Structuring the Argument

Language Faculty and Beyond

Internal and External Variation in Linguistics
Language Faculty and Beyond (LFAB) focuses on research that contributes to a deeper
understanding of the properties of languages as a result of the Language Faculty and
its interface with other domains of the mind/brain. While the series will pay particular
attention to the traditional tension between descriptive and explanatory adequacy, the
series will also address issues such as the level of linguistic design, through new lines
of inquiry often referred to as ‘physiological linguistics’ or ‘biolinguistics’. LFAB aims
to publish studies from the point of view of internal and external factors which bear on
the nature of micro- and macro-variation as, for example, understood in the minimalist
approach to language.
For an overview of all books published in this series, please see
http://benjamins.com/catalog/lfab

Editors
Kleanthes K. Grohmann Pierre Pica
University of Cyprus CNRS, Paris

Advisory Board
Paola Benincà Anders Holmberg
University of Padova, Italy University of Newcastle, UK
Cedric Boeckx Lyle Jenkins
ICREA/University of Barcelona, Spain Biolinguistics Institute, Cambridge, USA
Guglielmo Cinque Richard K. Larson
University of Venice, Italy Stony Brook University, USA
Noam Chomsky Andrew Ira Nevins
Massachusetts Institute of Technology, University College London, UK
Cambridge, USA Alain Rouveret
Stephen Crain University of Paris VII, France
Macquarie University, Sydney, Australia Esther Torrego
Marcel den Dikken University of Massachusetts, Boston USA
CUNY Graduate Center, New York, USA Anna Papafragou
Naama Friedmann University of Delaware, Newark, USA
Tel Aviv University, Israel Akira Watanabe
University of Tokyo, Japan

Volume 10
Structuring the Argument. Multidisciplinary research on verb argument structure
Edited by Asaf Bachrach, Isabelle Roy and Linnaea Stockall
Structuring the Argument
Multidisciplinary research
on verb argument structure

Edited by

Asaf Bachrach
Isabelle Roy
UMR 7023 Structures Formelles du Langage, Université Paris 8 - CNRS

Linnaea Stockall
Queen Mary, University of London

John Benjamins Publishing Company

Amsterdam / Philadelphia
 TM
The paper used in this publication meets the minimum requirements of
8

the American National Standard for Information Sciences – Permanence

of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data

Structuring the Argument : Multidisciplinary research on verb argument structure / Edited

by Asaf Bachrach, Isabelle Roy and Linnaea Stockall.
p. cm. (Language Faculty and Beyond, issn 1877-6531 ; v. 10)
Includes bibliographical references and index.
1. Grammar, Comparative and general--Verb phrase. 2. Grammar, Comparative and
general--Augmentatives. 3. Grammar, Comparative and general--Syntax.
I. Bachrach, Asaf, editor of compilation.
P291.S697 2014
415--dc23 2014008330
isbn 978 90 272 0827 9 (Hb ; alk. paper)
isbn 978 90 272 7010 8 (Eb)

© 2014 – John Benjamins B.V.

No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any
other means, without written permission from the publisher.
John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands
John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
 Table of contents

Acknowledgments vii

Introduction: Argumenting the structure 1

Asaf Bachrach, Isabelle Roy and Linnaea Stockall

Part I. The general issue: Verb argument structure

Can we dance without doing a dance? Two opposite views

on the integration of roots in the syntactic structure of the vP 23
Víctor Acedo-Matellán

Determining argument structure in sign languages 45

Carlo Geraci and Josep Quer

The processing and representation of light verb constructions 61

Eva Wittenberg, Ray Jackendoff, Gina Kuperberg,
Martin Paczynski, Jesse Snedeker and Heike Wiese

Part II. Non-canonical argument structure realization

Luigi piace a Laura? Electrophysiological evidence for thematic

reanalysis with Italian dative object experiencer verbs 83
Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

Causative nominalizations: Implications for the structure of psych verbs 119

Artemis Alexiadou and Gianina Iordăchioaia

Part III. Neurobiological models

Neurocognitive mechanisms of verb argument structure processing 141

Cynthia K. Thompson and Aya Meltzer-Asscher
vi Structuring the Argument

Argument structure: Between linguistics and neuroimaging 169

Einat Shetreet

Argument structure: Creating a productive space for theory

and experimentation 185
Gillian Ramchand

Language index 201

Subject index 203
 Acknowledgments

We thank the UMR 7023 Structures Formelles du Langage, CNRS / Université

Paris 8, and in particular Sophie Wauquier, Maya Hickman and Léa Nash, for their
support and help in bringing about the Workshop “Structuring the argument:
A multidisciplinary workshop on the mental representation of verbal argument
structure”, held in Paris, in October 2011. We also thank the workshop present-
ers, participants, commentators (Gillian Ramchand, Hamida Demirdache, John
Beavers, Sarah VanWagenen, Víctor Acedo-Matellán) and invited speakers (Alec
Marantz, Cynthia Thompson, Jesse Snedeker, John Beavers, Josep Quer) for lively
and inspiring discussions and presentations.
Financial support for the workshop was provided by UMR 7023 SFL, Isabelle
Roy’s CNRS Chaire d’Excellence and the project “Structure Argumentale et Struc-
ture Aspectuelle” (Fédération Typologie et Universaux du Langage).
We thank Pierre Pica and Kleanthes Grohmann, and the John Benjamins ed-
itors, for their initiative and continuous support for this volume. We would also
like to thank our many reviewers for their useful comments and feedback on the
manuscripts and of course the authors for their engagement and patience with
the process.

This volume is dedicated to our beloved friend and colleague Sarah Van­Wagenen.
 Introduction
Argumenting the structure

Asaf Bachrach, Isabelle Roy and Linnaea Stockall

This collection is the result of a workshop “Structuring the Argument/Structurer

l’argument” held in Paris in September, 2011. This workshop developed from an
interest in identifying and exploring practical, tangible points of intersection be-
tween theoretical linguists, psycholinguists and neurolinguists working on prob-
lems related to verb argument structure.
Argument structure is, of course, a well studied, foundational topic (indeed
there have been a number of other edited volumes and monographs on verb argu-
ment structure just in the past 5–6 years, including Bowerman and Brown 2008;
Roberge and Cuervo 2012; Duguine, Huidobro and Madariaga 2010; Everaert,
Marijana and Siloni 2012; Hoekstra, Sybesma, Barbiers, Den Dikken, Postma
and Vanden Wyngaerd 2008; Pylkkänen 2008; Ramchand 2008; Randall 2010;
Suihkonen, Comrie and Solovyev 2012). The argument structure of verbs has
long been a central issue in linguistic research of all varieties and continues to
be a vexed, and very central, area of research across a wide range of theoretical
and empirical approaches, as attested by the collection of papers in this volume.
However, in contrast to the richness of intra-disciplinary research and discussion,
the inter-disciplinary perspective and dialogue remains largely underexplored.
The goal of this book is to address that gap. We begin (and end) with many
of the most basic questions that are at the heart of this rich body of work, but our
primary aim is to juxtapose research on specific issues, and highlight the ways
in which different methodologies (ranging from the most ‘conventional’ tools of
careful, cross-linguistic analysis to the increasingly sophisticated tools of cogni-
tive neuroscience) are used to address similar questions. By doing so we showcase
the potential for innovative cross-disciplinary research. The structure of the book
echoes the structure of the workshop, and is organised around three core themes.
The first set of papers in this volume tackles questions about the basic building
blocks of verbal representations and modes of construction of the verb-argument
complex. The second set of papers are concerned with non-canonical argument
2 Asaf Bachrach, Isabelle Roy and Linnaea Stockall

structure realization, with a particular focus on object-experiencer psych verbs.

Finally, the third set offers three overview papers: two from neurobiological per-
spectives, reflecting on successes and weakness in recent and current research,
and the last from a theoretical perspective, focusing on the prospects for the fu-
ture of interdisciplinary research on verb argument structure.
The overall organization of the volume aims to accentuate the insight gained
by considering both theoretical and experimental perspectives on these issues,
and to inspire new work that spans these disciplinary boundaries.

1. The mapping question

It has long been assumed that the lexicon is the locus of the information regarding
which specific arguments a verb requires or allows, the particular configuration
of those arguments, and whether that verb participates in certain argument struc-
ture alternations (but not others). Pān�ini’s 6th century BCE Sanskrit grammar, for
instance, adopts what we would now call a strictly lexicalist position: verbal roots
and nominals are selected from the lexicon where they are stored with a set of
lexical semantic properties – syntactic structures are then derived on the basis of
these lexically stored properties and a set of mapping, or linking, rules (Kiparsky
2002). This lexicalist approach, in which the specific properties of an event are
learned and stored as part of the lexical representation of each individual verb,
continues to form the core of many of the most influential works on argument
structure in more recent times (Jackendoff 1990; Koontz-Garboden 2005; Beavers
and Koontz-Garboden 2012; Levin and Rappaport Hovav 1995, 2005; Pustejovsky
1995). Generalizations, such as the fact that the causers of eventualities are typical-
ly expressed as subjects, while entities that undergo a change of state are expressed
as direct objects, are accounted for by rules governing the mapping of lexical ele-
ments to syntactic structures as in (1) from Levin and Rappaport Hovav (1995).
(1) a. break : [[x do-something] cause [y become broken]]
b. Immediate Cause Linking Rule:
The argument of a verb that denotes the immediate cause of the eventu-
ality described by that verb is its external argument.
c. Direct Change Linking Rule:
The argument of the verb that corresponds to the entity undergoing the
directed change described by that verb is its internal argument.
 (Levin and Rappaport Hovav 1995: 135–146)

Argument structure alternations, such as the passive (3) or the causative/incho-

ative (4), are accounted for by relative linking rules, which appeal to hierarchies,
 Introduction 3

such as Fillmore, Bach and Harms (1968)’s early Subject Selection Rule (2) (for a
recent formulation of this kind of model, see Randall 2010).
(2) Subject Selection Rule:
If there is an A [= Agent], it becomes the subject (S); otherwise, if there is
an I [= Instrument], it becomes the subject; otherwise, the subject is the O
[= Objective, i.e., Theme (T)/Patient (P)].  (Fillmore 1968: 33)
(3) a. Sophia bit the daschund. [A → S, P → direct obj]
b. The daschund was bitten (by Sophia). [P → S]
(4) a. The door opened. [T → S]
b. Sophia opened the door. [A → S, T → direct obj]
c. The battering ram opened the door. [I → S, T → direct obj]
d. Sophia opened the door with the battering ram. [A → S, T → direct obj,
I → indirect obj]
e. *The door opened by Sophia.
f.   *The battering ram opened the door by Sophia.

In approaches of this kind, the syntax doesn’t play any role in actually generating
the argument structure relations between verbs and their arguments, it simply
realizes them in particular grammatical structures.
In the early to mid 90’s a confluence of factors lead to the emergence of a
diametrically opposed position on the relative contributions of the lexicon and
the syntax to the determination of argument structure. These factors included
Chomsky’s proposal of his new Minimalist Program (Chomsky 1993), and partic-
ularly its commitment to Bare Phrase Structure (Chomsky 1999), in which theta
theory and a generative lexicon were exposed as clearly non-minimalist compo-
nents; Hale and Keyser (1993)’s theory of l-syntax, which proposed syntactic ac-
counts for causative/inchoative alternations and denominal verb structure, and
which derived thematic roles from syntactic structures, rather than the other
way around; and Kratzer (1996)’s game-changing proposal that the subject wasn’t
properly the argument of the verb at all, but instead the argument of a higher
functional head, Voice (see Harley 2010 for a more thorough review of these de-
velopments).
Halle and Marantz (1993)’s articulation of the theory of Distributed Morphol-
ogy, which emerged as a response to these events, represents the most radical
variant of the family of approaches in which argument structure properties are de-
termined by the specific functional syntactic structure that the lexical verb occurs
in, with the lexical verb root itself contributing no, or minimal, grammatical fea-
tures to the event interpretation (Arad 2003; Barner and Bale 2005; Embick and
Noyer 2007; Harley 1995; Harley and Noyer 1999; Harley 1999; Marantz 1997,
2007; see also Borer 1994, 1998, 2003, 2005 for an approach with a similar divi-
4 Asaf Bachrach, Isabelle Roy and Linnaea Stockall

sion of labor between the lexicon and the syntax). In this kind of constructionalist
model, generalizations about argument structure are due to the systematic ways in
which particular syntactic structures are interpreted by the conceptual/semantic
system, not to any lexical features of the verbal roots. The fact that some roots
appear not to be grammatical in some syntactic contexts, as in (5), is explained by
critical differences in the fine-grained syntactic structure (Borer 2005; Folli and
Harley 2006).
(5) a. # The city destroyed.
b. # The captain arrived the ship.

In these systems, verbs (or event descriptions) stand in a near-homomorphic rela-

tionship to the syntactic structure: a categoriless root is merged with a functional
head that determines the lexical category (√ROOT + v = ‘verb’), and many other
properties of the verb phrase, including argument structure. There are no stored
lexical verbs in this system, and so no place for a syntax-independent linguistic
representation of event or argument structure.
The three papers in this section are all concerned directly with this debate
about the nature of the mapping between the conceptual or semantic event struc-
ture and the syntactic structure which conveys it.
The first paper, by Acedo-Matellán, takes the radical constructionist approach
as a starting point. In recent work, Marantz (2010, 2011) argues for a particular
version of this model where verbal roots themselves are actually only ever ad-
juncts (modifiers) of a verbal functional head (v) (see Borer 2005 for an earlier
articulation of this kind of analysis). The roots neither select for nor are select-
ed by functional heads or components. This approach is particularly radical in
the context of debates about the relative contributions of lexical and syntactic
information to argument structure, since it totally divorces the stored conceptual
knowledge (the root) from any structured representation of events or their ar-
guments, and completes the surgery begun with Kratzer’s ‘Severing the External
Argument from the Verb’. In Marantz (2011)’s proposal, the verbal root takes no
arguments at all, and the nominal elements that we normally have in mind when
we talk about arguments are in fact arguments of separate sub-event-denoting
functional heads as in (6).
(6) [[v √bake][the cake]]

Marantz (2011) argues for this approach based on data from the semantic scope
of re-prefixation, the interpretation of denominal verbs and the possible forms of
compound verbs. Critically, however, Marantz (2011)’s arguments are based on a
set of generalizations about the possible interpretations of verbal roots in English.
 Introduction 5

Acedo-Matellán reviews Marantz’s arguments and data, and shows that

cross-linguistic and corpus data raises questions about the robustness of these
generalizations. Acedo-Matellán concludes that while the roots and functional
heads approach to lexical categories is not challenged by this data, the status of
roots as modifiers is problematic. Instead, Acedo-Matellán argues for a version
of the model where roots are taken to be selected arguments of functional heads.
This more moderate version of the model opens the possibility that verb classes
are, at least partially, represented in the basic lexicon (determined by selectional
restrictions), and thus takes a position closer to the one argued for by Ramchand
(2008). The apparently narrow question of whether there is only one, or more than
one, way to combine a root with a DP, is of course, not narrow at all, and in fact
this question has broad implications for the much larger question of how encyclo-
paedic/conceptual information is integrated into the grammatical system and to
future neurolinguistic investigations of the mapping question. Methodologically,
this paper is a clear example of how and why careful cross-linguistic comparison,
and the integration of corpus data within mainstream linguistic analysis, remain
vital, alongside psycho and neurolinguistic research.
The second paper in this section, by Geraci and Quer, addresses the mapping
question from a different perspective, one emerging out of sign language research.
The characteristic iconicity of sign languages has brought certain researchers to
consider a very different solution to the mapping question, one where argument
and event realizations are represented ‘directly’ or iconically (using space, point-
ing and gestures) rather than via an arbitrary lexico-syntactic interface, and thus
where the issue of how to map between conceptual and morpho-syntactic rep-
resentations simply doesn’t arise. This perspective rests on the assumption that
there is a deep, fundamental difference between oral and signed languages, mo-
tivating a completely different architectural framework for understanding how
they function. Geraci and Quer argue against this position and instead suggest
ways in which the treatment of argument structure and event structure realization
in sign languages can be reconciled with the treatment of these issues in spoken
languages.
Like Acedo-Matellán, Geraci and Quer offer analyses that are compatible with
the Distributed Morphology framework specifically, and with the constructionist
approach to argument structure in general. The overall thrust of their chapter is
that sign languages offer an under-exploited resource for addressing questions
about how verb meanings and argument structure are determined. They show
that the same kinds of alternations that occur in spoken languages, such as the
unaccusative/unergative and unaccusative/transitive alternations, are well attest-
ed in a range of sign languages as well, but that the grammatical reflex of these
alternations appears in classifier predicates, not the lexical verbs. Geraci and Quer
6 Asaf Bachrach, Isabelle Roy and Linnaea Stockall

explore the extent to which the familiar tests for valency from spoken languages
can be extended to sign languages, and what these tests reveal about the syntax
and verbal agreement patterns in a number of sign languages. Geraci and Quer’s
proposals are, necessarily, preliminary, given the state of research on the mor-
pho-syntax of argument structure realization in sign languages, but they conclude
that there is at least no prima facie reason to believe sign and oral languages are
fundamentally incommensurable with respect to the ways in which they encode
argument structure relations. Geraci and Quer end by identifying a number of ar-
eas where further research is required, and can be, they believe, fruitfully carried
out. They are ultimately optimistic about the prospects for research in this vein.
The third paper, by Wittenberg, Jackendoff, Kuperberg, Paczynski, Snedeker
and Wiese presents a series of experiments pitting various alternative models of
the basic syntax of verb argument structures against one another. Wittenberg et al.
summarise a body of research that attempts to adjudicate between a set of possible
general models that divide the labour of representing and computing event inter-
pretations in different ways, using evidence from sentence processing. In addition
to the basic lexical storage and syntactic composition alternatives, summarized
above, that have for decades constituted the two main options for accounting for
verb argument structure differences, Wittenberg et al. also consider more recent
variants on these two basic options such as the Construction Grammar approach
(Goldberg 1995), and the Parallel Architecture model (Jackendoff 2009).
In Construction Grammar, whole constructions (for example a ‘passive con-
struction’ comprising a patient argument, verb, and optional by-phrase agent
argument) are stored as ‘lexical’ elements, in addition to the verbs and nouns
and such that slot into these constructions, and thus this model constitutes a
hyper-lexicalist position (even the syntactic structures themselves are stored as
lexical elements, not generated in a separate syntactic structure building compo-
nent). Jackendoff ’s Parallel Architecture approach, on the other hand, diverges
from the other models in that it includes an additional separate generative seman-
tics component, thereby raising the possibility that some aspects of verbal argu-
ment structure are determined by this semantic composition rather than either
the lexical representation of the verb or the syntactic structure that verb occurs in.
The test case used to set the different approaches against each other is the
light verb construction. Light verb constructions such as take a look seem to be
syntactically identical to non-light constructions such as take a book, yet have a
very different mapping from the semantics of the verbal and nominal roots to
the event interpretation. Across the set of experiments Wittenberg and colleagues
have conducted, sentences containing light verbs are systematically associated
with greater processing costs than matched non-light-verb sentences. This pro-
cessing evidence strongly suggests that light verb constructions involve additional
 Introduction 7

representational and or processing complexity as compared to non-light con-

structions. In their discussion of these results Wittenberg et al argue that the spe-
cific pattern of complexity differences is best explained by (and hence argues for)
the Parallel Architecture.
The particular interest of considering these three papers together does not lie
in the fact that they represent (at times radically) diverging theoretical perspec-
tives, but in that they make use of very different tools and empirical phenomena
to address one and the same question, that of the nature of the architecture under-
lying the mapping of conceptual structure into linguistic expressions.
Any theory of argument structure representation generates predictions not
only regarding the nature of the basic representations of verbs and verb phrases
but also about the distribution and properties of argument structure alternations
or non canonical realizations. The second set of papers tackles this precise issue,
which has been at the heart of work on verb argument structure for decades: how
best to account for argument structure mappings that seem to violate the princi-
ples and generalizations about precedence of the sort articulated in Fillmore et al.
(1968)’s Subject Selection Rule (see (2), above), or Baker (1985)’s Uniformity of
Theta Assignment Hypothesis.

2. Non canonical argument structure realization

In phrasal syntax, non-local dependencies are a long-standing empirical chal-

lenge for models of grammar; and the issue of how to represent the variability in
the relation between grammatical relations and relative surface positions is the
source of ongoing debate. In the domain of argument structure, alternations in
the mapping of conceptual event participants to syntactic verbal arguments rep-
resent the analogue (and possibly related) challenge.
Event participants are often classified into theta-roles, commonly defined
on the basis of the semantic role of the participant (e.g., agent, patient, theme,
source, beneficiary, location, and so on), or as clusters of such roles, cf., Dowty
(1991)’s proto-roles (proto-Agent, proto-Patient). As discussed above, particular
theta-roles are typically associated with particular grammatical/syntactic verbal
arguments, such that agents are most commonly realized as external arguments
(i.e., subjects), patients/themes as internal arguments (i.e., direct objects) and
others such as beneficiary or location, for instance, are generally expressed as
oblique objects (i.e., indirect objects).
Although the exact correspondence between argument role and syntactic
structure might differ across languages and even across verbs within a single lan-
guage, research adopting some version of the lexicalist position has identified/
8 Asaf Bachrach, Isabelle Roy and Linnaea Stockall

proposed a range of global principles of canonical mapping in terms of hierarchy

matching (see Fillmore et al. 1968’s Subject Selection Rule above; Baker 1985’s
Uniformity of Theta Assignment Hypothesis (UTAH); Tenny 1994’s Aspectual
Mapping Hypothesis; Levin and Rappaport Hovav 1995’s linking rules; among
others). These views assume that thematic roles and verbal arguments are intrin-
sically ordered; thematic roles in terms of cognitive event saliency, and verbal
arguments in terms of relative hierarchical position in the syntactic structure (or
their relative prominence in terms of scope or accessibility to transformations). In
a canonical mapping, the relative syntactic order between two verbal arguments
reflects their relative thematic ordering.
However, this ‘canonical’ mapping is far from absolute, since the same verb
(or verbal root) often allows for alternations in the mapping where there is a mis-
match between the argument hierarchy and the thematic hierarchy. One well-
known example of this non canonicity in mappings is passive voice, where the
agent is realized as an oblique argument while the patient/theme (which is the-
matically ‘lower’ than the agent) is realized as the external argument (which is
‘higher’ than the oblique) (7).
(7) a. John ate the chocolate cookies. (active voice)
b. The chocolate cookies were eaten by John. (passive voice)

Another case of mismatch concerns unaccusative verbs. While patients/themes

normally map into the internal argument position, unaccusative verbs are char-
acterized as having an external argument realizing a patient/theme role. Unaccu-
satives contrast with unergatives, another class of intransitive verbs that involves
typical agent to external argument mapping (compare (8) and (9)).
(8) a. The ice melted.
b. The sun melted the ice.
c. the melted ice
(9) a. The baby slept.
b. *The mother slept the baby.
c. *the slept baby

Certain unaccusative verbs are identifiable in English by the existence of a ‘ho-

mophonous’ verb with a ‘canonical’ transitive mapping (where the patient/theme
is realized as the object; e.g., melt, break, dry). However, in other cases, such al-
ternation is not observed (e.g., fall, arrive). In these cases, semantic evidence, and
in languages other than English, morphological evidence, can be used to identify
unaccusativity (for example the choice of auxiliary in Italian; Burzio 1986, (10)).
 Introduction 9

(10) a. E arrivato
(lit. (he) is arrived)
b. Ha telefonato
(lit. (he) has phoned)

Unaccusativity is different from passive voice in a number of important ways.

First and foremost, while in the passive the agent role is conserved (either via an
oblique or implicitly), in the unaccusative realization of a verb the agent role is
absent from the conceptual structure itself: The cookies were eaten in (7) entails
that there was an agent (an eater for the cookie), even if the by-phrase is not
pronounced, but, according to some, (8a) does not entail that there was an agent
of the melting event, and a by-phrase explicitly naming an agent is not possi-
ble. Second, unaccusative/transitive alternations are not always possible, while
for any passive there is always a related active form. Furthermore, unaccusativity
is associated with particular event structures in terms of verbal aspect (unaccu-
satives tend to be telic), while passive voice can apply across the board to almost
any transitive verb. Finally, at least in English, unaccusativity is not marked mor-
phologically. This specific construction, and the long and lively debate around its
analysis are dealt with extensively in the third section of this book.
The two papers that compose the second section of this book focus, albeit
from very different perspectives, on a third case of non-canonical mapping, that
of psychological (or ‘psych’) verbs. Psych verbs are stative verbs that express a
psychological state and assign the role ‘experiencer’ (of that psychological state)
to one of its arguments. Two groups of psych verbs, or constructions, are usually
contrasted depending on whether the experiencer is mapped into the subject or
the object position. The former are referred to as Subject Experiencer (SE) verbs,
and are considered to be the canonical expression of psychological events since
experiencers share many common features with agents, such as saliency and ani-
macy. The latter, Object Experiencer (OE) verbs, are considered the non-canoni-
cal counterpart. There is general debate as to how they should be analyzed: either
as instances of unaccusatives more generally (Belletti and Rizzi 1988; Bennis
2004; Landau 2010) or as a type of causative construction (Pesetsky 1995; Baker
1997). Under either view, they are assumed to involve more syntactic structural
complexity than the Subject Experiencers.
(11) John fears/likes/enjoys our performance. (SE psych verbs)
(12) The prisoner’s parole frightens/pleases/appalls the local community. (OE
psych verbs)
10 Asaf Bachrach, Isabelle Roy and Linnaea Stockall

Dröge, Maffongelli & Bornkessel-Schlesewsky’s and Alexiadou & Iordăchioaia’s

chapters offer two different ways to probe the structure and complexity of ob-
ject-experiencer psych verbs.
In their chapter Dröge et al. tackle the question of the complexity of Object
Experiencer psych verbs from the processing angle. Their paper offers a detailed
study of Italian object experiencer verbs which subcategorize for a dative experi-
encer (e.g., piacere ‘please’). This verb class is compared to a class of active verbs
which also subcategorize for a dative argument. Importantly, both verb classes
allow both O(dat)VS and SVO(dat) word orders as shown in (13).
(13) a. A Gianni piace questo.
to Gianni pleases this
b. Questo piace a Gianni.
this pleases to Gianni

They first provide a detailed review of past and current analyses of object expe-
riencer verbs and word order in Italian. They accept the argument made in these
analyses that Object Experiencer verbs are associated with a complex structure
where the experiencer is linked to dative case while the subject bears the low-
er role (i.e., the theme). Using questionnaires they demonstrate that while both
word orders are possible with psych and active verbs, speakers associate OSV
order with psych verbs and SVO with active verbs. Dröge et al. conduct an ERP
experiment where they cross verb class and word order. ERPs to the verb and to
the post verbal argument are analyzed. At the verb, the authors find 3 distinct dif-
ferences across conditions. First, when a sentence starts with a dative argument,
an active verb produces an N400 response. A later positivity is observed both in
response to an active verb following a dative argument and to a psych verb follow-
ing a nominative argument. While the N400 effect most likely reflects a surprisal
effect (since an initial dative argument strongly predicts a psych verb) the later
positivity more likely reflects the structural cost associated with reanalysis. The
third observed difference is in response to the post verbal argument where nom-
inative arguments following an active verb produce an anterior negativity effect
(compared to the same arguments following a psych verb). The authors interpret
this result as the consequence of the specific information structure associated
with these verbs (focused subject). However, an alternative explanation could be
in terms of movement (the requirement to associate the preverbal dative with a
postverbal gap). Put together, the ERP results offer an interesting way to tease
apart processing expectations, structural reanalysis and structural complexity.
Alexiadou and Iordăchioaia enter into the (cross-linguistic) debate on the
structure of object-experiencer psych verbs from a very different perspective,
 Introduction 11

namely nominalizations. Nominalization offers a nice, yet often neglected, tool

to look at argument structure. Deverbal nominalizations have an argument struc-
ture when they retain the event interpretation that is directly inherited from their
verbal base. Argument structure inheritance is cross-linguistically constrained,
however. Since Remarks on Nominalizations (Chomsky 1970), it has been noted
that nominalizations may take an external causer (cf., the destruction of the city
by the enemy) but not an internal causer (cf., *the growth of tomatoes (by Mary)).
Similarly, nominalizations may be used to tease apart stative vs. dynamic verb
bases, as in recent work by (Alexiadou 2011), and inform our understanding of
verb syntax and semantics.
Alexiadou and Iordăchioaia here present new arguments in favor of distin-
guishing several psych verb constructions in a range of unrelated languages which
are at first sight indiscernible. In particular, one aspect of the behavior of Object
Experiencer verbs that should be tied to their special internal structure is their in-
ability to form nominalizations with a causer external argument even though they
allow agents. From a cross-linguistic perspective, investigating new data from
Greek and Romanian, they observe that there are two types of languages. Some,
including Greek and Romanian, can nominalize either the Subject Experienc-
er anti-causative or (the passive of) the Object Experiencer transitive structure.
Others, e.g., English, only nominalize the Object Experiencer transitive form, as
the verbs lack the anticausative structure. Their results thus suggest a structural
difference between the two classes of languages, only visible through the lens of
nominal formation.
The two papers in this section differ both with respect to the empirical data
and puzzles they investigate and the theoretical and empirical approaches they
employ. Put together, they highlight the importance of using different yet comple-
mentary methodologies to continue addressing long standing theoretical linguis-
tic issues. Just as it is necessary to bring the tools of psychology and neuroscience
to bear in addressing the long standing theoretical questions, it is also crucial to
continue to apply more traditional methods based on native speaker judgement,
especially including less studied languages and the perspective of cross-linguistic
analysis. For instance, future research on Object Experiencer pysch verbs may
built on such theoretical work to generate new paradigms for ERP investigations.
Despite their differences, both papers bring into play the notion of structural
complexity as an important part of the explanation of the investigated phenome-
na. In the next section, devoted to neurological and neuroimaging studies of ar-
gument structure, the notion of complexity, and its manipulation, will take center
stage.
12 Asaf Bachrach, Isabelle Roy and Linnaea Stockall

3. Neurobiological models

The three chapters comprising the third and last section of the book, taken to-
gether, provide a multi-perspective assessment and critical discussion of the state
of the art of our understanding of the neurological underpinnings of argument
structure representation and processing. Given the nature of the data provided by
neurological, and in particular neuro-imaging, techniques, most of the experi-
mental questions discussed in these three chapters address the relative complexity
or the (possibly different) sources of complexity in the representation and process-
ing of argument structure. We have already discussed the issue of representational
complexity (e.g., which of two structures is more complex) and its relation to pro-
cessing complexity (measured in terms of difficulty, time or brain response), as in
the case of psych verbs (Dröge et al.), and the relation between representational
and processing complexity, as in the case of light verb constructions (Wittenberg
et al.). At times, two theories might agree on the relative complexity of a structure
but disagree on the source or nature of that complexity (whether it is semantic,
syntactic or lexical, as in the case of Wittenberg et al.’s chapter on light verbs). In
the terms used throughout the rest of this section, the two theories agree on the
quantitative difference across conditions but disagree regarding the qualitative
interpretation of the difference (the nature of the cognitive process whose role is
quantitatively manipulated).
Thompson and Meltzer-Asscher’s chapter represents the fruits of a large
scale research project investigating the structural complexity of argument struc-
ture representation bringing together patient and fMRI data. Subjects with brain
lesions which have an effect on their language performance (e.g. aphasia) are a
traditional source of neurological data. Lesion data is specifically useful in deter-
mining whether two linguistic structures or processes share common resourc-
es, as indicated by co-pathology. For example, preliminary data from work by
Martínez-Ferreiro, Bachrach, Sánchez and Picallo (2012) indicates that Spanish
aphasic patients who show deficits in the processing of the passive construction,
also show deficits in the processing of unaccusative and Object Experiencer psych
verbs. This co-pathology suggests a common mechanism for these three struc-
tures (possibly, but not necessarily, syntactic movement).
Patient data, however, cannot provide answers to all our questions (for exam-
ple, relative complexity is not necessarily easy to address) as it is often off-line and
depends on the specific and arbitrary distribution of lesions. Furthermore, given
that lesions are rarely spatially constrained, it is often hard to evaluate which is
the relevant brain region supporting a particular cognitive function. Online mea-
sures, such as reaction time or ERPs, can give us a direct estimation of complexity
in healthy subjects, but are highly local in time and can be of low dimensionality
 Introduction 13

(behavioral measures like reaction time, for example, often provide only one data
point per item). Given the temporal locality of such measures, it is not straight
forward to evaluate claims about the structural complexity of temporally extended
structures. Since linguistic theory is mostly interested in representational issues,
and abstracts away from the implementation or the processing of the postulated
structures, any attempt to produce temporally local predictions would require an
explicit parsing/processing theory, which is not a trivial matter (Walenski 2002).
While reaction time measures are able to distinguish the relative complexity of
two structures, their uni-dimensionality makes it difficult to know if the source
nature of their complexity is the same or different.
Functional Magnetic Resonance Imaging (fMRI), which is today’s major neu-
roimaging tool, provides a promising on-line avenue to address issues of (repre-
sentational) complexity with healthy subjects and address some of the limits of
patient data, ERP or reaction time data. fMRI is anatomically multidimensional
so can potentially distinguish different sources of complexity. While providing
online measures, the fMRI signal is smooth (on the order of seconds rather than
milliseconds as in ERP) and so reflects an overall measure of processing cost over
whole structures which can relax the requirement for temporally specific predic-
tions as in ERP.
The neuroimaging work reported in this chapter is predicated on the general
assumption of a causal relation between the relative complexity of a cognitive
process or representation and the amount or extent of neural activity (in a specific
region). In other words, the authors describe the contrast between conditions in
quantitative rather than qualitative terms (see Caplan 2009 for a thorough dis-
cussion of the ways in which neuropsychological or neuroimaging studies can
make comparisons across conditions that are either qualitative or quantitative in
nature).
The work reported in Thompson and Meltzer-Asscher’s chapter relates to
both the issues discussed in the previous sections. One set of experiments ma-
nipulated the ‘size’ of the structure (comparing, for example, transitive verbs to
intransitive ones), reflecting the issues discussed in (§1), while the other set of
experiments specifically address the neural representation of unaccusative verbs
compared to unergative or transitive verbs where the source of complexity is the
non-canonical argument realization (§2).
While the authors are specifically interested in representational complexity
of different argument realization templates, the experiments and findings pre-
sented clearly indicate the difficulty in teasing apart representational from pro-
cessing complexity. For example, the fact that verbs with multiple frames (e.g.
transitive, intransitive, etc.) induce more frontal brain activity than verbs with
only one frame is interpreted (quite reasonably) by the authors as an ambiguity
14 Asaf Bachrach, Isabelle Roy and Linnaea Stockall

effect on processing rather than a representational complexity effect. However, the

increase in fMRI activation (Shetreet, Friedmann and Hadar 2010b) (or decrease
in performance by Aphasic patients, Lee and Thompson 2004) caused by unaccu-
sative verbs compared to unergative verbs can be interpreted as a representational
complexity effect (for example, due to the reflex of the additional movement) or
as a processing effect (perhaps due to the mismatch between the syntactic posi-
tion and thematic role). While neuroimaging could eventually tease apart these
two interpretations (which reflect the traditional linguistic debate regarding the
derivation of unaccusativity discussed in §1), given the current state of the art,
discussed in detail in the chapter, the question remains open.
The chapter ends with a sketch of a neurolinguistic model of argument struc-
ture representation and processing which builds on the authors’ own results and
a survey of related work, as well as anatomical findings regarding white matter
connectivity. Apart from the traditional division of labor between frontal (syn-
tax) and posterior (semantic) regions, the model postulates more fine grained
distinctions within each. Distinct frontal regions are associated with 3 different
functions: resolution of lexical ambiguity, building of basic syntactic structure,
and processing of complex syntactic structures. In the more posterior part of the
network, the authors associate the inferior parietal region with the representa-
tion of the verb’s lexically encoded argument structure information, while tissue
within the superior and medial temporal gyrus supports integration of this lexical
information with the syntactic structure built in the inferior frontal gyrus.
The two final chapters of the section, and the book, are invited commentaries
by speakers at the 2011 Structuring the Argument workshop. A cognitive neu-
roscientist (Shetreet) and a theoretical linguist (Ramchand), both authorities in
their respective fields on the issue at hand, share their personal and critical ap-
preciation of the state of the art in the interdisciplinary investigation of verb ar-
gument structure.
In her chapter, Shetreet presents a critical view of fMRI research on argument
structure. Shetreet is well placed to provide such a critical review since she and her
collaborators have produced an important body of work on the topic (Shetreet,
Palti, Friedmann and Hadar 2007; Friedmann, Taranto, Shapiro and Swinney
2008; Shetreet, Friedmann and Hadar 2009, 2010a, b). Shetreet specifically high-
lights the difficulty in dissociating representational and processing complexity,
and questions a number of the assumptions much of the research on the topic
(including her own work and that of Thompson and colleagues) is built on. The
author clearly makes the point regarding the importance of the interconnection
between theoretical linguistics and brain research. Work in theoretical linguis-
tics can provide well defined questions that can be addressed by neuroimaging
techniques and help avoid running experiments that address ill-posed questions.
 Introduction 15

On the other hand, results from neuroimaging can potentially inform theoretical
debates such as the proper analysis of argument structure templates. For example,
Shetreet et al. (2010a) using fMRI, have provided evidence for a single lexical entry
for optionally transitive verbs such as ‘eat’ (and against a theory where such verbs
have two, transitive and intransitive, frames) and for the presence in the syntax of
a covert argument when such verbs appear without an overt complement.
In the second part of her chapter, Shetreet uses one of her own recent stud-
ies, an elegant experiment on the topic of reflexive verbs, as a demonstration of
the different methodological issues raised in the first part and how her group at-
tempts to address them. She addresses the articulation of the theoretical question,
the difficulties in the design of the experimental paradigm and choice of stimuli
as well as considerations regarding the task and format of representation. One
issue, also raised by Thompson and Meltzer-Asscher, has to do with the choice of
presenting a verb in isolation versus in the context of a sentence. Shetreet ques-
tions the ecological validity of single word presentation and the extent to which
one can draw conclusions from the processing of single words to the processing
of words in their naturalistic phrasal context. She provides a critical summary of
her results and while clearly pointing to the limits on the inference one can make
from them, she discusses their potential theoretical implications. For example,
the lack of activity in the left Inferior Frontal Gyrus for the processing of reflex-
ive verbs (as compared to unaccusatives) is cautiously interpreted as an evidence
against a movement analysis of reflexives.
Ramchand concludes this volume with a second critical take on research
on verb-argument structure, this time from a theoretical linguistics perspective.
Ramchand addresses the current state of research on verb argument structures,
especially argument structure alternations, such as the transitive/unnaccusative
alternation discussed throughout this volume, across multiple disciplines. She of-
fers her thoughts about the possibilities for building on this existing research in
future interdisciplinary collaborations. Ramchand is optimistic about the pros-
pects for advances through cross-disciplinary research but also highlights signif-
icant challenges to overcome and pitfalls to avoid. Ramchand is concerned with
a tendency to over-interpret results from processing and neuroscience as defini-
tively resolving theoretical debates. She points out that the relationship between
the experimental data and theoretical issues is rarely simple and transparent. By
specifically focusing on unaccusativity, Ramchand emphasizes just how nuanced
some of the theoretical issues can be, but she nonetheless proposes concrete and
feasible ways of addressing some of them.
The author frames the theoretical issue in terms of causation. For Ramchand,
a central open question to be resolved (potentially via the use of psycholinguistic
and neurolinguistic tools) is whether unaccusative verbs (or as she might put it,
16 Asaf Bachrach, Isabelle Roy and Linnaea Stockall

‘unaccusative syntactic structures’) contain a cause component. While it is intu-

itively clear that the transitive variant of a verb such as ‘break’ contains such a
component (Mary broke the vase = Mary caused the vase to break), it is under
debate whether the intransitive frame also contains such a semantic cause com-
ponent.
As pointed out by the author, simply finding quantitative differences be-
tween the processing of the transitive and the intransitive variants does not tell
us whether it is the (absence of a) cause component that triggers that difference.
Ramchand also criticizes the ‘lexicalist’ approach of most of the psycholinguistic
literature on the topic of argument structure. While this literature often views
argument structure as a property of specific words (verbs), Ramchand follows
Marantz in viewing argument structure as a read-out of the syntactic structure
(what Thompson and Meltzer-Asscher label the constructivist perspective). For
Ramchand, the lexicalist perspective (which attempts to identify coherent verb
classes) confuses and distorts the relevant theoretical questions and distinctions.
For example, the presence or absence of a cause component does not correspond
to the usual verb class distinctions.
Given her critical assessment of the literature and the possibility of responding
to questions framed in terms of verb classes, Ramchand puts forward a different
research question as a realistic research agenda. She proposes to find out wheth-
er verb types form natural classes with respect to different current experimental
metrics of complexity (or difficulty in processing). The author suggests that an
answer to this question could form the base of a more informed investigation
and testing of theoretically more sophisticated debates. Ramchand concludes that
“creating a productive multidisciplinary space is […] not a luxury but a necessity,
if the field is to advance.” Our hope is that this volume serves as an important step
towards creating this space.

References

Alexiadou, A., 2011. Statives and nominalization. Recherches Linguistiques de Vincennes 40,
25–52. DOI: 10.4000/rlv.1991
Arad, M., 2003. Locality constraints on the interpretation of roots: The case of Hebrew denominal
verbs. Natural Language & Linguistic Theory 21, 737–778. DOI: 10.1023/A:1025533719905
Baker, M. C., 1985. Incorporation: a theory of grammatical function changing. Number v. 1 in
Incorporation: A Theory of Grammatical Function Changing, Massachusetts Institute of
Technology.
Baker, M. C., 1997. Thematic roles and syntactic structures, in: Haegeman, L. (Ed.), Elements of
Grammar. Kluwer, Dordrecht, Netherlands, pp. 73–137.
DOI: 10.1007/978-94-011-5420-8_2
 Introduction 17

Barner, D., Bale, A. C., 2005. No nouns, no verbs? A rejoinder to Panagiotidis. Lingua 115,
1169–1179. DOI: 10.1016/j.lingua.2004.05.001
Beavers, J., Koontz-Garboden, A., 2012. Manner and Result in the Roots of Verbal Meaning.
Linguistic Inquiry 43, 331–369. DOI: 10.1162/LING_a_00093
Belletti, A., Rizzi, L., 1988. Psych verbs and θ-theory. Natural Language and Linguistic Theory 6,
291–352. DOI: 10.1007/BF00133902
Bennis, H., 2004. Unergative adjectives and psych verbs, in: Alexiadou, A., Anagnostopoulou,
E., Everaert, M. (Eds.), The Unaccusative Puzzle: Explorations at the Syntax-Lexicon Inter-
face. Oxford University Press, Oxford. Number 5 in Oxford Studies in Theoretical Syntax,
pp. 84–114.
Borer, H., 1994. The projection of arguments, in: Benedicto, E., Runner, J. (Eds.), University
of Massachusetts Occasional Papers in Linguistics 17. UMOPL, Amherst, MA: GLSA, U.
Massachusetts, pp. 19–47.
Borer, H., 1998. Passive without theta grids, in: Lapointe, S., Brentari, D., Farrell, P. (Eds.),
Morphology and its Relations to Phonology and Syntax. CSLI Publications, Stanford, CA,
pp. 60–99.
Borer, H., 2003. Exo-skeletal vs. endo-skeletal explanations: syntactic projections and the
lexicon, in: Polinsky, M., Moore, J. (Eds.), The nature of explanation in linguistic theory.
Chicago University Press, Chicago, pp. 1–35.
Borer, H., 2005. Structuring sense. Vol. 2, The normal course of events. Oxford University Press,
Oxford. DOI: 10.1093/acprof:oso/9780199263929.001.0001
Bowerman, M., Brown, P., 2008. Crosslinguistic perspectives on argument structure. Lawrence
Erlbaum Associates, New York.
Burzio, L., 1986. Italian Syntax. Reidel, Dordrecht, Netherlands.
DOI: 10.1007/978-94-009-4522-7
Caplan, D., 2009. Experimental design and interpretation of functional neuroimaging studies
of cognitive processes. Human Brain Mapping 30, 59–77. DOI: 10.1002/hbm.20489
Chomsky, N., 1970. Remarks on nominalization, in: Roderick, A., Rosenbaum, P. (Eds.), Read-
ings in English Transformational Grammar. Ginn, Waltham, MA, pp. 184–221.
Chomsky, N., 1993. A minimalist program for linguistic theory, in: Hale, K., Keyser, S. (Eds.),
The View from Building 20. MIT Press, Cambridge, MA, USA, pp. 1–52.
Chomsky, N., 1999. Derivation by phase. MIT Working Papers in Linguistics, Cambridge, MA.
Dowty, D., 1991. Thematic Proto-Roles and Argument Selection. Language 67, 547–619.
Duguine, M., Huidobro, S., Madariaga, N., 2010. Argument structure and syntactic relations: a
cross-linguistic perspective. John Benjamins Pub. Co., Amsterdam; Philadelphia, Pa.
DOI: 10.1075/la.158
Embick, D., Noyer, R., 2007. Distributed morphology and the syntax/morphology interface, in:
The Oxford handbook of linguistic interfaces. Oxford University Press, pp. 289–324.
Everaert, M., Marijana, M., Siloni, T., 2012. The Theta system: argument structure at the inter-
face. Oxford University Press, Oxford; New York.
DOI: 10.1093/acprof:oso/9780199602513.001.0001
Fillmore, C. J., Bach, E., Harms, R. T., 1968. The case for case. Holt, Rinehart, and Winston,
New York.
Folli, R., Harley, H., 2006. On the licensing of causatives of directed motion: Waltzing Matilda
all over. Studia Linguistica 60, 121–155. DOI: 10.1111/j.1467-9582.2006.00135.x
18 Asaf Bachrach, Isabelle Roy and Linnaea Stockall

Friedmann, N., Taranto, G., Shapiro, L. P., Swinney, D., 2008. The leaf fell (the leaf): The online
processing of unaccusatives. Linguistic Inquiry 39, 355–377.
DOI: 10.1162/ling.2008.39.3.355
Goldberg, A. E., 1995. Constructions: A Construction Grammar Approach to Argument Struc-
ture. Cognitive Theory of Language and Culture. University of Chicago Press.
Hale, K., Keyser, S., 1993. On Argument Structure and the Lexical Expression of Syntactic Re-
lations, in: The View from Building 20, pp. 53–109.
Halle, M., Marantz, A., 1993. Distributed Morphology and the Pieces of Inflection, in: Hale, K.,
Keyser, S. (Eds.), The View from Building 20. MIT Press, Cambridge, MA, USA, pp. 111–
176.
Harley, H., 1995. Subjects, Events and Licensing. Ph.D. Massachusetts Institute of Technology.
Harley, H., 1999. Denominal verbs and aktionsart. MIT Working Papers in Linguistics 35, 73–85.
Harley, H., 2010. A Minimalist Approach to Argument Structure, in: The Oxford Handbook of
Linguistic Minimalism. Oxford University Press, pp. 426–447.
Harley, H., Noyer, R., 1999. Distributed morphology. Glot International 4.
Hoekstra, T., Sybesma, R., Barbiers, S., Den Dikken, M., Postma, G., Vanden Wyngaerd, G.,
2008. Arguments and Structure Studies on the Architecture of the Sentence. Walter De
Gruyter. DOI: 10.1515/9783110197358
Jackendoff, R., 1990. Semantic structures. MIT Press, Cambridge, MA.
Jackendoff, R., 2009. Meaning and the Lexicon: The Parallel Architecture 1975–2010. Oxford
University Press, USA.
Kiparsky, P., 2002. On the architecture of pānini’s grammar, in: Three lectures delivered at the
Hyderabad Conference on the Architecture of Grammar, pp. 1–59.
Koontz-Garboden, A., 2005. On the typology of state/change of state alternations, in: Booij, G.,
Marle, J.v. (Eds.), Yearbook of Morphology 2005, pp. 1–28.
Kratzer, A., 1996. Severing the External Argument from Its Verb, in: Rooryck, J., Zaring, L.
(Eds.), Phrase Structure and the Lexicon. Kluwer Academic Publishers, Dordrecht, Nether-
lands, pp. 109–137. DOI: 10.1007/978-94-015-8617-7_5
Landau, I., 2010. The locative syntax of experiencers. MIT Press, Cambridge, MA.
Lee, M., Thompson, C. K., 2004. Agrammatic aphasic production and comprehension of unac-
cusative verbs in sentence contexts. Journal of Neurolinguistics 17, 315–330.
DOI: 10.1016/S0911-6044(03)00062-9
Levin, B., Rappaport Hovav, M., 1995. Unaccusativity: at the syntax-lexical semantics interface.
MIT Press, Cambridge, MA.
Levin, B., Rappaport Hovav, M., 2005. Argument realization. Cambridge Univ. Press,
Cambridge: U.K. DOI: 10.1017/CBO9780511610479
Marantz, A., 1997. No escape from syntax: Don’t try morphological analysis in the privacy of
your own lexicon. University of Pennsylvania working papers in linguistics 4, 201–225.
Marantz, A., 2007. Phases and words. Phases in the theory of grammar, 191–222.
Marantz, A., 2011. Syntactic approaches to argument structure without incorporation: Doing
the (anti-lexicalist) dance without doing the dance, in: Structuring the Argument: A mul-
tidisciplinary workshop on the mental representation of verbal argument structure, Paris.
Martínez-Ferreiro, S., Bachrach, A., Sánchez, A., Picallo, C., 2012. Violating canonicity in span-
ish agrammatism, in: Congreso de lingüística clínica de Málaga. Málaga, 15–17 Novembre
2011.
Pesetsky, D., 1995. Zero Syntax: Experiencers and Cascades. MIT Press, Cambridge, MA.
Pustejovsky, J., 1995. The generative lexicon. MIT Press, Cambridge, MA, USA.
 Introduction 19

Pylkkänen, L., 2008. Introducing arguments. MIT Press, Cambridge, MA.

DOI: 10.7551/mitpress/9780262162548.001.0001
Ramchand, G., 2008. Verb Meaning and the Lexicon: A First Phase Syntax. March, Cambridge
Univ. Press. DOI: 10.1017/CBO9780511486319
Randall, J. H., 2010. Linking: the geometry of argument structure. Springer, Dordrecht, Nether-
lands.
Roberge, Y., Cuervo, M. C., 2012. The end of argument structure? Emerald Group Pub., Bingley.
DOI: 10.1163/9781780523774
Shetreet, E., Friedmann, N., Hadar, U., 2009. The neural correlates of linguistic distinctions:
unaccusative and unergative verbs. Journal of Cognitive Neuroscience 22, 2306–2315.
DOI: 10.1162/jocn.2009.21371
Shetreet, E., Friedmann, N., Hadar, U., 2010a. Cortical representation of verbs with optional
complements: the theoretical contribution of fMRI. Human Brain Mapping 31, 770–785.
PMID: 19890846. DOI: 10.1002/hbm.20904
Shetreet, E., Friedmann, N., Hadar, U., 2010b. The neural correlates of linguistic distinctions:
Unaccusative and unergative verbs. Journal of Cognitive Neuroscience 22, 2306–2315.
DOI: 10.1162/jocn.2009.21371
Shetreet, E., Palti, D., Friedmann, N., Hadar, U., 2007. Cortical representation of verb process-
ing in sentence comprehension: number of complements, subcategorization, and thematic
frames. Cerebral Cortex 17, 1958–1969. DOI: 10.1093/cercor/bhl105
Suihkonen, P., Comrie, B., Solovyev, V. D., 2012. Argument structure and grammatical relations
a crosslinguistic typology.
Tenny, C., 1994. Aspectual roles and the syntax-semantics interface. Kluwer Academic,
Dordrecht, Boston. DOI: 10.1007/978-94-011-1150-8
Walenski, M., 2002. Relating parsers and grammars: On the structure and real-time compre-
hension of English infinitival complements. Ph.D. thesis. UCSD.
Part I

The general issue

Verb argument structure
 Can we dance without doing a dance?
Two opposite views on the integration of roots
in the syntactic structure of the vP*

Víctor Acedo-Matellán
Centre de Lingüística Teòrica, Universitat Autònoma de Barcelona

This paper presents two confronted theories of how roots – elements that, by
definition, convey exclusively conceptual content – are integrated in the syn-
tactic structure of the vP. In the first theory, defended by Marantz (2011), roots
can only be adjuncts (modifiers), either of an abstract verbal head or of a DP
merged as the object. In the second theory, defended by Acedo-Matellán (2010,
2011a), roots can occupy an adjunct position, but they can also be comple-
ments of an abstract verbal head or of an (abstract) preposition. Empirical and
theoretical arguments are adduced to evaluate both proposals, which, however,
share the same fundamental idea that (verbal) argument structure is a syntactic
configuration.

1. Introduction: Roots in theories of argument structure

One of the most recurring issues in recent theoretical discussions on argument

structure is that of the nature of roots, that is, of the elements that, by assump-
tion, convey the purely conceptual dimension of the meaning of linguistic expres-
sions (for one of the most recent works on the subject, see Alexiadou, Borer and
Schäfer, to appear). This issue is particularly important in theories where argu-
ment structure is undistinguishable from the mere syntactic configuration of the
vP, and roots, thereby, are to be attributed status as syntactic objects. Although all
such theories assume a fundamental distinction between roots and the functional
material which builds up the structure, they do not necessarily converge on which
positions roots may occupy in that structure. Thus, while in Ramchand (2008)
roots are exclusively merged as heads, for Borer (2005) and Marantz (2011, 2013)
roots are modifiers. Moreover, in Harley (2005), Acedo-Matellán (2010) and
Acedo-Matellán and Mateu (to appear) roots can be complements or adjuncts,
but only Harley (2005, to appear) allows roots to also be complement-taking
24 Víctor Acedo-Matellán

heads. Finally, De Belder (2011), De Belder and Van Craenenbroeck (2011) and,
following them, Acedo-Matellán (2011b, 2013), propose that roots are exclusively
complements. In this paper I focus on two such theories. In particular, I present
Marantz’s (2011) theory, where roots can only be adjuncts (and, consequently,
interpreted exclusively as event modifiers), and confront it to Acedo-Matellán’s
(2010, 2011a) theory, where roots are allowed to be either adjuncts (and interpret-
ed as event modifiers) or complements.1
In Section 2, I will confront both theories within the three empirical domains
that Marantz (2011) capitalises on: re-prefixation, the interpretation of denominal
verbs, and the properties of compound verbs in English.
In Section 3, I will provide three additional reasons to allow roots in comple-
ment position. The first is of theoretical nature: if roots cannot be complements it
is not clear how to understand the structure of unergative verbs in a minimalist
syntax. The other two additional reasons are of empirical nature and involve the
complement/adjunct distinction for roots and its impact on cross-linguistic vari-
ation and on the inner-aspectual properties of the predicate.
Finally, conclusions are drawn in Section 4.

2. Marantz (2011): Roots can only be adjuncts. A critical reply

Marantz (2011) observes that some syntactic analyses of the verb phrase derive
intransitive unergative verbs like (1a) from a transitive predicate like (1a’), and
accomplishment transitives like (1b) from a small-clause predicate like (1b’):
(1) a. John danced.
a’. John did a dance.
b. John whitened his teeth.
b’. John made/brushed his teeth white.

In these analyses the verbal root (say √dance in (1a) and √white in (1b)) is
generally understood as merged in complement position, either as a verbal com-
plement (in (1a); cf. a dance in (1a’)) or as the complement in a small-clause con-
figuration (in (1b); cf. white in (1b’)), something along the lines depicted in (2):
(2) a. John [vP v √dance] (cf. John danced)
b. John [vP v [SC his teeth √white]] (cf. John whitened his teeth)

The phonological material of the roots would incorporate into the null verbal
head to yield the surface form of the predicate (see Hale & Keyser 2002; Mateu
2002; Harley 2004; Haugen 2009; Acedo-Matellán 2010; among others). Marantz
proposes that such analyses are wrong, so that predicates like (1a) are plain
 Can we dance without doing a dance? 25

intransitive (unergative) predicates and that predicates like (1b) are plain transi-
tive predicates, involving no small-clause configuration. Verbal roots are, thereby,
banned from complement position and must be analysed, in all kind of verbal
predicates, as event modifying adjuncts.
In order to support his position, Marantz (2011) wields arguments from three
empirical domains of English morphosyntax: the pattern of re-prefixation, the
non-object interpretation of roots in denominal verbs and the restrictions on the
interpretation of verbal compounds. In the next three subsections I expose and
reply to each of these arguments.

2.1 On re-prefixation

The first argument has to do with the syntax of the restitutive prefix re-, in En-
glish. Semantically, this verbal prefix adds the presupposition that there has been
some previous stage in which the entity referred to by the direct object was in the
same state as that indicated by the verb:
(3) Sally reopened the door.
Assertion: “Sally caused the door to be open.”
Pressuposition: “Before Sally opened the door there had been some previous
stage when the door was open.”

