VAGUE LANGUAGE:

SOME VAGUE EXPRESSIONS IN ENGLISH

by

Joanna Mary Channell

Thesis presented in partial fulfilment

of the requirements for the degree of

Doctor of Philosophy
in the Department of Language
University of York

January 1983
 Contents

Chapter 1 Introduction 1
.... .... ..

1.1 The topic 1

.... .... "" ..
1.2 The rationale 1
.. "" "" "" "" "" ""
1.3 The results 6
." "" "" "" "" ""
1.3.1 Status of the results .... ."6
1.3.2 Psychological 7
reality .... ....
1.4 General 12
theoretical approach .... ..
1.4.1 Idealisation 12
.... .... ..
1.4.2 A note "meaning" 13
on .... ..
1.4.3 Grammar 16
.... .... ..
1.5 Methodology 23
"" "" "" "" ""
1.5.2 The data 27
.... .... ..
1.5.3 Transcription of-recorded data 28

1.5.4 Elicitation tests 28

. ". ". ..
1.6 the thesis 30
Organisation of .... ..
31
Chapter 2 Vagueness and Language .... .... ..
2.1 Vagueness 31
vs. ambiguity .... .... ..
2.2 Previous 36
work on vagueness .... ....
2.3 Vagueness 48
and communication .... ....
2.4 Different kinds of vagueness 51
..... ....
2.4.1 Vague additives 51
.... ....
2.4.2 Vague words 53
.... .... ....
2.4.3 Vagueness by omission 55
". .... ..
2.4.4 Vagueness by implicature 58
....
2.5 Summary 61
.... ....

Chapter 3 Number Approximations 62

.... .... ..
3.1 Introduction 62
.... .... .. ."
3.2 Conversation 64
examples .... .... ..
3.3 The elicitation test 65
.... ....
65
3.3.1 Objectives .... .... ..
3.3.2 Method 65
.... .... ..
3.3.3 Results 67
.... .... ....
3.3.4 Discussion 67
.... .. ""
3.4 Number approximations 69
.... .... ..
3.4.1 69
About/around/round .. "" ""
3.4.2 Approximately 71
". "" "" "" ""
3.4.3 75
norm .... .... ....
 3.4.4 n or so 82
.... .... ....
3.5 Combinations of approximators 85
.... ..
3.6 Numbers in approximations 87
.... ....
3.6.1 Certain numbers are favoured 87
....
3.6.2 "Round" numbers = reference point numbers 88

3.6.3 Reference point vs. non-reference

point approximations 93
.... ....
3.7 Implicatures and entailments of approximator
+n structures 95
.... .... ..
3.8 A note on partial specifiers 97
.... ....
3.9 Summary 101
.... .... .... ..
Chapter 4 Vague Category Identifiers 103
.... ....
4.1 Introduction 103
.... .... .... ..
4.2 The elicitation test 106
.... .... ..
4.2.1 Objectives 106
.... ..
4.2.2 Method 106
.... .... ....
4.2.3 Subjects 107
.... .... ..
4.2.4 General approach to the data ....
108

4.3 Results and discussion 113

.... ....
4.3.1 "Rosch" type test items 113
..
4.3.2 Attested test items 118
.... ....
4.3.3 General conclusions from the results 122
..
4.4 The structure of the tags 123
.... ....
4.5 Structure of exemplar + tag constructions 124
..
4.5.1 NP + tag 125
.... .... ..
4.5.2 VP + tag 126
.... ....
4.5.3 Adverbial + tag 126
.... .... ..
4.6 Intonation and vague tagging 126
.... ....
4.7 Co-occurrence restrictions 132
.... ....
4.8 Summary 137
.... .... .... ....

Chapter 5 Being Vague 138

.... .... .... ..
5.1 Introduction 138
.... .... .... ..
5.2 Analytical approach 138
.... ....
5.3 Three scenarios for vagueness 139
.... ..
5.3.1 "Eighty or so pence" 139
.... ....
5.3.2 "Elaborated code and things like that.. 141

