r e d i c t a b l e w h e n
not in the d o m a i n of a trigger, the value c a n n o t b e attributed to a default rule a n d
therefore m u s t b e part of the representation of the s e g m e n t , that is, m u s t b e speci-
fied. S i n c e b o t h I si a n d I si u n d e r g o h a r m o n y , the h a r m o n y rule m u s t c h a n g e u n -
derlyingly present feature values.

3. C H U M A S H P R E - C O R O N A L L A M I N A L I Z A T I O N

In addition to Sibilant H a r m o n y , C h u m a s h h a s a s e c o n d rule affecting t h e 1am-

inality of sibilants, w h i c h I call P r e - C o r o n a l L a m i n a l i z a t i o n ( P C L ) . T h i s rule,
stated in (8), m a k e s a sibilant laminal w h e n it i m m e d i a t e l y p r e c e d e s o n e of the
n o n - s t r i d e n t c o r o n a l s /t/, / l / , or Inl. T h e o p e r a t i o n of the rule is illustrated by the
e x a m p l e s in (9). In e a c h c a s e the third p e r s o n subject prefix I si (apical) b e c o m e s
[s] (laminal) before a n o n s t r i d e n t c o r o n a l .

(8) [ + c o r , -f stri] —> [ + d i s t ] / [ + c o r , - stri]

(9) man? Is + n a n ? / 'he goes'

stepu? Is + t e p u ? / 'he gambles'
sloxit? Is + l o x i t ? / ' h e surpasses m e '

P C L creates a systematic class of e x c e p t i o n s to the generalization that all sibi-

lants in a w o r d agree in laminality with the last. Sibilants w h o s e laminality is
d e t e r m i n e d by P C L are o p a q u e to Sibilant H a r m o n y . In (10a) w e see that the I si
318 William J. Poser

created by P C L fails to h a r m o n i z e with the Isl of / u s / . In (10b) not only d o e s the

Isl created by P C L fail to h a r m o n i z e with the t w o I sis of /sisin/, but it serves as a
trigger with respect to the Isl that p r e c e d e s it.

(10) a. stiyepus / s + ti + y e p + u s / ' h e tells h i m '

b. sislusisin Is + is + lu 4- sisin/ 'they t w o are g o n e a w r y '

In P o s e r ( 1 9 8 2 ) I a r g u e d that the opacity of Isl created b y P C L is i n c o m p a t i b l e

with theories in w h i c h feature-changing rules directly c h a n g e feature values. In
such theories, if P C L applies before Sibilant H a r m o n y , Isl created b y P C L should
h a r m o n i z e j u s t like u n d e r l y i n g / s / , since Sibilant H a r m o n y has n o w a y of distin-
g u i s h i n g b e t w e e n u n d e r l y i n g Isl a n d t h o s e derived b y P C L . If P C L applies after
Sibilant H a r m o n y , / s / s created b y P C L will fail to h a r m o n i z e but will not b e
o p a q u e , for p r e c e d i n g sibilants will h a r m o n i z e with the last sibilant in the w o r d
and will not h a r m o n i z e w i t h the Isl created b y P C L . In other w o r d s , if P C L applies
after Sibilant H a r m o n y , P C L will create islands, not o p a q u e s e g m e n t s .
T h e s a m e p r o b l e m arises in the putatively n o n - f e a t u r e - c h a n g i n g analysis of
Avery & R i c e ( 1 9 8 9 ) . In their discussion of C h u m a s h (pp. 193 f.) they d o not
directly address the e v i d e n c e that C h u m a s h is feature-changing, and they describe
their analysis as not f e a t u r e - c h a n g i n g . In o n e sense their analysis is not feature-
c h a n g i n g , for it d o e s n o t m a k e u s e of explicit d e l i n k i n g or deletion rules or di-
rectly c h a n g e o n e feature value into another. T h e i r analysis is, however, feature-
c h a n g i n g in a n o t h e r sense, namely, in that the h a r m o n y rule deletes u n d e r l y i n g l y
present information, a n d it is only for this r e a s o n that it can a c c o m o d a t e the evi-
d e n c e p r e s e n t e d a b o v e that C h u m a s h h a r m o n y is feature-changing. T h e i r har-
m o n y rule fuses the P l a c e n o d e of the last sibilant w i t h t h o s e of the sibilants before
it. T h e y explain the operation of fusion as follows (p. 181):

Fusion is an operation which takes identical primary content nodes and fuses them pro-
vided that the nodes are non-distinct; i.e., both nodes do not dominate different secondary
nodes. We assume that fusion is headed in that the secondary features of the triggering
segment are maintained.

A s the e x a m p l e s d i s c u s s e d on p . 182 m a k e clear, not only are the s e c o n d a r y fea-

tures of the trigger m a i n t a i n e d , b u t the s e c o n d a r y features of the h a r m o n i z i n g
s e g m e n t are deleted. T h u s , fusion of a [ + d i s t ] trigger with a [ — dist] u n d e r g o e r
m a k e s b o t h [ + d i s t ] , w h i c h is the equivalent of a feature-filling spreading opera-
tion, b u t fusion of a [ - d i s t ] trigger w i t h a [ + d i s t ] u n d e r g o e r c a u s e s deletion of
the s e c o n d a r y features of the undergoer, m a k i n g b o t h s e g m e n t s [ - d i s t ] , the
equivalent of spreading a c c o m p a n i e d by feature deletion. In the relevant sense,
then, their analysis is feature-changing, and like m o r e traditional feature-changing
analyses, c a n n o t a c c o u n t for the opacity of Isl created b y P C L , w h i c h they d o not
discuss.
T h e opacity of Isl created by P C L c a n b e a c c o u n t e d for in a theory like that of
P o s e r ( 1 9 8 2 ) in w h i c h feature-changing is the result of t w o distinct rules, o n e
Are Strict Cycle Effects Derivable? 319

delinking or deleting feature specifications, the other inserting n e w feature speci-

fications or s p r e a d i n g existing o n e s . T h e p r o p o s e d t w o - s t a g e analysis of C h u -
3

m a s h Sibilant H a r m o n y orders P C L b e t w e e n D i s t r i b u t e d D e l i n k i n g , w h i c h d e -
links n o n r i g h t m o s t specifications of the feature [dist], a n d D i s t r i b u t e d S p r e a d i n g ,
w h i c h s p r e a d s t h e r e m a i n i n g r i g h t m o s t specification of [dist] leftward.

(11) O R D E R OF RULES:
Distributed Delinking
Pre-Coronal Laminalization
Distributed Spreading

Ills c r e a t e d b y P C L fail to u n d e r g o Sibilant H a r m o n y b e c a u s e D i s t r i b u t e d D e -

linking gets n o o p p o r t u n i t y to delink t h e m , a n d D i s t r i b u t e d S p r e a d i n g is purely
feature-filling. 4

W e n o w c o m e to the crux of o u r a r g u m e n t , n a m e l y the fact that P C L is subject

to a d e r i v e d e n v i r o n m e n t condition. A s illustrated in (12), it fails to apply in tauto-
m o r p h e m i c clusters.

(12) stumukun 'mistletoe'

slow? 'eagle'
wastu? 'pleat'

Of c o u r s e , it is p o s s i b l e for [s] to a p p e a r before n o n s t r i d e n t c o r o n a l s m o r p h e m e

internally, as in the m o r p h e m e / w a s t i / 'of a flow, of liquid in m o t i o n ' . U n l i k e
instances of III d e r i v e d b y P C L , w h i c h as w e h a v e seen are o p a q u e to Sibilant
H a r m o n y , such u n d e r l y i n g / s / s , w h i c h are not created b y P C L , u n d e r g o Sibilant
H a r m o n y , as illustrated in (13) a n d (14). (13) illustrates the u n d e r l y i n g III that
surfaces w h e n n o I si follows. (14) s h o w s that this III is not o p a q u e to Sibilant
Harmony.

(13) wastinan? /wasti + n a n ? / ' t o spill'

(14) swastilok?inus Is + wasti + l o k ? i n + u s / ' t h e flow stops o n h i m '

T h i s is precisely w h a t m y analysis predicts: since u n d e r l y i n g [ + d i s t ] specifi-

cations are w i p e d out b y D i s t r i b u t e d D e l i n k i n g , unless P C L applies these sibilants
r e m a i n unspecified for [dist] a n d therefore u n d e r g o D i s t r i b u t e d S p r e a d i n g .
T h e d e r i v e d e n v i r o n m e n t condition on P C L is p r e d i c t e d b y the strict c y c l e c o n -
dition, since it applies only w h e n its e n v i r o n m e n t is satisfied as the result of m o r -
p h o l o g i c a l c o m p o s i t i o n . B u t it c a n n o t b e d e r i v e d from constraints on u n d e r l y i n g
feature specification since, o n m y analysis of Sibilant H a r m o n y , at the p o i n t at
w h i c h P C L applies, b o t h t h e sibilants that u n d e r g o P C L a n d t h o s e that fail to
u n d e r g o P C L are unspecified for the feature [dist] as a result of the prior applica-
tion of D i s t r i b u t e d D e l i n k i n g . It therefore a p p e a r s to b e i m p o s s i b l e to d e r i v e strict
cycle effects in all c a s e s from underspecification.
T h i s leaves o p e n the question of w h e t h e r the strict c y c l e condition m u s t b e
320 William J. Poser

m a i n t a i n e d in its entirety, for it has t w o clauses. O n e clause classifies an environ-

m e n t as derived if the structural description of the rule c o m e s to b e satisfied as a
result of m o r p h o l o g i c a l c o m p o s i t i o n . T h e other classifies an e n v i r o n m e n t as d e -
rived if the structural description c o m e s to b e satisfied as a result of prior appli-
cation of a p h o n o l o g i c a l rule. K i p a r s k y ' s p r o p o s a l appears to derive successfully
the k n o w n cases of derived e n v i r o n m e n t effects d u e to prior application of a p h o -
nological rule. T h e force of the C h u m a s h e x a m p l e m a y b e that it is necessary to
retain the first clause of the strict cycle condition, by w h i c h derived e n v i r o n m e n t s
result from m o r p h o l o g i c a l c o m p o s i t i o n , w h i l e d i s p e n s i n g with the second. T h i s is
a natural distinction, for this latter criterion requires information about the deri-
vation of a sort not present in the representation, in contrast to information about
m o r p h o l o g i c a l structure, w h i c h arguably is present in the representation.

NOTES

1
Let me take this opportunity to correct an error in Poser (1982), where I glossed has-
xintilawas as 'his former Indian name'. Actually, this form is the adjective meaning 'his
former gentile' and has the given meaning only when combined with the noun masti
"name", xintila is a loan from Spanish gentil 'gentile, heathen'.
2
I n Poser (1982) I treated the distinction between Chumash Isl and III as a distinction
between [4-ant] and [ — ant]. Here I adopt the proposal of Lieber (1987:147) that it should
be characterized as a distinction between [ — dist] and [H-dist], that is, as a distinction be-
tween apical and laminal.
3
Lieber (1987) mistakenly credits herself with this proposal and attributes to Poser
(1982) the metrical view of which it is a critique.
4
A full formalization of this analysis requires a decision as to how to handle the trans-
parency of the nonstrident coronals, which fortunately does not appear to be relevant to the
points at issue here. For discussion of the transparency issue see Shaw (1991).

REFERENCES

Applegate, R. B. (1972). Ineseno Chumash Grammar. Doctoral dissertation, University of

California, Berkeley.
Avery, P., and Rice, K. (1989). Segment structure and coronal underspecification. Pho-
nology 6, 179-200.
Harrington, J. P. (1974). Sibilants in Ventureno. International Journal of American Lin-
guistics 40, 1-9.
Kean, M.-L. (1974). The strict cycle in phonology. Linguistic Inquiry 5, 1 7 9 - 2 0 3 .
Lieber, R. (1987). An Integrated Theory of Autosegmental Processes. State University of
New York Press, Albany.
Are Strict Cycle Effects Derivable? 321

Mascaro, J. (1976). Catalan Phonology and the Phonological Cycle. Doctoral dissertation,
Massachusetts Institute of Technology, Cambridge.
Poser, W. J. (1982). Phonological representation and action-at-a-distance. In The Structure
of Phonological Representations, part 2 (H. van der Hulst and N. Smith, eds.),
pp. 1 2 1 - 1 5 8 . Foris, Dordrecht.
Shaw, P. A. (1991). Consonant harmony systems: The special status of coronal harmony.
In Phonetics and Phonology, Vol. 2 (C. Paradis and J.-F. Prunet, eds.), pp. 1 2 5 - 1 5 7 .
Academic Press, San Diego.
THE CHRONOLOGY AND STATUS
OF ANGLIAN SMOOTHING

B. ELAN DRESHER
Department of Linguistics
University of Toronto
Toronto, Ontario, Canada M5S 1A1

1. I N T R O D U C T I O N

T h e O l d E n g l i s h s o u n d c h a n g e k n o w n as S m o o t h i n g h a s b e e n p r o b l e m a t i c in
t e r m s of b o t h its c h r o n o l o g y and status. Descriptively, S m o o t h i n g is a n a m e given
to the p r o c e s s w h e r e b y , in certain dialects, d i p h t h o n g s w e r e m o n o p h t h o n g i z e d
before the velar c o n s o n a n t s k, g, and x, w h e n t h e s e followed either directly or with
an i n t e r v e n i n g r or /. T h e o b s e r v e d c h a n g e s are listed in (1).

(1) cea ce
e *" ' "r\
1

eo > -> < I

\l
Jo i ^ [ 4 - stress]

+cons
=/ +son 4-obst
[ 4 - stress] _ - nasal_ +back

T h e rule is s o m e w h a t u n e x p e c t e d , since s o m e of these c o n t e x t s w e r e o n e s that h a d

formerly c a u s e d d i p h t h o n g i z a t i o n . H e n c e S m o o t h i n g a p p e a r s to reverse the effects
of earlier rules, w h i c h leads to the suspicion that o n e or the o t h e r is n o t a natural
p r o c e s s . F r o m the p o i n t of v i e w of p h o n o l o g i c a l theory, the rule, t o g e t h e r w i t h the
p r o c e s s e s it interacts w i t h , p r o v o k e s a n u m b e r of interesting q u e s t i o n s . M y m a i n
c o n c e r n h e r e is w i t h the results of the interaction of S m o o t h i n g w i t h other p r o -
c e s s e s and the extent to w h i c h these interactions b e a r on the theory of lexical
p h o n o l o g y a n d vice versa.
325
Phonetics and Phonology, Volume 4 Copyright © 1993 by Academic Press, Inc.
Studies in Lexical Phonology All rights of reproduction in any form reserved.
326 B. Elan Dresher

In brief, I a r g u e that there are a n u m b e r of rules in M e r c i a n O l d E n g l i s h w h i c h

apply quite regularly in w e a k verbs and n o u n s but not in strong verbs. U p o n closer
inspection, w e find that the m o r p h o l o g i c a l c o n t e x t s of the apparent exceptionality
h a v e a structural basis, in that certain rules systematically d o not a p p l y w h e n their
triggering c o n t e x t includes verb a g r e e m e n t suffixes. In t e r m s of lexical p h o n o l o g y ,
this suggests that the p h o n o l o g y is partitioned into at least t w o levels. M o r e o v e r ,
the rules limited to the earlier level all h a v e in c o m m o n that they c a n b e treated as
persistent constraints. T h e facts suggest that certain types of rules are susceptible
to b e reinterpreted as c o n s t r a i n t s , t h e r e b y exhibiting c o m p l e x interactions with
other p r o c e s s e s .
First, s o m e p r e l i m i n a r i e s about the l a n g u a g e u n d e r discussion. S m o o t h i n g ap-
plied in the A n g l i a n dialects of O l d E n g l i s h , that is, not in W e s t S a x o n , w h i c h is
the dialect m o s t familiar from the standard h a n d b o o k s . T h e t e r m " A n g l i a n "
g r o u p s t o g e t h e r a n u m b e r of dialects; the o n e w e will b e l o o k i n g at is t h e M e r c i a n
dialect. O u r k n o w l e d g e of this dialect, a n d the various c h a n g e s it u n d e r w e n t ,
c o m e s from a series of texts from different p e r i o d s . T h e largest and m o s t c o n s i s -
tent text is the Vespasian Psalter [ P s ( A ) ] , w h i c h dates from a r o u n d 8 2 5 . Earlier
than P s ( A ) are s o m e glossaries that date from a r o u n d 7 0 0 - 8 0 0 ; quite a bit later
than it are a g r o u p of early M i d d l e E n g l i s h texts from the early 1 2 0 0 s . 1

A w o r d a b o u t the v o w e l s : O l d E n g l i s h v o w e l s as well as d i p h t h o n g s could be

either short or long. D i p h t h o n g s are all falling, with stress o n the first part, w h i c h
is a l w a y s a front v o w e l . T h e s e c o n d part of a d i p h t h o n g is always back, written o
w h e n the first p a r t is n o n l o w a n d a w h e n the first p a r t is low. W e can take it to b e
s o m e kind of s c h w a . 2

I b e g i n b y r e v i e w i n g s o m e p r o b l e m s in d e t e r m i n i n g the c h r o n o l o g y of S m o o t h -
ing. T h e s e p r o b l e m s are w o r t h c o n s i d e r i n g , even t h o u g h the c h r o n o l o g y of
S m o o t h i n g relative to o t h e r rules is in fact easily d e t e r m i n e d b y l o o k i n g at the
d o c u m e n t s . N e v e r t h e l e s s , m a n y traditional g r a m m a r i a n s did n o t believe w h a t the
texts a p p e a r to b e s h o w i n g , b e c a u s e the interaction of S m o o t h i n g with s o m e other
rules w a s not w h a t it should h a v e b e e n in t e r m s of their theory of l a n g u a g e c h a n g e ,
w h i c h itself w a s c o n n e c t e d to their t h e o r y of s y n c h r o n i c g r a m m a r . S o a study of
this c a s e b r i n g s o u t clearly the i n t e r d e p e n d e n c e b e t w e e n s y n c h r o n i c t h e o r y a n d
the interpretation of d i a c h r o n i c d e v e l o p m e n t s , as well as turning u p an interesting
p r o b l e m to w h i c h w e c a n apply c u r r e n t theories. L e t us turn, then, to the facts of
this case.

2. T H E C H R O N O L O G Y O F S M O O T H I N G

S m o o t h i n g applies fairly regularly to d i p h t h o n g s in early texts, a fact w h i c h

s u g g e s t s that it p r e d a t e s t h e m . A difficulty is p o s e d b y its interaction with the
Anglian Smoothing 327

relatively late rule of B a c k M u t a t i o n , also k n o w n as " b a c k u m l a u t " or " v e l a r

u m l a u t . " B a c k M u t a t i o n d i p h t h o n g i z e s short front v o w e l s if a b a c k vowel stands
in the following syllable; the rule is given in (2). (In the first part of this article, all
rules are written in informal linear format; these rules are m e a n t to b e d e s c r i p -
tive only.)

(2) ce cea
e eo * I C V
i io [+stress] [+back]

N o t e that B a c k M u t a t i o n potentially conflicts with S m o o t h i n g w h e n the interven-

ing c o n s o n a n t is velar. A traditional a s s u m p t i o n , g o i n g b a c k to the N e o g r a m m a r -
ians, is that a s o u n d c h a n g e , that is, w h a t is in o u r t e r m s a n e w rule such as B a c k
M u t a t i o n , applies to surface f o r m s . T h i s is e q u i v a l e n t to s u p p o s i n g that rules are
3

a d d e d only to the e n d of the g r a m m a r . Therefore, if S m o o t h i n g is historically

earlier than B a c k M u t a t i o n , as s u g g e s t e d b y the m a n u s c r i p t e v i d e n c e , w e w o u l d
e x p e c t to o b s e r v e the d i a c h r o n i c s e q u e n c e s h o w n in (3).

(3) a. b.
weorc dcegas Earlier forms
were — Smoothing
— dceagas Back Mutation

W e e x p e c t S m o o t h i n g to apply to old d i p h t h o n g s , as in (3a), b u t n o t to the n e w

d i p h t h o n g s created b y B a c k M u t a t i o n (3b); that is, B a c k M u t a t i o n should apply
freely before velars. In fact, t h o u g h , it d o e s not. W h a t w e actually find is variation
b e t w e e n s m o o t h e d a n d u n s m o o t h e d f o r m s , even at a t i m e w h e n B a c k M u t a t i o n
applies regularly before other (i.e., n o n v e l a r ) c o n s o n a n t s .
T h e interaction of S m o o t h i n g with B a c k M u t a t i o n thus d o e s n o t accord with
N e o g r a m m a r i a n a s s u m p t i o n s a b o u t s o u n d c h a n g e , and traditional g r a m m a r i a n s
h a v e h a d to p r o p o s e a variety of e x t r a m e c h a n i s m s to a c c o u n t for it. F o r e x a m p l e ,
to a c c o u n t for the e x i s t e n c e of s m o o t h e d f o r m s , L u i c k ( 1 9 6 4 : §235.a.3) p r o p o s e d
that B a c k M u t a t i o n w a s p r e v e n t e d b y b a c k c o n s o n a n t s in A n g l i a n . S u c h an as-
s u m p t i o n , however, is h a r d to reconcile with the e x i s t e n c e of f o r m s w h e r e it did
appply before b a c k c o n s o n a n t s . M o r e o v e r , the fact that b o t h S m o o t h i n g a n d B a c k
M u t a t i o n o b s e r v e the identical restriction on d i p h t h o n g s before b a c k c o n s o n a n t s
w o u l d b e u n e x p l a i n e d . F o r these r e a s o n s , C a m p b e l l ( 1 9 5 9 : § 2 4 7 ) a r g u e d for a
relatively late date for S m o o t h i n g , following B a c k M u t a t i o n . H e c l a i m s that the
" c o m p l e t e a b s e n c e in A n g l i a n e x c e p t in analogical forms of b a c k u m l a u t before
b a c k c o n s o n a n t s indicates that its effects w e r e r e m o v e d b y s m o o t h i n g , a n d that it
in fact t o o k p l a c e earlier than that c h a n g e . " C a m p b e l l therefore posits the dia-
c h r o n i c s e q u e n c e in (4).
328 B. Elan Dresher

(4) a. b.
dcegas sprecan Earlier forms
dceagas spreocan Back Mutation
dcegas sprecan Smoothing
spreocan Analogy

In this scenario, the earlier spellings with d i p h t h o n g s p r e d a t e S m o o t h i n g , while

the later o n e s represent an analogical restoration after S m o o t h i n g .
O n e p r o b l e m with this a c c o u n t is that there is n o c o n v i n c i n g basis for the ana-
logical c h a n g e in traditional t e r m s . A l s o , C a m p b e l l ' s view d o e s not accord with
the e v i d e n c e of the m a n u s c r i p t s , as h a s b e e n s h o w n b y Ball and Stiles ( 1 9 8 3 ) , w h o
find n o texts w h i c h s h o w B a c k M u t a t i o n but not S m o o t h i n g . T h e y c o n c l u d e that
S m o o t h i n g p r e c e d e d B a c k M u t a t i o n , a n d that B a c k M u t a t i o n of ce before velar
c o n s o n a n t s did take p l a c e " i n s o m e parts of the M e r c i a n a r e a . " T h u s they suggest
that the variation is d u e to dialect m i x t u r e , w h e r e f o r m s with d i p h t h o n g s (dceagas)
represent the c o n t r i b u t i o n of the area w h i c h did allow B a c k M u t a t i o n before ve-
lars, w h i l e forms with m o n o p h t h o n g s (dcegas) are from the area w h i c h did not.
This account, however, b e g s the question of w h y B a c k M u t a t i o n , the later rule,
should ever h a v e b e e n i m p e d e d before velar c o n s o n a n t s . Further, if the recorded
forms reflect dialect variation, w e w o u l d e x p e c t to find n o systematic pattern to
the variation; but the distribution of s m o o t h e d and u n s m o o t h e d forms in the s a m e
texts d o not support this view, as w e shall see.
T o o n ( 1 9 8 3 ) also d e m o n s t r a t e s that S m o o t h i n g p r e c e d e d B a c k M u t a t i o n in the
texts h e surveyed. H e s u m s up his position as follows (p. 184): " A l t h o u g h velar
u m l a u t and s m o o t h i n g w e r e s y n c h r o n i c forces in the s a m e texts, s m o o t h i n g b e g a n
earlier and w a s c o m p l e t e before velar u m l a u t . . . . S m o o t h i n g h a d c e a s e d to b e a
p r o d u c t i v e force in the l a n g u a g e of the Corpus Glossary: [aea] could d e v e l o p b e -
fore the velar spirant. It w a s o b v i o u s l y not a p r o d u c t i v e rule in the p h o n o l o g y of
the Vespasian Psalter s c r i b e . "
T h e situation in P s ( A ) d o e s not, however, support the view that S m o o t h i n g had
b e c o m e an u n p r o d u c t i v e rule. A l o o k at the distribution of s m o o t h e d and un-
s m o o t h e d forms reveals that the p r e s e n c e or a b s e n c e of d i p h t h o n g s h a s c o m e to
d e p e n d on m o r p h o l o g i c a l c a t e g o r y : strong verbs and their derivatives d o not in
general s h o w S m o o t h i n g in B a c k M u t a t i o n e n v i r o n m e n t s , b u t w e a k verbs, n o u n s ,
and adjectives d o . T h i s generalization h o l d s with very few e x c e p t i o n s . T h e n u m -
bers of forms of e a c h type are s h o w n in (5).

(5) B A C K M U T A T I O N BEFORE VELARS I N P S ( A ) :

Monophthongs Diphthongs

Nouns 67 5
W e a k verbs 15 0
S t r o n g verbs and derivatives 3 69
Anglian Smoothing 329

T h u s the s y n c h r o n i c rule o r d e r required to derive the strong verbs m i r r o r s the

d i a c h r o n i c o r d e r of S m o o t h i n g before B a c k M u t a t i o n (6a). H o w e v e r , a different
s y n c h r o n i c o r d e r of application is required in the other major lexical classes (6b).

(6) REORDERING OF SMOOTHING A N D B A C K M U T A T I O N :

a. S t r o n g verbs
sprecan Original Form
— Smoothing
spreocan Back Mutation
b. W e a k verbs and n o u n s
cwcecade dcegas Original F o r m
cwceacade dceagas Back Mutation
cwcecade dcegas Smoothing

Classical g e n e r a t i v e p h o n o l o g y , in the style of C h o m s k y a n d Halle ( 1 9 6 8 ) , p r o -

vides a r i c h e r a p p a r a t u s in t e r m s of w h i c h n e w rules can b e integrated into the
g r a m m a r . T h u s rules m a y b e a d d e d to the e n d of a g r a m m a r , or before the e n d ;
they m a y reorder, or b e c o m e modified or lost. In this case, S m o o t h i n g p r e c e d e d
B a c k M u t a t i o n chronologically. A d d i n g B a c k M u t a t i o n to the e n d of the g r a m m a r
w o u l d give the s e q u e n c e in (6a): B a c k M u t a t i o n should apply u n i m p e d e d by
S m o o t h i n g . T h i s m a y h a v e b e e n the initial situation; subsequently, the rules a p -
p e a r to h a v e r e o r d e r e d in m o s t lexical classes, creating the derivations s h o w n
in (6b).
R e o r d e r i n g b y itself, however, d o e s n o t a d e q u a t e l y a c c o u n t for the facts of the
case. First, the m o t i v a t i o n for the r e o r d e r i n g is unclear. In the putative original
order, B a c k M u t a t i o n counterfeeds S m o o t h i n g ; the r e o r d e r i n g is into feeding or-
der. To the extent that m a x i m a l application is a consideration, this o r d e r is an
i m p r o v e m e n t . H o w e v e r , the o r d e r r e m a i n s o p a q u e , in the sense of K i p a r s k y
( 1 9 7 3 ) . In the original order, B a c k M u t a t i o n creates surface violations of S m o o t h -
ing, so that it a p p e a r s that S m o o t h i n g has u n d e r a p p l i e d . R e o r d e r i n g r e m o v e s these
surface e x c e p t i o n s to S m o o t h i n g but creates surface e x c e p t i o n s to B a c k M u t a t i o n ,
m a k i n g it l o o k like it has u n d e r a p p l i e d . T h u s the synchronically later rule m a k e s
the earlier o n e o p a q u e , in either order. S e c o n d , w e c a n n o t explain w h y the reor-
dering w a s not carried t h r o u g h in the strong verbs. T h e r e are cases w h e r e an ana-
logical c h a n g e is i m p e d e d in a certain class of f o r m s , and w e could look for a
reason w h y strong verbs resisted the c h a n g e w h i l e w e a k verbs and n o u n s did not.
S t r o n g verbs h a d n o particular disposition to d i p h t h o n g a l forms, however, n o r did
w e a k verbs particularly avoid t h e m . In fact, it can b e s h o w n that a n a l o g y plays
virtually n o role in the distribution of d i p h t h o n g s in P s ( A ) , w h i c h can in a l m o s t
all cases b e a c c o u n t e d for p h o n o l o g i c a l l y ( D r e s h e r 1978, C h . 1). W e can c o n c l u d e
that rule r e o r d e r i n g d o e s not suffice to a c c o u n t for the o b s e r v e d variability in these
forms.
330 B. Elan Dresher

3. M O R P H O L O G I C A L C O N D I T I O N S

D r e s h e r (1978) a s s u m e d that B a c k M u t a t i o n and S m o o t h i n g h a d i n d e e d reor-

dered, that is, that the correct s y n c h r o n i c rule order for the P s ( A ) dialect w a s that
in (6b). T h e fact that S m o o t h i n g did not apply to the strong verbs w a s not ac-
c o u n t e d for by a different rule o r d e r i n g , as in (6a), but rather w a s attributed to a
failure of S m o o t h i n g to apply regularly in the strong verbs. T h i s failure forms part
of a w i d e r generalization. T h e r e are a n u m b e r of other rules in the P s ( A ) dialect
w h i c h distinguish b e t w e e n m o r p h o l o g i c a l categories in the s a m e way: they apply
regularly in w e a k verbs a n d n o u n s but not in strong verbs.
O n e such rule, like S m o o t h i n g , simplifies d i p h t h o n g s , this t i m e before the high
vowel HI. T h i s rule a c c o u n t s for the alternation w e find in the stem vowel of the
adjective ' s h a r p ' (7a) as o p p o s e d to the w e a k verb ' s h a r p e n ' (7b).

(7) a. Adjectives W e a k verbs

/scaerp+0/ /scaerp+i+d+un/ underlying
sccearp sccearp+i+d+un Breaking
sca?rp+i+d+un /-Monophthongization
^-Raising
sccerp+d+un Vowel Deletion
/-Lowering
sccearp scerptun other rules
c. S t r o n g verbs d. S t r o n g verbs
/-werp+id/ /swelg+id/ underlying
-weorp+id Breaking
FAILS /-Monophthongization
swilg+id ^-Raising
Vowel Deletion
-weorp+ed swilg+ed /-Lowering
-weorped swilged other rules

T h e d i p h t h o n g in the adjective is p r o d u c e d b y the rule of B r e a k i n g , given in (8a),

w h i c h d i p h t h o n g i z e s a front vowel before the b a c k continuants / w / and / x / , as well
as before the liquids III a n d Irl w h e n these are followed by a c o n s o n a n t .

(8) a. BREAKING:
->

ce cea
e eo y

i Jo [+stress]

=/
[+stress]
Anglian Smoothing 331

b. RETRACTION:

+son
ce —» a I
+back
[+stress] L

T h e adjective ' s h a r p ' exhibits B r e a k i n g before a cluster of / r / + c o n s o n a n t ; the

lack of a d i p h t h o n g in the verb can b e attributed to the effects of a rule of
/ - M o n o p h t h o n g i z a t i o n , given in (9); the p r e s e n c e of an lil in the m o r p h o l o g y of
w e a k verbs of this class can b e i n d e p e n d e n t l y m o t i v a t e d .

(9) 3->0/V C £ 0

B u t in the strong verbs, exemplified b y -weorpan ' t h r o w d o w n ' in (7c), /-

M o n o p h t h o n g i z a t i o n fails to apply, a l t h o u g h a potential triggering lil is present.
W e k n o w that the vowel of the third p e r s o n suffix -eth is u n d e r l y i n g l y HI b e c a u s e
it triggers other rules w h i c h apply before / i / , s u c h as raising of the s t e m v o w e l / e /
of forswelgan ' d e v o u r ' in (7d). A rule of / - L o w e r i n g lowers all u n d e r l y i n g u n -
stressed /i/s to e, g i v i n g the surface form of the suffix, -ed.
A n o t h e r rule w h i c h often fails to apply in the strong verbs is /-Mutation, w h i c h
fronts a stressed vowel w h e n an lil follows in the n e x t syllable.