On the syntactic side, Marantz (2005, 2011) shows, first, that the prefix re- is re-
stricted to predicates with an overt object, either a transitive direct object or an
unaccusative subject, and rejects predicates without an object (examples from
Marantz 2011: 5):
(4) a. Sue recaught the cat, The door reopened, etc.
b. ??I remosked, ??I relaughed, ??I resang

The above restriction, (Horn’s 1980 generalisation), conspires with Wechsler’s

(1989) observation that re- obeys the Sole Complement Generalisation, whereby
this prefix cannot attach to verbs that require two complements (like put). As-
suming that this kind of verb selects a small clause, and adding the fact that re- is
disallowed also in resultatives and similar constructions, Marantz (2011) derives
the generalisation that re- is not allowed in any construction that overtly involves
a small clause:
(5) a. put-verbs: Sue (*re)put [SC the vase on the table].
b. low applicatives: Sue (*re)gave [SC the keys to John].2
c. change-of-location alternants of the locative alternation:
Sue (*re)sprayed [SC the paint onto the wall].
26 Víctor Acedo-Matellán

d. strong resultatives (i.e., with unselected object):

Sue (*re)drank [SC the teapot dry].
e. make-resultatives: Sue (*re)made [SC her teeth clean].

From all the facts above Marantz (2011) concludes that re- is a modifier that at-
taches directly to object DPs semantically coerced into a change-of-state inter-
pretation.
In which way does the pattern of re-syntax bear on the question whether
roots have to be precluded from complement position? For unergative predicates
like (1a), Marantz argues that, since re- cannot target them (see (4b)), this is a
proof that those predicates do not involve an underlying object. However, it seems
that, from his own perspective, the problem is not whether unergatives are under-
lyingly transitive with a root at complement position or not; rather the problem
is that unergatives do not involve a DP object, which is the category selected by
re-. As for change-of-state predicates like (1b), Marantz’s argumentation is the
opposite: since they admit re-, that means that they cannot involve a small clause
(as proposed in analyses such as (2b)), given that re- strongly disfavours small
clauses (see (5)).
Marantz (2011: 3) expands on the issue of the combination of re- with change-
of-state predicates by considering the semantic interpretation of cases in which
an adjective is added, like open the door wide or paint the barn red, that is, weak
resultative constructions in Washio’s (1997) terminology. When re- is present in
these constructions, it semantically targets the end state represented by the root
(√open, √paint), leaving out the adjective:
(6) I repainted the barn red.
Assertion: “I caused the barn to be painted red.”
Presupposition: “The barn had been in a painted state before.”

For instance, as shown in (6), if I repainted the barn red, it is the case that I caused
the barn to be in the state of being painted red, presupposing that the barn was
in the state of being painted before, but not necessarily painted red. The redness,
then, is left out of the presupossition introduced by re-. So the truth conditions of
the utterance are maintained if the barn had been blue until I repainted it:
(7) The barn has always been blue and now I am repainting it red.

Thus, the adjective of a weak resultative construction is out of the scope of re-.
In particular, re-prefixation shows that in weak resultative constructions the root
specifies the end state while the adjective merely restricts that end state. For in-
stance, the adjective red restricts the paintedness of the barn to that of “red-paint-
edness”. Marantz acknowledges this different contribution made by the adjective
 Can we dance without doing a dance? 27

and the root and proposes that they are merged in different positions (from
Marantz 2011, adapted):
(8) [vP [v v √paint] [DP [DP re- [DP the barn]] red]]

In the above representation the prefix re- attaches as a modifier to the DP the
barn. The adjective red attaches, in turn, to this DP, yielding a DP yet again. This
DP is taken as complement by v, to which is adjoined the agentive root √paint.
Although I agree that the adjective and the root make different semantic contri-
butions in these constructions, I think that it is precisely the different contribu-
tion to the semantic computation of the event which suggests that the root is a
real predicate, sitting in the complement position of a small-clause constituent,
while the adjective is merged as an adjunct above that small clause. Crucially, this
distinction is not captured by Marantz’s analysis, since he proposes that, although
the root and the adjective occupy different positions, they are both modifiers of
the event structure.
In sum, while re-prefixation shows that the root and the adjective in a weak
resultative construction do not occupy the same position, it does not show that
the root is a modifier in these constructions. The properties of re- are perfectly
compatible with a scenario where the verbal root may occupy the complement
position of a small-clause constituent, where it is interpreted as a predicate for the
inner subject, which surfaces as an object or as an unaccusative subject. What we
need to show now is that such a scenario is also compatible with Marantz’s empir-
ical claim that re- cannot appear with overt small clauses, as shown in (5). I start
off by observing that Marantz’s empirical claim in (5) can be argued to show that
re- is in complementary distribution with prepositions and with the little a head
that introduces adjectives. So I tentatively take the data in (5) to suggest that the
prefix re-, rather than being a modifier of DPs, merges exactly at the functional
head that creates small clauses, and that I call Place. In turn, Spec-Place, which is
interpreted as a Figure, is the subject of the predicative relation, that is, the surface
object/unaccusative subject:
(9) [PlaceP DP [Place’ Place DP/√]

Adjectives and prepositions are also claimed to involve a PlaceP encoding a pred-
icative relation (see Mateu 2002 and Kayne 2009, among others). In the deriva-
tion of an AP, Place is endowed with uninterpretable φ-features that overtly agree
with those of the DP at Spec-Place in some languages like French or Catalan (see
(10)). In the derivation of PPs, Place appears associated with a root that specifies
the kind of predicative relation the Figure holds with the Ground (Compl-Place).
This is shown in (11), in which an additional functional head, Path, introduces
28 Víctor Acedo-Matellán

the semantics of transition and pulls up the nearest c-commanded DP onto its
specifier (the vase), where it is interpreted as a measurer of the event:
(10) Je considère [Place/aP Marie [Place’/a’ Place/a √intelligent]]
(French; ‘I consider Marie intelligent.’)
(11) Sue put [PathP [DP the vase] [Path’ [PlaceP [DP the vase] [Place’ [Place Place √on]
[DP the table]]]]

Assuming that re- is in complementary distribution with Place as an adjec-

tive-forming head or as a preposition-forming head, there are two possible lines
of analysis for this prefix: re- is either a direct instantiation of Place (see (12)) or
it is a root which merges as an adjunt to Place (see (13)), as in the case of prepo-
sitions illustrated in (11).
(12) Sue v [PathP [the door] [Path’ [PlaceP [the door] [Place’ Place (= re-) √open]]]]
(13) Sue v [PathP [the door] [Path’ [PlaceP [the door] [Place’ [Place Place √re] √open]]]]

In both analyses the prefix re- is incompatible with an adjectival Place or a prep-
ositional Place, accounting for the data in (5), and capturing the idea that re-
only appears with small clauses headed by a root, by hypothesis, the structures
involved in the formation of one-word transitive and unaccusative accomplish-
ments entailing a caused end state (see (4a)). As for the semantics of weak resul-
tative constructions in combination with the prefix re-, if one assumes, along the
lines of Marantz (2011), that the adjective in these constructions is an adjunct
merged above PlaceP, one can explain why re- does not scope over the adjective
and why the root and the adjective are interpreted in a very different way: the
former, a real predicate; the latter, a modifier, a restrictor of the state read off from
the combination of Place and the root:
(14) Sue v [PathP [the door] [Path’ [PlaceP [aP wide] [PlaceP [the door]
[Place’ Place (= re-) √open]]]]]

It is well known that re- can appear with creation predicates, as shown below:

(15) Sue redanced the mambo.

I argue that the same structure (and, consequently, the same structural semantics)
characterises these predicates:
(16) Sue v [PathP [the mambo] [Path’ [PlaceP [the mambo]
[Place’ Place (= re-) √dance]]]]
 Can we dance without doing a dance? 29

For these cases, then, I claim that re- forces the structure and the reading of a
change-of-state event (in the example, a mambo which was previously danced is
brought back to the “danced state”).
Finally, another prediction of Marantz’s theory is also born out in this alterna-
tive theory. Namely, that re- prefixation needs some overt DP: the prefix requires a
DP since it heads a small clause which necessarily projects an inner subject. This
rules out re-prefixation to unergatives, as shown in (17):3
(17) *Sue redanced.

2.2 The interpretation of the root in unergative denominal verbs

Marantz (2011) makes the observation, basing on work by Rimell (2010), that in
denominal verbs the theme reading of the root is avoided. He illustrates this with
potential verbs like to apple or to book, pointing out that in those verbs it is most
difficult to get the reading of “eat an apple/apples” or “read a book/books”, much
as the characteristic use of apples and books is to eat them and read them, respec-
tively. He takes this as evidence that the root cannot be merged as a complement:
(18) #Mary has been booking all week. (Intended: ‘reading/writing book(s)’)
(19) #Susan appled at lunch today. (Intended: ‘ate apple(s)’)

From the empirical side, and as a first reponse to this claim, I point out that there
are actually many unergative denominal verbs where the object interpretation of
the root is pretty evident, as shown in (20) through (24):4
(20) Birthing verbs: “to bear X”
calve, piglet, foal, spawn, pup, kitten, cub, etc.
(21) Dancing verbs: “perform X”
waltz, cha-cha, salsa, tango, merengue, etc.
(22) Verbs of bodily emission: “produce X from the body”
piss, drool, sweat, etc.
(23) Verbs of ingestion: “eat/drink X”; apud Clark and Clark (1979: 780)
tea, cheeseburger, win, liquor, booze
(24) Verbs of cropping “pick up or collect X”; apud Clark and Clark 1979: 780
blackberry in the woods, nut in the woods, crab, fish, shrimp, shark, whale,
pearl, sponge
30 Víctor Acedo-Matellán

We must not forget location verbs, where the root has been argued to be inter-
preted as the complement, albeit the complement of an abstract preposition – see
Hale and Keyser (1993f.) or Mateu (2002). These are extremely abundant (exam-
ples from Clark and Clark 1979: 772):
(25) ground the planes, beach the boats, land the boat, bench the players, string
the beads, skewer the meat, cloister the nuns, pot the begonias, can the fruit,
dock the boat, station the troops, etc.

Interestingly, in some languages location verbs show a prefix of prepositional or-

igin and semantics, suggesting the existence of an underlying structure where
there really is a preposition taking the root as its complement (Catalan example
from Acedo Matellán 2006: 44 and French example from Di Sciullo 1997: 63):
(26) a. Els nois han en-caps-at els llibres.
the boys have in-box-ptcp the books
‘The boys have boxed the books.’
b. Il l’a ac-croch-é.
He it=has at-hook-ptcp
‘He has hooked it.’

Still within the empirical domain, there are predicates in some languages that
can be argued to present a non-incorporated root in complement position. Con-
sider, for instance, Basque, well-known for its overtly transitive “unergatives”
(see, for instance, Hale and Keyser 2002: 117): dantza egin, lit. “dance do”, ‘dance’;
barre egin, lit. “laugh do”, ‘laugh’; eztul egin, lit. “cough do”, ‘cough’. In French this
non-verbal component shows properties that tell of its acategorial, hence root-
like status. Thus, in predicates like avoir froid, lit. “have cold”, ‘be cold’ and avoir
faim, lit. “have hunger”, ‘be hungry’, the non-verbal element froid/faim does not
have nominal properties, in particular, case and ability to undergo nominal quan-
tification (see (27) and (28)); on the other hand, these predicates cannot host a DP
object (see (29)), which suggests that the bare nominal is occupying the comple-
ment position of transitive avoir itself:5
(27) a. *Je l’ai, le froid./*J’en ai, du froid.
I it=have the cold I=partv have of.the cold
b. *Je l’ai, la faim./ *J’en ai, de la faim.
I it=have the hunger I=partv have of the hunger
(28) a. J’ai {très/ *beaucoup de} froid.
I=have very much cold
‘I am very cold.’
(Cf. {Beaucoup de/*très} sel/chaises, “many/*very of salt/chairs”)
 Can we dance without doing a dance? 31

b. J’ai {très/ *beaucoup de} faim.

I=have very much hunger
‘I am very hungry.’
(29) a. J’ai froid (*les mains).
I=have cold the hands
b. J’ai faim (*le/ *du fromage).
I=have hunger the/ of.the cheese

From the theoretical side, the fact that the verbs apple or book do not mean “eat an
apple” or “read a book” does not imply that there is no object reading of the root.
It only implies that the roots √apple and √book have not been endowed, in their
Encyclopaedia entries, with the acceptions “ingested object” and “read object”,
respectively, when they are taken as objects to v. What I am assuming is that the
root √apple/√book refers to a conceptual scene involving apples/books and that,
merged as the complement of a little v, it is interpreted as an Incremental Theme
(see Harley 2005). Crucially, √apple or √book or any other root, for that mat-
ter, do not denote entities by themselves. So the structure [vP v √apple/√book]
would be interpreted as “do an activity related to apples or books”.
(30) [vP v √apple/√book];
“Do an activity related to apples or books.”

Interestingly, Marantz (2011) points out that the interpretation of, say, dance is
not exactly as that of do a dance, and that this has to do with the fact that uner-
gative dance in fact does not involve any underlying object, as does do a dance.
However, while I agree that these two predicates receive a different interpretation
(in event structural terms, particularly), I do not think that this is due to the fact
that do a dance involves an object and dance does not. Rather, the difference stems
precisely from the different nature of the objects involved: a DP in do a dance
and a root in dance. Crucially, it is only in the former case that the object carries
(functional) structure that is to be semantically interpreted.
The exact nature of the activity denoted by the root in Compl-v position has
to be registered within its Encyclopaedia entry (see Marantz 1995, 1997), as it is
in the case of the verb wine “drink wine”. There must surely be reasons related to
world knowledge why to wine may mean “to drink wine” while to apple does not
mean “to eat apples”. Drinking wine has consequences outside the mere ingestion
of the liquid: for instance, it is related to social interaction or to the attainment
of drunkenness. These other world-knowledge implications may have been suffi-
cient to license the ingestion reading of to wine.6
32 Víctor Acedo-Matellán

(31) Syntax: [vP v √wine]

Encyclopaedia Entry of √wine: “In the context [vP v √wine], it is to be inter-
preted as the activity of drinking wine.”

2.3 On *to truck drive

The third argument used by Marantz (2011) is the fact that English does not al-
low verbal compounds where the first component is interpreted as an object, like
*truck drive, with the interpretation “drive trucks”, while the nominal synthetic
compound truck driver is possible. On the contrary, productive formation of ver-
bal compounds in English involves a manner reading of the first component, as
evidenced by a verb like ninja walk:
(32) a. *Truck drive (i.e., “drive trucks”); cf. truck driver
b. Ninja walk (i.e., “walk like a Ninja”)

Harley (2009), who derives synthetic nominal compounds as truck driver from a
structure where a noun truck is merged as complement of the root √drive (in the
line of Lieber’s 1983 or Booij’s 2009 analyses), proposes that the absence of com-
pounds like *truck drive is to be explained in phonological terms: in English, the
light verbal head cannot host two (stress-bearing) roots. By contrast, truck driver
does not involve a v head, but an n (nominal) head, with no such phonological
constraints. Needless to say, Harley’s explanation does not account for examples
like ninja walk.
There might be syntactic reasons underlying the unavailability of compounds
like *truck drive, but not the ones that Marantz (2011) advocates. For instance,
assuming Hale and Keyser’s (2002: 44) theory, I take transitive activities to involve
a small-clause-like configuration, as shown in (33): the surface object is the in-
ner subject, and the root sits at the complement position of the small clause. The
interpretation of such a structure is that of “provide the truck with a drive”, “give
the truck a drive”:
(33) a. [v [PlaceP [the truck] [Place’ Place √drive]]
b. “Provide the truck with a drive”, “give the truck a drive”

Within this framework, in order for truck drive to be interpreted as “drive trucks”
we should have a root √truck in the specifier position of the small clause:
(34) [v [PlaceP √truck [Place’ Place √drive]]
 Can we dance without doing a dance? 33

However, roots are generally precluded from specifier position, either on syntac-
tic accounts (Hale and Keyser 1993, 2002; Mateu 2002) or on morphophonolog-
ical accounts (Acedo-Matellán 2010). For example, Hale and Keyser (1993: 63)
argue that the anomaly of (35a) is due to the fact that the nominal element spear
has been incorporated into the null verb V from its specifier position within the
embedded VP constituent [spear straight] (see (35b)):
(35) a. #He speared straight. [Intended meaning: “He straightened (the) spear”.]
b. [VP V [VP [NP spear] [V’ V [AP straight]]]]

The account could cover, with some modifications, other potential verbal com-
pounds formed out of predicates that by hypothesis require a small-clause struc-
ture, like *to pavement pound (cf. pavement pounding) or *to bug kill (cf. bug
killer).7 However, how do we obtain ninja walk and truck driver? A tentative an-
swer is that we do obtain them through adjunction. With respect to the first case,
I envisage two hypotheses. In one of them, the roots √ninja and √walk adjoin to
each other forming a complex predicate through predicate composition, as Borer
(2005: 225f.) proposes for complex resultative constructions. The complex root is,
in turn, taken as complement by a verbalising head:
(36) [vP v √ninja√walk]

This hypothesis, however, does not explain the ordering facts – i.e., why it is we
say She ninja walked and not She walk ninjaed. Rather, it seems that walk is al-
ready verbal when ninja adjoins to it. Additionally, ninja seems to be nominal,
and not an acategorial root, as suggested by the fact that it bears an entity reading
(“walk like a ninja”):
(37) [V ninjaN [V walkV]]

As for truck driver, it can be analysed as the adjunction of the noun truck to driver,
the object interpretation being inferred, not entailed. That is, from this perspec-
tive a truck-driver is a “driver of trucks” (see (38b)) rather than “someone who
drives trucks” (see (38a)):
(38) a. [N [V truckN [driveV]]-erN]
b. [N truckN [driverN]]
34 Víctor Acedo-Matellán

3. Three additional reasons to allow roots in complement position

3.1 The derivation of unergative verbs within minimalist syntax

A theory of argument structure where roots cannot be complements encounters

a non-trivial theoretical problem when deriving unergative predicates within
the minimalist theory of syntax (Chomsky 1993f.) – a theory that D(istribut-
ed) M(orphology) has always claimed to be a part of (Halle and Marantz 1993;
Embick 2010). In the minimalist framework phrasal representations are the result
of a single operation, Merge, which is responsible for building up the structure.
Although there are different versions of Merge, this operation is generally accept-
ed to take two objects to form a new syntactic object out of them. In any deriva-
tion there is always a first instance of Merge, called First Merge, which takes two
objects, α and β, out of a set of items, and creates the first phrasal projection. By
definition, First Merge creates a structure involving a head (α in (39)), which is
the projecting element, and a complement (β in (39)):
(39) First Merge (α, β) → [αP α β]

A further phrase merged with this first syntactic object will be a specifier or an
adjunct (if these two objects are to be distinguished), provided that it is still α that
projects:
(40) Merge (XP, [αP α β]) → [αP XP [α β]]

Crucially, this scenario is at odds with unergative predicates such as Sue dances,
which, at least overtly, present a specifier (the subject) and no complement:
(41) [vP [DP Sue] [dance(s)]]

However, in theories of the Halekeyserian type, where unergative verbs are un-
derlyingly transitive, and in more recent theories, where the “underlying” com-
plement of an unergative is taken to be the root, the derivation of unergative
predicates poses no problem. First Merge takes the root and the verbal head, v,
creating a first syntactic object to which further instances of Merge can then apply:
(42) [vP [DP Sue] [v √dance]]

In a theory distinguishing between roots and functional material but where roots
cannot occupy complement positions and are thus not eligible as input for First
Merge, it is not clear how unergative verbs can be derived, assuming, as DM does,
a minimalist framework.8
 Can we dance without doing a dance? 35

3.2 Cross-linguistic variation involved in the complement/adjunct

distinction

The interpretation of roots as modifiers is prohibited for certain contexts in many

languages, as is well-known in the literature (cf. Talmy 1991, 2000; Snyder 1995f.;
Mateu 2002; Acedo-Matellán 2010; among others). Consider, as an example, a
comparison between English and Catalan. The English sentences in (43) and (44)
arguably involve an interpretation dubbed “lexical subordination” by Levin and
Rapoport (1988), which is expressed by the respective paraphrases (examples
from Levin and Rapoport 1988: 275–283):
(43) a. Sally waltzed into Phillip’s arms.
“Sally ended up in Phillip’s arms by waltzing.”
b. Denise hammered the metal flat.
“Denise got the metal flat by hammering.”
c. Pauline smiled her thanks.
“Pauline expressed her thanks with a smile.”

Crucially, the Catalan rendition of these English predicates is out:

(44) a. #La Sally valsejà als braços d’en Phillip. [Non-directional reading only.]
b. #En Denise martellejà el metall pla. [Depictive reading only.]
c. *La Sue somrigué les gràcies.

On the other hand, the interpretation of the root as a complement, either the
complement of the verb or of an abstract preposition, does not seem to be pre-
cluded cross-linguistically, as far as I know. Thus, both English-type languages
and Catalan-type languages possess so-called deadjectival verbs and denominal
verbs of the unergative type and of the location/locatum type:
(45) English and Catalan; deadjectival verbs
a. Sue whitened her teeth.
b. La Sue s’emblanquí les dents.
(46) English and Catalan; denominal unergative verbs
a. The cow calved yesterday.
b. La vaca vedellà ahir.
(47) English and Catalan; denominal location verbs
a. Sue corralled the horse.
b. La Sue encorralà el cavall.
(48) English and Catalan; denominal locatum verbs
a. Sue saddled the horse.
b. La Sue ensellà el cavall.
36 Víctor Acedo-Matellán

If the modifier/complement difference is allowed to hold for roots, it can be posit-

ed that the adjunction or direct merger of a root to v is impossible in certain con-
texts for some languages, accounting for data like (44). This idea is the common
trait of some analyses of the cross-linguistic variation described in this section:
see Snyder (1995f.), McIntyre (2004), Zubizarreta and Oh (2007) or Mateu (2012),
among others. However, in a theory where universally roots can only be modifi-
ers, it is difficult to see how the cross-linguistic variation could be accounted for.9

3.3 Contribution of roots to (a)telicity and the complement/adjunct

distinction

Specifically for the distinction between a root being merged as an adjunct to v

and as a complement to v, there is evidence from (a)telicity tests that the dis-
tinction holds. First, in predicates where the root is, by hypothesis, merged as a
complement to v – that is, unergative activity predicates –, it might be interpreted
as either bounded or unbounded, giving rise to telicity or atelicity, respectively:
(49) The mare foaled in an hour. [A foal produced.]
(50) #The mare foaled for hours. [A series of foals produced.]
(51) The kid pissed in five minutes.
(52) The kid pissed for five minutes.

This (a)telicity is not grammatically represented; rather, it is regulated strictly by

world knowledge. For instance, the root √piss may be interpreted as referring to
an unbounded mass of liquid or to a bounded quantity of liquid equivalent to the
contents of the bladder, giving rise, respectively, to the readings in (51) and (52).
Since there is nothing in the structure forcing a telic or an atelic reading of the
predicate, it can be interpreted as telic or atelic depending on the construction of
the meaning of the root. That this (a)telicity does not depend on the structure is
evidenced, for instance, by the behaviour of telic birthing verbs with respect to
auxiliary selection in languages like Italian. Although a telic reading is certain-
ly possible (and pragmatically most felicitous) for intransitive foal, intransitive
birthing predicates of this type do not trigger be-selection in this language (see
(53), from Mateu 2008), while telic intransitives are otherwise be-selecting, as
shown in (54) (from Sorace 2000: 873):
(53) La giumenta {ha figliato/ *è figliata} in due ore.
the mare has foaled is foaled in two hours
‘The mare has foaled in two hours.’
 Can we dance without doing a dance? 37

(54) I bambini *hanno/sono arrivato.

the children have/are arrived

Within a Halekeyserian framework, Harley (2005) proposes, from data such as

those presented in (49) through (52), that roots can indeed be merged as comple-
ments to v, since they behave like object DPs with respect to the impact they have
on the (a)telicity of the predicate. Compare the following examples, where the
DP triggers telicity or atelicity depending on its boundedness or unboundedness,
respectively:
(55) The mare bore a foal in/*for an hour.
(56) The mare bore foals {for hours/*in an hour}.

The Compl-v position, whether occupied by a root or a DP, seems to have an im-
pact on the (a)telicity of the predicate (Harley 2005). It is only in the latter case,
where (un)boundedness is grammatically represented through DP structure, that
the predicate is unambiguously telic or atelic. Quite on the contrary, the inner
aspect of constructions where the root is, by hypothesis, merged as an adjunct to
v is orthogonal to the interpretation of the root as bounded or unbounded. For
instance, in the resultative constructions of (57) and (58) telicity depends on the
construction itself – in Acedo-Matellán’s (2010) terms, on the presence of Path
and its licensing by the quantity DP herself. Crucially, the root merged with the
verb as a modifier (interpreted as “by foaling” and “by spawning”, respectively)
cannot affect the inner-aspectual interpretation of the predicate:
(57) The mare foaled herself exhausted in/*for one and a half hour.
(58) The salmon spawned herself exhausted in/*for an hour.
(Cf. The salmon spawned for some minutes.)
(59) [vP [The salmon] [v’ [v v √spawn] [PathP [Path’ Path [PlaceP [herself]
[Place’ Place √exhausted]]]]]]

This scenario is expected if we assume that the computation of (a)telicity does not
take into account adjunct material. As a result, impact on inner aspect seems to
distinguish complement from adjunct roots.

4. Conclusions

Marantz’s (2011) proposal that roots can only be modifiers of the structure, and
not complements, is based on arguments from three empirical domains: the se-
mantic and syntactic properties of re-prefixation, the interpretation of denominal
38 Víctor Acedo-Matellán

verbs and the restrictions on verbal compounding in English. I have shown that
these arguments are not free of problems, a nontrivial one being that they are
exclusively based on English data. Theoretically, allowing roots at complement
position, which constitutes the null hypothesis, makes it possible to derive un-
ergative verbs in a minimalist framework, to which DM adheres (cf. Halle and
Marantz 1993). On the contrary, if roots are not allowed at complement position
the syntactic analysis of unergatives should be revised.
There are additional empirical reasons to maintain a complement/adjunct
distinction for roots. On the one hand, this distinction can be argued to be the
locus of notable cross-linguistic variation; on the other hand, it also seems to be
relevant to the inner-aspectual computation of the predicate.
The controversy expounded in this paper is to be framed within the more
general problem of the nature of roots and their relation to grammar, which has
received attention mainly in theoretical discussions. Hopefully interdisciplinary
approaches combining theoretical insights and psycholinguistic and neurolin-
guistic methodologies will, in the future, shed more light on this very intricate
issue of the grammar-lexicon interface.

Notes

* I would like to thank Alec Marantz for valuable and insightful discussion of the ideas and
argumentation presented in this work. M. Teresa Espinal, Jaume Mateu and Gemma Rigau
helped me with some of the data in Catalan and provided useful comments. Finally, I am grate-
ful to the organisers and audience of the Structuring the argument workshop (5–7 September
2011), where a version of this work was presented, and two anonymous reviewers, who provid-
ed valuable comments for the final version. All errors are of course my own responsibility. This
work has benefitted from project FFI2010-20634, funded by the Spanish Ministerio de Ciencia
e Innovación.

1. See Marantz (2013) for theoretical arguments in favour of the position that roots are mod-
ifiers of the little head categorisers, like v.

2. An anonymous reviewer points out that the sentence Mary resent the application to Paul. is
fine and that re- scopes over the PP to Paul. If this is true, the sentence would be a counterex-
ample for Marantz’s empirical claims, since re- is here compatible with a predicate involving a
small-clause configuration (send). It would also constitute a counterexample for the alternative
theory presented here, which is based on the complementary distribution of re- and preposi-
tional and adjectival heads. As an aside, I note that Marantz’s proposal that re- cannot appear
with small-clause predicates contradicts his own claims on the structure of some morpholog-
ically complex predicates in English. In a nutshell, Marantz (2003, 2005) has argued, on em-
pirical and theoretical grounds, that prefixed verbs like con-struct or de-stroy are syntactically
complex in that the prefix they feature is an underlying predicate for the surface direct object,
the root being an adjunct to the little v head (example from Marantz 2003: 4, adapted):
 Can we dance without doing a dance? 39

(i) [vP [v v √stroy] [SC [the city] de-]]: Destroy the city.
However, if Marantz’s analysis for verbs like construct is correct, his assumptions on the selec-
tional properties of re- should not be, since this prefix attaches unproblematically to construct
or destroy (examples from Google search, June 2012):
(ii) Another architectural contest was held, and the winner, Paul Baumgarten, recon-
structed the building from 1961–1964.
(iii) If the universe recreates and redestroys itself over and over, I think there is at least a
possibility that I may have existed many times before […].

3. For other recent treatments of re-prefixation see Harley (2004), and Săvescu Ciucivara &
Wood (2013).

4. Some examples from Google search (February 2012):

(i) “We teaed in the garden and then I worked […].”

(ii) “We had arranged to go crabbing on the beach with the bunch this morning […].”
(Winans, Fonville. 2011. Cruise of the Pintail. A Journal. Louisiana State University
Press, p. 36.)
(iii) “A pause while we liquored.” (Simms, William Gilmore. 1996. Tales of the South.
Columbia, South Carolina: University of South Carolina, p. 323)
(iv) “First we dined; and then after we dined, we wined; and then after we wined, everybody
seemed to have something to say.” (Western Machinery World, vol. 6)

5. As an anonymous reviewer points out, cognate languages like Catalan or Spanish feature
similar predicates that, however, do not present the same behaviour. Thus, in Catalan Tenir
fred ‘Be cold’, lit. “Have cold”, the nominal segment bears case, as proved by its pronominalisa-
tion: Tenir-ne, lit. “Have of it”. The data show that in Catalan and Spanish these nouns are not
bare roots, but nominals with some (minimal) functional structure. However, it is possible that
other light-verb predicates in Catalan, built on fer ‘do’, involve simple roots, like Els tovallons
de colors fan festa ‘Coloured napkins create a party-like atmosphere’, lit. “Coloured napkins do
party”. In these cases the bare noun does not seem to bear case: *Els tovallons de colors en fan,
de festa, lit. “Coloured napkins of it do, of party”. Additionally, these fer-constructions are not
picky about the category of their non-verbal constituent, suggesting that, in fact, this does not
have a category: fer bonic ‘be nice’, lit. “do nice”, fer tard ‘be late’, lit. “do late”. I leave this inter-
esting issue for future research. For recent discussion on bare singular nominals in Catalan and
Spanish see Espinal (2010), Espinal and Mateu (2011) or Espinal and McNally (2011).

6. An anonymous reviewer suggests that wine and tea – and (s)he adds fish-and-chip – are
possible in the ingestion sense since the corresponding root may have an event reading, unlike
apple or book. Clearly this is related to world-knowledge: while there is an established event of
drinking tea, which is even associated with a particular time, there is no established event of
eating apples.

7. As a matter of fact, we find some examples of N-V compounds with an object reading of N
(examples from Google search, July 2012):
40 Víctor Acedo-Matellán

(i) “They salsa danced the whole way, stopping in pubs and shops to request donations
for the victims of the hurricane.”
(ii) “The packed room was full of hunks, drunks and lunks getting ready to pop cham-
pagne and blow their own horns as they polka danced in the New Year.”
(iii) “[…] they salmon fish from 10–15 hours a day between May and October.”
(iv) “Mike (the owner of the shop) generally can tell you what’s been working best as he
cod fishes fairly often […].”
(v) They like tight places like that, evidently. And the welp itches like mad! I don’t get too
many, but, my Dad gets eaten up with them, if he berry picks.”
Interestingly, in most of these cases there is a hyponimic relation between the noun and the
verb: for instance, a polka is a dance, or a salmon is a fish. These examples could actually be
predicted by the present account, since the left-hand N member of the compound could be
argued to originate in complement position. In a nutshell, for hyponimic object contructions of
the kind of dance polka, we assume a basic [V √root] structure, following a trend originating
in Hale and Keyser (1993f.). Assuming Haugen’s (2009) theory of cognate object predicates, the
root incorporates into the verbal head, leaving a copy at the base position. At Vocabulary In-
sertion two different Vocabulary Items are inserted at the head and the tail of the chain: [dance
polka]; in the case of verbal compounds, the object Vocabulary Item (polka) moves further
and becomes the left-hand member of the N-V compound. I leave this intriguing prediction
for a future occasion. The reader is referred to Lieber (2009) for more discussion on this kind
of compounds.

8. See De Belder (2011) and De Belder and Van Craenenbroeck (2011) for a thorough dis-
cussion on the syntax of roots and for the conclusion that roots must in fact always be com-
plements. See Acedo-Matellán (2011b, 2013) for the application of this theory of roots to the
difference between conflation and incorporation processes.

9. Note that the claim here is not that roots cannot be modifiers in certain languages, but that
while the complement position seems to be cross-linguistically available without restrictions,
the modifier position presents restrictions in certain languages. As a matter of fact, in languages
of the Catalan-type certain existential predicates allow the interpretation of roots as modifiers
(Catalan example from Mateu 2002: 188):
(i) En aquesta coral n’hi canten molts, de nens.
in this choir of_them=there sing.3pl many of children
‘There are many children who sing in this choir.’
In the above predicate the verb canten does not license its prototypical activity interpretation,
and, in fact, the predicate shows unaccusative behaviour, as evidenced by ne-extraction. Rather,
the interpretation of the whole sentence is existential, the root √cant being limited to labelling,
as a modifier, the (static) relationship asserted between the choir and the children. See Rigau
(1997), Mateu (2002) and Borer (2005) for relevant discussion in Catalan. See Centineo (1996)
and Mendikoetxea (1999) for similar facts and discussion thereof in Italian and Spanish, re-
spectively.
 Can we dance without doing a dance? 41

References

Acedo-Matellán, Víctor. 2006. “Una aproximació sintàctica als verbs prefixats en català”. Estu-
dios Catalanes 4: 41–78.
Acedo-Matellán, Víctor. 2010. Argument Structure and the Syntax-Morphology Interface. A
Case Study in Latin and other Languages. Barcelona: Universitat de Barcelona dissertation.
http://filcat.uab.cat/clt/publicacions/tesis/pdf/AcedoMatellan2010PhDDissertation.pdf
Acedo-Matellán, Víctor. 2011a. “A comment on Marantz’s “Syntactic approaches to argument
structure without incorporation””. Presented at the “Structuring the Argument: A multidis-
ciplinary workshop on the mental representation of verbal argument structure”. Structures
Formelles du Langage – UMR 7023 Paris 8/CNRS, Paris, 5–7 september 2011.
Acedo-Matellán, Víctor. 2011b. “A layered-derivation approach to the conflation/incorporation
distinction”. Presented at the “Workshop on verbal elasticity. Framing the verb/satellite dis-
tinction from a biolinguistic perspective”. Centre de Lingüística Teòrica, Universitat Autòno-
ma de Barcelona, Bellaterra, 3–5 october 2011.
Acedo-Matellán, Víctor. 2013. “Merging Roots in Bare Phrase Structure and the Conflation/
Incorporation Distinction”. NELS 42(1): 1–14.
Acedo-Matellán, Víctor and Jaume Mateu. To appear. “From syntax to roots: a syntactic ap-
proach to root interpretation”. In The syntax of roots and the roots of syntax, Artemis
Alexiadou, Hagit Borer and Florian Schäfer (eds.). Oxford: Oxford University Press.
Alexiadou, Artemis, Hagit Borer and Florian Schäfer (eds.). To appear. The syntax of roots and
the roots of syntax. Oxford: Oxford University Press.
Booij, Geert. 2009. “Construction morphology and compounding”. In The Oxford Handbook of
Compounding, Rochelle Lieber and Pavol Štekauer (eds.), 201–216. Oxford: Oxford Uni-
versity Press.
Borer, Hagit. 2005. Structuring Sense, vol. 2: The normal course of events. Oxford: Oxford Uni-
versity Press. DOI: 10.1093/acprof:oso/9780199263929.001.0001
Centineo, Giulia. 1996. “A lexical theory of auxiliary selection in Italian”. Probus 8(3): 223–271.
DOI: 10.1515/prbs.1996.8.3.223
Chomsky, Noam. 1993. “A minimalist program for linguistic theory”. In The View from Build-
ing 20: Essays in Linguistics in Honor of Sylvain Bromberger, Kenneth Hale and Samuel J.
Keyser (eds.), 1–52. Cambridge, MA: MIT Press.
Clark, Eve V. and Herbert H. Clark. 1979. “When nouns surface as verbs”. Language 55(4):
767–810. DOI: 10.2307/412745
De Belder, Marijke. 2011. Roots and Affixes. Eliminating Lexical Categories from Syntax. Utrecht:
Utrecht University/Uil-OTS and HUBrussel/CRISSP PhD thesis. http://ling.auf.net/ling-
Buzz/001280
De Belder, Marijke and Jeroen van Craenenbroeck. 2011. “How to merge a root” (ms.).
CRISSP/HUB/UiL-OTS/Utrecht and CRISSP/HUB/FUSL/KUL. http://ling.auf.net/ling-
Buzz/001226
Di Sciullo, Anna-Maria. 1997. “Prefixed verbs and adjunct-identification”. In Projections and
Interface Conditions. Essays on Modularity, Anna-Maria Di Sciullo (ed.), 52–74. New York/
Oxford: Oxford University Press.
Embick, David. 2010. Localism versus globalism in morphology and phonology. Cambridge, MA:
MIT Press. DOI: 10.7551/mitpress/9780262014229.001.0001
Espinal, M. Teresa. 2010. “Bare nominals in Catalan and Spanish. Their structure and meaning”.
Lingua 120(4): 984–1009. DOI: 10.1016/j.lingua.2009.06.002
42 Víctor Acedo-Matellán

Espinal, M. Teresa and Jaume Mateu. 2011. “Bare nominals and argument structure in Catalan
and Spanish”. The Linguistic Review 28: 1–39. DOI: 10.1515/tlir.2011.001
Espinal, M. Teresa and Louise McNally. 2011. “Bare singular nominals and incorporating verbs in
Spanish and Catalan”. Journal of Linguistics 47: 87–128. DOI: 10.1017/S0022226710000228
Hale, Kenneth and Samuel J. Keyser. 1993. “On argument structure and the lexical expression
of syntactic relations”. In The View from Building 20: Essays in Linguistics in Honor of Syl-
vain Bromberger, Kenneth Hale and Samuel J. Keyser (eds.), 53–109. Cambridge, MA: MIT
Press.
Hale, Kenneth and Samuel J. Keyser. 2002. Prolegomenon to a Theory of Argument Structure.
Cambridge, MA: MIT Press.
Halle, Morris and Alec Marantz. 1993. “Distributed Morphology and the pieces of inflection”.
In The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, Hale,
Kenneth and Samuel J. Keyser (eds.), 111–176. Cambridge, MA: MIT Press.
Harley, Heidi. 2004. “Merge, conflation, and head movement: The First Sister Principle revisit-
ed”. NELS 34(1): 239–254.
Harley, Heidi. 2005. “How do verbs get their names? Denominal verbs, Manner incorporation
and the ontology of verb roots in English”. In The Syntax of Aspect. Deriving Thematic and
Aspectual Interpretation, Nomi Erteschik-Shir and Tova Rapoport (eds.), 42–64. Oxford:
Oxford University Press.
Harley, Heidi. 2009. “Compounding in Distributed Morphology”. In Oxford Handbook of Com-
pounding, Rochelle Lieber and Pavol Stekauer (eds.), 129–144. Oxford: Oxford University
Press.
Harley, Heidi. To appear. “On the identity of roots”. Theoretical Linguistics. http://ling.auf.net/
lingBuzz/001527
Haugen, Jason D. 2009. “Hyponymous objects and Late Insertion”. Lingua 119(2): 242–262.
DOI: 10.1016/j.lingua.2007.10.008
Horn, Laurence R. 1980. “Affixation and the Unaccusative Hypothesis”. CLS 16: 134–146.
Kayne, Richard. 2009. “Antisymmetry and the Lexicon”. Linguistic Variation Yearbook 8(1):
1–31. DOI: 10.1075/livy.8.01kay
Levin, Beth and Tova R. Rapoport. 1988. “Lexical Subordination”. CLS 24: 275–289.
Lieber, Rochelle. 1983. “Argument Linking and Compounds in English”. Linguistic Inquiry 14:
251–285.
Lieber, Rochelle. 2009. “A Lexical Semantic Approach to Compounding”. In The Oxford Hand-
book of Compounding, Rochelle Lieber and Pavol Štekauer (eds.), 78–104. Oxford: Oxford
University Press.
Marantz, Alec. 1995. “Cat as a phrasal idiom” (ms.). Cambridge, MA: Massachusetts Institute
of Technology.
Marantz, Alec. 1997. “No escape from syntax: Don’t try morphological analysis in the privacy
of your own lexicon”. Penn Linguistics Colloquium (PLC) 21: 201–225.
Marantz, Alec. 2003. “Subjects and objects” (ms.). Cambridge, MA: Massachusetts Institute of
Technology.
Marantz, Alec. 2005. “Objects out of the lexicon: Objects as events” (ms.). Cambridge, MA:
Massachusetts Institute of Technology.
Marantz, Alec. 2011. “Syntactic approaches to argument structure without incorporation: Do-
ing the (anti-lexicalist) dance without doing the dance”. Presented at the “Structuring the
Argument: A multidisciplinary workshop on the mental representation of verbal argument
structure”. Structures Formelles du Langage – UMR 7023 Paris 8/CNRS, Paris, 5–7 Septem-
ber 2011.
 Can we dance without doing a dance? 43

Marantz, Alec. 2013. “Locality domains for contextual allomorphy across the interfaces”. In
Distributed Morphology Today. Morphemes for Morris Halle, Alec Marantz and Ora
Matushansky (eds.), 95–115. Cambridge, MA: MIT Press.
Mateu, Jaume. 2002. Argument Structure. Relational Construal at the Syntax-Semantics Inter-
face. Bellaterra: Universitat Autònoma de Barcelona dissertation. http://seneca.uab.cat/clt/
publicacions/tesis/index.html.
Mateu, Jaume. 2008. “Argument structure and denominal verbs” (ms.). Bellaterra: Centre de
Lingüística Teòrica-Universitat Autònoma de Barcelona.
Mateu, Jaume. 2012. “Conflation and incorporation processes in resultative constructions”. In
Telicity, Change, and State: A Cross-Categorial View of Event Structure, Violeta Demonte
and Louise McNally (eds.), 252–278. Oxford: Oxford University Press.
DOI: 10.1093/acprof:oso/9780199693498.003.0010
McIntyre, Andrew. 2004. “Event paths, conflation, argument structure, and VP shells”. Linguis-
tics 42(3): 523–71. DOI: 10.1515/ling.2004.018
Mendikoetxea, Amaya. 1999. “Construcciones inacusativas y pasivas”. In Gramática Descripti-
va de la Lengua Española, vol. 2: Las Construccions Sintácticas Fundamentales. Relaciones
Temporales, Aspectuales y Modales, Ignacio Bosque and Violeta Demonte (eds.), 1575–
1629. Madrid: Real Academia Española and Espasa Calpe.
Ramchand, Gillian C. 2008. Verb meaning and the lexicon. A first phase syntax. Cambridge:
Cambridge University Press. DOI: 10.1017/CBO9780511486319
Rigau, Gemma. 1997. “Locative sentences and related constructions in Catalan: ésser/haver al-
ternation”. In Theoretical Issues at the Morphology-Syntax Interface, Amaya Mendikoetxea
and Myriam Uribe-Etxebarria (eds.), 395–421. Bilbao and Donostia-San Sebastián: Uni-
versidad del País Vasco and Diputación Foral de Gipuzkoa.
Rimell, Laura. 2010. “Event structure and verb meaning possibilities for denominal verbs”. Pre-
sented at the “Workshop on the Lexicon-Syntax Interface: Verb Meaning, Event Semantics
and Argument Structure”, Universitat Autònoma de Barcelona, Bellaterra, 2–3 December
2010.
Savescu Ciucivara, Oana and Jim Wood. 2013. “Re-prefixation and Talmy’s parameter”. NELS
41(1): 193–204.
Sorace, Antonella. 2000. “Gradients in auxiliary selection with intransitive verbs”. Language
76(4): 859–890. DOI: 10.2307/417202
Snyder, William. 1995. Language Acquisition and Language Variation: The Role of Morphology.
Cambridge, MA: Massachusetts Institute of Technology PhD thesis.
Talmy, Leonard. 1991. “Path to realization: A typology of event conflation”. Berkeley Linguistic
Society (BLS) 17: 480-519.
Talmy, Leonard. 2000. Toward a Cognitive Semantics, vol. 2: Typology and Process in Concept
Structuring. Cambridge, MA: MIT Press.
Washio, Ryuichi. 1997. “Resultatives, compositionality and language variation”. Journal of East
Asian Linguistics 6: 1–49. DOI: 10.1023/A:1008257704110
Wechsler, Stephen. 1989. “Accomplishments and the prefix re-”. NELS 19: 419–434.
Zubizarreta, María Luisa and Eunjeong Oh. 2007. On the Syntactic Composition of Manner and
Motion. Cambridge, MA: MIT Press.
 Determining argument structure
in sign languages

Carlo Geraci and Josep Quer

CNRS / ICREA-Universitat Pompeu Fabra

In this paper we offer an overview of existing analyses of argument structure

that sets the stage for further inquiry into this domain. The particular structure
of the lexicon in sign languages (SLs) is introduced, with special attention to the
agreement patterns found in lexical predicates, as overt agreement marking in
the set of verbs that can realize it offers a window into verb meaning and overt
argument realization. Classifier predicates, on the other hand, have proven to be
a very rich domain for research on argument structure: unaccusative/unergative
and unaccusative/transitive alternations have been identified in American Sign
Language (ASL) classifier constructions, and replicated in other SLs. As expect-
ed, the validity of valency tests is sometimes limited to one language, but the
alternations are attested crosslinguistically and can be applied to lexical verbs as
well. Specially interesting is the traditional divide between agreement marking
in lexical predicates and spatial agreement marking in classifier constructions,
often seen as having a different nature. Given the fact that the morphological
exponence of agreement is superficially the same (i.e. the path or trajectory that
the verbal sign crosses in signing space), the divide must be motivated on em-
pirical arguments, which are not always compatible or consistent with a broad
empirical coverage. We identify a number of areas where research should be
carried out in order to advance our ounderstanding of argument structure in
languages in the visual-gestural modality, in order to determine which of the
observed properties is really modality-specific.

1. Introduction1

As natural languages, sign languages should contribute fresh empirical evidence

to the theoretical discussion between lexicalist vs. neoconstructionist approach-
es about how verb meanings and argument structure are determined. However,
research in this domain is still rather scarce within Sign Linguistics. This is in
part due to the still limited available research on particular sign languages (SLs),
46 Carlo Geraci and Josep Quer

but also to the fact that some core properties of the lexicon of SLs are still only
partially understood. We start by offering in Section 2 an overview of the general
structure of the lexicon in SLs, and by characterizing the predicate classes that
are traditionally recognized by their agreement and semantic properties. In Sec-
tion 3 we briefly review two accounts of argument encoding within a Distributed
Morphology approach. Section 4 reports on the research carried out on argument
structure of classifier predicates. Section 5 discusses some of the open questions
that the existing research brings to the fore and points to a number of future
avenues of research that SL research offers for our further understanding of argu-
ment structure issues in natural language.

2. The lexicon of SLs: A brief overview

The nature of the lexicon in SLs is quite composite and differs in many respects
from that of spoken languages. In this section, we illustrate some of the main
properties which will be crucial for the discussion of the coming sections. Specif-
ically, we focus on two classes of predicates, the class of lexical verbs/predicates
and the class of spatial and classifier predicates, the latter representing a special
form of predication in SLs (Sandler and Lillo-Martin 2006). Extending Brentari
and Padden’s (2001) proposal for ASL to the general case of SLs, we consider these
two classes of predicates as being part of the core/native portion of SL lexicon,
while non-native signs are forms imported either from other SLs or from spoken
languages (as loan translations from the ambient spoken language, via the man-
ual alphabet as fingerspelled forms, or as forms whose handshape incorporates
manual alphabet letters of the word in the lexicon of the spoken language, most
frequently the initial letter of the corresponding spoken word).

2.1 Lexical predicates

The class of lexical predicates is traditionally divided between agreeing and plain
(i.e. non-agreeing) predicates, depending on whether or not they allow for per-
son (and number) agreement to be morphologically realized (Padden 1990). In a
nutshell, agreement morphology exploits two articulatory properties of sign lan-
guages, namely that discourse referents can be localized in signing space and that
the movement component of a sign may take trajectories between two referential
locations already established in signing space. Thus, in a sentence from Italian
Sign Language (LIS) like (1), the locations in signing space for the two referents
of the NPs (GIANNI and MARIA) are linked by the trajectory of the movement of
 Determining argument structure in sign languages 47

the predicate SHOOT (as indicated by the indices below the glosses of the exam-
ple). The direction of the movement component of the predicate (from the posi-
tion where GIANNI is localized to the one where MARIA is) marks the syntactic
functions of subject and object, respectively. Therefore, person agreement in SL
manifests itself as sharing the same spatial location between verb arguments and
the start and ending points of its movement, creating a “path” between the two
arguments (see Sandler and Lillo-Martin 2006 for a more detailed description).2

(1) GIANNI1 MARIA2 1SHOOT2

‘Gianni shot Maria.’

As for three-argument predicates, agreement morphology generally marks sub-

ject and indirect object, leaving aside marking the direct object, as illustrated by
the example in (2). In this case the verb DONATE in LIS starts from the subject
position and ends in the position in the space where the indirect object is located,
as indicated by the indices (crucially, in this example there is no overt agreement
with the direct object).
(2) GIANNI1 MARIA2 RING3 1DONATE2
‘Gianni donated a ring to Maria.’

However, some three-argument predicates display an additional morphological

process of agreement affecting the verb and its direct object. This process is only
found with classifier predicates and will be briefly introduced below.
Due to their morphophonological specifications, which prevent them from
realizing path movement, a number of lexical predicates known as plain verbs
does not show overt agreement morphology. One of the most interesting strate-
gies that SLs may adopt to bypass the impossibility of showing agreement mor-
phology is that of employing an auxiliary sign to mark person agreement. This
option is attested in Catalan SL (LSC) and German SL (DGS), for instance. The
crucial property shared among the auxiliaries in these SLs is that of having a path
movement, linking subjects and objects.3 The example in (3) from LSC involves
the predicate LOVE, which is produced on the torso and does not have path
movement. The person auxiliary links the subject and object positions, thus sup-
plementing the lexical predicate with agreement information.
48 Carlo Geraci and Josep Quer

(3) 3AUX1 LOVE

‘S/he loves me.’

2.2 Spatial and classifier predicates

Spatial predicates agree with their locative arguments or adjuncts. The morpho-
logical device these predicates employ is the same as the one found with agree-
ing predicates, namely the trajectory of the movement component of the motion
predicate targets the locative arguments as their initial and final points, if both are
realized (alternatively, only one of them). Otherwise, when the predicate express-
es location, it is articulated where the relevant argument has been localized. How-
ever, Padden (1988 [1983]) shows that in ASL spatial predicates obey different
constraints in the case of subject agreement and also in the referential value of the
locus of agreement. Specifically, locations identified by spatial predicates are to be
maintained more rigidly than those established by agreeing predicates because in
the former, geometrical distance among signs maps the physical distance among
participants to the event in the real world. To illustrate, consider the contrast in
LIS between DONATE (analyzed as a lexical verb) and PUT (analyzed as a spatial
predicate), shown in (4) and (5). While the two versions of DONATE do not have
differences in meaning, the two versions of PUT do have different interpretations.
Specifically, the locations and the geometric distance in the (plural) object agree-
ment mark two different arrangements of the objects only in the examples in (5).

(4) a. 0 DONATE i, exhaustive b. 0DONATEi, exhaustive

(5) a. i PUT j PUT kPUT b. iPUT jPUT lPUT
 Determining argument structure in sign languages 49

Let us now turn to classifier (CL) predicates. These predicates are commonly an-
alyzed as morphologically complex signs (opposed to lexical signs, which are for
the majority of the cases monosyllabic and monomorphemic; cf. Brentari 1998),
with a strong iconic component. Morphological complexity relies on the fact that
each of the articulatory components (movement, location and handshape) of a
CL predicate is endowed with meaning. The movement component is taken to be
the ‘root’ of the complex sign indicating the type of event (e.g. motion, manner,
position, or extension in Engberg-Pedersen’s 1993 typology). Usually, locations
identify specific loci in the signing space as a topographic projection of the po-
sitions of the arguments’ referents in the real world (more or less like in spatial
predicates). Handshapes are the real classifiers in that they classify over entities
according to various properties (e.g. whole entity, handling, extension and sur-
face, or body part in Engberg-Pedersen’s 1993 typology). The example in (6) from
LIS illustrates the case of whole entity classifiers. The flat B [ ] handshape indi-
cates the class of vehicles in LIS, while the movement component indicates the
trajectory followed by the car in its passage.

(6) CAR Clvehicle: drive-past
‘The car drove past.’

While cases like (6) may lead us to assimilate CL predicates to other spatial pred-
icates, other cases point toward a stricter connection with agreement in lexical
predicates. This seems to be the case of the predicate GIVE in DGS. In the exam-
ples in (7), adapted from Glück and Pfau (1999), the trajectories linking the NPs
CHILD and MOTHER are reminiscent of standard cases of agreement (modulo
the discussion of examples (4) and (5) above). However, the classifier nature of the
predicate GIVE is clear from the handshape alternation. In (7a) the handshape of
the predicate classifies over thin objects (like the straw of a flower), while in (7b)
it classifies over round objects (like apples).
(7) a. CHILD1 MOTHER2 FLOWER Clthin-object-1GIVE2
‘The child gives a flower to the mother.’
b. CHILD1 MOTHER2 APPLE Clround-object-1GIVE2
‘The child gives an apple to the mother.’
50 Carlo Geraci and Josep Quer

Although the phonological realization of this process is intrinsically different

from the cases of agreement already discussed, Glück and Pfau (1999) offered an
analysis of handshape alternation in terms of agreement, arguing that what hap-
pens to the predicate stem is actually a phonological manifestation of the mor-
phological process of (object) agreement (see Zwitserlood 2003b for an analysis in
terms of gender agreement and Section 3 below for a formal approach).4
The class of lexical predicates and that of spatial and classifier predicates may
show different properties especially when it turns to the referential value of the
spatial locations of their arguments (see Padden 1990); still, they share the general
morphological mechanism for agreement marking. In this sense, it is not imme-
diately obvious how to draw neat lines between lexical predicates, spatial predi-
cates, frozen (almost lexical) forms of CL and productive CL predicates. Rather, it
seems that there is a somewhat more gradual continuum between classifier forms
and lexical forms, as suggested in works by Glück and Pfau (1999), Brentari and
Padden (2001) and Zwitserlood (2003a).

3. Distributed Morphology accounts

Although there may be lexical idiosyncrasies, language specific constraints and

strategies in the grammatical use of spatial resources to mark agreement (like the
use of person auxiliaries or that of non-manual components), the striking aspect
of the agreement system illustrated in Section 2 is that it is widely shared across
SLs, i.e. almost all SLs studied up to now seem to conform to the classification of
the predicate system into agreeing, plain, spatial and classifier predicates (Sandler
and Lillo-Martin 2006). In this section, we present an overview of how the basic
facts of agreement can be derived within the framework of Distributed Morphol-
ogy (Halle and Marantz 1993; Harley and Noyer 1999; and Embick and Noyer
2007) by reviewing Glück and Pfau’s (1999) and Zwitserlood’s (2003a) proposals
for DGS and NGT, respectively. Although the two analyses ground their empirical
base on two languages only, they can be extended, as far as we can see, to simi-
lar facts in other SLs. For concreteness, we illustrate the case of three-argument
predicates with full agreement pattern, which according to Glück and Pfau (1999)
belong to the class of CL predicates. Direct object agreement is expressed via
handshape alternation (see the examples in (7) above), while spatial linking of the
starting and end point of the path movement of the predicate marks subject and
indirect object agreement. This is the case of the classifier version of the predicate
GIVE in DGS, LIS, LSC and many other sign languages. Glück and Pfau (1999)
identify three syntactic projections where the agreement systems split, namely
 Determining argument structure in sign languages 51

the verb phrase, the aspectual phrase and the tense phrase, respectively hosting
direct object, indirect object and subject agreement. The abstract (X’-)schema in
(8) summarizes the proposal.
(8) Template for agreement system in SL
Tns
qp
Tns Asp
3 qp
Tns AgrS V Asp
3 3
V AgrDO AgrIO Asp

Agreement heads are inserted in the Morphological Structure component of the

derivation, after the relevant syntactic structure has been spelled-out. Ideally, the
order of head affixation reflects the sequential requirement imposed by the path
trajectory of the predicate, resulting in the alignment of subject agreement with
the starting point and indirect object agreement with the end point of the trajec-
tory. Handshape alternations in CL predicates, as illustrated in the examples in (7)
above, are realized at the lower level of the structure (within the VP).
Working in the same framework as Glück and Pfau, and basing her account
on feature competition, Zwitserlood (2003a) proposes an even stronger mapping
between the order of merger and the realization of the agreement patterns. Fea-
ture competition manifests itself in terms of precedence. Roughly, highly speci-
fied features are to be morphologically (and hence phonologically) realized in the
lower nodes of the structure as opposed to less specified features. Each argument
comes with two types of features: a gender feature and a location feature; howev-
er, only one may be realized overtly. Zwitserlood’s analysis is based on the idea
that handshape alternation is the manifestation of gender agreement5 in SL and
on the following two assumptions: (i) gender features are more highly specified
than location features (i.e. they take precedence in terms of morphosyntactic and
morphophonological realization); (ii) handshape change is generally not allowed
in CL predicates.6 The fact that direct object agreement always shows up as hand-
shape alternation in ditransitive predicates and that indirect objects and subjects
entertain location agreement is then easily derivable from the merge order of the
predicate’s argument. At the lower level of the structure, predicates entertain an
agreement relation with its direct object, resulting in gender agreement (realized
as handshape alternation). At the higher levels only location agreement may show
up, since by assumption handshape alternation cannot be realized as an addition-
al handshape.
52 Carlo Geraci and Josep Quer

4. Classifier predicates and argument structure

The first attempt at identifying argument alternations across verbs classes was
undertaken in Benedicto and Brentari (2004) for ASL.7 Their study concentrates
on classifier predicates, but the tests they developed to identify valency are argued
to extend to lexical verbs. The correlations they establish between classifier pred-
icates and their argument structure are as follows:
Predicates with a handling classifier are transitive (with an external and
(9) i.
an internal argument);
ii. Predicates with a whole entity classifier8 are intransitive unaccusative
(one single internal argument) and
iii. Predicates with a body part classifier are intransitive unergative (one
single external argument).