5.3.3 Humorous effects 145

5.4 147
.... .... .... ....
5.4.1 Giving the right amount of
information ..
147
.... ....
5.4.2 Withholding 150
.... .... ..
 5.4.3 Saying what you don't know
how to say 151
.... .... ....
5.4.4 Talking about things you aren't sure
about .... .... ....
155
..
5.4.5 Self-protection 157
.... .... ..
5.4.6 Deference 159
.... .... ..
5.4.7 Informality and atmosphere 162
.... ..
5.4.8 Women's language 163
.... .... ..
5.5 Conclusions 165
.... .... .... ..
Chapter 6 Lexical Aspects 167
.... .... ....
6.1 Introduction 167
.... .... ....
6.2 Lexical descriptions 167
.... .... ..
6.2.1 Or 167
.... .... ....
6.2.2 And :... 170
.... .... ..
6.6.2.3 So 171
.... .... .... ..
6.2.4 Something 172
.... .... ....
6.2.5 Like 172
.... .... .... ..
6.2.6 Things 173
.... .... ....
6.3 Idiomaticity 173
.... .... ....
6.4 Lexical meaning and phrase meaning 177
....
6.4.1 n or m .... 177
.... .... ..
6.4.2 n or so 179
.... .... ....
6.4.3 Or something 180
.... .... ..
6.4.4 And things 183
.... .... ....
6.4.5 And that 184
.... .... ....
6.5 The single word number approximators 185
....
6.5.1 About 186
.... .... .... ..
6.5.2 Around/round 186
.... ....
6.5.3 Approximately 186
....
6.5.4 Discussion 187
.... .... ....
6.6 Conclusions 188
.... .... ....
Chapter 7 Semantics, Pragmatics and Vague Meanings 189
..
7.1 Approach 189
.... .... .... ....
7.2 Vague meanings 189
....
7.2.1 Observed meanings 189
.... ....
7.2.2 A unified approach 191
....
7.3 Semantics and Pragmatics 196
.... ....
7.3.1 Literal meaning 196
.... .... ..
7.3.2 Bivalent truth-conditional semantics 202
..
 7.3.3 Many-valued truth-conditional
semantics 210
.... .... ..
7.3.4 Cognitive/referential vs emotive/non-
referential meaning 213
.... ....
7.3.5 Normative vs. descriptive 213
7.4 Meaning, semantics and pragmatics 214
.... ..
Chapter 8 Implications for the Study of Meaning 215
....
8.1 Summary of: findings 215
.... .... ..
8.2 Psycholinguistic/psychological implications 216
..
8.3 Consequences for linguistic theory 217
....
8.4 An approach to meaning 217
.... .... ..
Appendix 221
.... .... .... ....

References .... .... .... .. 223

Tables

The tables of data relevant to Chapters 3 and 4 are interleaved

between the coloured sheets at the end of these chapters

Table 1 Notational Conventions used in the Transcriptinn

of Conversation Extracts Y
.... ..
Table 2 Conventions used for Transcription of Intonation vi
Table 3.1 List of items used in number approximations
elicitation test foil. 102
.... .... ....
Table 3.2 Number approximation test: general results 102
..
Table 3.3 Frequency distribution: about/around 102
....
Table 3.4 Special results for n or m 102
.... ....
Table 3.5 Frequency distribution: n or so 102
.... ..
Table 4.1 Stimulus items: tag informant test 137
....
Table'4.2 Tag informant test: summary of results 137
..
Table 4.3 Null responses 137
.... .... ....
Table 4.4 Rosch-type test items, comparison 137
....
Table 4.5 Rosch-type test items, type 2 responses 137
..
Table 4.6 Attested test items: results 137
.... ..

Figures

Figure 1 89
.... .... .... ..
Figure 2 89
.... .... .... .... ..
Figure 3 The frequency of the forms one to twenty 92
."
Figure 4 102
.... .... .... .... ..
Figure 5 Vichy advertisement 139
Figure 6 Doonesbury cartoons 145
.... .... ..
 Acknowledgements

I am grateful to very many people in many different ways for

their help with this thesis. Firstly, Patrick Griffiths supervised

the research, and the writing of the thesis. He gave generously of

his time and of his ideas, in many talks over the last four years.

Discussions, or correspondence, with Jens Allwood, A. P. Cowie,

D. A. Cruse, Elizabeth Dines, R. B. Le Page, Michael Lumsden, and

Yorick Wilkes, were important in shaping the development of my

views on vague language. Tom Wachtel's work on number approximations

provided inspiration at a crucial time, and our talks and

correspondence have been uniquely valuable, even though we hardly

ever agree.

Particular people helped with specific aspects of the work.