(10) V -> [ - back] / C 0 i

C o m p a r e the derivations in (11) of the p r e s e n t indicative 3rd p e r s o n singular of

the strong v e r b haldan ' h o l d ' and the w e a k v e r b onhceldan 'lean d o w n ' . A rule of
Retraction (8b) retracts ce to a before a b a c k sonorant; then /-Mutation fronts the
a in (1 l b ) b u t fails in (11a).

(11) a. S t r o n g verbs b. W e a k verbs

/haeld+id/ /-haeld+i+id/ underlying
hald+id -hald+i+id Retraction
FAILS -hceld+i+id /-Mutation
halded -hcelded other rules

W e h a v e thus seen a n u m b e r of rules that fail in the strong verbs w h i l e a p p l y i n g

in the w e a k verbs, n o u n s , a n d adjectives. If w e a s s u m e n o n c y c l i c derivations of
the sort illustrated a b o v e , it a p p e a r s that this exceptional b e h a v i o r is simply a
p r o p e r t y of the class of strong verbs: n o p h o n o l o g i c a l r e a s o n suggests itself. B u t
there are interesting structural differences b e t w e e n strong a n d w e a k verbs w h i c h
can b e e x p l o i t e d b y a lexical p h o n o l o g y . A s is well k n o w n , strong verbs form their
past t e n s e b y modification of the root (ablaut), w h i l e w e a k verbs add a dental past
t e n s e m a r k e r . M o r e i m p o r t a n t to o u r analysis is a n o t h e r difference: w e a k verbs
require a s t e m - f o r m i n g e x t e n s i o n in the form of a vowel. T h i s e x t e n s i o n is lil in
w e a k C l a s s I, a n d / a / in w e a k C l a s s II.
332 B. Elan Dresher

(12) Present Past

Strong: Root + AGR Root (+Ablaut) + AGR
Weak: Root + E X T + A G R Root + E X T + TNS + A G R
E X T = lil in C l a s s I, / a / in C l a s s II; T N S - Idl

Inspection of the various rules I h a v e b e e n d i s c u s s i n g reveals a generalization:

the rules in question fail w h e n their e n v i r o n m e n t crucially involves a g r e e m e n t
( A G R ) ; they d o not fail w h e n their c o n t e x t is the stem, that is, the root with exten-
sion if there is o n e . This is illustrated in (13a,b) for S m o o t h i n g .

(13) a. S t r o n g verbs b. W e a k verbs Nouns

spree wcec-a dceg stem d o m a i n
wceac-a Back Mutation
wceca Smoothing
sprec-an wceca-an dceg-as word domain
spreoc-an dceag-as Back Mutation
FAILS dceg-as Smoothing
spreocan wcecian dcegas other rules

T h e s a m e is true for the other rules, as s h o w n in (14) a n d (15).

(14) a. S t r o n g verbs b W e a k verbs

werp sccerp-i stem d o m a i n
weorp sccearp-i Breaking
sccerp-i /-Monophthongization
weorp-id sccerpi-d-un word domain
Breaking
FAILS /-Monophthongization
weorped scerptun other rules

(15) a. S t r o n g verbs b. W e a k verbs

hceld -hceld-i stem d o m a i n
hald -hald-i Retraction
-hceld-i /-Mutation
hald-id -hceldi-id word domain
Retraction
FAILS /-Mutation
halded -hcelded other rules

W e can a c c o u n t for the failure of the rules at w o r d level by a s s u m i n g that they d o

not operate at that level (cf. Kiparsky, 1985), as in (16).
Anglian Smoothing 333

(16) LEXICAL PHONOLOGY OF M E R C I A N : 4

Morphology Phonology
Roots, Stems B r e a k , Retrct, B c k M u t E-stem
N o u n affixes / - M o n o , /-Mut, S m o o t h Level

Word
Verb A G R B r e a k , Retrct, B c k M u t
Level

S o m e t h i n g further m u s t b e said to a c c o u n t for the fact that n o u n s a n d adjectives

u n d e r g o all the rules even w h e n their c o n t e x t includes inflectional suffixes, as
s h o w n in ( 1 3 c ) . N o u n s a n d adjectives take suffixes indicating n u m b e r , gender, and
c a s e . N o u n inflectional suffixes act like v e r b stem e x t e n s i o n s , in that S m o o t h i n g
applies regularly. Evidently, the g r a m m a r d o e s not treat n o u n inflection on a p a r
with v e r b inflection. T h u s for o u r p u r p o s e s there is a d o m a i n w h i c h i n c l u d e s ev-
e r y t h i n g b u t a g r e e m e n t ( A G R ) , that is, v e r b s t e m s and n o m i n a l s w i t h n u m b e r ,
gender, a n d p e r s o n m a r k i n g s ; I call this the E - S T E M (for e x t e n d e d stem) level, as
s h o w n in ( 1 7 ) .

(17) PHONOLOGICAL DOMAINS:

S t r o n g verbs W e a k verbs Nouns
E - s t e m level: spree wcec-a dceg-as
W o r d level: spree-an wceca-an dcegas

T o s u m u p , a s y n c h r o n i c analysis of the P s ( A ) dialect can a c c o u n t for m u c h of

the e x c e p t i o n a l i t y of strong verbs b y r e c o g n i z i n g t w o lexical levels. T h e rules
of B r e a k i n g , R e t r a c t i o n , and B a c k M u t a t i o n apply at b o t h levels, w h i l e the
rules of S m o o t h i n g , / - M o n o p h t h o n g i z a t i o n , and /-Mutation a p p l y only at the
E-stem Level.

4. R U L E S I N N O N L I N E A R P H O N O L O G Y

It is n o t e w o r t h y that the three rules restricted to the E - s t e m level all share an

interesting p r o p e r t y : they all serve to u n d o the effects of other rules, a fact w h i c h
b e c o m e s a p p a r e n t w h e n w e c o n s i d e r their effects in parallel.

(18) R U L E INTERACTION: E-STEM LEVEL:

a. Input Back Mutation Smoothing
wceca wceaca wceca
b. Input Breaking /-Monophthongization
sccerpi sccearpi sccerpi
334 B. Elan Dresher

c. Input Retraction /-Mutation

-hceldi -haldi -hceldi

T h e parallelism g o e s further, a l t h o u g h it is o b s c u r e d b y the linear format in w h i c h

the rules h a v e b e e n formulated. C o n s i d e r the rule of Retraction (8b): this is a
straightforward b a c k n e s s assimilation, in w h i c h the front low vowel ce is retracted
to a before b a c k s o n o r a n t c o n s o n a n t s . C o m p a r e n o w B r e a k i n g (8a), w h i c h in a
linear format l o o k s quite different. However, d i p h t h o n g i z a t i o n represents the ad-
dition of a b a c k c o m p o n e n t to a front vowel w h e n a b a c k c o n s o n a n t follows. T h i s ,
too, is a variant of b a c k n e s s assimilation. B r e a k i n g and Retraction o c c u r u n d e r
similar c o n d i t i o n s : for e x a m p l e , w h e r e the M e r c i a n dialect h a s Retraction before
l\l + c o n s o n a n t , as in (19a), West S a x o n h a s b r e a k i n g ; h e n c e M e r c i a n aid, but
W e s t S a x o n ceald, b o t h from *celd:

(19) a. BREAKING A N D RETRACTION :

Prim. OE: *celd
Retraction: aid (Mercian)
Breaking: ceald (West S a x o n )
b. a-RESTORATiON A N D B A C K M U T A T I O N :
Prim. OE: *hcefuc
a-Restoration: hafuc (West S a x o n )
Back Mutation: hceafuc (Mercian)

Similar c o n s i d e r a t i o n s h o l d for B a c k M u t a t i o n (2), w h i c h also creates diph-

t h o n g s before b a c k s e g m e n t s : in this case, the triggers are b a c k v o w e l s , and the
rule can apply across an i n t e r v e n i n g c o n s o n a n t . Just as B r e a k i n g is paralleled b y
Retraction, there is a rule in West S a x o n w h i c h o p e r a t e s in the s a m e c o n t e x t as
B a c k M u t a t i o n ; but instead of turning a front vowel into a d i p h t h o n g , it retracts it.
T h i s rule is called a - R e s t o r a t i o n , a n d an e x a m p l e is given in (19b).
It is clear, then, that the d i p h t h o n g i z a t i o n rules in question are a form of b a c k -
ness assimilation. In the c a s e of Retraction (20), the assimilation is direct, in that
the feature [ B A C K ] spreads to the vowel. B r e a k i n g is similar (20b), t h e difference
b e i n g that the s p r e a d i n g [ B A C K ] is adjoined to the vowel, creating a d i p h t h o n g .
In these representations, A represents a low vowel apart from its specification for
[BACK]. 5

(20) a. RETRACTION:

[V] [SON] [V] [SON]

I I I I
A I d • A I d

I \ I
[BACK] [BACK]
Anglian Smoothing 335

b. BREAKING:

[V] [CONT] [V] [CONT]

I I /\ I
A I d A V / d

I \ i
[BACK] [BACK]
Similarly, ^ - R e s t o r a t i o n and B a c k M u t a t i o n exemplify the t w o styles of p r o p a g a -
tion of [ B A C K ] from a following vowel.

(21) a. #-RESTORATION:

[V] [V] [V] [V]

f A t u
I I
f A t u

[BACK] \y
b. B A C K MUTATION:

[BACK]
[V] [V] [V]

I I / A
A V t u

f A t u - [V]
[BACK]

L e t us turn n o w [BACK]to the rules w h i c h apply only at the E - s t e m level of the lexical
p h o n o l o g y . W e o b s e r v e i m m e d i a t e l y that / - M o n o p h t h o n g i z a t i o n a n d /-Mutation
can n o w b e s u b s u m e d u n d e r a single generalization, namely, d e l i n k a specification
[ B A C K ] to the left of lil. T h i s will h a v e the effect of fronting a b a c k v o w e l a n d
voiding the s e c o n d part of a d i p h t h o n g . (Recall that the s e c o n d e l e m e n t of a diph-
t h o n g o c c u p i e s n o slot of its o w n [i.e., h a s n o m o r a i c v a l u e ] , so the loss of [ B A C K ]
c a u s e s it to m e l d into the front vowel.) W e will call this unified rule / - U m l a u t ; its
effects are illustrated in (22).

(22) /-UMLAUT:

a. /-Mutation: b. /-Monophthongization:

A
h A l d i sc A V r p i

N
[BACK] 0
N
[BACK] 0
336 B. Elan Dresher

T h i s leaves S m o o t h i n g . U p to now, all the rules w e have b e e n l o o k i n g at have

involved s o m e kind of b a c k n e s s assimilation. B y contrast, S m o o t h i n g r e m o v e s a
b a c k d i p h t h o n g a l e l e m e n t before a b a c k c o n s o n a n t and so appears to b e a dissimi-
lation rule, a p p l y i n g m o r e o v e r , in s o m e of the s a m e c o n t e x t s w h i c h c a u s e b a c k i n g .
H o g g ( 1 9 9 2 : § § 9 3 - 1 0 2 ) , in an interesting discussion of this issue, p r o p o s e s
that the c o n s o n a n t s w h i c h c a u s e d S m o o t h i n g w e r e b y this t i m e n o longer velar but
palatal, specified [ - b a c k ] . S m o o t h i n g is thus b r o u g h t into the fold of natural as-
similation rules: m o n o p h t h o n g i z a t i o n is treated as a type of fronting, triggered by
a following [ — b a c k ] feature.
T h e r e are a n u m b e r of p r o b l e m s with this a p p r o a c h , however. First, H o g g ' s
analysis n e e d s a rather b a r o q u e m e c h a n i s m of feature spreading to spread the fea-
ture [ - b a c k ] from the first part of the d i p h t h o n g o n t o the following velar c o n s o -
nant, vaulting over the intervening [ + b a c k ] of the s e c o n d part of the d i p h t h o n g .
S e c o n d , there is n o clear e v i d e n c e that the required palatalization ever occurred.
C o u n t i n g against it is the fact that in the s a m e p e r i o d w h e n these c o n s o n a n t s w e r e
c a u s i n g S m o o t h i n g , they w e r e also b e g i n n i n g to exercise a b a c k i n g influence on
low v o w e l s , as s h o w n b y the f o r m s in (23) from P s ( A ) .

(23) I N C I P I E N T G E N E R A L I Z E D R E T R A C T I O N BEFORE VELARS [ P S ( A ) ]

N u m b e r of t o k e n s with
Form ce e a
mceht 18 3
dceg 41 6 1
asagas — 1
mcegan 5 1
*plcegian 2 1
ah — 44

W e find the w o r d mceht 18 t i m e s with the e x p e c t e d ce, b u t 3 t i m e s with a; asagas

appears o n c e , and it has a before g rather than the e x p e c t e d ce, and so on. T h e
n u m b e r of instances of a is not great, b u t there are m o r e than w e w o u l d e x p e c t as
simply the result of scribal error. T h e s e o c c u r r e n c e s gain in significance w h e n w e
c o n s i d e r that in the M i d d l e E n g l i s h dialect of The Life and the Passion of Ste.
Juliana, c o n s i d e r e d a d e s c e n d a n t of the P s ( A ) dialect, w e find that earlier ce gen-
erally b e c a m e a before b a c k c o n s o n a n t s : cwakien ' q u a k e ' [Ps(A) cwcecian], ahte
' w e a l t h ' [Ps(A) cehte], mahte ' m i g h t ' [Ps(A) mceht\, and so on ( d ' A r d e n n e 1961 :
182). Therefore w e can c o n s i d e r the sporadic a spellings to indicate the b e g i n n i n g
of the generalization of Retraction: originally triggered by b a c k sonorants, R e t r a c -
tion is b e i n g generalized to o c c u r before all b a c k c o n s o n a n t s . T h i s a c c o u n t of
G e n e r a l i z e d R e t r a c t i o n requires that the s a m e c o n s o n a n t s w h i c h are r e s p o n s i b l e
Anglian Smoothing 337

for S m o o t h i n g r e m a i n e d velar. T h e r e f o r e a s s u m i n g an i n t e r m e d i a t e stage of pala-

talized c o n s o n a n t s to a c c o u n t for S m o o t h i n g will prevent us from g i v i n g a natural
a c c o u n t of G e n e r a l i z e d Retraction.
Let us return, then, to the idea that the c o n s o n a n t s c a u s i n g S m o o t h i n g did in-
d e e d r e m a i n velar, that is, [ B A C K ] , a n d r e c o n s i d e r rather the e q u a t i o n of m o n o -
p h t h o n g i z a t i o n with fronting. W h i l e it is true that d i p h t h o n g i z a t i o n is a species of
b a c k i n g , it d o e s not thereby follow that m o n o p h t h o n g i z a t i o n is always a species
of fronting. It is i m p o r t a n t to n o t e that, unlike /-Umlaut, the c o n s o n a n t s w h i c h
c a u s e S m o o t h i n g d o not ever c a u s e fronting of a v o w e l . In t e r m s of n o n l i n e a r
r e p r e s e n t a t i o n s , S m o o t h i n g never c a u s e s the feature [ B A C K ] to b e erased; rather,
it simply delinks the feature from a d i p h t h o n g , c a u s i n g the loss of its s e c o n d part,
as s h o w n in (24).

(24) [V] [OBST] [V] [OBST]

/\ I s A x
s A V x

N [BACK]
[BACK]
T h i s d e l i n k i n g of a multiply-linked feature is a k i n d of O C P effect, in that it sim-
plifies the transition from a front to a b a c k e l e m e n t . T h i s is n o t an u n n a t u r a l
6

p r o c e s s , a n d so it is n o t n e c e s s a r y to s u p p o s e that S m o o t h i n g before a b a c k c o n -
sonant is p a r a d o x i c a l or p r o b l e m a t i c .
7

5. R U L E S O R C O N S T R A I N T S

H a v i n g n o w recast the various rules into n o n l i n e a r t e r m s , let us return to the

E - s t e m level a n d the rules that are limited to that level. I b e g a n w i t h three rules. I
then united the t w o rules c o n d i t i o n e d b y l\l into o n e rule of / - U m l a u t . I can n o w
g e n e r a l i z e the parallelism in the derivations in (18): in e a c h c a s e a rule w h i c h
spreads [ B A C K ] to a stressed vowel is reversed, the [ B A C K ] feature b e i n g
delinked.

(25) Input Add BACK Remove BACK

a. wceca wceaca wceca
b. sccerpi sccearpi sccerpi
c. -hceldi -haldi -hceldi

T h e nature of t h e s e derivations suggests a reanalysis: instead of s p r e a d i n g

[ B A C K ] and then d e l i n k i n g it, w h y not b l o c k the s p r e a d i n g to b e g i n w i t h ? In 8

classical g e n e r a t i v e p h o n o l o g y , there are a n u m b e r of w a y s of d o i n g this, b u t they

338 B. Elan Dresher

all h a v e d r a w b a c k s . O n e w a y is to simply e x c l u d e the restricted c o n t e x t from the

rule directly, but this usually requires a m o r e c o m p l i c a t e d rule. In this case, w e
c a n stipulate in the formulation of B a c k M u t a t i o n , for e x a m p l e , that spreading
m a y not o c c u r across a b a c k c o n s o n a n t , thus incorporating S m o o t h i n g into the
p r o c e s s . B u t this not only c o m p l i c a t e s the rule; it also m i s s e s the generalization
that n o d i p h t h o n g s from any source are allowed in that context.
A n o t h e r a p p r o a c h is to i n v o k e m e t a r u l e s , persistent rules, filters, constraints, or,
as they w e r e o n c e called, conspiracies. W o r k in lexical p h o n o l o g y h a s suggested
a n u m b e r of constraints o n derivations, such as structure preservation, and has
articulated a notion of levels or d o m a i n s within w h i c h such constraints are in force
or are turned off. F r o m a n o t h e r direction, the representations and operations of
n o n l i n e a r p h o n o l o g y lend t h e m s e l v e s quite naturally to a theory in w h i c h configu-
rational constraints and w e l l - f o r m e d n e s s c o n d i t i o n s play a large r o l e . 9

L e t us return, then, to the rules w h i c h are restricted to the E - s t e m level, n a m e l y

S m o o t h i n g and /-Umlaut. W e o b s e r v e that w e c a n reinterpret t h e m as constraints,
w h e r e b y the configurations in (26) are ruled out.

(26) a. Smoothing b. /.-Umlaut

* [V] [OBST] * [BACK] ... i

[BACK]

W e a s s u m e that these constraints serve to b l o c k any rule that w o u l d create a vio-

lation of t h e m . T h e s e constraints hold in the lexicon at the E - s t e m level b u t are
turned off at the W o r d level, d e r i v i n g the facts w e o b s e r v e in P s ( A ) .

6. L E X I C A L P H O N O L O G Y A N D D I A C H R O N Y

W e h a v e thus arrived at an analysis w h i c h gives a better a c c o u n t of the syn-

c h r o n i c facts as w e find t h e m , a n d the stage is n o w set for a better u n d e r s t a n d i n g
of the d i a c h r o n i c d e v e l o p m e n t s as well. T h e following scenario suggests itself:
S m o o t h i n g , w h a t e v e r its original status, c a m e to b e interpreted as a constraint in
the lexical p h o n o l o g y , that is, as a persistent rule, w h i c h like other such rules
r e m a i n s in force u p to the W o r d level. B a c k M u t a t i o n w a s then a d d e d to the g r a m -
m a r ; if it w a s not a lexical rule to b e g i n with, it s o o n b e c a m e o n e . A s such, it w a s
subject to the S m o o t h i n g constraint u p to the W o r d level but w a s u n i m p e d e d by it
thereafter.
Anglian Smoothing 339

7. CONCLUSION

T h i s analysis m a k e s sense of the s y n c h r o n i c a n d d i a c h r o n i c facts. M a n y q u e s -

tions r e m a i n , however, a n d I only raise three of t h e m here.
First, it w o u l d b e desirable to find i n d e p e n d e n t e v i d e n c e for the E - s t e m d o -
m a i n , in t e r m s of either cross-linguistic or l a n g u a g e - p a r t i c u l a r e v i d e n c e from O l d
English.
A m o r e general q u e s t i o n c o n c e r n s persistent rules, w h i c h h a v e b e e n a p p e a r i n g
m o r e a n d m o r e in the literature. Basically, the difference b e t w e e n a rule and a
persistent rule, or constraint, is that a rule h a s o n e c h a n c e to apply in a cycle, w h i l e
a constraint can apply after, or d u r i n g , every other rule. B u t in a theory with b o t h
k i n d s of rules, h o w d o w e distinguish o n e from t h e o t h e r ? W h y not m a k e all t h e
b a c k i n g rules constraints also, say * [ B A C K ] u n l i n k e d to stressed V in certain
configurations? O f c o u r s e , if e v e r y t h i n g is a constraint, w e c a n n o l o n g e r a c c o u n t
for the a p p a r e n t o r d e r i n g , or the differences in level of application ( E - s t e m and
W o r d levels for b a c k i n g rules, E - s t e m only for the others). W e o b s e r v e that, as
formulated, t h e difference is b e t w e e n rules that spread a feature a n d rules that d o
not. P e r h a p s that plays a role in m a k i n g the distinction.
Third, the constraints in (26) have b e e n cast as filters, b l o c k i n g d e v i c e s w h i c h
prevent certain configurations from arising. It is also p o s s i b l e to h a v e persistent
rules in the sense of M y e r s ( 1 9 9 1 ) : these h a v e the format of rules but apply w h e n -
ever their structural description is m e t . R a t h e r than b l o c k t h e application of other
rules, s u c h rules act to repair any constraint violations as soon as they occur. It is
difficult to d e t e r m i n e w h e t h e r the constraints in (26) h a v e a b l o c k i n g or repair
function, as t h e t w o m o d e s of application lead to the s a m e results in the cases
discussed here.
D e s p i t e t h e s e u n r e s o l v e d issues, it is n e v e r t h e l e s s clear that c u r r e n t t h e o r i e s
offer n e w a n d interesting p e r s p e c t i v e s on d i a c h r o n i c c h a n g e .

ACKNOWLEDGMENTS

I would like to thank Peter Avery, Alana Johns, Keren Rice, Tom Wilson, and an anony-
mous referee for helpful comments.

NOTES

For more on the language and texts, see Brunner (1965), Campbell (1959), Dresher
l

(1978), Hogg (1992), Kuhn (1965), Luick (1964), and Toon (1983).
340 B. Elan Dresher

2
N o t e that Old English texts typically spell the low diphthong [aea] as ea. There is little
doubt that this was purely an orthographic convention, and for ease of exposition, I indicate
it as cea.
3
See Kiparsky (1988) for a review of Neogrammarian and contemporary issues in pho-
nological change.
4
T h e diagram allows for cyclic levels, but I have not shown this; the Word Level espe-
cially may not be cyclic.
5
The representations in (20) and subsequently are schematic, and much structure is
omitted. Gussenhoven and van de Weijer (1990) consider the spreading feature to be
[+dorsal].
6
Strictly speaking, this is what McCarthy (1989) calls anti-spreading, which often works
together with the OCP to enforce cooccurrence restrictions.
7
An anonymous referee suggests that Smoothing may have begun in fast speech, where
the listener might be likely to interpret the second part of the diphthong as a mere formant
transition to the consonant. Translating this perception backward to phonemic representa-
tions gives the effect of monophthongization. It is noteworthy in this connection that Brun-
ner (1965) proposed that Smoothing is orthographic only, because the backness of the sec-
ond part of the diphthong was already conveyed by the following consonant. I see no
reason, however, to limit this effect to the orthography.
8
We thus return in a way to Luick's suggestion that Back Mutation was prevented before
velar consonants.
9
See Singh (1987), Paradis (1988), Goldsmith (1990: Ch. 6), and Myers (1991) for
recent proposals.

REFERENCES

d'Ardenne, S. R. T. O. (1961). be liflade ant te passiun of seinte luliene (Early English

Text Society 248). Oxford University Press, London.
Ball, C. J. E., and Stiles, P. (1983). The derivation of Old English geolu 'yellow', and the
relative chronology of smoothing and back-mutation. Anglia 101, 5 - 2 8 .
Brunner, K. (1965). Altenglische Grammatik: Nach der Angelsachsischen Grammatik von
Eduard Sievers, 3rd ed. Niemeyer, Tubingen.
Campbell, A. (1959). Old English Grammar. Clarendon, Oxford.
Chomsky, N., and Halle, M. (1968). The Sound Pattern of English. Harper and Row, New
York.
Dresher, B. E. (1978). Old English and the Theory of Phonology. Doctoral dissertation,
University of Massachusetts, Amherst.
Goldsmith, J. A. (1990). Autosegmental and Metrical Phonology. Blackwell, Oxford.
Gussenhoven, C , and van de Weijer, J. (1990). On V-place spreading vs. feature spreading
in English historical phonology. The Linguistic Review 7, 311 - 3 3 2 .
Hogg, R. M. (1992). A Grammar of Old English. Vol. I: Phonology. Blackwell, Oxford.
Kiparsky, P. (1973). Abstractness, opacity, and global rules. Part 2 of "Phonological Re-
Anglian Smoothing 341

presentations." In Three Dimensions of Linguistic Theory (O. Fujimura, ed.),

pp. 5 7 - 8 6 . TEC, Tokyo.
Kiparsky, P. (1985). Some consequences of lexical phonology. Phonology Yearbook 2,
85-138.
Kiparsky, P. (1988). Phonological change. In Linguistics: The Cambridge Survey. Vol. 1:
Linguistic Theory: Foundations (F. J. Newmeyer, ed.), pp. 3 6 3 - 4 1 5 . Cambridge Uni-
versity Press, Cambridge.
Kuhn, S. M. (1965). The Vespasian Psalter. University of Michigan Press, Ann Arbor.
Luick, K. (1964). Historische Grammatik der englischen Sprache. Blackwell, Oxford.
McCarthy, J. J. (1989). Guttural Phonology. Unpublished manuscript, University of Mas-
sachusetts, Amherst.
Myers, S. (1991). Persistent rules. Linguistic Inquiry 22, 3 1 5 - 3 4 4 .
Paradis, C. (1988). On constraints and repair strategies. The Linguistic Review 6, 7 1 - 9 7 .
Singh, R. (1987). Well-formedness conditions and phonological theory. In Phonologica
1984 (W. U. Dressier et al., eds.). Cambridge University Press, Cambridge.
Toon, T. E. (1983). The Politics of Early Old English Sound Change. Academic Press, New
York.
RULE REORDERING AND RULE
GENERALIZATION IN LEXICAL PHONOLOGY:
A RECONSIDERATION

ELLEN M. KAISSE
Department of Linguistics
University of Washington
Seattle, Washington 98195

1. INTRODUCTION

H o w d o g r a m m a r s c h a n g e from o n e g e n e r a t i o n to the n e x t ? T h e q u e s t i o n en-

g a g e d g e n e r a t i v e linguists d e e p l y in the late 1960s and early 1970s. Interest in the
theoretical characterization of s o u n d c h a n g e h a s since d i m i n i s h e d , as the larger
investigative p a r a d i g m h a s turned from rules, derivations, abstractness, a n d u n -
derlying f o r m s to p h o n o l o g i c a l representations and constraints. B u t m a n y i m p o r -
tant and, in s o m e cases, surprising results w e r e achieved in the seventies a n d it is
t i m e to r e c o n s i d e r t h e m in the light of a p a r a d i g m w h i c h h a s m u c h n e w to tell us
a b o u t rule o r d e r i n g a n d rule t y p e s : the t h e o r y of lexical p h o n o l o g y , w h i c h f o r m s
the subject of this v o l u m e . T h e i m p e t u s for m y r e e x a m i n a t i o n of historical p h e -
n o m e n a c o m e s from K i p a r s k y ' s (1988) article surveying p h o n o l o g i c a l c h a n g e
from the p e r s p e c t i v e of recent d e v e l o p m e n t s in p h o n o l o g i c a l theory. In the latter
part of this article, K i p a r s k y resolves the " N e o g r a m m a r i a n c o n t r o v e r s y " in a very
satisfying w a y : h e a r g u e s that c h a n g e s in postlexical rules will result in s o u n d
c h a n g e s w h i c h are regular, p h o n o l o g i c a l l y m o t i v a t e d , and in general o b e d i e n t to
the N e o g r a m m a r i a n h y p o t h e s i s . B u t c h a n g e s a n d additions to lexical rules will
yield s o u n d c h a n g e s with e x c e p t i o n s , sensitivity to w o r d structure, and so on, the
h a l l m a r k s of lexical diffusion. If lexical p h o n o l o g y can strike to the h e a r t of issues
like this, w h i c h h a v e c a u s e d c o n s t e r n a t i o n a n d strife a m o n g linguists for over a

343
Phonetics and Phonology, Volume 4 Copyright © 1993 by Academic Press, Inc.
Studies in Lexical Phonology All rights of reproduction in any form reserved.
344 Ellen M. Kaisse

century, it is o b v i o u s l y t i m e to r e c o n s i d e r the other things w e h a v e learned about

historical c h a n g e in the t e r m s of the theory. I m a k e a small b e g i n n i n g here.
I c o n s i d e r t w o results of the d i a c h r o n i c investigations of the last twenty years
and a t t e m p t to refine or explain t h e m within the theory of lexical p h o n o l o g y . T h e
first, w h i c h I refine, is the result that rules can be reordered with respect to o n e
a n o t h e r a n d thus can b e found out of their c h r o n o l o g i c a l l y correct a r r a n g e m e n t .
W e shall see that lexical p h o n o l o g y , with its ability to c h a r a c t e r i z e a rule's position
in the p h o n o l o g y by its characteristics, allows us a w i n d o w into h o w and w h e r e
rules m o v e . Investigators of t w o d e c a d e s a g o c o u l d not ask s o m e rather interesting
questions about rule r e o r d e r i n g , namely, w h i c h rule is m o v i n g , a n d w h e r e exactly
is it e n d i n g u p ? B u t the divisions b e t w e e n strata and b e t w e e n lexical and postlex-
ical rules n o w allow us to identify a rule's location not only by its order with
respect to o t h e r rules but b y a w h o l e suite of o t h e r characteristics. W e can n o w
d e t e r m i n e if a rule has m o v e d from o n e stratum to a n o t h e r or even, in the case
of a r e o r d e r i n g in C y p r i o t G r e e k , from the lexical c o m p o n e n t to the postlexical
c o m p o n e n t or the w o r d level.
T h e s e c o n d result that I r e e x a m i n e is the s o m e w h a t astonishing conclusion,
i n d e p e n d e n t l y arrived at by t w o investigators ( B l e y - V r o m a n , 1975; R o b i n s o n ,
1976), that w h e n a rule g e n e r a l i z e s , it d o e s not d o so in p l a c e , but rather at the e n d
of the g r a m m a r , leaving its u n g e n e r a l i z e d p r o g e n i t o r b e h i n d in its original posi-
tion. R o b i n s o n t e r m s a rule w h i c h h a s generalized in this w a y SCATTERED. I show
that this apparently p a r a d o x i c a l rule ordering, with its o d d d i s p l a c e m e n t of the
g e n e r a l i z e d form of a rule, in fact flows rather naturally from the c o n c e p t i o n of
the g r a m m a r that is central to the lexical p h o n o l o g y m o d e l and is related to the
increased access to feature specifications that rules m a y h a v e as they apply at later
levels. O n c e w e h a v e established that rules can m o v e " d o w n " in the g r a m m a r , and
even out of the lexicon, w e shall h a v e an e x p l a n a t i o n for w h y rules d o not a p p e a r
to generalize in place. Ultimately, w e can h o p e to explain a large class of rule
generalizations as the result of r e o r d e r i n g and the p r o g r e s s i v e specification of fea-
tures. M o r e o v e r , w e will h a v e a clearer view of the p r o g r e s s of a rule t h r o u g h the
g r a m m a r over time.
T h e direction of m o v e m e n t of a rule over t i m e h a s b e e n w i d e l y and correctly
agreed to b e " u p w a r d " (see Kiparsky, 1984, 1988, and references cited in the
latter, and Z e e , this v o l u m e ) : s o u n d c h a n g e s begin as variable rules of p h o n e t i c
i m p l e m e n t a t i o n , are gradually g r a m m a t i c i z e d as postlexical rules, and m o v e into
the lexicon as they are i n c o r p o r a t e d into the g r a m m a r , acquiring e x c e p t i o n s , ref-
e r e n c e to m o r p h o l o g i c a l information, a cyclic m o d e of application, and so forth.
W h e n w e look in detail at the history of a rule of c o n t i n u a n c y dissimilation in
C y p r i o t G r e e k , w e shall see an e x a m p l e of a rule that, h a v i n g u n d e r g o n e the usual
u p w a r d p r o g r e s s i o n , h a s n o w m o v e d D O W N again in the g r a m m a r , t h o u g h reor-
d e r i n g . U s i n g the theory of underspecification, w e shall see that a reordered rule
s h o u l d b e e x p e c t e d to c o m e to refer to feature values unavailable in the early strata
Rule Reordering and Generalization 345

of the lexicon ( p e r h a p s available only postlexically). Finally, w e briefly r e e x a m i n e

o n e p u b l i s h e d c a s e of rule " s c a t t e r i n g " ( R o b i n s o n , 1976) a n d o n e less familiar
o n e (Kaisse, 1976) in t e r m s of underspecification and lexical p h o n o l o g y a n d c o n -
c l u d e that m o v e m e n t to a later s t r a t u m or c o m p o n e n t c a n e x p l a i n the p h e n o m e n o n
of rule generalization that leaves b e h i n d the old version of the r u l e . 1

It is possible that the t w o directions of m o v e m e n t for rules that I postulate can

r e c o n c i l e t h e o p p o s i t e c l a i m s of K i p a r s k y ( 1 9 8 4 ) a n d H a l l e a n d M o h a n a n ( 1 9 8 5 )
with respect to the strong d o m a i n h y p o t h e s i s . Halle and M o h a n a n (p. 58) state the
following " P r i n c i p l e s of D o m a i n A s s i g n m e n t " :

a. In the absence of counterevidence, assign the smallest number of strata as the do-
main of a rule.
b. In the absence of counterevidence, assign the highest possible stratum as the do-
main of a rule (where 'lowest' = stratum 1).