The main empirical arguments for this correlations are based on a number of tests
targeting internal and external arguments in ASL: the grammatical combination
of a predicate with the distributive morpheme [dist] and the negation NOTHING
signal the presence of an internal argument (either the object of a transitive or the
subject of an unaccusative predicate); at the same time, the possibility to combine
a predicate with the negative imperative FINISH! or with an agent oriented ad-
verb like WILLINGLY are taken to indicate that there is an agent in the structure,
and consequently an external argument. The argument structure alternations that
Benedicto and Brentari (2004) identify for ASL are the following ones:
(10) a. unergative/unaccusative alternation: body part CL / whole entity CL
b. transitive/intransitive alternation: handling CL / whole entity CL

Examples of each type of alternation can be found in (11) and the different choice
of handshape results in a different type of predicate (unergative vs. unaccusative
in (11); transitive vs. intransitive unaccusative in (12)):
(11) a. ROSIE S9+BOW
Rosie headBPCL+bow
‘Rosie bowed.’
b. ROSIE 1+BOW
Rosie upright_beingw/e+bow
‘Rosie bowed.’
(12) a. [ø] BOOK C+MOVE
pron.3sg book obj_grabhdlg+move_vert.>hor.
‘S/he took the (standing) book and laid it down on its side.’
 Determining argument structure in sign languages 53

b. BOOK B+MOVE
book 2D_flat_objw/e+move_vert.>hor.
‘The (standing) book fell down on its side.’

As an illustration of the type of evidence for the alternations, consider the min-
imal pair in (13): applying the negative imperative to related classifier construc-
tions of the handling type (13a) and the whole entity type (13b) gives opposite
results: sentence (13a) is grammatical because it involves a transitive construction
with an agent that can be the target of a negative command; in contrast, (13b) is
ungrammatical because it features a whole entity classifier predicate, with a sin-
gle internal theme argument. This minimal pair shows the same reaction to the
WILLINGLY test.
(13) a. [ø] BOOK C+move FINISH
pro book obj_grabhdlg+move_vert>hor STOP_IMPER
‘Stop putting the book down on its side!’
b. *BOOK B+move FINISH
book 2D_flat_objw/e+move_vert>hor STOP_IMPER
#‘Book, stop falling on your side!’ (Benedicto and Brentari 2004: 771–772)

Moreover, when the handling classifier (13a) and whole entity classifier (13b) are
combined with the distributive morpheme and the negative marker NOTHING,
both have scope over the internal argument in each case, and never over an exter-
nal one.10 In contrast, body part classifiers in ASL are argued to react negatively
to the same tests, as they are claimed to realize intransitives lacking an internal
argument, that is unergative predicates. For this very reason they yield a gram-
matical result when the negative imperative and the WILLINGLY test are applied
to them.
Crosslinguistic work following the same kind of approach confirmed its gen-
erality, but it also raised additional questions. Benedicto et al. (2007) applied the
same strategies to classifier predicates in Argentinian SL (LSA) and Catalan SL
(LSC) and concluded that hypotheses (i) and (ii) formulated for handling and
whole entity classifier predicates in (9) above are confirmed for those two lan-
guages as well, thus supporting their crosslinguistic validity. Hypothesis (iii) was
not confirmed, though. This study suggests that the inconclusiveness of the re-
sults for body part classifiers might be due to the language-specific character of
the tests used. However, in their reanalysis of this type of classifier predicates in
ASL, Grose et al. (2007) argue that they should be analyzed as transitive predi-
cates where the classifier handshape stands for the body part of the external argu-
ment. In addition, when instrumental classifiers are incorporated to the picture
new questions arise, as Grose et al. (2007) and de Lint (2010) have pointed out.11
54 Carlo Geraci and Josep Quer

Despite the natural refinements and reformulations resulting from broader test-
ing of the hypotheses, it seems clear that Benedicto and Brentari’s approach opens
up a very fruitful and promising line of inquiry into the argument structure in
SLs. It should be mentioned here that Mathur and Rathmann (2007) follow a dif-
ferent path based on Reinhart’s Theta-system (Reinhart 2002) and put the weight
of the explanation on the featural characterization of root morphemes in classi-
fier predicates (HANDLE, MANIPULATE and LOC/MOV, that correspond to
instrumental, handling and motion/location classifiers, respectively).
Interesting extensions in the argument structure properties of classifier predi-
cates can be found in the expression of manner with motion verbs. Supalla (1990)
identified constructions such as the one illustrated in the example in (14) in ASL
as a serial verb expressing a single motion event by combining a manner predi-
cate (RUN) with a motion verb encoding path (GO_UP_ZIGZAG), that ordering
being the only grammatical one.
(14) PERSONy 1-1dwny+RUN 1y+GO_UP_ZIGZAG
person legsbody_part_cl+run human_beingwhole_entity+go_up_zigzag
‘A person (is) running zigzag up(hill).’  (adapted from Supalla 1990)

Interestingly, further crosslinguistic work has shown that parallel serial verb
structures exist in LSA and LSC as well (Benedicto et al. 2008), but they do not
display the ordering restriction just mentioned. In fact, a broader range of pos-
sibilities is attested. For instance, the verb expressing manner can incorporate
the path. The identified patterns are explained as different constraints operating
on a VP-shell structure that is taken to realize the serial construction at play. An
alternative view on this type of data is offered in Tang and Yang (2007: 1235) for
Hong Kong SL, who reject the serial verb analysis and propose that they are “mor-
phological V-V compounds the composition of which is based on a universal
conceptual schema.”12
What becomes clear from the work on argumental patterns in classifier con-
structions is that fine-grained morphosyntactic (and phonological) analysis is
able to unveil the intricacies of SL predicates in a domain that has often been
thought of as less constrained by grammar and more by iconic or representational
properties.

5. Discussion and future prospects

The approach to agreement facts and argument structure in SLs presented here
might look intriguing at face value, but it is clear that once their abstract prop-
erties are identified, we are faced with the patterns normally found in natural
 Determining argument structure in sign languages 55

language; however, there are also aspects of the various proposals that remain
open. We would like to highlight some of them as open issues for future research.
A first aspect has to do with the class of predicates showing backward agree-
ment. There are signs in which a path trajectory links objects and subjects in the
reverse order, like in the case of TAKE and INVITE. Any approach that tries to
derive the directionality of agreement from the hierarchical structure encounters
serious troubles when faced with cases like these.13
Another significant aspect concerns the analysis of handshape alternation in
terms of (gender) agreement in SLs. If this approach is correct, then what we are
observing is that the part of the SL lexicon which is commonly considered to be
more iconic and less governed by grammatical principles has developed a con-
sistent, highly grammatical, modality universal system to mark agreement. This
system does not seem to be confined to the cases of direct object of ditransitive
predicates, since it shows up also with intransitive predicates (see the example in
(6) above, but also (12b)). While handshape alternation tends not to overlap with
spatial agreement in case of ditransitive predicates, what happens in the case of
intransitive or simple transitive CL predicates is less obvious, since the possibility
is left for the classic agreement pattern to show up in addition to gender agree-
ment. More work is needed in this direction, in order to understand how the two
types of agreement interact with each other.
An overall result that emerges from most of the works reviewed here is that
SLs seem to make a strong case for the view that takes the lexicon just to be a
collection of features on which syntax and morphology operate. This can be seen
in the regularities emerging by handshape modification in ditransitive predicates,
which are analyzed as morphological gender markers in Zwitselood’s works; or by
argument restrictions in classifiers predicates, which are the manifestation of the
unergative/unaccusative alternation in a specific domain of the SLs (that of the
classifier predicate) where grammatical phenomena in the strict sense of the term
where expected not to be found. Given the centrality of aspectual composition
for accounts of argument structure such as Ramchand (2008) or Borer (2005),
the kind of work undertaken within Wilbur’s (2008) Event Visibility Hypothesis
(EVH) turns out to be especially relevant. This hypothesis builds on the regu-
larities identified in the mapping between the semantic components of events
and their morphophonological realizations in a visual-gestural language like ASL.
Specifically, the claim is that the way in which SLs encode events in their predicate
system is directly reflected by their morphophonological organization, therefore
arguing for the need of an interface mediating the semantic component and the
morphophonological component, in the spirit of Jackendoff (1997, 2008). This re-
lies on the very strong parallel hypothesis that the components of predicate signs
are grammaticalized from universally available physics of motion and geometry
56 Carlo Geraci and Josep Quer

of space. For example, it is a physical property of movements in SLs that they can
have a rapid deceleration to an end point in path movements. This physical fact
is immediately observable in the production of many predicates and is normally
associated to telicity (e.g. it is taken as a marker for telicity). Typically, atelic ac-
tivities like bicycling tend not to have path movements across SLs and therefore
they are incompatible with the telic marker. Such a proposal presupposes a strong
modality effect, but it does not lead to conclude that event composition is uni-
form across sign languages. In fact, crosslinguistic variation has been determined
when the EVH has been applied to a language other than ASL, namely Austrian
SL (ÖGS) (Schalber 2006). It is important to keep in mind that recruiting prop-
erties of the visual-gestural medium through a grammaticalization process does
not mean that we must adopt a purely localist view on the morphosemantics of
predicates for SLs (cf. Kegl 1990): once grammaticalized, the morphemes make
part of a formal system with language-particular properties. As such, the EVH
opens up a new window into predicate structure by assuming that event proper-
ties are encoded overtly and are thus identifiable for their compositional analysis.
In a different vein, but still putting the emphasis on the relevance of aspectual
properties for argument structure, Mathur and Rathmann (2007) dispense with
the lexical specification of SOURCE and GOAL theta-roles in classifier predicates
by deriving them from the telicity of the situation type, being inherently telic or
coerced into a telic interpretation.
Although both person and locative agreement have been argued to be the
same phenomenon (Quadros and Quer 2008), it remains to be properly under-
stood why locative agreement marking takes precedence over person agreement
marking, as in the LSC example (15):14
(15) IX1 LONDONx PARISy xFLYy BUSINESS.
‘I flew from London to Paris in business class.’

As mentioned above, one of the central questions that must be addressed is to

what extent the distinction between lexical predicates and classifier predicates can
be maintained, as has been traditionally assumed, and whether it should rather
be dispensed with. At bottom, what we still lack is a proper understanding of the
distinction between grammatical use of space (with lexical predicates) and topo-
graphic use (with classifier predicates). To the extent that the latter is part of the
grammatical system as well, we need to find a satisfactory way to integrate it into
the accounts of linguistic structure.
 Determining argument structure in sign languages 57

6. Conclusion

In this paper we offered an overview of the research on argument structure in SLs

and on other aspects of SL grammars such as agreement patterns that hinge on
and are linked to the expression of arguments. The study of classifier predicates,
often considered to go beyond our view of the stable lexicon of a language, has
been surprisingly useful to identify argument structure properties and alterna-
tions that are well established for spoken languages. In this respect, a reexam-
ination of what seems to be the most motivated/iconic and less frozen part of
SLs with the formal tools provided by current approaches to argument structure
turns out to be the most promising approach to establishing modality effects (or
lack thereof) on solid grounds. At the same time, a compositional neoconstruc-
tionist approach to argument structure appears to be reinforced by the results that
have been so far made available by SL research (although systematic works in this
framework have never been proposed in the literature of SLs yet). The intriguing
aspects of argument structure in SLs briefly reported upon here cannot but be-
come the focus of interest for any work targeting this domain in natural language
in its broad sense.

Notes

1. The research was partly made possible by a grant awarded by the Spanish Ministry of Econ-
omy and Competitiveness to Josep Quer (FFI2012-36238), by the Govern de la Generalitat de
Catalunya (2009SGR00763) and by SignGram Cost Action IS1006.

2. This is an oversimplified picture of how person agreement works in SL. Orientation or

facing of the hand towards an argument locus can also constitute the exponence of agree-
ment. Agreeing verbs can be further divided in subclasses obeying different morphological
constraints (Sandler and Lillo-Martin 2006). A non trivial property of overt person agreement
in SL is its surface optionality, especially in the case of subject agreement (see Meier 1981 and
subsequent works). In addition to overt modification of the movement component of the pred-
icate, SLs may adopt non-manual strategies to signal agreement (Neidle et al. 2000).

3. Interestingly, person agreement auxiliaries may encode aspectual information, but they
never encode tense or mood information.

4. More traditional approaches (see for instance the discussion in Sandler and Lillo-Martin
2006) analyzed these phenomena as cases of argument (generally object) incorporation similar
to those found in spoken languages (Baker 1988). To our knowledge, there is no systematic
comparison between the approach presented in the text and the one based on noun incorpo-
ration that may help us deciding between the two. Meir (2001) actually resorts to two different
types of noun incorporation to explain the different behavior of what she calls theme classifiers
58 Carlo Geraci and Josep Quer

and instrumental classifiers in Israeli SL. She analyzes the former as cases of noun incorpora-
tion into the verbal root that does not saturate the theme argument, thus allowing doubling by
an NP that can also be stranded. By contrast, in her account instrumental classifiers are cases of
compound noun incorporation, where the argument is saturated lexically in the V-N complex,
which explains the unavailability of doubling and stranding.

5. The term gender agreement may be not immediately clear. Here is the characterization
proposed in Zwitserlood (2003a: 392–393): “The handshape is taken from a small set of mean-
ingful handshapes and reflects a characteristic (often the shape) of the referent. For instance,
in intransitive verbs, the [ ] handshape usually represents long and thin, or animate referents
(pens, knives and poles, and humans and animals) […] I call this type of agreement gender
agreement, because it is reminiscent of gender agreement in Bantu languages”.

6. This second assumption is not explicit in Zwitserlood (2003a), where the impossibility of
two handshapes entertaining two separate and visible agreement relations is ruled out by the
sign language interface.

7. Kegl (1990) constitutes the only previous treatment of argument structure and alternation
in ASL.

8. In whole entity classifiers, the handshape is analyzed as a morpheme referring to the entire
size of an object, as in the case of classifier predicates for cars in LIS illustrated in the example
(6), above.

9. The symbols S, 1, C and B used in the glosses refer to the handshapes used in these type of
classifiers and correspond to the manual alphabet letters used in fingerspelling. The shape S is
a close fist, indicating the head of a person, the shape 1 indicates a standing individual in its
entirety, the shape C indicates the size of a thick object, and the shape B indicates a flat object.

10. For details, see Benedicto and Brentari (2004).

11. De Lint (2010) argues that instrumental classifiers do not enter the handling/whole entity
classifier alternation, as predicted by Benedicto and Brentari (2004).

12. In this connection, see their analysis of causative predications as either lexicalized causative
verbs or complex predicates.

13. Meir’s (1998, 2002) proposal to analyze the agreement pattern of these predicates as involv-
ing two independent mechanisms of agreement may reach an adequate description of the facts,
but does not offer a concrete explanation of why such class of verbs exists crosslinguistically in
SL. For the problems raised by Meir’s thematic approach to agreement in SL, see Quadros and
Quer (2008).

14. See Janis (1995) for a proposal in terms of agreement patterns as controlled by a case hier-
archy.
 Determining argument structure in sign languages 59

References

Baker, M. C. 1988. Incorporation. A theory of grammatical function changing. Chicago: The Uni-
versity of Chicago Press.
Benedicto, E., and Brentari, D. 2004. Where did all the arguments go? Argument-changing
properties of classifiers in ASL. Natural Language and Linguistic Theory 22: 743–810.
DOI: 10.1007/s11049-003-4698-2
Benedicto, E., Cvejanov, S., and Quer, J. 2007. Valency in classifier predicates: A syntactic Anal-
ysis. Lingua 117: 1202–1215. DOI: 10.1016/j.lingua.2005.06.012
Benedicto, E., Cvejanov, S. and J. Quer. 2008. The morphosyntax of verbs of motion in serial
constructions: a crosslinguistic study in three signed languages. In J. Quer (ed.), Signs of
the Time: Selected papers from TISLR 2004, 111–132. Seedorf: Signum Verlag.
Borer, H. 2005. Structuring Sense (2 vol.). Oxford: Oxford University Press.
DOI: 10.1093/acprof:oso/9780199263929.001.0001
Brentari, D. 1998. A prosodic model of sign language phonology. Cambridge, MA: MIT Press.
Brentari, D., and Padden, C. 2001. A Language with Multiple Origins: Native and Foreign Vo-
cabulary in American Sign Language. In D. Brentari (ed.), Foreign Vocabulary in Sign Lan-
guage: A Cross-linguistic Investigation of Word Formation, 87–119. Mahwah, NJ: Lawrence
Erlbaum Associates.
Embick, D., and Rolf, N. 2007. Distributed Morphology and the Syntax/morphology Interface.
In G. Catriona Ramchand (ed.), The Oxford Handbook of Linguistic Interfaces, 289–324.
Oxford: Oxford University Press.
Engberg-Pedersen, E. 1993. Space in Danish Sign Language. Hamburg: Signum Verlag.
Glück, S., and Pfau, R. 1999. A Distributed Morphology account of verbal inflection in German
Sign Language. In T. Cambier-Langeveld, A. Lipták, M. Redford, and E. J. v.d. Torre (eds.),
Proceedings of Console VI, 65–80.
Grose, D., Wilbur, R. B. and K. Schalber. 2007. Events and telicity in classifier predicates: A
reanalysis of body part classifier predicates in ASL. Lingua 117: 1258–1284.
DOI: 10.1016/j.lingua.2005.06.014
Halle, M., and Marantz, A. 1993. Distributed morphology and the pieces of inflection. In
K. Hale and S. J. Keyser (eds.), The View from Building 20, 111–176. Cambridge, MA: MIT
Press.
Harley, H. and Noyer, R. 1999. Distributed Morphology. Glot International 4(4): 3-9.
Jackendoff, R. 1997. The Architecture of the Language Faculty. Communication. Cambridge, MA:
MIT Press.
Jackendoff, R. 2008. Construction After Construction and Its Theoretical Challenges. Language
84 (1): 8–28. DOI: 10.1353/lan.2008.0058
Janis, W. D. 1995. A crosslinguistic perspective on ASL verb agreement. In Language, Gesture,
and Space. Karen Emmorey and Judy S. Reilly (eds.), 255–286. Hillsdale, NJ: Lawrence
Erlbaum Associates, Publishers.
Kegl, J. 1990. Predicate Argument Structure and Verb-Class Organization in the ASL Lexicon.
In C. Lucas (ed.), Sign Language Research: Theoretical Issues, 149–175. Washington, D.C.:
Gallaudet University Press.
Lint, V. de. 2010. Argument Structure in Classifier Constructions in ASL: an Experimental
approach. MA dissertation, Universiteit Utrecht.
60 Carlo Geraci and Josep Quer

Mathur, G. and Rathmann, C. 2007. The argument structure of classifier predicates in Ameri-
can Sign Language. In A. Rose Deal (ed.), Proceedings of the Fourth Meeting of Semantics
of Underrepresented Languages of Americas. Amherst, MA: GLSA (Graduate Linguistic
Students Association).
Meier, R. P. 1981. Icons and morphemes: Models of the acquisition of verb agreement in ASL.
Papers and reports on child language development 20: 92–99.
Meir, I. 1998. Syntactic-semantic interaction in Israeli Sign Language verbs: The case of back-
wards verbs. Sign Language and Linguistics 1.1: 3–37. DOI: 10.1075/sll.1.1.03mei
Meir, I. 2001. Verb classifiers as noun incorporation in Israeli Sign Language. Yearbook of Mor-
phology 1999: 295–315.
Meir, I. 2002. A Cross-Modality Perspective on Verb Agreement. Natural Language and Lin-
guistic Theory 20: 413–450. DOI: 10.1023/A:1015041113514
Neidle, C. et al. 2000. The Syntax of American Sign Language. Functional Categories and Hierar-
chical Structure. Cambridge, MA: MIT Press.
Padden, C. A. 1988. Interaction of morphology and syntax in American Sign Language (Out-
standing Dissertations in Linguistics, series IV). New York: Garland Press.
Padden, Carol A. 1990. The relation between space and grammar in ASL verb morphology.
In C. Lucas (ed.), Sign Language Research. Theoretical Issues, 118–132. Washington, DC:
Gallaudet University Press.
Quadros, R. M. de, and Quer, J. 2008. Back to backwards and moving on: on agreement, aux-
iliaries and verb classes in sign languages. In Sign Languages: spinning and unraveling the
past, present and future. TISLR9, forty five papers and three posters from the 9th. Theoret-
ical Issues in Sign Language Research Conference, Florianópolis, Brazil, December 2006,
530–551. Petrópolis, RJ: Editora Arara Azul.
Ramchand, G. C. 2008. Verb meaning and the lexicon. A first phase syntax. Cambridge:
Cambridge University Press. DOI: 10.1017/CBO9780511486319
Reinhart, T. 2002. The Theta System: An overview. Theoretical Linguistics 28: 229–290.
Sandler, W., and Lillo-Martin, D. 2006. Sign Language and Linguistic Universals. Cambridge,
UK: Cambridge Univerisity Press. DOI: 10.1017/CBO9781139163910
Schalber, K. 2006. Event visibility in Austrian Sign Language (ÖGS). Sign Language & Linguis-
tics 1/2: 207–231. DOI: 10.1075/sll.9.1.11sch
Supalla, T. 1990. Serial verbs of motion in American Sign Language. In S. Fischer (ed.), Theoret-
ical Issues in Sign Language Research, 129–152. University of Chicago Press.
Tang, G., and Yang. 2007. Events of motion and causation in Hong Kong Sign Language. Lingua
117(7): 1216–1257. DOI: 10.1016/j.lingua.2006.01.007
Wilbur. R. 2008. Complex Predicates involving Events, Time and Aspect: Is this why sign lan-
guages look so similar? In J. Quer (ed.), Signs of the time, 217–250. Seedorf: Signum Verlag.
Zwitserlood, I. 2003a. Word formation below and above little x: Evidence from Sign Language
of the Netherlands. In A. Dahl, K. Bentzen and P. Svenonius (eds.), Nordlyd Tromsø Uni-
versity Working Papers on Language and Linguistics, Vol. 31, 488–502.
Zwitserlood, I. 2003b. Classifying hand configurations in Nederlandse Gebarentaal (Sign Lan-
guage of the Netherlands). Doctoral dissertation, Universiteit Utrecht.
 The processing and representation
of light verb constructions

Eva Wittenberg,1,2,3 Ray Jackendoff,1 Gina Kuperberg,1

Martin Paczynski,1 Jesse Snedeker2 and Heike Wiese3
1 Tufts University / 2 Harvard University / 3 Potsdam University

This article gives an overview of our ongoing research on the processing and
representation of light verb constructions. Light verb constructions consist of
a verb that is semantically bleached, and an event nominal, which identifies
the kind of event. Together the noun and the verb determine the structure of
that event (the number of participants and their roles). Critically, in light verb
constructions the canonical mapping from surface syntactic structure to event
structure is disrupted. The present studies examine this phenomenon through
the lens of language processing. We summarize several behavioral and neurolin-
guistic studies that show that the interpretation of light verb constructions relies
on noncanonical mappings between syntax and semantics, while their syntactic
structure is not different from non-light constructions.

1. Introduction

Light verb constructions are complex predicates in which the verb is semantically
bleached. It merely expresses aspect, directionality or aktionsart of the predicate,
while the bulk of the predicative meaning stems from an event nominal within
the construction (Butt, 2010; Wiese, 2006).1 For example, in a sentence like Henry
took a walk, the character associated with the subject (Henry) is not transferring a
concrete thing into his possession, as in Henry took a spoon. Instead, take a walk
describes the same kind of event as the verb walk. The event nominal walk is part
of the predicate and assigns semantic roles to the subject, just like take. Thus, the
subject of the sentence is not only understood as the Agent of the verb take, but
also as the Agent of the event nominal walk – a phenomenon known as “argument
sharing” (Baker, 1989; Durie, 1988; Jackendoff, 1974; see also Alsina, 1996; Butt,
1995).2
62 Eva Wittenberg et al.

Light verb constructions have inspired research in a large number of languag-

es, from a vast array of theoretical perspectives, and with a variety of goals, re-
sulting in a heterogeneous set of terminology, definitions, and analyses (Winhart,
2002; Butt, 2010). A detailed examination of this literature is beyond the scope of
this paper. Here, we focus on the syntactic and semantic structure of some of the
most clear cut cases of light verb constructions, providing a summary of studies
that test how they are processed in English and German, and discussing the im-
plications of these studies for theoretical accounts.
In languages such as English and German, the surface syntax of light verb
constructions usually does not differ from the surface syntax of non-light con-
structions using the same verb. The subcategorization frame of the light verb
generally determines the syntactic argument structure of the sentence, just as in
non-light constructions. The event nominal occupies a syntactic argument posi-
tion within this subcategorization frame, usually one associated with the semantic
role Theme (see Winhart, 2002, for discussion). For instance, take is normally
a transitive verb whose object is the Theme; in the light verb construction take
a walk, a walk appears in direct object position, though it is not the Theme of
the event, but rather part of the predicate. Similarly, give is normally a ditran-
sitive verb that can appear with its Theme in one of two places depending on
whether the double object or prepositional object construction is used. Most light
verb constructions with give participate in the dative alternation, with the event
nominal always appearing where the Theme would be (give a hug to Harry/give
Harry a hug).
In contrast, the semantic structure of light verb constructions is clearly dis-
tinct from that of non-light constructions. Compare (1a–d).
(1) a. Henry gave a book to Elsa. [non-light]
b. Henry described a kiss to Elsa. [non-light]
c. Henry gave a kiss to Elsa. [light]
d. Henry kissed Elsa. [non-light]

First consider event structure: In the non-light construction (1a), the book is a
physical object, independent from the act of giving, but in the light verb con-
struction (1c), the kiss is an event type. Non-light constructions can also contain
verbs followed by event nominals like (1b). However, such constructions refer to
two independent actions, a describing and a kissing (which is being described),
while the light construction (1c) does not refer to two independent actions, one
of giving and one of kissing, but only one, just like (1d).
Next, consider the semantic roles in these events. In (1a), Henry is Agent, a
book is Theme, and Elsa is Recipient or Beneficiary. In the non-light construction
(1b), the event nominal kiss implies a kisser and kissee, but their identities are
 Processing and representation of light verb constructions 63

indeterminate. By contrast, in (1c), Henry is clearly the kisser (Agent) and Elsa is
the kissee (Patient), just as with the verb kiss in (1d). This difference between (1b)
and (1c) is the manifestation of argument sharing in the light verb construction.
The issues raised by light verb constructions, then, are (a) how the combina-
tion of light verb and event nominal is interpreted as denoting a single event, and
(b) how the shared semantic roles of the event nominal are determined. An ap-
proach that we find attractive (though we will discuss others below) is that when
a verb appears in a light verb construction with an event nominal, the event nom-
inal is not assigned a standard thematic role such as Theme or Patient, but what
we might call a “co-event” role. A co-event, unlike a Theme or Patient, is not a
semantic argument of the event denoted by the verb, but rather a further specifi-
cation of the event type itself: the event denoted by (1c) is both a giving (of sorts)
and a kissing.3 The difference between a verb in its light and in its non-light use,
then, is that, in its light use, it has a co-event instead of a Theme, and thus there is
a noncanonical mapping between syntax and semantics. We will call this proposal
the co-event hypothesis.
The co-event hypothesis leads directly to an account of argument sharing. If
give and kiss in (1c) together describe a single event, this event must have a single
grid of thematic roles, incorporating the thematic roles individually specified by
give and kiss. Thus the thematic roles assigned to kiss depend (at least in part)
on the thematic roles of the main verb. This prediction is correct: contrast (2a
and 2b).
(2) a. Henry gave Elsa a kiss.
roles of give: Agent Beneficiary co-event
roles of kiss: Agent Patient
b. Elsa received a kiss from Henry
roles of receive: Beneficiary co-event Agent
roles of kiss: Patient Agent

(2) illustrates a general principle governing the thematic grids of light verbs and
event nominals: Agents align with Agents, and Patients align with Patients or
Beneficiaries – regardless of how these roles are expressed syntactically. (3) and
(4) are further illustrations of variation in alignment that depend on the thematic
roles assigned by the light verb.
(3) a. Joan did an operation on Harry. (= ‘Joan operated on Harry’)
b. Harry had an operation. (= ‘someone operated on Harry’)
(4) a. The dinner gave Bill pleasure. (= ‘the dinner pleased Bill’)
b. Bill got pleasure from the dinner. (= ‘the dinner pleased Bill’)
64 Eva Wittenberg et al.

The light verb can also affect the number of thematic roles of the event: (5a) is a
simple one-character event, but (5b) adds a causative agent.
(5) a. Olive took a bath. (= ‘Olive bathed’)
b. Tom gave Olive a bath. (= ‘Tom bathed Olive’)4

In addition, the light verb influences the aspectual properties of the overall event.
For instance, a kissing event may be telic (cf. She kissed him in 5 minutes) or atelic
(cf. She kissed him for 5 minutes). But since give is telic, give a kiss can only be telic
(She gave him a kiss in 5 minutes/*for 5 minutes).5 (See Folli, Harley, and Karimi,
2002, for a discussion of these points for Persian light verbs; McGinnis, 2002, for
similar behavior in VP idioms). Thus the light verb plays an important role in the
overall construal of the event.
Overall, then, the co-event hypothesis permits a rather natural account of the
event structure of light verb constructions and of argument sharing (see Wiese,
2006, for a formal semantic representation that captures this, and Culicover and
Jackendoff, 2005, Section 6.5.1, for a different formal approach based on similar
intuitions).
A further complication in of light verb constructions is that not any light verb
can combine with any event nominal (e.g. *make a shower, *give a jog). Some
combinations and limitations are idiosyncratic (e.g. American/British make a
decision vs. British-only take a decision). Nevertheless, there are pockets of (rel-
ative) productivity, defined by fine-grained semantic constraints. For example,
give readily combines with any kind of noun denoting deliberate contact (Andrew
gave Holly a push/kick/kiss/hug). Likewise, have combines with event nominals in
constructions where the sentential subject is construed as Patient or Experiencer
of the event nominal (Joe had an operation/accident/collision), while it does not
combine so readily with contact event nominals (*Fanny had a kick/stab/punch/
kiss at Gerry; see Wierzbicka, 1982; Brugman, 2001; Newman, 1996 for detailed
case studies). Thus these pockets of productivity depend on the interaction be-
tween the light verb and the meaning of the event noun. Furthermore, at least
some instances of the construction must be stored in memory as quasi-idioms.
The problems of co-event structure, argument sharing, and semi-productivity
pose interesting problems for theories of grammar; we will discuss some of these
below. But they also raise questions about sentence processing. On the co-event
hypothesis, light verb constructions have the same syntactic structure as non-
light constructions but the mapping between syntax and semantics differs. Specif-
ically, the argument that would typically be assigned the role of Theme is assigned
the role of co-event, inducing a process in which the semantic roles of both pred-
icates are aligned. Thus, on this hypothesis, light verb constructions make use
of the same syntactic structures and structure building operations as non-light
 Processing and representation of light verb constructions 65

constructions, but they require an additional operation to align the semantic roles
of the verb and the event nominal: Jan gave Julius an order has the same syntactic
form as Jan gave Julius an orange, but the event representation differs. The predic-
tions that we make about the processing of light verb constructions will depend
upon the factors that we deem most relevant to comprehension or production.

– Since the verbs in question are more frequent as light verbs than as non-light
verbs, processing effort might be decreased (see remarks below.)
– Aligning the thematic grids of the light verb and event nominal may require
more processing resources. Since under the co-event hypothesis, the proper
alignment depends on the semantics of the light verb and event nominal rath-
er than (or at least more than) their syntax, the processing burden should be
primarily in the course of semantic composition, not in syntactic parsing.

The experiments reported in the next section investigate these predictions.

2. Investigating the processing of light verb constructions

As in any processing study, one factor that must be addressed is the frequency of
the construction under investigation. Both lexical frequency and cloze probability
are inversely related to processing effort (Bicknell and Levy, 2012). In most lan-
guages, light verbs are actually among the most frequent verbs in the lexicon. For
instance, the light verbs take, have, make, do, and give are among the twenty most
frequent verbs in English (PropBank corpus, Palmer, Gildea, and Kingsbury,
2005). Also, the frequency of particular combinations of light verbs and event
nominals is significantly higher than that of non-light verb–noun combinations,
and a some verbs such as give are more frequent in light constructions than in
non-light constructions (Piñango, Mack, and Jackendoff, to appear; Wittenberg
and Piñango, 2011). Consequently, we might expect that light verb constructions
would be processed more easily than non-light constructions.
In the first psycholinguistic study investigating the processing of light verb
constructions, Piñango et al. (to appear) used a cross-modal lexical decision task.
Participants listened to light verb constructions such as Mr. Olson gave an order
to the produce guy, as well as to non-light constructions using the same verbs
(gave an orange) or the same nouns (typed an order). After the object noun (or-
der/orange) was heard, letter-string probes appeared on the screen. Participants
were required to make a lexical decision about whether these strings were words
or non-words. Piñango et al. found that participants were slower to respond to
probes appearing 300 ms after the end of a light verb construction than after a
66 Eva Wittenberg et al.

non-light construction using the same object. No differences were seen when the
probes were presented immediately after the offset of the object noun.
In a follow-up experiment, Wittenberg and Piñango (2011) used the same
methodology, using German subordinate clauses, which have a verb-final sen-
tence structure (6):
(6) Während der Demonstrant ‘While the protester
a. einen Vortrag hielt, a speech held, [light]
b. eine Fahne hielt, a flag held, [non-light, same verb]
c. einen Vortrag hörte, a speech heard, [non-light, same noun]
schritt die Polizei ein intervened the police’

In this context, all arguments could be presented before the verb. When the probe
appeared 300ms after the offset of the verb, the reaction times for light verb con-
structions were longer than those for both of the other constructions. Again,
there was no difference in reaction times when the probe appeared immediately
at the end of the verb.
The increased reaction times to light verb constructions were taken as evi-
dence for increased processing costs. That these costs only arose after a certain
time, and not immediately after the construction, was interpreted as an effect of
complex operations in the mapping between syntactic and semantic argument
structures, as predicted by the co-event hypothesis, and not as a consequence of
arguably faster processes, such as lexical access or a first-pass parsing (Embick,
Hackl, Schaeffer, Kelepir, and Marantz, 2001; Boland, 1997; McElree and Griffith,
1995).
The results of both studies are important first steps for our understanding
of how light verb constructions are processed. However, while the findings do
show that these constructions incur processing costs by 300 ms after the end of
the construction, nothing can be said about the more fine-grained mechanics of
processing.
One study that could have shed light on this was conducted by Briem,
Balliel, Rockstroh, Butt, Schulte im Walde, and Assadollahi (2010). They carried
out three MEG experiments in German, contrasting potential light verbs like geb-
en (‘give’) with non-light verbs like erwarten (‘expect’). Experiment 1 presented
isolated verbs in third person singular present tense. They found that non-light
verbs were associated with more activity than light verbs in a central occipito-pa-
rietal region. Experiment 2 presented these verbs together with a subject pro-
noun (Er gibt, ‘he gives’). The same effect was found, as well as an increased signal
for non-light verbs between 160–200 ms in a left visual region. In Experiment 3,
verbs in non-light verb constructions evoked more activity than verbs in light
constructions between 270–340 ms in left temporal regions.
 Processing and representation of light verb constructions 67

Thus, in all three experiments, non-light verbs elicited more neural activi-
ty than light verbs. These findings seem to directly contradict the predictions of
the co-event hypothesis, as well as the behavioral findings of Piñango et al. (to
appear) and Wittenberg and Piñango (2011) described above, namely more cost
associated with light verb constructions than with non-light constructions. How-
ever, as Wittenberg, Paczynski, Wiese, Jackendoff and Kuperberg (under review)
discuss in detail, several confounds restrict the interpretation of Briem and col-
leagues’ study: verbs were imbalanced in terms of length and morphological com-
plexity, several of the non-light items were of questionable grammaticality, and
the two classes of verbs differed in imageability. Moreover, in Experiment 3, the
stimuli used object-verb-subject order, which is a marked word order for isolated
sentences in German.
However, for our present purpose, the most critical limitation of these studies
is that they were not designed to address processes related to argument sharing,
which necessarily involves the interaction of the verb, the event nominal, and all
arguments, rather than just the verb alone. Experiments 1 and 2 did not include
the event nominal; the results may simply reflect the higher frequency of light
verbs. Experiment 3 did include all arguments, but critically the researchers did
not analyze activity after the verb, which is where the studies described above
found evidence for greater processing effort.
To help close this gap, Wittenberg et al. (2014) investigated the processing
of light verb constructions using Event-Related Potentials, which measure brain
activity during sentence comprehension. In this study, participants first saw a
context sentence like (7), presented as a whole. Then they saw a verb-final subor-
dinate clause that was either a light verb construction (8a), a non-light construc-
tion using the same verb (8b), or an anomalous construction using the same verb
(8c). This was followed by the matrix clause (9). Both the subordinate and matrix
clause were presented word-by-word. Cloze probabilities, as determined in a sep-
arate test, were highest for the light condition, lower for the non-light condition,
and zero for the anomalous condition. Our analyses focused on the response to
the verb (underlined in the examples).
(7) Das Flugzeug war bereits hoch über den Wolken.
The airplane was already high over the clouds.
‘The airplane was already high in the sky.’
(8) a. Als die Stewardess eine Ansage machte,
When the stewardess an announcement made
‘When the stewardess made an announcement’
68 Eva Wittenberg et al.

b. Als die Stewardess einen Kaffee machte,

When the stewardess a coffee made
‘When the stewardess made a coffee’
c. *Als die Stewardess ein Gespräch machte,
When the stewardess a conversation made
‘When the stewardess made a conversation’ (unacceptable in German)
(9) ging gerade die Sonne auf.
went just the sun up
‘the sun was just rising.’

Examining the waveforms at the verb, we found no differences between sentence

types in the classic N400 time window (300–500 ms). The anomalous sentences
evoked a posteriorly-distributed positivity effect (a P600) relative to the other two
sentence types. The light verb constructions, however, in contrast to the other two
constructions, evoked a widespread negativity from 500–900 ms with an anterior
focus. Since the same verb was used for all three sentence types, we can rule out
the possibility that lexical factors drove these effects. This cannot be an N400 re-
sponse to difference in cloze probability; if it were the negativity would be smaller
for the light verbs, since they are more predictable.
As we discuss in Wittenberg et al. (2014), these findings are consistent with
the behavioral results from Piñango et al. (to appear) and Wittenberg and Piñango
(2011). Both studies found longer reaction times for making a lexical decision to a
probe after light verb constructions than after non-light constructions. Crucially,
the probes in the behavioral experiments were placed 300 ms after the offset of the
constructions. Assuming that the critical word lasted 200–300 ms, this slowdown
correlates with the onset of negativity effect that was observed 500–600 ms after
the onset of the critical word in the ERP study. The negativity evoked by the light
verb constructions could reflect the mapping operations involved in argument
sharing. During sentence comprehension, predictions are made about the roles
that the arguments are to receive; once the verb is encountered and a light verb
construction detected, the semantic roles have to be distributed both from the
light verb and the event nominal. Specifically, under the co-event hypothesis, the
event nominal has to be integrated as a co-event with the main verb, and the the-
matic grids of the two predicates have to be aligned with the resulting composite
event.6
The results in these studies offer an interesting parallel with coercion opera-
tions such as aspectual coercion (10).

(10) The light flashed until dawn. (= ‘The light flashed repeatedly until dawn’)
 Processing and representation of light verb constructions 69

In this case, it has been argued that extra semantic material (underlined in the
gloss) is introduced in the course of mapping from syntax to semantics (Talmy,
1978 and Jackendoff, 1991, among others). Experiments on aspectual coercion
have found effects that have similar scalp distributions and similar timing to the
effects that we found for light verb constructions, a pattern which is different from
the classic N400 pattern (Bott, 2010; Paczynski and Kuperberg, to appear). Thus
our ERP results add to a growing body of data showing processing costs for con-
structions that involve noncanonical mappings between syntax and semantics.
The co-event hypothesis claims that while non-light and light verb construc-
tions have different syntax-semantics mappings, they share the same syntactic
structure. Wittenberg and Snedeker (in prep.) used structural priming to test this
claim. This paradigm exploits the fact that during language production, people
tend to automatically repeat structures that they have recently encountered. In
particular, it has been shown that hearing Double Object (DO) word order (Henry
gave Elsa a rose) primes production of semantically unrelated DO constructions
(Joan showed Harry her stamp collection), and hearing Prepositional Object (PO)
word order (Henry gave a rose to Elsa) primes production of semantically un-
related PO constructions (Joan showed her stamp collection to Harry). Crucially,
syntactic priming during production appears to be attributable to the syntactic
surface structure of the sentence (Bock, 1986, 1989; Bock and Loebell, 1990).
Light verb constructions with give undergo the dative alternation just like
non-light ditransitive constructions. However, prior studies of dative priming
have focused exclusively on non-light datives. We tested whether light verb di-
transitives would prime the word order of non-light ditransitives as effectively
as other non-light ditransitives. We reasoned that if the difference between light
and non-light verbs is one of semantic structure, and not syntactic structure, then
their priming behavior should not differ. On the other hand, if light verb con-
structions have a different syntactic structure from non-light constructions, they
should serve as less effective primes for non-light constructions, since the degree
of representational overlap is decreased.
Participants read out loud prime sentences that were either light or non-light
and that employed either DO word order (Henry gives Elsa a kiss/rose) or PO
word order (Henry gives a kiss/rose to Elsa). Then they described target pictures
that could be described equally well with a PO or a DO construction. To disguise
the purpose of the experiment, a distractor memory task was used (cf. Bock and
Loebell, 1990).
We found that both light and non-light DO primes resulted in more DO tar-
gets than did PO primes, with a robust main effect of dative type that was reliable
for both the light and the non-light sentence types. There was no significant in-
teraction between sentence type and prime type. Thus, even though the semantic
70 Eva Wittenberg et al.

argument structures of light and non-light verb constructions differ dramatically,

they both prime non-light targets equally, suggesting that light and non-light con-
structions have the same syntactic form.
The interpretation of these findings depends on our understanding of struc-
tural priming. As we noted above, most of the prior research suggests that struc-
tural priming during production primarily results from an overlap in syntactic
structure of the utterance. Critically, priming can occur between utterances which
have similar surface structures despite having very different semantic structures
(Bock and Loebell, 1990; see also Pickering and Ferreira, 2008, for a review). Our
findings are fully consistent with this literature: light dative sentences prime non-
light datives because they share the same syntactic structure, even though their
semantic structures appear to be quite different.
This is not the pattern that we would expect to find if structural priming
during production primarily reflected semantic structure or how it maps onto
syntactic structure. The light verb primes necessarily had less semantic overlap
with the non-light targets than the non-light primes: the argument that plays the
Theme role in the non-light constructions plays the role of co-event in the light
constructions, and, in the light constructions, argument sharing introduces new
thematic roles from the event nominal which are not present in canonical non-
light constructions. For example, between a non-light prime sentence like The
grandfather is reading the book to the toddler and a target sentence like The girl is
tossing the ball to the boy, the syntactic and thematic structures are identical, in-
volving in both cases an Agent, Theme, and Recipient, in that order. On the other
hand, if the prime sentence is the light verb construction The husband is giving a
kiss to his wife, then the direct object is a co-event rather than a Theme, and the
object of to is a Patient, rather than a Recipient, resulting in less overlap with the
target The girl is tossing the ball to the boy. Thus, if structural priming was highly
influenced by thematic roles, one would have expected different results.
At first glance, these findings may appear to contradict a small set of pri-
or studies, which demonstrate that priming at the level of semantic structure or
thematic mappings can occur (Chang, Bock, and Goldberg, 2003; Thothathiri
and Snedeker, 2012). But this contradiction disappears when we look careful-
ly at the contexts in which thematic priming appears. Thematic priming has
been observed in contexts in tasks that involve comprehension, either as a step
toward production (the RSVP task, Chang et al., 2003) or as the ultimate mea-
sure (Thothathiri and Snedeker, 2012). Picture description tasks, like the present
one, which do not involve comprehension of the sentence to be produced, appear
to be more sensitive to syntax than to semantics (e.g., Bock and Loebell, 1990).
This may reflect the different pathway that information travels along during the
two processes (from meaning to form in production, from form to meaning in
 Processing and representation of light verb constructions 71

comprehension). In addition, effects of thematic structure during production

may emerge only when the effects of syntax have been neutralized. The Chang
study explored locative priming, where both forms share the same surface syntax
and differ only in the ordering of thematic roles. In contrast, in studies like the
present one (where syntactic variation cuts across differences in semantic form)
researchers have consistently found robust effects of syntax on priming which
are not mediated by differences in meaning (see Pickering and Ferreira, 2008, for
review).
To summarize, the experimental studies reviewed here provide data that con-
strain our understanding of light-verb constructions and how they are processed:
The cross-modal lexical decision studies and the ERP study suggest that light verb
constructions call for more processing resources. The cross-modal lexical deci-
sion studies suggest that this extra effort occurs late in processing; results from
syntactic priming suggest that light verb constructions do not differ from non-
light in their syntax. Altogether, these studies are consistent with the co-event hy-
pothesis. Below we consider the degree to which these findings uniquely support
the co-event hypothesis by examining how they would be explained under other
theories about the representation of light verbs.

3. Repercussions for grammatical theories of light verb constructions

In this section we explore how these experimental results might bear on three
different grammatical theories of the light verb construction. Differentiating be-
tween syntactic and semantic mechanisms in processing crucially depends on
which model of the linguistic architecture one is assuming. Phenomena that are
firmly anchored in the syntactic waters of one theory are often regarded as se-
mantic in the next. The connection between theoretical linguistic models and
psycholinguistic data is notoriously hard to tie down (Phillips & Lewis, to ap-
pear), although successful experimental tests of linguistic theories can and do
occur (e.g. Pinker, 1999 on the existence of morphological rules; Hofmeister and
Sag, 2010 on extraction constraints, among many others). For the purposes of
bridging the gap between approaches strictly focusing on linguistic theory and
those that investigate the linguistic system by observing language processing, we
believe it is critical to evaluate any theory of representation in terms of processing.
Our interpretation of the experimental results has been based on the co-
event hypothesis, which is rooted in Jackendoff ’s Parallel Architecture frame-
work (Jackendoff, 1997, 2002; Culicover and Jackendoff, 2005). In this theory, the
structure of a sentence is a triple of phonological, syntactic, and semantic struc-
tures, each characterized by its own set of generative principles. The relationship
72 Eva Wittenberg et al.

among these structures is established by interface principles that link pieces from
multiple components. In particular, a word is thought of as an interface rule that
links a small piece of phonology, a set of syntactic features, and a piece of seman-
tics. Above the level of words, the canonical mapping between syntax and seman-
tics says that semantic functions are associated with syntactic heads, semantic
arguments are associated with syntactic positions such as subject and object, and
semantic modifiers are associated with syntactic adjuncts. However, there are also
many noncanonical mappings between syntax and semantics. For instance, in the
phrase that gem of a theory, the syntactic head is gem, but the semantic head is
theory, as can be seen from the paraphrase in more canonical form: ‘that theory,
which is a gem.’ Thus this construction requires a special interface rule to effect
such a linking.
A light verb construction is another such noncanonical mapping between
syntax and semantics. According to the co-event hypothesis, the direct object is
not interpreted as a canonical semantic argument, but rather as a co-event with
the light verb; in effect the two words map into a single semantic constituent. We
have proposed here that argument sharing is an automatic consequence of this
mapping, in that the single semantic constituent can have only one set of thematic
roles. We have observed that the Agent of the light verb aligns with the Agent of
the event nominal, and that the Patient of the light verb aligns with the Patient of
the event nominal – regardless of their syntactic position in the clause. The result
is that the syntactic arguments of the main verb acquire thematic roles associated
with the event nominal, based not on the syntax, but on the semantics of the light
verb and the nominal. The co-event hypothesis has been partially formalized in
the Parallel Architecture framework by Culicover and Jackendoff (Section 6.5.1)
though they do not work out the details of argument sharing.
The format of the Parallel Architecture lends itself to a direct relation be-
tween grammatical theory and theories of processing (Jackendoff, 2002, 2007). In
particular, noncanonical interface relations in which the syntactic and semantic
structure diverge are predicted to create greater processing load. This is what we
find in the studies on the light verb construction described here.
Another account of light verb constructions is rooted in the framework of
Construction Grammar. Construction Grammar is an umbrella term for a range
of theories that share the basic assumption that the primary units of grammar are
constructions: stored pairings of form and function (Croft, 2001; Goldberg, 1995,
2009; Kay, 1995). There are no independent modules of grammar interacting with
each other; all composition is in terms of full constructions (Fried and Östman,
2004). Thus there is no necessary distinction between canonical and noncanoni-
cal pairings of form and function, except in terms of frequency.
 Processing and representation of light verb constructions 73

Goldberg (2003) works out an analysis of Persian light verb constructions in

this framework. She concludes that light verb constructions must be stored as lin-
guistic units: each noun that can enter a light verb construction is stored together
with its respective light verb, and each of these stored structures is associated with
a distinct meaning; alternatively, groups of light verb constructions can be rep-
resented as sub-constructions of other constructions in an inheritance hierarchy
(Family, 2009).
In terms of processing, the most important factor for Construction Grammar
has been frequency: the more frequent a construction, the easier it should be to
process. We noted above that the light verb constructions under consideration
are more frequent than non-light constructions with the same verbs. This would
suggest a prediction that light verb constructions should require less processing
effort than cognate non-light constructions (and Goldberg has verified [p.c.] that
this is what her analysis predicts).7 The experimental results reported here falsify
this prediction.
However, this prediction is based on the specific co-occurrence frequency
(corresponding to cloze probability) reported in Piñango et al. (to appear) and
Wittenberg and Piñango (2011). Which measures of frequency are most relevant
to this aspect of sentence processing will depend on the underlying theory of
representation. If light verb constructions are represented as broader mappings
that generalize across event nominal (give a kiss, give a punch, give a pinch), then
processing costs could depend on frequency measures that are pooled across this
class, total token frequency across types, the number of types, and the distribu-
tion of token frequencies across these types. Some of these measures could gen-
erate the prediction that light verbs would have greater processing costs – just
like the co-event hypothesis. For example, light verbs have less variability, result-
ing in fewer types. Unlike the co-event hypothesis, however, that cost would not
arise from complex mapping operations, because in the Construction Grammar
framework, all syntactic structure is taken to be a consequence of precompiled
form-meaning pairings. Thus, the higher cost for light verb construction would
entirely arise in the realm of lexical access – a process that is thought to be rap-
id (Embick et al., 2001), contradicting the behavioral data in Piñango et al. (to
appear) and Wittenberg and Piñango (2011), and, in the ERP literature, asso-
ciated with a modulation of the N400 signature (contradicting the findings in
Wittenberg et al., 2014).
A third account of the grammar of light verbs is rooted in Principles and
Parameters theory and the Minimalist Program (Chomsky, 1981, 1995), in par-
ticular the work of Hale and Keyser (1993, 2002). In this configurationally de-
fined approach to argument structure, semantic roles in a sentence are assigned
based on the arguments’ position in the syntactic tree, and the correspondence
74 Eva Wittenberg et al.

between syntactic and semantic roles is strictly homomorphic (Levin and

Rappaport Hovav, 2005). According to this model, a surface verb like walk is de-
rived from an underlying structure [V [DP walk]], which Hale and Keyser gloss
with a light-verb-like construction ‘DO a walk’ or ‘TAKE a walk.’ The verb is an
“abstract V” or, in some formulations, “little v” – the same sort of abstract verb
involved in Larson’s (1988) VP-shells. By head-to-head raising, walk is incorpo-
rated or “conflated” with the abstract verb to form the surface verb in its surface
position.
Notice how this approach bears some resemblance to our treatment of the
light verb and event nominal as specifying co-events: the abstract verb, whose
only semantic content is its specification of argument roles, is combined with the
nominal, which provides the meat of the surface verbs’ semantic content. Howev-
er, in Hale and Keyser’s model, this combination takes place in syntactic structure
rather than in the mapping between syntax and semantics.
The conflation operation is treated as a rule that takes place “in the lexicon”
(Hale and Keyser, 2002: 47). It is not clear whether the conflation is conceived of
as precompiled, so that for all intents and purposes there is a lexical verb walk, or
whether the conflation is considered to be a step that takes place prior to lexical
insertion in the derivation of a sentence. (See Culicover and Jackendoff, 2005, for
a critique of this approach.)
Hale and Keyser (1993, 2002) do not make specific proposals about the
derivation of the light verb construction, but their analysis can be extended in
a straightforward way to account for these constructions. In fact, just such an
extension has been proposed by Folli, Harley, and Karimi (2004) to account for
Persian light verb constructions. Their starting point, however, deviates from
Hale and Keyser’s in one respect: they treat the operation of conflation not as a
“lexical” operation but as part of the syntactic derivation, following Larson (1988)
and many others (see also Jung, 2002, on Korean).
Hale and Keyser’s theory is not intended to make explicit predictions about
the processing of light verb constructions, and neither is Folli et al.’s extension of
it. Nevertheless, we believe that a theory of the linguistic system should have a
bearing on how it is actually put in use through language processing. If we look
at this account from this perspective, it is reasonable to suppose that derivational
complexity in this case should correspond to processing complexity.
Example (11) compares the derivation of a light verb construction with two
non-light verb constructions under Folli et al.’s approach; many details are sim-
plified.
(11) a. Light verb construction:
[V [DP order]] → [give [DP (an) order]] [by spell-out]
 Processing and representation of light verb constructions 75

b. Non-light verb construction with order

[V [DP order]] → [[V+orderi] [DP ti]] [by head movement]
c. Non-light verb construction with give
[V [DP give] [DP an orange]] → [[V+givei] [DP ti] [DP (an) orange]]
[by head movement]

Assuming the spell-out of the abstract V as give in (11a) is a phonological oper-

ation, light verb constructions actually have a simpler syntactic derivation than
non-light constructions, since they involve no operation of head movement. On
this account, then, a non-light verb construction such as walkV to the park would
be syntactically more complex than a light verb construction such as take a walk
to the park, in that it has undergone the derivational step of head-to-head rais-
ing. If derivational complexity were taken to correspond to processing effort, this
would predict that light verb constructions would be easier to process than non-
light. The experiments reported here falsify that prediction; the reverse is the case.
Also, the head movement account says that light and non-light constructions
have different syntax, both in underlying and in surface structure, as can be seen
by comparing (11a) and (11c). However, the structural priming results give us
no evidence that the degree of structural overlap is reduced in these cases. In
short, to the extent that we can derive predictions about processing from the head
movement account, they are all not borne out.