John Local commented on the transcription of intonation in Chapters 3

and 4; David Pimm and Rolph Schwartzenburger tried to give me a

more mathematical perception of number; Thomas Baldwin was kind

enough to comment on sections of Chapters 3 and 7. John Illingworth

the University of York Computing Service,

of advised me on data

handling, and adapted a sorting program which sorted the data in

Chapter 4.

A linguist must owe a special debt to those who agree to have

their speech used as data, and to those who act as subjects in

informant tests. In this connection I thank undergraduate and

graduate students of the Department of Language and the Department of

Psychology of the University of York, 1978-79 and 1979-80; the

University of York Graduate Students' Association; students of the

Department of French, College of Ripon and York, St John; and Mr

Arthur Harrison of Nunthorpe Grammar School, and the Sixth Form,

1979-80. Marion Owen kindly from her corpus

provided me with examples

of data. In knowing my interest,

addition, many friends, of

i
collected written and spoken examples of vague language for me.

Two years of research were supported by a Department of

Education and Science Major Award. Part-time work at the College

of Ripon and York, St John, and the Linguistics Department of

the University of Hull, enabled me to survive while continuing

work on the thesis, and I am grateful to both institutions on

that account.

The thesis was prepared using the Queensland version of

RUNOFF on the University of York DEC-10 Computer. I would like to

thank Suky Thompson and Ulla Wiberg for their help with formatting

and editing, and all the Computing Service staff for their

friendliness, and patience with my often unreasonable demands.

Muriel Pirozek typed the references for me. Peter Monk read the

whole draft, and tried to simplify and clarify my convoluted style.

Where I refused his advice, it will be my readers' loss. I am

grateful to him for his patience, and, as always, for his

unswerving support.

Joanna Channell

York, January 1983

The cartoons in Chapter 5 are reproduced from The People's

Doonesbury: Notes from Underfoot, 1978-81, by G. B. Trudeau,

published 1981 by Wildwood House, London

ii
 Declaration

Parts of Chapters 3 and 5 have appeared in a slightly

different form as a paper: 'More on Approximations: a reply

to Wachtel', in the Journal of Pragmatics 4 (1980).

111
 Abstract

There is evidence in the literature which suggests that the

linguistic analysis of vague language poses problems. In particular,

some vague expressions create difficulties as to which aspects of

their meaning should be handled by semantics, and which by pragmatics.

In this thesis, the-general-phenomenon of vague language is

investigated through a detailed examination of-two types of vague

expressions: number approximations (eg about ten), and vague category

identifiers (eg a film or something). The results of informant tests,

designed to discover how these vague expressions are understood, show

that hearers-interpret them as identifying fuzzy sets whose membership

is defined by (a) the form and content of the vague expression; (b)

linguistic context and situation of utterance; and. (c) world knowledge.

Informants a high degree of agreement in their responses.

exhibit

Conversation data (and some written data) froma variety of

sources are analysed in order to ascertain what are the conversational

effects arising from the use of vague language, and what interactional

problems are solved by its use. Intonational characteristics, lexical

aspects; and idiomaticity are also covered.

The subsequent theoretical discussion confronts various existing

proposals for drawing the distinction between semantics and pragmatics

with these data, and finds no valid ways of establishing the

distinction. In the concluding chapter an alternative strategy for

the study of meaning is put forward in programmatic outline.

iv
 Table 1

Table 1

Notational Conventions used in the Transcription

of Conversation Extracts

+ attested example

A: B: stands for different speakers

(laughs) brackets enclose description of non-linguistic

communication

(. ) untimed short pause

[ indicates speech or vocal noise concurrent with that

above

// encloses interjected utterance from a

different speaker

ýýý indecipherable speech

/ marked pitch deviation, upwards/downwards

if precedes emphatically stressed syllable

[I enclose transcriber/situational comment

any prosodic features heard as delineating a

sense group

I have used apostrophes in don't, can't, etc, simply because it seems

unnecessarily upsetting to miss them out. They are, however, omitted,

where intonation transcription is included, because they would

interfere with reading the intonation symbols. I have not used

capital letters at the beginning of turns, but in individual short

examples I have used them, again because it seems typographicallly

upsetting not to.