G i v e n t h e s e a s s u m p t i o n s , in the u n m a r k e d c a s e all p h o n o l o g i c a l rules apply at the

postlexical s t r a t u m (b) a n d only at that s t r a t u m (a). W e d o not k n o w at p r e s e n t
w h e t h e r there in fact exist l a n g u a g e s w h e r e all p h o n o l o g i c a l rules are restricted to
the postlexical stratum. A footnote r e a d s :

Principles la,b are of course not the only ones conceivable. Alternatively, one might think
of the following principles: (a) In the absence of counterevidence, assign the maximum
number of strata as the domain of a rule, (b) In the absence of counterevidence, assign the
lowest stratum as the domain of a rule.

T h e footnote implies that H a l l e a n d M o h a n a n h a v e c h o s e n their p r i n c i p l e s of d o -

m a i n a s s i g n m e n t w i t h s o m e uncertainty. In any case, they g i v e neither e m p i r i c a l
not theoretical a r g u m e n t s for the c h o i c e they h a v e m a d e . In fact, K i p a r s k y ( 1 9 8 4 :
141ff.) m a k e s a l m o s t precisely the o p p o s i t e c h o i c e , a s s u m i n g that the u n m a r k e d
rule applies at Level 1 a n d that rules d o not b e l o n g to particular d o m a i n s unless
explicitly restricted.
T h e s e issues, I w o u l d argue, c a n n o t be u n d e r s t o o d n o r r e s o l v e d w i t h o u t taking
historical factors into a c c o u n t . Halle a n d M o h a n a n are a l m o s t certainly c o r r e c t if
w e are talking a b o u t rules n e a r the b e g i n n i n g of their histories. K i p a r s k y indeed
agrees that n e w rules are a d d e d as e x c e p t i o n l e s s postlexical o n e s . B u t as a rule
ages, it e x t e n d s its d o m a i n u p w a r d to earlier strata within the lexicon, p e r h a p s
curtailing its postlexical application as it m o v e s . T h u s the e n d p o i n t of a p h o n o -
logical rule is to b e e x p e c t e d to b e stratum 1, with the earlier strata the rule p a s s e d
t h r o u g h p o s s i b l y also subject to its action. T h u s for o l d e r rules, K i p a r s k y is p r o b -
ably correct. M o r e o v e r I s u b m i t that parts (a) and (b) of H a l l e and M o h a n a n ' s
principles are i n d e p e n d e n t : w e d o not k n o w the c i r c u m s t a n c e s u n d e r w h i c h a rule
will m a i n t a i n its c h r o n o l o g i c a l l y earlier, low-level d o m a i n s of application w h i l e
a d d i n g lexical o n e s . A rule m i g h t e x t e n d its d o m a i n u p w a r d w h i l e s i m u l t a n e o u s l y
shrinking its applicability in later strata. Similarly, rules m i g h t r e - d e s c e n d t h r o u g h
the g r a m m a r , either e x t e n d i n g their d o m a i n s , so as to apply in early a n d late strata
346 Ellen M. Kaisse

as well, or, instead, forsaking the older, earlier strata. T h e first type of descent
should result in " s a n d w i c h " t y p e rule application, w h e r e a n e w l y e x t e n d e d rule
can apply b o t h before a n d after a n o t h e r rule. T h i s is a characteristic part of the
" s c a t t e r e d " rule p h e n o m e n o n . T h e s e c o n d t y p e of d e s c e n t will result in a rule
w h i c h w e d e s c r i b e as h a v i n g " m o v e d " or " r e o r d e r e d , " for instance the C y p r i o t
c a s e of Section 3.

2. E X T E N S I O N D O W N W A R D O F M O R P H O L O G I C A L D O M A I N :
KASKA ^-DELETION

Before w e turn to our central e x a m p l e from C y p r i o t G r e e k , let us c o n s i d e r a

similar b u t s i m p l e r c a s e (since it involves only o n e rule) from the A t h a b a s k a n
l a n g u a g e K a s k a ( H a r g u s , p e r s o n a l c o m m u n i c a t i o n 1989, b a s e d on data from
M o o r e , 1988). L i k e other A t h a b a s k a n l a n g u a g e s , K a s k a h a s an intricate set of
prefixes w h i c h A t h a b a s k a n i s t s divide into t w o sets: disjunct prefixes (those far-
thest from the stem, consisting largely of adverbial material) and conjunct pre-
fixes, those closer to the stem, w h i c h include a g r e e m e n t a n d tense m o r p h e m e s .
T h e conjunct prefixes are further s u b d i v i d e d into sets, partially o n the basis of
their p h o n o l o g i c a l behavior. ( M o s t A t h a b a s k a n i s t s r e c o g n i z e t w o divisions within
the conjunct prefixes, b u t H a r g u s a r g u e s for three in Sekani.) H a r g u s (1988) ar-
g u e s that the o u t e r m o s t set of Sekani conjunct prefixes is a d d e d at level IV w h i l e
the inner sets are a d d e d at levels II a n d III.

(1) y [disjunct I V [conjunct m [conjunct n [conjunct j [stem

In all the l a n g u a g e s of the family e x c e p t K a s k a , a rule of intervocalic deletion

w h i c h deletes the Is/ of the conjunct prefix *?s(9) ( > s(o) p l u s a distinctive t o n e
in the d a u g h t e r l a n g u a g e s ) applies only if the / s / a n d t h e d e t e r m i n a n t v o w e l s o c c u r
within stratum II; vowel-final prefixes a d d e d later, at stratum III, IV, or V d o not
supply the intervocalic e n v i r o n m e n t for the rule to operate. T h u s in the following
e x a m p l e s from S e k a n i , / s / is deleted in (2a,b) but not in (2c); in the former cases,
the t h e m a t i c prefix / z o / and the inceptive prefix / d o / are s h o w n b y H a r g u s (1988)
to fall within the level II m o r p h o l o g y , w h i l e in the last case, the p r e c e d i n g prefix,
/ n o / ' 2 - s g . o b j e c t ' , is in the level III conjunct p r e f i x e s . 2

(2) a. /zo - V i - h - x j / —> [zehxj] T kill (sg. obj.)'

thm-cnj-lsPf-clf-kill
b. / d o - V i - y a / —» [deya] T started to w a l k '
incp-cnj-lsPf-go
c. / n o - V i - c ' o / —» [ n o s i c ' p ] T shot you d e a d '
2sObj-cnj-1 sPf-shoot'
Rule Reordering and Generalization 347

T h u s the extent of ^-Deletion's d o m a i n in Sekani a n d the other A t h a b a s k a n lan-

g u a g e s is as illustrated in (3).

(3) I. stem
S-Deletion
II. conjunct prefixes
III. conjunct prefixes
IV. conjunct prefixes
V. disjunct prefixes

In K a s k a , however, the d o m a i n of ^-deletion is e x t e n d e d to the e n v i r o n m e n t b e -

t w e e n level III conjunct prefixes a n d the following material as well. Object p r o -
n o u n s in level III n o w trigger deletion of a following conjunct / s / . T h e critical
c o m p a r i s o n is b e t w e e n (2c) a n d the K a s k a c o g n a t e in (4c). (4a,b) s h o w that the
K a s k a rule c o n t i n u e s to apply, as in S e k a n i , b e t w e e n a t h e m a t i c level II prefix and
the conjunct prefix, w h i c h M o o r e transcribes as / s e / . ( T h e [s] w h i c h a p p e a r s in
(4a,c) is the first p e r s o n singular marker. I i n c l u d e a s e c o n d p e r s o n f o r m in (4b)
for c o m p a r i s o n . )

(4) a. /de-'se-s-h-hln/—> deshin

thm-cnj-lsSub-clf-kill T kill sg. object'
b. d e - ' s e - n - t s l n / —> dentsin
thm-cnj-2sSub-make ' y o u m a d e it for y o u r s e l f
c. /ne-^se-'s-un'/—> nes'un
2sgObj-cnj-lsgSub-shoot T shot y o u '

S o w e see that it is p o s s i b l e for a p h o n o l o g i c a l rule to reverse the n o r m a l h i s -

torical p r o g r e s s i o n u p w a r d to earlier strata a n d e x t e n d its d o m a i n . Indeed, in a
survey of such c h a n g e s in A t h a b a s k a n , H a r g u s (personal c o m m u n i c a t i o n , 1989)
finds n o d o m a i n s h r i n k i n g , several rules w h o s e d o m a i n h a s b e e n e x t e n d e d , several
w h i c h are lost, and a few w h i c h h a v e a c q u i r e d m o r p h o l o g i c a l restrictions but not
via d o m a i n shrinking. In fact, extension of a rule's m o r p h o l o g i c a l d o m a i n m a k e s
sense in t e r m s of an increase in t r a n s p a r e n c y : the rule applies increasingly w h e r e
its p h o n o l o g i c a l structural description is met. W e shall see that t r a n s p a r e n c y is
also m a x i m i z e d in the C y p r i o t dialect by a r e o r d e r i n g into feeding order.
D u m a s ' s (1981) a c c o u n t of the d e v e l o p m e n t of vowel l e n g t h e n i n g in Q u e b e c o i s
F r e n c h a p p e a r s to b e a n o t h e r e x a m p l e of the d o w n w a r d m o v e m e n t of a p h o n o -
logical rule, in this c a s e into the postlexical c o m p o n e n t . W h i l e standard F r e n c h
3

lengthens the class of v o w e l s w h i c h D u m a s t e r m s " t e n s e " (higher m i d v o w e l s and

nasalized vowels) in c l o s e d syllables only, Q u e b e c o i s l e n g t h e n s these v o w e l s in
any syllable whatsoever, s o long as the v o w e l is not absolutely sentence-final.
T h u s the rule has generalized, in losing the closed syllable r e q u i r e m e n t from its
e n v i r o n m e n t and, crucially, has b e c o m e postlexical, since information about p o -
348 Ellen M. Kaisse

sition in the utterance is not available within the lexicon. W e see the logical c o n -
nection b e t w e e n generalization a n d d o w n w a r d m o v e m e n t in later sections.

3. R U L E R E O R D E R I N G A N D G E N E R A L I Z A T I O N IN
CYPRIOT GREEK

3.1. Continuant Dissimilation

A fairly simple generalization g o v e r n s the distribution of obstruent clusters in

all M o d e r n G r e e k dialects: adjacent o b s t r u e n t s m a y not agree in continuancy.
T h e r e is m u c h to say a b o u t this rule (see Kaisse, 1989), b u t for o u r p u r p o s e s , all
w e n e e d to k n o w is that it takes a n y s e q u e n c e of stop p l u s stop o r of fricative plus
fricative a n d turns it into a s e q u e n c e of fricative p l u s stop. M o r e o v e r , in the rare
cases w h e r e a stop p l u s fricative s e q u e n c e arises, both c o n t i n u a n c y values are
c h a n g e d , yielding fricative plus stop; this fact suggests the e x i s t e n c e of a t e m p l a t e
i m p o s e d as in (5) and indicates that the rule is not only dissimilatory, t h o u g h this
is certainly the effect it h a s in the vast majority of cases.

(5)

FRICATIVE + STOP

O n e ' s initial g u e s s m i g h t b e that C o n t i n u a n t D i s s i m i l a t i o n is a syllable-based ad-

j u s t m e n t in sonority, but b e c a u s e it h a p p e n s in onset clusters as well as b e t w e e n
syllables, I reject this idea.
Since the subrules affecting stops a n d affecting fricatives c a n n o t b e collapsed
using n o r m a l abbreviatory d e v i c e s , and since, as I noted, s e q u e n c e s of stop plus
fricative are also altered to fricative p l u s stop, I a r g u e in K a i s s e ( 1 9 8 9 ) that the
rule should b e formalized as in (6).

(6) a c o n t fteont —> + c o n t —cont

H e r e are e x a m p l e s of the application of the e x p a n s i o n of the rule w h e r e both a

a n d P are set to m i n u s , that is, w h e r e t w o stops c o m e together.

(7) a. ek-timo —»extimo 'esteem'

out-honor
b. pep-to -> pefto 'fall'
fall-vbl
d. *okto > oxto 'eight'
Rule Reordering and Generalization 349

Fricativization of stops is clearly a lexical rule in standard S p o k e n M o d e r n G r e e k ,

for it applies differently in different m o r p h o l o g i c a l e n v i r o n m e n t s and has e x c e p -
tions in s o m e of t h e m . T h e rule is obligatory b e t w e e n stem and suffix and has
applied invariably as a historical c h a n g e within m o r p h e m e s . ( T h e fact that it is
thus a m o r p h e m e structure c o n d i t i o n on w o r d s of the s p o k e n l a n g u a g e thus argues
for both values of c o n t i n u a n c y b e i n g available in u n d e r l y i n g representations.)
H o w e v e r , b e t w e e n prefix and stem, fricativization applies to only s o m e w o r d s .
C o n s i d e r the realizations of the prefix / e k / exemplified in (7a). W h i l e the [x] there
is the preferred colloquial p r o n u n c i a t i o n , p h o n o l o g i c a l l y similar w o r d s like ektelo
'carry o u t ' c a n n o t b e spirantized for m y informant. (I w o u l d e x p e c t j u s t w h i c h
w o r d s will and will not spirantize to differ a m o n g speakers to s o m e extent.)
T h e operation of Despirantization, that is, the e x p a n s i o n of C o n t i n u a n t D i s -
similation w h e n a a n d (3 are set to p l u s , is illustrated in (8).

(8) a. yraf-dike -> yraftike 'it w a s written'

b. ef-xaristo —» efkaristo 'thank you'

A g a i n , the rule o p e r a t e s differently at p r e f i x - s t e m than at s t e m - s u f f i x b o u n d a r y .

It is obligatory for all speakers (in colloquial speech) in the f o r m e r e n v i r o n m e n t ,
but only s o m e s p e a k e r s d o it b e t w e e n prefix and stem. ( T h u m b , 1964, reports the
form I give for ' t h a n k y o u ' and I h a v e h e a r d this myself, but m y major i n f o r m a n t
s h u d d e r s at it.) T h e despirantization portion of C o n t i n u a n t D i s s i m i l a t i o n m a y b e
lexical b y the test of failure to apply b e t w e e n w o r d s , but o n e is hard p r e s s e d to
find e x a m p l e s w h i c h can a r g u e this o n e w a y or the other. Unhappily, G r e e k w o r d s
d o not n o r m a l l y e n d in any obstruent e x c e p t / s / , and / s / is p r o b l e m a t i c a l . H o w -
4

ever, if w e are willing to c o n s i d e r despirantization after / s / to b e effected by the

C o n t i n u a n t Dissimilation rule, w e o b s e r v e that interword application fails to o c -
cur: tis Salassas ' t h e s e a s ' c a n n o t b e c o m e *tis talassas.
H e r e are a few e x a m p l e s with reversal of stop p l u s fricative.

(9) a. / p l e k - 6 i k e / - > plextike 'it w a s k n i t t e d '

(cf. plekete 'it is k n i t t e d ' )
b. / p a r a - l i p - 6 i k a / —> paraliftika T was neglected'
(cf. paralipome T am neglected')

It is critical for the d e v e l o p m e n t of o u r a r g u m e n t that in every dialect of M o d e r n

G r e e k of w h i c h I a m a w a r e ( T h u m b , 1964, a n d N e w t o n , 1972b, are r e m a r k a b l y
c o m p l e t e a n d reliable sources), C o n t i n u a n t Dissimilation fails to apply in t w o
cases w h e r e o n e m i g h t h a v e e x p e c t e d it w o u l d : it d o e s not apply to voiced frica-
tives a n d it d o e s not apply to a n y t h i n g other than obstruents. T h e sole a n d telling
exception to this generalization is Cypriot, to w h i c h I turn in the next section. T h e
following e x a m p l e s from the standard dialect s h o w s e q u e n c e s of voiced fricatives
w h i c h regularly fail to s h o w dissimilation and s e q u e n c e s of Irl + fricative, w h e r e
350 Ellen M. Kaisse

Ixl is p r e s u m a b l y [ + c o n t i n u a n t ] o n t h e surface (Halle a n d C l e m e n t s , 1983) yet

d o e s not c o n d i t i o n dissimilation.

(10) a. avyo 'egg'

b. vy-o 'go out'
c. erxete 'he comes'

In K a i s s e ( 1 9 8 9 ) I s h o w that b o t h t h e s e restrictions follow from t h e s a m e thing:

neither v o i c e d fricatives n o r s o n o r a n t s are u n d e r l y i n g l y specified for continuancy.
T h i s in turn follows from t h e t h e o r y of contrastive underspecification (Steriade,
1987). T h e lack of specification for s o n o r a n t s is p e r h a p s the m o r e o b v i o u s of these
t w o : M o d e r n G r e e k , like m o s t l a n g u a g e s , d o e s n o t u s e c o n t i n u a n c y as t h e sole
feature w h i c h d i s t i n g u i s h e s a n y pair of s o n o r a n t s . T h e r e are, for instance, n o frica-
tive laterals o r nasals in t h e l a n g u a g e . N a s a l s , / l / , a n d Ixl are t h u s distinguishable
b y the features [nasal] a n d [lateral]. B u t w h a t of m y assertion that c o n t i n u a n c y is
not distinctive for v o i c e d fricatives? T h i s follows from t h e rather m a r k e d system
of o b s t r u e n t s in G r e e k . T h e r e a r e u n d e r l y i n g voiceless stops a n d there a r e b o t h
voiced a n d voiceless fricatives, b u t there a r e n o voiced s t o p s . T h u s voicing is 5

distinctive for fricatives (but n o t for stops), w h i l e c o n t i n u a n c y is distinctive for

voiceless o b s t r u e n t s b u t not for voiced o n e s . A n y v o i c e d o b s t r u e n t is r e d u n d a n t l y
[4-continuant].

(11) C O N T R A S T I V E L Y U N D E R S P E C I F I E D MATRICES FOR G R E E K ( l e x i c o n ) A N D

F U L L Y SPECIFIED ( p o s t l e x i c o n ) :

lexical postlexical
t e 6 n 1 r t 0 d n 1 r

vc — + - — + + + +
cnt - 4- - + 4- - - +
son - - - 4- + + _ _ - + 4- 4-
lat - 4- - - - - - 4- -
nas + - - - - - + - -

(12) R E D U N D A N C Y RULES:

a. 4-son —> + v c
f.
b
- r -lat -son j
—» — v c
- s o n —> -cont!
J
L~ n a s

i —» 4-cont
c. 4-nas —> —cont -son
4-vc
d. + l a t —> - c o n t
-lat
e. —» 4-cont
4- s o n
Rule Reordering and Generalization 351

T h e r e d u n d a n t specifications for c o n t i n u a n c y of the liquids (12d,e) will b e i m -

p o r t a n t in the u p c o m i n g d i s c u s s i o n a n d d e s e r v e s o m e m e n t i o n h e r e . A s I said,
H a l l e a n d C l e m e n t s ( 1 9 8 3 ) , by slightly m o d if y in g t h e C h o m s k y and H a l l e ( 1 9 6 8 )
definition of c o n t i n u a n c y , define III as [ — c o n t i n u a n t ] w h i l e Ixl is [ 4 - c o n t i n u a n t ] .
T h o u g h they d o n o t p r o v i d e e m p i r i c a l e v i d e n c e , I i m a g i n e that they h a v e the fol-
l o w i n g sorts of p h e n o m e n a in m i n d . In m a n y dialects of S p a n i s h , v o i c e d stops are
realized as fricatives b e t w e e n v o w e l s a n d also b e t w e e n an /r/ and a vowel. T h e
p r o c e s s t h u s a p p e a r s to b e a spread of continuancy. B e t w e e n n a s a l s a n d a v o w e l
or, crucially, b e t w e e n III a n d a vowel, the stops e m e r g e as s t o p s . T h u s the lateral,
6

like the nasals, b e h a v e s as a n o n c o n t i n u a n t sonorant. Similarly, in K o r e a n , under-

lying III is r e a l i z e d as Ixl before a v o w e l . H e r e , Ixl a p p e a r s as t h e result of contin-
u a n c y spread. A n d C l e m e n t s has h i m s e l f w o r k e d on the distribution of intrusive
stops in E n g l i s h , w h i c h , in s o m e dialects, a p p e a r after a nasal or lateral b u t not
after Ixl. W e shall see that the C y p r i o t liquids also g i v e e v i d e n c e for t h e s e o p p o s -
ing va lue s of c o n t i n u a n c y for III a n d Irl.

3 . 2 . T h e O b s t r u e n t i z a t i o n of lyl in C y p r i o t

O n e of the striking facts a b o u t the C y p r i o t dialect is that it turns lyl to

the voiceless palatal stop [k^] after a c o n s o n a n t . (After /r/, lyl b e c o m e s [ k ] ) . 7

Informally,

(13) y^WIC

T h e following e x a m p l e s s h o w alternations i n d u c e d b y the rule, involving, as they

d o , u n d e r l y i n g / i / s w h i c h lose their syllabicity before a n o t h e r v o w e l a n d t h u s b e -
c o m e subject to obstruentization.

(14) a. adelfi 'brother'

b. / a d e l f i + a / (—> adelfya) —> adelfk>a 'brothers'
c. teri ' o n e of a p a i r '
d. / t e r i + a z o / ( - > teryazo) —> terkazo 'to m a t c h '

Just w h a t n e e d s to b e specified in the formalization of the rule in u n c e r t a i n — f o r

instance, is the c o n s o n a n t a l i t y of the p r e c e d i n g C s p r e a d i n g with all the other
c h a n g e s following from r e d u n d a n c y r u l e s ? Is c o n t i n u a n c y the critical c h a n g e ?
S i n c e lyl is o b s t r u e n t i z e d after Ixl as well as after o t h e r o b s t r u e n t s , the p u r e s t form
of the rule a p p e a r s to involve spread of the feature [consonantal] only, with other
features e x p l a i n e d t h r o u g h r e d u n d a n c y or o t h e r m e a n s . T h e n o n c o n t i n u a n c y of
the o u t p u t m a y even stem from the further operation of the c o n t i n u a n c y t e m p l a t e
u p o n this n e w c o n s o n a n t , w h i c h is, of c o u r s e , a l w a y s in the s e c o n d position of a
c o n s o n a n t cluster. H o w e v e r , these issues are o r t h o g o n a l to o u r m a i n point, and I
leave t h e m u n r e s o l v e d h e r e .
8
352 Ellen M. Kaisse

(15) O B S T R U E N T I Z A T I O N O F y:
~— c o n s
— obst +cons
+cont —> +obst
+hi — cont
_-bk _ vc

N o w c o n s i d e r w h a t h a p p e n s w h e n the c o n s o n a n t triggering Obstruentization is

a stop: that stop itself b e c o m e s available as a focus for C o n t i n u a n t Dissimilation.
In other w o r d s , there is a new, innovative feeding order b e t w e e n Obstruentization
and C o n t i n u a n t Dissimilation.

(16) a. mati 'eye' maQWa 'eyes'

b. spiti 'house' spiOkya 'houses'
c. /pi/ 'drink' nafk o
y
'let m e d r i n k '

T h e following derivation illustrates this n e w order.

(17) /spiti + a/
spitya Glide Formation
spitkya O b s t r u e n t i z a t i o n of y
spinel C o n t i n u a n t Dissimilation

C o n t i n u a n t D i s s i m i l a t i o n is an old p a n - G r e e k rule, while O b s t r u e n t i z a t i o n of y is

a relatively n e w innovation of C y p r u s only. B e c a u s e it is a w e l l - r e c o g n i z e d result
of generative historical linguistics that rules are not a d d e d in the m i d d l e s of g r a m -
m a r s ( K i n g 1 9 7 3 ) , and that n e w rules are a d d e d at the e n d s of g r a m m a r s , w e
9

c o n c l u d e that the s y n c h r o n i c order reflects a classic (Kiparsky, 1965, 1968) reor-

dering into feeding order. B u t w h i c h rule m o v e d w h e r e ? D i d Obstruentization
m o v e u p or did Dissimilation m o v e d o w n ? O r did they switch? If C o n t i n u a n t
Dissimilation m o v e d , h o w far " d o w n " did it g o ? T h e a n s w e r I shall give is: D i s -
similation m o v e d d o w n , m o s t likely into the postlexical c o m p o n e n t , at least into
a very late level of the lexicon w h e r e r e d u n d a n t feature values b e c o m e available.
(18) also incorporates the speculation that Obstruentization is lexical, since it a p -
plies before r e d u n d a n t values for voicing are assigned. T h i s p l a c e m e n t is u p h e l d
by the failure of any source o n the dialect to cite application of Obstruentization
between words.

(18) I (Pan-Greek) II (Proto-Cypriot) III (Cypriot)

Lexicon: Dissimilation Dissimilation
Obstruentization

Obstruentization
Postlexicon: Dissimilation
Rule Reordering and Generalization 353

3 . 3 . M o t i v a t i o n for C o n t i n u a n t D i s s i m i l a t i o n in t h e Postlexical C o m p o n e n t

F o u r s e e m i n g l y unrelated c h a n g e s in the C o n t i n u a n t Dissimilation rule in C y p -

riot a r g u e that it has m o v e d later in the g r a m m a r than in other dialects. T h e rule
h a s g e n e r a l i z e d with voiced fricatives as either focus or d e t e r m i n a n t [illustrated
in (19)] a n d with Ixl as d e t e r m i n a n t (20). M o r e o v e r , t w o p r e v i o u s l y apparently
unrelated peculiarities a b o u t C y p r i o t c o m e into focus: w h e r e other G r e e k dialects
h a v e / before a c o n s o n a n t (or w h e r e C y p r i o t h a s a b o r r o w e d w o r d with the ety-
m o l o g i c a l s e q u e n c e / + C ) , C y p r i o t h a s r (21). F u r t h e r m o r e , there are sporadic
c h a n g e s that g o in the o p p o s i t e direction, turning r to /. B u t those c h a n g e s take
p l a c e only w h e n the original r is in the s e c o n d (or final) position of a c o n s o n a n t
cluster (22). I n o t e the last p h e n o m e n o n t h r o u g h a close r e a d i n g of K a h a n e and
K a h a n e ' s ( 1 9 8 7 ) survey of the C y p r i o t vocabulary, /'s a n d r's in other environ-
m e n t s are stable in all 127 C y p r i o t i s m s cited in K a h a n e a n d K a h a n e , of w h i c h
a b o u t 9 2 c o n t a i n at least o n e instance of / or r. In o t h e r w o r d s , C y p r i o t d o e s not
s h o w a r a n d o m confusion of b o r r o w e d / and r. T h e only p l a c e s w h e r e o n e is
substituted for the other is in c o n s o n a n t clusters w h e r e the t e m p l a t e [ + c o n t i n u a n t ]
[ — continuant] will b e instantiated [or, in the case of (21e, 2 2 ) , a p p r o x i m a t e d b y
giving the liquid the correct value for its position in the c o n s o n a n t cluster]. Inci-
dentally, K a h a n e and K a h a n e speculate that the [1] in spliverin (22a) c o m e s from
dissimilation from the following [r]. B u t m y r e a d i n g of their v o c a b u l a r y yields
several w o r d s with t w o /r/s w h i c h d o N O T dissimilate. T h e n o n d i s s i m i l a t i n g /r/s
are in clusters w h e r e their c o n t i n u a n c y value is already correct {kourkourizi
' r u m b l e ' ) . I c o n c l u d e that dissimilation is neither a n e c e s s a r y n o r a sufficient c o n -
dition for the c h a n g e of a liquid.
W h e r e n o source l a n g u a g e is given below, the form on the left represents the
reflexes found in n o n - C y p r i o t G r e e k dialects, w h i c h are c o n s e r v a t i v e with respect
to t h e c o n s o n a n t i s m . T h e final / n / s found in the C y p r i o t f o r m s are a m o r p h o l o g i c a l
c o n s e r v a t i s m , not relevant to o u r p o i n t here.

(19) a. ravdi > ravdin 'stick'

b. avyo > avgon 'egg'
c. evdomada > evdomada 'week'

(20) a. erxete > erkete 'he comes'

b. arxi > arki ~ arci 'beginning'

(21) a. belki (Turk.) > berki 'perhaps'

b. balkone (Ital.) > parkonin 'balcony'
c. poltos ( A n c . G k . ) > portos 'sweetmeat'
d. coltello (Ital.) > kurtela ' h a n d l e - l e s s knife'
e. adelfi > aderfi 'sibling'
354 Ellen M. Kaisse

(22) a. esprevier (O.Fr.) > spliverin 'mosquito net'

b. kriOari > kliOOarin 'barley'

T h e "fricativization" of the liquids in ( 2 1 ) results in the fulfilment of the first part

of the c o n t i n u a n c y t e m p l a t e : the first c o n s o n a n t b e c o m e s [ + c o n t i n u a n t ] . W h e t h e r
or not the full t e m p l a t e is instantiated through this c h a n g e apparently d e p e n d s on
w h e t h e r the input s e q u e n c e h a p p e n e d to already h a v e a stop in s e c o n d position.
Ideally w e w o u l d h a v e e x p e c t e d [ a d e r p i ] . N e w t o n cites only this o n e w o r d with
an inherited r 4- fricative s e q u e n c e , and I d o not yet k n o w w h a t to m a k e of the
a p p a r e n t failure of the lil to b e c o m e [ — c o n t i n u a n t ] . Similarly, w h i l e the failure
of the Ipl of esplevier to spirantize m a y b e e x p l a i n e d b y the p r e c e d i n g / s / , I d o not
k n o w w h y the Ikl of kliOOarin d o e s not spirantize. T h e i m p o r t of the l/r alterna-
tion is n o n e t h e l e s s clear: it is a l m o s t certainly N O T an i n d e p e n d e n t and unrelated
peculiarity of C y p r i o t b u t yet a n o t h e r e x a m p l e of the availability of the r e d u n d a n t
values for c o n t i n u a n c y to the C o n t i n u a n t Dissimilation rule and h e n c e an argu-
m e n t that Dissimilation is postlexical or word-level in this d i a l e c t . 10

N o t e that in e x p l a i n i n g rule generalization, I a m not taking the formal tack of

earlier generative p h o n o l o g i s t s , w h o saw generalization as the r e m o v a l of features
from a rule, so that it applied to a b r o a d e r class of s e g m e n t s or in a b r o a d e r class
of e n v i r o n m e n t s . F r o m that viewpoint, o u r C y p r i o t generalization w o u l d h a v e
consisted in the c h a n g e illustrated in ( 2 3 ) .