4. Conclusions

This chapter has presented a summary of ongoing research on the processing and
representation of light verb constructions. In these constructions, the light verb
and the event nominal establish a shared argument structure, while the syntactic
structure is indistinguishable from non-light sentences. Light verb constructions
are challenging for linguistic theories because they violate the general pattern of
the verb as the sole predicate, resulting in complex event structures and a mis-
match between syntactic and semantic structure.
Different theoretical attempts have been made to model the mechanism of ar-
gument sharing. One solution is to posit hidden syntactic structure which estab-
lishes a one-to-one mapping from syntactic to semantic arguments and derives
the surface form by head movement (Hale and Keyser, 1993, 2002). Alternatively,
one can conceive of light verb constructions as units that are stored in memory
with fully specified syntactic and semantic information (Goldberg, 2003). Final-
ly, a third approach acknowledges the independence of syntactic and semantic
structure, allowing the mechanism of argument sharing to be a complex semantic
76 Eva Wittenberg et al.

operation, not affecting syntactic structure (Jackendoff, 2002; Culicover and

Jackendoff, 2005).
Each of these representational models can be used to make different pre-
dictions about the processing of light verb constructions: The head movement
model suggests that light verb constructions should be easier to process than non-
light constructions, since in the syntactic interpretation of that theory, light verb
constructions are associated with less structural and derivational complexity. A
constructionalist model makes the same prediction, but bases it on the higher
frequency of the light verb construction, which should facilitate processing. The
Parallel Architecture, however, predicts that the more complex semantic opera-
tions involved in argument sharing would lead to more processing load for un-
derstanding light verb constructions.
Behavioral and electrophysiological processing studies in English and Ger-
man show that the particular way in which light verb constructions are processed
in comprehension is rooted in complex semantics-to-syntax mapping operations.
These studies report increased reaction times (Piñango, Mack, and Jackendoff,
to appear; Wittenberg and Piñango, 2011) and sustained negativities in Event-­
Related Potentials (Wittenberg et al., 2014) to light verb constructions, compared
to standard verb-object constructions. The production priming study (Wittenberg
and Snedeker, in prep.) and electrophysiological evidence (Wittenberg et al., un-
der review) support the view that there is no syntactic cost associated with light
verb constructions, and their syntactic structure does not differ from non-light
constructions.
These findings raise some interesting questions for future research. First,
most of the studies so far have focused on the comprehension of light verb con-
structions, while production has not yet been addressed extensively. It is crucial
to understand what factors lead to the frequent usage of light verb constructions,
despite the complex semantic operations that characterize them. One consider-
ation could be information structure or utterance planning preferences. We are
planning to investigate this by looking at circumstances of light verb production
in corpora. Other aspects include very subtle differences in meaning between a
light verb construction (to give a kiss) and its underlying base verb construction
(to kiss), such as aspectual features, directionality, agentivity, or aktionsart. We
are also investigating this possibility, looking closer into the conceptualization of
events described by light verb constructions.
In short, light verb constructions give us crucial insight into the interface
of syntax and semantics. The richness of theoretical problems associated with
them, together with their centrality in everyday language use, make them a su-
perb object of theoretical investigation. Moreover, light verb constructions pose
interesting questions relevant to research on language processing and its ties to
 Processing and representation of light verb constructions 77

how we conceptualize and express events. We have shown that psycholinguistic

techniques can shed light on these questions, both in terms of processing and
representation.

Notes

1. We thank the participants of the workshop “Structuring the Argument” (Paris 2011), the
editors of this volume and two anonymous reviewers for helpful comments and suggestions.

2. We use the term “event nominal” for convenience; recognizing that the nominal in some
light verb expressions does not denote an event, e.g. have an ability.

3. To what degree is the composite event an act of giving? A great deal of the content of non-
light give is “bleached out” from light give. What remains is at least the grid of thematic roles
and its aspectual/aktionsart features. What there is beyond that is a question we will not try to
resolve here, but we have begun to explore this issue in our ongoing work testing the concep-
tualization of light verb constructions as compared to non-light constructions (Wittenberg and
Snedeker, 2013).

4. Notice that the noun bath cannot itself take a causative argument: *Tom’s bath of Olive is
out, and Tom’s bath can only mean the bath that Tom took, not the bath that he gave to Olive.

5. Note that the telic event can be iterated by pluralizing the event nominal: She gave him kiss-
es for 5 minutes. This parallels similar effects in non-light constructions: She handed him tools
for 5 minutes.

6. As a reviewer points out, it is not clear whether this extra processing should be construed
as a repair or simply as an alternative way of mapping a direct object plus verb into semantics.
In either case, the normal route to interpreting the syntactic combination has to be overridden
and the thematic grids must be aligned. However, the fact that the P600 is not elevated with
grammatical light verbs, unlike anomalous light verbs, suggests that the extra processes are
distinct from those involved in resolving a (syntactic) anomaly.

7. We noted above that some light verb constructions (such as make a decision) have to be
stored in the lexicon – but not all, as Goldberg’s analysis posits. Under the co-event hypothe-
sis, it is an open question whether stored light verb constructions have to undergo argument
sharing, or whether they store the result “pre-compiled.” If the latter, stored light verb construc-
tions might behave differently from those that are processed online. Alternatively, just as for
some frequent multimorphemic words, parallel access and computation routes are conceivable
(Baayen, Dijkstra, and Schreuder, 1999).

References

Alsina, Alex. 1996. The Role of Argument Structure in Grammar. Stanford, California: CSLI
Publications.
78 Eva Wittenberg et al.

Baayen, R. R, Dijkstra, T., and Schreuder, R. 1997. Singulars and plurals in Dutch: Evidence for
a parallel dual route model. Journal of Memory and Language, 36: 94–117.
DOI: 10.1006/jmla.1997.2509
Baker, Mark C. 1989. Incorporation: a theory of grammatical function changing. Chicago: Uni-
versity of Chicago Press.
Bicknell, Klinton and Roger Levy. 2012. Word predictability and frequency effects in a rational
model of reading. Proceedings of the 34th Annual Meeting of the Cognitive Science Confer-
ence.
Bock, Kay. 1986. Syntactic persistence in language production. Cognitive Psychology, 18: 355–
387. DOI: 10.1016/0010-0285(86)90004-6
Bock, Kay. 1989. Closed-class immanence in sentence production. Cognition, 31: 163–186.
DOI: 10.1016/0010-0277(89)90022-X
Bock, Kay, and Loebell, Helga. 1990. Framing sentences. Cognition, 35: 1–39.
DOI: 10.1016/0010-0277(90)90035-I
Boland, Julie. 1997. The relationship between syntactic and semantic processes in sentence
comprehension. Language and Cognitive Processes, 12(4): 423–484.
DOI: 10.1080/016909697386808
Bott, Oliver. 2010. The Processing of Events. Philadelphia, PA: John Benjamins Publishing Com-
pany. DOI: 10.1075/la.162
Briem, Daniela, Balliel, Britta, Rockstroh, Brigitte, Butt, Miriam, Schulte im Walde, Sabine and
Assadollahi, Ramin. 2010. Distinct processing of function verb categories in the human
brain. Brain Research, 1249: 173–180. DOI: 10.1016/j.brainres.2008.10.027
Brugman, Claudia. 2001. Lights verbs and polysemy. Language Sciences, 23: 551–578.
DOI: 10.1016/S0388-0001(00)00036-X
Butt, Miriam. 1995. The Structure of Complex Predicates in Urdu. Stanford, California: CSLI
Publications.
Butt, Miriam. 2010. The Light Verb Jungle: Still Hacking Away. In Complex Predicates in
Cross-Linguistic Perspective, Mengistu Amberber, Mark Harvey and Brett Baker (eds),
48–78. Cambridge University Press.
Chang, Franklin, Bock, Kay, and Goldberg, Adele. 2003. Can thematic roles leave traces of their
places? Cognition, 90: 29–49. DOI: 10.1016/S0010-0277(03)00123-9
Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, Noam. 1995. The Minimalist Program. MIT Press.
Croft, William. 2001. Radical construction grammar: syntactic theory in typological perspective.
Oxford; New York: Oxford University Press.
DOI: 10.1093/acprof:oso/9780198299554.001.0001
Culicover, Peter, and Jackendoff, Ray. 2005. Simpler Syntax. Oxford University Press.
DOI: 10.1093/acprof:oso/9780199271092.001.0001
Durie, Mark. 1988. “Verb serialization and verbal prepositions”. Oceanic Linguistics, 27: 1–23.
DOI: 10.2307/3623147
Embick, D., Hackl, M., Schaeffer, J., Kelepir, M., & Marantz, A. 2001. A magnetoencephalo-
graphic component whose latency reflects lexical frequency. Cognitive Brain Research,
10(3): 345–348. DOI: 10.1016/S0926-6410(00)00053-7
Family, Neiloufar. 2009. Mapping semantic spaces: A constructionist account of the light verb
eat in Persian. In From polysemy to semantic change: Towards a typology of lexical semantic
associations. M. Vanhove (ed.). Amsterdam: John Benjamins Publishing Company.
 Processing and representation of light verb constructions 79

Folli, Raffaella, Harley, Heidi and Karimi, Simin. 2004. Determinants of event type in Persian
complex predicates. Lingua, 115(10): 1365–1401. DOI: 10.1016/j.lingua.2004.06.002
Fried, Mirjam and Östman, Jan-Ola. 2004. Construction Grammar: A thumbnail sketch. In
Construction Grammar in a cross-language perspective. Mirjam Fried and Jan-Ola Östman.
11–86. Amsterdam: John Benjamins. DOI: 10.1075/cal.2
Goldberg, Adele. 1995. Constructions: A Construction Grammar Approach to Argument Struc-
ture. University of Chicago Press.
Goldberg, Adele. 2003. Words by default: The Persian complex predicate construction. In
Mismatch: form-function incongruity and the architecture of grammar, Elaine Francis and
Laura Michaelis (eds.), 117–149. Stanford, CA: CSLI Publications.
Goldberg, Adele. 2009. The Nature of Generalization in Language. [Target Article] Cognitive
Linguistics, 20(1): 93–127.
Hale, Kenneth L., and Keyser, Samuel J. 1993. On argument structure and the lexical expression
of syntactic relations. The view from building 20: Essays in linguistics in honor of Sylvain
Bromberger, 53–100. Cambridge, MA: MIT Press.
Hale, Kenneth L., and Keyser, Samuel J. 2002. Prolegomenon to a theory of argument structure.
Cambridge, MA: MIT Press.
Hofmeister, Philip and Sag, Ivan. 2010. Cognitive constraints and island effects. Language, 86:
366–415. DOI: 10.1353/lan.0.0223
Jackendoff, Ray. 1974. A deep structure projection rule. Linguistic Inquiry, 5(4): 481–505.
Jackendoff, Ray. 1991. Parts and boundaries. Cognition 41: 9–45.
DOI: 10.1016/0010-0277(91)90031-X
Jackendoff, Ray. 1997. The Architecture of the language faculty. Cambridge: MIT Press.
Jackendoff, Ray. 2002. Foundations of Language. Oxford University Press.
DOI: 10.1093/acprof:oso/9780198270126.001.0001
Jackendoff, Ray. 2007. A Parallel Architecture Perspective on Language Processing. Brain Re-
search, 1146: 2–22. DOI: 10.1016/j.brainres.2006.08.111
Jung, Dukkyo. 2002. Light Verb just as a Little v. Kansas Working Papers in Linguistics, 59–74.
Kay, Paul. 1995. Construction Grammar. In Handbook of Pragmatics. Jef Verschueren, Jan-Ola
Östman, and Jan Blommaert (eds), 171–177. Amsterdam/Philadelphia: John Benjamins.
DOI: 10.1075/hop.m.con2
Larson, R. 1988. On the Double Object Construction. Linguistic Inquiry, 19: 3.
Levin, Beth and Malka Rappaport Hovav. 2005. Argument Realization. Cambridge: Cambridge
University Press. DOI: 10.1017/CBO9780511610479
McElree, Brian, and Griffith, Teresa. 1995. Syntactic and thematic processing in sentence com-
prehension: Evidence for a temporal dissociation. Journal of Experimental Psychology:
Learning, Memory and Cognition, 21(1): 134–157. DOI: 10.1037/0278-7393.21.1.134
McGinnis, M. 2002. On the systematic aspect of idioms. Linguistic Inquiry, 33.4: 665–672.
DOI: 10.1162/ling.2002.33.4.665
Newman, John. 1996. Give: A Cognitive Linguistic Study. (Cognitive Linguistics Research 7.)
Berlin: Mouton de Gruyter.
Paczynski, Martin, and Kuperberg, Gina. To appear. When events change their nature.
Palmer, Martha, Gildea, David, and Kingsbury, Paul. 2005. The proposition bank: An annotated
corpus of semantic roles. Computational Linguistics, 31(1): 71–206.
DOI: 10.1162/0891201053630264
80 Eva Wittenberg et al.

Phillips, Colin, and Lewis, C. To appear. Derivational order in syntax: Evidence and architec-
tural consequences. In Directions in Derivations, C. Chesi (ed.). Elsevier.
DOI: 10.4153/CJM-1981-105-7
Pickering, Martin, and Ferreira, Victor. 2008. Structural priming: A critical review. Psychologi-
cal Bulletin, 134: 427–459. DOI: 10.1037/0033-2909.134.3.427
Piñango, Maria M., Mack, Jennifer, and Jackendoff, Ray. To appear. Semantic combinatorial
processes in argument structure: Evidence from light verbs. Proceedings of the annual
meeting of the Berkeley Linguistics Society, 2006.
Pinker, Steven. 1999. Words and Rules. New York: Perennial.
Talmy, Leonard. 1978. The relation of grammar to cognition: A synopsis. Theoretical Issues in
Natural Language Processing 2, D. Waltz (ed.). New York: Association for Computing Ma-
chinery. Revised version in Talmy, Toward a Cognitive Semantics, MIT Press.
Thothathiri, Malathi and Snedeker, Jesse. 2012. The Role of Thematic Roles in Sentence Pro-
cessing: Evidence from Structural Priming in Young Children. Unpublished Manuscript.
Wierzbicka, Anna. 1982. Why can you have a drink when you can’t *have an eat? Language,
58(4): 753–799. DOI: 10.2307/413956
Wiese, Heike. 2006. „Ich mach dich Messer“ – Grammatische Produktivität in Kiez-Sprache
(‚Kanak Sprak‘). Linguistische Berichte, 207: 245–273.
Winhart, Heike. 2002. Funktionsverbgefüge im Deutschen. Zur Verbindung von Verben und
Nominalisierungen. Dissertation, Universität Tübingen.
Wittenberg, Eva, and Piñango, Maria M. 2011. Processing light verb constructions. The Mental
Lexicon, 393–413. DOI: 10.1075/ml.6.3.03wit
Wittenberg, Eva, and Snedeker, Jesse. 2013. It takes two to kiss, but does it take three to give a
kiss? Categorization based on thematic roles. Language and Cognitive Processes, (ahead-
of-print), 1–7.
Wittenberg, Eva, and Snedeker, Jesse. In preparation. Syntactic Priming Across Constructions:
Light Verbs.
Wittenberg, Eva, Paczynski, Martin, Wiese, Heike, Jackendoff, Ray and Kuperberg, Gina. 2014.
The difference between “giving a rose” and “giving a kiss”: Sustained neural activity to the
light verb construction. Journal of Memory and Language, 73C: 31–42.
Part Ii

Non-canonical argument structure realization

 Luigi piace a Laura?
Electrophysiological evidence for thematic reanalysis
with Italian dative object experiencer verbs

Alexander Dröge,1 Laura Maffongelli2

and Ina Bornkessel-Schlesewsky1,3
1 Department of Germanic Linguistics, University of Marburg,
Marburg, Germany / 2 RBCS – Robotics, Brain and Cognitive Sciences,
IIT – Italian Institute of Technology, Genova, Italy / 3 School of Psychology,
Social Work and Social Policy, University of South Australia, Adelaide,
Australia

The syntactic properties of psych verbs have been debated in theoretical lin-
guistics since the seminal paper by Belletti and Rizzi (1988). However, surpris-
ingly little is known about the neural processes underlying the comprehension
of psych verb constructions. Here, we report an electrophysiological study
on Italian piacere-class verbs, which were presented in sentences with sub-
ject-verb-object (SVO) and object-verb-subject (OVS) orders and contrasted
with agent-theme (“active”) verbs. At the verb position, we observed a biphasic
N400–late positivity pattern for active versus piacere-class verbs in object-initial
orders and a late positivity for piacere-class versus active verbs in subject-ini-
tial orders. These results demonstrate that thematic expectations are generated
incrementally and may be based upon only a single argument. They further
support the idea of structural differences between piacere-class verbs and other
verb classes in Italian and suggest that these are used rapidly to inform language
processing.

1. Introduction

In linguistic research, psychological (“psych”) verbs are of great importance both

from a theoretical and a cognitive perspective. In contrast to agentive verbs such
as kill or write, psych verbs do not assign the thematic roles agent and patient, but
rather express some psychological state and take an experiencer as one of their ar-
guments (Primus 2004: 377). The roles agent and experiencer are assumed to rank
84 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

higher in the thematic hierarchy than the patient/theme role (e.g., Grimshaw 1990;
Pesetsky 1995; Primus 1999). Depending on the type of psych verb, argument link-
ing differs substantially. Belletti and Rizzi (1988: 291f.) distinguish three classes of
psych verbs in Italian:1 psych verbs with the experiencer mapping onto the subject,
called temere-class (1a), those with the experiencer surfacing as an accusative ob-
ject, called preoccupare-class (1b), and those with the experiencer being realized as
a dative object, called piacere-class (1c–d).2 As shown in (1c–d), piacere-­class verbs
are the only ones that allow for both SVO and OVS orders in Italian.
(1) Classes of psych verbs (Belletti and Rizzi 1988: 291f.):
a. Gianni teme questo.
Gianni fears this
b. Questo preoccupa Gianni.
this worries Gianni
c. A Gianni piace questo.
to Gianni pleases this
d. Questo piace a Gianni.
this pleases to Gianni

Psych verbs have also been the focus of a number of previous psycholinguistic
studies. In contrast to the theoretical literature, however, these have primarily
investigated the processing of psych verbs in languages such as English or Ger-
man (e.g., Bornkessel, McElree, Schlesewsky and Friederici 2004; Bornkessel,
Schlesewsky and Friederici 2003; Bornkessel, Zysset, Friederici, von Cramon
and Schlesewsky 2005; Corrigan 1988; Ferreira 1994; Kretzschmar, Bornkessel-­
Schlesewsky, Staub, Roehm and Schlesewsky 2012; Schlesewsky and Bornkessel
2006; Thompson and Lee 2009). By contrast, though they are well investigated in
the theoretical literature, little is known about the processing of psych verb con-
structions in Italian, or Romance languages in general (for Spanish, see Gattei,
Vasishth and Dickey 2011).
In this chapter, we attempt to bridge the gap between previous work on psych
verbs in theoretical linguistics and psycho­linguistics/neuro­linguistics by exam-
ining how piacere-class verb constructions differ from constructions with dative
active verbs (i.e., verbs which project the higher thematic role onto the grammat-
ical subject) and how this influences online language comprehension. Our goal
is to link linguistic theory to neurolinguistic data. To this end, we first provide
a detailed theoretical description of piacere-class verbs and dative active verbs
in Italian, taking into account syntactic, semantic and information structural
factors (Section 2).3 In Section 3, these theoretical bases – in combination with
a cross-linguistically oriented model of language comprehension – are used to
formulate hypotheses regarding online comprehension. These were tested in an
 Thematic reanalysis with Italian dative object experiencer verbs 85

experimental study using event-related brain potentials (ERPs), which is report-

ed in Section 4. Our results show a close correspondence between the theoret-
ical analysis of piacere-class versus dative active verbs in Italian and the way in
which these constructions are processed in real time. In particular, they provide
evidence that the language processing system endeavors to (a) analyze the first
argument it encounters as bearing the highest ranking thematic role (differing
between active and piacere-class verbs); and (b) to assume a neutral information
structure in the absence of evidence to the contrary, as revealed by additional
processing costs for OVS orders with active verbs. The chapter concludes with a
discussion of these results in Section 5.

2. Theoretical considerations

It is instructive to first examine some general structural differences between Ital-

ian and English. While SVO is the canonical constituent order of both languages,
Italian displays a greater range of word order variations. It has been shown in
psycholinguistic experiments that English speakers mostly rely on word order
to identify semantic roles, whereas interpretation of Italian sentences is strong-
ly dependent on agreement relations, animacy and stress (Bates, McNew, Mac­
Whinney, Devescovi and Smith 1982; MacWhinney, Bates and Kliegl 1984).
English and Italian have no inflectional case marking on nouns, but in contrast
to English, Italian has a rich system of verbal morphology allowing overt agree-
ment in all persons. Italian verbal morphology has been taken as one factor that
licenses pro-drop, i.e., non-emphasized subject pronouns are not realized overtly
(cf. Rizzi 1982; 1986). In Italian, the empty category pro is also an important de-
vice to license post-verbal subjects (for a critical discussion, see Belletti 2001).

2.1 Analysis of psych verb constructions

Subject experiencer constructions have generally been assumed to have a struc-

ture similar to regular transitive stative verbs such as know (e.g., Bennis 2004).
However, there is some dissent concerning the representation of object expe-
riencer constructions. While some scholars regard object experiencer verbs as
an instance of unaccusative/ergative verbs (e.g., Belletti and Rizzi 1988; Bennis
2004), others argue that at least accusative object experiencer verbs should be
analyzed as a type of causative (e.g., Baker 1997; cf. also discussions in Landau
2010; Pesetsky 1995). This difference has substantial semantic and syntactic im-
plications because unaccusatives and causatives are assumed to involve different
thematic roles, and unaccusatives lack a v-projection and an external argument.
86 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

In the following, we focus mainly on the analysis of piacere-class verbs in com-

parison to subject experiencer or active verbs, since this distinction is most rele-
vant for the present study. The analysis of accusative object experiencer verbs, by
contrast, will only be discussed in passing (for a more detailed discussion of these
verbs, see, e.g., Bennis 2004; Landau 2010).
Working within the Government and Binding framework (see Chomsky 1981;
and subsequent work), Belletti and Rizzi (1988) argue that subject and object expe-
riencer verb constructions employ the same θ-roles. The theme is θ-marked direct-
ly by the verb while the experiencer is θ-marked compositionally by the verb and
the theme. The verb classes differ in their Case-grids and the respective θ-relations
between inherent Case and θ-role. Subject experiencer verbs, i.e., temere-class
verbs, do not have any inherent Case; thus, the verb assigns structural accusative
Case to its complement. Object experiencer verbs, on the other hand, contain an
entry for inherent Case in their Case-grid: accusative with preoccupare-class verbs
and dative with piacere-class verbs (Belletti and Rizzi 1988: 344). The lexical en-
try for subject experiencer verbs also marks the experiencer role as the external
argument. With object experiencer verbs there is no external argument, but both
arguments map onto positions within VP. Belletti and Rizzi assume that the expe-
riencer must always c-command the theme; hence, in the case of object experienc-
er verbs, the experiencer is projected to a configurational position higher in the VP
than the theme, as illustrated in Figure 1 (from Belletti and Rizzi 1988: 293).
In the phrase marker in Figure 1, the complement of V should receive struc-
tural accusative Case; however, following Belletti and Rizzi (1988: 332), in the case
of object experiencer verbs Case assignment is blocked due to a reformulation of
Burzio’s generalization (see (2); cf. Burzio 1986). Thus, object experiencer verbs
are a type of unaccusative verbs, for they lack an external argument and do not
assign structural Case to their complement.

NP VP

ec Vʹ NP

V NP Gianni
a Gianni
preoccupa questo
piace

Figure 1. Phrase marker of D-structures of preoccupare-class and piacere-class verbs

(from Belletti and Rizzi 1988: 293)
 Thematic reanalysis with Italian dative object experiencer verbs 87

(2) Reformulation of Burzio’s generalization (Belletti and Rizzi 1988: 332):

V is a structural Case assigner iff it has an external argument.

In an analysis couched in a Minimalist framework (see Chomsky 1995; 2000;

2001; and other work), Bennis (2004: 88) analyzes dative object experiencer (i.e.,
piacere-class) constructions as what he terms “simplex ergative” constructions.
These lack a vP-shell (i.e., have no external argument) and cannot assign accusa-
tive Case to the internal argument.4
In accordance with the analyses by Belletti and Rizzi (1988) and Bennis (2004),
we assume here that the experiencer role of piacere-class verbs is linked to the da-
tive object, while the subject bears the lower role, i.e., the theme. Thus, the experi-
encer is the thematically more prominent role, bears some subject-like properties
and is projected onto a structurally higher position in the phrase marker.
This assumption, which is crucial for the hypotheses of our current experi-
ment, finds support in neurolinguistic studies on the processing of psych verbs in
German: EEG results show that dative object experiencer verbs, which are com-
parable to the Italian piacere-class, engender thematic reanalysis effects when en-
countered clause-finally (Bornkessel et al. 2003) and render a reanalysis towards
an object-initial reading less costly in locally ambiguous sentences (Bornkessel et
al. 2004; Schlesewsky and Bornkessel 2006; for converging results from fMRI and
eye-tracking, see Bornkessel et al. 2005; Kretzschmar et al. 2012, respectively).

2.2 Implications for word order

Having argued that piacere-class verbs differ from dative active verbs in that they
project the higher thematic role onto the object as opposed to the subject, we will
now discuss the implications of this difference for word order variations and their
structural analysis. Belletti and Rizzi (1988) point out that, of all psych verbs, only
piacere-class verbs may surface with either SVO or OVS order in Italian. This also
holds for constructions involving dative active verbs. Thus, fronting of a dative
object is generally possible because it is realized as an indirect object with inher-
ent Case that is licensed by a semantically void preposition, e.g., a Laura (Belletti
and Rizzi 1988: 336).5 However, as we will discuss in the following, the two orders
differ in terms of syntax as well as information structure depending on the type of
verb involved. These differences will be crucial for our hypotheses regarding the
processing of such constructions.
According to Belletti and Rizzi (1988) and Rizzi (2005; 2006), dative-initial
sentences show a different syntactic behavior with a piacere-class verb as opposed
to a dative active verb. Specifically, island effects indicate that a fronted dative
occupies a different position depending on the verb class. The fronted dative
88 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

experiencer in (3a) seems to constitute the unmarked order, while the other initial
datives in (3b–c) lead to less acceptable structures (Rizzi 2006: 120f.).
(3) Fronted datives showing island effects (Rizzi 2006: 120f.):
a. Le idee che a Gianni piacciono di più sono queste.
The ideas that to Gianni please most are these
b. ??Le idee che a Gianni Maria raccomanda sono queste.
The ideas that to Gianni Maria recommends are these
c. ?Le idee che a Gianni raccomandiamo sono queste.
The ideas that to Gianni we-recommend are these

Following Bennis’s (2004) analysis, we assume that piacere-class verbs instanti-

ate “simplex ergative” constructions lacking a vP-layer and an external argument;
thus, both the theme and the experiencer are projected to VP-internal positions.
Belletti and Rizzi (1988) locate the initial dative of experiencer verbs in the struc-
tural subject position, i.e., the specifier of S, which would be [Spec,TP] in current
terminology.6 The phrase marker of an object-initial piacere-construction is given
in Figure 2.7
By contrast, the possibility of fronting the dative argument of an active verb
is constrained by information structural properties and, specifically, requires a
discourse context that renders the dative a topic (Rizzi 2005). In accordance with
this observation, Rizzi (2005; 2006) proposes that the fronted dative surfaces in
a topic position in the CP-layer and an expletive pro is merged into [Spec,TP] to
value the EPP-feature (cf. Rizzi 1997; 2004; 2005; 2006). Therefore, initial datives
end up in structurally distinct positions for the two verb classes under consid-
eration, forming an A-chain in piacere-constructions and an A′-chain in active
verb constructions (cf. Rizzi 2005). This structural difference reflects the basic

TP

Spec Tʹ

a Laura V+T VP

piace Spec Vʹ

a Laura V Compl

piace Luigi

Figure 2. Phrase marker of an OVS sentence containing a piacere-class verb

 Thematic reanalysis with Italian dative object experiencer verbs 89

observation that, only for piacere-class verbs, the clause-initial positioning of a

dative is licensed by thematic properties, while, for dative active verbs, it is con-
strained by information structure. As it has long been assumed in psycholinguis-
tic research that the language processing system prefers readings which require
fewer assumptions about the preceding discourse (Altmann and Steedman 1988;
Crain and Steedman 1985), this difference has important implications for the
processing choices that may be elicited by an initial dative argument.
Having discussed the position of initial datives in OVS sentences, we still need
to explicate where the post-verbal subject is located in these constructions. One
possible assumption is that post-verbal subjects remain in their base-generated
position (i.e., [Spec,vP] for active verbs and complement of VP for piacere-class
verbs, respectively), with their nominative Case licensed via Agree. This analysis
is possible with piacere-class verbs, but it would not adequately reflect the specif-
ic information structural requirements for OVS orders with dative active verbs.
Thus, Belletti (2001; 2004) argues in favor of a Focus category above the highest
VP shell to host the post-verbal subject, since these subjects convey new, focused
information. Including this TP-internal Focus projection into the configuration,
we can build up the phrase marker for object-initial active verb sentences given
in Figure 3.8
For the sake of completeness, the two phrase markers for SVO sentences with
experiencer verbs and active verbs are given in Figures 4 and 5, respectively. The
indirect object stays in situ (its Case is inherent and need not be checked/valued),

TopP

Spec Topʹ

a Laura Top TP

Spec Tʹ

proexpl V+v+T FocP

scrive Spec Focʹ

Luigi Foc vP

Luigi scrive a Laura

Figure 3. Phrase marker of an OVS sentence containing a dative active verb

90 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

TP

Spec T′

Luigi V+T VP

piace Spec V′

a Laura V Compl

piace Luigi

Figure 4. Phrase marker of an SVO sentence containing a piacere-class verb

TP

Spec T′

Luigi V+v+T vP

scrive Spec V′

Luigi V+v VP

scrive V Compl

scrive a Laura

Figure 5. Phrase marker of an SVO sentence containing a dative active verb

and the subject-DP moves to the subject position [Spec,TP], where the EPP-fea-
ture is valued and nominative Case is licensed.
To summarize the structural differences between OVS and SVO orders with
dative active and dative object experiencer verbs, the most complex configuration
is OVS with an active verb, because – for information structural reasons – an
EPP-feature in Top attracts the indirect object, which therefore needs to move
to [Spec,TopP]; the focused subject additionally moves from [Spec,vP] to [Spec,
FocP]. The unmarked structure of active verbs in an SVO order is less complex,
since no topicalization or focalization is required. In sentences with piacere-class
 Thematic reanalysis with Italian dative object experiencer verbs 91

verbs, the configurations of both word order variants are quite similar, because
either the experiencer or the theme surfaces in the structural subject position
[Spec,TP] while the other constituent stays in situ. However, only in OVS order,
the thematically higher experiencer argument c-commands the thematically low-
er theme argument; hence, OVS is expected to be the unmarked order for experi-
encer verbs, in accordance with Belletti and Rizzi’s (1988) analysis.

3. Processing sentences with object experiencer verbs

Having provided a relatively elaborate syntactic description of sentences with

dative active verbs and dative object experiencer verbs in Italian, we will now
use these assumptions to formulate hypotheses regarding the processing of these
constructions in real time. Crucially, sentence comprehension is an incremental
process, i.e., the comprehension system does not wait until the end of the sentence
to begin interpretation, but rather integrates each element as soon as it is avail-
able (cf. Crocker 1994). Furthermore, incremental processing involves setting up
predictions for upcoming elements (e.g., Kamide 2008; Kutas, DeLong and Smith
2011; Van Berkum, Brown, Zwitserlood, Kooijman and Hagoort 2005; Wicha,
Moreno and Kutas 2004). The central question to be examined here is thus: how
are predictive and integrative processes during incremental interpretation influ-
enced by verb class and, in particular, do piacere-class verbs show any special
properties in this regard?
Building on the preceding section, we assume that the following observa-
tions are relevant to this question: (a) OVS sentences with dative active verbs are
structurally more complex than their SVO counterparts and SVO/OVS orders
with piacere-class verbs; (b) piacere-class verbs map the higher θ-role (experi-
encer) to the object, while dative active verbs map the higher θ-role (agent) to
the subject. As we will argue in the following, both of these properties should
be expected to play a crucial role in the online processing of dative active and
piacere-class constructions. On the one hand, the assumption that the process-
ing system prefers structurally less complex readings is well established in the
psycholinguistic literature (e.g., Bornkessel and Schlesewsky 2006; Fodor 1998;
Frazier and Fodor 1978). On the other hand, evidence for a preference to analyze
the first argument as bearing the highest role stems from ERP results in a number
of languages with flexible word order, including German (Bornkessel et al. 2003)
and Japanese (Wolff, Schlesewsky, Hirotani and Bornkessel-Schlesewsky 2008).
Thus, we expect effects for OVS versus SVO sentences with dative active verbs due
to the information structurally-driven complex sentence structure of the OVS
order, whereas a thematic effect is anticipated for the difference between SVO and
92 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

OVS orders with piacere-class constructions, which are of comparable syntactic

complexity.
We formulate our predictions within the framework of the extended Argu-
ment Dependency Model (eADM; Bornkessel and Schlesewsky 2006; Bornkessel-­
Schlesewsky and Schlesewsky 2008a; 2009a), a neurocognitive model of
cross-linguistic sentence comprehension. Since a full description of the model is
beyond the scope of this chapter, we focus only on the most relevant assumptions
(pertaining to stage 2 of the overall model). The eADM posits that arguments
encountered in the absence of verbal information are assigned a generalized se-
mantic role (actor or undergoer) in accordance with their relative prominence
(“Compute Prominence”). Prominence is established via a range of semantic and
morphosyntactic features (e.g., animacy, definiteness, case marking, word order)
and their language-specific weighting. When verb-based information is available,
by contrast, verbs and arguments are integrated with one another (“Compute
Linking”). To this end, the arguments are linked via their (actor/undergoer) roles
to the decomposed semantic structure assumed to be stored in the lexical entry
of the verb.
Three aspects of this overall interpretation process are crucial for present
purposes:

a. The processing system follows a “least effort” principle (Minimality;

Bornkessel and Schlesewsky 2006: 790), according to which it assumes the
simplest possible structure and interpretation compatible with the input.
b. With regard to role assignments, the actor is privileged, i.e., the system assumes
that the first argument encountered bears the highest role (for a comprehen-
sive discussion and cross-linguistic evidence, see Bornkessel-Schlesewsky
and Schlesewsky 2009a).
c. The prominence information associated with the arguments is used to as-
sess role prototypicality (e.g., an inanimate initial argument will initially be
taken as the actor in accordance with the actor preference, but assessed as a
non-prototypical actor). This information is used to predict semantic proper-
ties of the verb, i.e., verb class (see Bornkessel-Schlesewsky and Schlesewsky
2008b; 2009a).

These assumptions make the following predictions for the processing of Italian
sentences: an initial argument should be interpreted as bearing the highest role ir-
respective of whether it is marked for dative or not; at the same time, the process-
ing system should opt for the simplest possible structure, i.e., the object-initial
structure for active verbs should be dispreferred. If these assumptions are cor-
rect, an initial dative (e.g., a Laura) should lead the processing system to expect a
piacere-class verb rather than an active verb. By contrast, no such expectation
 Thematic reanalysis with Italian dative object experiencer verbs 93

should be set up in a subject-initial structure. These hypotheses were tested in the

present study.

4. The present study

We conducted an ERP experiment and two internet-based questionnaire studies

to investigate the processing of Italian dative experiencer constructions. To this
end, we used piacere-class experiencer verbs (EXP) and active verbs (ACT) sub-
categorizing for a dative object. In the ERP experiment, we manipulated word
order and verb class yielding the four critical conditions illustrated in Table 1.
As described above, based on previous findings in German and Japanese
(Bornkessel et al. 2003; Wolff et al. 2008), we assume that the first argument is
always assigned the highest thematic role. Thus, the conditions OVS+EXP and
SVO+ACT should not cause any problems for role assignment when the verb
is reached. Both conditions are thematically unmarked because the higher role
precedes (and c-commands) the lower role; they are also syntactically and infor-
mation structurally unmarked.
Condition SVO+EXP shows the canonical order of subject and object, but
produces a problem with respect to the thematic hierarchy because the theme
precedes the experiencer. Therefore, this condition should be rather acceptable
but elicit effects due to a thematic mismatch (required thematic reanalysis of the
initial argument to the lower role) at the position of the verb.
Condition OVS+ACT calls for a licensing context. Here, the initial dative
should lead to the expectation for an object experiencer verb, hence yielding a
thematic reanalysis once the verb is reached. In addition, the structure is highly
marked since it requires a topicalized object and a focused post-verbal subject.
In summary, we expect to observe a thematic mismatch/thematic reanalysis
effect at the verb position for SVO+EXP and OVS+ACT. In accordance with the

Table 1. Example sentences for the four critical conditions in the ERP study
Condition Example sentence
SVO+EXP Piero piace a Matilde.
Piero appeals to Matilde
OVS+EXP A Piero piace Matilde.
to Piero appeals Matilde
SVO+ACT Piero scrive a Matilde.
Piero writes to Matilde
OVS+ACT A Piero scrive Matilde.
to Piero writes Matilde
94 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

eADM, this effect, which is due to “Compute Linking”, should be expected to

manifest itself as an N400. Particularly in OVS+ACT, it may be followed by a late
positivity which reflects the reduced wellformedness of the overall structure.
There is one further complicating factor, namely that predictions during
language comprehension not only involve abstract linguistic information such
as morphosyntactic categories or thematic roles, but can also be very specific in
terms of lexical meaning. If the prior context leads to the expectation of a cer-
tain item and this prediction is not borne out, this typically yields an N400 (e.g.,
DeLong, Urbach and Kutas 2005; Federmeier and Kutas 1999; Kutas and Hillyard
1984). Since the dative object experiencer class only contains a small number of
verbs, we assume that an initial dative may impose a stronger degree of lexical
constraint on the following verb than an initial nominative and that this may
affect our ERP results.
In the following, we first report two questionnaire studies, which were de-
signed to assess whether our assumptions about lexical and thematic predictabili-
ty and word order preferences are correct, before turning to the ERP study.

4.1 Questionnaire study 1 – Sentence completion

Questionnaire 1 used a very simple sentence completion task to examine the de-
gree of lexical and thematic predictability imposed by an initial dative argument
in contrast to an initial nominative argument in Italian. The internet-based ques-
tionnaire consisted of six items, each comprising a single phrase. The task was to
complete each item to form a simple but complete and meaningful sentence. For
example, given a single phrase such as Sonia, a sentence like Sonia mangia un
panino “Sonia is eating a sandwich” could be constructed.

Participants
Fifty-five participants (all monolingual native speakers of Italian; 23 female; mean
age 29.8 years; age range 20–64 years) entered the final data analysis. Two addi-
tional participants were excluded because they did not produce complete (finite)
sentences.

Materials and procedure

For the items including nominative and dative phrases, we only used proper
names in order to exclude effects due to differences in animacy or definiteness.
The participants saw one item at a time, i.e., a single phrase such as Sonia (nomi-
native) or a Sonia (dative). There were four versions of the questionnaire that were
balanced with respect to the relative order of dative and nominative items and the
 Thematic reanalysis with Italian dative object experiencer verbs 95

gender of dative and nominative arguments. Every version contained six items:
one nominative phrase, one dative phrase and four filler phrases, e.g., an adverbial
like in primavera “in spring”. For illustration purposes, all items used in version 1
are listed in Table A1 in Appendix.

Results
Out of the 55 initial dative phrases, 52 were completed using a dative experiencer
verb (EXP; see Figure 6). Of these, 49 were a form of the verb piacere “to appeal”,
the others being mancare “to miss” (two cases) and riuscire “to succeed” (one
case). In three cases, ditransitive or reflexive active verbs were used (dire “to tell”
in one case; regalare “to give as a present” in one case; rimproverarsi “to blame
oneself ” in one case). The low relative entropy (Hrel = 0.287) indicates that the
choice of lexical item was highly biased in favor of the verb piacere.
All 55 nominative items were completed with some sort of active verb (ACT),
i.e., a non-object experiencer verb, including subject experiencer verbs. However,
the completion showed a great range of lexical and syntactic diversity. A copula
construction containing the verb essere “to be” was used in most of the cases (13
occurrences), followed by the transitive verb mangiare “to eat” (6 occurrences)
and the motion verb andare “to go” (5 occurrences). Six verbs occurred two or
three times; 17 verbs had only single occurrences. A high value of relative entropy
(Hrel = 0.871) suggests that there is no clear lexical prediction after an initial nom-
inative. Interestingly, despite the variety of active verbs used after a nominative,

60

50

40
Occurrences

EXP verb
30
ACT verb

20

10

0
Dative Nominative
First phrase

Figure 6. Verb classes selected after the initial item

96 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

there was no occurrence of a bivalent dative active verb, and only a single occur-
rence of a trivalent active verb, namely prestare “to lend”.
Figure 6 shows the distribution of the verb classes selected after the first item.
Complete lists of all verbs chosen by the participants are given in Tables A2 and
A3 in Appendix.

Discussion
The results of the completion task confirm our hypothesis that an initial dative
leads to the expectation for an object experiencer verb. By contrast, the high vari-
ability of continuations in the nominative condition shows that a sentence-initial
subject is not very constraining. Finally, in addition to predicting the verb class,
initial datives also induce a higher degree of lexical predictability than initial
nominatives, as demonstrated by the high number of piacere continuations in
this condition. We assume that, even though these results were produced in a
production study, they are nevertheless informative for sentence comprehension
(see Gennari and MacDonald 2008, for the close relation between sentence com-
pletion and sentence comprehension results).

4.2 Questionnaire study 2 – Sentence building

Questionnaire 2 was a permutation task consisting of 36 sentences. Each sentence

was given as a list of three phrases, with two nominal arguments and one verb.
The task of the internet-based questionnaire was to order the elements to yield a
grammatical and meaningful sentence. With this, we wanted to assess word order
preferences for different verb classes.

Participants
Twenty-six participants (all monolingual native speakers of Italian; 14 female;
mean age 31.2 years; age range 20–61 years) entered the final data analysis. Two
additional participants were excluded (one because of reported bilingualism, one
because of misinterpretation of the task).

Materials and procedure

Each sentence was to be constructed from three given items. The lists of items
were presented vertically, with the verb either the first or last element in order to
avoid any bias, since all outcome sentences were expected to be verb-medial. For
example, given a set like manca / Raimondo / a Marisa, a sentence like A Marisa
manca Raimondo “Marisa misses Raimondo” could be constructed. The critical
conditions contained either a piacere-class experiencer verb (EXP) or a dative
 Thematic reanalysis with Italian dative object experiencer verbs 97

active verb (ACT) in third person singular form, and two proper names appear-
ing as the dative phrase and the nominative phrase. The dative object experiencer
verbs chosen for the study were piacere “to appeal”, mancare “to be missing”, in-
teressare “to take an interest”, and importare “to be important”; the dative active
verbs were scrivere “to write”, badare “to take care”, mentire “to lie”, and credere “to
believe”.
Experiencer verbs were chosen from a list of piacere-class verbs provided by
Kailuweit (2005). The list of verbs of this class is very small, and especially verbs
that allow two animate arguments are very rare. Furthermore, many verbs in his
list are regionally marked or very low in frequency. Therefore, we could only use
four verbs from Kailuweit’s list, and these do not show perfectly homogeneous
syntactic behavior: the verb interessare “to take an interest” allows for two dif-
ferent subcategorization frames, selecting either a dative or an accusative, and is
used with the auxiliary avere “to have” rather than essere “to be”. The subcatego-
rization frames of all critical verbs (experiencer and active verbs) were checked
in the PONS Dictionary of Italian Verbs (Blumenthal and Rovere 1998) to ensure
that the verbs may take two arguments: a subject and a prepositional (“dative”)
object.9
There were six versions of the questionnaire, each containing 16 critical items
(8 EXP items and 8 ACT items). In order to avoid syntactic priming effects, the
relative order of the arguments was balanced within and across questionnaire ver-
sions, and 20 filler items were included that contained lexical material different
from the critical conditions such as inanimate nouns and transitive accusative
verbs. All lists were pseudo-randomized.

Results
When confronted with an active verb, participants constructed SVO sentences
in 204 cases, OVS order occurred in 3 sentences. The constituent order was more
balanced for the experiencer condition, which showed 78 sentences with SVO
order and 129 sentences with OVS order. These word order proportions differ
significantly for the two verb classes, as shown by Pearson’s chi-squared test with
Yates’s continuity correction for a 2 x 2 (verb class x word order) contingency
table: χ2(1) = 173.78, p < 0.001. Separate chi-squared tests for each verb class
revealed significant effects for both ACT (χ2(1) = 195.17, p < 0.001) and EXP
(χ2(1) = 12.57, p < 0.001).
The individual lexical items within each verb class do not show marked differ-
ences in their order distribution (see Figures 7 and 8); however, it is interesting to
note that all three instances of OVS sentences with active verbs contain the verb
badare “to take care”.
98 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

60

50

40
Occurrences

SVO
30
OVS

20

10

0
scrive bada mente crede
ACT verbs

Figure 7. Word orders chosen for each dative active verb

60

50

40
Occurrences

SVO
30
OVS

20

10

0
piace manca interessa importa
EXP verbs

Figure 8. Word orders chosen for each piacere-class verb

Discussion
The results show that in a production task with given elements, SVO is the pre-
ferred word order with dative active verbs. For dative object experiencer verbs, by
contrast, OVS is preferred, though the asymmetry between word orders is not as
pronounced as for the active verbs. This data pattern supports the theoretical as-
sumptions concerning thematic and information structural markedness that were
outlined in Section 2. In addition, and importantly for the following ERP study,
the homogeneous pattern of results for individual verbs within a class suggests
that the chosen verbs do, indeed, constitute uniform verb groups.
 Thematic reanalysis with Italian dative object experiencer verbs 99

4.3 ERP study

We conducted an experiment using electroencephalography (EEG) to record the

electrophysiological brain responses of participants to Italian sentences with ac-
tive verbs and psych verbs, using event-related potentials (ERPs) as the depen-
dent measure. This method provides a non-invasive means of measuring brain
activity related to language processing. It offers a high temporal resolution (in the
range of milliseconds) and can, moreover, reveal qualitative differences between
different effects based on their polarity, latency, amplitude, and topography. For
introductions to this method, see Kutas, Van Petten and Kluender (2006) and
Bornkessel-Schlesewsky and Schlesewsky (2009b).

Participants
Twenty-nine native speakers of Italian, most of them university students (Ital-
ian exchange students at the University of Marburg), participated in our study.
All participants were right-handed, with normal or corrected-to-normal visual
acuity. They signed an informed consent document before the experiment and
were paid for their participation. Twenty-two participants entered the final data
analysis (all monolingual; 10 female; mean age 25.8 years; age range 20–39 years);
the other seven had to be excluded because of EEG artifacts and/or an insufficient
accuracy in the behavioral tasks.

Materials
We used the same selection of dative active verbs (ACT) and piacere-class verbs
(EXP) as in Questionnaire 2 (see Section 4.2). 50% of the sentences with expe-
riencer verbs and active verbs respectively appeared in SVO order, 50% in OVS
order. 240 proper names (120 male names and 120 female names) were used as
the arguments of the verbs. Each sentence included one male and one female
name, with the assignment of genders to the subject and object roles counterbal-
anced across items. The materials thus consisted of a total of 480 sentences, i.e.,
120 lexical sets of the four conditions in Table 1, which were subdivided into four
pseudo-randomized lists of 120 sentences each (30 per condition) using a Latin
square design. Each list included 380 additional filler items, thus resulting in a
total of 500 sentences per list.

Procedure
Participants were instructed to rate the acceptability of each sentence (i.e., wheth-
er a sentence sounded natural or odd to them) on a 2-point scale. In addition,
they performed a word-recognition task that required them to decide whether a
certain word was contained in the previous sentence.
100 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

Sentences were presented visually on a 17-inch computer screen using Pre-

sentation (Neurobehavioral Systems, Inc.). Prepositional phrases were presented
together, all other elements were shown as single words. The presentation time
was 300 ms followed by a blank screen of 200 ms. After each sentence, three ques-
tion marks (???) cued participants to judge sentence acceptability (maximal reac-
tion time, RT: 2000 ms). The judgment task was followed by the word-recognition
task, signaled by appearance of a single word (maximal RT: 2000 ms). Responses
were given by pressing one of two buttons on a game controller and assignment of
“yes” and “no” responses to the left and right buttons was counterbalanced across
participants.
The experiment began with a training session of twenty sentences. The follow-
ing experimental session was divided into 10 blocks, between which participants
took short breaks. Total time of EEG recording (including electrode preparation)
was about three hours.

EEG recording
The EEG was recorded with a BrainAmp EEG amplifier (Brain Products GmbH)
from 24 sintered Ag/AgCl electrodes referenced to the left mastoid (rereferenced
to linked mastoids offline). AFZ served as ground. The electrooculogram (EOG)
was monitored from electrodes placed at the outer canthi of each eye and above
and below the participant’s left eye. All electrode impedances were kept below
5 kΩ. The EEG was recorded with a digitization rate of 500 Hz and filtered offline
with a 0.3–20.0 Hz band pass filter to exclude slow signal drifts. After computing
statistical analysis, an additional 8.5 Hz low pass filter was applied to smoothen
the plots.

Data analysis
For the behavioral data, acceptability ratings and reaction times were calculated
per condition; all trials with incorrectly answered word-recognition tasks were
excluded. We computed repeated-measure analyses of variance (ANOVAs) with
the within factors VERB (dative active vs. dative experiencer verb) and ORDER
(SVO vs. OVS) by participants (F1) and by items (F2).
For the analysis of the ERPs, we calculated single-participant averages for
each condition in the time window 200 ms before onset of the critical phrase to
1200 ms after onset. Subsequently, grand averages were computed over all partic-
ipants. For statistical analysis, time windows were chosen on the basis of findings
from previous studies and visual inspection of the data. We calculated ANOVAs
for mean amplitude values per time window involving the within-participant fac-
tors VERB (dative active vs. dative experiencer verb), ORDER (SVO vs. OVS) and
topographical region of interest (ROI). Analyses were calculated both for lateral
 Thematic reanalysis with Italian dative object experiencer verbs 101

ROIs (left-anterior: F7, F3, FC5, FC1; right-anterior: F8, F4, FC6, FC2; left-poste-
rior: CP1, CP5, P3, P7; right-posterior: CP2, CP6, P4, P8) and midline ROIs (one
electrode per ROI: FZ; FCZ; CZ; CPZ; PZ; POZ). The correction of Huynh and
Feldt (1970) was applied whenever there was more than one degree of freedom in
the numerator and Mauchly’s sphericity test (Mauchly 1940) had reached signif-
icance. Trials with incorrectly answered word-recognition tasks or EEG artifacts
(such as eye blinks) at the critical positions were excluded from the data before
analysis (the EOG rejection criterion was 40 μV).

Behavioral results
The accuracy in the word-recognition task for all critical stimuli was 96.41% (in-
correct: 3.4%; time-out: 0.19%), indicating that participants processed the stimuli
attentively. There were no significant differences across conditions.
The acceptability judgments (see Figure 9) showed that sentences with OVS
order and active verbs, though grammatical, were strongly disfavored compared
to all other three conditions. Both OVS and SVO orders with experiencer verbs
were rated as acceptable, but lower than SVO with active verbs.
The ANOVA confirmed these descriptive impressions, revealing significant
main effects of ORDER (F1(1,21) = 14.59, p < 0.001; F2(1,119) = 43.35, p < 0.001)
and VERB (F1(1,21) = 4.42, p < 0.05; F2(1,119) = 17.33, p < 0.001) as well as a
significant interaction ORDER x VERB (F1(1,21) = 10.93, p < 0.01; F2(1,119) =
108.19, p < 0.001). Resolving the interaction showed a significant effect ORDER

1.0

0.8
Mean values

0.6
by subjects
by items
0.4

0.2

0.0
OVS OVS SVO SVO
ACT EXP ACT EXP
Conditions

Figure 9. Mean values of acceptability ratings by subjects (n = 22) and by items

(n = 120). 1.0 is acceptable; 0.0 is unacceptable. Error bars show the 95% confidence
interval
102 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

1000

Mean values 800

600
by subjects
by items
400

200

0
OVS OVS SVO SVO
ACT EXP ACT EXP
Conditions

Figure 10. Mean values (in milliseconds) of reaction times by subjects (n = 22)
and by items (n = 120). Error bars show the 95% confidence interval

(F1(1,21) = 13.43, p < 0.01; F2(1,119) = 117.58, p < 0.001) for active verbs, but no
such effect for experiencer verbs (F1/F2 < 1) .
The reaction times (RTs) for the ratings also differ across conditions (see Fig-
ure 10). RTs were fastest in the thematically unmarked conditions SVO+ACT and
OVS+EXP.
The ANOVA for the RTs showed no significant main effects but a signifi-
cant interaction ORDER x VERB (F1(1,21) = 11.17, p < 0.01; F2(1,119) = 22.17,
p < 0.001). Resolving the interaction showed a significant effect ORDER for both
active (F1(1,21) = 7.91, p < 0.02; F2(1,119) = 15.73, p < 0.001) and experiencer
verbs (F1(1,21) = 4.7, p < 0.05; F2(1,119) = 3.96, p < 0.05).
In summary, the results of the judgment task show that OVS orders are re-
duced in acceptability in comparison to SVO orders for active verbs, while both
orders are equally acceptable for dative experiencer verbs. Reaction times revealed
an advantage for thematically unmarked orders (SVO+ACT and OVS+EXP).

ERP results
Verb. Figures 11 and 12 show grand average ERPs at the position of the verb for
SVO and OVS orders, respectively, at selected electrodes. Visual inspection re-
veals a negativity with a peak at approximately 400 ms for EXP vs. ACT in SVO
order, and for ACT vs. EXP in OVS order, which indicates that a verb class effect
interacts with word order. However, the negativity in the SVO contrast is smaller
and restricted to anterior and central electrodes, whereas the effect is more pro-
nounced in the OVS contrast where it appears also in parietal regions. The nega-
tivity appears to be followed by a positivity with a maximum around 700 ms for
 Thematic reanalysis with Italian dative object experiencer verbs 103

FZ

FC1 FC2

CZ

–4 µV

S ACT
0.5 1.0 EXP
4

CP1 CP2

PZ

late pos.

Figure 11. Grand average (n = 22) ERPs at selected electrodes at the position of the verb
(onset at the vertical bar) for active vs. experiencer verbs in SVO sentences. Negativity is
plotted upwards

the same contrasts (i.e., EXP vs. ACT for SVO and ACT vs. EXP for OVS). Two
time windows were chosen for the analysis of these effects: 350–500 ms for the
negativity and 650–800 ms for the positivity.

Verb (time window: 350–500 ms). In the earlier time window, ANOVAs showed
a significant interaction VERB x ROI (Flat(3,63) = 8.55, p < 0.002; Fmid(5,105) =
8.03, p < 0.004). In addition, the interaction ORDER x VERB x ROI reached sig-
nificance in lateral ROIs (Flat(3,63) = 3.46, p < 0.04; Fmid(5,105) = 1.03, p = 0.4),
and the interaction VERB x ORDER in midline ROIs (Flat(1,21) = 3.46, p = 0.08;
Fmid(1,21) = 5.36, p < 0.04). Resolving the latter revealed no significant effects of
verb class in either word order.
Resolving the three-way interaction for the lateral electrodes by ROI showed
a significant interaction of ORDER x VERB in posterior regions (left: F(1,21) =
6.14, p < 0.03; right: F(1,21) = 6.46, p < 0.02). Within these two ROIs, the effect
104 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

FZ

FC1 FC2

CZ

–4 µV

S ACT
0.5 1.0 EXP
4

N400 CP2
CP1

PZ

late pos.

Figure 12. Grand average (n = 22) ERPs at selected electrodes at the position of the verb
(onset at the vertical bar) for active vs. experiencer verbs in OVS sentences. Negativity is
plotted upwards

of VERB was significant only for OVS (left-posterior: F(1,21) = 4.12, p = 0.05;
right-posterior: F(1,21) = 5.42, p < 0.03).
Thus, this early time window at the position of the verb revealed a posterior
negativity (N400) for active versus object experiencer verbs in OVS orders.

Verb (time window: 650–800 ms). The ANOVAs for the later time window
showed significant interactions ORDER x VERB (Flat(1,21) = 35.12, p < 0.001;
Fmid(1,21) = 49.66, p < 0.001), and ORDER x VERB x ROI (Flat(3,63) = 3.56,
p < 0.03; Fmid(5,105) = 5.64, p < 0.02).
Resolutions by ROI revealed significant interactions of ORDER x VERB in all
ROIs (all ps < 0.01), but with a maximum at posterior sites (lateral: FMAX = 52.07
in the right-posterior ROI; FMIN = 9.37 in the left-anterior ROI; midline: FMAX =
79.34 at POZ; FMIN = 12.62 at FZ). Resolving further by ORDER, we find a sig-
nificant effect of VERB for OVS in all ROIs (all ps < 0.05; lateral: FMAX = 12.36 in
 Thematic reanalysis with Italian dative object experiencer verbs 105

the left-posterior ROI; FMIN = 4.63 in the left-anterior ROI; midline: FMAX = 19.81
at POZ; FMIN = 5.32 at CZ), and for SVO in both posterior lateral ROIs and all
midline electrodes except FZ (all ps < 0.05; lateral: FMAX = 14.69 in the left-pos-
terior ROI; FMIN = 12.58 in the right-posterior ROI; midline: FMAX = 20.07 at PZ;
FMIN = 4.48 at FCZ).
In summary, in this later time window at the position of the verb, the data
showed a broadly distributed late positivity for active versus object experiencer
verbs in OVS orders and for object experiencer versus active verbs in SVO orders.