V
 Table 2

Conventions used for transcription of Intonation

onset of tone unit

ýý
end of tone unit

stressed syllable

heavy stress

fall

rise

rise-fall

fall-rise

(. ) untimed pause

sudden pitch change for beginning of next

segment

V1
 Chapter 1

Chapter 1

Introduction

1.1 The Topic

"Some of the most interesting questions ire raised by the

study of words whose job it is to make things fuzzier or less
fuzzy" (Lakoff, 1972: 195)

The general subject of this thesis is vagueness in the use of

language. There are many ways in which speakers can be vague: they

can employ "hedges" ( in the sense of Lakoff, 1972) such as virtually

or it seems that, or they can use agentless passives (Fowler and

Kress, 1979: 30ff) to give just two examples. What are the linguistic

and non-linguistic mechanisms which enable speakers to be. vague, and

which enable hearers to understand them when they are? Why and when

are speakers vague? These are some of the questions which will be

discussed.

The present work contributes to the study of vague language use

through an investigation of two types of vague expressions. The

expressions studied in detail are:

Type A: Number Approximations

(n and m= members of the set of Real Numbers)

about n
approximately n
(a)round n
n or so
norm

Type B: 'Tag' Approximations

(X = typically NP or VP)

X and things like that

X and things
X and that
X or anything like that
X or anything
X or something like that
X or something
 Chapter 1

In Chapter 2a classification of different types of vagueness is

presented - and other vague expressions are introduced and described

at various points where they are relevant.

The pretheoretical notion of vagueness will be refined and

defined throughout the thesis, but a general working definition would

be that a vague utterance is one which cannot be assigned an exact (or

in the case of ambiguity being present, several discrete but exact)

meaning(s), even with recourse to context. One example to suggest the

general notion:

[Context gloss: discussion of academics who publish many articles]

(1) +B: and they repeat themselves each time. You find that
...
you get five or six articles and they're really
very much the same [11.21.2]

The hearer of (1), even with recourse to all he[l] knows of the

speaker's knowledge, and likely intentions, to the discourse

situation, and any other relevant or irrelevant pragmatic factors,

cannot know from the information provided whether he should take the

reading of five or six articles to be "exactly five" or "exactly six"

or exactly any other number or interval of numbers. Importantly,

however, this apparent vagueness causes no difficulty to the hearer or

to the progress of the conversation.

1.2 The Rationale

"What' then is
a reasonable field of endeavour for,
linguistics? We would claim that the study of meaning is
vitally important but that meaning must be studied in a new
light, namely with respect to the actual usage of speakers. "
[Schank and Wilks, 1974: 312]

1 Where-there is no antecedent marked for he,

which is clearly gender,
him and his should be read as referring to people of either sex.

2
 Chapter 1

Recent linguistic work on "meaning" has tended to broaden its scope in

the way suggested by Schank and Wilks, in order to look at more

non-linguistic (pragmatic) aspects of meaning; and the interface

between semantics and pragmatics especially has received and is

continuing to receive a good deal of attention (typified for example

by Wilson, 1975, Wilson and Sperber, 1982; Gazdar, 1979, Levinson,

1979; Allwood, 1981).

_

The vague expressions investigated here are particularly relevant

to this area because they seem to be more inherently 'pragmatic' in

nature than many others, indeed it has been suggested that some of the

expressions are almost semantically empty and that their

interpretation has to be handled entirely within pragmatics (Sadock,

1977). However we evaluate this particular view, what is clear is

that these expressions can probably not be handled satisfactorily

except by looking at both semantics and pragmatics.

There is evidence in the literature to suggest that the

linguistic analysis of vague expressions poses problems (Lakoff, 1972;

Sadock, 1977; Wachtel, 1980,1981; Channell, 1980 Fauconnier, 1976,

Klein, 1980,1982; Danell, 1978). In particular, some vague

expressions create difficulties as to which aspects of their meaning

should be handled by semantics and which by pragmatics. Partial

theoretical analyses have been proposed for a few expressions, but

remain to be substantiated. There is as yet no major study, of

linguistic vagueness, and no generally agreed approach to, it. This

study of vague language is intended to be on the one hand

theoretically relevant in contributing to the continuing debate on the

semantics /pragmatics interface, descriptively useful

and on the other,

in making available body data concerned with a hitherto

a of

3
 Chapter 1

underinvestigated area.