(23) a. U N G E N E R A L I Z E D FORM:
-fobst +obst
- voice — voice —> + c o n t —cont
acont Pcont

b. G E N E R A L I Z E D FORM:
a c o n t p c o n t —» + c o n t - c o n t

Instead, w e say that the rule has always h a d the " g e n e r a l " form of (23b). T h e
class of s e g m e n t s affected b y the rule b e c o m e s larger u n d e r generalization not
b e c a u s e the rule has c h a n g e d , but b e c a u s e of the characteristics of the stratum it
has m o v e d to. O u r a r g u m e n t follows the path laid d o w n in K i p a r s k y (1985). Ki-
p a r s k y a r g u e s that R u s s i a n Voicing A s s i m i l a t i o n applies m o r e b r o a d l y in its post-
lexical applications than its lexical o n e s b e c a u s e values for voice b e c o m e assign-
able to a n d s p r e a d a b l e from nondistinctively voiced sonorants. Similarly, w e say
that C y p r i o t C o n t i n u a n t D i s s i m i l a t i o n affects a n d is triggered b y the n o n d i s t i n c -
tively c o n t i n u a n t s e g m e n t s w h e n it m o v e s to a later stratum. T h e innovation in m y
a p p r o a c h is to r e c o g n i z e this b r o a d e n i n g as a p r i m a r y m e c h a n i s m of rule gener-
alization. It m a y well b e that not all rule generalizations can b e r e d u c e d to this
t y p e — r u l e generalization b y the actual r e m o v a l of features from rules m a y yet b e
Rule Reordering and Generalization 355

a valid m e c h a n i s m of p h o n o l o g i c a l c h a n g e . B u t in the next t w o sections of this

article, I sketch h o w t w o other rule generalizations, w h i c h h a d b e e n treated as
feature r e m o v a l , can instead b e seen as the result of the filling in of nondistinctive
features.
M y a r g u m e n t is, of c o u r s e , only as strong as the theory of underspecification
on w h i c h it rests, a n d if w e are incorrect in thinking that n o n c o n t r a s t i v e and re-
d u n d a n t feature v a l u e s only b e c o m e available late in derivations, m o s t likely post-
lexically, this k i n d of a r g u m e n t c a n n o t g o t h r o u g h . H o w e v e r , the a r g u m e n t s for
s o m e version of underspecification s e e m to m e very strong ( A r c h a n g e l i , 1 9 8 8 ;
Steriade, 1987). T h u s w e see that the availability of n o n c o n t r a s t i v e features in the
g e n e r a l i z e d o r r e o r d e r e d f o r m of a rule a r g u e s powerfully for its relocation in late
strata.
W e h a v e seen several strands of e v i d e n c e indicating that C o n t i n u a n t D i s s i m i -
lation h a s g a i n e d a c c e s s to nondistinctive features. W h a t w e are lacking, and w h a t
is not available in t h e d a t a r e p o r t e d in the literature, is c o r r o b o r a t i n g e v i d e n c e of
other types w h i c h w o u l d pin d o w n the position that C o n t i n u a n t D i s s i m i l a t i o n h a s
a s s u m e d in t h e g r a m m a r . I anticipate h a v i n g a c c e s s to further C y p r i o t m a t e r i a l s
and to i n f o r m a n t s shortly to gather the m i s s i n g d a t a . If C o n t i n u a n t D i s s i m i l a t i o n
1 1

h a s b e c o m e postlexical, I e x p e c t to find t h e following:

1. U n l i k e in other dialects, C y p r i o t C o n t i n u a n t Dissimilation should apply b e -

t w e e n w o r d s . B e c a u s e the only word-final o b s t r u e n t in M o d e r n G r e e k is / s / , o n e
w o u l d e x p e c t to o b s e r v e this effect m o s t clearly in the defricativization case, that
is, in e x a m p l e s of the form tis Salasses (—> tis talasses), t h o u g h the p r o b l e m a t i c
nature of e x a m p l e s c o n t a i n i n g / s / h a s b e e n m e n t i o n e d . N a s a l s can also b e w o r d
final, but w e h a v e seen that nasals d o not participate in the generalization of C o n -
tinuant D i s s i m i l a t i o n in C y p r i o t . A few w o r d s e n d in /r/, so w e m i g h t also l o o k
for defricativization of word-initial c o n s o n a n t s after r-final w o r d s . T h e r e m a y
also b e b o r r o w e d , clipped, or a c r o n y m i c w o r d s w h i c h e n d in c o n s o n a n t s other
than / s / , /r/, a n d / n / w h o s e b e h a v i o r in external sandhi I will investigate.
2. T h e e x c e p t i o n s to C o n t i n u a n t D i s s i m i l a t i o n w h i c h I r e p o r t e d a b o v e for m y
A t h e n i a n i n f o r m a n t s h o u l d b e absent in C y p r i o t .
3. T h e m o r p h o l o g i c a l c o n d i t i o n i n g o n the rule, w h e r e b y application b e t w e e n
stem and suffix is o b l i g a t o r y but application b e t w e e n prefix and stem is optional
and lexically g o v e r n e d , s h o u l d disappear.

W e d o not yet u n d e r s t a n d e n o u g h about the o r d e r i n g of r e d u n d a n c y rules to be

certain w h e n features like c o n t i n u a n c y in sonorants b e c o m e available. Possibly
they are filled in at the w o r d level, p o s s i b l y at the postlexical level. (See B o r o w s k y ,
this v o l u m e , for e v i d e n c e that structure preservation m a y shut off at the w o r d
level.) C o r r o b o r a t i n g e v i d e n c e of the type I h o p e to collect will h e l p to ascertain
w h e n r e d u n d a n c y rules o p e r a t e .
356 Ellen M. Kaisse

4. A S C A T T E R E D R U L E IN S W I S S G E R M A N

It is a p p r o p r i a t e that w e take u p h e r e an e x a m p l e first m a d e illustrious in Kipar-

sky's early ( 1 9 6 5 , 1968) w o r k o n rule r e o r d e r i n g , namely, the relation b e t w e e n the
rules of U m l a u t and L o w e r i n g in several dialects of S w i s s G e r m a n . K i p a r s k y ' s
original o b s e r v a t i o n s on this r e o r d e r i n g w e r e s h a r p e n e d a n d e x p a n d e d b y R o b i n -
son ( 1 9 7 6 ) , w h o a r g u e s as follows. T h e g e n e r a l i z e d form of a rule l o w e r i n g the
m i d b a c k v o w e l , w h i c h applied in c o n s e r v a t i v e S w i s s dialects only before Ixl
( 4 - c o r , —nasal, — lat), is g e n e r a l i z e d in the Schaffhausen dialect to apply before
c o r o n a l stops a n d fricatives as well (but not before III). T h e surprising fact u n c o v -
ered by R o b i n s o n is that the applications before Ixl c o n t i n u e to apply in their c o n -
servative position before U m l a u t , w h i l e the innovative, g e n e r a l i z e d applications
before c o r o n a l o b s t r u e n t s , a p p l y after U m l a u t . T h u s U m l a u t b l e e d s the innovative
applications, b y fronting the focus v o w e l before it c a n b e lowered, but counter-
b l e e d s the older, u n g e n e r a l i z e d f o r m of l o w e r i n g .

(24) L o w e r i n g before Ixl

Umlaut
General Lowering

R o b i n s o n ( 1 9 7 6 : 1 4 8 , 151) f o r m u l a t e s the u n g e n e r a l i z e d a n d g e n e r a l i z e d versions

of L o w e r i n g as follows.

(25) a. UNGENERALIZED:
V + son
-high -» [+low] / +cor
+back - nasal
- long — lateral

b. GENERALIZED:
V +cor
-high - » [ + low] / - nasal
+back — lateral
- long_

O u r n e w v i e w p o i n t leads us to realize that in a s i m p l e c o n s o n a n t inventory like

that of G e r m a n , the features [-nasal] and [-lateral] are distinctive only for the so-
norant, /r/, not for the o b s t r u e n t s . (I will n o t p a u s e h e r e to a t t e m p t to further un-
derspecify the s o n o r a n t s for t h e s e t w o features; the a l g o r i t h m in Steriade, 1987, is
i n d e t e r m i n a t e for n a r r o w i n g the D - c l a s s for lateral and nasal further than to the
s o n o r a n t s . I d i s c u s s the issue m o r e in the following section.)

(26) r 1 n t s

Nasal - - + 0 0
Lateral - + - 0 0
Rule Reordering and Generalization 357

T h e features [lateral and [nasal] n o r m a l l y distinguish only sonorants e x c e p t in

l a n g u a g e s with lateral affricates, nasal fricatives, prenasalized stops, and the like
T h e y should b e c o m e available for o b s t r u e n t s only in that portion of the g r a m m a r
w h e r e r e d u n d a n t feature values m a y b e referred to. A s with the C y p r i o t case, the
rule n e e d not have c h a n g e d but can a l w a y s h a v e h a d the generalized form in (25b).
T h e " s c a t t e r e d " effect, w h e r e L o w e r i n g before Ixl is c o u n t e r b l e d by U m l a u t w h i l e
L o w e r i n g before other c o r o n a l s is bled by it, follows from the spread of L o w e r -
ing's d o m a i n to the postlexical c o m p o n e n t .
W e can n o w see h o w R o b i n s o n ' s r e s u l t — t h a t g e n e r a l i z e d forms of rules o c c u r
at the e n d of the g r a m m a r , separated from the u n g e n e r a l i z e d rule from w h i c h they
s p r i n g — f o l l o w s from the lexical p h o n o l o g y m o d e l . It is not generalization that
b r e e d s scattering, as R o b i n s o n t h o u g h t . Rather, the extension d o w n w a r d of the
d o m a i n of a rule, as in the case of L o w e r i n g , automatically entrains generalization,
for the rule has a c q u i r e d a d o m a i n in w h i c h m o r e information b e c o m e s available:
in this c a s e , nondistinctive features. Generalization is not the c a u s e but the effect
of r e o r d e r i n g into a later stratum. T h e scattering effect results w h e n instead of
p i c k i n g u p and m o v i n g its d o m a i n d o w n w a r d , a rule m e r e l y e x p a n d s its d o m a i n s ,
a p p l y i n g in b o t h lexical and postlexical strata.

5. L I Q U I D D E L E T I O N I N S A M O T H R A K I G R E E K

T h e q u e s t i o n r e m a i n s w h e t h e r d o m a i n e x t e n s i o n and the c o n c o m i t a n t availabil-

ity of r e d u n d a n t features is the sole m e a n s of rule generalization, or w h e t h e r the
m o r e traditional e x p l a n a t i o n s m u s t still b e called on for s o m e c a s e s . B e c a u s e it is
difficult to d e t e r m i n e the properly underspecified m a t r i x for a given dialect with-
out k n o w i n g its p h o n o l o g y thoroughly, I p r e s e n t the following e x a m p l e not as a
definitive analysis but as a case w h e r e an underspecification solution, t h o u g h pur-
suable, is not entirely c o n v i n c i n g .
In the dialects of M o d e r n G r e e k s p o k e n in S a m o t h r a c e ( N e w t o n , 1972b; K a i s s e ,
1976), a scattering similar to the S w i s s G e r m a n o n e has taken p l a c e . In the rela-
tively c o n s e r v a t i v e city dialect, intervocalic Ixl is deleted. In m o r e innovative
c o u n t r y dialects, b o t h III a n d Ixl are deleted. A vowel height dissimilation rule
applies to the v o w e l s e q u e n c e s created b y r-deletion b u t not b y /-deletion.
(27) a. or-a (—» od) —> ua 'time'
b. kol-a —» koa 'starch'
Just as in S w i s s G e r m a n , w e can posit an a p p a r e n t generalization at the e n d of the
grammar.
(28) r-Loss
Vowel H e i g h t D i s s i m i l a t i o n
General Liquid Loss
358 Ellen M. Kaisse

L e t us a t t e m p t to w r i t e b o t h r - L o s s and its g e n e r a l i z e d counterpart, L i q u i d D e -

letion, as a single rule w h i c h applies m o r e g e n e r a l l y w h e n it applies later.

(29) 4-son
4-cor - > 0 / V .
-nas

S i n c e G r e e k and G e r m a n h a v e the s a m e inventory of s o n o r a n t s , the simplest hy-

p o t h e s i s is that they h a v e the s a m e feature m a t r i x for t h o s e c o n s o n a n t s .

(30) r 1 n

Nasal — — +
Lateral — 4- —

W e did n o t think carefully a b o u t the specification of the sonorants in G e r m a n ,

n o t i n g only that the features [nasal] a n d [lateral] w e r e o b v i o u s l y n e c e s s a r y in
s o m e w a y to distinguish t h e m from o n e another, w h i l e t h o s e features w e r e not
relevant for d i s t i n g u i s h i n g o b s t r u e n t s . A s I m e n t i o n e d in the last section, Steri-
a d e 's a l g o r i t h m is i n d e t e r m i n a t e for d e c i d i n g h o w fine a distinction o n e w a n t s to
m a k e in d e t e r m i n i n g D(istinctive) a n d R ( e d u n d a n t ) c l a s s e s . W e c o u l d m e r e l y di-
v i d e the c o n s o n a n t s into o b s t r u e n t s and s o n o r a n t s , as I did a b o v e , a n d state that
the s o n o r a n t s are the D - c l a s s for [lateral] a n d [nasal]. B u t it is p o s s i b l e to m a k e
finer distinctions, a n d this is the tack o n e w o u l d h a v e to take in p u r s u i n g an un-
derspecification solution a l o n g the lines w e a d v o c a t e d for C y p r i o t and S w i s s
German.
W e k n o w that [lateral] is distinctive for the class of liquids, particularly since,
like the other G r e e k dialects, S a m o t h r a k i d o e s n o t i m p o s e the c o n t i n u a n c y t e m -
plate on s o n o r a n t s . T h u s c o n t i n u a n c y c a n n o t b e the feature w h i c h distinguishes
Ixl from / l / , and [lateral] is p r o p e r l y specified for b o t h s e g m e n t s . B u t the rest is
n o t so o b v i o u s ; (30) still c o n t a i n s s o m e r e d u n d a n t information. A m o r e thor-
o u g h l y underspecified m a t r i x , if w e follow S t e r i a d e ' s a l g o r i t h m for d e t e r m i n i n g
D-classes, would be:

(31) r 1 n

Nasal - +
Lateral — +

I d e r i v e d (31) as follows. T h e feature that d i s t i n g u i s h e s III from Ixl is [lat-

eral]. T h u s they are a D - c l a s s for lateral a n d receive p l u s and m i n u s features
respectively.

(32) 1 r
Lateral 4- —
Rule Reordering and Generalization 359

S t e r i a d e states that an R - c l a s s for a feature is that c l a s s of s o u n d s for w h i c h o n e

value of a distinctive feature is systematically e x c l u d e d . F o r e x a m p l e , s o n o r a n t s
are u s u a l l y an R-class for voicing b e c a u s e the value [ — v o i c e ] is e x c l u d e d for
t h e m . F o l l o w i n g this r e a s o n i n g , w e see that w h i l e liquids are a D - c l a s s for [lat-
e r a l ] , b o t h features of lateral b e i n g a s s i g n a b l e to t h e m , n a s a l s are an R-class for
lateral, as the features [ + n a s a l ] [ + l a t e r a l ] are i n c o m p a t i b l e in n o r m a l feature in-
v e n t o r i e s . T h u s Inl is [Olateral]. C o n t i n u i n g this r e a s o n i n g for the feature [nasal],
w e see that [ 4 - n a s a l ] is e x c l u d e d for III b e c a u s e , as w e h a v e said, [ + l a t e r a l ] is
i n c o m p a t i b l e with [ 4 - n a s a l ] . H o w e v e r , n e i t h e r [ + n a s a l ] n o r [ - nasal] is e x c l u d e d
from s e g m e n t s like Irl w h i c h are [ — lateral], a n d i n d e e d w e n e e d [ — nasal] to
distinguish Irl from Inl.
T h e r e a d e r c a n n o w see w h e r e this a r g u m e n t is l e a d i n g : rule (29) r e q u i r e s the
feature [ — nasal] in o r d e r to apply to a s o n o r a n t . In c o n t r a s t i v e underspecification
t h e o r y t h e r e is n o t h i n g special a b o u t a 4- v a l u e over a — v a l u e ; a r u l e c a n r e q u i r e
either e q u a l l y well. In lexical applications, therefore, n e i t h e r n a s a l s ( [ + n a s a l ] ) n o r
laterals ([Onasal]) m e e t the structural d e s c r i p t i o n of the rule, a n d o n l y Irl u n d e r -
g o e s deletion. T h e postlexical e x t e n s i o n of L i q u i d D i s s i m i l a t i o n will result in the
n o n n a s a l i t y of III b e c o m i n g available, a n d III t h u s will b e c o m e subject to deletion.
S u c h an e x p l a n a t i o n m a y turn out to b e t e n a b l e . H o w e v e r , I m u s t p o i n t o u t
several w e a k n e s s e s in it. If the underspecification a n a l y s i s is rejected d u e to t h e s e
w e a k n e s s e s , w e will h a v e to a d m i t that there is m o r e t h a n o n e m e c h a n i s m b y
w h i c h rules m a y g e n e r a l i z e : o u r underspecification m e c h a n i s m , and the m o r e tra-
ditional m e t h o d b y w h i c h features are d r o p p e d from rules, a l l o w i n g t h e m to apply
to a b r o a d e r class of s e g m e n t s .
First, w e m u s t us a s k w h e t h e r Inl a n d III are distinct in (31). P u s h i n g the c o n -
trastive underspecification a l g o r i t h m led us to assign n o — value to o n e for w h i c h
a c o r r e s p o n d i n g + w a s a s s i g n e d to the other. S o w e m u s t a s k w h e t h e r [ 4 - n a s a l ]
c o n t r a s t s with [Onasal] a n d / o r w h e t h e r [ + l a t e r a l ] c o n t r a s t s w i t h [Olateral]. T h e
o n l y s t r o n g a r g u m e n t s for ternarily c o n t r a s t i n g features c o m e from situations t y p i -
cal to t o n e l a n g u a g e s , w h e r e L - a n d H - t o n e d m o r p h e m e s c o n t r a s t w i t h alternating,
0 - t o n e d m o r p h e m e s or, similarly, from h a r m o n y s y s t e m s , w h e r e 4 - F a n d — F v o w -
els c o n t r a s t w i t h alternating, OF v o w e l s . G r e e k p r o v i d e s n o e v i d e n c e I k n o w of for
a ternary distinction within the feature [nasal] or [lateral]. T h e r e f o r e , I tentatively
c o n c l u d e that an early r e d u n d a n c y rule m u s t fill in at least the value [ - lateral] on
Inl or [ — nasal] on III. W h i c h o n e is filled in a p p e a r s to b e arbitrary, or w e m i g h t
c h o o s e to fill in both. B u t of c o u r s e o u r rule g e n e r a l i z a t i o n will only w o r k if w e
c h o o s e to fill in the value for [lateral] on Inl early a n d leave III unspecified for
nasality until the postlexical c o m p o n e n t . O n l y in that w a y will III e s c a p e deletion
p r i o r to H e i g h t D i s s i m i l a t i o n and u n d e r g o deletion after H e i g h t D i s s i m i l a t i o n . I
find this i n d e t e r m i n a c y w o r r i s o m e . W i t h o u t further e v i d e n c e , w e m u s t w o n d e r
w h e t h e r the availability of an underspecification solution for the S a m o t h r a k i c a s e
is m e r e l y a result of the latitude p r o v i d e d b y c u r r e n t t h e o r i e s of underspecification.
360 Ellen M. Kaisse

A s e c o n d w o r r i s o m e aspect of the solution sketched above c o n c e r n s the status

of the feature [nasal]. Steriade (personal c o m m u n i c a t i o n , 1990) n o w suspects that
[nasal] is a privative feature, not a binary o n e . H e r e v i d e n c e involves the a b s e n c e
of rules spreading [ - n a s a l ] and the frequency of rules spreading [ + n a s a l ] . If
[ — nasal] c a n n o t b e referred to by p h o n o l o g i e s , L i q u i d Deletion (27) m u s t be re-
formulated to apply to sonorants with n o nasal n o d e , that is, to the liquids. At n o
time d o e s l\l h a v e a nasal n o d e . Therefore, the n o n g e n e r a l i z e d form of the rule,
w h i c h deletes only /r/, m u s t specifically include the feature [ — lateral], and we
are b a c k to the old analysis of rule generalization via the d r o p p i n g of features
from rules.

6. CONCLUSION

T h e theory of underspecification within lexical p h o n o l o g y allows us to under-

stand the m e c h a n i s m of rule generalization with greater clarity than w a s previ-
ously p o s s i b l e . It further allows us to link rule generalization to an u n d e r a p -
preciated direction of rule r e o r d e r i n g : m o v e m e n t d o w n w a r d in the g r a m m a r ,
including even re-entry into the postlexical c o m p o n e n t . W e h a v e seen cases of
d o w n w a r d m o v e m e n t in K a s k a , in Q u e b e c o i s F r e n c h , in C y p r i o t G r e e k , in the
Schaffhausen dialect of S w i s s G e r m a n , and possibly i S a m o t h r a k i G r e e k as well.
t

In the K a s k a c a s e , m o v e m e n t takes the form of e x t e n s i o i of a rule's m o r p h o l o g i c a l

d o m a i n o n e lexical stratum further, so that m o r e s e q u e n c e s m e e t i n g the p h o n o -
logical description of the rule actually u n d e r g o it. T h e rule's transparency is
thereby increased. T h e K a s k a c a s e w o u l d n o t typically b e t e r m e d a rule general-
ization b e c a u s e it d o e s not involve an extension of the natural class of s e g m e n t s
u n d e r g o i n g or c o n d i t i o n i n g the rule. T h e case in Q u e b e c o i s F r e n c h involves ex-
tension or m o v e m e n t into the postlexical c o m p o n e n t , since a Vowel Deletion rule
gains access to information c o n c e r n i n g the e n d of a sentence, that is, to the p r o -
sodic organization of constituents into intonational p h r a s e s and utterances. At the
s a m e t i m e , the rule has acquired a generalized e n v i r o n m e n t in n o longer requiring
its focus to a p p e a r in a c l o s e d syllable. T h e Cypriot, G e r m a n , a n d S a m o t h r a k i
cases s h o w e d (in o r d e r of decreasing certainty on the part of this author) that a
rule w h i c h is out of its chronologically correct o r d e r m a y gain access to r e d u n d a n t
feature values. F r o m this w e c o n c l u d e d that such rules h a v e m o v e d d o w n w a r d
with respect to the c h r o n o l o g i c a l l y later rules they n o w follow. T h e analysis of
Cypriot, for w h i c h w e w e r e able to bring the m o s t e v i d e n c e to bear, h a d the ad-
vantage of unifying several disparate facts about t h e dialect u n d e r t h e rubric of
rule generalization. M o r e o v e r , it suggested that w e m i g h t eliminate the subtraction
of features from foci and d e t e r m i n a n t s as a p r i m a r y m e c h a n i s m of p h o n o l o g i c a l
c h a n g e . Instead, such generalization of rules can b e s u b s u m e d u n d e r the inde-
Rule Reordering and Generalization 361

p e n d e n t l y n e e d e d m e c h a n i s m of rule r e o r d e r i n g . W i t h i n a theory w h i c h adopts

level o r d e r i n g , this r e o r d e r i n g m e c h a n i s m m i g h t further b e e x p e c t e d to result in
s a n d w i c h - t y p e rule interactions, w h e r e an old rule both p r e c e d e s a n d follows a
n e w o n e . T h e C y p r i o t case d o e s not a p p e a r to involve this sort of o r d e r i n g p a r a -
d o x , the type that led R o b i n s o n to coin the t e r m " s c a t t e r e d r u l e . " B u t l o o k i n g at
" s c a t t e r i n g " in S w i s s G e r m a n and in S a m o t h r a k i G r e e k , I c o n c l u d e d that at least
s o m e cases of rule generalization w h i c h exhibit s a n d w i c h i n g fall out from the
theory of rule r e o r d e r i n g a n d underspecification I h a v e d e v e l o p e d here. F u r t h e r
investigation will tell us if all rule generalization can b e r e d u c e d to a m e r e c o n s e -
q u e n c e of late rule application or if the elimination of features from structural
descriptions m u s t r e m a i n as an i n d e p e n d e n t m e c h a n i s m of p h o n o l o g i c a l c h a n g e .

NOTES

The other case of rule scattering in the literature known to me is Bley-Vroman's (1975)
1

treatment of Old Norse. In that language, umlaut of vowels in extra-heavy syllables applies
before a vowel syncope rule. The generalized umlaut of vowels in any type of syllable
applies after syncope. The treatment of this phenomenon in terms of rule movement is not
obvious to me but might emerge from a close study of the form syllabification takes at
different strata in Old Norse or of syllabification's ordering with respect to umlaut. The
Quebecois French case discussed at the end of this section also involves loss of a closed
syllable requirement.
2
In Sekani transcriptions, a vowel unmarked for tone is high, HI is an allomorph of the
lsg prefix used in the perfective of verbs having a zero or -h- classifier ('elf'). The grave
accent preceding the conjunct prefix indicates a floating tone, realized as Low in Sekani
and as High in Kaska.
3
1 am grateful to Doug Pulleyblank for directing me to this example. An anonymous
referee suggests another interpretation of Dumas's facts, namely that the vowel lengthening
rule has become optionally cyclic, so that it can optionally lengthen vowels that precede
voiced fricatives that are word final prior to suffixation.
4
The spirant partner of lil is / 8 / , not Isl. And indeed Isl acts asymmetrically with respect
to other continuants. It cannot despirantize itself and causes despirantization of any frica-
tive which precedes it, in violation of the continuancy template. Despirantization before
Isl is a very old rule, one found in Attic Greek. It is therefore unclear whether phenomena
involving Isl should be used as evidence in discussions of Continuant Dissimilation, and I
will avoid such examples.
5
Actually, the phonemic status of voiced stops is a more vexed question than my simple
statement indicates, and Modern Greek scholars have argued over it at length. Voiced stops
do appear as the result of voicing spread from a preceding nasal, and when that nasal is
optionally lost, an apparently unconditioned voiced stop can appear on the surface. Voiced
stops also appear in loanwords, though an etymologically incorrect nasal occasionally ap-
pears along with them. I cannot hope to resolve the issue here, but the reader may be
362 Ellen M. Kaisse

assured that the distribution of voiced stops is defective in Modern Greek and that credible
analyses where these segments are lacking abound. Moreover, the evidence for the derived
character of voiced stops is particularly weighty in Cypriot, where they are invariably pre-
ceded by nasals.
6
An anonymous reviewer reminds me that Idl does not spirantize after /l/, though / b /
and Igl do. The complication may result from the fact that III and Idl are homorganic; the
linking convention may prevent spirantization of stops sharing a place node with an adja-
cent consonant.
7
Data in this and the following sections are taken from the excellent work of Newton
(1972a,b), except where otherwise noted. My analysis of some rules differs. I discuss the
spread of consonantality in Kaisse (1992).
8
Pat Shaw points out to me that the emergence of y from this rule as a voiceless conso-
nant indicates that the rule applies before the redundant values for voicing of glides and
stops have been assigned. Obstruentization of y is therefore likely to be a lexical rule in
Cypriot, and the value of voicing can be omitted from the actual structural change of the
rule. The placement of Obstruentization in the lexicon is consistent with my claim that
Continuant Dissimilation, which follows Obstruentization, is postlexical. It also suggests
weakly that not only did Continuant Dissimilation move down, but that Obstruentization
moved up. However, the movement of Obstruentization into the lexical component could
of course have been independent of its reordering with Continuant Dissimilation. As we
have said, the normal course of development for any new rule is for it to move into the
lexicon.
9
There are still a few controversial cases where rule insertion appears the best analysis,
the most celebrated being Lachmann's Law.
10
N a s a l s never seem to participate in the continuancy template; in fact, fricatives are
disfavored after presumably [ - continuant] nasals, since the normal historical development
of aspirated and voiced stops to fricatives did not occur after nasals. I conclude that the
value for continuancy for nasals is never specified within the phonology of Greek.
The Gulf War prevented my planned visit to Cyprus in January 1991. I apologize to
1 1

the reader for leaving you in suspense.

REFERENCES

Archangeli, D. (1988). Aspects of underspecification theory. Phonology 5, 183-207.

Bley-Vroman, R. (1975). Opacity and interrupted rule schemata. Papers from the Regional
Meeting of the Chicago Linguistics Society 11, 7 3 - 8 0 .
Chomsky, N., and Halle, M. (1968). The Sound Pattern of English. Harper and Row, New
York.
Dumas, D. (1981). Structure de la diphthongaison quebecoise. Canadian Journal of Lin-
guistics 26, 1 - 6 0 .
Halle, M., and Clements, G. N. (1983). Problem Book in Phonology. MIT Press, Cam-
bridge, Mass.
Halle, M., and Mohanan, K. P. (1985). Segmental phonology of Modern English. Linguis-
tic Inquiry 16, 57 - 1 1 6 .
Rule Reordering and Generalization 363

Hargus, S. (1988). The Lexical Phonology of Sekani. Garland, New York.

Kahane, H , and Kahane, R. (1987). A Cypriot etymologicum: comments to the glossary
of Georgios Loukas. Mediterranean Language Review 3, 71 - 1 0 4 .
Kaisse, E. (1976). Hiatus in Modern Greek. Doctoral dissertation, Harvard University,
Cambridge, Mass.
Kaisse, E. (1989). Modern Greek Continuant Dissimilation and the OCP. Unpublished
manuscript, University of Washington, Seattle.
Kaisse, E. (1992). Can [consonantal] spread? Language 68, 313-332.
King, R. (1973). Rule insertion. Language 49, 5 5 1 - 5 7 8 .
Kiparsky, P. (1965). Phonological Change. Doctoral dissertation, Massachusetts Institute
of Technology, Cambridge.
Kiparsky, P. (1968). Linguistic universals and linguistic change. In Universals in Linguistic
Theory (E. Bach and R. Harms, eds.), pp. 1 7 1 - 2 0 2 . Holt, Rinehart and Winston, New
York.
Kiparsky, P. (1984). On the lexical phonology of Icelandic. In Nordic Prosody III (C. C.
Elert et al., eds.), pp. 135-164. University of Umea, Umea, Sweden.
Kiparsky, P. (1985). Some consequences of lexical phonology. Phonology Yearbook 2,
82-136.
Kiparsky, P. (1988). Phonological change. In The Cambridge Survey of Linguistics, vol. 1,
Linguistic Theory: Foundations (F. J. Newmeyer, ed.), pp. 3 6 3 - 4 1 5 . Cambridge Uni-
versity Press, Cambridge.
Moore, P. (1988). Kaska Verb Workshop. Unpublished manuscript, Yukon Native Lan-
guage Centre, Whitehorse.
Newton, B. (1972a). Cypriot Greek: Its Phonology and Inflection. Mouton, The Hague.
Newton, B. (1972b). The Generative Interpretation of Dialect: A Study of Modern Greek
Phonology. Cambridge University Press, Cambridge.
Robinson, O. W. (1976). A "scattered" rule in Swiss German. Language 52, 1 4 8 - 1 6 1 .
Steriade, D. (1987). Redundant values. Papers from the Parasession on Autosegmental and
Metrical Phonology, pp. 3 3 9 - 3 6 2 . Chicago Linguistic Society, Chicago.
Thumb, A. (1964). A Handbook of the Modern Greek Language. Argonaut, Chicago.
RULE DOMAINS AND PHONOLOGICAL CHANGE

DRAGA ZEC
Department of Modern Languages and Linguistics
Cornell University
Ithaca, New York 14853

1. INTRODUCTION

T h i s p a p e r a d d r e s s e s a p r o b l e m in historical linguistics, the p r o b l e m of rule

addition, from the p e r s p e c t i v e of lexical p h o n o l o g y and m o r p h o l o g y ( L P M ) . R u l e
addition played an i m p o r t a n t role in the d e b a t e of several d e c a d e s a g o on the
characterization of l a n g u a g e c h a n g e in the generative f r a m e w o r k (Halle, 1962;
Kiparsky, 1965; K i n g , 1969, 1973). C e n t r a l in this d e b a t e w a s the o r d e r i n g of a
n e w l y a d d e d rule with respect to other rules in the g r a m m a r . W i t h i n the L P M
m o d e l , rule addition can b e a p p r o a c h e d from a s o m e w h a t different viewpoint.
This m o d e l p r o p o s e s a set of possible rule d o m a i n s , as well as a c o n s t r a i n e d m a p -
ping b e t w e e n rules and their respective d o m a i n s . T h u s , c o u c h e d within the L P M
m o d e l , the p r o b l e m of rule addition can b e formulated as follows: w h i c h of the
d o m a i n s available in the g r a m m a r m a y b e selected by the n e w l y a d d e d rule?
T h e c a s e of rule addition to b e studied h e r e is accent retraction in the S t o k a v i a n
dialects of S e r b o - C r o a t i a n . T h i s historical p r o c e s s , dated a r o u n d the 15th c e n t u r y
1

(Belie, 1 9 5 6 : 1 6 1 ) , significantly altered the accentual properties of the Stokavian

dialects. H o w e v e r , not all S t o k a v i a n dialects h a v e b e e n affected by this p r o c e s s in
the s a m e fashion. In the N e o - S t o k a v i a n g r o u p of dialects this p r o c e s s h a s affected
all accents a n d as a result entirely r e p l a c e d the old accentual system w i t h the n e w
o n e . T h e O l d S t o k a v i a n dialects fall into t w o s u b g r o u p s . T h e c o n s e r v a t i v e dialects
preserve the accentual situation prior to accent retraction. T h e less c o n s e r v a t i v e
dialects h a v e b e e n affected b y accent retraction only partially, so that they retain

365
Phonetics and Phonology, Volume 4 Copyright © 1993 by Academic Press, Inc.
Studies in Lexical Phonology All rights of reproduction in any form reserved.
366 Draga Zee

certain p r o p e r t i e s of the old accentual system w h i l e also a c q u i r i n g properties of

the n e w o n e . M y focus is on those dialects w h i c h h a v e u n d e r g o n e accent retrac-
tion. T h e variation a c r o s s this g r o u p of dialects p r o v i d e s an insight into various
stages of this historical c h a n g e . W e will b e able to o b s e r v e h o w the addition of a
rule interacts with p h o n o l o g i c a l s y s t e m s of essentially the s a m e t y p e yet differing
in certain relevant details. In particular, dialects vary as to w h a t rule d o m a i n s are
selected b y accent retraction. M o r e o v e r , the cross-dialectal variation follows a
clear pattern, w h i c h receives a straightforward a c c o u n t within the m o d e l of L P M .
I n o w turn to t h o s e aspects of the m o d e l w h i c h are crucial to m y analysis. T h e
version of the L P M m o d e l I use h e r e is that p r o p o s e d in K i p a r s k y ( 1 9 8 4 , 1985).
In this m o d e l , the g r a m m a r is o r g a n i z e d into distinct levels w h i c h p r o v i d e suc-
cessively larger rule d o m a i n s ; in S e r b o - C r o a t i a n w e n e e d to posit at least those
levels listed in (1).