Second argument. In addition to examining ERP responses at the verb, we also

calculated ERPs for the second argument. This comparison was expected to be
particularly interesting for the OVS+ACT condition, in which the marked infor-
mation structure calls for a focused post-verbal subject. As shown in Figures 13
and 14 for SVO and OVS orders respectively, grand average ERPs indeed suggest

FZ

FC1 FC2

CZ

–4 µV

S ACT
0.5 1.0 EXP
4

CP1 CP2

PZ

Figure 13. Grand average (n = 22) ERPs at selected electrodes at the position of the sec-
ond argument, i.e., the object (onset at the vertical bar) for active vs. experiencer verbs in
SVO sentences. Negativity is plotted upwards
106 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

AN FZ

FC1 FC2

CZ

–4 µV

S ACT
0.5 1.0 EXP
4

CP1 CP2

PZ

Figure 14. Grand average (n = 22) ERPs at selected electrodes at the position of the sec-
ond argument, i.e., the subject (onset at the vertical bar) for active vs. experiencer verbs
in OVS sentences. Negativity is plotted upwards

that condition OVS+ACT engendered an anterior negativity (AN) in compar-

ison to all other conditions. This effect was examined in a time window from
300–550 ms.

Second argument (time window: 300–550 ms). The statistical analysis revealed a
significant main effect of ORDER (Flat(1,21) = 7.9, p < 0.02; Fmid(1,21) = 5.92,
p < 0.03) and a significant interaction ORDER x VERB x ROI (Flat(3,63) = 5.91,
p < 0.01; Fmid(5,105) = 8.13, p < 0.01). Resolving the interaction by ROI showed
a significant interaction ORDER x VERB only in anterior regions (left-anterior:
F(1,21) = 7.64, p < 0.02; right-anterior: F(1,21) = 5.88, p < 0.03; FZ: F(1,21) = 8.78,
p < 0.01; FCZ: F(1,21) = 4.08, p = 0.05). All regions showing a significant interac-
tion also showed a significant VERB effect for OVS order, due to a negativity for
ACT vs. EXP (left-anterior: F(1,21) = 7.65, p < 0.02; right-anterior: F(1,21) = 6.69,
p < 0.02; FZ: F(1,21) = 9.16, p < 0.01; FCZ: F(1,21) = 5.52, p < 0.03).
 Thematic reanalysis with Italian dative object experiencer verbs 107

Thus, for the position of the second argument, the statistical analysis con-
firmed the presence of an anterior negativity for OVS orders with active verbs.

Discussion
The acceptability ratings support the theoretical assumptions outlined in Section 2
as well as our hypotheses: the reduced acceptability of OVS+ACT is predicted by
the thematic and information structural markedness of this construction, while
the higher RTs for thematically marked conditions (SVO+EXP, OVS+ACT) can
be interpreted as reflecting the thematic reanalysis that is required in these sen-
tences.
The ERP data also show that verb class and word order interact during online
processing. At the position of the verb, ACT engendered an N400 in comparison
to EXP in object-initial word orders, which can be interpreted as reflecting the
thematic mismatch/reanalysis that is required in the eADM’s “Compute Linking”
when the expectation for a particular verb class that was set up on the basis of the
first argument is not met. Crucially, since this effect interacted with word order
it cannot simply reflect a lexical difference between active and object experiencer
verbs. The following late positivity could be viewed as a reflex of the syntactic
reanalysis that is required as a consequence of the verb class change (see, for ex-
ample, Kutas et al. 2006, for a review of reanalysis-related late positivity findings)
or of the lower degree of wellformedness of sentences involving a mismatch be-
tween word order and the thematic hierarchy (Bornkessel and Schlesewsky 2006).
This account of the biphasic pattern appears relatively straightforward. However,
it does not explain why there was no comparable, statistically reliable negativity
for SVO+EXP vs. SVO+ACT. We will return to this point in the general discus-
sion (in Section 5).
The anterior negativity for OVS+ACT at the second argument position can
plausibly be explained in terms of the information structural requirements of
these sentences in comparison to the other sentence conditions examined here.
Both SVO and OVS are information structurally unmarked with experiencer
verbs, as is SVO with active verbs, while the post-verbal subject of active verbs
has to be focused. Thus, the anterior negativity could reflect the increased costs of
computing this focalization in the absence of a suitable context.

5. General discussion

We presented two questionnaire studies and an ERP experiment which clearly

indicate that verb class interacts with word order in the processing of Italian sen-
tences. Questionnaire 1 revealed a preference to complete sentence fragments
108 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

beginning with a dative argument using dative object experiencer verbs (pref-
erably piacere), while subject-initial sentence fragments were completed using a
variety of active verbs. The observation of different word order preferences for pi-
acere-class verbs and dative active verbs is further supported by Questionnaire 2,
the results of which showed a clear preference to construct SVO sentences with
active verbs and a less pronounced, but significant preference for OVS with pi-
acere-class verbs. Finally, the present ERP study showed clear interactions be-
tween word order and verb class during online processing: at the position of the
verb, OVS sentences showed a biphasic N400–late positivity pattern for dative
active versus dative object experiencer verbs, while SVO sentences showed a late
positivity for dative object experiencer versus dative active verbs. At the position
of the post-verbal argument, OVS sentences with active verbs additionally en-
gendered an anterior negativity in comparison to their counterparts with object
experiencer verbs.
These findings from the processing domain are compatible with the theoret-
ical assumptions about the different verb classes that were outlined in Section 2.
Sentences with object experiencer verbs have a fundamentally different structure
from sentences with active verbs despite appearing very similar on the surface.
Verb-argument linking works differently because the thematically higher role
maps onto the subject with active verbs but onto the object with piacere-class
verbs. In addition, information structure is highly marked in OVS sentences with
active verbs because the object is in a topic position and the post-verbal subject
must be focused, while the same order with object experiencer verbs is not only
thematically but also structurally unmarked.
In the following, we first briefly discuss the consequences of the present re-
sults for the relationship between syntactic structure and the thematic hierarchy
during language processing, before turning to functional interpretations of the
ERP effects observed.

5.1 Syntactic structure and the thematic hierarchy

The results of the permutation task in Questionnaire 2 and the high accept-
ability ratings in the ERP experiment suggest that OVS is the basic word order
for piacere-constructions in Italian. Beginning a sentence with an object rather
than a subject seems to be a marked and unexpected choice. However, there is
considerable psycholinguistic evidence that the parser tries to avoid a violation
of the thematic hierarchy even if this leads to a marked syntax. In a sentence
production task for English, Ferreira (1994) found that participants tended to
produce more passive constructions with object experiencer verbs compared to
 Thematic reanalysis with Italian dative object experiencer verbs 109

active verbs. Thus, a construction that is normally dispreferred, viz. passive, was
chosen in order to respect the thematic hierarchy. In a study by Thompson and
Lee (2009), participants with agrammatic aphasia had less difficulty in producing
and understanding passive sentences with object experiencer verbs compared to
subject experiencer verbs, even though passives were more demanding in general.
For German, several online comprehension studies (e.g., Bornkessel et al. 2004;
Kretzschmar et al. 2012) showed that reanalysis towards a dative-before-nomina-
tive order is less costly for object experiencer verbs. Thus, the thematic hierarchy
seems to be of importance for the processing system across languages.
Nevertheless, the way in which the thematic hierarchy interacts with other in-
formation types appears to depend somewhat upon language-specific properties.
Using ERPs, Bornkessel et al. (2003) contrasted dative object experiencer verbs
(comparable to the Italian piacere-class) with dative active verbs in verb-final sen-
tences in German. They found a thematic reanalysis effect for the experiencer verbs
irrespective of word order (nominative-before-dative or dative-before-nomina-
tive), suggesting that the processing system generally assigned the higher-ranking
thematic role to the nominative argument.10 Thus, a dative-initial linearization in
the pre-verbal region in German (the so-called “Middlefield”) does not give rise
to a clear prediction in favor of an object experiencer verb, which is in contrast to
our data from Italian. Italian word order therefore seems to be more constraining
than German word order in enabling the parser to build up thematic expecta-
tions, even though word order is considerably less strict than in a language such
as English.

5.2 Thematic expectations and the N400

The N400 component is known to be modulated by unexpected items even if they

are semantically plausible (e.g., Kamide 2008; Kutas et al. 2011), and it has been
reported when a thematic mismatch induces a linking violation (e.g., Bornkessel
et al. 2004). Thus, we assume that the N400 effects observed at the position of the
verb in our experiment resulted from the interplay of two different factors: the
thematic hierarchy and lexical predictability.
An N400 effect was elicited at the position of the verb when an initial dative
was followed by an active verb (e.g., A Piero scrive Matilde “Matilde writes to
Piero”) because of a mismatch in the thematic hierarchy and the need for ar-
gument reanalysis. This assumption follows from a processing strategy in which
the first argument is generally expected to be higher in the thematic hierarchy,
i.e., more agent-like if not explicitly marked otherwise. In terms of the eADM,
the N400 effect can be derived as a reflex of the “Compute Linking” step, which
110 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

results when the expectations about the verb’s logical structure that were set up by
the pre-verbal argument are not met.
However, the statistical analysis revealed no comparable N400 for the the-
matic mismatch in the SVO+EXP condition (e.g., Piero piace a Matilde “Piero ap-
peals to Matilde”) compared to the thematically unmarked SVO+ACT condition
(e.g., Piero scrive a Matilde “Piero writes to Matilde”). We assume that the absence
of a clear N400 effect in this position can be attributed to the lack of predictability
following an initial nominative. In contrast to the high predictability for a dative
object experiencer verb following an initial dative (recall the results of Question-
naire 1), an initial subject does not lead the processing system to anticipate a da-
tive active verb. Hence, the lower degree of lexical predictability leads to an N400
increase in SVO+ACT, thus masking the thematic N400 effect in SVO+EXP.
In summary, the overall pattern of N400 effects at the verb provides compel-
ling evidence that the human language comprehension system sets up expecta-
tions about the class of an upcoming verb during online processing.

5.3 Reanalysis / wellformedness and the late positivity

In contrast to the N400 effect, the late positivity effects at the position of the verb
are symmetrical for both word orders, i.e., experiencer verbs showed a positivity
effect in comparison to active verbs for SVO and vice versa for OVS. Thus, the late
positivity does not appear to reflect processing expectations in the same way as
the N400 in the current study; rather, it is elicited whenever there is a mismatch
between the thematic hierarchy and the preferred syntactic structure. This result is
compatible with several functional interpretations of late positivity effects in lan-
guage comprehension: the modulation of the positivity could reflect (a) reanalysis
towards a dispreferred syntactic structure (e.g., Osterhout and Holcomb 1992);
(b) higher syntactic integration costs (e.g., Kaan, Harris, Gibson and Holcomb
2000), which may be expected as a result of a mismatch with the thematic hierar-
chy; or (c) a wellformedness mismatch (e.g., Bornkessel and Schlesewsky 2006)
or even a reflex of general conflict monitoring (e.g., Van de Meerendonk, Kolk,
Chwilla and Vissers 2009), again due to the syntax-thematic hierarchy mismatch.
All of these interpretations are compatible with the theoretical assumptions about
object experiencer verbs that were discussed in Section 2. However, as we cannot
differentiate between the different possibilities on the basis of the present findings,
we will not discuss them in any further detail.
 Thematic reanalysis with Italian dative object experiencer verbs 111

5.4 Information structure and the anterior negativity

Finally, consider the anterior negativity observed for the second argument in OVS
structures with active verbs (e.g., A Piero scrive Matilde “Matilde writes to Piero”).
Recall that these constructions have a marked information structure, requiring a
focused post-verbal subject. The finding of an anterior negativity at the position
of a focused constituent is in line with previous results from a study on prosody
and information structure in German (Hruska and Alter 2004). Hruska and Alter
observed anterior negativities at all focused positions in the answer sentences of
simple question-answer pairs and interpreted this effect as reflecting the process-
ing of contextually-expected focus information.

6. Conclusion

Our study on dative object experiencer verbs and dative active verbs in Italian
clearly shows that verb classes – and their thematic, syntactic and information
structural ramifications – affect real time sentence processing, including the in-
cremental build up of expectations. In accordance with previous results, the pres-
ent findings attest to the importance of respecting the thematic hierarchy during
online processing (i.e., the preference to interpret the first argument as bearing
the highest-ranking thematic role). They also suggest, however, that this gener-
al preference interacts with language-specific characteristics, thus leading to a
stronger effect of word order for Italian as opposed to German. Finally, the fact
that the interaction between word order and verb class manifested itself in several
qualitatively dissociable ERP effects indicates that verb class information exerts
an influence on multiple levels of the language processing architecture, affecting
the interface between syntax, semantics and discourse (information structure).

Acknowledgments

Parts of the research reported here were supported by the German Research
Foundation (grant BO 2471/3-2). We are very grateful to Walter Bisang,
Richard Wiese, Alec Marantz, and the organizers and the audience of the work-
shop “Structuring the Argument” at Paris for helpful comments and valuable dis-
cussion. We also thank Phillip Alday, Sabine Frenzel, Viviana Haase, Yu-Chen
Hung, Johannes Knaus, R. Muralikrishnan, Valentina Reggio, Anna-Lena Rumpf,
Jona Sassenhagen, Lea Schäfer, and Sarah Tune for their comments and sugges-
tions, support during the preparation, and help in the lab.
112 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

Notes

1. Kasper (2008) suggests an even finer-grained classification of psych verbs, discriminating

five classes.

2. It should be noted that there is no inflectional case marking in Italian nouns; the terms
accusative object and dative object are used here to refer to direct and indirect objects, respec-
tively.

3. Most of the relevant theoretical literature on psych verbs in Italian and their syntactic prop-
erties is couched in a generative framework, as is our syntactic description in Section 2. Note,
however, that the implications of the theoretical description for our experimental hypotheses
are not tied to a generative perspective. We aim to present psycholinguistic evidence for seman-
tic and syntactic properties of Italian piacere-class verbs rather than for the technical details of
a particular theory.

4. By contrast, Bennis (2004: 88) considers accusative object experiencer verb constructions,
i.e., preoccupare-class verbs, an instance of what he terms “complex ergative” constructions.
These do have a v-projection, but, due to an intrinsic property, v fails to assign a θ-role. In
contrast to Belletti and Rizzi’s (1988) proposal that accusative object experiencer verbs assign a
lexical accusative Case, Bennis assumes that accusative Case is assigned in a structural position,
i.e., outer Spec of v. However, no external argument is projected, and both arguments are gen-
erated within the lower VP. Like Belletti and Rizzi, but in contrast to Baker’s UTAH (cf. Baker
1988; 1997), Bennis assumes that the experiencer projects to a configurational position higher
in the structure, i.e., [Spec,VP] in his analysis.

5. The prepositional dative in Italian may be analyzed as a Kase Phrase, e.g., [KP a [DP ø [NP
Laura]]] (see, e.g., Gabriel and Müller 2008; for a KP-analysis of German datives, see Bayer,
Bader and Meng 2001).

6. Cardinaletti (1997) agrees with the analysis that pre-verbal datives of experiencer verbs
are not moved to a topic position but occupy a subject position in the unmarked case; she
mentions, however, that topicalization of dative experiencers is possible in general, and dative
experiencers may move to a topicalized position like indirect objects of other verbs.

7. Rizzi (2005; 2006) dissociates the subject position that carries the Case and φ-features from
another subject position which hosts a prepositional phrase like the experiencer of piacere-class
verbs. He refers to a proposal by Cardinaletti (1997; 2001; 2004) who points out that Belletti
and Rizzi’s (1988) original analysis cannot be entirely accurate because piacere-constructions
with a fronted dative and a dropped subject project an expletive pro that must precede the verb.
Therefore, Cardinaletti (2004) distinguishes the functional projections SubjP, eppP, and AgrSP.
The category Subj is equipped with a subject-of-predication feature, thus it attracts for example
indirect objects that have certain subject properties like those in the piacere-constructions. The
category AgrS is responsible for checking Case and φ-features, and the category epp for check-
ing the EPP-feature.

8. Note that the given phrase markers represent an oversimplification of the syntactic struc-
ture because we leave aside the issue of remnant VP movement here as it does not seem of
immediate relevance for the current discussion (but see, e.g., Belletti 2001).
 Thematic reanalysis with Italian dative object experiencer verbs 113

9. As the only exception, the stimulus verb scrivere “to write” is classified as a three-place
verb with an optional direct object in Blumenthal and Rovere (1998); however, sentences with
scrivere containing an overtly realized indirect object and a non-realized direct object are con-
sidered complete and fully acceptable by native speaker informants.

10. Note that Bornkessel et al. (2003) also observed a pre-verbal thematic reanalysis effect at
the position of a nominative argument following an initial dative. They thus argue that, when
the initial dative is first encountered, it is analyzed as the highest-ranking argument, but that
this assumption is revised when an animate nominative is encountered next.

References

Altmann, Gerry T. M. and Steedman, Mark. 1988. “Interaction with context during human
sentence processing.” Cognition 30: 191–238. DOI: 10.1016/0010-0277(88)90020-0
Baker, Mark C. 1988. Incorporation: A Theory of Grammatical Function Changing. Chicago: The
University of Chicago Press.
Baker, Mark C. 1997. “Thematic Roles and Syntactic Structure.” In Elements of Grammar,
Liliane Haegeman (ed.), 73–137. Dordrecht: Kluwer. DOI: 10.1007/978-94-011-5420-8_2
Bates, Elizabeth, McNew, Sandra, MacWhinney, Brain, Devescovi, Antonella and Smith, Stan.
1982. “Functional constraints on sentence processing: A cross-linguistic study.” Cognition
11: 245–299. DOI: 10.1016/0010-0277(82)90017-8
Bayer, Joseph, Bader, Markus and Meng, Michael. 2001. “Morphological underspecification
meets oblique case: Syntactic and processing effects in German.” Lingua 111: 465–514.
DOI: 10.1016/S0024-3841(00)00041-3
Belletti, Adriana. 2001. “Inversion as Focalization.” In Subject Inversion in Romance and the
Theory of Universal Grammar, Aafke Hulk and Jean-Yves Pollock (eds), 60–90. Oxford:
Oxford University Press.
Belletti, Adriana. 2004. “Aspects of the Low IP Area.” In The Structure of CP and IP [The Car-
tography of Syntactic Structures 2], Luigi Rizzi (ed.), 16–51. New York: Oxford University
Press.
Belletti, Adriana and Rizzi, Luigi. 1988. “Psych-verbs and θ-theory.” Natural Language and Lin-
guistic Theory 6: 291–352. DOI: 10.1007/BF00133902
Bennis, Hans. 2004. “Unergative Adjectives and Psych Verbs.” In The Unaccusativity Puzzle:
Explorations of the Syntax-Lexicon Interface [Oxford Studies in Theoretical Linguistics 5],
Artemis Alexiadou, Elena Anagnostopoulou and Martin Everaert (eds), 84–114. Oxford:
Oxford University Press. DOI: 10.1093/acprof:oso/9780199257652.003.0004
Blumenthal, Peter and Rovere, Giovanni. 1998. PONS: Wörterbuch der italienischen Verben:
Konstruktionen, Bedeutungen, Übersetzungen. Stuttgart: Klett.
Bornkessel, Ina, McElree, Brian, Schlesewsky, Matthias and Friederici, Angela D. 2004.
“Multi-dimensional contributions to garden path strength: Dissociating phrase structure
from case marking.” Journal of Memory and Language 51: 495–522.
DOI: 10.1016/j.jml.2004.06.011
Bornkessel, Ina and Schlesewsky, Matthias. 2006. “The extended Argument Dependency Mod-
el: A neurocognitive approach to sentence comprehension across languages.” Psychological
Review 113: 787–821. DOI: 10.1037/0033-295X.113.4.787
114 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

Bornkessel, Ina, Schlesewsky, Matthias and Friederici, Angela D. 2003. “Eliciting thematic re-
analysis effects: The role of syntax-independent information during parsing.” Language
and Cognitive Processes 18: 268–298. DOI: 10.1080/01690960244000018
Bornkessel, Ina, Zysset, Stefan, Friederici, Angela D., von Cramon, D. Yves and Schlesewsky,
Matthias. 2005. “Who did what to whom? The neural basis of argument hierarchies during
language comprehension.” NeuroImage 26: 221–233.
DOI: 10.1016/j.neuroimage.2005.01.032
Bornkessel-Schlesewsky, Ina and Schlesewsky, Matthias. 2008a. “An alternative perspective on
‘semantic P600’ effects in language comprehension.” Brain Research Reviews 59: 55–73.
DOI: 10.1016/j.brainresrev.2008.05.003
Bornkessel-Schlesewsky, Ina and Schlesewsky, Matthias. 2008b. “Unmarked transitivity: a pro-
cessing constraint on linking.” In Investigations of the syntax-semantics-pragmatics inter-
face, Robert Van Valin, Jr. (ed.), 413–434. Amsterdam: John Benjamins.
Bornkessel-Schlesewsky, Ina and Schlesewsky, Matthias. 2009a. “The Role of Prominence In-
formation in the Real-Time Comprehension of Transitive Constructions: A Cross-Lin-
guistic Approach.” Language and Linguistics Compass 3: 19–58.
DOI: 10.1111/j.1749-818X.2008.00099.x
Bornkessel-Schlesewsky, Ina and Schlesewsky, Matthias. 2009b. Processing Syntax and Mor-
phology. A Neurocognitive Perspective. Oxford: Oxford University Press.
Burzio, Luigi. 1986. Italian Syntax. Dordrecht: Reidel. DOI: 10.1007/978-94-009-4522-7
Cardinaletti, Anna. 1997. “Subjects and Clause Structure.” In The New Comparative Syntax,
Liliane Haegeman (ed.), 33–63. London: Longman.
Cardinaletti, Anna. 2001. “A second thought on emarginazione: Destressing vs. ‘Right Disloca-
tion’.” In Current Studies in Italian Syntax: Essays Offered to Lorenzo Renzi [North-Holland
Linguistic Series: Linguistic Variations 59], Guglielmo Cinque and Giampaolo Salvi (eds),
117–136. Amsterdam: Elsevier Science.
Cardinaletti, Anna. 2004. “Toward a cartography of subject positions.” In The Structure of CP
and IP [The Cartography of Syntactic Structures 2], Luigi Rizzi (ed.), 115–165. New York:
Oxford University Press.
Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris Publications.
Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press.
Chomsky, Noam. 2000. “Minimalist Inquiries: The Framework.” In Step by Step: Essays on
Minimalist Syntax in Honor of Howard Lasnik, Roger Martin, David Michaels and Juan
Uriagereka (eds), 89–155. Cambridge, MA: MIT Press.
Chomsky, Noam. 2001. “Derivation by Phase.” In Ken Hale: A Life in Language, Michael
Kenstowicz (ed.), 1–52. Cambridge, MA: MIT Press.
Corrigan, Roberta. 1988. “Who Dun It? The Influence of Actor-Patient Animacy and Type of
Verb in the Making of Causal Attributions.” Journal of Memory and Language 27: 447–465.
DOI: 10.1016/0749-596X(88)90067-8
Crain, Stephen and Steedman, Mark. 1985. “On not being led up the garden path: The use of
context by the psychological parser.” In Natural language parsing: Psychological, compu-
tational, and theoretical perspectives, David Dowty, Lauri Karttunen and Arnold Zwicky
(eds), 320–357. Cambridge: Cambridge University Press.
Crocker, Matthew W. 1994. “On the nature of the principle-based sentence processor.” In Per-
spectives on sentence processing, Charles Clifton, Lyn Frazier and Keith Rayner (eds), 245–
266. Hillsdale, NJ: Lawrence Erlbaum Associates Inc.
 Thematic reanalysis with Italian dative object experiencer verbs 115

DeLong, Katherine A., Urbach, Thomas P. and Kutas, Marta. 2005. “Probabilistic word pre-ac-
tivation during language comprehension inferred from electrical brain activity.” Nature
Neuroscience 8: 1117–1121. DOI: 10.1038/nn1504
Federmeier, Kara D. and Kutas, Marta. 1999. “A Rose by Any Other Name: Long-Term Memory
Structure and Sentence Processing.” Journal of Memory and Language 41: 469–495.
DOI: 10.1006/jmla.1999.2660
Ferreira, Fernanda. 1994. “Choice of Passive Voice Is Affected by Verb Type and Animacy.”
Journal of Memory and Language 33: 715–736. DOI: 10.1006/jmla.1994.1034
Fodor, Janet D. 1998. “Learning to parse.” Journal of Psycholinguistic Research 27: 285–319.
DOI: 10.1023/A:1023258301588
Frazier, Lyn and Fodor, Janet D. 1978. “The sausage machine: A new two-stage parsing model.”
Cognition 6: 291–326. DOI: 10.1016/0010-0277(78)90002-1
Gabriel, Christoph and Müller, Natascha. 2008. Grundlagen der generativen Syntax: Franzö-
sisch, Italienisch, Spanisch. Tübingen: Niemeyer.
Gattei, Carolina, Vasishth, Shravan and Dickey, Michael W. 2011. The Role of Semantic Argu-
ments Order and the Syntax-Semantics Interface in Spanish Sentence Comprehension. Post-
er presented at Structuring the Argument/Structurer l’argument (CNRS/Paris 8), Paris,
France, September 5–7.
Gennari, Silvia P. and MacDonald, Maryellen C. 2008. “Semantic indeterminacy in object rela-
tive clauses.” Journal of Memory and Language 58: 161–187. DOI: 10.1016/j.jml.2007.07.004
Grimshaw, Jane. 1990. Argument Structure. Cambridge, MA: MIT Press.
Hruska, Claudia and Alter, Kai. 2004. “Prosody in dialogues and single sentences: How proso-
dy can influence sentence perception.” In Information structure: Theoretical and empirical
evidence, Anita Steube (ed.), 211–226. Berlin: Walter de Gruyter.
Huynh, Huynh and Feldt, Leonard S. 1970. “Conditions under which the mean square ratios in
repeated measurement designs have exact F-distributions.” Journal of the American Statis-
tical Association 65: 1582–1589. DOI: 10.1080/01621459.1970.10481187
Kaan, Edith, Harris, Anthony, Gibson, Edward and Holcomb, Phillip. 2000. “The P600 as an
index of syntactic integration difficulty.” Language and Cognitive Processes 15: 159–201.
DOI: 10.1080/016909600386084
Kailuweit, Rolf. 2005. Linking: Syntax und Semantik französischer und italienischer Gefühlsver-
ben. Tübingen: Niemeyer. DOI: 10.1515/9783110943351
Kamide, Yuki. 2008. “Anticipatory Processes in Sentence Processing.” Language and Linguistics
Compass 2: 647–670. DOI: 10.1111/j.1749-818X.2008.00072.x
Kasper, Simon. 2008. A comparison of ‘thematic role’ theories. Unpublished Master’s Thesis.
Philipps-Universität, Marburg.
Kretzschmar, Franziska, Bornkessel-Schlesewsky, Ina, Staub, Adrian, Roehm, Dietmar and
Schlesewsky, Matthias. 2012. “Prominence facilitates ambiguity resolution: On the interac-
tion between referentiality, thematic roles and word order in syntactic reanalysis.” In Case,
word order and prominence. Interacting cues in language production and comprehension
[Studies in Theoretical Psycholinguistics 40], Monique Lamers and Peter de Swart (eds),
239–271. Berlin: Springer. DOI: 10.1007/978-94-007-1463-2_11
Kutas, Marta, DeLong, Katherine A. and Smith, Nathaniel J. 2011. “A Look around at What
Lies Ahead: Prediction and Predictability in Language Processing.” In Predictions in the
Brain: Using Our Past to Generate a Future, Moshe Bar (ed.), 190–207. New York: Oxford
University Press.
116 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

Kutas, Marta and Hillyard, Steven A. 1984. “Brain potentials during reading reflect word expec-
tancy and semantic association.” Nature 307: 161–163. DOI: 10.1038/307161a0
Kutas, Marta, Van Petten, Cyma and Kluender, Robert. 2006. “Psycholinguistics electrified
II (1994–2005).” In Handbook of Psycholinguistics, 2nd edition, Matthew J. Traxler and
Morton A. Gernsbacher (eds), 659–724. London: Elsevier.
DOI: 10.1016/B978-012369374-7/50018-3
Landau, Idan. 2010. The Locative Syntax of Experiencers. Cambridge, MA: MIT Press.
MacWhinney, Brian, Bates, Elizabeth and Kliegl, Reinhold. 1984. “Cue Validity and Sentence
Interpretation in English, German, and Italian.” Journal of Verbal Learning and Verbal Be-
havior 23: 127–150. DOI: 10.1016/S0022-5371(84)90093-8
Mauchly, John W. 1940. “Significance test for sphericity of a normal n-variate distribution.”
Annual of Mathematical Statistics 11: 204–209. DOI: 10.1214/aoms/1177731915
Osterhout, L. and Holcomb, P. 1992. “Event-related brain potentials elicited by syntactic anom-
aly.” Journal of Memory and Language 31: 785–806. DOI: 10.1016/0749-596X(92)90039-Z
Pesetsky, David. 1995. Zero Syntax: Experiencers and Cascades. Cambridge, MA: MIT Press.
Primus, Beatrice. 1999. Cases and Thematic Roles – Ergative, Accusative and Active. Tübingen:
Niemeyer. DOI: 10.1515/9783110912463
Primus, Beatrice. 2004. “Protorollen und Verbtyp: Kasusvariation bei psychischen Verben.” In
Semantische Rollen, Rolf Kailuweit and Martin Hummel (eds), 377–401. Tübingen: Narr.
Rizzi, Luigi. 1982. Issues in Italian Syntax. Dordrecht: Foris. DOI: 10.1515/9783110883718
Rizzi, Luigi. 1986. “Null Objects in Italian and the Theory of pro.” Linguistic Inquiry 17: 501–
557.
Rizzi, Luigi. 1997. “The Fine Structure of the Left Periphery.” In Elements of Grammar, Liliane
Haegeman (ed.), 281–337. Dordrecht: Kluwer. DOI: 10.1007/978-94-011-5420-8_7
Rizzi, Luigi. 2004. “Locality and Left Periphery.” In Structures and Beyond [The Cartography
of Syntactic Structures 3], Adriana Belletti (ed.), 223–251. New York: Oxford University
Press.
Rizzi, Luigi. 2005. “On Some Properties of Subjects and Topics.” In Contributions to the 30th
Incontro di Grammatica Generativa, Laura Brugè, Giuliana Giusti, Nicola Munaro, Walter
Schweikert and Giuseppina Turano (eds), 203–224. Venezia: Università Ca’ Foscari Vene-
zia.
Rizzi, Luigi. 2006. “On the Form of Chains: Criterial Positions and ECP Effects.” In Wh-Move-
ment: Moving On, Lisa Cheng and Norbert Corver (eds), 97–134. Cambridge, MA: MIT
Press.
Schlesewsky, Matthias and Bornkessel, Ina. 2006. “Context-sensitive neural responses to con-
flict resolution: Electrophysiological evidence from subject-object ambiguities in language
comprehension.” Brain Research 1098: 139–152. DOI: 10.1016/j.brainres.2006.04.080
Thompson, Cynthia K. and Lee, Miseon. 2009. “Psych verb production and comprehension in
agrammatic Broca’s aphasia.” Journal of Neurolinguistics 22: 354–369.
DOI: 10.1016/j.jneuroling.2008.11.003
Van Berkum, Jos J. A., Brown, Colin M., Zwitserlood, Pienie, Kooijman, Valesca and Hagoort,
Peter. 2005. “Anticipating Upcoming Words in Discourse: Evidence From ERPs and Read-
ing Times.” Journal of Experimental Psychology: Learning, Memory, and Cognition 31: 443–
467. DOI: 10.1037/0278-7393.31.3.443
 Thematic reanalysis with Italian dative object experiencer verbs 117

Van de Meerendonk, Nan, Kolk, Herman H. J., Chwilla, Dorothee J. and Vissers, Constance Th.
W. M. 2009. “Monitoring in language perception.” Language and Linguistics Compass 3:
1211–1224. DOI: 10.1111/j.1749-818X.2009.00163.x
Wicha, Nicole Y. Y., Moreno, Eva M. and Kutas, Marta. 2004. “Anticipating Words and Their
Gender: An Event-related Brain Potential Study of Semantic Integration, Gender Expec-
tancy, and Gender Agreement in Spanish Sentence Reading.” Journal of Cognitive Neuro-
science 16: 1272–1288. DOI: 10.1162/0898929041920487
Wolff, Susann, Schlesewsky, Matthias, Hirotani, Masako and Bornkessel-Schlesewsky, Ina. 2008.
“The neural mechanisms of word order processing revisited: Electrophysiological evidence
from Japanese.” Brain and Language 107: 133–157. DOI: 10.1016/j.bandl.2008.06.003

Appendix

Table A1. Stimuli (critical and filler items) of version 1 of Questionnaire Study 1.
Participants saw one item at a time. English translations have been added here
Condition Item in questionnaire
Filler In primavera “in spring” [blank to complete sentence]
Nominative Claudio [blank to complete sentence]
Filler Il tecnico “the technician” [blank to complete sentence]
Filler Ieri “yesterday” [blank to complete sentence]
Dative A Sonia [blank to complete sentence]
Filler I bicchieri “the glasses” [blank to complete sentence]

Table A2. Lexical items chosen after an initial dative phrase in Questionnaire study 1
Verb Occurrences Verb class
piacere “to appeal” 49
mancare “to be missing” 2 EXP verb
riuscire “to succeed” 1
dire “to tell” 1
regalare “to give as a present” 1 ACT verb
rimproverarsi “to blame onself ” 1
118 Alexander Dröge, Laura Maffongelli and Ina Bornkessel-Schlesewsky

Table A3. Lexical items chosen after an initial nominative phrase in Questionnaire
study 1
Verb Occurrences Verb class
essere “to be” 13
mangiare “to eat” 6
andare “to go” 5
avere “to have” 3
fare “to do” 3
lavorare “to work” 2
amare “to love” 2
sapere “to know” 2
uscire “to go out” 2
compiere gli anni “to have a birthday” 1
comprare “to buy” 1
correre “to run” 1
cucinare “to cook” 1
ACT verb
diventare “to become” 1
finire “to finish” 1
giocare “to play” 1
lavare “to wash” 1
morire “to die” 1
parlare “to talk” 1
piangere “to cry” 1
prestare “to lend” 1
ricordarsi “to remember” 1
rompersi “to crack” 1
studiare “to study” 1
telefonare “to call” 1
venire “to come” 1
 Causative nominalizations
Implications for the structure of psych verbs*

Artemis Alexiadou and Gianina Iordăchioaia

We investigate so-called causative psych nominalizations (CPNs), i.e., nomi-

nalizations of object experiencer (OE) verbs that realize non-agentive causers
as external arguments. While they are ruled out in English (Grimshaw 1990;
Iwata 1995; Pesetsky 1995) and have been suggested to be cross-linguistically
banned (Landau 2010), we show that Romanian and Greek derive CPNs from
the subject experiencer (SE) form of alternating OE verbs. We analyze them as
nominalizations of the anticausative SE form of these verbs. Our results suggest
a structural difference between Romanian/Greek and English psych nominal-
izations: the former can nominalize either the SE anticausative or (the passive
of) the OE transitive structure (Anagnostopoulou 1999), while the latter only
nominalize the OE transitive form, as the verbs lack the anticausative structure.

1. Introduction

The structure of OE verbs has been subject to many cross-linguistic debates

(Belletti & Rizzi 1988; Grimshaw 1990; Pesetsky 1995; Landau 2010). A property
that is often associated with their special behavior is the fact that nominalizations
derived from them allow agentive causers, but exclude non-agentive ones as exter-
nal arguments (see Grimshaw 1990: 118–123; Iwata 1995; Pesetsky 1995: 71–73).
Despite the compatibility of the verb with both an agentive and a non-agentive
causer, the nominalization in (1b/c) is only compatible with an agent. We assume
here a dichotomy between agents (i.e., agentive causers) and non-agentive caus-
ers: the action of the former can be characterized as (non-)deliberate, while that
of the latter cannot. While (human) animates can function both as agents and
non-agentive causers, inanimates cannot acts as agents. In what follows, we use
examples with a human DP to illustrate an agent thematic role and we use an
inanimate DP for the non-agentive causer.
120 Artemis Alexiadou and Gianina Iordăchioaia

(1) a. John/the event humiliated Mary.

b. John’s/*the event’s humiliation of Mary
c. Mary’s humiliation by John/*by the event

As a consequence of this, nominalizations of non-agentive psych verbs like amaze

cannot realize the external argument at all, as shown in (2).
(2) a. John/the event (*deliberately) amazed Mary.
b. *John’s/*the event’s amazement of Mary
c. *Mary’s amazement by John/by the event

The lack of causative force in psych nominalizations was already observed by

Lakoff (1970), who notices that despite the OE-SE variation of psych verbs and
adjectives in (3), the nominalization is always SE (in (4)).
(3) a. What he did amused/surprised me. (OE-verb)
b. I was amused/surprised at what he did. (SE-adjective)
c. Movies are enjoyable to me. (OE-adjective)
d. I enjoy movies. (SE-verb)
(4) a. my amusement/surprise at what he did
b. my enjoyment of movies

Following Lakoff ’s intuition, Pesetsky (1995) argues that OE verbs are derived
from a SE root that is causativized by a zero-morpheme CAUS. Thus the nomi-
nalization pattern that is ruled out is the one in (5b), where a structure including
CAUS is nominalized. The simple SE nominalization in (5a), corresponding to
the examples in (4), is available.
(5) a. [[√SE-predicate V] nominalizer]
b. *[[√SE-predicate V] CAUSV] nominalizer]

Pesetsky excludes (5b) by making appeal to Myers’s (1984) generalization in (6).

The presence of the null morpheme CAUS prevents the attachment of the nomi-
nalizer in (5b).
(6) Zero-derived words do not permit affixation of further derivational
morphemes.

Myers’s generalization runs into problems with affixes like -able and -er in En-
glish, which provide many counter-examples (see, e.g., accentable and accenter,
derived from the denominal verb to accent, Pesetsky 1995: 76–77). As Landau
(2010: 143–149) points out, the only way to solve this conflict is to argue that the
generalization constrains the type of affixes that can attach to CAUS. However,
this idea misses the cross-linguistic picture: languages that have an overt CAUS
 Causative nominalizations and psych verbs 121

like Hebrew still disallow CPNs (see, e.g., (7)). From this, Landau concludes that
the ban on CPNs is universal and indifferent to morphology.
(7) *zi´azua Rina/ha-xadašot et Gil
shock.CAUS.NMZ Rina/the-news ACC Gil
‘Rina’s/the news’ shock of Gil’

In this paper we show that Romanian and Greek do have a kind of CPNs which
are derived from the SE form of OE verbs that alternate. Since the SE verb form
receives an anticausative structure (see Alexiadou, Anagnostopoulou & Schäfer
2006), the nominalizations we discuss are structurally different from the ones in
(1b) and (2b), which nominalize the transitive structure of the OE verb form.
However, our data show that there is no ban on the realization of non-agentive
causers in psych nominalizations. The correct generalization is that non-agen-
tive causers are banned in transitive psych nominalizations, a property which,
as Landau suggests, is not typical of psych roots only, but seems to also apply to
other transitive nominalizations, as well as to verbal Voice alternations in some
languages (see Sichel 2010, 2011; Alexiadou et al. 2013a, for discussion).
We start by presenting the ambiguous behavior of OE verbs, and the corre-
sponding nominalizations that realize non-agentive causers in Greek and Roma-
nian (Section 2). We show that these nominalizations are only possible with OE
verbs that have a SE alternate. We continue by arguing that the OE-SE alternation
is an instance of the (anti)causative alternation, and that non-agentive causer PPs
in psych nominalizations are inherited from the verb as arguments (Section 3).
In Section 4 we offer an account of CPNs in Greek and Romanian, in which they
are taken to be nominalizations of the anticausative structure that is associated
with the SE verb form. We formulate a few further remarks on why English does
not have CPNs and on the restriction on non-agentive causers with transitive and
passive psych nominalizations in Section 5, and conclude in Section 6.

2. OE verbs and causative nominalizations

Both in Greek and Romanian the OE form of a psych verb may have an agentive
((8a), (9a)), or a causative use ((8b), (9b)) (see also Belletti & Rizzi 1988; Pesetsky
1995; Arad 1998; Anagnostopoulou 1999; Pylkkänen 2000; Martin 2006; Landau
2010). The agentive use is always eventive and involves a change of state in the
experiencer. The causative use is ambiguous between an eventive/change of state
reading and a stative one (Arad 1998, 2002). Various factors like the past or the
present tense form of the verb makes the former, respectively, the latter reading
more salient in (8b) and (9b). The SE form of the verb is marked with non-­active
122 Artemis Alexiadou and Gianina Iordăchioaia

(NACT) morphology in Greek (8c) and with a reflexive (RF) in Romanian (9c).
It is usually eventive (only few SE forms are stative; see Section 3.1) and has been
claimed to be inchoative (Belletti & Rizzi 1988, for Italian; Marín & McNally
2011, for Spanish).1
(8) a. O Janis enohlise ti Maria epitides.
the John annoyed.3SG the Maria intentionally
‘John annoyed Maria intentionally.’
b. To kurema tis Marias (ton) enohlise/enohili to Jani.
the haircut the Maria.GEN him annoyed/annoys the John.ACC
‘Maria’s haircut annoyed/annoys John.’
c. O Janis enohlithike me ta nea.
the John annoyed.NACT.3SG with the news
‘John got annoyed with the news.’
(9) a. Ion a enervat-o pe Maria dinadins.
John has annoyed-her ACC Maria intentionally
‘John annoyed Mary intentionally.’
b. Freza lui Ion o enervează/a enervat-o pe Maria.
haircu John.GEN her annoys/has annoyed-her ACC Maria
‘John’s haircut annoys/annoyed Maria.’
c. Maria s-a enervat (pe Ion) de la joc.
Maria RF-has annoyed (at John) from game
‘Maria got annoyed at John from the game.’

We limit our attention to the eventive readings of psych verbs (and their nominal-
izations) with non-agentive causers in (8/9b, c). We will see that in these contexts
the nominalizations keep the argumental PPs that the verbs use. This is usually
not the case with stative readings and their nominalizations (see Section 3.1 be-
low and Iordăchioaia, Alexiadou & Soare to appear).

2.1 Greek and Romanian CPNs

Most OE verbs that take non-agentive causers in Greek and Romanian also build
nominalizations that can realize non-agentive causers (see (10) and (11)). Similar
examples can be constructed with other verbs, among which a alerta ‘to alert’, a
descuraja ‘to discourage’, a intimida ‘to intimidate’, a indigna ‘to offend’, a întrista
‘to sadden’, a îngrijora ‘to worry’, a înviora ‘to cheer up’, a supăra ‘to upset’ in Ro-
manian, and prosvalo ‘offend’, apelpizo ‘despair’, erehthizo ‘irritate’ in Greek.
(10) enervarea Mariei de către Ion/de la joc
annoy.INF.the Maria.GEN by John/from game
‘John’s annoying Maria/Maria’s getting annoyed with the game’
 Causative nominalizations and psych verbs 123

(11) i enohlisi tis Marias apo to Jani/me ta nea

the bothering the Maria.GEN by the John/with the news
‘John’s annoying Maria/Maria’s getting annoyed with the game’

Unlike in English, nominalizations in Greek and Romanian can realize external ar-
guments only as PPs. That is, both languages lack ‘transitive’ nominalizations of the
English type (see Alexiadou 2001; Alexiadou et al. 2009; Iordăchioaia 2008 for a
detailed discussion on the use of external argument prepositions in Greek and Ro-
manian nominalizations). However, by-phrases in English seem to obey the same
restrictions as the possessor, as illustrated in (1c) and (2c) (see also Sichel 2010,
2011 and Alexiadou et al. 2013b). This allows us to compare the three languages.
As indicated by the prepositions, (10) and (11) may have both agentive and
causative readings with de către/apo ‘by’, respectively, de la ‘from’/me ‘with’. How-
ever, the non-agentive causer seems to be allowed only with nominalizations from
OE verbs that have a SE counterpart. Nominalizations from non-alternating verbs
like a dezamăgi ‘to disappoint’, a jigni ‘to offend’, a încuraja ‘to encourage’, a ter-
oriza ‘to terrorize’ in Romanian and haropio ‘cheer up’ in Greek do not accept
non-agentive causers (see (12), for Romanian).
(12) încurajarea Mariei de către Ion/*de la știri
encourage.INF.the Maria.GEN by John/from news
‘John’s encouraging Maria/*Maria’s getting encouraged with the news’

(12) suggests that the nominalizations that realize non-agentive causers in (10)
and (11) must be derived from the SE verb form, i.e., they nominalize the anti-
causative psych verb. This is confirmed by the observation that in (10) (and also
in (11)) adjectival modifiers like brusc ‘sudden’, spontan ‘spontaneous’ make the
causative reading more natural, while the agentive one is otherwise the more sa-
lient one:
(13) enervarea spontană a Mariei de la joc/??de către Ion
annoy.INF.the spontaneous Maria.GEN from game/by John
‘Maria’s spontaneously getting annoyed with the game’

In conclusion, we will treat CPNs as nominalizations of the anticausative struc-

ture of SE verb forms. Before we go into the syntactic analysis of CPNs, however,
we will take a closer look at the OE-SE alternation, which, we will argue, is an
instance of the (anti)causative alternation. At the same time, we will show that the
non-agentive causer-PPs that appear in the verbal domain are maintained in the
CPNs, thus indicating that CPNs inherit the non-agentive causer argument from
the verbal anticausative structure.
124 Artemis Alexiadou and Gianina Iordăchioaia

3. Argumental PPs in the OE-SE alternation in Greek and Romanian

In Greek and Romanian most OE predicates have SE counterparts; Greek has

in fact much fewer exceptions than Romanian. Both languages usually mark the
OE-SE alternation morphologically: the SE variant bears non-active morphology
in Greek (see (14b)) and a reflexive in Romanian (see (15b)). It licenses non-agen-
tive causer PPs with me ‘with’ and de la ‘from’, but not agent PPs with apo/de către
‘by’. In this, they pattern like other anticausatives in the two languages, which we
illustrate in (16) and (17).2
(14) a. Ta nea enohlisan ti Maria.
the news annoyed.ACT the Maria
‘The news annoyed Maria.’
b. I Maria enohlithike me/*apo ta nea
the Maria annoyed.NACT with/by the news
‘Maria got annoyed with/*by the news.’
(15) a. Știrile au enervat-o pe Maria.
news.the have annoyed-her ACC Maria
‘The news annoyed Maria.’
b. Maria s-a enervat de la/*de către știri.
Maria RF-has annoyed from/*by news
‘Maria got annoyed with the news.’
(16) a. I dinati fotia ekapse ti supa
the strong fire burnt.ACT the soup
‘The strong fire burnt the food.’
b. I supa kaike me /??apo ti dinati fotia
the soup burnt.NACT with/??by the strong fire
‘The soup burnt because of the strong fire.’
(17) a. Focul puternic a ars mâncarea.
fire.the strong has burnt food.the
‘The strong fire burnt the food.’
b. Mâncarea s-a ars de la/*de către focul puternic.
food.the RF-has burnt from/by fire.the strong
‘The food burnt because of the strong fire.’

In Greek there are psych verbs as in (18) that do not mark the alternation mor-
phologically, but the same holds of the causative alternation as in (19):

(18) a. Ta nea thimosan to Jani

the news angered.3PL the John
‘The news angered John.’
 Causative nominalizations and psych verbs 125

b. O Janis thimose me/??apo ta nea3

the John angered.3SG with/from the news
‘John got angry with the news.’
(19) a. O ilios stegnose ta ruha
the sun dried.3SG the clothes
‘The sun dried the clothes.’
b. Ta ruha stegnosan me/??apo ton ilio
the clothes dried.3PL with/from the sun
‘The clothes dried from the sun.’

In Romanian, there are very few verbs that do not mark the OE-SE or the caus-
ative alternation morphologically (see (20) and (21), respectively).
(20) a. Soarele puternic l-a căpiat pe Ion.
sun.the strong him-has crazed ACC John
‘The strong sun made John crazy.’
b. Ion a căpiat de la/*de către soarele puternic.
John has crazed from/by sun.the strong
‘John got crazy because of the strong sun.’
(21) a. Soarele puternic a secat iazul.
sun.the strong has drained pool.the
‘The strong sun drained the pool.’
b. Iazul a secat de la/*de către soarele puternic.
pool.the has drained from/by sun.the strong
‘The pool drained because of the strong sun.’

Note that with the SE verbs in (14b), (15b), (18b) and (20b) the preposition that
introduces the non-agentive causer is the same as with the anticausative verbs
in (16b), (17b), (19b) and (21b), namely, me ‘with’ in Greek and de la ‘from’ in
Romanian. In agreement with Doron (2011), we take the presence of a causative
preposition to be suggestive of the causative nature of these verbs.

3.1 A note on non-agentive causers, subject matter and targets

One may observe that in taking anticausative verb forms to involve causativi-
ty and causers, we depart from the traditional view, which is also followed in
the literature that documents the ‘ban on causative psych nominalizations’ (e.g.,
Grimshaw 1990; Pesetsky 1995; Landau 2010) and which takes only the transitive
verb form to be causative. In this view, only the OE version of the verb is consid-
ered causative.
126 Artemis Alexiadou and Gianina Iordăchioaia

We follow Alexiadou, Anagnostopoulou & Schäfer (AAS 2006), who take an-
ticausatives to also involve a cause component (see also Chierchia 2004; Levin
& Rappaport 2005; Doron 2003, 2011; Koontz-Garboden 2009). According to
Alexiadou, Anagnostopoulou & Schäfer, causatives and anticausatives differ from
one another only in the presence of an external argument. That is, both causative
and anticausative variants of a predicate involve causation, but only the former in-
volves the bringing about of the event by an external argument. This is supported
in our examples by the compatibility of anticausatives with non-agentive causer
prepositions (see the b. examples in (14) to (21)).
Importantly, however, we take only eventive SE (and OE) forms to be caus-
ative, i.e., the ones that involve a change of state in the experiencer. As mentioned
in Section 2 and illustrated in (8b) and (9b), psych verbs are ambiguous between
stative and eventive readings. What is, in our view, misleading in the literature on
psych verbs is the fact that often both stative and eventive readings are described
as ‘causative’ (or involving a causer). This started with the linking problem asso-
ciated with psych verbs, which is noted in Belletti & Rizzi (1988), and continued
later, despite Pesetsky (1995), in Arad (1998), Pylkkänen (2000), Landau (2010),
where this issue is not directly addressed.
We think that Greek and Romanian shed more light into the realization
of non-agentive causers in psych verbs, as they employ prepositions to mark
non-agentive causers in eventive readings that are different from those that they
use in stative readings to mark the target or subject matter roles documented in
Pesetsky (1995). This difference can only be observed with SE verb forms, since in
the OE version, as Pesetsky notices, these theta roles are all realized as subjects. As
illustrated in (14), (15), (18), and (20), the non-agentive causer subject in the OE
example in (a) is realized as a non-agentive causer me/de la-PP in the SE example
in (b). All these examples are eventive, i.e., they involve a change of state in the
experiencer.
The few psych verbs that have a stative SE variant, however, use a different
preposition to realize the target/subject matter arguments that surface as a subject
in the OE version (see (22) and (23)). The verbs interest in Greek and gladden in
Romanian are stative on both OE and SE uses (see Alexiadou & Iordăchioaia to
appear for relevant aspectual tests).
(22) a. Ta fita endiaferun to Jani
the plants interest the John
‘Plants interest John.’
b. O Janis endiaferthike ja/*me ta fita
the John interested.NACT for/*with the plants
‘John was interested in plants.’
 Causative nominalizations and psych verbs 127

(23) a. Succesul Mariei îl bucură pe Ion.

success.the Maria.GEN him gladdens ACC John
‘Mary’s success makes John glad.’
b. Ion se bucură de (*la) succesul Mariei.
John RF gladdens of (at) success Maria.GEN
‘John is glad at Maria’s success.’

Greek systematically uses the preposition ja ‘for’ to mark the subject matter ar-
gument with the SE verb, while Romanian employs a few such prepositions (de-
pending on the verb) including de ‘of ’, which differs from the complex preposition
de la ‘from’, that is used for non-agentive causers. A non-agentive causer PP is
completely ruled out in (22b) and (23b), which we relate to the fact that these
verbs are unambiguously stative. Other stative SE constructions like a se teme
‘to fear’, a se gândi ‘to think’, a reflecta ‘to reflect’, a fi bucuros ‘to be happy’ in Ro-
manian also reject non-agentive causer PPs (see also Iordăchioaia, Alexiadou &
Soare to appear). It should be noted that the nominalizations of stative SE verbs
reject non-agentive causer PPs, too:
(24) to endiaferon tu Jani ja ta fita/*me ta fita
the interest the John for the plants/with the plants
‘John’s interest for plants’
(25) bucuria lui față de/*de la succesul Mariei
joy his towards/from success Maria.GEN
‘His joy towards Maria’s success’

3.2 Interim conclusion

We conclude that Greek and Romanian me/de la prepositions mark non-agentive

causers in change of state verbs. Thus the eventive SE verb form is just as caus-
ative as the corresponding eventive OE form. Moreover, it should be noted that
it is only with respect to the eventive reading of psych verbs that we take the OE-
SE alternation to be a subcase of the (anti)causative alternation. The parallelism
between the two alternations crucially hinges on the realization of non-agentive
causer PPs, which are only allowed under eventive/change of state readings. This
is perfectly compatible with the (anti)causative alternation which, it is known,
involves change of state verbs. The stative readings of the OE and SE verb forms
will be left aside, as they do not realize non-agentive causers.
128 Artemis Alexiadou and Gianina Iordăchioaia

4. A syntactic analysis for CPNs

Under our view that the realization of an argumental non-agentive causer PP

indicates causativity (following Alexiadou, Anagnostopoulou & Schäfer 2006),
nominalizations of eventive SE verbs that realize non-agentive causer arguments
are causative. In what follows we treat them as syntactically derived from the an-
ticausative structure of the SE verb form. We first adapt the analysis of the (anti)
causative alternation from Alexiadou, Anagnostopoulou & Schäfer (2006) to the
OE-SE alternation and then present CPNs as nominalizations of the anticausative
structure.

4.1 The syntax of the OE-SE alternation

Following Marantz (2005) and Alexiadou, Anagnostopoulou & Schäfer (2006),

we decompose change of state verbs like those in (16), (17), (19) and (21) into a
Voice, a v and Root/Resultant state component. Under the Voice Hypothesis put
forth in Kratzer (1996), the functional projection of Voice is responsible for the
introduction of external arguments. The same head introduces a DP in the active
and licenses a PP in the passive. The v layer corresponds to Ramchand’s 2008 pro-
cess head: it introduces event implications and in combination with a result state
it is interpreted as causative, and thus licenses non-agentive causers (see Embick
2004; Ramchand 2008; Schäfer 2012). We assume that transitive and anticausative
variants differ in that only the former contain a Voice projection that introduces
an external argument, as illustrated in (26), following Alexiadou, Anagnostopou-
lou & Schäfer (2006).
(26) a. [VoiceP [vP [RootP]]) agentive: de către/apo ‘by’
b. [vP [RootP]] anticausative: de la ‘from’/me ‘with’

The transitive verbs in (16a), (17a), (19a), and (21a) have a Voice projection that
hosts the external argument (i.e. the agent or non-agentive causer subject) as il-
lustrated in (26a). The same structure appears in the passive form of these con-
structions, where Voice hosts the external argument realized as a by-phrase. The
intransitive form of the verbs in (16b), (17b), (19b), and (21b) receives a structure
without Voice and only realizes the non-agentive causer argument as a me/de la-
PP that adjoins to vP as given in (26b). The lack of Voice in this structure accounts
for the unavailability of by-phrases in (16b) and (17b).
We take the structures in (26) to also characterize the OE-SE alternation with
eventive psych verbs as argued in Section 3. Assuming, just like Arad (1998) and
Doron (2011), that the root of psych verbs hosts the experiencer argument, we
 Causative nominalizations and psych verbs 129

associate the structure in (27) with the OE verbs exemplified in (14a), (15a), (18a)
and (20a). The SE verbs in (14b), (15b), (18b) and (20b) receive the structure
in (28).
(27) VoiceP
3
DP Voice′
3
Voice vP
3
v
√
3
√ Exp
(28) vP
3
v √
3
√ Exp

A note on Voice morphology and the OE-SE alternation is in place here. In Greek
several verbs surface with active morphology (i.e., without special marking on
the SE counterpart, see (18)). This is straightforwardly explained by (28), which
lacks Voice. But in Romanian there are only a few verbs that lack reflexive mark-
ing in the SE form (see (20)). The majority of the verbs in the SE form surface
with non-active morphology in Greek (and reflexive in Romanian), a form that
has been argued in the literature to be associated with a special kind of Voice
(expletive in Schäfer 2008; middle in Doron 2003 and Alexiadou & Doron 2012)
or absence of a specifier in v/Voice (Embick 1998; Alexiadou & Anagnostopoulou
2004). Crucially for our purposes, this kind of Voice is non-agentive, and thus
even if it is present in the verbal structure which feeds the nominalization, it does
not lead to an agentive interpretation.