Such work as there has been on vague language(see 2.2) has been

more theoretical than descriptive (these terms are distinguished as

described in Lyons 1977: 138), in the sense that it uses only limited

numbers of (often unattested). examples to substantiate particular

theoretical analyses. The work reported here begins to right the

imbalance between theoretical and descriptive work by proceeding in

the opposite direction. It is very largely descriptive, and moves

towards theoretical analyses through observation of a body of attested

data different sorts (see 1.5 Methodology). The expressions

of

studied, listed at 1.1, do not exhaustively cover even the narrow

field to which they belong. They are however seen as representative

in that the problems they raise for theory are essentially the same as

those raised by other vague expressions. The decision to study this

particular set of expressions, and not others, arose from a wish to

study only expressions which were attested from conversation or

written data. Random data, even in very large quantities, cannot be

expected to produce examples of all the phenomena the investigator has

identified as possibly related. His two choices then (given practical

limitations on his data collecting resources) are to make use of

1-1 unattested data, or to restrict his investigation. For theoretical

reasons which will be discussed in greater detail in Section 1.5 the

second alternative was preferred [2].

2 Nevertheless it was necessary, especially in discussing syntactic

restrictions, to invented data. In addition,
make use of some
examples drawn from the work of other authors are usually unattested,
or no information on their provenance is given.

4
 Chapter 1

Although the discusses only a restricted set of data, its

work

firm links with an existing body of theoretical work enable it to

avoid the pitfall noted by Teeter (1964):

methodological

"If we are willing to restrict our range of enquiry

sufficiently, we will be able to find some answers easily,
but they will have little bearing on anything of
importance. "(: 205)

He goes on to describe what he sees as the correct relationship

between generalisation and particularisation, as follows:

"If insist on dealing with large questions, our first

we
attempts at answers are likely to be mistaken, or untestable
without a great deal of further work. But our answers to
questions are only worthy of the name of science when they
both and testable. Theory divorced from the
are significant
best knowledge is irrelevant, a castle in the air,
available
interest to science. But knowledge
and by that token of no
"fails to rise to the level of theory is of equal lack
which
interest, an accumulation of unconnected trivia. Science
of
lies between these two dangers. " (1964: 205-206)

According to a conventional model, it will be the task of two

sets of rules, semantic rules and pragmatic rules, to account for the

observed meanings of the vague expressions described in this thesis.

Perhaps however the main point which emerges from trying to see how

this would be done, is to call into question the semantics/pragmatics

distinction, for the practical reason that it does not appear to

the account. Allwood (1981) has argued coherently on both

elucidate

theoretical and practical grounds that the distinction is not helpful.

He does however map out an alternative approach which could be

not

followed. It cannot be the task of this thesis, whose primary aim is

descriptive, to develop an entire new theory of meaning. I shall

therefore' as follows. The theoretical discussion of meaning

proceed

will take the approach. The various

place within semantics/pragmatics

possible ways drawing the distinction, outlined by Allwood, will

of

each be considered, in to the data presented herein. This

relation

will that the different will be of interest and

mean analyses

5
 Chapter 1

accessible to adherents of any of the theoretical positions mentioned,

as well as, to those who may incline to a more ethnomethodological

perspective.

1.3 The Results

1.3.1 Status of the results

Given the present lack of an accepted linguistic theory of vagueness,

plus the lack of a complete and generally agreed psychological theory

of production/comprehension, it is not possible to give the empirical

work reported in this thesis the status of "experiments to confirm or

disconfirm the hypothesis involved" (Wirth, 1975). The work must be

seen principally as exploratory. Wirth contends that the logical

relation between theory and observed data is as follows:

Hg & He --> Psb

Where Hg is a grammatical hypothesis (with 'grammatical' being

understood in its broadest sense), He a theory of speech production,

and Psb a proposition describing some observable aspect of speech

behaviour. The partial analyses I Hgs,

shall present are and the

from data Psbs, but

observations are a series of with no definite and

widely accepted He available, the Hgs are neither confirmed nor

disconfirmed by the relevant Psbs, and neither can they be by relevant

experimental evidence presented by other workers.

The rationale of the present work has therefore been to seek out

and systematize a set of Psbs relevant to a particular sub-area of

linguistic enquiry and to use these to-confront existing approaches to

meaning in order to see whether they can deal with vague language in

an acceptable way.

6
 Chapter 1

In the light of this logic, all the work reported herein must be

seen only as a contribution to the continuing increase in our

scientific knowledge of how language behaviour takes place. It

represents one step into the whole largely uninvestigated area of

vague language use.