(1) Level 1
Cyclic
Level 2
Level 3 Phonological word
Postcyclic
Level 4 Phonological phrase

T h e postcyclic levels posited for S e r b o - C r o a t i a n are in fact universal; they are

derived from, a l t h o u g h not i s o m o r p h i c with, syntactic d o m a i n s (Hayes, 1989a;
N e s p o r a n d Vogel, 1986; Selkirk, 1978, 1980). Postcyclic levels h a v e thus b e e n
e q u a t e d with the i n d e p e n d e n t l y m o t i v a t e d hierarchy of p r o s o d i c constituents such
as the p h o n o l o g i c a l w o r d , the p h o n o l o g i c a l p h r a s e , and so on. T h e levels p r e -
c e d i n g the postcyclic set can in p r i n c i p l e b e either cyclic or n o n c y c l i c . It so h a p -
p e n s that in S e r b o - C r o a t i a n all the earlier levels are cyclic (for a r g u m e n t s see
Zee, 1988). 2

A further p r o p e r t y of this m o d e l crucial for us is the lexical/postlexical segre-

gation. T h e lexical c o m p o n e n t i n c l u d e s levels s m a l l e r than the word, w h i l e the
postlexical c o m p o n e n t i n c l u d e s levels larger than the word. T h e w o r d is thus
caught, as it w e r e , at the intersection p o i n t b e t w e e n these t w o c o m p o n e n t s . T h i s is
d u e to the a m b i g u i t y of the w o r d d o m a i n , w h i c h b e l o n g s b o t h to the syntax and to
the m o r p h o l o g y . T h i s b e i n g the case, the p h o n o l o g i c a l w o r d could b e created ei-
ther lexically or postlexically. T h e facts presented h e r e clearly s h o w the dual char-
acter of the w o r d .
N o w , w h i c h of the levels in (1) b e l o n g to the lexical, a n d w h i c h to the postlexi-
cal c o m p o n e n t ? T h e situation in the Stokavian dialects is not uniform in this re-
spect; the p h o n o l o g i c a l w o r d m a y either b e created lexically, as in (2), or post-
lexically, as in (3).
Let us start with the situation in (2), in w h i c h the p h o n o l o g i c a l w o r d is created
lexically. It is n o l o n g e r a s s u m e d , as it w a s in the early versions of L P M , that all
Rule Domains and Phonological Change 367

postcyclic rules apply postlexically; at least o n e postcyclic d o m a i n , the level of

the w o r d , m a y h a v e lexical p r o p e r t i e s , as a r g u e d in K i p a r s k y (1985) and Booij
and R u b a c h ( 1 9 8 7 ) . A lexically created p h o n o l o g i c a l w o r d is also p r e s e n t p o s t l e x -
ically a n d m a y c o m b i n e with clitics. It is at this p o i n t that clitics e n t e r the structure
and are i n c l u d e d into the p h o n o l o g i c a l w o r d , thus increasing its s i z e . In the lexi-3

con, however, the p h o n o l o g i c a l w o r d can only a p p e a r in its smaller size, that is,
without clitics.

(2) Lexical Level 1

Cyclic
Level 2
Level 3 P h o n o l o g i c a l w o r d (word-sized)
Postcyclic
Postlexical Level 3 P h o n o l o g i c a l w o r d (clitic g r o u p )
Level 4 Phonological phrase

If created postlexically, as in (3), the p h o n o l o g i c a l w o r d is manifested only in

c o m b i n a t i o n with clitics, n e v e r in its smaller size.

(3) Lexical Level 1 ^ i-

Level 2 C y c l l C

Postlexical Level 3 P h o n o l o g i c a l w o r d (clitic g r o u p )

Level 4 Phonological phrase

W e thus o b s e r v e an a s y m m e t r y : if the p h o n o l o g i c a l w o r d is created lexically, it

c o m e s both in its s m a l l e r a n d in its larger size; if created postlexically, it c o m e s
only in its larger size. T h e d o m a i n often referred to in the literature as the clitic
g r o u p c o r r e s p o n d s to t h e p h o n o l o g i c a l w o r d in its postlexical guise. In fact, o n c e
the p r o s o d i c c o n s t i t u e n c y is d i v i d e d u p into its lexical a n d postlexical subparts, it
is p o s s i b l e to d i s p e n s e w i t h the clitic g r o u p as a p r o s o d i c level w h i l e retaining the
rule d o m a i n of t h e clitic g r o u p s i z e . 4

G i v e n this general d e s i g n of the g r a m m a r , w e will ask the following question:

W h i c h of the levels, or c o m p o n e n t s , is c h o s e n as the e n t r a n c e gate for a p h o n o -
logical rule that is n e w l y a d d e d to t h e g r a m m a r ? D o rules enter a g r a m m a r with
an arbitrarily specified d o m a i n , or d o e s rule addition follow a m o r e r e g u l a r pat-
tern? W h a t I p r o p o s e is c o m p a r a b l e to the old claim, d u e to K i n g ( 1 9 7 3 ) , that a
n e w rule is a d d e d at the e n d of a g r a m m a r . 5

(4) A n e w rule is a d d e d to the largest (i.e., highest) d o m a i n of a g r a m m a r .

Its further p r o p e r t i e s will then follow from the strong d o m a i n h y p o t h e s i s (Ki-

parsky, 1984), w h i c h essentially m a k e s the following c l a i m .

(5) S T R O N G DOMAIN HYPOTHESIS: If a rule applies at level n, it also applies at

level n — 1, b u t not necessarily vice versa.
368 Draga Zee

U n d e r optimal c o n d i t i o n s , the n e w l y a d d e d rule will then simply spread

t h r o u g h o u t the g r a m m a r . T h e c a s e of p h o n o l o g i c a l c h a n g e to b e studied here will
s h o w that the strong d o m a i n h y p o t h e s i s g o v e r n s both synchronic and diachronic
aspects of the rule c o m p o n e n t , with o n e n o t a b l e exception: rules of the d o m a i n
limit type diverge from the strong d o m a i n h y p o t h e s i s by c h o o s i n g only the post-
cyclic d o m a i n s .
T h e article is o r g a n i z e d as follows. Sections 2 - 4 p r e s e n t the set of facts rele-
vant for locating accent retraction within the S e r b o - C r o a t i a n accentual system. In
Section 2 accent retraction is a n a l y z e d as a t o n e spreading rule; Section 3 focuses
on the postcyclic application of this rule in a representative g r o u p of dialects,
while Section 4 p r o v i d e s a r g u m e n t s for its cyclic application. Section 5 develops
a c o n s t r a i n e d view of postcyclic levels, w h i c h a c c o u n t s for all cross-dialectal
variation associated with accent retraction in t e r m s of a single set of rule d o m a i n s .
Finally, Section 6 outlines a route of c h a n g e that can b e r e c o n s t r u c t e d for this case
within the L P M m o d e l . T h e general view of S e r b o - C r o a t i a n accent to b e p r e s e n t e d
here is that d e v e l o p e d in Inkelas and Z e e ( 1 9 8 8 ) and Z e e ( 1 9 8 8 ) . W h i l e Inkelas
and Z e e (1988) focus on late lexical as well as postlexical p r o c e s s e s , Z e e (1988)
looks into the relatively early lexical p r o c e s s e s . T h e t w o a c c o u n t s in m a n y w a y s
c o m p l e m e n t e a c h other; moreover, b o t h are b a s e d on L e h i s t e and Ivic's (1986)
i m p o r t a n t study of the p h o n e t i c p r o p e r t i e s of S e r b o - C r o a t i a n accent.

2. A C C E N T R E T R A C T I O N A S A T O N E - S P R E A D I N G R U L E

A c c o r d i n g to traditional descriptions, the accentual difference b e t w e e n the N e o -

and O l d Stokavian dialects is manifested in the n u m b e r of surface accent t y p e s :
the N e o - S t o k a v i a n h a s four pitch accents, t w o rising a n d t w o falling, w h i l e the
O l d Stokavian h a s only the t w o falling pitch a c c e n t s . T h e difference b e t w e e n the
6

t w o dialect g r o u p s has b e e n attributed to the effect of accent retraction. In this

section I outline the accentual properties of the t w o Stokavian dialect g r o u p s , and
then s h o w that the N e o - S t o k a v i a n dialects differ from the O l d Stokavian o n e s in
p o s s e s s i n g the p h o n o l o g i c a l rule w h i c h is the synchronic c o u n t e r p a r t of accent
retraction.

2.1. Neo-Stokavian Dialects

T h e N e o - S t o k a v i a n dialects are traditionally described as p o s s e s s i n g four pitch

a c c e n t s : short rising, short falling, long rising, a n d long falling. B e l o w are given
e x a m p l e s for each of the accent types, m a r k e d b y the traditionally used diacritics;
in the r i g h t h a n d c o l u m n are given the pitch p r o p e r t i e s associated with e a c h of the
accents:
Rule Domains and Phonological Change 369

(6) a. Short rising: marama mara\ma 'scarf

b. Short falling: kucica ku\cica 'house (dim.)'
c. L o n g rising: rdzlika ra\azli\ka 'difference'
d. L o n g falling: rddnica Tdladnica ' w o r k e r '
T h e relevant p h o n e t i c correlates of a c c e n t established by L e h i s t e and Ivic
(1986) are pitch and duration. T a k i n g duration to b e a correlate of stress, and pitch
a correlate of tone, Inkelas and Z e e ( 1 9 8 8 ) p r o p o s e to d e c o m p o s e accent into tone
and stress, as s h o w n in the following a u t o s e g m e n t a l r e p r e s e n t a t i o n .
7

(7) a. Short rising: marama

V I
H L

b. Short falling: kucica

I V
HL

c. L o n g rising: raazlika
IV I
LH L

d. L o n g falling: raadnica

H L
In (7), the pitch p r o p e r t i e s of e a c h of the four accents are r e p r e s e n t e d in t e r m s
of H i g h a n d L o w t o n e s . A s noted in L e h i s t e a n d Ivic ( 1 9 8 6 ) , pitch p r o m i n e n c e is
found on the only syllable b e a r i n g a falling accent a n d on the a c c e n t e d and the
p o s t a c c e n t u a l syllables of the rising accent. T h i s is c a p t u r e d in (7) in t e r m s of the
distribution of the H i g h tones: a falling a c c e n t is associated with a H i g h tone
linked to a single syllable, and a rising accent with a H i g h linked to t w o c o n s e c u -
tive syllables. T h e leftmost syllable linked to a H i g h tone bears stress, w h i c h is
m a r k e d with an asterisk.
T h e r e p r e s e n t a t i o n in (7) is s u g g e s t i v e in several respects. If w e factor out vowel
length, the four accents can be r e d u c e d to j u s t t w o — t h e falling H ( L ) and the rising
( L ) H H ( L ) t y p e . M o r e o v e r , the t w o a c c e n t t y p e s can b e k e p t distinct even if the
L o w tones are left out of the structure: the falling accents can simply b e c h a r a c -
terized as singly linked H i g h s , and the rising accents as d o u b l y linked H i g h s .
(8) a. Short rising: marama
V
H
b. Short falling: kucica
I
H

c. L o n g rising: raazlika
V
H

d. L o n g falling: raadnica
i
H
370 Draga Zee

It is thus sufficient to represent only H i g h t o n e s in the lexical c o m p o n e n t ; L o w

tones are i n t r o d u c e d b y a default rule, late in the postlexical c o m p o n e n t (for d e -
tails see Inkelas a n d Z e e , 1988).
T h e representation in (8) can b e simplified even further: o n e of the t w o associa-
tion lines linked to the H i g h is r e d u n d a n t a n d can therefore b e d e r i v e d by rule.
T h i s b r i n g s us to the rule r e s p o n s i b l e for accent retraction, the p h o n o l o g i c a l
c h a n g e focused on here. T h e s y n c h r o n i c c o u n t e r p a r t of this p r o c e s s is a tone-
spreading rule: a linked H i g h t o n e spreads to the i m m e d i a t e l y p r e c e d i n g m o r a ,
deriving the forms in (8) from those in ( 1 0 ) . 8

(9) SPREADING: u ji

\ l
H

(10) a. Short rising: marama

I
H
b. Short falling: kucica

H
c. L o n g rising: raazlika
I
H
c. L o n g falling: raadnica

A s formulated a b o v e , S p r e a d i n g applies r i g h t to left. In s u p p o r t of this, w e c a n

b r i n g u p pairs of f o r m s as in (11).

(11) a. raadnik 'worker'

b. neraadnik 'nonworker'
N
H

Crucial to o u r p o i n t is the fact that the singly linked H i g h in (1 l a ) spreads to the

prefix in (1 l b ) . T o n e linked to a stem-final syllable w o u l d not spread in a similar
9

fashion to the following suffix. T h e historical c h a n g e of accent retraction also

operated right to left, j u s t like its synchronic counterpart.
M o s t importantly, S p r e a d i n g alone a c c o u n t s for the distribution of accents in
the N e o - S t o k a v i a n . T h e distribution of the N e o - S t o k a v i a n accents is described in
the literature as follows: (1) the falling accents a p p e a r only on word-initial syl-
lables, (2) the rising accents a p p e a r on any syllable o t h e r than the final one, and
(3) m o n o s y l l a b i c w o r d s can only b e a r a falling accent (see B r o w n e a n d M c C a w -
Rule Domains and Phonological Change 371

ley, 1 9 6 5 ; Halle, 1 9 7 1 ; K e n s t o w i c z , 1974; L e h i s t e and Ivic, 1986, a n d the refer-

e n c e s therein). G i v e n the formulation of S p r e a d i n g , these distributional p r o p e r t i e s
follow naturally. T h e falling accents o c c u r o n the word-initial syllable b e c a u s e this
is the only position w h i c h can b e o c c u p i e d b y an u n s p r e a d H i g h t o n e . T h e stressed
syllable of a rising accent c a n n o t o c c u p y word-final position b e c a u s e the leftmost
syllable c o n t a i n i n g a H i g h tone will never b e word-final in forms that h a v e under-
g o n e S p r e a d i n g . Finally, rising accents, w h i c h are a result of S p r e a d i n g , c a n n o t
a p p e a r on m o n o s y l l a b i c w o r d s since they can only b e associated with f o r m s w h i c h
p o s s e s s at least t w o syllables.
S p r e a d i n g t h u s creates a system with four pitch accents typical of the N e o -
S t o k a v i a n dialects. In addition, S p r e a d i n g identifies the t o n e - b e a r i n g unit in
S e r b o - C r o a t i a n . S i n c e c o n t o u r tones are found o n long b u t not o n short v o w e l s ,
w e are led to c o n c l u d e that the t o n e - b e a r i n g unit is the m o r a a n d that tones are
linked to t o n e - b e a r i n g units in a o n e - t o - o n e fashion.

2.2. O l d S t o k a v i a n D i a l e c t s

T h e O l d S t o k a v i a n dialects h a v e e s c a p e d the effect of accent retraction. T h e

four N e o - S t o k a v i a n accents illustrated in (6) c o r r e s p o n d to the following O l d Sto-
kavian accents.

(12) a. Short falling: marama mafa\ma 'scarf'

b. S h o r t falling: tiucica ku\cica 'house (dim.)'
c. Short falling: raztika raaz\li\ka 'difference'
d. L o n g falling: rddnica Ta\adnica 'worker'

T h e O l d S t o k a v i a n thus p o s s e s s e s t w o a c c e n t t y p e s , the long falling a n d the

short falling; a n d if w e factor out vowel length, the r a n g e of accents is r e d u c e d to
( L ) H ( L ) as the only a c c e n t type. T h e p h o n e t i c p r o p e r t i e s of the pitch accents re-
m a i n c o n s t a n t across dialects: the falling O l d S t o k a v i a n accents can t h u s b e r e p -
resented in exactly t h e s a m e fashion as the falling N e o - S t o k a v i a n a c c e n t s .

(13) a. Short falling: marama

I I I
LH L

b. Short falling: kucica

I V
HL

c. Short falling: raazlika

V II
L HL

d. L o n g falling: raadnica
i v
H L
372 Draga Zee

O n c e the representation is simplified by taking the L o w tones out of the structure,

w e arrive at the lexical representations, w h i c h are identical for the t w o g r o u p s of
dialects. 10

marama
(14) a. Short falling

kucica
b. Short falling

raazlika
c. Short falling
H
c. L o n g falling: raadnica

T h e s e are precisely the forms w h i c h in the N e o - S t o k a v i a n serve as input to

S p r e a d i n g . S p r e a d i n g is thus solely r e s p o n s i b l e for the accentual differences b e -
t w e e n the t w o dialect g r o u p s . If w e factor out the effect of S p r e a d i n g , the t w o
dialect g r o u p s b e c o m e identical in their accentual p r o p e r t i e s .

2.3. Lexically Toneless Forms

In order to c o m p l e t e m y a c c o u n t of the surface accentual properties in the t w o

dialect g r o u p s , I n e e d to i n t r o d u c e the distinction b e t w e e n lexically toneless stems
and stems characterized b y lexical tone. N o t e the difference in tonal properties
b e t w e e n the sets of verbal f o r m s in ( 1 5 ) a n d ( 1 6 ) .

(15) F O R M S W I T H LEXICAL T O N E :
a. tone ' s i n k - 3 S g P r e s ' (imperfective)
I
H
b. potone ' s i n k - 3 S g P r e s ' (perfective)

(16) L E X I C A L L Y T O N E L E S S FORMS:
a. tonu ' s i n k - 3 S g A o r i s t ' (imperfective)

H
b. potonu 4
s i n k - 3 S g A o r i s t ' (perfective)

H
Rule Domains and Phonological Change 373

T h e situation in (15) is familiar: in the N e o - S t o k a v i a n dialects, the H i g h on the

stem in (15a) spreads to the prefix in (15b); in the O l d Stokavian, the H i g h r e m a i n s
on the stem a n d the prefix eventually receives a L o w t o n e . T h i s is in fact the
pattern exhibited b y t h o s e forms w h i c h are associated with lexical tone. T h e f o r m s
in (16a) and (16b), however, a p p e a r to u n d e r g o tonal rules separately. T h i s pattern
is to b e found with toneless s t e m s , that is, with t h o s e that lack lexical tone. T h e
distinction b e t w e e n the forms in (15) and in (16) c o r r e s p o n d s to the w e l l - k n o w n
bifurcation into accented a n d u n a c c e n t e d stems found in pitch-accent l a n g u a g e s
(see the a n a l y s e s of Sanskrit and L i t h u a n i a n in H a l l e a n d Kiparsky, 1 9 8 1 ; Kipar-
sky, 1 9 7 3 ; and K i p a r s k y and Halle, 1977; as well as Poser's 1984 analysis of
J a p a n e s e ) . T h e p r e s e n t tense verbal f o r m s , those in (15), h a v e lexical tone, w h i l e
the aorist f o r m s , t h o s e listed in (16), are toneless. Toneless forms receive tone b y
virtue of Initial H i g h Insertion, a rule w h i c h assigns a H i g h to their initial m o r a s
(for details, see Inkelas and Z e e , 1 9 8 8 ) . 11

(17) INITIAL H I G H INSERTION: [JUL —> [JJL

T h e effect of this rule is to e n s u r e that, eventually, all f o r m s obtain accent; in o u r

t e r m s , this m e a n s that lexically toneless f o r m s will h a v e to b e supplied w i t h a H i g h
t o n e at s o m e p o i n t of the derivation. In s u m , w e distinguish t w o k i n d s of H i g h
t o n e s — t h o s e that are associated with f o r m s lexically, a n d t h o s e that are assigned
by virtue of Initial H i g h Insertion.
T h e rules p r o p o s e d thus far a c c o u n t for the pitch p r o p e r t i e s that L e h i s t e and
Ivic attribute to accent types in various S t o k a v i a n dialects. L o c a t i o n of the syllable
w h o s e v o w e l exhibits increase in duration is fully predictable from pitch: in all
S t o k a v i a n dialects, this is the leftmost syllable that contains a H i g h t o n e . Stress
a s s i g n m e n t is thus identical for the t w o g r o u p s of dialects; in those dialects that
p o s s e s s S p r e a d i n g , this rule will h a v e to p r e c e d e the a s s i g n m e n t of stress.
W e n o w turn to the d o m a i n s within w h i c h S p r e a d i n g applies. It will b e s h o w n
that this rule is associated b o t h with the cyclic a n d with the postcyclic d o m a i n s .
Its postcyclic application is discussed in Section 3 , and its cyclic application, in
Section 4.

3. P O S T C Y C L I C A P P L I C A T I O N S O F S P R E A D I N G

In this section I focus on the postcyclic applications of S p r e a d i n g in three N e o -

Stokavian dialects, w h i c h differ in their c h o i c e of postcyclic d o m a i n s for this rule.
W h i l e all three dialects select the p h o n o l o g i c a l w o r d as a d o m a i n for S p r e a d i n g ,
o n e of the dialects also selects the p h o n o l o g i c a l p h r a s e . M o r e o v e r , S p r e a d i n g m a y
374 Draga Zee

apply either within the lexical or within the postlexical version of the p h o n o l o g i c a l
w o r d . T h e p h o n o l o g i c a l w o r d in its postlexical size is referred to here, informally,
as the clitic g r o u p ; this t e r m is taken to refer to a d o m a i n size rather than to an
actual p r o s o d i c constituent.
Before t u r n i n g to the p o s t c y c l i c d o m a i n s of S p r e a d i n g , I will digress briefly to
c o m m e n t on t h e p r o s o d i c p r o p e r t i e s of S e r b o - C r o a t i a n w o r d s . F o l l o w i n g Inkelas
( 1 9 8 9 ) , I classify lexical i t e m s in S e r b o - C r o a t i a n into those that are prosodically
salient and t h o s e that lack p r o s o d i c salience. Prosodically salient forms will m a p
into p h o n o l o g i c a l w o r d s . F o r m s that lack p r o s o d i c salience are all clitics in S e r b o -
Croatian; they include certain p e r s o n a l p r o n o u n s , prepositions, conjunctions, and
particles. S i n c e S e r b o - C r o a t i a n clitics s u b c a t e g o r i z e for the p h o n o l o g i c a l word,
they will b e i n c o r p o r a t e d into this particular p r o s o d i c constituent, f o r m i n g an en-
larged d o m a i n .
I first e x a m i n e the p o s t c y c l i c d o m a i n s of S p r e a d i n g for the three N e o - S t o k a v i a n
dialects to b e d i s c u s s e d h e r e a n d then c o m p a r e t h e m with those of Initial H i g h
Insertion. It will b e s h o w n that the d o m a i n s of the t w o rules c o i n c i d e in e a c h of
the dialects to b e e x a m i n e d below. T h e d o m a i n s of Initial H i g h Insertion will in
fact play an i m p o r t a n t role in d e t e r m i n i n g the d o m a i n s of S p r e a d i n g for e a c h of
the dialects.

3.1. D o m a i n s of Spreading

T h e three N e o - S t o k a v i a n dialects described h e r e are N e o - S t o k a v i a n 1 ( N S 1 ) , in

w h i c h S p r e a d i n g o p e r a t e s within the clitic g r o u p , N e o - S t o k a v i a n 2 ( N S 2 ) , in w h i c h
this rule c h o o s e s the d o m a i n of the p h o n o l o g i c a l word, a n d N e o - S t o k a v i a n 3
( N S 3 ) , in w h i c h S p r e a d i n g applies both within the p h o n o l o g i c a l w o r d a n d within
the clitic g r o u p . M o r e o v e r , in N S 1 S p r e a d i n g also applies within the d o m a i n of
1 2

the p h o n o l o g i c a l p h r a s e .
In N S 1 , a word-initial H i g h s p r e a d s o n t o the p r e c e d i n g clitic, as in (18).

(18) NS1: u kucu 'into (the) house'

N
H

If the p r e c e d i n g clitic is disyllabic, S p r e a d i n g will affect its s e c o n d syllable.

(19) N S 1'. ispred kuce 'in front of (the) house'

A s a c o n s e q u e n c e , w o r d s with an initial lexically assigned H i g h exhibit the

following alternation in N S 1.
Rule Domains and Phonological Change 375

(20) a. vidim kucu b. usao je u kucu

H H
N N
H H
see(lsg) house(Acc)
entered Aux in house
T see a house.
'He got into the house.'
T o n e s p r e a d i n g o n t o a proclitic is a h i g h l y general p h e n o m e n o n in this dialect.
In o r d e r to illustrate this, w e g i v e a few m o r e e x a m p l e s (taken from V u k o v i c ,
1940). In (21a), t o n e s p r e a d s o n t o a p r e p o s i t i o n from a p r e c e d i n g adjective;
and in (21b), a H i g h linked to the initial syllable of a verb spreads to the
complementizer.

(21) a. na brzem konju 'on (a) faster horse'

N N
H H
b . da vidiim 'that (I) see'

N
H
M o r e o v e r , in N S 1 w e also e n c o u n t e r applications of S p r e a d i n g within the p h o -
nological p h r a s e . E x a m p l e s in (22) are t a k e n from V u k o v i c ( 1 9 4 0 ) , and t h o s e in
(23) from R u z i c i c ( 1 9 2 7 ) .
(22) a. dvabrata 'two brothers'

N
b . tries ijedna kuca 'thirty one houses'

N
C. cetiri litre 'four liters'

N
H
'my brother (Vocative)'
d. moj brate
N
H
e. sve Ijeto 'all summer'

N
H
'all night'
f. svu nooc

N
H

N
g. svu dugu nooc 'all (the) long night'

H H
376 Draga Zee

(23) a. sedam kuuca 'seven houses'

\J
H
b . dvije kuce 'two houses'

N
H
C. tri knjige 'three books'
N
H
d. pet dinaaraa 'five dinars'

N
H
It should b e p o i n t e d out that S p r e a d i n g will not apply b e t w e e n j u s t any t w o w o r d s
in a p h o n o l o g i c a l p h r a s e . In all e x a m p l e s that h a v e b e e n reported, the target of
S p r e a d i n g is either a n u m e r a l or a nonclitic p r o n o u n . T h e s e forms receive tone
optionally in the dialect d e s c r i b e d b y V u k o v i c and R u z i c i c as well as in other
dialects d i s c u s s e d in this article. E x a m p l e s in (24), taken from V u k o v i c , illustrate
this point. In (24a) the p r o n o u n sav air is toneless, a n d in (24b) it is associated
4

with an initial H i g h tone; the n u m e r a l in (24c) is toneless, and that in (24d) is

supplied with tone.

(24) a. sve proljece 'all spring'

N
H
b . svi su . . . 'all are.

C. dvaes koonjaa 'twenty horses'

N
H
d. deseet koosaa 'ten scathes'

H H

O n l y those f o r m s that r e m a i n toneless will b e targets of S p r e a d i n g . T h e fact that

forms linked to tone d o not b e c o m e targets of S p r e a d i n g within the p h o n o l o g i c a l
p h r a s e is consistent with the generalization that a w o r d m a y b e associated with at
m o s t o n e H i g h t o n e (see n o t e 11). M o r e o v e r , S p r e a d i n g applies optionally within
the p h o n o l o g i c a l p h r a s e ; R u z i c i c lists alternants of (22f,g) in w h i c h the p r o n o u n
sav ' a l l ' a p p e a r s toneless.
In N S 2 , the next dialect w e will d e s c r i b e here, S p r e a d i n g is w o r d - b o u n d e d .
Word-initial H i g h s d o not u n d e r g o this rule, as s h o w n in (25); only t h o s e H i g h s
Rule Domains and Phonological Change 377

that are linked to a noninitial syllable of a w o r d are affected by S p r e a d i n g in this

dialect.

(25) NS2: u kucu i n t o (the) h o u s e '

ispred kuce 4
in front of (the) h o u s e '

H
In this dialect, therefore, w e e n c o u n t e r n o alternation of the sort illustrated in
(20), as s h o w n b y the following e x a m p l e s .
(26) a. victim kucu

I I
H H
see( 1 sg) house(Acc)
T see a house.'
b. usao je u kucu

\l I
H H
entered Aux in house
'He got into the house.'
Finally, N S 3 p o s s e s s e s alternant f o r m s like t h o s e listed in ( 2 7 ) - ( 2 8 ) (taken
from N i k o l i c , 1970, a n d M o s k o v l j e v i c , 1 9 2 7 - 2 8 ) , w h i c h strongly suggests that
S p r e a d i n g applies b o t h within the p h o n o l o g i c a l w o r d a n d within the clitic g r o u p .
1 3

(27) a. u banju 'to (the) spa'

I
H
b . u banju 'to (the) spa'

N
H
(28) a. u Sapcu 'in Sabac'

I
H
b . u Sapcu 'in Sabac'

N
H
Variation that c h a r a c t e r i z e s N S 3 will b e attributed to the optionality of rule appli-
cation: in this dialect S p r e a d i n g applies obligatorily within the p h o n o l o g i c a l w o r d ,
but optionally within the clitic g r o u p . 1 4
378 Draga Zee

T h e table in (29) s u m m a r i z e s the applications of S p r e a d i n g in the three N e o -

S t o k a v i a n dialects d i s c u s s e d h e r e .

(29)
NS1 NS2 NS3

Lexical ? Yes Yes

phonological word
(phonological w o r d )
Postlexical Yes No Yes(Opt)
phonological word
(clitic g r o u p )
Phonological phrase Yes(Opt) No No

I have s h o w n that in N S 2 S p r e a d i n g o p e r a t e s within the p h o n o l o g i c a l w o r d but

not within the clitic g r o u p ; a n d that in N S 3 S p r e a d i n g operates b o t h within the
p h o n o l o g i c a l w o r d and within the clitic g r o u p . L e s s clear at this p o i n t is w h a t
h a p p e n s in N S 1. W h i l e w e h a v e clear e v i d e n c e for the application of S p r e a d i n g
within the clitic g r o u p a n d the p h o n o l o g i c a l p h r a s e , there is little e v i d e n c e that
this rule also applies within the p h o n o l o g i c a l w o r d . This p r o b l e m is a d d r e s s e d in
Section 5.