4.2 CPNs

In principle, nominalizations should be available from both structures in (27)

and (28) and this is indeed the case in Greek and Romanian. As mentioned in
Section 2, the nominalizations in (10) and (11), repeated below, are ambiguous
between an agentive and a causative reading.
130 Artemis Alexiadou and Gianina Iordăchioaia

(10) enervarea Mariei de către Ion/de la joc

annoy.INF.the Maria.GEN by John/from game
‘John’s annoying Maria/Maria’s getting annoyed from the game’
(11) i enohlisi tis Marias apo to Jani/me ta nea
the bothering the Maria.GEN by the John/with the news
‘John’s bothering Maria/Maria’s getting bothered with the news’

The input to CPNs is the structure in (28), which realizes non-agentive causer PPs
and disallows agents; the whole nominalization structure is given in (29).
(29) nP
3
n vP
-re/-si 3
PP vP
de la/me 3
v √
-a/-i 3
√ Exp
enerv-/enohl- Maria

The other nominalization pattern is the one of passive nominals (see Grimshaw
1990), which nominalizes the transitive structure with Voice, but in its passive
form, given that it realizes the external argument as a by-phrase.4 This is given
in (30).
(30) nP
3
n VoiceP
-re/-si 3
PP VoiceP
de către/apo 3
Voice vP
3
v √
-a/-i 3
√ Exp
enerv-/enohl- Maria

The construction in (29) is predicted to be incompatible with apo/de către-PPs,

while the one in (30) should be incompatible with me/de la-PPs. Let us test this
prediction. There are not many eventive SE verbs that lack an OE counterpart
 Causative nominalizations and psych verbs 131

to unambiguously derive a SE nominalization that would allow non-agentive

causer PPs and rule out by-PPs. One example is the Romanian verb a se aprinde
‘(lit.) to inflame; (fig.) to burst’. To the extent that a corresponding nominaliza-
tion is accepted, it only allows de la-PPs, as expected (see (31)). There are several
non-alternating agentive OE verbs whose nominalizations disallow non-agentive
causer-PPs in favor of by-PPs, thus confirming the prediction: teroriza ‘terrorize’,
încuraja ‘encourage’, jigni ‘offend’ (see (32)).
(31) aprinderea bruscă a Mariei de la un fleac/*de către Ion
brust.INF.the sudden of Mary.GEN from a trifle/by John
‘Mary’s outburst because of a trifle/*by John’
(32) terorizarea Mariei de către Ion/*de la un fleac
terrorize.INF.the Mary.GEN by John/from a trifle
‘John’s terrorizing Mary/*Mary’s terrorizing from a trifle’

5. CPNs in English

After having identified two possible sources for CPNs, we now investigate their
availability in English and then approach the previous discussion in the literature
from the present perspective.
We saw that Greek and Romanian derive CPNs from the anticausative struc-
ture in (29) and the question that arises is whether English has such nominaliza-
tions as well. The psych nominalizations that have been argued in the literature to
lack causative force do not seem to allow an SE reading: the examples in (33) are
not interpreted as anticausative. Humiliation and embarrassment are ambiguous
between an interpretation in which Mary is the experiencer of the OE verb, and
a stative reading (i.e., Mary’s state of humiliation/embarrassment), while amuse-
ment prefers the stative reading.
(33) Mary’s humiliation/embarrassment/amusement
≠ Mary’s getting humiliated/embarrassed/amused

The explanation for this behavior can be found in the structure of the English lex-
icon, in particular, in Pesetsky’s (1995: 73, 96) observation that English has very
few SE verbs that have an OE alternate. Among the few SE verbs that enter the
OE-SE alternation we find worry, puzzle, grieve, delight. Importantly for us, these
are not change of state verbs as their incompatibility with an in-adverbial indi-
cates in (34);5 they are either activities, or states, or they are ambiguous between
the two. We thus do not expect the nominalizations worry, puzzle, grief, delight to
132 Artemis Alexiadou and Gianina Iordăchioaia

involve non-agentive causers in the way that Greek and Romanian CPNs do. This
is confirmed by (35).
(34) a. The television set worried John.
b. John worried about the television set (*in five minutes).
c. Sue’s remarks puzzled us.
d. We puzzled over Sue’s remarks (*in five minutes).
e. The court decision grieved Sue.
f. Sue grieved over/at the court decision (*in five minutes).
g. His new-found wealth delighted Bill.
h. Bill delighted in his new-found wealth (*in five minutes).
(35) *the TV set’s worry of John/*John’s worry from the TV set

In addition, English does not seem to have SE-only verbs that involve a change
of state, i.e. inchoative, which would also receive the structure in (28) and could
accommodate non-agentive causer PPs. Pesetsky (1995: 96) notes that SE verbs in
English are either states (e.g., like, love, adore, hate) or activities (e.g., fret, mourn,
rage, enjoy), but no SE verb can express inchoatives like become pleased/sad/
amused (see Alexiadou & Iordăchioaia to appear for more discussion on the lack
of a psych causative alternation in English).
In conclusion, it seems to be a peculiarity of the English lexicon that it does
not have inchoative psych verbs. English psych verbs thus lack the anticausative
structure in (28). Under our working assumption that deverbal nominalizations
only inherit argument structure together with the verbal structure that originally
licenses it in the verbal domain (Grimshaw 1990; Alexiadou & Grimshaw 2008),
we now understand why anticausative CPNs do not exist in English: this language
lacks anticausative psych verbs and, as a result, it also lacks the corresponding
nominalizations. Note, however, that the anticausative structure in (26b) is avail-
able in English non-psych verbs and nominalizations as in (36b, c).
(36) a. Inflation diminished his salary.
b. His salary diminished.
c. the diminishment of his salary

5.1 CPNs from OE verbs

We saw that English lacks anticausative SE verbs and, implicitly, also anticausative
CPNs. However, English verbs do exhibit the causative transitive structure in (27)
and this is expected to derive transitive and passive nominalizations, a predic-
tion that is also borne out: humiliation and embarrassment in (37) can realize the
 Causative nominalizations and psych verbs 133

external argument as a possessive or as a by-phrase, thus indicating a transitive

and a passive nominalization.6
(37) a. John humiliated/embarrassed Mary.
b. John’s humiliation/embarrassment of Mary
c. Mary’s humiliation/embarrassment by John

The crucial fact about these nominalizations, however, is that they cannot realize
non-agentive causers, instead of agents, despite the compatibility of non-agentive
causers with the corresponding verbs (see (1) above and (38)).
(38) a. The event embarrassed Mary.
b. *the event’s embarrassment of Mary
c. *Mary’s embarrassment by the event

In view of our discussion so far, we can conclude that ‘the ban on causative psych
nominalizations’ mentioned in the literature refers to a ban on the realization of
non-agentive causers in transitive and passive nominalizations, as illustrated in
(38). This seems to hold of the Romanian nominalizations in (10), as well, to the
extent that a non-agentive causer cannot be realized with de către:7
(39) a. Jocul/Ion a enervat-o pe Maria.
game.the/John has annoyed-her ACC Mary
‘The game/John annoyed Mary.’
b. enervarea Mariei de către Ion/*joc
annoy.INF.the Mary.GEN by John/game
‘John’s/the game’s annoying Mary’

However, as Pesetsky (1995) observed, the restriction on the realization of

non-agentive causers in transitive nominalizations does not apply only to psych
verbs, but also to other causative verbs: e.g., the nominalization of the transitive
verb in (36a) is unavailable in (40):
(40) *the inflation’s diminishment of his salary

This issue has been dealt with in the literature outside the context of psych verbs
(see, for instance, Grimshaw 1990; Marantz 1997; Harley & Noyer 2000; Sichel
2010, 2011) and it turns out to also be available in the verbal domain (see
Alexiadou et al. 2013b for an extended discussion and an analysis of these facts
with respect to the non-psych domain). In Greek and Hebrew passives, for in-
stance, it has been shown that non-agentive causers are allowed by the active, but
not by the passive Voice (cf. (41a–b) in Hebrew, and see more details in Doron
2003 and Alexiadou & Doron 2012).
134 Artemis Alexiadou and Gianina Iordăchioaia

(41) a. sibot kalkaliyot hošivu otam b-a-negev

economic reasons seat.CAUS ACC-them in-the-Negev
‘Economic reasons made them inhabit the Negev.’
b. hem hušvu b-a-negev al-yedey
they seat.CAUS.PASS in-the-Negev by
ha-šiltonot/*sibot kalkaliyot
the-authorities/economic reasons
‘They were seated in the Negev by the authorities/*by economic reasons.’

While it is beyond the aim of this paper to account for the complexity of this
phenomenon, we refer the reader to Alexiadou et al. (2013a) and Alexiadou et al.
(2013b) for discussion and a proposal for the various instantiations of this restric-
tion in different languages.

6. Conclusions and predictions

We have shown that Greek and Romanian can realize non-agentive causers in
psych nominalizations, despite previous predictions that psych nominalizations
might lack causative force cross-linguistically (Landau 2010). We explained,
however, that the literature that was extensively concerned with this issue (e.g.,
Grimshaw 1990; Pesetsky 1995) strictly refers to transitive/passive psych nom-
inalizations, which seem to resemble the corresponding nominalizations from
non-psych verbs in their restrictions on the realization of external arguments
(Harely & Noyer 2000; Sichel 2010, 2011; Alexiadou et al. 2013b).
Under the assumption that anticausatives also involve causation (Chierchia
2004; Doron 2003, 2011; Levin & Rappaport 2005; Alexiadou, Anagnostopoulou
& Schäfer 2006), we showed that Greek and Romanian can derive causative psych
nominalizations from the anticausative version of alternating psych verbs, to the
extent that they may employ argumental non-agentive causer PPs, just like their
verbal counterparts. The lack of such nominalizations in English was explained
via the lack of anticausative forms for psych verbs in this language (cf. Pesetsky
1995 and Alexiadou & Iordăchioaia to appear).
Our prediction is thus that every language whose OE psych verbs alternate (or
which has inchoative SE verbs) should be able to derive CPNs, just like Greek and
Romanian. This is supported by the Hebrew data in (42), which were provided to
us by Edit Doron and confirmed by Ivy Sichel. In addition, these nominalizations
exhibit middle morphology, just like the SE anticausative, thus supporting our
analysis that they are derived from this verbal form.
 Causative nominalizations and psych verbs 135

(42) a. ha-hit’acvut Selo me-ha-xadaSot

the-saddening.MID his from-the-news
b. ha-hit’odedut Selo me-ha-xadaSot
the-encouragement.MID his from-the-news
c. ha-hizda’ze’ut Selo me-ha-xadaSot
the-becoming-shocked.MID his from-the-news

Notes

* We thank Edit Doron, Florian Schäfer, Ivy Sichel, two anonymous reviewers, the editors of
this volume and the audience of the workshop ‘Structuring the Argument’ at the University of
Paris 8, for useful comments and remarks. Our research has been supported by a DFG grant to
project B1 within the Collaborative Research Center (SFB) 732 at the University of Stuttgart.

1. The precise aspectual nature of the eventive SE form in the sense of Marín & McNally
(2011) is independent of the realization of a causer argument, which is our focus here, so we
will not address this issue in further details.

2. To the extent that (17b) with an agent might be marginally available, it would receive an
impersonal passive reading.

3. See Note 7 below for discussion on apo.

4. In English, this may be a passive with by (e.g., ‘the city’s destruction by the enemy’) or
a transitive structure with a possessive, where the DP external argument occupies Spec, DP,
(e.g, ‘the enemy’s destruction of the city’). As already mentioned, since Greek and Romanian
cannot realize the external argument as a possessive, they only have the passive structure, see
Alexiadou (2001), Borer (2003). Borer (2003) argued that the transitive nominalization is pas-
sive, too, an analysis we are led to assume, given that the restriction that applies to the prenom-
inal genitive and the by-phrase is identical.

5. Here, we assume that all change of state verbs are telic and thus should be compatible with
an in-adverbial.

6. Thanks to Amanda Kahrsch and Christopher Piñón for judgements on the data.

7. The situation in Greek is slightly more complicated, as apo, unlike de către can be used to
introduce direct causers (see Alexiadou, Anagnostopoulou & Schäfer 2006; Alexiadou et al.
2013a; Alexiadou et al. 2013b).

References

Alexiadou, A. 2001. Functional structure in nominals: nominalization, and ergativity. Amster-

dam: John Benjamins. DOI: 10.1075/la.42
136 Artemis Alexiadou and Gianina Iordăchioaia

Alexiadou, A. & E. Anagnostopoulou. 2004. Voice morphology in the causative-inchoative Al-

ternation: evidence for a non unified structural analysis of unaccusatives. In A. Alexiadou,
E. Anagnostopoulou, and M. Everaert (eds.), The Unaccusativity puzzle. Oxford University
Press. DOI: 10.1093/acprof:oso/9780199257652.001.0001
Alexiadou, A., E. Anagnostopoulou & F. Schäfer. 2006. The properties of anticausatives
cross-linguistically. In M. Frascarelli (ed.), Phases of Interpretation, 187–212. Berlin:
Mouton. DOI: 10.1515/9783110197723.4.187
Alexiadou, A., E. Anagnostopoulou & F. Schäfer. 2009. PP licensing in nominalizations. Pro-
ceedings of NELS 38.
Alexiadou, A. & G. Iordăchioaia. to appear. The psych causative alternation. Lingua.
Alexiadou, A., G. Iordăchioaia, F. Martin, F. Schäfer & M. Cano. 2013a. ‘Direct participation’
and ‘agent exclusivity’ effects in derived nominals and beyond. In Iordăchioaia, G., Roy, I.,
Takamine, K. (eds.), Categorization and category change, 153–180. Newcastle upon Tyne:
Cambridge Scholars Publishing.
Alexiadou, A., G. Iordăchioaia, M. Cano, F. Martin & F. Schäfer. 2013b. The realization of ex-
ternal arguments in nominalizations. Journal of Comparative Germanic Linguistics 16(2):
73–95.
Alexiadou, A. & E. Doron. 2012. The syntactic construction of two non-active Voices: passive
and middle. Journal of Linguistics 48. DOI: 10.1017/S0022226711000338
Alexiadou, A. & J. Grimshaw. 2008. Verbs, nouns and affixation. In Schäfer, F. (ed.), SinSpeC (1):
Working Papers of the SFB 732, University of Stuttgart, 1–16.
Anagnostopoulou, E. 1999. On experiencers. In A. Alexiadou, G. Horrocks & M. Stavrou (eds),
Studies in Greek syntax, 67–93. Dordrecht: Kluwer. DOI: 10.1007/978-94-015-9177-5_4
Arad, M. 1998. Psych-notes. UCL Working Papers in Linguistics 10: 203–223.
Arad, M. 2002. Universal features and language particular morphemes. In A. Alexiadou (ed.)
Theoretical approaches to universals. Amsterdam: John Benjamins.
Belletti, A. & L. Rizzi. 1988. Psych-Verbs and Theta-Theory. Natural Language and Linguistic
Theory 6: 291–352 DOI: 10.1007/BF00133902
Borer, H. 2003. Exo-Skeletal vs. Endo-Skeletal Explanations: Syntactic Projections and the Lex-
icon. The Nature of Explanation in Linguistic Theory. J. Moore and M. Polinsky (eds.). CSLI
Publications.
Chierchia, G. 2004. A semantics for unaccusatives and its syntactic consequences. In The unac-
cusativity puzzle: explorations of the syntax-lexicon interface, A. Alexiadou, E. Anagnost-
opoulou, and M. Everaert (eds.), 22–59. Oxford: Oxford University Press.
DOI: 10.1093/acprof:oso/9780199257652.003.0002
Doron, E. 2003. Agency and voice: the semantics of the semitic templates. NLS 11: 1–67.
Doron, E. 2011. Agents and causers. Paper presented at the Roots III Workshop. Jerusalem, June
2011.
Embick, D. 1998. Voice Systems and the Syntax/Morphology Interface. In The Proceedings of the
Penn/MIT Workshop on Aspect, Argument Structure, and Events, H. Harley (ed.), MITWPL
Embick, D. 2004. ‘On the structure of resultative participles in English’, Linguistic Inquiry 35.3,
355–392. DOI: 10.1162/0024389041402634
Grimshaw, J. 1990. Argument Structure. Cambridge: MIT Press.
Harley, H. & Noyer, R. 2000. “Licensing in the non-lexicalist lexicon”. In B. Peeters, ed., The
Lexicon/Encyclopaedia Interface, 349–374. Amsterdam: Elsevier Press.
Iordăchioaia, G. 2008. External argument PPs in Romanian nominalizations. In F. Schäfer (ed.),
Working Papers of the SFB 732 (1), 71–84.
 Causative nominalizations and psych verbs 137

Iordăchioaia, G., A. Alexiadou & E. Soare. to appear. The structural source of stative nomi-
nalizations from psych verbs. In Cahiers Chronos 27. Amsterdam/Philadelphia: Rodopi.
Iwata, S. 1995. The Distinctive Character of Psych-Verbs as Causatives. Linguistic Analysis 1–2,
95–120.
Koontz-Garboden, A. 2009. Anticausativization. Natural Language and Linguistic Theory 27:
77–138. DOI: 10.1007/s11049-008-9058-9
Kratzer, A. 1996. Severing the External Argument from its Verb. Phrase Structure and the Lexicon.
J. Rooryck and L. Zaring. Dordrecht: Kluwer. 109–137. DOI: 10.1007/978-94-015-8617-7_5
Lakoff, G. 1970. Irregularity in Syntax. New York: Holt, Reinhart and Winston.
Landau, I. 2010. The Locative Syntax of Experiencers. Cambridge, Mass: MIT Press.
Levin, B. & M. Rappaport Hovav. 2005. Argument realization. Cambridge: Cambridge Univer-
sity Press. DOI: 10.1017/CBO9780511610479
Marantz, A. 1997. No escape from syntax: Don’t try morphological analysis in the privacy of
your own lexicon. In University of Pennsylvania Working Papers in Linguistics, Vol. 4.2,
A. Dimitriadis, L. Siegel, C. Surek-Clark, and A. Williams (eds.), 201–225. Philadelphia.
Marantz, A. 2005. Objects out of the lexicon: objects as event. Handout presented at the Uni-
versity of Vienna.
Marín, R. & McNally, L. 2011. Inchoativity, change of state, and telicity. Natural Language and
Linguistic Theory 29: 467–502.
Martin, F. 2006. Prédicats statifs, causatifs et résultatifs en discours. Sémantique des adjectifs
évaluatifs et des prédicats psychologiques [Stative, causative and resultative predicates in
discourse. Semantics of evaluative adjectives and psychological verbs], PhD, Université
libre de Bruxelles.
Myers, S. 1984. Zero-derivation and inflection. In MIT Working Papers in Linguistics 7: Papers
from the January 1984 MIT Workshop in Morphology 7: 53–69.
Pesetsky, D. 1995. Zero Syntax: Experiencers and Cascades. Cambridge: MIT Press.
Pylkkänen, L. 2000. On Stativity and Causation. In Tenny, C. L., and J. Pustejovsky (eds.), Events
as Grammatical Objects: The Converging Perspective of Lexical Semantics and Syntax. CSLI
Publications, Stanford CA, 417–444.
Ramchand, G. 2008. Verb Meaning and the Lexicon. Cambridge University Press, Cambridge.
DOI: 10.1017/CBO9780511486319
Schäfer, F. 2008. The Syntax of (Anti-)Causatives. External arguments in change-of-state con-
texts. Amsterdam/Philadelphia: John Benjamins. DOI: 10.1075/la.126
Schäfer, F. 2012. Two types of external argument licensing: the case of causers. Studia Linguis-
tica 66(2): 128–180.
Sichel, I. 2010. Event Structure Constraints in Nominalization. In A. Alexiadou & M. Rathert
(eds), The Syntax of Nominalizations across Languages and Frameworks, 151–190. Berlin:
Mouton.
Sichel, I. 2011. Nominalization, causativization, and category-free syntax. Paper presented at the
Roots III Workshop. Jerusalem, June 2011.
Part III

Neurobiological models
 Neurocognitive mechanisms of verb
argument structure processing*

Cynthia K. Thompson1,2,3 and Aya Meltzer-Asscher4,5

1 Ralph and Jean Sundin Professor, Department of Communication
Sciences and Disorders / 2 Department of Neurology / 3 Cognitive
Neurology and Alzheimer’s Disease Center, Northwestern University /
4 Department of Linguistics, Tel Aviv University / 5 Sagol School of

Neuroscience, Tel Aviv University

1. Introduction

Different verbs appear in different syntactic environments, as can be seen even

in simple sentences such as (1)–(3). The verbs sneeze, hit and hear differ in their
thematic requirements, i.e. the number and type of thematic roles they encode.
Sneeze has only one thematic role, an agent, and thus must be accompanied by
only one argument (1), whereas hit has two thematic roles, an agent and a theme,
and must be accompanied by two arguments (2a, b). Verbs also differ in their
subcategorization requirements, i.e. requirements with regard to the syntactic
realization of their arguments. Thus, the theme argument of hit can be realized
only as a noun phrase ((2b) vs. (2c)), while the theme argument of hear can be
realized by either a noun phrase or a clause (3b, c). This information about verbs
is often referred to as verb ‘argument structure’ information, and it is central to
what speakers know when they know a language.
(1) a. John sneezed.
b. *John sneezed [the doctor].
(2) a. *John hit.
b. John hit [the doctor].
c. *John hit [that the doctor was wrong].
(3) a. *John heard.
b. John heard [the doctor].
c. John heard [that the doctor was wrong].
142 Cynthia K. Thompson and Aya Meltzer-Asscher

This chapter addresses the neural mechanisms involved in the processing of verb
argument structure. We first discuss how current models of sentence processing
incorporate the concept of argument structure, and how these views fit with dif-
ferent theoretical linguistic approaches. We then present key results from neu-
ropsychological, neuroimaging and EEG experiments manipulating argument
structure complexity in different ways. These results reveal a distributed neural
network involved in the processing of argument structure information, leading to
a proposed neurocognitive model of such processing.

2. Argument structure in language processing models

and theoretical linguistics

Many current language processing models discuss the use of argument structure
information during sentence comprehension and production. Sentence produc-
tion models postulate levels of sentence formulation, including non-linguistic
(pre-verbal) conceptualization of an intended message and subsequent transfor-
mation of the message into a linguistically specified sentence via grammatical
and morpho-phonological encoding processes (Bock, 1995; Bock & Levelt, 2002;
Levelt, 1999; Ferreira & Slevc, 2007). Although conceptualized in the aforemen-
tioned models as serial processes, these same levels are postulated in interactive
models, which allow both forward and backward checking mechanisms (Chang,
Dell, & Bock, 2006; Dell & Sullivan, 2004; Stemberger 1985, and others). Gram-
matical encoding involves selection of lemmas, which entail both semantic and
syntactic/grammatical information, but are not specified for phonological form.
Verb lemmas include, among other things, argument structure information,
which is used for ensuing syntactic structure building. As lemmas become avail-
able, the process of building syntactic structure is initiated, generating a hierar-
chically structured functional relationship between lemmas. Hence, the syntactic
properties specified within each lemma’s entry enable the grammatical encod-
er to project phrasal nodes and concatenate those into a sentence structure. The
morpho-phonological encoder activates lexemes (i.e., the phonological form of
selected lemmas) and specifies the morphological form of function-assigned
lemmas. Therefore, the output of the morpho-phonological encoding process is a
linearly ordered string of words and morphemes with phonological specification.
In accordance with these models, there is psycholinguistic evidence showing that
argument structure information is utilized in pre-speech sentence planning (Lee
& Thompson, 2011; Meyer, 1996).
Sentence comprehension models include roughly the same processing levels
as sentence production, in reversed order. For example, Friederici (2002, 2011)
 Neurocognitive mechanisms of verb argument structure processing 143

suggests that comprehension begins with low level phonological processing, fol-
lowed by assignment of word-class categories and building of basic phrase struc-
ture. After this initial syntactic stage, argument structure information is accessed,
and serves to establish syntactic and semantic (thematic) relations between words.
In the final processing stage, a full-fledged semantic interpretation of the sentence
is built. If the initial relations postulated between words at previous stages turn
out to be erroneous, syntactic reanalysis takes place.
Many psycholinguistic studies show that argument structure information
encoded within a verb’s representation is automatically accessed when verbs are
encountered, and used in real time during sentence comprehension (Boland,
2005; Boland, Tannenhaus, & Garnsey, 1990; Ferretti, McRae, & Hatherell, 2001;
Friedmann, Taranto, Shapiro, & Swinney, 2008; MacDonald, Pearlmutter, &
Seidenberg, 1994; Trueswell & Kim, 1998; Trueswell, Tanenhaus, & Kello, 1993,
and many others). In a series of seminal papers, Shapiro and colleagues (Shapiro,
Brookins, Gordon, & Nagel, 1991; Shapiro, Zurif, & Grimshaw, 1987) showed, us-
ing a cross-modal lexical decision paradigm, that lexical decision times to visually
presented letter strings were longer after hearing verbs like hear, with multiple
subcategorization options, than after hearing verbs like hit, with a single option
(see examples (2)–(3) above). Gorrell (1991) reported similar findings for oblig-
atory transitive verbs (hit) as compared to intransitive verbs (sneeze) (see exam-
ples (1)–(2) above). These results are predicted if upon encountering a verb, its
argument structure properties are immediately activated; the more complex these
properties, the more processing they require, accounting for delays in concurrent
visual lexical decision.
As can be understood from the discussion above, current comprehension
and production models adopt the assumption that detailed argument structure
information is included in the lexical entries of verbs. This view is in line with tra-
ditional linguist approaches to argument structure (Jackendoff, 1972; Williams,
1981; Levin & Rappaport, 1986, among many others; and see more recent ver-
sions in Reinhart, 2002; Horvath & Siloni, 2011). For example, Chomsky (1981)
suggests that the lexical information associated with a verb contains a list of the
thematic roles that the verb assigns, as well as subcategorization frames. Accord-
ing to this view, which is often referred to as the lexicalist view, lexical informa-
tion associated with verbs determines the syntactic structure in which the verb
appears, as well as the interpretation of its accompanying arguments.
In the last two decades, alternative approaches to argument structure have
emerged in the linguistic literature. Under the radical constructivist approach,
lexical entries do not include any grammatically relevant information. This means
that thematic specifications do not exist in the lexicon, and that syntactic struc-
ture building is not guided by thematic information associated with verbs. Rather,
144 Cynthia K. Thompson and Aya Meltzer-Asscher

structure building is free, restricted only by world knowledge (Borer, 2005). As

noted by Ramchand (2008), this view is problematic, even when attempting to
explain simple selection phenomena such as those illustrated in (1)–(3) above.
In addition, this approach is not compatible with the psycholinguistic findings
mentioned above (as well as the imaging studies to be discussed below), showing
differences in processing between verbs with different argument structure speci-
fications. If argument structure information is not encoded together with verbal
lexical entries at all, such differences are not predicted to exist.1
Yet other theoretical approaches to argument structure offer an intermediate
view, falling between the aformentioned detailed lexical representation of argu-
ment structure in the lexicon and a complete lack thereof (e.g. Marantz 1997;
Embick 2004; Pylkkanen 2008). For example, as mentioned above, Ramchand
(2008) holds that in order to explain basic selection facts (as in (1)–(3) above)
some grammatical information must be encoded on lexical items, in contrast to
the constructivist view. However, Ramchand suggests that this grammatical infor-
mation can be kept to a minimum by including only an aspectual/event structure
template, and avoiding detailed thematic specifications. Whether these approach-
es can accommodate the results of psycho- and neurolinguistic studies and be
incorporated into models of sentence comprehension and production remains, at
this point, an open question.

3. The neural instantiation of argument structure processing

In the last two decades, researchers have sought to elucidate the neural underpin-
nings of argument structure processing. This was initially done in neuropsycho-
logical studies with aphasic individuals, where argument structure effects were
examined in processing and production across different patient neuropatholo-
gies, with particular focus on agrammatic individuals. The picture emerging from
these studies with regard to the network of brain regions involved in argument
structure processing and the time course of this processing has been further elab-
orated in recent years using event-related potentials (ERPs) and neuroimaging
techniques such as fMRI.
Because, as discussed above, argument structure processing takes place as an
inherent part of any instance of sentence production and comprehension, it is
challenging to distinguish the neural correlates of argument structure processing
from those of other processes involved in sentence processing, such as phrase
structure building or semantic integration. To overcome this problem, and tar-
get behavior or neural activation specifically associated with processing of argu-
ment structure information, many experiments explicitly contrast stimuli with
 Neurocognitive mechanisms of verb argument structure processing 145

less versus more complex argument structure, keeping other factors constant. In
this section, we discuss two argument structure complexity contrasts investigated
in recent years: those involving the number of thematic roles or number of the-
matic/subcategorization options (Section 3.1), and those involving unaccusativ-
ity vs. unergativity (Section 3.2).2 Another way to target processing of argument
structure information is by using grammaticality contrasts, with correct sentenc-
es compared to sentences including violations of argument structure information.
Studies utilizing this strategy are discussed in Section 3.3.

3.1 Brain structures associated with number of thematic roles

and thematic options

One straightforward aspect of argument structure complexity is the number of

thematic roles a verb encodes. For example, a transitive verb (e.g. hit, see (2)
above) is more complex than an intransitive verb (e.g. sneeze, (1) above). Like-
wise, a ditransitive verb (e.g. send as in John sent [the flowers] [to Mary]) is more
complex than a transitive verb. A similar but different aspect of complexity is
presented by the number of subcategorization options a verb has. For example,
although hear and hit are both transitive, hear, which has two subcategorization
options (it can select either a noun phrase or a clause, see (3) above), is more
complex than hit, which has only one option (it invariably selects a noun phrase,
(2) above).
Studies have shown that Broca’s aphasic individuals with (primarily) left an-
terior brain damage show normal access to subcategorization frames during on-
line sentence processing. That is, their reaction times (RTs) in a secondary task
are longer following verbs with multiple subcategorization options versus verbs
with only one such option, as they are in non-brain-damaged volunteers (Shapiro,
Gordon, Hack, & Killackey, 1993; Shapiro & Levine, 1990).3 In addition, as shown
in Kim & Thompson (2000, 2004), agrammatic individuals’ single verb compre-
hension is preserved, with no differences in comprehension between verbs with
different numbers of thematic roles (also see Thompson, Lukic, King, Mesulam,
& Weintraub, 2012).
In contrast to agrammatic aphasic patients’ relatively spared verb comprehen-
sion and access to argument structure information during sentence processing,
these patients show production deficits associated with argument structure com-
plexity. Several studies have shown that as the number of thematic roles increases,
verb production becomes more difficult for agrammatic aphasic patients; that is,
transitive and ditransitive verbs are more difficult to produce than intransitive
verbs. This pattern has been noted in English-speaking patients in verb naming
tasks (Kemmerer & Tranel, 2000; Kim & Thompson, 2000, 2004; Thompson,
146 Cynthia K. Thompson and Aya Meltzer-Asscher

Lange, Schneider, & Shapiro, 1997). For example, a study by Kim & Thompson
(2004) examined production of verbs with one, two or three thematic roles in nine
patients with agrammatic aphasia and 14 individuals with probable Alzheimer’s
disease. Results showed impaired verb naming for both patient groups. However,
only the agrammatic aphasic participants showed an effect of number of thematic
roles, exhibiting progressively greater difficulty with verbs associated with more
roles. Parallel results were obtained in verb naming tasks with agrammatic speak-
ers in German (DeBlesser & Kauschke, 2003), Hungarian (Kiss, 2000), and Italian
(Luzzatti et al., 2002).
Similar effects have been shown in sentence production tasks. Thompson et
al. (1997) found that fewer correct sentences were produced with ditransitive and
transitive verbs than with intransitive verbs (in this study, a sentence was counted
as correct only if it contained all of the arguments required by the verb). Parallel
results are reported for Russian in Dragoy & Bastiaanse (2010), where sentences
with two arguments were produced less accurately than sentences with one argu-
ment. Similarly, Bastiaanse & Jonkers (1998) found that in spontaneous speech,
agrammatic Dutch speakers produced fewer sentences with two and three argu-
ments than with one argument.4
The picture that emerges is that agrammatic individuals’ representation of the
number of thematic roles and subcategrization frames seems to be spared, as evi-
denced by their normal reaction times and comprehension patterns, but that inter-
action of this knowledge with production is impaired. Assuming that agrammatism
is often correlated with left inferior frontal lesions (as suggested by lesion analysis
studies such as Dronkers, Wilkins, Van Valin, Redfern, & Jaeger, 2004; Kertesz,
Harlock, & Coates, 1979), this may mean that these regions are not involved in the
representation of argument structure per se, but only come into play in production
tasks. The fact that Wernicke’s aphasic individuals do not exhibit normal sensitivity
to subcategorization options raises the possibility that this knowledge, and knowl-
edge about argument structure information in general, is instantiated in posterior
temporal and inferior parietal regions, which are often damaged in these patients
(Dronkers et al., 2004; Kertesz, 1977; Kertesz et al., 1979).
In a series of fMRI studies, we explicitly tested this possibility first in young
normal volunteers (Thompson et al., 2007) and later in a study with older un-
impaired and agrammatic aphasic participants (Thompson, Bonakdarpour, &
Fix, 2010). In both studies, we examined verb processing using a lexical decision
task. Thompson et al. (2007) tested intransitive, transitive and ditransitive verbs.
Stimuli of each type, along with nouns and pseudowords, were presented visually
using an event-related fMRI design. Results for 14 young normal participants
showed activation in the supramarginal and angular gyri, limited to the left hemi-
sphere, when transitive verbs were compared to intransitive ones. Additionally,
 Neurocognitive mechanisms of verb argument structure processing 147

when we compared both transitive and ditransitive verbs to intransitive verbs, we

found bilateral activation in these same regions. These patterns suggest that pos-
terior perisylvian regions are engaged for the processing of argument structure
information associated with verbs, in particular information referring to the
number of thematic roles, with increased neural tissue in inferior parietal regions
associated with an increased number of roles.
Similar patterns were derived in a follow-up study with older unimpaired
and aphasic listeners (Thompson et al., 2010). Fourteen older healthy listeners
(age 45–68) and five age-matched aphasic patients performed the same lexical
decision task. Results for the age-matched listeners largely replicated those for
the younger participants studied by Thompson et al. (2007): verbs with a greater
number of thematic roles engaged bilateral posterior perisylvian tissue (i.e., the
angular gyrus) more so than verbs with fewer thematic roles. Similar results were
obtained for the agrammatic aphasic patients. All performed the task with high
accuracy and, despite differences in lesion site and extent, recruited spared tissue
in the same regions as healthy controls. However, activation was unilateral (in the
right hemisphere for three participants) rather than bilateral, likely because these
patients’ lesions extended to the left temporoparietal region.
In another study manipulating argument structure complexity, Ben-Shachar,
Hendler, Kahn, Ben-Bashat, & Grodzinsky (2003) found increased bilateral su-
perior temporal sulcus (STS) activation as a function of the number of themat-
ic roles/arguments associated with a verb. However, in that study, verbs by type
were not directly compared with one another. Rather sentences containing verbs
with additively greater numbers of arguments were examined (i.e., sentences with
three versus two arguments in the matrix clause and two versus one arguments
in the embedded clause were compared). These data support the notion that the
posterior perisylvian region is crucially engaged for processing information relat-
ed to verb argument structure.5
With regard to the number of thematic options, a somewhat different picture
emerges. Shetreet, Palti, Friedmann, & Hadar (2007) found activation in the left
posterior superior temporal gyrus, as well as the inferior (BA 47) and middle
frontal gyrus (BA 9) for sentences containing verbs with more as opposed to few-
er thematic and subcategorization options. In a recent study, Meltzer-Asscher,
Schuchard, den Ouden, & Thompson (2013) found a somewhat similar pattern.
We compared alternating verbs, e.g. roll, which have two thematic options (i.e.,
they can be used both transitively and intransitively), with simple intransitive
verbs, e.g. walk, which have only one thematic option, in a lexical decision task.
The contrast of alternating over simple verbs yielded clusters of activation in the
angular gyrus (BAs 39, 40) as well as in the mid-superior frontal gyrus (BAs 8, 9),
bilaterally. Based on the results of Thompson et al. (2007, 2010), we attributed the
148 Cynthia K. Thompson and Aya Meltzer-Asscher

posterior activation to the difference in the number of thematic roles between the
two verb types. In contrast, the mid-superior frontal activation was hypothesized
to reflect differences in thematic options, namely the fact that alternating verbs
have two subcategorization/thematic frames, whereas simple verbs have only one.
The latter claim is in line with Shetreet et al. (2007), as well as previous studies
where mid-superior frontal regions were implicated in the processing of lexical
ambiguity for nouns (Chan et al., 2004; Mason & Just, 2007). Notably, verbs with
alternating transitivity are ambiguous with regard to their thematic options. By
hypothesis, when such verbs are encountered all thematic options for that verb
are automatically activated.6
As can be seen from this discussion, most imaging studies manipulating the
number of thematic roles and thematic options have used listening tasks. In one
study examining verb production, den Ouden, Fix, Parrish, & Thompson (2009)
found frontal, in addition to posterior perisylvian, activation associated with an
increase in the number of thematic roles required by the verb. As in the lexical
decision studies described above (Thompson et al., 2007, 2010), greater activation
was found for transitive versus intransitive verbs in posterior brain regions (i.e.,
bilateral activation in the posterior middle temporal, angular and supramarginal
gyri). However, activation was also detected in the left inferior frontal cortex, par-
ticularly the triangular and opercular parts of the inferior frontal gyrus (IFG, BA
45 and 44, respectively), as well as in the middle frontal gyrus (BA 6, 9, and 46). A
large number of studies (e.g., Grewe et al., 2005; Friederici, Ruschemeyer, Hayne,
& Fiebach, 2003 and others; see review in Friederici, 2011) implicate parts of the
inferior frontal gyrus for building initial phrase structure. The increased activa-
tion in this region associated with production of verbs with more thematic roles
may therefore be interpreted as resulting from a process in which access to the
argument structure of a verb projects the phrase structure required for that verb
for sentence production. The more thematic roles a verb has, the more complex
the syntactic structure built.
Considered collectively, the neuroimaging data discussed above indicate
that posterior perisylvian regions may serve to access information with regard to
number of thematic roles and thematic options for verbs appearing in isolation,
as summarized in Figure 1. Additionally, posterior regions may contribute to the
syntactic and semantic integration of a verb and its arguments, as indicated by
the activation observed in these regions in tasks involving argument structure
processing in sentence contexts. Anterior tissue may be crucial for sustaining ac-
tivation of multiple lexical options (subcategorization frames or argument struc-
tures), and is also engaged for naming actions, exhibiting increasing participation
for verbs with a greater number of thematic roles, possibly due to initial phrase
structure building processes.
 Neurocognitive mechanisms of verb argument structure processing 149

a.

b.

c.

Figure 1. Activations associated with number of thematic roles / thematic options in a

lexical decision task: (a) transitive and ditransitive over intransitive verbs, young healthy
participants (data from Thompson et al., 2007); (b) transitive and ditransitive over in-
transitive verbs, older healthy participants (Thompson et al., 2010); (c) alternating transi-
tivity over intransitive verbs, young healthy participants (Meltzer-Asscher et al., 2013)

3.2 Brain structures associated with unaccusativity

An additional aspect of argument structure complexity is concerned not with the

number of thematic roles a verb encodes, but rather with the mapping of these
roles to syntactic positions. One type of verb that has received some attention in
this respect is the class of unaccusative verbs, first discussed in Perlmutter (1978).
Consider for example (4) and (5).
(4) The girl AGENT walked. (unergative verb)
(5) The girl THEME disappeared. (unaccusative verb)
150 Cynthia K. Thompson and Aya Meltzer-Asscher

Although both unaccusative and unergative verbs are intransitive, as can be seen
in (4)–(5), they differ in that the sole argument of unergative verbs is an agent, a
volitional causer of the action denoted by the verb; The single argument of unac-
cusative verbs, in contrast, is a theme, undergoing – rather than instigating – the
action.7
Unaccusativity is treated differently in different linguistic frameworks. Ac-
cording to the influential Government & Binding (GB) framework of generative
linguistics (as exemplified for example in Burzio, 1986; Levin & Rappaport Hovav,
1995), unaccusative verbs are unique in that they do not assign accusative Case.
In the initial syntactic structure, the argument of unaccusative verbs, which is a
theme argument, is mapped canonically, namely – as an internal argument (just
like the theme of transitive verbs). However, since it cannot be assigned Case in
this position, it subsequently moves to the syntactic subject position, to receive
nominative Case, leaving behind a trace (t) in object position. In contrast, the
argument of unergative verbs, an agent, is base-generated in subject position, and
no movement operation is required, as seen in the contrast in (6)–(7), and illus-
trated in Figure 2.
(6) The girl walked. (unergative verb)
(7) The girli disappeared ti. (unaccusative verb)

Note, that although the initial mapping of the theme argument of unaccusative
verbs is to the canonical direct object position, the subsequent movement renders
the sentence structure noncanonical, in the sense that in this structure a theme
argument precedes, rather than follows the verb. Unaccusative verbs can thus be
considered more complex than unergative ones: when their single argument is
projected, an additional syntactic operation (i.e. movement) is obligatory in order
for a grammatical sentence to result.
The GB account of unaccusativity relies on the assumption that sentences
have basic syntactic structures and surface syntactic structures, which can be dif-
ferent as a result of movement operations. However, not all linguistic frameworks
share these assumptions. Other analyses, which allow only one level of syntac-
tic structure, suggest that unergative and unaccusative sentences do not differ
syntactically, only semantically. For example, Dowty (1991) argues that for any
intransitive verb, if its thematic role has mostly agentive properties (e.g. volitional
involvement, sentience, causing an event) the verb is unergative; if the thematic
role has mostly patient/theme properties (e.g. undergoing change, causally affect-
ed, stationary), the verb is unaccusative. However, the two verb types do not proj-
ect their argument in different syntactic positions; in both cases, the argument
is directly mapped to the syntactic subject position. Similarly, in the Role and
 Neurocognitive mechanisms of verb argument structure processing 151

a. Unergative c. Object relative

TP TP

NP T′ NP T′

The girl T VP Pete T VP

V′ V′

V V NP

walked saw N′
b. Unaccusative
N CP
TP

the girli whoi C′

NP T

C TP
The girli T VP

NP T′
V′

the boy T VP
V NP

was V′
disappeared ti

V NP

chasing ti

Figure 2. According to GB theory, unlike sentences with unergative verbs (a), sentences
with unaccusative verbs involve syntactic movement of the subject from object position
(b). Unaccusative sentences thus resemble other types of noncanonical sentences, such as
object relatives (c)

Reference Grammar framework (van Valin, 1990), an intransitive verb’s thematic

role is determined based on its aspectual classification (verbs denoting activities
select for actors, roughly parallel to agents, whereas other verbs select for under-
goers, roughly resembling themes), but this does not affect syntactic mapping. In
Head-driven Phrase Structure Grammar (HPSG), verbs select for their subjects
(SUBJ) and complements (COMPS) (Pollard & Sag, 1994). In an HPSG account
of the difference between unergative and unaccusative verbs, Dini (1995) pro-
poses that whereas unergative verbs are characterized by the presence of a noun
phrase in the verb’s lexically specified SUBJ list, unaccusative forms have an emp-
ty SUBJ list, and their surface subject is the first member of the COMPS list.
152 Cynthia K. Thompson and Aya Meltzer-Asscher

Thus, although these different theories do not assume that unaccusative verbs
necessitate syntactic movement, all indicate that unaccusative verb require non-
canonical argument mapping, namely mapping of a theme/patient/undergoer
argument to subject position, a position prototypically occupied by agents and
causers. The resulting representation is thus more marked than that of unergative
verbs.
A number of studies have tested production of unaccusative verbs and sen-
tences in agrammatic individuals. Based on spontaneous narrative speech, Kegl
(1995) and Thompson (2003) observed that these speakers produce significantly
fewer sentences with unaccusative verbs than unimpaired speakers. Other studies
investigating production of unaccusative and unergative verbs in sentence contexts
using constrained tasks (Bastiaanse & van Zonneveld, 2005; Lee & Thompson,
2004), have found similar results. For example, Lee & Thompson (2004) found that
unergative sentences were produced correctly significantly more often than unac-
cusative sentences by agrammatic participants, with a mean percentage correct of
the former at 98% (SD = 3.84) and the latter at 73.8% (SD = 23.87). In contrast,
the healthy volunteers who participated in the study had no difficulty producing
sentences with either verb type.
In a recent study using eye tracking, J. Lee & Thompson (2011), tested both
healthy speakers’ and agrammatic aphasic patients’ production of sentences with
either unaccusative or unergative verbs (e.g., The black tube is floating versus
The black dog is barking, respectively) while their eye movements were tracked.
Results showed differential fixation patterns to the subject noun phrase in the
unaccusative condition (in which the subject is a theme) compared to the uner-
gative condition (in which the subject is an agent) for both healthy and aphasic
participants. However, the aphasic speakers showed this effect prior to speech
production, whereas the normal speakers showed the effect during production.
This result suggests that aphasic speakers engage in prolonged planning before
sentence production in the case of unaccusative sentences.
Note, that production of a sentence with an unaccusative verb requires non-
canonical argument mapping (whether considered to be direct mapping of the
theme to a sentence-initial position or the result of syntactic movement, as exem-
plified in Figure 2 above). Production of sentences with noncanonical word or-
der is independently known to be difficult for agrammatic individuals, regardless
of unaccusativity (e.g. Cho-Reyes & Thompson, 2012; Caramazza & Zurif, 1976;
Caplan & Hanna, 1998; Grodzinsky, 1986). For example, these speakers are im-
paired on production of sentences with object relative clauses, such as ‘Pete saw
that girl that the boy was chasing’, in which the theme argument of chase appears
before the verb and its subject (see Figure 2(c)). Thus, based on unaccusative
sentence production tasks alone, it is unclear whether agrammatic individuals
 Neurocognitive mechanisms of verb argument structure processing 153

experience difficulty with unaccusative verbs above and beyond their indepen-
dent difficulty with movement operations or noncanonical argument mapping.
Importantly, however, several studies examining production of unaccusative
and unergative verbs in isolation found unaccusative verb production deficits.
For example, Thompson (2003) examined production of unaccuative and uner-
gative verbs in eight agrammatic and seven unimpaired age-matched speakers in
a constrained action-picture naming task. The aphasic, but not the unimpaired
speakers, showed significantly poorer naming of unaccusatives (M = 60% correct;
SD = 17.7%) than unergatives (M = 92.6% correct, SD = 9.0%). Similar findings
were reported in Kim (2005) and Luzzatti et al. (2002). These results demonstrate
that even when no overt argument mapping or syntactic movement takes place,
unaccusatives are more difficult for agrammatic speakers than unergatives, sug-
gesting that the entailed syntactic complexity associated with unaccusative verbs
is manifest even devoid of an overt syntactic environment.
As mentioned above, the left inferior frontal gyrus, especially Broca’s area, has
been repeatedly shown to be involved in syntactic processing, i.e. phrase structure
building. Moreover, activation in this region tends to co-vary with syntactic com-
plexity, such that complex, and in particular noncanonical sentence structures en-
gender greater activation in this region than simple structures (e.g. Ben-Shachar
et al., 2003; Just, Carpenter, Keller, Eddy, & Thulborn, 1996; Stromswold, Caplan,
Alpert, & Rauch, 1996; Constable et al., 2004; see Friederici, 2011, for review).
Under the assumption that agrammatism often involves at least some damage
to this region, the difficulty of agrammatic speakers with unaccusativity can be
explained as follows: processing of unaccusative verbs, even when they appear in
isolation, entails projection of a noncanonical syntactic structure, thus requiring
left IFG involvement. When this region is damaged, production of these verbs is
impaired.8
In a recent imaging study, Shetreet, Friedmann, & Hadar (2010) compared
brain activations in response to sentences containing unaccusative verbs and sen-
tences containing unergative or transitive verbs in Hebrew. The authors observed
differential activation in the left IFG (BA 45/46) and the left posterior middle
temporal gyrus (MTG, BA 21) in response to sentences containing unaccusatives.
Because the frontal activation detected in the experiment was slightly anterior to
Broca’s area, the authors did not interpret its role as supporting syntactic move-
ment. Rather, they suggested that it reflects the noncanonical semantic/thematic
processing required by sentences with unaccusative verbs, namely the association
of a theme with the subject position. The posterior MTG activation is attributed
by the authors to the operation of a lexical process deriving unaccusative verbs
from their transitive alternates (in Hebrew, this process is accompanied by mor-
phological marking, e.g. kivec ‘shrink.TR’ → hitkavec ‘shrink.UNACC’).9
154 Cynthia K. Thompson and Aya Meltzer-Asscher

To conclude, unaccusativity exemplifies a different dimension of argument

structure complexity, related not to the number of thematic roles, but to their
mapping into syntactic positions. Given their syntactic complexity, it is not sur-
prising that the region most consistently implicated in the processing of unaccu-
sative verbs and sentences is the left inferior frontal gyrus, which is independently
assumed to be involved in complex syntactic processing.

3.3 Argument structure violations

A third way to probe brain structures specifically involved in argument structure

processing has been through experiments including violations of the argument
structure requirements of verbs. Here, findings have been somewhat inconsistent.
In two studies, Kim & Thompson (2000, 2004) report the results of a grammat-
icality judgment task in which agrammatic individuals listened to grammatical
sentences, as well as sentences with argument structure violations, in which ei-
ther an obligatory argument was omitted (e.g. *The woman is giving the sandwich;
*The boy is carrying) or a noun phrase was added following an intransitive verb
(e.g. *The dog is barking the girl). In both studies, agrammatic participants per-
formed at near normal level (means 93.6% correct, 92.1% correct, respectively).
Grodzinsky & Finkel (1998) also tested acceptability judgments of sentences with
argument structure violations in their study of agrammatic individuals’ grammat-
ical abilities, with ungrammatical sentences containing either extra arguments
(*The children sang the football over the fence) or illicitly omitted arguments (*The
children threw). Percent correct responses of the agrammatic participants ranged
from 62%–100%, with a mean of 91% correct, and only one participant showed
a high error rate (> 25%). In contrast to these studies, demonstrating relative-
ly spared detection of argument structure anomalies, in a recent ERP study, we
(Kielar, Meltzer-Asscher, & Thompson, 2012) found a different pattern. Using a
grammaticality judgment task, the mean percent correct responses, for 15 agram-
matic participants, to argument structure violations, such as that exemplified in
(8), was only 59.7% (range: 8%–100%).
(8) *John sneezed the doctor and the nurse

The low accuracy on the argument structure condition in this study may be ex-
plained by the fact that argument structure violations in the experiment invari-
ably occurred at mid-sentence (whereas in previous studies some violations were
sentence final), and were followed by a grammatical five-word noun phrase. Ad-
ditionally, there were 800 msec of silence beginning at sentence offset, only after
which participants were allowed to respond. In contrast, in previous studies par-
ticipants could provide their judgments at any point during the presentation of
 Neurocognitive mechanisms of verb argument structure processing 155

the sentence. The delayed judgment required in our task may have allowed for
interference effects.
With regard to electrophysiological responses associated with argument
structure violations, several studies have been undertaken in healthy and aphasic
participants. In three studies with healthy German speakers (Friederici & Frisch,
2000; Frisch, Hahne, & Friederici, 2004; Friederici & Meyer, 2004), violations of
the correct number of arguments in a sentence (e.g. *The cousin dawdled the vio-
linist) elicited a biphasic N400-P600 pattern. This pattern has been interpreted to
reflect different aspects of lexical-thematic integration process. The N400 results
from difficulty in integration of an argument that violates the thematic informa-
tion carried by the verb, whereas the P600 reflects an attempt at syntactic repair,
following thematic integration failure. Precise localization of the generators of
the N400 and P600 is unclear. However, the N400 is assumed to be generated in
the vicinity of the auditory cortex, whereas the P600 has been localized to the
MTG and the posterior portion of the temporal cortex (Friederici, 2011). Thus,
processing of argument structure violations seems to involve tissue in the middle
and superior temporal gyrus, consistent with neuroimaging studies examining
argument structure processing.
As mentioned above, in a recent study, we investigated ERPs in response
to argument structure violations in young healthy speakers, older controls, and
agrammatic aphasic participants (Kielar et al., 2012). In contrast to the healthy
participants, who showed both an N400 and a P600 effect to these violations,
the agrammatic participants exhibited only an attenuated and restricted P600 ef-
fect, suggesting a loss of fast and automatic online processing of argument struc-
ture information assumed to take place during normal sentence comprehension
(Friedrici, 2002, 2001). Interestingly, although the participants in our study most-
ly exhibited frontal lesions, in many patients the lesions also extended to the tem-
poral lobe.
To summarize this section, at this point it is unclear to what extent judgments
with regard to argument structure are preserved in agrammatism, as well as what
brain regions are involved in the detection of argument structure violations. Ad-
ditional research with healthy as well as impaired participants will undoubtedly
shed light on these issues.

4. A neurocognitive model of verb argument structure processing

The picture emerging from the various experimental findings derived from both
normal and language impaired individuals is that verb argument structure pro-
cessing is subserved by a language network involving both anterior and posterior
156 Cynthia K. Thompson and Aya Meltzer-Asscher

perisylvian regions in the left hemisphere, as well as some focal regions in the
right hemisphere. Specifically, this network involves regions in the inferior pari-
etal lobule bilaterally, in the left posterior temporal gyrus, and in the left inferior
and middle frontal gyri. We discuss the role of each of these regions in more detail
below.

4.1 The role of posterior perisylvian regions

Data derived from both lesion-deficit and neuroimaging studies implicate tem-
poroparietal regions in argument structure processing. Neuropsychological stud-
ies indicate that posterior lesions affect patients’ ability to access verb argument
structure. Conversely, patients with Broca’s aphasia show normal access to ar-
gument structure during on-line sentence processing (Shapiro & Levine, 1990;
Shapiro et al., 1993), and generally do well at detecting argument structure viola-
tions (Kim & Thompson, 2000, 2004). Importantly, persons with Broca’s aphasia
often have lesions that spare the left posterior superior and middle temporal gyri
and surrounding area, including the angular and supramarginal gyri.
Neuroimaging studies also implicate the posterior perisylvian network for
verb argument structure processing. Specifically, two regions have repeatedly
been implicated in these studies: the angular gyrus, and the posterior superior/
middle temporal gyrus (STG/MTG). Although precise anatomical and functional
delineation of these regions is a complex matter, and labeling may vary between
individuals and between studies, the literature suggests that these two regions
might be involved in different processes (see Table 1).
The angular gyrus as well as the supramarginal gyrus, in both the left and
right hemisphere, have been found to show graded activation associated with
the number of thematic roles for both healthy and aphasic individuals (Meltzer-­
Asscher et al., 2013; Thompson et al., 2007, 2010). These results are in line with
many findings showing involvement of the bilateral inferior parietal lobule in
semantic tasks (see review in Binder, Desai, Graves, & Conant, 2009, whose me-
ta-analysis shows the left angular gyrus to be the region most consistently im-
plicated in semantic processing tasks). Because verbs with more thematic roles
denote more complex events, involving more participants, recruitment of inferior
parietal regions, associated with semantic complexity, is predicted.
In contrast to the angular and supramarginal gyri, activations in the left pos-
terior superior and middle temporal gyrus have been found mostly in experi-
ments investigating processing of increased number of arguments in sentence
contexts (e.g. Ben-Shachar et al., 2003). Other studies which examined process-
ing of argument structure in sentences, though not manipulating the number of
arguments, report similar results. For example, Bornkessel, Zyssett, Friederici,
 Neurocognitive mechanisms of verb argument structure processing 157

Table 1. fMRI studies manipulating argument structure complexity in lexical tasks and
sentence-level tasks. Activation peaks for lexical tasks are in the angular gyrus and infe-
rior parietal lobule (bilaterally in two out of the three studies). Activation peaks for tasks
involving sentences are mainly in the left temporal gyrus
Study Contrast Peak Brodmann
coordinates area1
(Talairach)
Lexical contrasts
Thompson et al. Transitive & ditransitive > intransitive verbs –42 –51 302 39
(2007) 50 –63 222 39
Thompson et al. Transitive & ditransitive > intransitive verbs –45 –63 312 39
(2010)
Meltzer-Asscher Alternating transitivity > intransitive verbs –39 –66 462 7
et al. (2013) 50 –58 262 39
Sentential contrasts
Ben Shachar et al. Sentences with 3 arguments > sentences –53 –42 73 22
(2003) with 2 arguments
Bornkessel et al. Sentences requiring noncanonical argument –52 –43 18 13
(2005) mapping > sentences requiring canonical –47 –58 24 39
argument mapping
Shetreet et al. Sentences containing verbs with multiple –58 –50 142 22
(2007) thematic options > sentences containing
verbs with less thematic options
1 Brodmann areas were determined using the Talairach client software (Lancaster et al., 2000).
2 Coordinates originally reported in MNI space.
3 Peak coordinates of ROI selected for analysis.

von Cramon, & Schlesewsky (2005) found activation in the posterior superior
temporal sulcus associated with processing sentences with noncanonical argu-
ment mapping (though one not derived by syntactic movement). Wu, Waller, &
Chatterjee (2007) found that patients with lesions in the lateral temporal cortex
had difficulty assigning thematic roles in simple sentences, using a sentence-pic-
ture matching task. These and other findings suggest that the left posterior STG/
STS is involved in the actual integration of arguments into the syntactic structure
(as proposed by Friederici, 2011).
We suggest, therefore, that activation in the angular and supramarginal gy-
rus is correlated with the semantic density of the verb, whereas activation in the
posterior middle and superior temporal gyrus is correlated with the complexity of
verb-argument integration. It is interesting to ask whether these results shed light
on the nature of argument structure information, i.e. whether it is lexical, syn-
tactic, or derived solely from world knowledge, as discussed in Section 2 above.
158 Cynthia K. Thompson and Aya Meltzer-Asscher

However, since the findings reported here were not derived from experiments
that explicitly set out to differentiate between these theories, we believe that at this
point, they do not definitively favor one theory over others. That is, angular gyrus
activation is predicted by any theory that ascribes more semantic information to
verbs with more complex thematic properties. This is clearly compatible with lex-
icalist approaches, but can also hold for “hybrid” approaches such as Ramchand
(2008). As for activations in the middle and superior temporal gyrus, integration
of a verb with its arguments can be, and most probably is, guided by different
types of information: lexical, syntactic and pragmatic. For example, Kuperberg et
al. (2000) found that verb argument mismatches elicited activation in the left su-
perior temporal gyrus, whether resulting from subcategorization (The young man
slept the guitar), semantic (The young man drank the guitar) or pragmatic (The
young man buried the guitar) violations. Posterior temporal activation associated
with argument structure processing in sentence contexts is thus compatible with
different theories of argument structure representation.