3 . 2 . C o m p a r i s o n w i t h Initial H i g h I n s e r t i o n

If w e n o w e x a m i n e the application of Initial H i g h Insertion, a very interesting

fact e m e r g e s . In e a c h of the dialects, the d o m a i n of Initial H i g h Insertion overlaps
with that of S p r e a d i n g : in N S 1 the t w o rules m e e t in the d o m a i n of the clitic g r o u p ,
in N S 2 they m e e t within the p h o n o l o g i c a l word, and in N S 3 they overlap both in
the p h o n o l o g i c a l w o r d and in the clitic g r o u p . T h e facts p r e s e n t e d b e l o w are in
m a n y w a y s a replication of those found with S p r e a d i n g .
In N S 1 , Initial H i g h Insertion assigns a H i g h t o n e to the proclitic, if there is o n e
in the structure.
(30) NS1: ugraad 'in (the) t o w n '

H
If the proclitic is disyllabic, t o n e is assigned to its first syllable. This follows
from the w a y the rule is f o r m u l a t e d in (17): t o n e is assigned to the leftmost m o r a
of the relevant d o m a i n .
(31) N S 1: iza graada ' b e h i n d (the) t o w n '

H
In N S 2 , however, the rule ignores the clitic and operates on the word-ini-
tial m o r a .
Rule Domains and Phonological Change 379

(32) NS2: u graad

W e thus find the following alternation in N S 1 b u t not in N S 2 .

(33) a. vidim graad b. otisao je u graad

I I \ \
H H H H
see(lsg) city (Acc) went Aux to city
T see a city' 'He went to the city.'
In N S 3 , w e e n c o u n t e r alternant f o r m s of the following kind, w h i c h I attrib-
ute to optional rule application in the lexical, a l t h o u g h not in the postlexical,
domain. 1 5

(34) NS3: u glaavu versus u glaavu i n t o (the) h e a d '

H H

(35) NS3: ispodleda versus ispodleda ' u n d e r (the) i c e '

I I
H H

T h e applications of Initial H i g h Insertion in the three N e o - s t o k a v i a n dialects

are s u m m a r i z e d in ( 3 6 ) . 1 6

(36)
NS1 NS2 NS3
Lexical No Yes Yes(Opt)
phonological word
(phonological word)
Postlexical Yes No Yes
phonological word
(clitic g r o u p )

3 . 3 . I n t e r a c t i o n of S p r e a d i n g a n d Initial H i g h I n s e r t i o n

3.3.1. INNS3

In N S 3 , S p r e a d i n g and Initial H i g h Insertion apply b o t h lexically a n d postlexi-

cally and could therefore potentially interact with e a c h other. In particular, a H i g h
assigned to a p h o n o l o g i c a l w o r d by virtue of Initial H i g h Insertion m a y spread to
a proclitic within the d o m a i n of the clitic g r o u p . T h e following alternant f o r m s ,
from Nikolic ( 1 9 7 0 ) , support this prediction.
380 Draga Zee

(37) a. u pasnjaaku 'in (the) pasture'

I
H
b . u pasnjaaku 'in (the) pasture'

N
H
T h e alternation s h o w n in (37) follows from the optionality of rule application:
Initial H i g h Insertion is optional within the lexical p h o n o l o g i c a l word. M o r e o v e r ,
w e also find three-way alternations as in (38) (again from N i k o l i c , 1970); (38c) is
d u e to the optionality of S p r e a d i n g within the postlexical p h o n o l o g i c a l w o r d (i.e.,
the clitic g r o u p ) .

(38) a. iz salaasa 'from (the) field'

H
b . iz salaasa 'from (the) field'

N
H
C. u salaasu 'in (the) field'

I
H

B e l o w are given the derivations of forms in (38) iz salaasa u salaasu

(39) iz salaasa
H
N H
H
salaasa salaasu
Lexical phonological word: salaasa

Spreading
salaasa salaasu
I I
Initial H i g h I j
H H

Postlexical p h o n o l o g i c a l w o r d : iz salaasa iz salaasa u salaasu

H H
Spreading
iz salaasa

H
Initial H i g h iz salaasa
I
I
Rule Domains and Phonological Change 381

Lexically toneless forms w h i c h spread their H i g h o n t o the proclitic are treated

in the literature as a n o m a l o u s and a c c o u n t e d for in t e r m s of a n a l o g y (Nikolic,
1970). U n d e r the analysis d e v e l o p e d here, however, there is n o n e e d to a s s u m e
that such forms are in any way exceptional.

3.3.2. INNSI

In N S 1 , S p r e a d i n g applies b o t h w i t h i n the p h o n o l o g i c a l w o r d a n d within the

p h o n o l o g i c a l p h r a s e , w h i l e Initial H i g h Insertion applies only within the p h o n o -
logical w o r d . A s a result, H i g h s assigned by Initial H i g h Insertion at the level of
the p h o n o l o g i c a l w o r d m a y u n d e r g o S p r e a d i n g within the p h o n o l o g i c a l p h r a s e .
V u k o v i c ( 1 9 4 0 : 2 7 1 ) lists several cases w h o s e tonal p a t t e r n s are a c c o u n t e d for
straightforwardly if v i e w e d as resulting from the interaction of Initial H i g h Inser-
tion a n d S p r e a d i n g . A m o n g the forms listed are pairs of f o r m s as in (40), in w h i c h
the w o r d noc ' n i g h t ' , w h i c h is lexically toneless a n d therefore receives t o n e by
virtue of Initial H i g h Insertion, triggers different tonal p a t t e r n s on the p r e c e d -
ing w o r d .

(40) a. za nooc 'in (a) night'

b . svu nooc 'all night'

N
H
This difference is p r e d i c t e d u n d e r the analysis p r o p o s e d h e r e . In N S 1 , Initial H i g h
Insertion applies only within the postlexical p h o n o l o g i c a l w o r d , that is, within the
clitic g r o u p . In (40a), this rule will apply to the entire prepositional p h r a s e , w h i c h
forms a postlexical p h o n o l o g i c a l word, a n d will link the H i g h to the proclitic za;
and in (40b) it will apply only to the form nooc, w h i c h by itself constitutes a
postlexical p h o n o l o g i c a l word. T h e toneless form svu, w h i c h is an a p p r o p r i a t e
target for S p r e a d i n g , u n d e r g o e s this rule within the p h o n o l o g i c a l p h r a s e , that is,
at the point at w h i c h it shares c o n s t i t u e n c y with nooc. T h e s e steps are s h o w n
explicitly in the following derivation.

(41) za nooc svu nooc

I N
H H

Phonological word: za nooc nooc

Spreading
Initial H i g h za nooc nooc
i i
j i
H H
382 Draga Zee

Phonological phrase: za nooc svu nooc

H H
Spreading svu nooc

4. C Y C L I C A P P L I C A T I O N O F S P R E A D I N G

T h u s far w e h a v e seen that the N e o - S t o k a v i a n dialects p o s s e s s the rule of

S p r e a d i n g w h i c h is absent from the O l d S t o k a v i a n g r o u p ; and that the N e o -
S t o k a v i a n dialects vary as to w h i c h of the postcyclic d o m a i n s is selected by this
rule. In this section I a r g u e that, in the N e o - S t o k a v i a n g r o u p of dialects, S p r e a d i n g
applies not only postcyclically, as w e saw in the p r e v i o u s section, but also cycli-
cally. T h e a r g u m e n t itself will take us s o m e w h a t afield. W e will again b r i n g in the
O l d - S t o k a v i a n dialects, as a p o i n t of c o m p a r i s o n for the N e o - S t o k a v i a n g r o u p .
E v i d e n c e for the cyclic application of S p r e a d i n g in the N e o - S t o k a v i a n is p r o v i d e d
b y accent shift in the genitive plural, w h i c h follows different patterns in the two
major dialect g r o u p s . T h e s e d i v e r g i n g patterns will receive a simple characteriza-
tion u n d e r the a s s u m p t i o n that S p r e a d i n g operates in the N e o - S t o k a v i a n dialects
but not in the O l d S t o k a v i a n o n e s . In o r d e r to d e m o n s t r a t e this, I n e e d to digress
into the territory of lexical t o n e . Sections 4.1 and 4.2 are d e v o t e d to the properties
of lexical t o n e shared by the t w o dialect g r o u p s . In Section 4.3 I turn to accent
shift in the genitive plural, w h i c h b r i n g s in accentual differences b e t w e e n the t w o
g r o u p s of dialects.

4.1. Lexical Tone

It h a s generally b e e n h e l d that the distribution of S e r b o - C r o a t i a n accents is

fairly free, if not for the m o s t p a r t r a n d o m ( B r o w n e and M c C a w l e y , 1965; Halle,
1 9 7 1 ; K e n s t o w i c z , 1974; L e h i s t e a n d Ivic, 1986). In k e e p i n g with this, Inkelas
and Z e e ( 1 9 8 8 ) p r o p o s e to r e d u c e w h a t a p p e a r s to b e r a n d o m distribution of a c -
cent to r a n d o m distribution of lexical t o n e ; in their analysis, H i g h tones are p r e -
linked in the u n d e r l y i n g form. L e x i c a l tone, however, is not devoid of pattern. If
tone is p r e l i n k e d in the u n d e r l y i n g form, then w e will h a v e to stipulate the a b s e n c e
of the following u n d e r l y i n g c o n f i g u r a t i o n .
17

(42)

H
Rule Domains and Phonological Change 383

T h e p r o h i b i t e d configuration in (42) points to a g a p in the distribution of lexical

tone: a H i g h t o n e m a y not b e linked to the r i g h t m o s t m o r a of a h e a v y syllable. In
fact, Inkelas and Z e e (1988) treat the prohibited configuration in (42) as a c o n -
straint o n u n d e r l y i n g form. T h i s m o v e , however, is r e d u n d a n t . T a k i n g this g a p as
a point of d e p a r t u r e , I will s h o w that the distribution of tone is fully predictable.
T h e tonal configurations associated with lexical t o n e are those given in (43),
a n d they are both c h a r a c t e r i z a b l e b y rule. I a c c o u n t for t h e configuration in (43a)
in Section 4 . 1 . 1 , b y p r o p o s i n g a cyclic t o n e linking rule. T h e configuration in
(43b) is attributed to a relinking rule to b e d i s c u s s e d in Section 4.1.2.

(43) a. a b. a

I A
I I
H H

4.1.1. T O N E LINKING R U L E

T h e accentual p r o p e r t i e s of various inflectional classes serve as principal m o -

tivation for the tonal rule w h i c h creates the configuration in (43a). Crucial for m y
a r g u m e n t is the c o n t r a s t b e t w e e n forms with a light stem-final syllable, as in (44),
and t h o s e with a h e a v y stem-final syllable, as in (45). T h e forms in (44) a n d (45)
are all n o m i n a l s t e m s , and this is the class of stems to w h i c h I limit m y s e l f h e r e ;
the c l a i m s to b e m a d e e x t e n d naturally to the adjectival a n d verbal stems (for
details see Z e e , 1988).
Let us first e x a m i n e the p r o p e r t i e s of s t e m s that e n d in a light syllable. All the
c a s e f o r m s in (44) (with the e x c e p t i o n of the genitive plural, w h i c h is d i s c u s s e d in
Section 4.3) h a v e lexical tone linked to the final syllable of the s t e m . 1 8

(44) Masculine Feminine Neuter

'deer' 'raspberry' 'knee'
Sg. Nom. jelen malin-a kolen-o
1
1
H H H
Gen. jelen-a malin-e kolen-a
1
1
H H H
Dat. jelen-u malin-i kolen-u
1
1
H H H
Acc. jelen-a malin-u kolen-o

H H H
384 Draga Zee

Instr. jelen-om malin-om kolen-om

H H H

Loc. jelen-u malin-i kolen-u

H H H

malin-e
PI. Nom. jelen-i kolen-a
I
H
H malin-a H
Gen. jelen-a kolen-a
H H H
Dat. jelen-ima malin-ama kolen-ima

H H H
Acc. kolen-a
jelen-e malin-e
I
H
Instr. H
jelen-ima H
malin-ama kolen-ima

H H H
Loc. jelen-ima malin-ama kolen-ima

H H H

F o r m s w h o s e stem-final syllable is h e a v y d o not bear lexical tone on this syl-

lable (with the e x c e p t i o n of the n o m i n a t i v e singular form, to b e discussed in S e c -
tion 4.2). Lexical tone a p p e a r s , instead, on the suffix.

(45) Masculine Feminine Neuter

'hero' 'guitar' 'principle'
Sg. Nom. junaak gitaar-a naceel-o
I
H H H
Gen. junaak-a
naceel-a
I gitaar-e
H
H H
Dat. junaak-u naceel-u
gitaar-i I
H H
H
naceel-o
Acc. junaak-a
gitaar-u

H I H
H
Rule Domains and Phonological Change 385

Instr. junaak-om gitaar-om naceel-om

H H H
Loc. junaak-u gitaar-i naceel-u
I I I
H H H

PI. Nom. junaac-i gitaar-e naceel-a

H H H
Gen. junaak-a gitaar-a naceel-a

H H H
Dat. junaac-ima gitaar-ama naceel-ima

H H H
Acc. junaak-e gitaar-e naceel-a
I I I
H H H
Instr. junaac-ima gitaar-ama naceel-ima

H H H
Loc. junaac-ima gitaar-ama naceel-ima

H H H

T h i s distinction is c a p t u r e d b y a cyclic rule of t o n e a s s i g n m e n t . T h e bifurcation

into a c c e n t e d and u n a c c e n t e d s t e m s m e n t i o n e d earlier b e c o m e s relevant at this
point. T h i s distinction is r e p r e s e n t e d u n d e r l y i n g l y as in (46).

(46) a. stem b. stem

H

A c c e n t e d s t e m s are affiliated with a floating H i g h in their u n d e r l y i n g form, as in

(46a), w h i c h k e e p s t h e m apart from the u n a c c e n t e d o n e s . T h e rule r e s p o n s i b l e for
the a s s i g n m e n t of lexical tone will o p e r a t e on accented s t e m s , in the following
fashion. 19

(47) T O N E LINKING R U L E (informal): L i n k the floating tone to the stem-final

syllable if and only if it d o e s not
branch.

T h i s rule will link a H i g h tone to the final syllable of the stem if this syllable is
light, as in (44). If the stem-final syllable is heavy, the tonal rule will fail to apply
on the stem c y c l e ; but it will reapply on the next h i g h e r cycle, assigning t o n e to
the suffix, as in (45). T h e relevant derivations are given in (48).
386 Draga Zee

(48)
C Y C L E 1: [junaak] [jelen]
H H
Tone Linking — [jelen]

H
C Y C L E 2: [[junaak] a] [[jelen] a]
H I
H
Tone Linking [[junaak] a]

T h e tonal rule operates only o n those forms w h i c h are associated with lexical
tone. In contrast, toneless f o r m s receive a H i g h t o n e on their initial syllable, b y
virtue of Initial H i g h Insertion (discussed in Section 2.3), as s h o w n in the deriva-
tions of t w o toneless m a s c u l i n e f o r m s , oblaak ' c l o u d ' and dever ' b r o t h e r - i n - l a w ' .

(49)
CYCLIC:
C Y C L E 1: [oblaak] [dever]
Tone Linking — —
C Y C L E 2: [oblaak] a] [dever] a]
Tone Linking — —
POSTCYCLIC: Initial H i g h [[oblaak] a] [[dever] a]

H H

W e thus o b s e r v e clear tonal differences b e t w e e n toneless s t e m s a n d s t e m s w h i c h

are lexically affiliated with t o n e .
T h e rule of T o n e L i n k i n g is also r e s p o n s i b l e for the distribution of accent in
derived f o r m s . W h e n the four s t e m s in ( 4 8 ) - ( 4 9 ) are c o m b i n e d with the p o s s e s -
sive suffix -ov, the pattern r e m a i n s unaltered. A g a i n , accented stems like jelen
receive tone o n the stem-final syllable, and those like junaak, on the first syllable
of the suffix; u n a c c e n t e d s t e m s receive t o n e by Initial H i g h I n s e r t i o n . 20

(50) [jelen] ov versus [dever] ov

H H

(51) [junaak] ov versus [oblaak] ov

H H

T h e distinction b e t w e e n stems e n d i n g in a short vowel and those e n d i n g in a long

o n e p e r v a d e s a g o o d portion of the accentual system; w e find it in fact in all the
Rule Domains and Phonological Change 387

forms w h i c h p o s s e s s lexical tone. F u r t h e r m o r e , this generalization has interesting

c o n s e q u e n c e s for m o n o s y l l a b l e s : T o n e L i n k i n g will assign t o n e to a m o n o s y l l a b i c
stem w h o s e stem syllable is light, as s h o w n in ( 5 2 a ) ; o t h e r w i s e the H i g h is linked
to the suffix, as in (52b).

(52) a. jad-a 4
woe(gen.sg.)'

b. dvoor-a 'court(gen.sg.)'

T o n e l e s s s t e m s obtain tone on the stem syllable by virtue of Initial H i g h Inser-

tion; as a c o n s e q u e n c e , only lexically toneless f o r m s h a v e t o n e linked to the h e a v y
syllable of a m o n o s y l l a b i c stem.

(53) graad-a 'city(gen.sg.)'

I
H

U n d e r the a s s u m p t i o n that the place of accent is u n p r e d i c t a b l e , such regularities

in the distribution of accent across inflectional p a r a d i g m s w o u l d h a v e to b e treated
as m e r e l y accidental.

4.1.2. M E T A T O N Y

T o n e L i n k i n g thus derives the w e l l - f o r m e d configuration in ( 4 3 a ) , in w h i c h t o n e

is linked to the only m o r a of a light syllable, w h i l e e x c l u d i n g the illicit a r r a n g e -
m e n t in (42) with t o n e linked to the n o n h e a d m o r a of a h e a v y syllable. W e n o w
turn to t h e o t h e r w e l l - f o r m e d configuration, that in ( 4 3 b ) , in w h i c h t o n e is linked
to the h e a d m o r a of a h e a v y syllable. E v i d e n c e for this will c o m e from the b e h a v -
ior of the n o m i n a t i v e m a s c u l i n e singular form in (49) a b o v e . T h e relevant e x a m p l e
is r e p e a t e d in (54).

(54) junaak t
hero(nom.sg.)'

This is the only c a s e form in the entire p a r a d i g m that lacks a p h o n o l o g i c a l l y

realized d e s i n e n c e ; furthermore, the H i g h links h e r e to the stem-final syllable al-
t h o u g h this syllable is heavy, w h i c h c o u l d b e seen as a c o u n t e r e x a m p l e to the
operation of T o n e L i n k i n g . It is justified, however, to posit a vocalic d e s i n e n c e for
the n o m i n a t i v e singular, in the form of a yer v o w e l , an abstract s e g m e n t generally
associated with the vocalic s y s t e m s of Slavic l a n g u a g e s . In S e r b o - C r o a t i a n this
vowel a p p e a r s o n the surface only u n d e r a n a r r o w l y defined set of c i r c u m s t a n c e s :
a yer is vocalized as a if followed by a n o t h e r yer, o t h e r w i s e it is lost. W e will thus
388 Draga Zee

a s s u m e that the n o m i n a t i v e singular e n d i n g is a yer vowel ( M a r k e d here as d), as

in ( 5 5 ) . 2 1

(55) [junaak] a

Yer v o w e l s play an i m p o r t a n t role in s h a p i n g the S e r b o - C r o a t i a n accentual sys-

t e m (for details, see Z e e , 1988). M o r a s h e a d e d by yer v o w e l s can be linked to tone
by virtue of T o n e L i n k i n g ; and if the yer vowel d i s a p p e a r s from the structure, tone
relinks in predictable w a y s . In particular, tone will relink to the left, to the nearest
m o r a that p o s s e s s e s an s label. I refer to this relinking p r o c e s s as M e t a t o n y . 22

(56) METATONY: |JL (JL

T h e accentual properties of junaak can n o w b e a c c o u n t e d for in the following

fashion: tone d o c k s on the yer in the d e s i n e n c e , b y virtue of T o n e L i n k i n g ; w h e n
the yer is lost from the structure (since it d o e s not vocalize in this context), the
m o r a it h e a d s also delinks, and so d o e s the t o n e linked to it. T h e H i g h then relinks
to the nearest m o r a that p o s s e s s e s an s label.
T h e derivation in (57) s h o w s explicitly h o w tone is assigned to junaak. M o r e -
over, in order to d e m o n s t r a t e that derivational suffixes c o n t a i n i n g yer vowels b e -
have in exactly the s a m e fashion as their inflectional counterparts, w e also give
the derivation of junaacki ' h e r o i c ' , f o r m e d with the suffix -ask w h i c h c o n v e r t s
n o m i n a l stems into adjectives.

(57) C Y C L E 1: [junaak] [ junaak]

H H
Tone Linking:

C Y C L E 2: [[ junaak]a] [ [ junaak]ask]
H H

Tone Linking: [ [ junaak]a] [[ junaak]ask]

i i
i
H H

i
Yer Delinking: \[junaak]k] [ [ junaak]ask]
I I
H H
Metatony: [\junaak\k] [[ junaak]ask]
\ 4 \ 4
H H

other rules

OUTPUT: junaak junaacki

I l
H H
Rule Domains and Phonological Change 389

Crucially, tone relinks to t h e head m o r a of a heavy syllable, skipping over a

nonhead mora.

4.2. T o n e L i n k i n g a n d S y l l a b l e S t r u c t u r e

It is i m p o r t a n t to note at this point that T o n e L i n k i n g and M e t a t o n y , o n the o n e

hand, differ in their m o d e of application from S p r e a d i n g and Initial H i g h Insertion,
on t h e other. B o t h Initial H i g h Insertion a n d S p r e a d i n g c a n b e stated solely in
t e r m s of m o r a s and could hardly b e stated in any other t e r m s . B e i n g oblivious to
syllable structure, these t w o rules support t h e earlier observation that the tone
b e a r i n g unit in S e r b o - C r o a t i a n is the m o r a . T h e Tone L i n k i n g rule, however, does
m a k e reference t o syllable structure: tone is linked t o a light b u t n o t to a heavy
syllable. M e t a t o n y also n e e d s to b e c h a r a c t e r i z e d in t e r m s of syllable structure:
tone relinks t o t h e only m o r a of a light syllable, o r t o t h e h e a d m o r a of a heavy
syllable. A w a y of sorting o u t these facts c o u l d b e t o say that T o n e L i n k i n g a n d
M e t a t o n y c h o o s e the syllable as the t o n e - b e a r i n g unit, w h i l e S p r e a d i n g and Initial
H i g h Insertion m a k e reference to t h e m o r a . T h i s distinction c a n then b e h a n d l e d
b y level o r d e r i n g , that is, b y assigning T o n e L i n k i n g a n d M e t a t o n y t o t h e cyclic
level, a n d S p r e a d i n g and Initial H i g h Insertion to the postcyclic level, as in (58).

(58) CYCLIC LEVELS: Tone Linking

Metatony
POSTCYCLIC L E V E L S : Spreading
Initial H i g h Insertion

T h e distribution across levels in ( 5 8 ) is c o m p a t i b l e with t h e o r d e r i n g of these

rules, since T o n e L i n k i n g and M e t a t o n y feed S p r e a d i n g , and Initial H i g h Insertion
has t o follow all the applications of T o n e L i n k i n g in o r d e r to " k n o w " w h i c h forms
r e m a i n toneless.
A n o b v i o u s solution, w h i c h ultimately proves unsatisfactory, is to stipulate that
at t h e cyclic levels t h e t o n e - b e a r i n g unit is t h e syllable, w h i l e at t h e postcyclic
levels it is the m o r a . T h i s w a s in fact p r o p o s e d in Inkelas and Z e e ( 1 9 8 8 ) , in o r d e r
to a c c o u n t for the fact that tone can only a p p e a r o n the leftmost m o r a . T h e expla-
nation offered w a s that, at the early levels, tone resides o n syllables but p e r c o l a t e s
d o w n t o t h e head m o r a . T h i s m o v e , however, p r o v i d e s only a partial solution;
w h i l e it could w o r k for M e t a t o n y , it d o e s not w o r k for T o n e L i n k i n g , since it d o e s
not explain w h y this rule operates only o n light syllables. I therefore p u r s u e a
different path here, retaining the segregation into levels in (58), yet p r o p o s i n g that
the m o r a is the t o n e - b e a r i n g unit at all p h o n o l o g i c a l levels. T h e limited distribu-
tion of lexical t o n e is c a p t u r e d b y the following positive constraint.

(59) fx s

I
H
390 Draga Zee

T h e constraint in (59) o p e r a t e s at the cyclic levels a n d is turned off postcycli-

cally. A s a result, lexical t o n e m a y link only to the ^-labeled m o r a , that is, to the
m o r a that acts as the syllable h e a d . M y claim, then, is that t o n e m a y b e a s s i g n e d
only to m o r a s , a l t h o u g h not necessarily to all m o r a s . T h e T o n e L i n k i n g rule, stated
informally in (47), c a n n o w b e restated as in (60): the floating H i g h is linked to
the r i g h t m o s t m o r a of the d o m a i n .

(60) ix]

H
T h i s rule will b e able to o p e r a t e only if the domain-final m o r a is also the h e a d
m o r a , that is, if it p o s s e s s e s an s label. If the r i g h t m o s t m o r a h a p p e n s to b e the
n o n h e a d m o r a , p o s s e s s i n g a w label, the rule will fail to apply, since the m o r a
targeted b y the rule will not b e able to receive tone. T h u s , b y assigning t o n e to
only a subset of m o r a s , w e c a n derive the fact that the H i g h that u n d e r g o e s T o n e
L i n k i n g can b e linked to a light but not to a h e a v y s y l l a b l e . M o r e o v e r , the con-
23

straint in (59) also e x p l a i n s the functioning of M e t a t o n y . S i n c e only h e a d m o r a s

m a y receive t o n e at the cyclic level, the delinked H i g h can b e relinked b y M e t a -
tony only to the leftmost m o r a of the syllable. However, in the e x a m p l e s of
S p r e a d i n g that w e saw earlier, either the head or the n o n h e a d m o r a can b e linked
to tone. T h i s is b e c a u s e S p r e a d i n g applies in the postcyclic c o m p o n e n t , w h i l e
constraint (59), w h i c h prohibits t o n e a s s i g n m e n t to a n o n h e a d m o r a , is operative
only cyclically a n d is t u r n e d off at the postcyclic level. H o w e v e r , the fact that
S p r e a d i n g is a postcyclic rule d o e s not e x c l u d e the possibility that it also applies
cyclically. T h a t this is i n d e e d the c a s e is s h o w n in the following section.

4.3. Spreading as a Cyclic Rule

In the cyclic c o m p o n e n t , o n l y the h e a d m o r a , that is, the leftmost m o r a of the

h e a v y syllable or the only m o r a of the light syllable, m a y b e linked to t o n e . If
S p r e a d i n g applies at the cyclic level, it will h a v e to operate within this level's
constraints: it c o u l d a p p l y as in (61a), b u t not as in (61b) o r ( 6 1 c ) .

(61) a. a a b. *a a C. *a a

I I A I A I
|x s |JL S H'S M-w m ' M^s
MW
\ l X
J
H H H

A p p l i c a t i o n in (61b) is b l o c k e d b y virtue of constraint (59), w h i c h prohibits t o n e

a s s i g n m e n t to a n o n h e a d m o r a . S p r e a d i n g c a n n o t o p e r a t e as s h o w n in (61c) either,
since skipping over a m o r a w o u l d lead to a violation of locality: a multiply linked
t o n e m a y only b e linked to adjacent t o n e - b e a r i n g units.
Rule Domains and Phonological Change 391

T h u s , if S p r e a d i n g applies at the cyclic level, the spread H i g h will b e able to

link only to a light syllable; those applications in w h i c h the spread H i g h links to
the s e c o n d m o r a of a h e a v y syllable will be i n c o m p a t i b l e with the cyclic c o m p o -
nent. A s s o c i a t i n g s o m e of the applications of S p r e a d i n g with the cyclic, a n d others
with the postcyclic, c o m p o n e n t will h a v e a theoretically desirable c o n s e q u e n c e :
w e a d h e r e to the p r i n c i p l e that a rule applies as early as p o s s i b l e . B u t a further
a d v a n t a g e of p l a c i n g S p r e a d i n g a m o n g cyclic rules is that this p r o v i d e s a s i m p l e
a c c o u n t of the a c c e n t shift i n d u c e d by the genitive plural e n d i n g -a. T h i s accent
24

shift o p e r a t e s differently in the O l d and N e o - S t o k a v i a n dialects. In the f o r m e r

g r o u p accent will shift to any p r e c e d i n g syllable, w h i l e in the latter g r o u p it will
shift only to a light p r e c e d i n g syllable.
I first p r e s e n t the situation in the O l d S t o k a v i a n dialects: the H i g h linked to the
final syllable of the stem retracts o n e syllable to the left in the genitive plural
form. T h e e x a m p l e s b e l o w are a m o n g t h o s e listed in R e s e t a r ( 1 9 0 0 ) a n d Steva-
novic(1940).

(62)

a. jezik ( n o m . s g . ) , jezik-a (gen.sg.) versus jeziik-a (gen.pl.) ' t o n g u e '

H H H
b . jelen ( n o m . s g . ) , jelen-a (gen.sg.) versus jeleen-a (gen.pl.) ' d e e r '
l I l
H H H
c. bubreg ( n o m . s g . ) , bubreg-a (gen.sg.) versus bubreeg-a (gen.pl.) ' k i d n e y '

H H H

(63)

a. naarod ( n o m . s g . ) , naarod-a (gen.sg.) versus naarood-a (gen.pl.) ' p e o p l e '

I I I
H H H
b. saabor ( n o m . s g . ) , saabor-a (gen.sg.) versus saaboor-a (gen.pl.) 'fair'
I I I
H H H
c. proorok ( n o m . s g . ) , proorok-a (gen.sg.) versus proorook-a (gen.pl.) ' p r o p h e t '
I I I
H H H

T h e pattern e x h i b i t e d b y the genitive shift c a n b e captured b y a m o r p h o l o g i c a l l y

c o n d i t i o n e d d e l i n k i n g rule, in the spirit of K e n s t o w i c z ' s ( 1 9 7 4 ) p r o p o s a l : a H i g h
linked to the stem-final syllable is d e l i n k e d in the e n v i r o n m e n t of the genitive
plural e n d i n g -a. 25
V)2 Draga Zee

(64) C J I N P L Di L I N K I N G (OS): JUL] - a G e n P 1

t
H
T h e delinked H i g h tone will then a s s u m e its shifted position by virtue of M e -
tatony, the relinking rule p r o p o s e d in the previous section. In (65) are given the
derivations of the genitive plural forms jeleena and naarooda.
(65) CYCLE i: lone Linking jelen naarod

H
CYCLE 2.- jelen]a naarod]a \
Gem GcnP

I I
H H

G e n PI Delinking ( O S ) jelen]a naarod\a

' \ \
Metatony H H
jelen] a

naarod]
OUTPUT:
H
jeleena H
"I naarooda
H I
H
T h i s analysis, however, d o e s not extend to the N e o - S t o k a v i a n dialects, w h i c h
exhibit a different pattern of the genitive shift: t o . . . shifts to a p r e c e d i n g light
syllable, but not to a p r e c e d i n g ! avy syllable. T h u s , -rms in (66) retract their
H i g h in the genitive plural, w h i l e i ' se in (67) fail to d o s o .
; 2 6

(66) a. jelen (nom.sg.), jelen-a ( or?.sg.) versus jeleet,

r .j (gen.pl.) ' d e e r '

H H H
b. malin-a (nom.sg.) versus maliin-a (gen.pl.) ' r a s p b e r r y '
I I
H H
c. kolen-o (nom.sg.) versus koleen-a (gen.pl.) ' k n e e '

H H
d. republi-a (nom.sg.) versus republiik-a (gen.pl.) ' r e p u b l i c '

H H
e. kategorij-a (nom.sg.) versus kategoriij-a (gen.pl.) ' c a t e g o r y '

H H

(67)
a. naarod (nom.sg.), naarod-a (gen.sg.) versus naarood-a (gen.pl.) ' p e o p l e '
I I I
H H H
Rule Domains and Phonological Change 393

b. saanjalo ( n o m . s g . ) , saanjal-a (gen.sg.) versus saanjaal-a (gen.pl.) ' d r e a m e r '

H H H
c. ucioonic-a ( n o m . s g . ) versus uciooniic-a (gen.pl.) ' c l a s s r o o m '

H H
d. organizaacij-a ( n o m . s g . ) versus organizaaciij-a (gen.pl.) 'organization'

H H

W e could c a p t u r e the difference b e t w e e n the stems in ( 6 6 ) and t h o s e in ( 6 7 ) by

i m p o s i n g an additional condition on the rule of G e n . P l . D e l i n k i n g : in the N e o -
Stokavian this rule applies only in the e n v i r o n m e n t of a light syllable. T h i s c o n -
dition, however, is highly suggestive. It distinguishes b e t w e e n f o r m s that are s u b -
j e c t to the cyclic version of S p r e a d i n g a n d those that w o u l d fail to u n d e r g o this
rule in the cyclic c o m p o n e n t . W e can in fact a c c o u n t for the difference b e t w e e n
the t w o sets of forms s i m p l y by including S p r e a d i n g in the set of cyclic rules.
C o n s i d e r the following p o s s i b l e scenario for d e r i v i n g the N e o - S t o k a v i a n g e n i -
tive plural forms. First, T o n e L i n k i n g links the H i g h tone to the stem-final syllable;
next, the linked t o n e s p r e a d s to the p r e c e d i n g syllable; and finally, a d o u b l y linked
H i g h t o n e loses its s e c o n d link in the e n v i r o n m e n t of the genitive plural suffix, b y
virtue of the following rule.

(68) GENPLDELINKING ( N S ) : u, JA] -tfGenPl

U n d e r this scenario, the N e o - S t o k a v i a n version of the genitive delinking will

not be able to apply to the forms in ( 6 3 ) and ( 6 7 ) , since these forms d o not u n d e r g o
S p r e a d i n g . T h i s s u c c e s s i o n of m o v e s is p r e s e n t e d formally in ( 6 9 ) .
2 7

(69) CYCLE 1: Tone Linking jelen naarod

I I
|
H
Spreading jelen

N
H
CYCLE 2 : jelen } a naarod 1 a

N
H
GenPl Delinking(NS)
jelen ] a

OUTPUT: H naarooda

jeleena

H
394 Draga Zee

In s u m , the O l d S t o k a v i a n version of G e n . P l . D e l i n k i n g w a s r e a n a l y z e d in the

N e o - S t o k a v i a n dialects d u e to the i m p a c t of S p r e a d i n g . O n c e S p r e a d i n g entered
the cyclic c o m p o n e n t of the g r a m m a r , it w a s m o s t likely r e o r d e r e d with respect to
G e n . P l . D e l i n k i n g . A n d o n c e this h a p p e n e d , t w o o u t c o m e s w e r e possible, b y vir-
tue of an effect d u e to inalterability ( H a y e s , 1986): either G e n . P l . D e l i n k i n g could
h a v e applied only to f o r m s that h a d not u n d e r g o n e S p r e a d i n g , or the rule itself
could h a v e m u t a t e d so as to apply only to those forms that h a d u n d e r g o n e Spread-
ing. It is this latter alternative that w a s realized in the N e o - S t o k a v i a n dialects.

5. S P R E A D I N G A N D T H E S T R O N G D O M A I N H Y P O T H E S I S

In the p r e v i o u s t w o sections I h a v e traced the rule of S p r e a d i n g t h r o u g h the

various c o m p o n e n t s of the g r a m m a r in the N e o - S t o k a v i a n dialects. In Section 4 it
w a s a r g u e d that, in all N e o - S t o k a v i a n dialects, S p r e a d i n g applies cyclically. S e c -
tion 3 s h o w e d that, a l t h o u g h all N e o - S t o k a v i a n dialects h a v e S p r e a d i n g in their
postcyclic c o m p o n e n t , in N S 1 this rule applies only postlexically, in N S 2 it a p -
plies only lexically, a n d in N S 3 it applies b o t h lexically a n d postlexically. (70)
s u m m a r i z e s the application of S p r e a d i n g in the three N e o - S t o k a v i a n dialects w e
have examined.

(70)
NS1 NS2 NS3

Cyclic Yes Yes Yes

Lexical P W (phonological word) ? Yes Yes
Postlexical P W (clitic g r o u p ) Yes No Yes
Phonological phrase Yes No No

N S 1 requires a further c o m m e n t . Recall that there are n o c o m p e l l i n g r e a s o n s to

a s s u m e that in this dialect S p r e a d i n g operates within the p h o n o l o g i c a l w o r d .
M o r e o v e r , in this s a m e dialect Initial H i g h Insertion clearly fails to apply in this
domain.

(71)
NS1 NS2 NS3

Lexical P W ( p h o n o l o g i c a l w o r d ) No Yes Yes

Postlexical P W (clitic g r o u p ) Yes No Yes

This strongly s u g g e s t s that, in N S 1 , both S p r e a d i n g and Initial H i g h Insertion

in s o m e sense ignore the w o r d - s i z e d d o m a i n . O n e possibility m i g h t b e that b o t h
S p r e a d i n g and Initial H i g h Insertion in fact skip over the w o r d - s i z e d d o m a i n , thus
violating the strong d o m a i n h y p o t h e s i s . B u t is it really likely that e a c h of the rules
Rule Domains and Phonological Change 395

i n d e p e n d e n t l y c h o o s e s its d o m a i n , and that they j u s t h a p p e n to g o h a n d in h a n d ?

T h i s p u z z l e is resolved u n d e r the view of rule d o m a i n s p r e s e n t e d in the intro-
d u c t o r y part of the article. In particular, g r a m m a r s m a y vary in w h e t h e r the p h o -
nological w o r d is created lexically or postlexically. T h i s is precisely the difference
b e t w e e n N S 2 a n d N S 3 on the o n e hand, and N S 1 on the other. In the first case,
the p h o n o l o g i c a l w o r d is created lexically, w h i c h yields (72) as the set of available
rule d o m a i n s .

(72) Lexical Level 1

Cyclic
Level 2
Level 3 p h o n o l o g i c a l w o r d (word-si zed)
Postcyclic
Postlexical Level 3 p h o n o l o g i c a l w o r d (clitic g r o u p )
Level 4 phonological phrase

In N S 1 , o n t h e o t h e r h a n d , the p h o n o l o g i c a l w o r d is c r e a t e d postlexically, w h i c h
results in the set of d o m a i n s listed in (73).

(73) Lexical Level 1 ^

Level 2 C y C , 1 C

Postlexical Level 3 p h o n o l o g i c a l w o r d (clitic g r o u p )

Level 4 phonological phrase

T h e situations d e p i c t e d in (72) a n d (73) are sufficient to a c c o u n t for the d o m a i n s

of S p r e a d i n g in the three dialects. T h u s , in N S 1 , in w h i c h the p h o n o l o g i c a l w o r d
is created postlexically, S p r e a d i n g applies as in (74).

(74) Lexical cyclic levels yes

Postlexical phonological word yes
phonological phrase yes

In other w o r d s , N S 1 simply lacks the w o r d - s i z e d d o m a i n , w h i c h e x p l a i n s w h y

both S p r e a d i n g a n d Initial H i g h Insertion b e h a v e in the s a m e fashion in this dia-
lect, failing to apply in this d o m a i n .
T h e set of d o m a i n s available in N S 2 and N S 3 is exactly the s a m e . B u t w h i l e in
N S 3 S p r e a d i n g applies in b o t h the lexical and the postlexical c o m p o n e n t s (just as
in N S 1 ) , in N S 2 it only applies lexically, as s h o w n in ( 7 5 ) . 2 8

(75) NS2 NS3

Lexical cyclic levels yes yes
phonological word yes yes

Postlexical phonological word no yes

phonological phrase no yes

I a m n o w in a position to explain w h y S p r e a d i n g and Initial H i g h Insertion g o

h a n d in h a n d in e a c h of the dialects. T h e t w o rules o v e r l a p in t h e d o m a i n of the
396 Draga Zee

p h o n o l o g i c a l word, and this d o m a i n varies from o n e dialect to the next. In N S 1 ,

Initial H i g h Insertion c a n n o t apply on the p h o n o l o g i c a l w o r d in its lexical size
b e c a u s e this d o m a i n is f o r m e d only postlexically. In N S 2 and N S 3 , however,
w h e r e the p h o n o l o g i c a l w o r d is available in its smaller size, Initial H i g h Insertion
d o e s apply lexically.
In s o m e versions of the p r o s o d i c p h o n o l o g y (Hayes, 1989a; N e s p o r and Vogel,
1986), the set of p r o s o d i c d o m a i n s i n c l u d e s the clitic g r o u p in addition to the
p h o n o l o g i c a l w o r d and the p h o n o l o g i c a l p h r a s e . T h e clitic g r o u p thus c o r r e s p o n d s
to w h a t w e call h e r e the p h o n o l o g i c a l w o r d in its postlexical g u i s e ; and the p h o -
nological w o r d is treated as a strictly word-sized d o m a i n . U n d e r this set of as-
s u m p t i o n s , however, the fact that the rules in N S 1 ignore the w o r d - s i z e d d o m a i n
w o u l d raise the question of h o w universal are p r o s o d i c constituents such as the
p h o n o l o g i c a l w o r d or the clitic g r o u p . In particular, N S 1 w o u l d n e e d to b e char-
acterized as lacking the p h o n o l o g i c a l w o r d d o m a i n , in w h i c h c a s e the p h o n o l o g i -
cal w o r d loses its universal status. T h e only w a y to salvage the universal status of
the p h o n o l o g i c a l w o r d w o u l d b e to a s s u m e its p r e s e n c e in N S 1 and then treat as
accidental the fact that b o t h S p r e a d i n g a n d Initial H i g h Insertion skip over this
d o m a i n , in violation of the strong d o m a i n h y p o t h e s i s . T h i s situation in w h i c h w e
h a v e to a b a n d o n either the strong d o m a i n h y p o t h e s i s or the universality of the
p h o n o l o g i c a l w o r d will persist as l o n g as w e k e e p the clitic g r o u p in the inventory
of p r o s o d i c constituents or of rule d o m a i n s .
B u t o n c e the p r o s o d i c c o n s t i t u e n c y is divided into its lexical and postlexical
subparts, it is possible to d i s p e n s e with the clitic g r o u p as a p r o s o d i c level w h i l e
retaining the rule d o m a i n of the clitic g r o u p size, as a r g u e d h e r e (and in Z e e ,
1988). T h e clitic g r o u p is in fact the p h o n o l o g i c a l w o r d in its postlexical g u i s e .
T h i s view of the clitic g r o u p is b a s e d o n Inkelas's (1989) theory of p r o s o d i c c o n -
stituency. B y associating clitics with p r o s o d i c subcategorization frames, this
f r a m e w o r k p r o v i d e s theoretical foundation for the intuition that clitics are indeed
" l e a n e r s . " S e r b o - C r o a t i a n clitics s u b c a t e g o r i z e for the p h o n o l o g i c a l w o r d , w h i c h
e x p l a i n s w h y this d o m a i n automatically includes clitics in the postlexical c o m p o -
nent. M o r e o v e r , w e h a v e an a c c o u n t for the a s y m m e t r y in the p o s s i b l e sets of rule
d o m a i n s . T h e f r a m e w o r k set u p h e r e predicts t w o types of c a s e s : the p h o n o l o g i c a l
w o r d is either manifested only in its larger size, or in b o t h its smaller and its larger
size. W e thus e x c l u d e the c a s e of the p h o n o l o g i c a l w o r d b e i n g manifested only in
its smaller size (as long as a l a n g u a g e p o s s e s s e s clitics that s u b c a t e g o r i z e for the
phonological word).
N o t e that, in N S 3 , S p r e a d i n g applies within the p h o n o l o g i c a l w o r d both lexi-
cally a n d postlexically. K a n e r v a ( 1 9 8 9 ) points to a similar c a s e of cyclicity in
C h i c h e w a , arguing that cyclicity of this sort follows from the specific properties
of the p h o n o l o g i c a l w o r d ; this is the only p r o s o d i c d o m a i n w h i c h is present in
m o r e than o n e c o m p o n e n t of the g r a m m a r .
Rule Domains and Phonological Change 397

6. T H E R O U T E O F C H A N G E

In the i n t r o d u c t o r y part of the article I p r o p o s e d that a new rule is a d d e d to the

highest level, w h i c h c o r r e s p o n d s to the largest d o m a i n of the g r a m m a r , and that it
then enters all the accessible lower levels (which h o u s e successively smaller d o -
m a i n s ) b y virtue of the strong d o m a i n h y p o t h e s i s . However, Kiparsky (1984) also
p r o p o s e s that a rule m a y b e t u r n e d off at a certain level of the g r a m m a r . I interpret
this as a w i t h d r a w a l of a rule from o n e or m o r e of its higher, that is, larger, d o -
m a i n s ; this will again b e v i e w e d as a type of phonological c h a n g e .
W i t h these b a c k g r o u n d a s s u m p t i o n s , we are n o w able to Follow u p t h e p h o n o -
logical c h a n g e s associated with accent retraction, or rather, with the addition of
S p r e a d i n g to the set of p h o n o l o g i c a l rules. Each of the ihree iNeo-Stokavian dia-
lects e x a m i n e d h e r e is v i e w e d as a stage in this process of c h a n g e . N S 1 stands for
the earliest stage, in w h i c h S p r e a d i n g simply p e r v a d e s all the c o m p o n e n t s of the
g r a m m a r . In N S 3 this rule has w i t h d r a w n from the higher levels of the g r a m m a r ,
2 9

c h o o s i n g the p h o n o l o g i c a l word as its highest level. A n d in N S 2 , the rule h a s

w i t h d r a w n entirely from the postlexical c o m p o n e n t .

(76) STAGES OF THE p h o n o l o g i c a < MANGE:

NS1 NS3 NS2

Lexical cyclic levels / J y
(phonological word) / y

Postlexical p h o n o l o g i c a l word / /
phonciugical phrase /

W e thus see that the lexical/postlexical b o u n d a r y is a relevant p o i n t at w h i c h

rules m a y be turned off. In other w o r d s , if a rule is lexicalized (Kiparsky, 1984,
1988), its crucial reference point will b e c o m e the lexical/postlexical b o u n d a r y ,
even if this b o u n d a r y h a p p e n s to cut across o n e of the p r o s o d i c constituents, as is
c l a i m e d here to be the case in S e r b o - C r o a t i a n .
A n o t h e r d i m e n s i o n a l o n g w h i c h the dialects differ is w h e t h e r the p h o n o l o g i c a l
v «jid is created lexically o r postlexically. W h i l e this difference c a n b e interpreted
^ another p h o n o l o g i c a l c h a n g e , I treat it as unrelated to S p r e a d i n g . T h i s view is
supported by the fact that the d o m a i n s of Initial H i g h Insertion, a rule w h i c h is
historically unrelated t o S p r e a d i n g , are also sensitive to t h e p r e s e n c e versus a b -
sence of the p h o n o l o g i c a l w o r d in the lexical c o m p o n e n t . Moreover, Initial H i g h
Insertion exhibits t h e s a m e type of variation in the O l d Stokavian dialects w h i c h
entirely lack S p r e a d i n g .
In order to c o m p l e t e the picture, I b r i n g in at this point the application of
S p r e a d i n g in the O l d S t o k a v i a n dialects. Several regional varieties p o s s e s s a rule
w h o s e effect is close to the N e o - S t o k a v i a n version of S p r e a d i n g , with o n e i m p o r -
398 Draga Zee

tant difference: in this case, S p r e a d i n g operates as a d o m a i n limit rule, affecting

only the H i g h tones linked to the r i g h t m o s t m o r a of the d o m a i n . T h e O l d Stoka-
vian version of S p r e a d i n g can b e formulated as in (77).

(77)

I illustrate briefly h o w S p r e a d i n g applies in the O l d Stokavian dialect d e s c r i b e d

in C u p i c ( 1 9 7 7 ) . T h e d o m a i n c h o s e n by S p r e a d i n g in this dialect is the p h o n o -
logical w o r d . T h e operation of this rule is exemplified b y the following p a r a d i g m .

(78) naarod (nom.sg.) naarod-a(gcn.sg.) naarod-ima (dat.pl.) 'people'

N I I
H H H

T h e rule applies only in the n o m i n a t i v e singular form, w h o s e H i g h tone is

linked to the final m o r a of the d o m a i n . It fails to apply in the genitive singular and
dative plural f o r m s , in w h i c h the H i g h is linked to a nonflnal m o r a .
F u r t h e r m o r e , since this dialect p o s s e s s e s only the p h o n o l o g i c a l w o r d in its post-
lexical size, w e also e n c o u n t e r alternations as in (79) and (80), in w h i c h the
p r e s e n c e of an enclitic b l o c k s the application of S p r e a d i n g (from C u p i c , 1 9 7 7 :
205, 210).

(79) a. [ dobro mi J [ dosli ] 'welcome to me'

I N
H H
well me(dat) come
b . svi da ste mi [ dobro J [ dosli ] 'welcome all of you'

I \ lN
H H H
all Comp Aux me(dat) well come
(80) a. [biiloje]. 'there was .

H
was Aux
b. [pa je biilo ] . . . 'and there was . . . '

N
H
and Aux was

Finally, t o n e will spread o n t o a proclitic only if its host is a m o n o s y l l a b i c word,

w h i c h again follows from the formulation of S p r e a d i n g in (77).
Rule Domains and Phonological Change 399

(81) a. u smrt 'in death'

N
H
b . pred rat 'before (the) war'

In this O l d S t o k a v i a n dialect, then, S p r e a d i n g selects the d o m a i n of the p h o n o -

logical w o r d . W e m a y a s s u m e that it applies at levels that h o u s e d o m a i n s larger
than the p h o n o l o g i c a l w o r d ; this, however, c a n n o t b e d e m o n s t r a t e d b e c a u s e the
e d g e of any of the larger d o m a i n s will c o i n c i d e with the e d g e of the p h o n o l o g i c a l
w o r d . H o w e v e r , rule (77) d o e s not apply in any of the d o m a i n s smaller than the
p h o n o l o g i c a l w o r d ; as w e saw earlier, the O l d S t o k a v i a n g r o u p of dialects lacks
S p r e a d i n g in the cyclic c o m p o n e n t . T h i s is clearly a d e p a r t u r e from the strong
d o m a i n h y p o t h e s i s . M o r e o v e r , S p r e a d i n g is restricted to the p o s t c y c l i c d o m a i n s
in j u s t t h o s e dialects in w h i c h it o p e r a t e s as a d o m a i n limit rule. M y h y p o t h e s i s is
that this d e p a r t u r e from the strong d o m a i n h y p o t h e s i s m a y well b e related to the
p r o p e r t i e s of d o m a i n limit rules in g e n e r a l ; it c o u l d b e that selecting o n e of the
e d g e s of a rule d o m a i n is p e r m i s s i b l e only in the postcyclic c o m p o n e n t . In s u p p o r t
of this, I b r i n g u p t w o other S e r b o - C r o a t i a n rules of the d o m a i n limit t y p e w h i c h
violate the strong d o m a i n h y p o t h e s i s . First, Initial H i g h Insertion, w h i c h m a k e s
reference to a d o m a i n e d g e , fails to apply at any levels w i t h d o m a i n s s m a l l e r than
the p h o n o l o g i c a l w o r d ; a n d second, a rule p r o p o s e d in I n k e l a s a n d Z e e ( 1 9 8 8 ) that
links a L o w t o n e at the right e d g e of o n e of the larger d o m a i n s , m o s t likely the
intonational p h r a s e , clearly fails to apply within any of the l o w e r l e v e l s . W h i l e 30

I h a v e n o e x p l a n a t i o n of w h y d o m a i n limit rules b e h a v e in this p e c u l i a r fashion, it

will at least b e p o s s i b l e to n a r r o w d o w n the class of rules w h i c h d e p a r t from the
strong d o m a i n h y p o t h e s i s to those of the d o m a i n limit t y p e . 3 1

B u t h o w are the O l d a n d the N e o - S t o k a v i a n versions of S p r e a d i n g related? T h a t

is, w h i c h of the t w o versions characterizes an older stage and w h i c h is an innova-
tion? W h a t s e e m s to b e a likely s c e n a r i o is that S p r e a d i n g e n t e r e d the g r a m m a r in
its O l d S t o k a v i a n form, that is, as a d o m a i n limit rule. T h e crucial c h a n g e that
o c c u r r e d in N e o - S t o k a v i a n is that S p r e a d i n g lost its d o m a i n limit specification;
and only then w a s it able to c o n f o r m to the strong d o m a i n h y p o t h e s i s , b y b e c o m -
ing o m n i p r e s e n t in the g r a m m a r . U n d e r the reversed scenario, S p r e a d i n g w o u l d
have to a b a n d o n the lexicon in its transition from the N e o - S t o k a v i a n to the O l d
S t o k a v i a n version. H a d this h a p p e n e d , w e w o u l d h a v e e x p e c t e d to find traces of
the cyclic version of S p r e a d i n g in at least s o m e of the O l d S t o k a v i a n dialects;
however, I a m not a w a r e of any e v i d e n c e that w o u l d s u p p o r t the latter alternative.
T h e principles of the L P M m o d e l h a v e thus e n a b l e d us to interpret the c r o s s -
dialectal variation a s s o c i a t e d with a c c e n t retraction as a succession of s y n c h r o n i c
400 Draga Zee

stages w h i c h characterize a p h o n o l o g i c a l c h a n g e . T h e Old Stokavian g r o u p of

dialects records an earlier stage than the N e o - S t o k a v i a n g r o u p . W i t h i n the O l d
Stokavian, the innovative dialects are those that possess S p r e a d i n g as a d o m a i n
limit rule; and in the N e o - S t o k a v i a n dialects, S p r e a d i n g lost its d o m a i n limit p r o p -
erties, w h i c h resulted in first its p e r v a s i o n of and then withdrawal from the various
subparts of the g r a m m a r . T h e set of stages posited here follow from the strong
d o m a i n h y p o t h e s i s and from the a s s u m p t i o n that rules are a d d e d at the e n d of the
postlexical c o m p o n e n t . Crucial for o u r c l a i m s , however, is a specific organization
of the L P M m o d e l into levels w h i c h serve as rule d o m a i n s .

ACKNOWLEDGMENTS

I am grateful to Sharon Inkelas, Ellen Kaisse, Paul Kiparsky, and an anonymous referee
for most helpful comments on earlier versions of this article.

NOTES

1
Serbo-Croatian is divided into three major dialects, the Stokavian, the Cakavian, and
the Kajkavian. I cover the accentual situation in the Stokavian dialects, with the exception
of those characterized by only one accent. The subdivisions within the Stokavian group of
dialects adopted here are those proposed in Ivic (1958, 1985).
2
1 assume that rule domains are provided by morphological constituents at the cyclic
levels, and by prosodic constituents at the postcyclic levels.
3
Here I follow Inkelas (1989), where clitics are analyzed as prosodically dependent
elements which subcategorize for prosodic domains. Serbo-Croatian clitics, both proclitics
and enclitics, subcategorize for the domain of the phonological word.
4
T h i s proposal is made in Zee (1988), to account for the applications of several post-
cyclic rules in Bulgarian. Kanerva (1989) adopts this perspective in his analysis of clitics
in Chichewa. Section 5 of the present article provides further arguments for this view of
the phonological word. This position is originally due to Selkirk (1986), who assumes that
clitics are included into the phonological word.
5
According to King (1973), a new rule is necessarily added at the end of a grammar,
that is, as the latest in the set of ordered rules (for other views in the debate on rule addition
see Halle, 1962; King, 1969; Kiparsky, 1965; as well as Kaisse, this volume). However,
while King focuses on the ordering of a newly added rule, my focus is on the new rule's
domain.
6
T h e situation in the Old Stokavian is somewhat more complex. While the more con-
servative Old Stokavian dialects are indeed characterized only by the falling accents (as in
the regional variety described in Stevanovic, 1940), the less conservative Old Stokavian
dialects may also possess the rising accents. According to Ivic (1985), the crucial difference
Rule Domains and Phonological Change 401

between the Old Stokavian and the Neo-Stokavian dialects is in the distribution of the pitch
accents, regardless of the number of accents they may possess. The less conservative Old
Stokavian dialects are discussed in Section 6.
7
Vowel length is marked throughout the article as gemination. This is the only departure
from the standard orthography.
8
Syllable weight in Serbo-Croatian is determined by vowel length: a syllable with a
short vowel is light, and a syllable containing a long vowel is heavy. Consonants do not
contribute to syllable weight. Furthermore, Serbo-Croatian does not have diphthongs; two
unlike adjacent vowels belong to different syllables and are separated by a hiatus (Lehiste
and Ivic, 1967).
9
T h a t the prefix ne- is not associated with a High tone if it escapes the effect of Spread-
ing is shown by the pair of forms pusaac 'smoker' and nepusaac 'nonsmoker', both having
lexical tone on their final syllables; the lexical High then spreads to the preceding syllable,
which in pusaac is also the initial syllable. The prefix ne- in nepusaac remains toneless.
I n Inkelas and Zee (1988), lexical tones are prelinked in the underlying form. In Zee
1 0

(1988), lexical tone is assigned by a cyclic tonal rule. This article adopts the latter analysis,
which is presented in Section 4.
1 1
In order to make sure that rule (17) applies only to toneless forms, we may invoke the
OCP in its passive version (McCarthy, 1986), which will prevent the application of (17) to
a form already linked to a High tone. Generally, Serbo-Croatian words may be associated
with at most one High tone.
N S 1 is represented by what is known as the Vukovian norm, described in Danicic
1 2

(1925); this norm is based on the east Herzegovian dialect (see Vukovic, 1940, and Ruzicic,
1927, for descriptions of two of its regional varieties). NS2 corresponds to the eastern
standard, spoken predominantly in Belgrade (Miletic, 1952). NS3 is spoken in the regions
of Srem and Macva (see the descriptions of regional varieties in Moskovljevic 1927-28
and Nikolic 1953-54, 1966, 1970).
1 3
1 have found no convincing evidence that Spreading applies within the phonological
phrase in this dialect. Nikolic (1970:51) lists a single case of this type, sto hektaara '(one)
hundred nectars' in which the High on the first syllable of hektara spreads onto sto.
1 4
I f Spreading applies lexically, it will not reapply in the postlexical component. The
failure of Spreading to apply more than once can be attributed to inalterability (Hayes,
1986): given the statement of the rule in (9), only a singly linked High will undergo
Spreading.
1 5
Note that in NS3 Initial High Insertion is optional lexically, but obligatory postlexi-
cally. In other words, a phonological word has to be assigned tone within the phonological
word domain. It was suggested by an anonymous referee that Initial High Insertion is a
metrical rather than a tonal rule (see Halle, 1971) since the High assigned by this rule
coincides with stress, and that it is metrical structure rather than tone that is obligatorily
assigned within the phonological word. However, the High assigned by Initial High Inser-
tion may undergo Spreading, as shown in Section 3.3, and in this case it will not coincide
with stress. In fact, a metrical analysis would have to posit a stress shift in this case that
would replicate Spreading. This demonstrates that a tonal analysis is to be preferred over a
metrical one.
1 have found no cases of the application of Initial High Insertion within the domain of
1 6

the phonological phrase. This is to be expected, I believe, since Initial High Insertion ap-
402 Draga Zee

plies only within toneless domains (see note 11), and a phonological phrase can never be
toneless by virtue of the fact that it has to contain at least one content word, which will
have received tone by this point.
1 7
Here I assume the moraic theory of syllable structure with a direct representation of
moras, as in Hyman (1984), Hayes (1989b), Ito (1989), and Zee (1988), among others. I
also adopt the labeling convention of marking the leftmost mora as s(trong), and the re-
maining moras as w(eak); the ^-labeled mora is interpreted here as the head of the syllable
it belongs to (for details see Zee, 1988).
1 8
The forms in (44) and (45), as well as all other forms in this and the following sub-
sections, are shared by the two dialect groups. In other words, we ignore the effect of
Spreading.
1 9
A formal version of the rule is provided in Section 4.2.
2 0
T h e situation is actually somewhat more complex. Just like stems, affixes may either
be toneless or have tone of their own. The suffix -ov is obviously toneless. If a toneless
stem combines with an affix with its own tone, the derived form will be associated with
tone. For example, the derived form oblacic 'cloud (dim.)' has a High on its final syllable;
this High comes from the suffix -ic, which is affiliated with tone in its underlying form.
When a toneless simplex stem is combined with a toneless suffix, the derived form remains
toneless.
2 1
It is worth noting that, at the earlier stages of the language, before the fall of yers, the
nominative ending had exactly this form.
2 2
Metatony is triggered by any process that delinks a High tone. A High may be de-
linked either by rule (see the case discussed in Section 4.3), or by virtue of vowel loss. This
latter case could be motivated as follows: once the vowel is delinked from its mora, the
mora in its turn delinks from the higher structure, which then leads to the delinking of tone.
Tone delinking caused by yer deletion is only one such case; other cases are found in the
verbal paradigm where a vowel other than yer is deleted when followed by another vowel
(Zee, 1988).
2 3
Zee (1988) posits two cyclic, that is, lexical levels, and associates Tone Linking with
Level 2. The failure of this rule to apply at Level 1 is attributed to the prosodic setup of this
level. Several phonological phenomena receive a unified account under the assumption that
Level 1 includes only moras while Level 2 includes both moras and syllables. Since con-
straint (59) is operative throughout the cyclic component, it will preclude the linking of
tone to a mora that does not bear the s marking, and such moras are not available before
Level 2. Absence of relevant structure at some point in the grammar is the standard type of
interference with the strict domain hypothesis (cf. Kiparsky, 1984).
2 4
T h i s type of accent shift is found only in the Stokavian dialects (Leskien, 1914:
2 2 1 - 2 2 2 ) . The Cakavian dialects, which in many ways preserve an older accentual situ-
ation than that in the Stokavian, lack entirely the accent shift in the genitive plural form.
Moreover, only the genitive plural forms taking the -a desinence are subject to this type of
accent shift; no accent shift is exhibited by forms taking the -i desinence.
2 5
This ending has another peculiar property: it lengthens the vowel of the immediately
preceding syllable. The lengthening process, however, appears not to be responsible for the
genitive accent, as the discussion below will make evident. This lengthening process is
ignored in the derivations given below.
Rule Domains and Phonological Change 403

1 focus here only on those forms which do not contain yer vowels in the stem. Forms
2 6

with yers, like komaarac (nom.sg.), komaarca (gen.sg.), komaraaca (gen.pl.) 'mosquito'
exhibit a shift in the genitive plural which I believe is of a different sort. In this class of
forms, tone is linked to the desinence in all case forms other than the genitive plural. Thus
tone is delinked here from the genitive plural desinence rather than from the stem-final
syllable. While I do not have an account for this set of forms, I will assume that the pattern
they follow is distinct from that discussed here.
2 7
We will have to assume that yer delinking operates before Spreading, in order to
account for the fact that this process is not sensitive to syllable weight.
2 8
Evidence for the postlexical version of the phonological word in NS2 comes from
postlexical applications of Initial High Insertion. One class of forms, indefinite pronouns,
is lexically marked as exceptional to Initial High Insertion and can thus undergo this rule
only postlexically; as a result, the High is assigned to a proclitic. Thus the negated prepo-
sitional phrase ni sa kim 'not with anyone' has a High tone on the negative particle ni,
which is a proclitic, rather than on the pronominal stem kim. Note that ni gets no tone in ni
sa Petrom, since Petrom possesses a High tone on its initial syllable.
1 have no evidence for the application of this rule within any of the levels larger than
2 9

the phonological phrase; however, it is reasonable to assume that the rule has bled itself, so
to speak, beyond the phonological phrase level.
3 0
This situation is not restricted to Serbo-Croatian. Hyman (1990) discusses several
tonal rules of the domain limit type which apply postlexically. Two such rules in Kinande,
for example, operate on the domains of the phonological and the intonational phrase, but
not on any of the smaller domains.
3 1
Kaisse (this volume) argues that a relatively new rule may be applying within larger
(most likely, postlexical) domains while still failing to descend to the lexical domains; this
most likely captures the progression of Spreading in the Neo-Stokavian dialects. However,
while I agree with Kaisse that a relatively new rule may be applying in the larger but not in
the smaller domains, I also believe that this kind of situation may become stable only in
the case of domain limit rules. Initial High Insertion, for example, is a very old rule with
deep Indo-European roots. As for the Old Stokavian version of Spreading, it is not clear
whether it dates as far back as its Neo-Stokavian counterpart; if it does, then its failure to
descend to smaller domains is due to its domain limit properties.

REFERENCES

Belie, A. (1956). Osnovi istorije srpskohrvatskog jezika, vol. 1, Fonetika. Naucna knjiga,
Beograd.
Booij, G., and Rubach, J. (1987). Postcyclic versus postlexical rules in lexical phonology.
Linguistic Inquiry 18, 1 - 4 4 .
Browne, W. E., and McCawley, J. (1965). Srpskohrvatski akcenat. Zbornik za filologiju i
lingvistiku 8, 1 4 7 - 1 5 1 .
Cupic, D. (1977). Govor Bjelopavlica. Srpski dijalektoloski zbornik 23, 1-226.
404 Draga Zee

Danicic, Dj. (1925). Srpski akcenti. Srpska Kraljevska Akademija, Beograd-Zemun.