4.2 The role of left frontal regions

Both inferior and more dorsal parts of the left frontal lobe have been implicated
in studies examining argument structure. Middle-superior frontal regions are ac-
tivated in response to verbs encoding more than one argument structure option
(Meltzer-Asscher et al., 2013; Shetreet et al., 2007), and can thus be argued to play
a role in the maintenance of multiple lexical options.
With regard to the inferior frontal gyrus, considering the findings presented
in this chapter, this region seems to subserve at least two distinct processes. First,
it has a role in the production of verbs with increased number of thematic roles.
This is shown by the fact that patients with inferior frontal lesions often show a
verb argument structure production hierarchical deficit, exhibiting more difficul-
ty producing verbs with more thematic roles than those with simpler argument
structure (Kim & Thompson, 2000, 2004). Likewise, neuroimaging data show that
when normal speakers produce transitive verbs, more tissue in the inferior frontal
region is engaged as compared to when intransitive action verbs are produced
(den Ouden et al., 2009). As mentioned above, numerous studies also show in-
volvement of the left IFG in syntactic structure building (Friederici, 2011). The
increased activation in this region associated with production of verbs with more
thematic roles may, therefore, be attributed to the more complex phrase structure
projected by these verbs, even when produced in isolation.
The inferior frontal gyrus also has been argued to be involved not only in
basic phrase structure building, but also in more complex syntactic processes.
 Neurocognitive mechanisms of verb argument structure processing 159

This is evidenced by the fact that greater activation in this region has been found
for sentences with increasing syntactic complexity, in particular noncanonical
sentence structures (Ben Shachar et al., 2003; Just et al., 1996; Stromswold et al.,
1996, among others). It is thus not surprising that this region has also been im-
plicated in studies investigating the production and processing of unaccusative
verbs, which entail noncanonical syntax, as discussed above.10 Recent research
also suggests that different subregions of the left inferior frontal gyrus are active
in basic structure building and in the processing of noncanonical structures, with
the frontal operculum involved in the former and the pars opercularis in the latter
(Friederici, 2011; Friederici et al., 2003).
In summary, the model we propose for argument structure processing is as
follows. In sentence production, speakers first access verb lemmas with grammat-
ically relevant semantic properties, including argument structure information.
This process is supported by inferior parietal regions in both hemispheres, sup-
porting also other aspects of lexical semantic complexity. Once the lemma with

Argument
Initial phrase
structure
Structure
Information
Building
retrieval

Verb/Argument
Integration

Angular gyrus / Supramarginal gyrus

Inferior frontal gyrus
Posterior middle / superior temporal gyrus

Figure 3. Neurocognitive model of verb argument structure processing in sentence pro-

duction. The angular and supramarginal gyri in both hemispheres are involved in lemma
access and retrieval of argument structure information; this information is used to gen-
erate initial phrase structure building processes in the left IFG. Sentence level syntactic
and semantic integration engage the left posterior MTG/STG
160 Cynthia K. Thompson and Aya Meltzer-Asscher

Argument
Initial phrase
structure
Structure
Information
Building
retrieval

Verb/Argument
Integration

Angular gyrus / Supramarginal gyrus

Inferior frontal gyrus
Posterior middle / superior temporal gyrus

Figure 4. Neurocognitive model of verb argument structure processing in sentence

comprehension. Initial syntactic parsing and structure building involves the left IFG.
Once verbs are activated, the bilateral angular/supramarginal gyri are engaged to support
retrieval of associated argument structure information. This information, along with the
initial structure, is transmitted to left temporal regions for sentence-level semantic and
syntactic integration

its argument structure information is accessed, this information triggers initial

phrase structure building operations in the IFG. In turn, left posterior regions,
namely the STG and MTG, are required for sentence production, as they support
integration of the verb with its arguments (Figure 3).
In sentence comprehension, processing begins with initial syntactic parsing
and phrase structure building in the IFG. Argument structure information as-
sociated with the verbs in the sentence is then retrieved, a process supported by
the angular and supramarginal gyri. The basic phrase structure is then fed, along
with the argument structure information, to posterior temporal regions, for in-
tegration. If a syntactic structure is built which was incompatible with the verb’s
argument structure requirements, repair processes will take place (Figure 4).
The left posterior superior temporal gyrus and inferior frontal regions, thus,
form a network for building clausal syntactic structures and integrating lexical
material, establishing syntactic and semantic relations between a verb and its
 Neurocognitive mechanisms of verb argument structure processing 161

arguments. Studies show that these two regions are connected by a dorsal path-
way consisting of the arcuate fasciculus and the superior longitudinal fascicle
(Brauer, Anwander, & Friederici, 2011). Interestingly, an additional, ventral path-
way connects the anterior IFG (BAs 45 and 47) with the anterior temporal lobe
via the uncinate fasciculus. This network is argued by Friederici (2011) to support
local structure building and semantic combinatorics. Despite the importance of
anterior temporal regions for sentence-level syntactic and semantic processing
(see review also in Hickok & Poeppel, 2007), experiments that explicitly manipu-
late argument structure processing demands have not elicited activation in these
regions. The role of the ventral pathway between frontal and temporal regions in
supporting argument integration is thus still unclear.

Notes

* Research supported by the National Institutes of Health, R01DC01948 and RO1DC007213.

1. It could be argued that if world knowledge successfully constrains the argument structure
of verbs, and if world knowledge relevant to verbs is stored and accessed when verbal entries
are retrieved, the extant psycholiguistic (and neurolinguistic) findings are consistent with con-
structivism. On such an account world knowledge would be retrieved together with verbs (even
for verbs appearing in isolation), thus limiting its argument structure options. We note, howev-
er, that for this mechanism to work, namely to account for the psycholinguistic findings, verbs
with more argument structure options would have to be associated with greater (stored) world
knowledge information (since verbs with complex argument structures are slower to process
and require more neural resources). This, however, is not the common view of proponents of
the constructivist approach, who view the building of a verb phrase as a free operation con-
strained by world knowledge, i.e., verbs with less argument structure options are actually as-
sumed in these approaches to be associated with more constraints. In contrast, in the lexicalist
approach, verbs with less complex argument structures are indeed stored with less information.
In addition, it is not clear to what extent world knowledge sufficiently substitutes for the lexical
argument structure properties of verbs (see also Ramchand, 2008). For example, in English, fall
cannot be used transitively: ‘*John fell the cup’. It is difficult to explain this phenomenon based
on world knowledge, i.e. to find a language-external reason why the sentence cannot mean
‘John caused the cup to fall’, especially given the fact that fall can be used transitively in other
languages, e.g. Hebrew.

2. Another verb type potentially relevant to complexity contrasts is psych- (or experiencer)
verbs, which denote psychological states or events (see Thompson & Lee, 2009). However, be-
cause of the theoretical challenges posed by this verb class (see e.g. Arad, 1998; Pesetsky, 1995;
Reinhart, 2002) psych verb processing is not discussed in this chapter.

3. In contrast, these studies show that Wernicke’s aphasic patients with primary damage to
posterior, rather than anterior, language regions do not show differential RTs to words follow-
ing verbs with different complexity on this factor, indicating a lack of sensitivity to subcatego-
rization information associated with verbs.
162 Cynthia K. Thompson and Aya Meltzer-Asscher

4. The authors of this study do not report whether the sentences produced with one argument
were correct sentences (i.e. containing intransitive verbs), or incorrect sentences (i.e. contain-
ing transitive or ditransitive verbs, with argument omission).

5. Palti, Ben-Shachar, Hendler, & Hadar (2007) found a similar activation when contrasting
the processing of verbs to that of nouns. The authors hypothesized that this activation might be
due to the fact that verbs have thematic roles, whereas (concrete) nouns do not.
One exception to the involvement of posterior perisylvian regions in the processing of mul-
tiple thematic roles was reported by Shetreet, Palti, Friedmann, & Hadar (2007). Examining
verb processing in sentence contexts, the authors found that the medial precuneus and the
anterior cingulate, not considered to be ‘traditional’ language regions, were sensitive to the
number of arguments in a sentence.

6. More generally, these results are consistent with the commonly held assumption that dor-
solateral prefrontal regions in the two hemispheres play a role in general working memory and
maintenance processes (see D’esposito, 2001; Curtis & D’esposito, 2003 for reviews).

7. An exception to this generalization is offered by the group of so-called ‘Theme-unergatives’

(Reinhart, 2002), which consists of verbs that map their sole thematic role externally, although
this role is not agentive (see Levin & Rappaport-Hovav, 1995).

8. In general comprehension of unaccusative sentences is preserved in agrammatism (Lee &

Thompson, 2004). This is not surprising given that an unaccusative sentence includes one ac-
tion (denoted by the unaccusative verb) and one participant (the argument noun phrase), and
can thus be understood without syntactic computation.

9. It is debated whether this derivation process indeed takes place upon each encounter with
an unaccusative verb (see Horvath & Siloni, 2009 for evidence that it does not). Thus, the role
of the posterior MTG in unaccusative sentence processing is not completely clear.

10. Several studies have suggested that the involvement of Broca’s area in the processing of
complex sentences is related to working memory demands required for maintaining a dis-
placed element over a prolonged distance until it can be successfully integrated with the sen-
tence. Whether the process involved is a syntactic-specific form of working memory or an
ordinary, domain-general form of working memory (as in Baddeley, 1992), is the focus of an
on-going debate (Caplan, Alpert, Waters, & Olivieri, 2000; Rogalsky, Matchin, & Hickok, 2008;
Santi & Grodzinsky, 2007; see Thompson & Kielar, in press, for review).

References

Arad, M. (1998). Psych-notes. UCL Working Papers in Linguistics 10 (pp. 203–223), University
College London.
Baddeley, A. (1992). Working memory. Science, 31, 556–559. DOI: 10.1126/science.1736359
Bastiaanse, R., & Jonkers, R. (1998). Verb retrieval in action naming and spontaneous speech in
agrammatic and anomic aphasia. Aphasiology, 12, 951–969.
DOI: 10.1080/02687039808249463
 Neurocognitive mechanisms of verb argument structure processing 163

Bastiaanse, R., & van Zonneveld, R. (2005). Sentence production with verbs of alternating tran-
sitivity in agrammatic Broca’s aphasia. Journal of Neurolinguistics, 18(1), 57–66.
DOI: 10.1016/j.jneuroling.2004.11.006
Ben-Shachar, M., Hendler, T., Kahn, I., Ben-Bashat, D., & Grodzinsky, Y. (2003). The neural
reality of syntactic transformation – evidence from fMRI. Psychological Science, 14(5),
433–440. DOI: 10.1111/1467-9280.01459
Binder, J.R., Desai, R. H., Graves, W. W., & Conant, L. L. (2009). Where is the semantic system?
A critical review and meta-analysis of 120 functional neuroimaging studies. Cerebral Cor-
tex, 19, 2767–2796. DOI: 10.1093/cercor/bhp055
Bock, K. (1995). Sentence production: From mind to mouth. In J. L. Miller, & P. D. Eimas
(Eds.), Handbook of Perception and Cognition. Vol 11: Speech, Language, and Communica-
tion (pp. 181–216). Orlando, FL: Academic Press.
Bock, K., & Levelt, W. J. M. (2002). Language production: Grammatical encoding. In G. T. M.
Altmann (Ed.), Psycholinguistics: Critical Concept in Psychology (pp. 405–452). London:
Routledge.
Boland, J. E. (2005). Visual arguments. Cognition, 95, 237–274. DOI: 10.1016/j.cognition.
2004.01.008
Boland, J. E., Tanenhaus, M. K., & Garnsey, S. M. (1990). Evidence for the immediate use of
verb control information in sentence processing. Journal of Memory and Language, 29,
413–432. DOI: 10.1016/0749-596X(90)90064-7
Borer, H. (2005). The Normal Course of Events. Structuring Sense, Volume II. Oxford: Oxford
University Press. DOI: 10.1093/acprof:oso/9780199263929.001.0001
Bornkessel, I., Zysset, S., von Cramon, D. Y., Friederici, A. D. & Schlesewsky, M. (2005). Who
did what to whom? The neural basis of argument hierarchies during language comprehen-
sion. Neuroimage, 26, 221–233. DOI: 10.1016/j.neuroimage.2005.01.032
Brauer, J., Anwander, A., & Friederici, A. D. (2011). Neuroanatomical prerequisites for lan-
guage functions in the maturing brain. Cerebral Cortex, 21, 459–466.
DOI: 10.1093/cercor/bhq108
Burzio, L. (1986). Italian Syntax. Dordrecht, The Netherlands: Reidel.
DOI: 10.1007/978-94-009-4522-7
Caplan, D., Alpert, N., Waters, G., & Olivieri, A. (2000). Activation of Broca’s area by syntactic
processing under conditions of concurrent articulation. Human Brain Mapping, 9, 65–71.
DOI: 10.1002/(SICI)1097-0193(200002)9:2<65::AID-HBM1>3.0.CO;2-4
Caplan, D. & Hanna, J. E. (1998). Sentence production by aphasic patients in a constrained task.
Brain and Language, 63, 184–218. DOI: 10.1006/brln.1998.1930
Caramazza, A. & Zurif, E. B. (1976). Dissociation of algorithmic and heuristic processes in
language comprehension: Evidence from Aphasia. Brain and Language, 3, 572–582.
DOI: 10.1016/0093-934X(76)90048-1
Chan, A. H. D., Liu, H., Yip, V., Fox, P. T., Gao, J., & Tan, L. H. (2004). Neural systems for word
meaning modulated by semantic ambiguity. Neuroimage, 22, 1128–1133.
DOI: 10.1016/j.neuroimage.2004.02.034
Chang, F., Dell, G. S., & Bock, K. (2006). Becoming syntactic. Psychological review, 113, 234–
272. DOI: 10.1037/0033-295X.113.2.234
Cho-Reyes, S. & Thompson, C. K. (2012). Verb and sentence production and comprehension:
Northwestern Assessment of Verbs and Sentences (NAVS). Aphasiology, 26, 1250–1277.
Chomsky, N. (1981). Lectures on government and binding: The Pisa lectures. Dordrecht, The
Netherlands: Foris.
164 Cynthia K. Thompson and Aya Meltzer-Asscher

Constable, R. T., Pugh, K. P., Berroya, E., Menci, W. E., Westerveld, M., Ni, W. et al. (2004). Sen-
tence complexity and input modality effects in sentence comprehension: an fMRI study.
Neuroimage, 22, 11–21. DOI: 10.1016/j.neuroimage.2004.01.001
Curtis, C. E. & D’esposito, M. (2003). Persistent activity in the prefrontal cortex during working
memory. Trends in Cognitive Sciences, 7, 415–423. DOI: 10.1016/S1364-6613(03)00197-9
De Bleser, R., & Kauschke, C. (2003). Acquisition and loss of nouns and verbs: parallel or diver-
gent patterns? Journal of Neurolinguistics, 16(2–3), 213–229.
DOI: 10.1016/S0911-6044(02)00015-5
Dell, G. S. & Sullivan, J. M. (2004). Speech errors and language production: Neuropsychological
and connectionist perspectives. In B. H. Ross (Ed.), The Psychology of Learning and Moti-
vation (pp. 63–108). San Diego: Elsevier.
den Ouden, D., Fix, S., Parrish, T., & Thompson, C. (2009). Argument structure effects in action
verb naming in static and dynamic conditions. Journal of Neurolinguistics, 22, 196–215.
DOI: 10.1016/j.jneuroling.2008.10.004
D’Esposito, M. (2001). Functional neuroimaging of working memory. In R. Cabeza &
A. Kingstone (Eds.), Handbook of Functional Neuroimaging of Cognition (pp. 293–328).
Cambridge, MA: MIT Press.
Dini, L. (1995). Unaccusative behaviors. Quaderni del Laboratorio di Linguistica, 7, 92–122.
Dowty, D. R. (1991). Thematic proto-roles and argument selection. Language, 67, 547–619.
Dragoy, O., & Bastiaanse, R. (2010). Verb production and word order in Russian agrammatic
speakers. Aphasiology, 24, 28–55. DOI: 10.1080/02687030802586902
Dronkers, N. F., Wilkins, D. P., Van Valin Jr., R. D., Redfern, B. B., & Jaeger, J. J. (2004). Lesion
analysis of the brain areas involved in language comprehension. Cognition, 92, 145–177.
DOI: 10.1016/j.cognition.2003.11.002
Embick, D. (2004). Unaccusative syntax and verbal alternations. In A. Alexiadou,
E. Anagnostopoulou and M. Everaert (Eds.), The Unaccusativity Puzzle: Explorations of the
Syntax-Lexicon Interface (pp. 137–158). Oxford: Oxford University Press.
DOI: 10.1093/acprof:oso/9780199257652.003.0006
Ferreira, V. S. & Slevc, L. R. (2007). Grammatical encoding. In M. Gareth Gaskell (Ed.), The
Oxford Handbook of Psycholinguistics (pp. 453–469). Oxford, England: Oxford University
Press.
Ferretti, T. R., McRae, K., & Hatherell, A. (2001). Integrating verbs, situation schemas, and
thematic role concepts. Journal of Memory and Language, 44(4), 516–547.
DOI: 10.1006/jmla.2000.2728
Friederici, A. D. (2002). Towards a neural basis of auditory sentence processing. Trends in Cog-
nitive Science, 6, 78–84. DOI: 10.1016/S1364-6613(00)01839-8
Friederici, A. D. (2011). The brain basis of language processing: From structure to function.
Physiological Reviews, 91, 1357–1392. DOI: 10.1152/physrev.00006.2011
Friederici, A. D. & Frisch, S. (2000). Verb argument structure processing: The role of verb-spe-
cific and argument-specific information. Journal of Memory and Language, 43, 476–507.
DOI: 10.1006/jmla.2000.2709
Friederici, A. D. & Meyer, M. (2004). The brain knows the difference: two types of grammatical
violations. Brain Research, 1000, 72–77. DOI: 10.1016/j.brainres.2003.10.057
Friederici, A. D., Rüschemeyer, S. A., Hahne, A., & Fiebach, C. J. (2003). The role of left inferior
frontal and superior temporal cortex in sentence comprehension: localizing syntactic and
semantic processes. Cerebral Cortex, 13, 170–177. DOI: 10.1093/cercor/13.2.170
 Neurocognitive mechanisms of verb argument structure processing 165

Friedmann, N., Taranto, G., Shapiro, L. P., & Swinney, D. (2008). The leaf fell (the leaf): The
online processing of unaccusatives. Linguistic Inquiry, 39, 355–377.
DOI: 10.1162/ling.2008.39.3.355
Frisch, S., Hahne, A., & Friederici, A. D. (2004). Word category and verb-argument structure
information in the dynamics of parsing. Cognition, 91, 191–219.
DOI: 10.1016/j.cognition.2003.09.009
Gorrell, P. (1991). Subcategorization and sentence processing. In R. C. Berwick, S. P. Abney,
& C. Tenny (Eds.), Principle based parsing (pp. 279–300). Dordrecht, The Netherlands:
Kluwer Academic Press. DOI: 10.1007/978-94-011-3474-3_11
Grewe, T., Bornkessel, I., Zysset, S., Wiese, R., von Cramon, D. Y., & Schlesewsky, M. (2005).
The emergence of the unmarked: A new perspective on the language-specific function of
Broca’s area. Human Brain Mapping, 26, 178–190. DOI: 10.1002/hbm.20154
Grodzisnky, Y. (1986). Language deficits and the theory of syntax. Brain and Language, 27,
135–159. DOI: 10.1016/0093-934X(86)90009-X
Grodzinsky, Y. & Finkel, L. (1998). The neurology of empty categories. Journal of Cognitive
Neuroscience, 10, 281–292.
Hickok, G. & Poeppel, D. (2007). The cortical organization of speech processing. Nature Re-
views Neuroscience, 8, 393–402. DOI: 10.1038/nrn2113
Horvath, J. & Siloni, T. (2009). Hebrew idioms: The organization of the lexical component.
Brill’s Annual of Afroasiatic Languages and Linguistics, 1, 283–310.
DOI: 10.1163/187666309X12491131130666
Horvath, J. & Siloni, T. (2011). Causatives across components. Natural Language and Linguistic
Theory, 29, 657–704. DOI: 10.1007/s11049-011-9135-3
Jackendoff, R. (1972). Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT
Press.
Just, M. A., Carpenter, P. A., Keller, T. A., Eddy, W. F., & Thulborn, K. R. (1996). Brain activation
modulated by sentence comprehension. Science, 274, 114–116.
DOI: 10.1126/science.274.5284.114
Kegl, J. (1995). Levels of representation and units of access relevant to agrammatism. Brain and
Language, 50(2), 151–200. DOI: 10.1006/brln.1995.1044
Kemmerer, D., & Tranel, D. (2000). Verb retrieval in brain-damaged subjects: 1. Analysis of
stimulus, lexical, and conceptual factors. Brain and Language, 73, 347–392.
DOI: 10.1006/brln.2000.2311
Kertesz, A. (1977). Isotope localization of infarcts in aphasia. Archives of Neurology, 34, 590–
601. DOI: 10.1001/archneur.1977.00500220024004
Kertesz, A., Harlock, W., & Coates, R. (1979). Computer tomographic localization, lesion size,
and prognosis in aphasia and nonverbal impairment. Brain and Language, 8, 34–50.
DOI: 10.1016/0093-934X(79)90038-5
Kielar, A., Meltzer-Asscher, A., & Thompson, C. K. (2012). Electrophysiological responses to
argument structure violations in healthy adults and individuals with agrammatic aphasia.
Neuropsychologia, 50, 3320–3337.
Kim, K.-Y. (2005). Verb production and argument structures in aphasics. MA thesis, Yonsei
University; Seoul, Korea. (in Korean).
Kim, M., & Thompson, C. K. (2000). Patterns of comprehension and production of nouns and
verbs in agrammatism: implications for lexical organization. Brain and Language, 74, 1–
25. DOI: 10.1006/brln.2000.2315
166 Cynthia K. Thompson and Aya Meltzer-Asscher

Kim, M., & Thompson, C. K. (2004). Verb deficits in Alzheimer’s disease and agrammatism:
implications for lexical organization. Brain and Language, 88, 1–20.
DOI: 10.1016/S0093-934X(03)00147-0
Kiss, K. (2000). Effects of verb complexity on agrammatic aphasics’ sentence production. In
R. Bastiaanse & Y. Grodzinsky (Eds.), Grammatical disorders in aphasia. London: Whurr
Publishers.
Kuperberg, G. R., McGuire, P. K., Bullmore, E. T., Brammer, M. J., Rabe-Hesketh, S., Wright,
I. C. et al. (2000). Common and distinct neural substrates for pragmatic, semantic and syn-
tactic processing of spoken sentences: An fMRI study. Journal of Cognitive Neuroscience,
12, 321–341. DOI: 10.1162/089892900562138
Lancaster, J. L., Woldorff, M. G., Parsons, L. M., Liotti, M., Freitas, C. S., Rainey, L., et al. (2000).
Automated Talairach Atlas labels for functional brain mapping. Human Brain Mapping,
10, 120–131. DOI: 10.1002/1097-0193(200007)10:3<120::AID-HBM30>3.0.CO;2-8
Lee, J., & Thompson, C. K. (2011). Real-time production of unergative and unaccusative sen-
tences in normal and agrammatic speakers: An eyetracking study. Aphasiology.
DOI: 10.1080/02687038.2010.542563.
Lee, M., & Thompson, C. K. (2004). Agrammatic aphasic production and comprehension of
unaccusative verbs in sentence contexts. Journal of Neurolinguistics, 17(4), 315–330.
DOI: 10.1016/S0911-6044(03)00062-9
Levelt, W. J. M. (1999). Producing spoken language: A blueprint of the speaker. In C. Brown
& P. Hagoort (Eds.), The neurocognition of language. New York: Oxford University Press.
Levin, B. & Rappaport, M. (1986). The formation of adjectival passives. Linguistic Inquiry, 17,
623–661.
Levin, B. & Rappaport Hovav, M. (1995). Unaccusativity: At the Syntax-Lexical Semantics Inter-
face. Cambridge, MA: MIT Press.
Luzzatti, C., Raggi, R., Zonca, G., Pistarini, C., Contardi, A., & Pinna, G. D. (2002). Verb-noun
double dissociation in aphasic lexical impairments: the role of word frequency and image-
ability. Brain and Language, 81(1–3), 432–444. DOI: 10.1006/brln.2001.2536
MacDonald, M. C., Perlmutter, N. J., & Seidenberg, M. S. (1994). The lexical nature of syntactic
ambiguity resolution. Psychological Review, 101(4), 676–703.
DOI: 10.1037/0033-295X.101.4.676
Marantz, A. (1997). No escape from syntax: Don’t try a morphological analysis in the privacy
of your own lexicon. University of Pennsylvania Working Papers in Linguistics, 4, 201–225.
Mason, R., & Just, M. (2007). Lexical ambiguity in sentence comprehension. Brain Research,
1146, 115–127. DOI: 10.1016/j.brainres.2007.02.076
Meltzer-Asscher, A., Schuchard, J., den Ouden, D. B., & Thompson, C. K. (2013). The neural
substrates of complex argument structure representations: Processing of “alternating tran-
sitivity” verbs. Language and Cognitive Processes, 28, 1154–1168.
Meyer, A. (1996). Lexical access in phrase and sentence production: Results from picture-word
inference experiments. Journal of Memory and Language, 35, 477–496.
DOI: 10.1006/jmla.1996.0026
Palti, D., Ben-Shachar, M., Hendler, T., & Hadar, U. (2007). The cortical correlates of grammat-
ical category differences: An fMRI study of nouns and verbs. Human Brain Mapping, 28,
303–314. DOI: 10.1002/hbm.20280
Perlmutter, D. (1978). Impersonal passives and the unaccusative hypotheses. Paper presented at
the Fourth Annual Meeting of the Berkeley Linguistics Society, 1978.
Pesetsky, D. (1995). Zero syntax: Experiencers and cascades. Cambridge, MA: MIT Press.
 Neurocognitive mechanisms of verb argument structure processing 167

Pollard, C., & Sag, I. (1994). Head-Driven Phrase Structure Grammar. Chicago: University of
Chicago Press.
Pylkkänen, L. (2008). Introducing Arguments. Cambridge, MA: MIT Press.
DOI: 10.7551/mitpress/9780262162548.001.0001
Ramchand, G. (2008). Verb Meaning and the Lexicon: A First Phase Syntax. Cambridge:
Cambridge University Press. DOI: 10.1017/CBO9780511486319
Reinhart, T. (2002). The Theta System – an overview. Theoretical Linguistics, 28, 229–290.
Rogalsky, C., Matchin, W., & Hickok, G. (2008). Broca’s area, sentence comprehension, and
working memory: An fMRI study. Frontiers in Human Neuroscience, 2, 14.
DOI: 10.3389/neuro.09.014.2008
Santi, A. & Grodzinsky, Y. (2007). Working memory and syntax interact in Broca’s area. Neuro-
image, 37, 8–17. DOI: 10.1016/j.neuroimage.2007.04.047
Shapiro, L.P., Brookins, B., Gordon, B., & Nagel, N. (1991). Verb effects during sentence pro-
cessing. Journal of Experimental Psychology, 17(5), 983–996.
Shapiro, L.P., Gordon, B., Hack, N., & Killackey, J. (1993). Verb-argument structure processing
in complex sentences in Broca’s and Wernicke’s aphasia. Brain and Language, 45, 423–447.
DOI: 10.1006/brln.1993.1053
Shapiro, L. P., & Levine, B. A. (1990). Verb processing during sentence comprehension in apha-
sia. Brain and Language, 38(1), 21–47. DOI: 10.1016/0093-934X(90)90100-U
Shapiro, L. P., Zurif, E. B., & Grimshaw, J. (1987). Sentence processing and the mental represen-
tations of verbs. Cognition, 27, 219–246. DOI: 10.1016/S0010-0277(87)80010-0
Shetreet, E., Friedmann, N., & Hadar, U. (2010). The neural correlates of linguistic distinctions:
Unaccusative and unergative verbs. Journal of Cognitive Neuroscience, 22, 2306–2315.
DOI: 10.1162/jocn.2009.21371
Shetreet, E., Palti, D., Friedmann, N., & Hadar, U. (2007). Cortical representation of verb pro-
cessing in sentence comprehension: Number of complements, subcategorization, and the-
matic frames. Cerebral Cortex, 17(8), 1958–1969. DOI: 10.1093/cercor/bhl105
Stemberger, J. (1985). An interactive activation model of language production. In A. Ellis (Ed.),
Progress in the Psychology of Language. London: Erlbaum.
Stromswold, K., Caplan, D., Alpert, N., & Rauch, S. (1996). Localization of syntactic compre-
hension by positron emission tomography. Brain and Language, 52, 452–473.
DOI: 10.1006/brln.1996.0024
Thompson, C. K. (2003). Unaccusative verb production in agrammatic aphasia: A syntactic
account of verb production deficits. Journal of Neurolinguistics, 16, 151–167.
DOI: 10.1016/S0911-6044(02)00014-3
Thompson, C. K., Bonakdarpour, B., & Fix, S. F. (2010). Neural mechanisms of verb argument
structure processing in agrammatic aphasic and healthy age-matched listeners. Journal of
Cognitive Neuroscience.
Thompson, C. K., Bonakdarpour, B., Fix, S. C., Blumenfeld, H. K., Parrish, T. B., Gitelman,
D. R., & Mesulam, M. M. (2007). Neural correlates of verb argument structure processing.
Journal of Cognitive Neuroscience, 19(11), 1753–1767. DOI: 10.1162/jocn.2007.19.11.1753
Thompson, C. K., & Kielar, A. (in press). Neural bases of sentence processing: Evidence from
neurolinguistic and neuroimaging studies. In Ferreria, V., Goldrick, M., & Miozzo, M.
(Eds), The Oxford Handbook of Language Production. Oxford: Oxford University Press.
DOI: 10.1093/oxfordhb/9780199735471.013.020
168 Cynthia K. Thompson and Aya Meltzer-Asscher

Thompson, C. K., Lange, K. L., Schneider, S. L., & Shapiro, L. P. (1997). Agrammatic and non-
brain damaged subjects’ verb and verb argument structure production. Aphasiology, 11,
473–490. DOI: 10.1080/02687039708248485
Thompson, C. K., & Lee, M. (2009). Psych verb production and comprehension in agrammatic
Broca’s aphasia. Journal of Neurolinguistics, 22, 354–369.
DOI: 10.1016/j.jneuroling.2008.11.003
Thompson, C. K., Lukic, S., King, M., Mesulam, M.-M., & Weintraub, S. (2012). Verb and noun
deficits in stroke-induced and primary progressive aphasia: The Northwestern Naming
Battery. Aphasiology, 26, 632–655.
Trueswell, J. C., & Kim, A. E. (1998). How to prune a garden path by nipping it in the bud: Fast
priming of verb argument structure. Journal of Memory and Language, 39, 102–123.
DOI: 10.1006/jmla.1998.2565
Trueswell, J. C., Tanenhaus, M. K., & Kello, C. (1993). Verb-specific constraints in sentence pro-
cessing: Separating effects of lexical preference from garden-paths. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 19(3), 528–553.
DOI: 10.1037/0278-7393.19.3.528
Van Valin, R. (1990). Semantic parameters on split intransitivity. Language, 66, 221–260.
DOI: 10.2307/414886
Williams, E. (1981). Argument structure and morphology. The Linguistic Review, 1, 81–114.
DOI: 10.1515/tlir.1981.1.1.81
Wu, D., Waller, S., & Chatterjee, A. (2007). The functional neuroanatomy of thematic role and
locative relational knowledge. Journal of Cognitive Neuroscience, 19, 1542–1555.
DOI: 10.1162/jocn.2007.19.9.1542
 Argument structure
Between linguistics and neuroimaging*

Einat Shetreet

“A verb is like the sun … and a noun is like a planet” (Chafe 1970). This metaphor
describes the critical role of verbs in sentence processing. Within a sentence, the
verb is the central element around which nouns cluster with their own relation
to the event described by the verb. Thus, verbs have been the subject of many
theoretical and experimental investigations. Linguists have defined several the-
oretical concepts related to verbs, based on the assumption that verbs carry lex-
ical information that determines the sentential environment in which they can
appear (e.g., Chomsky 1965; Grimshaw 1979; van Valin 2001). However, despite
the extensive description of verbs within the linguistic literature, many questions
have been left unanswered: the types of information that are actually represented
within the lexical entry of a verb, the form of representation of this information,
and even whether or not any information is stored at all.
Experimental investigations may shed light on some of these questions. To
study the processing of verbs, psycholinguists use various types of offline and on-
line methods, including judgment tasks, self-paced-reading, eye-movement par-
adigms or priming studies. Neuroimaging methods, such as fMRI, EEG or MEG,
are also used. fMRI allows online detection of brain activations with a fairly good
spatial resolution, and is used to explore the neural architecture of language by
mapping linguistic functions to specific brain regions. Recently, criticisms have
been made regarding the contribution of neuroimaging, and specifically fMRI,
to the study of brain and language, arguing that it has not made substantial con-
tributions to cognitive theory (linguistics e.g., Coltheart 2006), or neuroscience
(e.g., Poeppel and Embick 2005). Although some of these criticisms are justified,
I will argue in this paper that the problem is not inherent to neuroimaging meth-
ods nor that it is due to an intellectual barrier between linguistics and cognitive
neuroscience, but rather it is due to the specific questions that have been asked.
In order to maximize the advantages of fMRI in linguistic research, linguistical-
ly-sophisticated and fine-grained questions should be asked. Questions of this
170 Einat Shetreet

kind can potentially bear substantial contribution to our understanding of both

language and the brain.

Defining a question

To ask fine-grained questions, one should consult the linguistic literature. This
literature demonstrates the central role of verbs in sentence processing, as well
as their linguistic complexity. Various linguistic theories break down this com-
plexity into examinable components. These theories describe different types of
verbs (such as transitive, unaccusative, or psych verbs) and define various types
of information associated with verbs, on both syntactic and semantic dimensions
(such as thematic roles, subcategorization or selectional restrictions). Linguistic
theories also spell out the alternatives for the formation of the sentential envi-
ronment in which the same verb can appear (e.g., alternation of verbs like break
between transitive and unaccusative forms, or alternation of verbs like eat be-
tween transitive and intransitive forms). These well-defined concepts should lay
the ground for any experimental investigation, including neuroimaging.
Many questions have already been specified within the linguistic literature.
Is any lexical information stored with the verb at all? Are unaccusatives derived
from their transitive counterparts? Is syntactic subcategorization redundant in
light of semantic selection? Importantly, the linguistic literature not only pos-
es the questions, but often also provides the means to explore the answers. Lin-
guists frequently offer competing theories to the same question. These theories
may be contrasted by examining the predictions they make regarding differences
and similarities between verbs. If the contrast between the predictions of two (or
more) linguistic theories can be translated into a contrast between their predic-
tions regarding brain activation, neuroimaging can prove a useful tool to help
decide between the theories.
A simple example concerns the question of whether any information is stored
in the lexicon. Certain linguists argue that the arguments of the verb are struc-
turally assigned so that no or limited grammatical information is lexically repre-
sented with the verb (e.g., Borer 2004; van Hout 1992), whereas others assume
that the different sentential environments of the verb are determined by informa-
tion stored in the verb’s lexical entry (e.g., Chierchia 2004; Reinhart 2002). The
structural theory (Borer 2004; van Hout 1992) predicts that verbs that can appear
in multiple sentence types and verbs that can appear in limited sentence types
should be similar with regard to their lexical “weight”. By contrast, the lexical
account (Chierchia 2004; Reinhart 2002) predicts that the two verb classes should
 Between linguistics and neuroimaging 171

differ, because verbs that can appear in multiple sentence types have a richer (pol-
ysemous) argument structure in their lexical entry. Therefore, a comparison be-
tween verbs that can appear in limited sentential environments and those that can
appear in multiple sentential environments can inform this debate. Similar brain
patterns for both verb classes will support the structural approach, whereas in-
creased activations for the rich-argument-structure verbs will support the lexical
approach (for some experimental results pertaining to this questions see: Shetreet
et al. 2007; Thompson et al. 2007; for experiments comparing competing theories
on optional arguments and on reflexive derivation see Shetreet et al. 2010b and
Shetreet and Friedmann 2012 respectively).
Some less straight-forward examinations can also inform this debate. In
Shetreet et al. (2009), we asked whether lexical complexity, represented by the
syntactic type of the verbal arguments, affects brain activations. To test this, we
compared verbs that can appear with syntactically complex arguments (i.e., CP
argument) and verbs that can appear with simpler arguments (i.e., NP argument).
Importantly, we used the same syntactic structure (with a PP argument) for both
verb types. For example, both the verb complain that can take a CP argument and
the verb nibble that can take an NP argument appeared in sentences with a PP
argument (as in John complained about the cold soup and John nibbled at the big
cake). We found that even when appearing in the same sentential environment,
the verbs had differential brain patterns, showing increased activations for the
syntactically-complex verbs. This result is significant for the theoretical debate be-
tween the structural and the lexical approaches. Notably, it has implications also
for the psycholinguistic literature as it provides insights for models of language
processing by pointing out which components of verb representation should be
incorporated into the parsing model. Our study indicates that information at the
lexical entry affects the access to a verb even when the information is not included
in the sentence. This suggests that when modeling sentence comprehension, one
should consider some lexical elements that are not explicitly mentioned in the
sentence.
Up until now I have preceded without saying one word about the localization
of the activations. To make inferences that two verb classes are different, it is not
required to show activation in specific brain regions (Henson 2006). Truly, if acti-
vations are observed in brain areas that are consistent with results from previous
neurolinguistic examinations or lesion studies of aphasia of the same linguistic
process, the inferences will be more compelling. However, activations may show
in unexpected brain areas. Such areas may subserve the tested linguistic functions
even though their location seems unconventional, e.g., outside the “canonical lan-
guage system” (see for example Bachrach (2008), Moro et al., (2001), or Shetreet
172 Einat Shetreet

et al., (2009) which suggest that the precuneus, a non-canonical language brain
region, has a role in syntactic processing). Therefore, the most important result
with regard to differences between verb classes is not in certain brain localiza-
tions, but in differential brain patterns, which will presumably suggest different
cognitive processes for each verb class.
Nonetheless, fMRI enables mapping of linguistic functions to brain regions
(which is also the more standard use of this method). It is therefore common in
neuroimaging studies to attribute function to the activated regions. I have argued
here that this should not be the sole goal of neurolinguistic studies and that neu-
roimaging could be used fruitfully to adjudicate between competing linguistic
theories or psycholingusitic models. Furthermore, I will argue (below) that deter-
mining what cognitive or linguistic process is responsible for an observed brain
activation is not in any way trivial. Speculations about the roles of specific brain
regions can be made based on the theoretical differences between the experimen-
tal conditions, as well as based on findings from previous studies that have also
identified the same brain regions. However, these speculations are by no means
conclusive due to some limitations related to the method and the traditional prac-
tice within it (as discussed below). Additionally, neuroimaging results can indi-
cate which brain areas take part in the processing of a specific aspect of language,
but they cannot separate the areas that are necessarily involved in the processing
from the areas that only have a supporting role. To identify brain areas that are
required for a specific linguistic process, lesion studies of individuals with apha-
sia are required. For example, left IFG activations were observed with relation
to the processing of multiple complementation options in a neuroimaging study
(Shereet et al. 2007), although individuals with Broca’s aphasia showed sensitivity
to the number of options in a cross-modal lexical decision task (Shapiro et al.
1993). This suggests that the IFG participates in the processing of multiple op-
tions but it does not have a critical role in this process.

Methodological considerations

One key methodological concern that limits the potential to conclusively inter-
pret neuroimaging results comes from the dominant experimental paradigm in
fMRI studies, the subtraction paradigm. Within this paradigm, the experimental
conditions should differ on one dimension only. If the conditions differ on more
than one dimension, it is difficult to determine which dimension is responsible
for the elevated activations. Hence, any inference regarding the function of brain
areas made based on a design in which the experimental conditions are different
on two or more dimensions is weak (Caplan 2007). For example, we have found
 Between linguistics and neuroimaging 173

differential brain activations for unaccusative and unergative verbs (Shetreet et al.
2010a). However, the only firm conclusion that we could offer was that different
processes are involved when these verbs are encountered. This is because under
any linguistic theory of unaccusative derivation, unaccusatives and unergatives
differ on more than one dimension (e.g., their formation, their semantics, their
likelihood to appear with an animate subject and so on).
To help uncover the role of an activated brain region, neuroscientists often
consult the existing neuroimaging literature. If a brain region was found active
in one study that has manipulated a certain linguistic function/dimension, it
strengthens the interpretation of an activation found in the same brain region
in another study in terms of the same function or dimension. However, relying
on past literature for interpretation can be tricky, as several studies cannot con-
clusively determine the source of the activations they found (due to some of the
limitations discussed in this paper). For example, we have found increased acti-
vation in the left IFG (which is in the vicinity of Broca’s area) with unaccusatives
compared to unergative (Shetreet et al. 2010a). We ascribed this activation to
the syntactic movement associated with unaccusatives and not with unergatives
(Burzio 1986; Levin and Rappaport-Hovav 1995; Perlmutter 1978; Perlmutter and
Postal 1984). This interpretation was based on previous neuroimaging findings
attributing a role in syntactic processing and specifically in syntactic movement
to this area (e.g., Ben-Shachar et al. 2003; Grodzinsky and Friederici 2006; Just et
al. 1996; Stromswold et al. 1996). However, recent claims suggest that the left IFG
is involved in processing of functions other than linguistic complexity (e.g., work-
ing memory load, Kaan and Swaab 2002). Such an account obviously weakens our
conclusion that the activation in the left IFG is due to syntactic movement.
Thus, to form experimental settings that allow less ambiguous interpretation
of the results, strict control of several linguistic and psycholinguistic dimensions
should be practiced. The need for control highlights the importance of consider-
ing theoretical concepts defined within the linguistic literature. It is not sufficient
to control for psycholinguistic and cognitive dimensions such as verb frequency,
length, or imageability. One should also consider the various linguistic aspects of
verb representation (such as the number of arguments, possible subcategoriza-
tion frames, or alternations between forms). These should be kept constant across
the experimental conditions to avoid confounds. Additionally, when verbs are
presented in sentences, the syntactic structure should be controlled across condi-
tions, as well as the sentence duration.
Unfortunately, manipulating one dimension of verb representation, while
perfectly controlling for all other dimensions, can prove rather challenging. Due
to the nature of the linguistic stimuli, dissociating certain dimensions may often
be difficult. Going back, for example, to our study of lexical complexity (Shetreet
174 Einat Shetreet

et al. 2009), we have manipulated the syntactic complexity of the verb classes (con-
trolling for number of subcategorization frames, number of arguments, sentence
structure frequency, duration, etc.). However, this manipulation also affected the
semantics of the two verbs classes. Verbs that allow syntactically complex (CP) ar-
guments, such as complain or think, usually denote mental activities (which may
require a sentential complement). This is not the case for verbs with syntactically
simpler arguments, which are often action verbs, such as destroy or push. There-
fore, we could not conclusively determine whether the activation we observed
when comparing the two verb classes (using the same sentence structure) was
due to the lexical-syntactic or the lexical-semantic differences between the verbs.
(Nonetheless, we could unequivocally argue that there is a lexical difference be-
tween the two verb classes). Dissociation between linguistic dimensions is thus
another concern posed by the subtraction paradigm and should be considered in
the interpretation of the results.
Because it is extremely difficult (if not impossible) to create verb classes that
differ on one dimension only, compromises on some aspects of the experimental
control are deserving. The dimensions on which the compromises apply should
be selected based on the study question. In the lexical complexity study that I just
mentioned above (Shetreet et al. 2009), for example, dissociating the lexical-se-
mantic dimension from the lexical-syntactic dimension was secondary. Our aim
was to test whether lexical complexity affected sentence processing. For that, we
specifically dissociated lexical and sentential properties of our stimuli. We con-
trolled for the syntactic structure in which the verbs were presented by using a
subcategorization frame shared by the two verb classes. That is, we used a PP
argument for both the syntactically-complex verbs (e.g., complain) and the syn-
tactically-simpler verbs (e.g., nibble). This allowed us to show effects related to
the lexical properties of the verbs but not to the syntactic structure. In this case,
using the CP argument for the syntactically-complex verbs and the NP argument
for the simple verbs would have created a difference between the verb classes
on both lexical and syntactic dimensions. That would have therefore produced a
confound between the effects of syntactic structure and the effects of the lexical
representation.
Another concern that researchers face when designing an experiment in-
volves the form of presentation and the task. Many neuroimaging investigations
of verb processing have used tasks with single words (e.g., den Ouden et al. 2009;
Thompson et al. 2007). This is beneficial in order to study the lexical aspects of
verb representations while avoiding structural confounds that can arise from
embedding verbs within sentences. However, studying isolated verbs does not
capture the entire impact of the lexical representation on language processing,
 Between linguistics and neuroimaging 175

because it does not tell us how the lexical information is employed when incor-
porating the verb into a sentence and what effects it carries when the sentential
environment for the verb is already set. For example, we have shown that lexical
ambiguities (i.e., with regard to the subcategorization frame in multiple-frame
verbs) affect sentence processing even when the ambiguity is resolved by the syn-
tactic structure of the sentence (Shetreet et al. 2007) and even when it is resolved
prior to encountering the verb (Shetreet et al. 2011). Furthermore, it is unclear
whether the processing of an isolated word involves the same linguistic and cog-
nitive mechanisms as processing a whole sentence. In real life, outside the ex-
perimental settings, language users encounter verbs mostly within a sentential
context.
The need to reveal the mechanisms that participate in language use in every-
day life should also constrain the selection of the task. Although passive reading
or listening to words and sentences may closely replicate everyday language use, it
is usually avoided in neuroimaging studies. Experimenters try to ensure that sub-
jects are awake and attentive (as subjects lie inside the MRI scanner away from the
experimenter). Thus, a task by which the experimenter can evaluate the attentive-
ness of subjects is frequently used. Experimental investigations are rarely natural-
istic, however, some tasks better replicate language processing in the real world.
Tasks that require judgments on the correctness of the linguistic form of the stim-
uli (i.e., grammatically judgments or lexical decision task) may involve linguistic
and cognitive processes that are not usually employed when using language in
everyday life. Furthermore, they may influence the cognitive performance and
thus, the brain activations. This is specifically crucial when using the subtraction
paradigm for the comparison of words and non-words or grammatical and un-
grammatical sentences, a method often employed in neuroimaging studies (e.g.,
Friederici and Frisch 2000; Thompson et al. 2007). It is not clear what processes
are revealed when subtracting ungrammatical from grammatical utterances be-
cause there is no model that describes what are the mechanisms that underlie
ungrammaticalities in language use and whether the same mechanisms subserve
grammatical and ungrammatical utterances. Using tasks that involve explicit
judgments on the tested linguistic dimension (e.g., identifying the type of verb)
may also be problematic. In such tasks, the attention is drawn to specific aspects
of the stimuli in ways that do not occur in everyday language use, and thus, may
lead to the employment of processes that usually do not participate in language
processing. Therefore, tasks should have a more natural context and should be
irrelevant to the linguistic dimension that is being studied (e.g., making judgment
on the content of the sentence (e.g., Shetreet et al. 2007) or asking comprehension
questions following the sentences (e.g., Ben-Shachar et al. 2003)).
176 Einat Shetreet

A case study: The derivation of reflexive verbs

In the sections above, I have discussed the importance of consulting with the
linguistic literature when defining the experimental question, as well as when pre-
paring the stimuli. I specifically distinguished between studies that contrast lin-
guistic and cognitive theories and studies that aim to map cognitive functions to
brain regions. In this section, I will illustrate the approach I have been advocating
for by thoroughly reviewing a study that focuses on a linguistic contrast. For that,
I will use our study that examined the derivation of reflexive verbs (Shetreet and
Friedmann 2012).
It is standard to maintain that different types of intransitive verbs are derived
differently. Certain accounts concerning this propose that unergatives, such as
smile, are retrieved from the lexicon as is, whereas unaccusatives, such as fall,
are derived from the transitive form by a reduction of the external argument.
Following the reduction, the internal argument at the post-verbal position of the
transitive form is moved to the subject position (e.g., Levin and Rappaport-Hovav
1995; Perlmutter 1978; Reinhart 2002; Reinhart and Siloni 2005). A simplified ex-
ample is in the sentence “the cup broke” which is generated from the form “(some-
one) broke the cup”, where the external argument (someone) is removed and object
“the cup” is moved to the subject position. Thus, unergatives and unaccusatives
differ on two aspects: the reduction operation and the syntactic movement.
The picture regarding the derivation of reflexive verbs, such as shave, under
these accounts, is more controversial. The reflexive form is assumed to be derived
from a transitive form where the subject (e.g., Johns) and the object (e.g., Johno)
are identical. However, it is debated which argument is being reduced. Some ar-
gue for an unaccusative-like derivation. According to this account, an operation
reduces the external argument from the transitive form, and thus the internal
argument is moved from its original position to the subject position (Bouchard
1984; Kayne 1988; Marantz 1984; Pesetsky 1995). Thus, the subject of a reflexive
verb is actually the object like in the sentence “Johno shaved”. Others argue that
the internal argument is the one to be reduced. Therefore, no further operation
is needed and no movement is included in reflexive derivation (Chierchia 1989;
Grimshaw 1982; Reinhart 1996; Reinhart and Siloni 2004; Siloni 2008). That is,
the object is deleted to create the sentence “Johns shaved”.
This debate can be readily informed from an fMRI study, as the two accounts
make different predictions regarding the similarities between unaccusatives and
reflexives (Table 1). The main difference between the two approaches is in the
presence of syntactic movement. The unaccusative-like approach assumes that a
syntactic movement is involved in the derivation of reflexives, whereas the other
approach assumes that no such movement is involved. Another difference is in
 Between linguistics and neuroimaging 177

Table 1. Predicated brain similarities and differences between unaccusatives

and reflexives under the external-argument and the internal-argument reduction
approaches
External-argument Internal-argument reduction approach
reduction approach
Same mechanism Different mechanisms
for reduction for reductions
Syntactic movement same different different
Reduction operation same same different

the reduction operation. Although the two approaches predict that a reduction
operation is involved in the derivation process, they make different assumptions
regarding the targeted argument: the unaccusative-like approach assumes that
the external argument is reduced and the other approach assumes that the in-
ternal argument is reduced. It is possible that all reduction operations occur via
the same mechanism and thus no difference between the approaches is predicted
with regard to this feature. Another possibility is that different reduction opera-
tions (external or internal argument) are generated in different ways. This would
suggest greater differences between reflexives and unaccusatives under the in-
ternal-argument reduction approach, because different argument positions are
targeted in the two verb classes.
The above differences can be translated into differences in predicted brain
activation patterns (Table 1). The unaccusative-like approach should predict sim-
ilar activation patterns for unaccusatives and reflexives, because both verb classes
share the same operations in their derivation process. By contrast, the internal-ar-
gument approach predicts greater differences between the activation patterns of
the two verb classes, as their derivation is different with regard to the inclusion of
the movement operation and possibly also with regard to the type of reduction
operation.
To test this, we contrasted reflexive and unaccusative verbs. We also includ-
ed unergative verbs to set the baseline activation for non-derivational verbs. In
Hebrew, reflexive verbs appear in a specific template, the template “hitpael”. Un-
accusatives (and unergatives) can appear in various templates including “hitpael”.
Because it is unknown how Hebrew morphological templates affect processing,
only verbs of the “hitpael” template were included in our experiment. Verbs were
selected according to semantic and syntactic criteria (see Shetreet and Friedmann
2012 for further information).
The selected verbs were inserted into short sentences that included a subject
and two additional elements (that were added for task purposes). Reflexive verbs
require animate subjects, thus such subjects were used with all the verbs in the
178 Einat Shetreet

experiment to avoid any effects that may be related to animacy. In addition to

controlling for verb template and subject animacy, we also controlled for cogni-
tive and perceptual factors such as verb frequency, sentence duration, volume,
and plausibility. We also controlled for the sentence structure by using the same
elements in all the conditions (e.g., the adjunct “in-the-main street” was added to
the reflexive verb “scratch-self ”, the unaccusative verb “collapse” and the unerga-
tive verb “jump”).
In this study, like in other studies we have conducted, we used an irrelevant
comprehension task. We asked the participants to decide whether the event de-
scribed in the sentence was more likely to happen at home or not. This was done
to ensure that participants properly listen to the sentences and processed them
fully. Furthermore, this task did not uncover the tested feature (as was indicated
by a post-experiment interview with the participants). Our stimuli were present-
ed in blocks of four sentences, and therefore we varied the number of “yes” and
“no” responses between blocks to avoid any response strategies.
Although we did not fully control for all linguistic and cognitive features, as
reflexives and unaccusatives differ on more than one dimension, we could inform
the theory using our neuroimaging results. The comparison of unaccusatives and
reflexives (where unaccusatives > reflexives) showed activations in the left IFG
and the left MTG. Importantly, the left IFG and the left MTG were also found in
the baseline comparison of unaccusatives and unergatives (where unaccusatives
> unergatives). Furthermore, in a previous study, we had identified the left IFG
and the left MTG as key areas for the neural distinction between unaccusative
and unergatives (Shetreet et al. 2010a). Finally, the comparison of reflexives with
unergatives (where reflexives > unergatives) did not reveal activations in the left
IFG nor the left MTG.
Thus, our results show that the activation pattern associated with process-
ing reflexive verbs is extensively different from the activation pattern associated
with unaccusative verbs. This indicates that reflexives differ from unaccusatives,
similarly to the way that unergatives differ from unaccusatives. Because the inter-
nal-argument approach predicts great differences between reflexives and unaccu-
satives (based on the different derivational operations), our results lend support
to this approach.
Our results show a clear difference in brain activations between unaccusa-
tives and reflexives, as well as unaccusatives and unergatives. Bear in mind that
activations in the left IFG and left MTG with unaccusatives were identified in
two independent studies with different sets of verbs. Although the results have
been replicated, we could not exclusively determine the roles of the areas in the
processing of unaccusative.
 Between linguistics and neuroimaging 179

With regard to the activation in the left IFG, we had previously speculated
that it is linked to the syntactic movement of the internal argument to the subject
position in unaccusatives (compared with unergatives and transitive, Shetreet et
al. 2010a). We based our hypothesis on findings from previous neuroimaging ex-
periments and studies with aphasic patients. If this assumption is correct, then the
absence of activation in the left IFG with reflexive verbs suggests that there is no
syntactic movement. This further suggests that it is the internal argument that is
reduced with reflexives. However, we cannot conclusively assert this, because we
cannot isolate the cognitive or linguistic process that elicit the activation in the
left IFG.
In Shetreet et al. (2010a), which compared unaccusatives with unergatives,
we speculated that the MTG is activated with unaccusatives due to the reduc-
tion operation. Although the internal-argument approach, which is supported
by the brain patterns observed in our study, assumes that the derivation of re-
flexives involves a reduction operation, no activation in the MTG was found with
the processing of these verbs. One can offer various explanations for the lack of
MTG activation in the processing of reflexive verbs. This may simply suggest that
different reduction operations occur via different mechanisms in different brain
regions. Alternatively, we should consider Reinhart and Siloni’s suggestion (2005)
regarding a “bundling” operation with reflexives. According to this suggestion,
the derivation of reflexives does not involve reduction of an argument, but rather
an operation that merges the two thematic roles of the external and internal argu-
ments into one to allow its assignment to one argument. That is, the representa-
tion of reflexive verbs includes two thematic roles. Thus, if the MTG activation is
related to a reduction operation (or to lexical valence change), it is not expected
to show with reflexives. Another possibility is that the reduction operation does
not occur every time the verb is encountered (Horvath and Siloni 2009). Instead,
the derived lexical form of unaccusative and reflexives is stored with their acqui-
sition. If so, no brain activation related to the derivation process is expected with
unaccusatives or reflexives. Under such an assumption, the activation in the MTG
may be linked to the syntactic movement. We have argued that this possibility is
not likely based on previous studies that consistently showed brain activation in
the MTG with lexical processing.
As I mentioned above, various reasons can account for the MTG activation
with unaccusatives and its absences with reflexives. Our study design does not
enable us to reveal the source of the difference between the verb classes. Within the
limits of our design, we could suggest different interpretations based on findings
from previous neuroimaging studies. Linguistic theories could further constrain
our interpretation by defining the derivation process of the verb classes. However,
180 Einat Shetreet

the reader should be aware that these are merely speculations and further research
is needed in order to determine the role of the MTG with unaccusatives.

Concluding remarks

A final comment on the relationship between linguistics and neurolinguistic is in

place. Neuroimaging results, like any other experimental results, cannot uncon-
ditionally and unequivocally prove or disprove a linguistic theory. They can, how-
ever, provide support for one theory over another. Furthermore, neurolinguistic
results should be considered in light of the theoretical linguistic analysis, as well
as be weighed against findings regarding the same topic from other neuroimag-
ing studies that used other tasks, from psycholinguistic behavioral studies, and
from studies with individuals with aphasia or other language impairments. These
together can paint a more comprehensive empirical picture against which the the-
ories under discussion should be evaluated.
In conclusion, by considering linguistic concepts and theories of verb rep-
resentation when approaching a new neurolinguistic study, one can ask fine-
grained and sophisticated questions, as well as create well-designed experiments
that allow a more accurate interpretation of the results. By doing so, neurolinguis-
tics can bring insights into our understanding of both brain and language. On one
direction, differential brain activation patterns can inform theoretical linguistic
doubts and debates. On the other direction, linguistic ideas can be used to con-
strain the interpretation of neuroimaging results and enhance our understanding
of how the brain processes verbs.

Note

* I thank Rama Novogrodsky for a helpful discussion and her comments on this manuscript,
and Asaf Bachrach for encouraging me to write it. I also thank Naama Friedmann for teaching
me to think linguistically and critically, and for her comments on the manuscript too.

References

Bachrach, Asaf. 2008. Imaging neural correlates of syntactic complexity in a naturalistic context.
Doctoral dissertation, MIT.
Ben-Shachar, Michal, Hendler, Talma, Kahn, Itamar, Ben-Bashat, Dafna and Grodzinsky, Yosef.
2003. The neural reality of syntactic transformations: Evidence from fMRI. Psychological
Science 14: 433–440. DOI: 10.1111/1467-9280.01459
 Between linguistics and neuroimaging 181

Borer, Hagit. 2004. The grammar machine. In The Unaccusativity Puzzle, Artemis Alexiadou,
Elena Anagnostopoulou and Martin Everaert (eds), 288–331. Oxford: Oxford University
Press. DOI: 10.1093/acprof:oso/9780199257652.003.0012
Bouchard, Dan. 1984. On the content of empty categories. Foris, Dordrecht.
Burzio, Luigi. 1986. Italian syntax. Dordrecht: Reidel. DOI: 10.1007/978-94-009-4522-7
Caplan, David. 2007. Functional neuroimaging studies of syntactic processing in sentence
comprehension: A critical selective review. Language and Linguistic compass 1: 32–47.
DOI: 10.1111/j.1749-818X.2007.00005.x
Chafe, Wallace L. 1970. Meaning and the Structure of Language. Chicago: University of Chicago
Press.
Chomsky, Noam. 1965. Aspects of the theory of syntax. Cambridge, MA: MIT Press.
Chierchia, Gennaro. 1989. A semantics for unaccusatives and its syntactic consequences. Ms.,
Cornell University.
Chierchia, Gennaro. 2004. A semantics for unaccusatives and its syntactic consequences. In The
Unaccusativity Puzzle, Artemis Alexiadou, Elena Anagnostopoulou and Martin Everaert
(eds), pp. 22–59. Oxford: Oxford University Press.
DOI: 10.1093/acprof:oso/9780199257652.003.0002
Coltheart Max. 2006. What has functional neuroimaging told us about the mind (so far)? Cor-
tex 42: 323–333. DOI: 10.1016/S0010-9452(08)70358-7
Friederici, Angela D. and Frisch, Stefan. 2000. Verb argument structure processing: The role
of verb-specific and argument specific information. Journal of Language and Memory 43:
476–507. DOI: 10.1006/jmla.2000.2709
Grimshaw, Jane. 1979. Complement selection and the lexicon. Linguistic Inquiry 10: 279–326.
Grimshaw, Jane. 1982. On the lexical representation of Romance reflexive clitics. In The mental
representation of grammatical relations, Joan Bresnan (ed). MIT Press, Cambridge, MA.
Grodzinsky, Yosef and Friederici, Angela. D. 2006. Neuroimaging of syntax and syntactic pro-
cessing. Current Opinion in Neurobiology 16: 240–246. DOI: 10.1016/j.conb.2006.03.007
Henson, Richard. 2006. Forward inference using functional neuroimaging: Dissociations ver-
sus associations. Trends in Cognitive Sciences 10: 64–69. DOI: 10.1016/j.tics.2005.12.005
Horvath, Julia and Siloni Tal. 2009. Hebrew idioms: The organization of the lexical component.
Brill’s Annual of Afroasiatic Languages and Linguistics 1: 283–310.
DOI: 10.1163/187666309X12491131130666
van Hout, Angeliek. 1992. Linking and projection based on event structure. Ms, Tilburg: Tilburg
University.
Just, Marcel A., Carpenter, Patricia A., Keller, Timothy A., Eddy, William F. and Thulborn,
Keith R. 1996. Brain activation modulated by sentence comprehension. Science 274: 114–
116. DOI: 10.1126/science.274.5284.114
Kaan, Edith and Swaab, Tamara. 2002. The brain circuitry of syntactic comprehension. Trends
in cognitive neuroscience 6: 350–356. DOI: 10.1016/S1364-6613(02)01947-2
Kayne, Richard. 1988. Romance se/si. Abstract presented at the GLOW 11 Colloquium, Buda-
pest, GLOW newsletter 20: 33.
Levin, Beth A. and Rappaport-Hovav, Malka. 1995. Unaccusativity. Cambridge, MA: MIT Press
Marantz, Alec. 1984. On the nature of grammatical relations. MIT Press, Cambridge, MA.
Moro, A., Tettamanti, M., Perani, D., Donati, C., Cappa, S. F., & Fazio, F. 2001. Syntax and the
brain: Disentangling grammar by selective anomalies. Neuroimage 13: 110–118.
182 Einat Shetreet

den Ouden, Dirk-Bart, Fix, Steve, Parrish, Todd B. and Thompson, Cynthia K. 2009. Argument
Structure effects in action verb naming in static and dynamic conditions. Journal of Neu-
rolinguistics 22: 196–215. DOI: 10.1016/j.jneuroling.2008.10.004
Poeppel, David and Embick, David. 2005. Defining the relation between linguistics and neu-
roscience. In Twenty-first century psycholinguistics: Four cornerstones, Anne Cutler (ed).
Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
Perlmutter, David M. 1978. Impersonal passives and the unaccusative hypothesis. In Proceed-
ings of the Fourth Annual Meeting of the Berkeley Linguistic Society, 159–189. Berkeley, CA:
Berkeley Linguistics Society.
Perlmutter, David M. and Postal, Paul M. 1984. The 1-advancement exclusiveness hypothesis.
In Studies in relational grammar 2, David M. Perlmutter and Carol Rosen (eds), 81–126.
Chicago: University of Chicago Press.
Pesetsky, David M. 1995. Zero syntax: Experiencer and cascades. MIT Press, Cambridge, MA.
Reinhart, Tanya. 1996. Syntactic effects of lexical operations: Reflexives and unaccusatives. OTS
Working Papers in Linguistics. Utrecht University, OTS, Utrecht.
Reinhart, Tanya. 2002. The theta system: An overview. Theoretical Linguistics 28: 229–290.
Reinhart, Tanya and Siloni, Tali. 2004. Against the unaccusative analysis of reflexives. In The
Unaccusativity Puzzle, Artemis Alexiadou, Elena Anagnostopoulou and Martin Everaert
(eds), pp. 159–180. Oxford: Oxford University Press.
Reinhart, Tanya and Siloni, Tali. 2005. The lexicon-syntax parameter: Reflexivization and other
arity operations. Linguistic Inquiry 36: 389–436.
Shapiro, Lewis P., Gordon, Betty, Hack, Nancy and Killackey, Jennifer. 1993. Verb argument
processing in complex sentences in Broca and Wernicke’s aphasia. Brain and Language 45:
423–447.
Shetreet, Einat and Friedmann, Naama. 2012. Stretched, jumped and fell: An fMRI investiga-
tion of reflexive verbs and other intransitives. NeuroImage.
Shetreet, Einat, Friedmann, Naama and Hadar, Uri. 2009. An fMRI study of syntactic layers:
Sentential and lexical aspects of embedding. Neuroimage 48: 707–716.
Shetreet, Einat, Friedmann, Naama and Hadar, Uri. 2010a. The neural correlates of linguis-
tic distinctions: Unaccusative and unergative verbs. Journal of Cognitive Neuroscience 22:
2306–2315.
Shetreet, Einat, Friedmann, Naama and Hadar, Uri. 2010b. The representation of optional com-
plements: The theoretical contribution of fMRI. Human Brain Mapping 31: 770–785.
Shetreet, Einat, Linzen, Tal and Friedmann, Naama. 2011. The effects of complement predict-
ability on the processing of verb’s complementation options. Poster presented at the 3rd
Neurobiology of Language Conference, Annapolis, MD, USA.
Shetreet, Einat, Palti, Dafna, Friedmann, Naama and Hadar Uri. 2007. Cortical representation
of verb processing in sentence comprehension: Number of complements, subcategoriza-
tion and thematic frames. Cerebral Cortex 17: 1958–1969.
Siloni, T. 2008. On the Hitpa’el template. In Theoretical Hebrew Linguistics, Galia Hatav (ed),
pp. 111–138. Jerusalem: The Hebrew University Magnes Press. (In Hebrew)
Stromswold, Karin, Caplan, David, Alpert, Nathaniel and Rauch, Scott. 1996. Localization
of syntactic comprehension by positron emission tomography. Brain and Language 52:
452–473.
 Between linguistics and neuroimaging 183

Thompson, Cynthia K., Bonakdarpour, Borna, Fix, Stephen C., Blumenfeld, Henrike K.,
Parrish, Todd B., Gitelman, Darren R. and Mesulam, M.-Marsel. 2007. Neural correlates
of verb argument structure processing. Journal of Cognitive Neuroscience 19: 1753–1767.
van Valin, Robert D. 2001. An introduction to syntax. Cambridge, UK: Cambridge University
Press.
 Argument structure
Creating a productive space for theory
and experimentation

Gillian Ramchand
University of Tromsø / CASTL

1. Introduction

What do we know when we know a verb in our language? Are the radical con-
structivists right that (verbal) roots are essentially asyntactic, containing only
conceptual information and not even category information let alone argument
structure? Or are the lexicalists right that there is a rich set of event structure and
argument information present in the entries of verbs?
In this paper, I use the concrete example about the debate about the lexical
representation of unaccusatives to explore the question of how psycholinguis-
tic and neurolinguistic approaches to argument structure can potentially inform
questions of linguistic analysis. The results of this short paper will be both nega-
tive and positive. On the negative side, there will unfortunately be no straightfor-
wardly smoking gun that these other methodologies can uncover. On the positive
side, the discussion will allow us to be specific about the nature of the problem
space, where solutions are being constrained by our emerging understanding of
data from all these different methodologies.
It will also allow us to be explicit about the ways in which we can improve
the extent to which we learn from all these different sources and move towards
genuine scientific advance. In terms of the contribution of theoreticians to the
productive workspace, I will argue that the focus on frame classes such as ‘un-
accusative’ sidesteps the important question of form-frame alternations which
current theoretical research on argument structure has now shown to be central
to understanding the relationship between verbal lexical items and syntactic rep-
resentations (Marantz 2013; Ramchand 2013; Levin and Rappaport-Hovav 2005).
186 Gillian Ramchand

Psycholinguists, more generally, are in a position to investigate overarching

issues of cognitive processing and integration that can begin to place reliable
boundary conditions on the specific linguistic theories that are developed. To-
gether, the hope is that one can build cumulative databases of correlations be-
tween linguistic representations and processing behaviour across different tasks
that will allow us to make progress on the deepest issues of symbolic thought.
In the end, I see this short paper as a kind of modest manifesto for continued
productive collaboration between theoreticians and psycholinguists, in an area
where it is far too easy to pursue respective research agendas in a way that makes
it impossible or irrelevant to compare results.

2. The causative-inchoative alternation

One of the most common types of verbal alternation across languages is one be-
tween an intransitive verbal concept and a transitivized or causativized version.
In many languages, the alternation is mediated by regular morphology and the
transitive contains an extra morpheme (e.g. Hindi/Urdu, Indonesian). In other
languages, the alternation can be mediated by morphology, but it is the intran-
sitive variant that has the extra morpheme (e.g. Italian, Norwegian, where what
is added is the ‘reflexive’ clitic si, or seg respectively). In yet other languages, like
English, both transitive and intransitive variants have exactly the same morpho-
logical form. See Haspelmath (1993) for a typological overview.
(1) a. makaan ban-aa Hindi
house make-PERF.M.SG
‘The house was built.’
b. anjum-ne makaan ban-aa-yaa
Anjum-ERG house make-aa-PERF.M.SG
‘Anjum built a house.’  (from Butt 2003)
(2) a. Il vento ha rotto la finestra Italian
the wind has broken the window
‘The wind broke the window.’
b. La finestra *(si) è rotta
The window REFL is broken
‘The window broke.’  (from Folli 2001)
(3) a. The wind broke the window. English
b. The window broke.
 Creating a productive space for theory and experimentation 187

The phenomenon of verbal alternations raises deep questions for the organiza-
tion of grammar. In particular, recent debate has centred around the question of
how much internal structure needs to be represented in verbal lexical items, and
whether there are lexicon internal rules and relations in the grammar in addition
to those given in the syntax. Regular alternations, especially those that are cross-
linguistically pervasive, are important because it is implausible to represent them
using multiple unrelated but homonymous lexical items.
One of the major theoretical debates that the transitivity alternation shown
above has raised is the question of which of the two alternants, if any, is semanti-
cally primary, and which is derived. As we have seen, the morphology of different
languages does not give us unambiguous results in this case, and it gives us no
clear information in the case of English. In fact, the most popular current theory
in generative grammar is to take the intransitive as being derived from the transi-
tive alternant and to take even the intransitive’s conceptual structure to be under-
lyingly extremely complex (Levin and Rappaport Hovav 1995; Chierchia 2004).
The important thing about the transitivity alternation is that it is extremely
common and pervasive crosslinguistically. Semantically, it seems that the addi-
tional expression of a ‘causer’ is what makes the difference between the transi-
tive and the intransitive version. Most structured theories of the lexicon assume
event structure templates that include primitives such as CAUSE and PROCESS and
represent ‘caused’ events as more complex than ‘uncaused’ events (see Levin and
Rappaport Hovav 1995), even to the extent of the former containing an additional
subevent in their representation (Croft 1998, Parsons 1990, Pustejovsky 1995).
Schematically, the minimal difference between caused and uncaused events that
otherwise share a root, should be the addition of a CAUSE conceptual component
to the event structure representation.1
(4) i. Caused Event of Change: CAUSE < CHANGE
ii. Uncaused Event of Change: CHANGE

However, as we have seen, there is a debate in the literature concerning the direc-
tion of the causative-inchoative alternation, and indeed concerning whether the
surface intransitive truly is simpler from an event structure point of view than the
corresponding transitive. Levin and Rappaport Hovav (1995), Chierchia (2004)
and Reinhart (2002) all agree in deriving the inchoative alternant from a lexically
causative base. For example, Levin and Rappaport Hovav (1995) argue that the
transitive is the base form, and that the intransitive is derived by a lexical suppres-
sion of the CAUSE component in the item’s lexical conceptual structure. Since not
all transitive verbs with a CAUSE component actually have intransitive counter-
parts, a lexicon internal condition must be placed on the suppression mechanism.
188 Gillian Ramchand

Thus, for many researchers, both members of the verbs in English that alternate
for transitivity have the lexical template in (i).
Regardless of what one thinks about the causative-inchoative alternation, it
is certainly the case that some intransitives in English fall under the proposed
template in (ii). These intransitives would include the uncausativizable fall and
arrive in English. At the same time, there are verbs that have the template in (i),
but which never surface in intransitive form so that the rule of CAUSE suppression
never applies to them. Examples in this category would include destroy and reject
in English. Another verb type whose event structure is subject to some controver-
sy are the verbs of ‘internal causation’ in the terms of Levin and Rappaport Hovav
(1995). These include verbs like bloom and stink in English, which seem to lack
an agent per se, but where the action seems to derive from the inherent properties
of the subject in some way. In addition, the more general class of unergatives is
also important here, since they are claimed to be intransitives that have a more
uncontroversial CAUSE in their representation (verbs such as dance and swim in
English).
It seems to me that the theoretical discussion has arrived at something of a
stalemate, with farreaching architectural decisions hinging on evaluating the in-
ternal lexical complexity of various verb forms. Given that simple grammaticality
judgements and individual introspection do not give us any real purchase on this
problem, it is not surprising that theoretical linguists are increasingly turning to
other methodologies to gain a more realistic psychological and processing-based
answer to these questions. The psycholinguistic literature in the mean time has
had a long history of attention to the structure of the lexical database. We now
have a lot of data about the speed of response times in various kinds of lexical
decision tasks and lexical processing. For example in priming experiments, it has
been shown that lexical access to individual items is facilitated by words that are
semantically related (semantic primes), by words that are phonologically related
(phonological primes), and those that are in a regular morphological relationship
to the target (see Altmann 1997 for a general introduction). Interestingly, in the
latter category, regular morphological relationships function differently from ir-
regular but paradigmatic morphology (Pinker 1999).
So what about internal lexical complexity? Can we detect processing effects
independent of frequency and availability of cohorts in phonological and seman-
tic domains, that can be directly ascribed to conceptual or representational com-
plexity? Recently, in collaboration with theoretical linguists, there have been a
number of investigations of a psycholinguistic nature attempting to control for
the other known factors influencing lexical processing time to isolate this inde-
pendent factor of complexity in lexical representation (McKoon and Macfarland
2002; Gennari and Poeppel 2003; Husband et al. 2008), and the work of Cynthia
 Creating a productive space for theory and experimentation 189

Thompson and her colleagues has explicitly targetted argument structure com-
plexity in their work with Aphasic patients (Thompson 2003; Thompson and Lee
2009).
Given the emerging evidence then, maybe we can get answers from psycho-
linguistic methodologies for a specific question argued over in the theoretical lit-
erature, phrased as in (5) below:
(5) Does the lexical semantic representation of unaccusatives contain a CAUSE
component, making it thereby just as event structurally complex as a transi-
tive verb with a CAUSE < CHANGE template?

2.1 Psycholinguistic correlates of verbal complexity

As indicated above, recent psycholinguistic research has started to ask the question of
whether there are measurable processing correlates of verbal complexity. Let us briefly
summarize the initial indications. A number of recent studies have suggested that fac-
tors of event structure complexity indeed play a role in the lexical specification of ver-
bal items in a way that affects the speed with which they are accessed and integrated
in sentence comprehension. Brennan and Pylkkänen (2010) use a self paced reading
experiment (supplemented with MEG) to show that Object Experiencer Psych verbs,
which express the causation of a mental state in some experiencer, are associated with
slower reaction times than Subject Experiencer verbs which have no causational tran-
sition and which are simply stative. These findings accord with Gennari and Poeppel
(2003), in which eventive causational verbs like break in English are compared to
simple states in self paced reading and lexical decision tasks. They found that events
were associated with significantly slower reading times and decision times than states.
In a series of experiments that compared different types of events, McKoon and
Macfarland (2000) and McKoon and Macfarland (2002) compared classes of verbs
called internally vs. externally caused in the theory of Levin and Rappaport Hovav
(1995) where only the latter involved a separable CAUSE subevent in the lexical repre-
sentation. Once again, under whole sentence reading tasks and lexical decision, they
found that the more causally complex verbs (the externally caused class) provoked
significantly longer reading and reaction times than the verbs without an additional
causing event. While these studies concern different subsets of the verbal vocabulary,
they are consistent in that they show an effect of ‘caused state’ vs. ‘state’, with the
former showing longer reaction times. Similarly, what we know from the experi-
ments with aphasics is that the larger the number of arguments, the harder it is for
an agrammatic aphasic to generate a particular verb (Thompson 2003).
How does the class of unaccusatives fit into this picture of lexical represen-
tational complexity? Do they show evidence of an extra CAUSE component? Psy-
cholinguistic experiments by Bard et al. 2010 have shown that unaccusative vs.
190 Gillian Ramchand

unergative verb types have testable effects on processing even at the sentence level,
confirming that the distinction is relevant to the syntax. McKoon and Macfarland
report longer reaction times for the class of verbs in Levin and Rappaport’s ex-
ternally caused intransitives than for internally caused intransitives. But this still
does not properly zero in on what precisely the cause of that extra complexity is.
Recall that Levin and Rappaport Hovav’s class of alternating verbs are called verbs
of ‘external causation’ because they argue that CAUSE may be suppressed as part of
a lexical rule to give the intransitive version (although the complexity of the lex-
ical representation still remains underlyingly). This is the complexity that McK-
oon and Macfarland (2000; 2002) invoke to explain their experimental results.
However, the contrast set for McKoon and Macfarland was Levin and Rappaport
Hovav’s ‘internally caused verbs’ whose status with respect to the CAUSE com-
ponent is controversial. They did not embrace the larger category of unergatives,
nor did they include non-alternating transitives and intransitives. Thus the base
line behaviour for verbs which are uncontroversial exemplars of the templates in
(i) and (ii) above were never tested. This makes it difficult to interpret the results
of their study or place the complexity of causative-inchoative verbs and inter-
nally caused verbs in the context of other more uncontroversial verb types. Nev-
ertheless, what evidence there is from adult speakers of English seems to show
increased processing times for unaccusatives both in lexical decision tasks and in
sentence processing, even when frequency was explicitly controlled for. We also
know that unaccusatives are more difficult to access in verbal naming tasks than
unergatives for agrammatic aphasics (Thompson and Lee 2004), and that unaccu-
satives are rarely spontaneously produced.
But this does not actually settle the specific question posed by the theoretical
linguist concerning whether a verb like break contains a CAUSE component in its
lexical representation. Even if it turns out to be true that unaccusatives induce
longer processing times, we still don’t know what it is it about unaccusatives that
makes them so hard. It turns out that there are many such potential reasons. On
page 356 for example,Thompson and Lee (2009) offer all of the following as po-
tential reasons:

A. Extra Movement Processes: unaccusatives have an extra movement in their

syntactic derivation (with the sole argument moving from internal position
to subject position)
B. Telicity: unaccusatives are semantically more complex because they are usual-
ly telic, while unergatives are atelic.
C. Multiple Selectional Frames: unaccusatives (most of them) are alternating
verbs, which might introduce extra complexity due to the potential for more
interpretations.
 Creating a productive space for theory and experimentation 191

D. Causation: unaccusatives have an extra CAUSE component in their event

structure representation as compared to unergatives, even though it is not
necessarily expressed.
E. Non-Canonical Mapping between Participant type and Grammatical Subject:
in unaccusatives, there is a non-canonical mapping between event structure
and topic/subjecthood.

In fact, there is independent evidence that non-canonical mappings between

conceptual structure and syntactic structure are hard in general for agrammatic
aphasics, and that for a population that is plausibly using cognitive heuristics to
make up for their disrupted syntactic component, deviations from non-canoni-
cality are extremely damaging. Here are some facts about aphasic behaviour that
all seem to be consistent with the hypothesis that this is something that they find
difficult: (i) Agrammatic Aphasics are worse at passives than they are at actives;
(ii) agrammatic aphasics are worse at Object Experiencer verbs than Subject ex-
periencer verbs; (iii) agrammatic aphasics are actually better at passivized Object
Experiencer verbs than active ones! (iv) agrammatic aphasic are better at unerga-
tives than unaccusatives (Thompson 2003; Thompson and Lee 2009).
We also have to be careful when we look at language impaired populations
because what might be difficult for them might not be the thing that has increased
complexity for us. An example from different kinds of deficits is instructive here.
Kim and Thompson (2004) show that conceptually rich ‘heavy’ verbs are hard
for Alzheimer’s patients, and they produce ‘light’ verbs more easily. In the same
task, Agrammatic Aphasics and normals showed a preference/ease of naming
with heavy verbs. So are conceptually rich verbs more complex than conceptu-
ally ‘light’ verbs? Or the other way around? Basically, the word ‘complex’ at the
moment has too many hidden contextual variables ‘Complex (for whom) (doing
what task) (for what reason)’. In particular, difficulty in production does not nec-
essarily entail lexical representational complexity
Still, there may be additional effects of event structure complexity that arise
simply because of causal complexity and it would be relevant to know whether
unaccusatives pattern with non-caused dynamic eventualities, or eventualities of
direct or indirect causation when other factors are controlled for. Unfortunately,
the processing correlates of causal complexity have not yet been tested in the con-
text of a simple crosslinguistic comparision, where the two languages vary with
respect to the morphology used to mediate the transitivity alternation. Recall, the
typological literature (see Haspelmath 1993) tells us that some languages create
transitives from intransitives by the addition of morphology (e.g. Hindi/Urdu),
others create intransitives from transitives by the addition of extra morpholo-
gy (e.g. Italian, Norwegian), others primarily use a labile, non-morphologically
192 Gillian Ramchand

mediated, alternation (e.g. English). We need to make sure that the results from
comparing verb types are not most directly affected by the fact that the morpho-
logical forms in question are actually systematically ambiguous. Essentially, to
truly test whether the inchoative members of the causative-inchoative alternation
are as lexically complex as transitive verbs, we need to control for the ambiguity
effect, in addition to comparing unaccusatives not just with unergatives but with
base line transitives.
The interpretational gap arises partly because the different possibilities have
not yet been controlled for as discussed above, but also partly because we under-
stand so little about the processes that occur in the brain in the deployment of
lexical items in the building of sentences. The specific details of the theoretical
debate are simply too fine grained to be adjudicated by the sort of the data we
can currently get from from reaction times or brain imaging. We also need to
ask ourselves whether the differences in ‘theory’ we are so attached to actual-
ly willl survive as substantive issues once a deeper understanding of the brain
are achieved. I think it is undeniable that as we learn more about the brain and
cognitive processing, the very nature of the questions we linguists ask will be
transformed and differences that might seem huge under one kind of conception
dissolve into differences of packaging and notation. This is not to say that there
are not real architectural questions at stake. It is just that it is sometimes difficult
to anticipate where they will be.
The data reported by Thompson et al (this volume) concerning the neural
correlates of argument structure complexity are interestingly suggestive. They
show clear activation differences both for sheer valency increase and for differ-
ences in flexibility, as well as deviations from non-canonicality. However, the re-
gions localized for the two sorts of effects seem to be different. So indeed it takes
a longer time to process a verb that has multiple ‘argument frames’ at its disposal,
and longer to process a non-canonical mapping to subject, and all things being
equal it takes longer to process a verb with more arguments (3 vs. 2 being the
clearest case). However, the reported brain regions for this differential increase in
activation appear to be different for these different kinds of complexity.

3. A blueprint for successful interaction

So we can’t really decide conclusively on debates in the terms in which the warring
theoreticians frame them, at least for the particular question I have posed here in
(5). At least not yet. Does this mean that the theoreticians’ hopes for productive
collaboration are misplaced? I think not. What is absolutely clear however is that
a lot of ground work needs to be done to construct a common scientific space
 Creating a productive space for theory and experimentation 193

which is relevant and productive for both sides. In particular, psycholinguists

cannot and should not attempt to answer questions relevant to theoreticians in
their own highly specific terms. As we have seen, some of the debates entertained
by syntacticians with regard to unaccusatives are at a far higher level of specific-
ity than would be reasonable to attempt to test given our current knowledge of
mental processing.
So communication between theoretician and psycholinguist should not con-
sist of a steady flow of experiments designed to test the latest ‘new theory’ of
argument structure. Such experiments are doomed to inconclusiveness since our
knowledge of both the logical structure of human language and psycholinguistic
behaviour are still so sparse that the very terms of art are likely to change and
render all such overly specific experiments obsolete and uninterpretable after five
years. Psycholinguists need to continue to pursue questions that answer to very
general facts about the patterning of behaviour in response to linguistic data, and
these in turn can be used to inform very general framing architectures for the
more fine grained theoretical debates.
However, I think it is important to emphasize that the theoretical literature
also has important insight to pass on to the psycholinguistics, but once again at
a general and non-parochial level. Here I am specifically thinking of the point
raised by Alec Marantz in his recent article on psycholinguistics and argument
structure, which is particularly relevant because the problem I sought an answer
to in this paper was indeed one involving argument structure. Marantz (2013)
makes the important point that thirty years of theoretical research on argument
structure have produced mounting evidence that argument structure generaliza-
tions are generalizations over syntactic representations, and that it is increasingly
difficult to match argument frames in a one to one fashion with a particular lex-
ical exponent of it (see also Ramchand 2013). This is because every verbal lexical
item that we can think of in English, and even in other less flexible languages, has
more than one such argument structure frame at its disposal. Much psycholin-
guistic experimentation in this area still shows a strong lexicalist bias and tries
to put aside the ‘problem’ of alternations as a bothersome issue that needs to be
sidestepped experimentally. The testing of an argument frame in this world be-
comes synonymous with testing a particular verb class. However, it seems rather
that alternations are the core fact that we need to be able to deal with, and if this
is so then the whole nature of the problem changes. In other words, the central
problem as I see it is the question of how the mental lexicon (the database) is pro-
cessed and integrated with syntactic information, the latter of which is the locus
of reliable event-argument relations.
The lexical item bears certain information relevant to its integration in syn-
tactic structure, but the theoretical literature is very divided about the nature
194 Gillian Ramchand

and terms of that information. The options in the theoretical literature about the
specific mechanisms of lexicalization (e.g. projection of event structure informa-
tion (Levin and Hovav 2005); insertion of acategorial roots at the base of the tree
(Borer 2005); adjunction of roots to functional heads (Marantz to appear); span-
ning in a parallel architecture (Ramchand 2008)) are all far too specific to adjudi-
cate at this point. However, psycholinguistic experimentation can potentially do
a lot to narrow down the space of possible implementations if it starts to ask the
question in the same sorts of terms. I think that the immense amount of progress
that has been made in understanding the patterns and flexibilities of argument
structure relationships transcends the theory internal debates and the knowledge
gained can contribute a lot to changing the way that the questions are asked in the
psycholinguistic arena as well.
However, once the notion of argument frame flexibility is properly acknowl-
edged, it importantly affects the way in which experiments are constructed. To
take the empirical issue under discussion here as an example, alternation be-
haviour cannot and should not be ‘controlled for’, in order to simply focus on un-
accusative vs. unergative ‘verbs’. Most of the verbs that diagnose as unaccusative
in English according to the participial modification test for example, also par-
ticipate in the causative-inchoative alternation. Alternation and varying degrees
of flexibility of various lexical items are part of the issue that we wish to under-
stand, which concerns both representation and the integration of various con-
tributions to meaning. Psycholinguistic experimentation on argument structure
has to explicitly tackle the processing effects of flexibility and underspecification
and the interplay between grammatical and conceptual factors and investigate the
word form across behaviours, instead of particular versions of it in specific lexical
frames. Syntactic form gives reliable semantic entailments, but the ways in which
lexical items in the database contribute to and constrain the syntactic forms they
appear in are still to be properly understood.

3.1 Hypothesis testing and levels of specificity

At the psycholinguistic level, hypotheses should be constructed in collaboration

with linguists whose specialities are involved, but who are willing to suspend
some particularities of implementation in favour of investigating broader effects.
We need to arrive at an intermediate level of specificity for the hypotheses to
be tested so that what we are testing is a question whose answer is relevant to
the theoretical literature, but which is not so theory-internal or parochial that it
would be impossible to answer or out of date once the terms of the debate change.
Both types of linguists are required in order to zero in on the right level of testable
 Creating a productive space for theory and experimentation 195

generality here. Once again, taking the specific problem space discussed in this
paper, we could imagine an hypothesis tested experimentally couched in the fol-
lowing different sorts of terms.
(6) I. Are unaccusatives derived from transitives, or the other way around?
II. Are unaccusatives more conceptually complex than unergatives?
III. Are internally caused change of state verbs more internally complex than
externally caused change of state verbs?
IV. With respect to difficulty of processing, in tasks of type X, do verb
types fall into clear natural classes (where we test for e.g. dominant
valency, tendency to alternate, thematic nature of subject, transparent
morphological complexity etc.)?

From my perspective, questions of type I, II and III are all problematic to various
degrees. In III, the category of internally caused vs. externally caused change of
state verbs is an interpretational category division proposed by some specific the-
oretical linguists (Levin and Rappaport Hovav 1995) and has neither clear diag-
nostics nor clear robust acceptance in the literature (This was the organizational
category imposed on the data by McKoon and Macfarland 2000, for example). In
II, the categories are more neutrally robust descriptively speaking, but the notion
of complexity found here and in III is actually too vague, given that we know
that different kinds of complexity (some not even narrowly syntactic) have effects
on processing time (for example, the non-canonicality of having a non-agentive
subject). Question I is even worse because it compounds the category problem
with the notion of derivation, which has a different status in different theories and
which so far we have no clear psycholinguistic fingerprint for.
Question IV is an attempt at formulating the kind of question I would like to
know the answer to, and which I would trust as a solid non-parochial result that
my own theories would have to take notice of. (Of course, if the only natural class-
es within verbs that emerged were ones related to frequency or syllable structure,
or even morphological complexity then I would have less reason to pay attention,
but then at least I would know.) Specifically, in question IV the parameter that
pertains to flexibility vs. more constructional specificity is explicitly part of the
factors to be tested, in line with our recent understanding from the theoretical lit-
erature. In question IV, in response to the emerging richness of the experimental
paradigms being tested at the psycholinguistic level, the question has also been
relativized to the nature of the task. This is because we have seen from the liter-
ature that different kinds of complexity affect different tasks, and with different
time profiles for their effects. The results of an experiment looking for answers to
question IV, would still pose an interpretational gap with respect to the questions
that theoretical linguists are seeking answers to, but at least they would be a start.
196 Gillian Ramchand

3.2 Correlational effects and cumulative advance

Essentially, we do not have enough knowledge of the details of linguistic pro-

cessing to be able to interpret in a direct way what data like reaction time or
eye-tracking mean, either in terms of linguistic theory or of symbolic computa-
tion in the mind more generally. The interpretational gap problem however is not
fatal to progress. It just means that research needs to progress incrementally and
cumulatively by slowly amassing data which establishes clear correlational effects.
For example, if I know that semantic infelicity (as opposed to ungrammaticality)
reliably gives rise to slower reaction times in certain kinds of tasks, or to EEG
effects of a certain type at a particular time after data exposure, then I start to
build up a behavioural fingerprint for a certain natural class of phenomena. If
my own particular experimental paradigm throws up that very same behavioural
fingerprint, I am then justified in assigning the variable I was testing to that par-
ticular natural class of phenomena. All this is possible and remains an important
result, even if we later turn out to be entirely wrong about how we described or
‘explained’ the reasons for those effects at the beginning. We might even be wrong
about the way we characterized that natural class of phenomena theoretically, or
how we chose to label it. The point is not that we should be wishy-washy about
attempting to interpret, but rather that we should at the same time concentrate
on building up a body of correlational results that transcend interpretational fads,
overzealousness, and outright mistakes. This kind of methodology ensures that in
the long run, the sheer cumulation of correlational patterns will allow us to better
understand what is going on.
In the particular case I have been discussing in this short article, it would be
important to subject argument structure complexity effects to the kind of barrage
of different experiments that would isolate what natural classes of phenomena
they pattern with. Does a violation of a verb’s argument structure induce the same
kind of processing effect as a semantically ‘strange’ lexical choice? Or is it the
same as the complexity that arises from the attempt to resolve lexical ambiguity?
Or does it rather have the behavioural fingerprint of a syntactic violation? An
important question to ask is whether all types of argument structure variability
effects behave the same in this respect, or is the causative inchoative alternation
somehow special? These are very basic questions that we are in a position to try
and answer, and which are not dependent either on our theoretical beliefs or on
the ultimate interpretation of the behavioural effects we track.
 Creating a productive space for theory and experimentation 197

3.3 Mapping from externalized performance to brain activity

Most would admit that researchers are still in a pretty profound state of ignorance
about how the brain works and how different functions might be localized. But
once again this does not mean that useful information cannot already be gleaned.
Neurolinguistically, there are important methodologies like EEG that measure
brain potentials and can track effects with good time discrimination in response
to linguistic stimulus. fMRI on the other hand gives good localization data for
activity within a human brain, but has limited temporal resolution. In the field
today these methodologies are used in tandem to build up a complete picture of
brain response. Neurolinguistics as a whole also interacts closely with psycholin-
guistics in a productive way. There are many different types of psycholinguistic
experiment that are done, which probably involve many and different ancillary
brain functionality depending on the task. Some involve reaction time measure-
ments to different kinds of tasks, others use eye-tracking, or even pure judge-
ments. It is feasible to think that as we build up more and more information about
the behavioural/performance correlates of brain activity that we will start to get
firmer generalizations that map from a particular type of externalized response to
a linguistic phenomenon to information about what part of the brain is activated
during that response.
Eventually the hope is that stepwise methodologies can thus in principle even-
tually allow us to map formal representations of the linguistic symbolic system at
work to actual brain profiling. To bridge the gap between linguistic modelling
and actual brain activity, we need to construct experiments that link linguistic
phenomena to replicable and measurable behavioural effects, and then in turn,
we need to learn how and if those measurable behavioural effects map onto brain
activity. Crucially, testable behavioural properties are vital for establishing reliable
links eventually between linguistic symbolic facts and the brain.
Once again, systematic and replicable correlations are valuable even in the
absence of the complete knowledge that we would require to interpret the cor-
relations in an absolute sense. Building up a reliable picture depends crucially on
linguists, psycholinguists and neurolinguists using the same basic granularity of
variables to be tested, replicating and building on each other’s results to establish
new correlational generalizations.

4. Conclusion

I started this short paper by raising the possibility that experimental method-
ologies could begin to adjudicate important questions in a theoretical debate
198 Gillian Ramchand

over points of linguistic analysis. In the particular test case I was hopeful about,
I found that there was a huge interpretational gap between actual experimental
results and their implications for theoretical representation. However, diagnosing
the problem showed that it was not a principled one, but rather one that existed
because of a lack of agreement across disciplines about the specificity and the
granularity of the hypotheses to be tested. If the psycholinguist tests hypotheses
that make no use of the primes uncovered by theoretical investigation, then they
will not produce any results that are interpretable for that theory. On the other
hand, if the hypotheses are too specific and theory-internal then the results are
subject to massive interpretational indeterminacies that render the results of the
experimentation difficult to interpret once the terms of the debate change (which
they inevitably do). In the case of the problem of unaccusative verbs, I suggested a
particular set of testable hypotheses that I thought might be a useful place to start.
I argued that theoretical progress in the domain of argument structure has
importantly changed the way in which the terms of the basic hypotheses should
be set up (see specifically Marantz 2013). On the other hand, the data from un-
conscious behaviours and brain activity (as opposed to conscious behaviours like
judgements) can no longer be ignored as a potentially rich source of information
about the structure of the grammar.
Creating a productive multidisciplinary space is therefore not a luxury but a
necessity, if the field is to advance. In Section 3, I spelled out the common sense
logic behind such a multidisciplinary space. The details and the practice however
will have to emerge from actual collaborations, both within institutions but also
across them in the form of academic fora and workshops where research results
and methodologies are shared.

Note

1. In the simplified representations above, we abstract away from the difference between
‘complete’ and ‘incomplete’ CHANGE (telicity). This is an important, and possibly completely
independent factor which would need to be investigated independently.

References

Altmann, Gerry T. 1997. The Ascent of Babel. Oxford University Press.

Bard, E., C. Frenck-Mestre, and A. Sorace. 2010. Processing auxiliary selection with italian
intransitive verbs. Linguistics 48 2: 325–361. DOI: 10.1515/ling.2010.011
 Creating a productive space for theory and experimentation 199

Borer, Hagit. 2005. Structuring Sense: An Exo-Skeletal Trilogy. Oxford University Press, New
York.
Brennan, Jonathan and Liina Pylkkanen. 2010. Processing psych verbs: Behavioural and MEG
measures of two different types of semantic complexity. Language and Cognitive Processes
25 6: 777–807. DOI: 10.1080/01690961003616840
Butt, Miriam. 2003. The morpheme that wouldn’t go away. Handout on Urdu causatives. Ms.,
University of Konstanz (ling.uni-konstanz.de/pages/home/butt).
Chierchia, Gennaro. 2004. A semantics for unaccusatives and its syntactic consequences. In
The Unaccusativity Puzzle: Explorations of the Syntax-Lexicon Interface, edited by Artemis
Alexiadou, Elena Anagnostopoulou, and Martin Everaert, pp. 22–59. Oxford University
Press, Oxford. (first circulated in 1989). DOI: 10.1093/acprof:oso/9780199257652.003.0002
Croft, William. 1998. Event structure in argument linking. In The Projection of Arguments, edit-
ed by Miriam Butt and Wilhelm Geuder. CSLI Publications, Stanford, CA.
Folli, Raffaella. 2001. On the relation of priority between causative and inchoative construc-
tions. In Romance Languages and Modern Linguistic Theory 1999, edited by Johan Rooryck
Yves D’Hulst and Jan Schroten, pp. 143–166. John Benjamins.
Gennari, Silvia and David Poeppel. 2003. Processing correlates of lexical semantic complexity.
Cognition 89 2: 27–41. DOI: 10.1016/S0010-0277(03)00069-6
Haspelmath, Martin. 1993. More on the typology of inchoative/causative verb alternations. In
Causatives and Transitivity, edited by Bernard Comrie and Maria Polinsky, pp. 87–120.
John Benjamins, Amsterdam/Philadelphia.
Husband, E. M., L. Stockall, and A. Beretta. 2008. Vp-internal event composition: Processing
evidence for phrase-level event interpretation. Ms. Michigan State University, Hampshire
College.
Kim, Mikyong and Cynthia Thompson. 2004. Verb deficits in alzheimer’s disease and agram-
matism: Implications for lexical organization. Brain and Language 88 1: 1–20.
DOI: 10.1016/S0093-934X(03)00147-0
Levin, Beth and Malka Rappaport Hovav. 2005. Argument Realization. Research Surveys in Lin-
guistics. Cambridge University Press, Cambridge. DOI: 10.1017/CBO9780511610479
Levin, Beth and Malka Rappaport Hovav. 1995. Unaccusativity: At the Syntax-Lexical Semantics
Interface. MIT Press, Cambridge, MA.
Marantz, Alec. 2013. Verbal argument structure: Events and participants. Lingua 130: 152–168.
DOI: 10.1016/j.lingua.2012.10.012
McKoon, G. and T. Macfarland. 2000. Externally and internally caused change of state verbs.
Language 76: 833–858. DOI: 10.2307/417201
McKoon, G. and T. Macfarland. 2002. Event templates in the lexical representations of verbs.
Cognitive Psychology 45: 1–44. DOI: 10.1016/S0010-0285(02)00004-X
Parsons, Terence. 1990. Events in the Semantics of English: A Study in Subatomic Semantics. MIT
Press, Cambridge, MA.
Pinker, Steven. 1999. Words and Rules. New York, Basic Books.
Pustejovsky, James. 1995. The Generative Lexicon. MIT Press, Cambridge, MA.
Ramchand, Gillian. 2008. Verb Meaning and the Lexicon. Cambridge University Press.
DOI: 10.1017/CBO9780511486319
Ramchand, Gillian. 2013. Argument structure and argument structure alternations. In Cam-
bridge Handbook of Generative Syntax, edited by Marcel den Dikken, pp. 265–321.
Cambridge University Press.
200 Gillian Ramchand

Reinhart, T. 2002. The theta system – an overview. Theoretical Linguistics 28 3.

Thompson, Cynthia. 2003. Unaccusative verb production in agrammatic aphasia: A syntactic
account of verb production deficits. Journal of Neurolinguistics 16: 151–167.
DOI: 10.1016/S0911-6044(02)00014-3
Thompson, Cynthia and M. Lee. 2009. Psych verb production and comprehension in agram-
matic broca’s aphasia. Journal of Neurolinguistics 22 4: 354–369.
DOI: 10.1016/j.jneuroling.2008.11.003
 Language index

A F I
Argentinian Sign Language French 27–28, 30 Italian 8, 10, 36, 40, 46, 83–85,
(LSA) 53–54 87, 91–94, 96–97, 99, 107–109,
Austrian Sign Language (ÖGS) G 111–112, 122, 146, 186, 191
56, 60 German 62, 66–68, 76, 84, 87, Italian Sign Language (LIS)
91, 93, 109, 111–112, 146, 155 46–50, 58
B German Sign Language (DGS)
Basque 30 47, 49–50 J
Greek 11, 119, 121–127, 129, Japanese 91, 93
C 131–135
Catalan 27, 30, 35, 38–40 R
Catalan Sign Language (LSC) H Romanian 11, 119, 121–127, 129,
47, 50, 53–54, 56 Hebrew 121, 133–134, 153, 161, 131–135
177
E Hong Kong Sign Language 54 S
English 4, 8–9, 11, 24–25, 32, 35, Spanish 12, 38–40, 57, 84, 122
38, 62, 65, 76, 84–85, 108–109,
119–121, 123, 131–132, 134, 145,
161, 186–190, 192–194
 Subject index

A argument structure complexity classifier predicates (sign

accomplishment (verbs) 24 9–10, 12–14, 76, 92, 142, 145, language) 5, 45–50, 52–58
activity (verbs) 31–32, 36, 40, 147, 149, 153–154, 157, 189–192, coercion 26, 56, 68–69
66 196 co-event hypothesis 63–69,
adjectives 26–28, 120 argument structure information 71–73, 77
adjuncts 23–25, 27–28, 34–38, 14, 142–146, 155, 157, 159–160 cognate objects 40
48, 72, 178 atelic(ity) 36–37, 56, 64, 190 complements 15, 23–27, 29–38,
adjunction 33, 36, 194 see also telicity 40, 86, 89, 151, 172, 174
agents 3, 6–11, 52–53, 61–64, auxiliary selection 36 complex predicates 33, 38,
70, 72, 83, 91, 109, 119, 124, 58, 74
128, 130, 135, 141, 149, 152, 188 B compound verbs/verbal
agrammatic aphasia (speakers, BA 45 (Broadmann area) 148, compounds 4, 24–25, 32–33,
subjects, patients…) 153 38, 40, 54
109, 144–147, 152–155, see also frontal gyrus conflation 40, 74
189–191 BA 9 (Broadmann area) 147 construction/constructionism/
see also aphasia see also frontal gyrus constructionist/constructivist
agreement birthing verbs 29, 36 4–6, 16, 73, 143–144, 161
backward – 55 body part classifiers (sign corpus/corpora 5, 65, 76
– morphology 46–47 language) 52 cross-linguistic variation 35–
locative – 48, 56 boundedness 36–37 36, 38, 56
person – 47, 50, 56–57, 85 Broca’s area 153, 162 cross-modal lexical decision
sign language 46 see also frontal gyrus task 65, 147, 172, 189–190
alternations (argument
C
strcture) 2, 7–8, 15, 52 D
case (grammatical) 10, 39,
see also causative alternation, dative alternation 62, 69, 121
85–87, 89–90, 92, 112, 150
dative alternation, hand deadjectival verbs 35
causatives/causation/causers
shape alternations, denominal verbs 4, 24–25, 29,
2–3, 9, 15, 17, 58, 85, 119,
psychological verb 35, 120
123, 125–131, 133–134,
alternation, unegrative/ Distributed Morphology (DM)
186–191
unaccusative alternation 3, 5, 34, 38, 46, 50
(non-)agentive causers 119,
angular gyrus 146–148, ditransitive verbs/predicates
121–123, 125–128, 130–134,
156–160 51, 55, 62, 69, 95, 145, 147,
195
anterior cingulate 162 156, 157
anticausative structure 11,
anterior negativity (ERP) 10,
119, 121, 123–126, 128,
68, 106–108, 111 E
131–132, 134
aphasia 12, 109, 146, 171–172, encyclopaedia 31–32
causative alternations
180 entity classifier (sign language)
anti causative 123–125,
Broca’s – 156, 172 49, 52–54, 58
127–129, 131
Wernicke’s – 146, 160 event(ive) 2–9, 11, 24–25,
causative-inchoative 186–
see also agrammatic aphasia 27–29, 31, 39, 48–49,
188, 190–192, 194, 196
argument sharing 61, 63–64, 54–56, 61–77, 120–128, 130,
change-of-state predicates 26,
67–68, 70, 72, 75–76 133, 135, 144, 146, 150, 156,
29
204 Structuring the Argument

161, 169, 178, 185, 187–189, inner aspect 24, 37–38 N

191, 193–194 internal argument 2, 7–8, N400 (ERP) 10, 68–69, 73, 83,
structure 4–5, 9, 27 61–62, 52–53, 87, 150, 176–177, 179, 94, 104, 107–110, 155
64, 75, 144, 185, 187–189, 190 neuroimaging 11, 13–15, 142,
191, 194 intransitives 8, 13, 15–16, 24–25, 144, 148, 155–156, 158, 169–
Event-Related Potentials (ERPs) 36, 52–53, 55, 58, 128, 143, 170, 172–175, 178–180
10, 13, 67–69, 73, 76, 85, 145–151, 154, 157–158, 162, 170, neurolinguistics 5, 14–15, 38,
93–94, 99–100, 102–106, 109, 176, 186–188, 190–191 61, 84, 87, 144, 161, 171–172,
144, 154–155 180, 185, 197
Event Visibility Hypothesis 55 L non(-) canonical/non-canonicity
expectation 10, 92–94, 96, 107, late positivity (ERP) 10, 83, 94, 1, 7, 9, 13, 61, 63, 69, 72, 150–
109–111 103, 105, 107–108, 110 153, 157, 159, 172, 191–192, 195
experiencers 2, 9–12, 64, lexical ambiguity 13–14, 148,
83–91, 93–112, 119, 121, 126, 175, 192, 196 O
128, 131, 161, 189, 191 lexical complexity 159, 171, object experiencers 2, 9–12,
external arguments 2, 4, 7, 8, 173–174, 188, 189, 190, 191 83–88, 90–91, 93–98, 104–105,
11,52–53, 85–88, 112, 119–120, lexical decision (task) 65, 68, 107–112, 119, 189, 191
123, 126, 128, 130, 135, 176–177, 71, 143, 146–149, 172, 175,
179 188–190 P
eye tracking 152, 196, 197 lexicalist/lexicalism 2, 6–7, 16, P600 (ERP) 68, 77, 155
45, 143, 158, 161, 187, 193 Parallel architecture 71–72, 76
F lexical subordination 35 passive 2, 6, 8–9, 11–12, 108–
figure (~ground) 27 lexicon 2–5, 38, 45–46, 55, 57, 109, 119, 121, 128, 130, 132–135,
frontal gyrus 65, 74, 77, 131–132, 143–144, 175, 191
inferior (IFG) 14–15, 170, 176 path 27–28, 37, 45, 47, 50–51,
147–148, 153–154, 156, light verbs/light verb 54–56
158–160 constructions 6, 12, 32, 39, patients 3, 6–8, 63–64, 70, 72,
middle 147–148, 156, 158 61–77, 80, 191 83–84, 144, 150, 152
Functional Magnetic Resonance location/locatum verbs 30, 35 posterior perisylvian regions
Imaging (fMRI) 12–15, 87, 147–148, 156, 162
144, 146, 157, 169, 172, 176, 197 M precuneus 162, 172
Magnetoencephalography prepositions 23, 27–28, 30, 35,
G (MEG) 66, 169, 189 87, 112, 123, 125–127
Government & Binding 86, 150 mapping/linking 2, 4–10, 46– priming 69–71, 75–76, 97,
handling classifiers (sign 51, 55, 61, 63–64, 66, 68–70, 169, 188
langugae) 52–54, 58 72–77, 84, 86–87, 91–92, 94, psych nominalizations 119–121,
hand shape alternations (sign 107–109, 126, 149–154, 157, 125, 131, 133–134
language) 49–52, 55, 58 162, 169, 172, 176,191–192, 197 psycholinguistics 15–16, 38, 65,
Head-driven Phrase Structure Merge 4, 23–24, 27–29, 31–32, 71, 77, 84–85, 89, 91, 108, 112,
Grammar (HPSG) 151 34, 36–37, 42, 51, 88 142–144, 161, 171, 173, 180, 185,
head movement 75–76 minimalist framework/ 188–189, 193–195, 197
hierarchy (semantic/thematic) minimalist syntax/ psychological/psych verbs 2,
3, 7–8, 56, 61–65, 68, 70–74, minimalism 3, 24, 34, 38, 9–12, 112, 119–124, 126–128,
77, 83–86, 92–94, 107–111, 113, 73, 87 132–134, 170, 189
126, 141, 143, 145–149, 154, morphology/morpheme 3, – alternations 123–125,
156–158, 162, 170, 179 5, 25, 33, 45, 85, 120–122, 127–129, 131
124–125, 129, 134–135, 142, 177, piacere-class 10, 83–93,
I 186–188, 191–192, 195 95–99, 108–109, 112
incorporation 40, 57–58 movement 10, 12, 14–15, 75–76, phrase structure 86–90
information structure 10, 76, 112, 150–153, 157, 173, 176–177, – processing 83–85, 87, 99,
85, 87–91, 105, 108, 111 179, 190 108–111
 Subject index 205

R T U
reflexives/reflexive verbs 15, 95, telic/telicity/atelic/atelicity 9, unaccusative/unacusativity 5,
122, 124, 129, 171, 176–179, 36–37, 56, 64, 77, 135, 190, 198 8–9, 12–16, 25, 27–28, 40 45,
186 temporal gyrus 14, 66, 146– 52, 55, 85–86, 145, 149–154,
– processing 178–179 148, 153, 155–161, 197 159, 162, 166–167, 170, 173,
re-prefixation 4, 24–27, 29, superior temporal sulcus 176–180, 185, 189–195, 198
37, 39 147, 157 unergative/unaccusative
representations/representational posterior middle temporal alternation 5, 8, 45, 52, 55,
complexity 1–2, 4–5, 7, gyrus 148, 153, 156–157, 145, 194
12–14, 34, 54, 61, 64–65, 69, 159–160 unergative/unergativity 5, 8,
71, 75–76, 85, 143–144, 146, posterior superior temporal 13–14, 24–26, 29–31, 34–36,
158, 169, 173–174, 185–191, 193, gyrus 147, 155–157, 38, 45, 52–53, 55, 113, 145,
197–198 159–160 149–153, 173, 176–178, 188,
resultatives 25–27, 33, 37 thematic hierarchy 190–192, 194–195
Role and Reference Grammar see hierarchy
150–151 thematic reanalysis 87, 93, 107, V
roots 2–5, 8, 23–40, 49, 54, 58, 109, 113–114 verb-argument integration 148,
71–73, 76, 120, 128, 186–187, thematic/theta roles 3, 7–8, 14, 155, 157–161
194 65, 63–64, 70–72, 77, 83–85, verb naming tasks 145–146, 153
87, 91, 93–94, 108–109, 111, 119, verb production 145, 148, 153
S 126, 141, 143, 145–151, 153–154, violations
sentence comprehension models 156–158, 162 thematic – 108
142–144, 171 theme 3, 7–8, 10, 29, 31, 53, linking – 109
sentence production models 57–58, 62–64, 70, 83–84, argument structure – 145,
142–144 86–88, 91, 93, 141, 149–150, 154–156
serial verbs 54 152–153, 162 vP 23–24, 27, 31–34, 37–38,
small clauses 4, 29, 32–33, 38 incremental – 31 87–90, 128–130
Sole Complement transitive structure 11, 13, 15,
Generalisation 25 28, 32, 34, 52–53, 62, 69, 85, W
spatial predicates 46, 48–50 119, 121, 128, 130, 132, 135, 143, word order 10, 67, 69, 85,
specifiers 28, 32–34, 88, 129 145–149, 153, 157, 161, 187, 87, 91–94, 96–98, 102–103,
subcategorization 10, 62, 97, 191–192 107–109, 111
141, 143, 145–148, 158, 161, 170, transitive verb 8–9, 13, 15–16,
173–175, 182 24–28, 30, 32, 51–53, 55, 62,
subject experiencers 9, 11, 85, 95, 97, 119–121, 123, 125,
64, 84–86, 95, 109, 119–132, 128, 130, 123–135, 143, 145–149,
134–135, 189, 191 153–154, 157–158, 161–162, 170,
supramarginal (gyrus) 146, 176, 186–192, 195
148, 156–157, 159–160