Halle, M. (1962). Phonology in generative grammar. Word 18, 5 4 - 7 2 . Reprinted in The
Structure of Language: Readings in the Philosophy of Language. (J. A. Fodor and
J. Katz, eds., 1964), pp. 3 3 4 - 3 5 2 . Prentice-Hall, Englewood Cliffs, N.J.
Halle, M. (1971). Remarks on Slavic accentology. Linguistic Inquiry 2, 1-19.
Halle, M., and Kiparsky, P. (1981). Review of Histoire de Vaccentuation slave by Paul
Garde. Language 57, 150 - 1 8 1 .
Hayes, B. (1986). Inalterability in CV phonology. Language 62, 3 2 1 - 3 5 1 .
Hayes, B. (1989a). The prosodic hierarchy in meter. In Rhythm and Meter (P. Kiparsky and
G. Youmans, eds.), pp. 2 0 1 - 2 6 0 . Academic Press, San Diego.
Hayes, B. (1989b). Compensatory lengthening in moraic phonology. Linguistic Inquiry20,
253-306.
Hyman, L. (1984). A Theory of Phonological Weight. Foris, Dordrecht.
Hyman, L. (1990). Boundary tonology and the prosodic hierarchy. In The Phonology-
Syntax Connection (S. Inkelas and D. Zee, eds.), pp. 1 0 9 - 1 2 5 . CSLI Publications and
University of Chicago Press, Chicago.
Inkelas, S. (1989). Prosodic Constituency in the Lexicon. Doctoral dissertation, Stanford
University, Stanford, Calif.
Inkelas, S., and Zee, D. (1988). Serbo-Croatian pitch accent: The interactions of tone,
stress, and intonation. Language 64, 2 2 7 - 2 4 8 .
Ito, J. (1989). A prosodic theory of epenthesis. Natural Language and Linguistic Theory 7,
217-259.
Ivic, P. (1958). Die serbokroatischen Dialekte: Ihre Struktur und Entwicklung, vol. 1, All-
gemeines und die stokavischen Dialektgruppe. Mouton, The Hague.
Ivic, P. (1985). Dijalektologija srpskohrvatskog jezika: Uvod i stokavsko narecje. Matica
srpska, Novi Sad.
Kanerva, J. (1989). Focus and Phrasing in Chichewa Phonology. Doctoral dissertation,
Stanford University, Stanford, Calif.
Kenstowicz, M. (1974). Inflectional accent of the Serbo-Croatian noun. Studies in the Lin-
guistic Sciences 4, 8 0 - 1 0 6 .
King, R. D. (1969). Historical Linguistics and Generative Grammar. Prentice-Hall, Engle-
wood Cliffs, N.J.
King, R. D. (1973). Rule insertion. Language 49, 5 5 1 - 5 7 8 .
Kiparsky, P. (1965). Phonological Change. Doctoral dissertation, Massachusetts Institute
of Technology, Cambridge.
Kiparsky, P. (1973). The inflectional accent of Indo-European. Language 49, 7 9 4 - 8 4 9 .
Kiparsky, P. (1984). On the lexical phonology of Icelandic. In Nordic Prosody III (C. Elert,
I. Johansson, and E. Strangert, eds.), pp. 135-164. University of Umea, Umea,
Sweden.
Kiparsky, P. (1985). Some consequences of lexical phonology. In Phonology Yearbook 2,
85-138.
Kiparsky, P. (1988). Phonological change. In Linguistics: The Cambridge Survey, vol. 1,
Linguistic Theory: Foundations (F. J. Newmeyer, ed.), pp. 3 6 3 - 4 1 5 . Cambridge Uni-
versity Press, Cambridge.
Kiparsky, P., and Halle, M. (1977). Towards a reconstruction of the Indo-European accent.
Rule Domains and Phonological Change 405

In Studies in Stress and Accent (Southern California Occasional Papers in Linguistics

4) (L. Hyman, ed.), pp. 2 0 9 - 2 3 8 . University of Southern California, Los Angeles.
Lehiste, I., and Ivic, P. (1967). Some problems concerning the syllable in Serbocroatian.
Glossa 1/2, 126-136.
Lehiste, I., and Ivic, P. (1986). Word and Sentence Prosody in Serbocroatian. MIT Press,
Cambridge, Mass.
Leskien, A. (1914). Grammatik der Serbo-Kroatischen Sprache. Winter, Heidelberg.
McCarthy, J. (1986). OCP effects: Germination and antigemination. Linguistic Inquiry 17,
207-263.
Miletic, B. (1952). Osnovi fonetike srpskog jezika. Znanje, Beograd.
Moskovljevic, M. (1927-28). Akcenti pocerskog govora. Juznoslovenskifilolog 7, 5 - 6 8 .
Nespor, M., and Vogel, I. (1986). Prosodic Phonology. Foris, Dordrecht.
Nikolic, B. (1953-54). O govoru Srema. Juznoslovenski filolog 20, 2 7 3 - 2 8 7 .
Nikolic, B. (1966). Macvanski govori. Srpski dijalektoloski zbornik 16, 1 7 9 - 3 1 3 .
Nikolic, B. (1970). Osnovi mladje novostokavske akcentuacije. Institut za srpskohrvatski
jezik, Beograd.
Poser, W. J. (1984). The Phonetics and Phonology of Tone and Intonation in Japanese.
Doctoral dissertation, Massachusetts Institute of Technology, Cambridge.
Resetar, M. (1900). Die serbocroatische Betonung siidwestlicher Mundarten. Vienna.
Ruzicic, G. (1927). Akcenatski sistem pljevaljskog govora. Srpski dijalektoloski zbornik 3,
115-176.
Selkirk, E. O. (1978). On prosodic structure and its relation to syntactic structure. In Nordic
Prosody II (T. Fretheim, ed.), pp. 111 - 1 4 0 . TAPIR, Trondheim, Norway.
Selkirk, E. O. (1980). Prosodic domains in phonology: Sanskrit revisited. In Juncture (M.
Aronoff and M.-L. Kean, eds.), pp. 107-129. Anma Libri, Saratoga, Calif.
Selkirk, E. O. (1986). On derived domains in sentence phonology. Phonology Yearbook 3,
371-405.
Stevanovic, M. (1940). Sistem akcentuacije u piperskom govoru. Srpski dijalektoloski
zbornik 10, 6 7 - 1 8 4 .
Vukovic, J. L. (1940). Akcenat govora Pive i Drobnjaka. Srpski dijalektoloski zbornik 10,
185-417.
Zee, D. (1988). Sonority Constraints on Prosodic Structure. Doctoral dissertation, Stan-
ford University, Stanford, Calif.
LANGUAGE INDEX

A C

Abkhaz, y deletion, 127 Carib, stress, 9 1 - 9 8 , 1 0 6 - 1 0 7

Ahtna Catalan
morphological structure, 48, 151 nasal assimilation, 15, 255, 2 5 6 - 2 6 2
phonology, 152 cluster simplification, 2 5 6 - 2 6 0
Amele, final consonant invisibility, 8 4 - 8 5 , 95, syllabification, 2 5 8 - 2 5 9
105 underspecification of coronals and nasals,
Arabic, Classical 256-262
closed syllable shortening, 1 2 4 - 1 2 5 vowel lowering, 2 9 3 - 2 9 5
glide elision, 1 2 4 - 1 2 7 , 135 Chamorro
possessive allomorphy (1-singular diph- free element condition, 302
thongization) 1 2 6 - 1 2 7 , 1 3 4 - 1 3 5 vowel lowering, 2 9 9 - 3 0 1
vowel elision, 1 2 4 - 1 2 7 , 135 Chichewa, focal phrase, 104
Arabic, Levantine, free element condition, 302 Chimwiini, /-alveolarization and passive forma-
Arabic, Maltese tion, 69, 1 3 5 - 1 3 7
apocope, 1 1 5 - 1 1 6 , 139-141 Chinese, see Shanghai
domain assignment of pronominal suffixes, 4, Choctaw, free element condition, 302
114-115, 139-140 Chumash
lengthening, 140-141 pre-coronal laminalization, 2 9 7 - 2 9 9 ,
stress, 1 1 5 - 1 1 6 , 138-141 317-320
Athabaskan sibilant harmony, 2 9 7 - 2 9 9 , 3 1 6 - 3 2 0
functional element order, 157
morphological structure, 1 4 5 - 1 4 6 , 1 5 2 - 1 5 5 ,
161,346-347
phonological domains, 1 5 0 - 1 5 5 , 162, 163, D
346-347
s deletion 3 4 6 - 3 4 7 Dagbani 235ff
lexical tone rules, 239, 250
postlexical tone rules, 1 1 - 1 3 , 2 3 9 - 2 4 4 ,
B 247-248
structure preservation in, 1 1 - 1 2 , 235, 241,
Basque, and revised alternation condition, 264 245,247, 251
Beaver, morphological structure, 149, 152 Dakota, free element condition, 302

407
408 Language Index

Danish, imperative formation, 48, 54, 69, stress assignment, 3 5 - 3 7 , 4 8 - 4 9 , 216,

127-128, 131-132 2 1 7 - 2 1 8 , 302, 303
Diola Fogny, homorganic clusters, 258 syllabification, 2 2 1 - 2 2 5
Diyari trisyllabic shortening, 10
free element condition, 302 underspecification of anterior coronals, 257
metrical structure and affixation, 91, 106 vowel rounding, Adelaide dialect, 227
Dutch vowel shortening, 266
clitic pronoun ie, 3 7 - 3 9 , 42 word-level rules, 2 0 2 - 2 0 4
compounds, 78, 83 English, Old
compound stress in nouns, 2 7 - 2 8 agreement, 164
-eurlris allomorphy, 48, 54, 78, 1 3 7 - 1 3 8 Anglian smoothing, 15, 3 2 5 - 3 3 8
-ischlief allomorphy, 25, 4 1 - 4 2 , 138 a restoration (West Saxon), 334, 335
stress, 41, 138 back mutation, 3 2 7 - 3 3 0 , 3 3 2 - 3 3 8
breaking, 330, 333, 334, 335
e raising, 330
E / lowering, 330
/ monophthongization, 330, 331, 332, 333,
English, Modern 335
-able, 8 i mutation, 331, 332, 333, 335
adjective extrametricality, 27 retraction, 331, 332, 333, 334
ae tensing, 2 2 6 - 2 2 7 vowel deletion, 330
Aitkin's Law, Scottish and Irish dialects,
2 1 0 - 2 1 2 , 223,225
-al (adjective-forming), 88 F
-al (noun-forming), 6 6 - 6 7 , 71, 137, 143
anaphoric islands and reference into words, Finnish
1 8 1 - 1 8 7 , 191 consonant gradation, 279, 2 8 3 - 2 8 4 , 2 9 5 - 2 9 7
comparative suffix, 3 4 - 3 5 , 3 9 functional element order, 157
dentalization, Belfast dialect, 2 0 9 - 2 1 0 , 223 interaction of stress and suffixation, 4 8 - 4 9
flapping, 219 t assibilation, 2 8 2 - 2 8 3 , 2 8 6 - 2 8 8 , 2 9 1 - 2 9 2 ,
free element condition, 302, 303 315-316
functional categories, 157, 191 t deletion, 2 8 0 - 2 8 2
g deletion, 202, 2 1 5 - 2 1 6 , 222 vowel coalescence, 2 8 0 - 2 8 2
geminates, 2 1 8 - 2 1 9 vowel harmony, 14
/-laxing, Southern English dialects, 228 French, Quebecois vowel lengthening,
interaction of stress and expletive infixation, 347-348,360-361
48-49
invisible suffixes, 105
-/Yy, 8, 3 5 - 3 6 , 87 G
/-darkening variables, 1 7 5 - 1 7 9 , 189, 191
lexical phonology and phonetic implementa- Georgian, portmanteau morphemes, 149
tion in, 174, 178, 1 8 0 - 1 8 1 , 1 8 8 - 1 9 0 German
/-vocalization, London dialect, 2 1 2 - 2 1 3 allophonic rules, 2 0 8 - 2 1 4
-mit, 87 compounds, 78, 83
monophthongization, Northern Irish dialect, fricative assimilation, 2 0 5 - 2 0 8 , 225
213 g deletion, 2 0 4 - 2 0 5
nasal assimilation, 2 1 6 - 2 1 7 interaction of -eilerei allmorphy and stress,
nominal extrametricality, 85 48-49, 69-70
o lowering, London dialect, 2 2 7 - 2 2 8 interaction of g deletion and suffixation, 48,
Pre-/ breaking, 228 5 5 - 5 6 , 69
-'j (clitic), 40, 87 Schaffhausen dialect, 356, 360
sonorant syllabification, 2 0 2 - 2 0 4 schwa insertion in nouns, 28
Language Index 409

sonorant syllabification, 2 0 4 - 2 0 5 phonological overapplication in reduplication,

Swiss dialects, 3 5 6 - 3 5 7 127
umlaut, 129, 3 5 6 - 3 5 7 Juaneno, 58
word level rules, 2 0 4 - 2 0 8
Greek, Ancient
continuant dissimilation, 361 K
free element condition, 302
Greek, Modern Kaska
compounds, 83 s deletion, 3 4 6 - 7 , 360
continuant dissimilation, 3 4 8 - 3 5 1 , 3 5 2 - 3 5 5 , tone, 361
362 Kihehe
Cypriot dialect, 344, 3 5 1 - 3 5 5 , 360, 362 prosodic word, 130-131
definite article, 39 reduplication and glide formation, 48, 53,
functional element order, 157 127,129-131
minimal phrase, 105 Kimatuumbi
obstruentization of y, 3 5 1 - 3 5 2 , 362 glide formation, 118 - 1 2 2
phonological domain construction, 105 initial tone insertion, 1 2 1 - 1 2 3
Samothraki dialect, 3 5 7 - 3 6 0 levels, 119
voiced stops, 3 6 1 - 3 6 2 locative truncation, 1 3 2 - 1 3 3
shortening, 118, 1 2 0 - 1 2 1 , 123, 129
Koasati, rhyme deletion (plural formation),
H 4 8 - 4 9 , 69, 1 3 1 - 1 3 2 , 134
Korean
Hausa, phonological domain construction, 105
aspiration coalescence, 269
Hebrew, interaction of root extraction and me-
default vowel, 143
tathesis, 4 8 - 4 9 , 5 2 - 5 3 , 7 0
nominative destructuring (ilka allo-
Hungarian, vowel harmony, 273
morphy), 1 3 3 - 1 3 4
obstruent neutralization, 269
palatalization, 10, 255, 2 6 3 - 2 6 4 , 2 6 6 - 2 6 7 ,
I
289-291
phonological domains, 7, 8, 105
Icelandic, Modern
underspecification of coronals, 257, 2 6 6 - 2 7 0
a deletion in verbs, 48, 54, 69, 127
Kwakwala, determiner particle, 39
M-umlaut, 2 9 2 - 2 9 3
Indonesian
bracketing paradoxes, 34, 78
L
compounds, 83
free element condition, 302
lexical phonological domains, 79 Lapp, interaction of suffixal allomorphy and
Italian, compounds, 83, 105 stress, 4 9 - 5 2 , 70
Latin, portmanteau morphemes, 149
Luganda, apparent evidence for interactionism,
J 137
Luiseno, interaction of spirantization and word
Japanese formation, 48, 54, 5 7 - 6 0 , 71
invisible suffixes, 105
minimality, 105, 304
root-root compounding, 106 M
Javanese
a raising and habitual-repetitive, 6 7 - 6 9 , 71 Malayalam
interaction of a raising and word formation, compounds, 78, 83
48,54, 60-66,71 Maltese, see Arabic, Maltese
410 Language Index

Manam Serbo-Croatian
free element condition, 302 genitive plural delinking, 3 9 1 - 3 9 4
stress and clitics, 101-103 high tone spreading, 14, 365, 3 6 8 - 4 0 0
Margi, tone assignment and suffixation, 91 initial high insertion, 373, 3 7 8 - 3 8 1
Mende, interaction of mutation and reduplica- metatony, 3 8 7 - 3 9 0
tion, 48, 53 Neo-Stokavian dialects, 365, 3 6 8 - 3 7 1 ,
373-400
Old Stokavian dialects, 368, 3 7 1 - 3 7 3 , 382,
N 394, 3 9 7 - 4 0 0
postcyclic lexical level, 100
Navajo tone linking, 3 8 3 - 3 9 0
functional element order, 157 Shanghai, phonological domains, 105
Shi, apparent evidence for interactionism, 137
morphological structure, 1 4 8 - 1 4 9
Slave
Nganhcara, pronominal clitics, 3 8 - 3 9
morphological structure, 1 4 6 - 1 5 1 , 153-161
Norse, Old, scattered rule, 361
phonological domains, 9, 153, 1 6 2 - 1 6 6
Spanish
interaction of la/el allomorphy and stress,
P
4 - 5 , 69
interaction of stress, e raising, and adverb for-
Polish mation, 6
comparative suffix allomorphy, 26, 42 spirantization, 351, 362
imperative suffix allomorphy, 2 6 - 2 7 Sundanese, nasal harmony and plural infixation,
non-isomorphic phonological and morpho- 48, 5 4 - 5 5
logical structure, 6 - 7
postcyclic lexical level, 100
T

R Tagalog, reduplication, 29, 3 2 - 3 4 , 42, 48, 53

Tigrinya, default vowel, 143
Russian Tiv, tone assignment and suffixation, 91
devoicing, 15 Turkish
voicing assimilation, 14, 255, 262, 354 final devoicing, 284
minimality, 105
revised alternation condition, 264, 284
S vowel harmony, 3 0 6 - 3 0 7

Sanskrit W
compounds, 83
infixation and ruki rule, 48, 54 Warlpiri, stress, 106
ruki rule, 2 8 4 - 2 8 5
Vedic, 2 8 4 - 2 8 5
Sekani Y
morphological structure, 1 4 8 - 1 4 9
s deletion, 3 4 6 - 3 4 7 Yawelmani Yokuts, vowel harmony, 272
subject prefix allomorphy, 361 Yidiny reduplication, 3 0 - 3 2 , 42
tone, 361 Yoruba, vowel harmony, 305
SUBJECT INDEX

A 9 4 - 9 6 , 9 8 - 1 0 3 , 2 1 9 , 331
and nonderived environment blocking, 278
Absolute neutralization, see Phonological rule, postcyclic level, 200, 3 6 6 - 3 6 7 , 3 7 3 - 3 7 4
neutralizing precyclic rule application, 1 4 1 - 1 4 2
Across-the-board application, see Postlexical as property of affix or level, 88, 1 0 2 - 1 0 3 ,
rule, characteristics of 200, 201
Allophonic rule, see Phonological rule, of rule application, 5, 16, 7 5 - 1 0 4 , 112,
allophonic 1 1 5 - 1 1 , 2 8 0 - 2 8 2 , 382, 3 8 5 - 3 9 3
Allomorphy, 4 - 5 , 2 5 - 2 7 , 4 8 - 5 2 , 69, 126-127,
1 3 3 - 1 3 5 , 1 3 7 - 1 3 8 , 361
Analogical change, 3 2 7 - 3 2 8 , 329 D
Anaphoric island, see Lexical integrity
Derived environment condition
blocking in non-derived environment, 9,
277-307
B
as characteristic of lexical rules, 16, 255, 267,
Bound root, see Morphological constituent 270
Boundary strength, 1 7 5 - 1 7 9 as characteristic of structure-preserving rules,
Bracketing, morphological 255,263-265,267
erasure, 81, 173, 188 explanation via strict cycle condition, 315,
paradoxes, 3 4 - 3 7 , 40, 105 319
visibility to phonology, 2 - 3 , 16, 29, 77, 81, explanation via underspecification, 319
88, 90, 173, 180 phonological versus morphological, 9, 320
in vacuously derived environment, 2 9 1 - 2 9 5
Domain, of rule
C and historical change, 1 4 - 1 5 , 3 4 4 - 3 4 8 ,
353-361,365-366
Clitic, 3 7 - 4 0 , 8 6 - 8 7 , 101, 367, 396 noncyclic, see Cyclicity
Clitic group, see Domain, prosodic, clitic group predictability of, see Strong domain
Constraint, persistent, 326, 338, 339, see also hypothesis
specific constraint as property of individual rule, 15, 255, 258.
Cyclicity 262, 265
versus noncyclic rule application, 8 9 - 9 0 , word-internal, see Level, word

411
412 Subject Index

Domain, prosodic, 76, 78, 150, 152, 162. See H

also Level
clitic group, 76, 79, 104, 367, 374, 377, 378, Head operation, 2 8 - 2 9 , 34, 40
380, 3 9 4 - 3 9 6
construction of, 24, 77, 8 0 - 8 2 , 8 4 - 8 6 ,
8 8 - 9 1 , 9 7 , 0 5 - 1 0 6 , 151, 153, 1 6 3 - 1 6 6 I
hierarchy, 24, 7 6 - 7 7 , 80, 164
Indirect reference hypothesis, 77
mismatch with metrical structure, 76, 78, 80,
Inflection, 1 4 5 - 1 4 6 , 149, 151, 153, 155, 156,
168
160, 164,333
mismatch with morphological structure, 2,
Interaction of phonology and morphology, 2,
6 - 7 , 24, 29, 42, 76, 78, 8 2 - 8 3 , 85, 104,
4 5 - 7 0 , 77, 111, 117, 199-201
146, 151, 188
apparent evidence for, 6 6 - 6 8 , 1 2 4 - 1 2 9 ,
mot, see Domain, prosodic, phonological
1 3 7 - 1 3 8 , 142
word
morpheme outside of phonological domain, 3,
p-constituent, p-structure 76, 78
48, 5 4 - 5 6 , 9 1 , 3 3 0 - 3 3 2
phonological phrase, 78, 1 6 4 - 1 6 5 , 366,
morphological sensitivity to derived phono-
375-376, 378-379, 381-382, 394-396
logical property, 3, 2 5 - 2 7 , 4 8 - 5 3 , 76,
phonological word, 1 4 - 1 6 , 3 0 - 4 0 , 48,
87,127
7 8 - 7 9 , 101-102, 1 3 0 - 1 3 1 , 1 6 4 - 1 6 5 ,
overapplication of phonology in reduplica-
180, 201, 2 1 6 - 2 1 7 , 304, 3 6 6 - 3 6 7 , 374,
tion, 29, 48, 5 3 - 5 4 , 127, 129-131
3 7 7 - 3 8 1 , 3 9 4 - 3 9 6 , 3 9 7 - 3 9 9 . See also
in prosodic morphology, 3 - 4 , 30, 35, 48, 69
Level, word
Interaction of phonology and syntax, 112, 121.
small word, 1 6 4 - 1 6 5 , 167
See also Indirect reference hypothesis;
subcategorization of affix for, 2 5 - 2 6 , 2 9 - 3 5 ,
Lexical rule
38, 80, 8 6 - 8 7 , 9 0 , 1 3 9 - 1 4 0 , 146,
Invisibility, see Extrametricality
1 5 2 - 1 5 3 , 163
subcategorization of clitic for, 3 7 - 4 0 , 8 6 - 8 7 ,
102, 107
L
Domain limit rule, 399, 400
Language acquisition, 2 7 8 - 2 7 9
Level, 112, 114, 366
E cyclicity of, see Cyclicity
discontinuous, 1 6 1 - 1 6 2
Elsewhere condition, 26, 86, 277, 278
lexical, 2 - 3 , 5 - 9 , 4 6 - 4 7 , 7 8 - 8 0 , 82,
Exceptions
3 6 6 - 3 6 8 , 374, 3 9 4 - 3 9 6
explained via derived environment condition,
noncyclic, see Cyclicity
279
postcyclic, see Cyclicity
explained via level ordering, 333
postlexical, 282, 2 8 4 - 2 8 5 , 3 4 4 - 3 4 5 , 347,
property of affix, 87
3 5 3 - 3 5 6 , 3 6 6 - 3 6 8 , 374, 3 9 4 - 3 9 6 , 397,
property of lexical rules, 28, 255, 343
400
property of postlexical rules, 16
stem, 199, 200, 2 1 8 - 2 2 1 , 332, 339
Extrametricality, 2 6 - 2 8 , 8 4 - 8 5 , 8 7 - 8 8 , 93,
word, 78, 82, 99, 199-228, 2 8 2 - 2 8 4 , 332,
9 7 - 9 8 , 100-101
338, 355
Level ordering hypothesis, 2, 6 - 7
Lexical diffusion, 17, 343
F Lexical entry, 147, 150, 152, 161, 165
Free element condition, 3 0 1 - 3 0 3 Lexical integrity, 173, 189, 192
Functional category, see Domain, prosodic, con- and reference into words, 181 - 1 8 6
struction of; see also specific language in Lexical phonology, versions of
Language Index extended, 117, 1 2 3 - 1 2 4
Subject Index 413

noninteractive, 112, 114, 1 2 2 - 1 2 4 , 127-128, P

132, 137, 142
prosodic, 7 5 - 7 6 , 8 0 - 8 3 , 8 8 - 9 1 , 9 9 - 1 0 0 ,
103 Peripherality condition, see Extrametricality
standard, 3, 46 Persistent rule, see Constraint, persistent
Lexical rule Phonetic implementation, 2, 247
defined, 191
characteristics of, 2 - 3 , 9, 11, 16, 112, 255,
343, 366 sensitivity to metrical structure, 1 9 1 - 1 9 2
having access to output of later level, sensitivity to morphological structure, 174,
138-141 178, 1 8 0 - 1 8 1 , 1 8 8 - 1 9 0
having access to output of syntax, 112, 117, sensitivity to syntactic structure, 180, 192
1 2 1 - 1 2 2 , 124 Phonological change, 4 - 1 5
rule addition, 329, 338, 344, 3 6 5 - 3 6 6 ,
3 6 7 - 3 6 8 , 400
M rule domain change, 1 4 - 1 5 , 3 9 7 - 4 0 0
rule generalization, 344, 346, 352, 354, 360
rule reordering, 329, 3 4 4 - 3 4 6 , 352, 394
Minimal word, see Domain, prosodic, phono- Phonological phrase, see Domain, prosodic,
logical word phonological phrase
Morphological constituent, 76, 81 Phonological rule
discontinuous, 145, 149, 153 addition, see Phonological change
hierarchy, 87 allophonic, 11, 16, 287
phrasal affix, 40 allophonic application in lexicon, 2 0 8 - 2 1 4
root, bound, 8 6 - 8 7 , 8 9 - 9 0 , 106 allophonic application in postlexical compo-
stem, 8 1 , 8 6 , 8 9 - 9 0 nent, see Postlexical rule, and redundant
subcategorization for, 24, 29, 3 2 - 3 4 , 8 6 - 8 7 , features
90, 145, 159, 166 application in different components, 235, 244,
Morphological rule 250, 255, 263, 264, 270
defined, 4, 6 8 - 6 9 , 1 2 8 - 1 2 9 application to affixes, 2 1 4 - 2 1 6
reduplication as, 1 3 0 - 1 3 1 , 1 4 2 - 1 4 3 automatic, 278
subtractive, 1 2 7 - 1 2 8 , 1 3 1 - 1 3 3 , 135 biplanar, 2 7 - 2 8
Morphological template, 1 4 5 - 1 4 6 , 1 5 2 - 1 5 3 , cyclic, see Cyclicity
1 6 0 - 1 6 1 , 166 extension of, see Phonological change, rule
Morphology-phonology interaction, see Inter- generalization
action of phonology and morphology generalization, see Phonological change, rule
generalization
gradient application, 17, 191
N lexical, see Level, lexical; Lexical rule
morphologically conditioned, 133, 137
Neogrammarian hypothesis, 327, 343 neutralizing, 3, 9 - 1 1 , 96, 269, 2 7 7 - 2 7 8 ,
Neutralization, absolute, see Phonological rule, 307
neutralizing obligatory, 2 7 7 - 2 7 8 , 377
Non-derived environment, see Derived environ- optional, 2 7 7 - 2 7 8 , 287, 307, 377, 380
ment condition ordering, see Rule order
postcyclic, see Cyclicity
postlexical, see Level, postlexical; Postlexical
O rule
reordering, see Phonological change, rule
reordering
Obligatory contour principle, 260, 262, 337 structure-building application, 8 6 - 8 7 , 9 3 - 9 5 ,
Opacity, 10, 360 277, 279, 285, 2 8 6 - 2 8 8 , 291, 307
414 Subject Index

Phonological rule (continued) Rule reordering, see Phonological change, rule

structure-changing application, 285, reordering
2 8 6 - 2 8 8 , 2 9 5 - 3 0 1 , 307, 317, 318
word level, see Level, word
Phonological word, see Level, word; Domain, S
prosodic, phonological word
Phonology/syntax interface, see Interaction of Scattered rule, 344, 3 5 6 - 3 5 9
phonology and syntax Scope, morphological, 153, 168
Postcyclic level, see Cyclicity as determinant of morpheme order, 1 5 7 - 1 6 1 ,
Postlexical rule 165-166
characteristics of, 2, 1 6 - 1 7 , 1 7 3 - 1 7 4 , 187, Second language, transferability of postlexical
190, 253, 255, 278, 2 8 4 - 2 8 5 , 343, 353, rule to, 17
366-367 Sound change, see Phonological change
domains, 2 4 6 - 2 4 7 Stratum, see Level
PI, P2 rules, 12, 245 Stress rule, see specific language in Language
and postlexical syllabification, 2 5 8 - 2 5 9 Index
preceding lexical rules, 1 2 0 - 1 2 2 Strict cycle condition
and redundant features, 255, 2 5 8 - 2 6 2 as cause of derived environment effects, 278,
Precompilation, 2, 4, 117, 1 2 2 - 1 2 3 , 141 307 315
Principle of Phonology-Morphology Segrega- as characteristic of lexical rules, 201
tion, 6 9 , 1 2 9 domain of , 9 , 201
Prosodic, definition of, 7 5 - 7 6 elimination of, 277, 278, 304, 3 0 6 - 3 0 7 , 315
Prosodic domain, see Domain, prosodic and neutralization rules, 9, 2 6 3 - 2 6 5
Prosodic hierarchy, see Domain, prosodic and structure-changing rules, 11
Prosodic subcategorization, see and underspecification, 9, 15
Subcategorization Strict layer hypothesis, 7, 164
Prosodic word, see Domain, prosodic, phono- Strong domain hypothesis, 15, 167, 236, 345,
logical word; Level, word 367-368, 394-397, 399-400
Structure-building application, see Phonological
rule, structure-building application
R Structure preservation
applicability to linked structures, 1 2 - 1 4 ,
262-263
Relatedness Paradox, 186-187, 192 as arbitrary characteristic of rules, 12, 249
Revised alternation condition, 9, 11, 264, 277, as characteristic of lexical rules, 16, 9 2 - 9 4 ,
307 99, 106, 201, 2 5 9 - 2 6 0 , 2 6 4 - 2 6 5 , 270
Root, bound, see Morphological constituent as constraint on derivations, 167, 338
Rule, see Phonological rule and derived environment condition, 255, 270
Rule addition, see Phonological change, rule domain of, 11, 14, 236, 249, 253
addition Subcategorization, see also Domain, prosodic,
Rule extension, see Phonological change, rule subcategorization of affix for; Morphologi-
generalization cal constituent, subcategorization of affix
Rule generalization, see Phonological change, for
rule generalization biplanar, 3 0 - 3 5 , 4 0 , 8 6 - 8 7 , 1 0 6
Rule order
counterfeeding, 289, 329
feeding, 329, 356 T
irreflexivity, 257, 258
opaque, 329. See also Opacity Transparency, see Opacity
paradoxical, 255, 258 Truncation, see Morphological rule, subtractive
Subject Index 415

and nonderived environment blocking, 9 - 1 0 ,

277, 279, 285ff, 295, 307
Underspecification radical, 2 5 7 - 2 5 8 , 266, 268, 270, 285
context-free versus context-sensitive, and structure-building rules, 277, 279, 285ff,
265-270 295
contrastive specification, 349, 3 5 6 - 3 6 0
default rule, 285, 289, 370
as diagnostic of change in rule domain, 15, W
344-345,350-351
Word level, see Level, word
PHONETICS AND PHONOLOGY

V o l u m e 1: R h y t h m a n d M e t e r
edited by Paul Kiparsky and Gilbert Y o u m a n s
V o l u m e 2 : T h e Special S t a t u s of C o r o n a l s : Internal a n d E x t e r n a l E v i d e n c e
edited by C a r o l e Paradis and J e a n - F r a n c o i s Prunet
V o l u m e 3 : C u r r e n t I s s u e s in A S L P h o n o l o g y
edited by Geoffrey R. C o u l t e r
V o l u m e 4 : S t u d i e s in L e x i c a l P h o n o l o g y
edited by Sharon H a r g u s and Ellen M . Kaisse