1.3.2 Psychological Reality

Clearly the business of engaging in linguistic behaviour- involves a

variety of cognitive skills, and both the nature of these skills and

the allocation of different activities to them are as yet not very

well understood. For the descriptive linguist, a necessary assumption

is the hypothetical division of cognitive skills into linguistic and

non-linguistic skills. The linguist's (as opposed to the

psychologist's) investigation of the linguistic skills usually makes

the further assumption that linguistic behaviour is the result of the

existence of a shared system or systems, the necessary elements of

which can be ascertained by studying linguistic data. The linguist's

task is construed as making a description of the elements of the

system(s). The, work presented here offers a description, valid for

speakers of Standard English (see below), of some elements which must

be present in the system(s) of English in order to account for vague

language behaviour.

To what extent can the theoretical descriptions presented here be

said to have "psychological reality"? I think it important to make

this clear since "psychological reality" is often brought to bear as

an evaluative criterion in regard to particular linguistic constructs.

In the case of semantics, it has even been suggested that

formulation of "psychologically semantic constructs is the only

real"

possible goal for semantics:

7
 Chapter 1

The main problem in linguistic semantics is that of empirical

validity. To construct a revealing semantic representation
of an utterance is to show what thought this utterance
corresponds to. When we propose a semantic representation we
want to be sure that our formula actually has some
psychological reality - that it is not a more or less elegant
fantasy but rather a faithful portrait of a thought. "
(Wierzbicka, 1978: 118)

First we must be clear about what it would mean to claim

"psychological reality" for a linguistic construct. To say that a

construct is psychologically real is to say that the construct as

described is a component of a system the brain uses, and to predict

that if we could examine the operation of people's brains, we would be'

able to recognise the particular cognitive construct in the processes

going on in the brain from the description previously provided. A

useful analogy is to say that a construct is psychologically real if

it is part of the software held in the brain and used for cognitive

processing. Since this is not the right setting for lengthy

discussion of this question, I want briefly first to make the point

that "psychological reality" so characterised is not. a reasonable goal

for semantics, and neither therefore for this study, present, and
at

second to describe a relationship between theory-building and

cognitive considerations which is reasonable, and is the one adopted

in the present work. Finally I shall summarize the general

psychological claims which can be made for this work.

Kosslyn (1978)- in describing why, in general, the act of

investigating how the brain works, verges on the impossible, shows us

why 'psychologically real semantics' is not a reasonable goal:

"This situation of the [that

investigator] is rather like one
where an unknown solid object is placed in a dark box and our
task is to describe the surface of the object by shooting in
BBs and observing how they bounce back. There are three
important aspects to this situation: first, there is the
nature of the hidden object of study; next, there is the
data, the angle at which BBs bounce back when they are shot
in at various angles; and finally, there is the systematic

8
 t, napLer 1

relation between the angle of incidence and the angle of

reflection. It is clear from this example that one needs to
know two of these things in order to determine the third. If
one does not know that the angle of incidence equals the
angle of reflection, one cannot use the data to deduce the
shape. Similarly, if one wants to discover the
incidence/reflection relation, one must have knowledge of
both the data (the results of shooting BBs at particular
angles) and the surface characteristics of the object.
Consider the sorrow of one who knows only how BBs bounce out
of our dark box, and knows nothing of the law relating the
angles of incidence and reflection, nor anything about the
concealed object. Somehow he must infer both the
incidence/reflection relation and the nature of the surface
of the object from the data, but he needs prior knowledge of
the relation to infer the surface characteristics and vice
versa! This pitiful creature is the cognitive psychologist. "
(1978: 218)

If we accept Kosslyn's analogy[31, the goal set out by Wierzbicka for

semantics is unrealisable in the present circumstances. The situation

is that outlined by Wirth (Section 1.3.1); if three variables are

,
involved in a function, and we know only the value of one, we cannot

deduce either of the other two. Given this, it is misguided to use

the criterion of psychological reality in criticising semantic

constructs.

This means that linguistic model making can be quite unconcerned

by questions of the relationship of the model to the psychological

mechanisms in the brain, if its proponents so choose. This attitude

is made explicit, and to a certain extent has been viewed as

sanctioned by, Chomsky's early statement that (1965: 9) "a generative

grammar is not a model for a speaker or a hearer". A linguist may,

3 And we-might not, since as Peter Monk has pointed out to me, an
investigator in this position can use simple trigonometry to plot
points which must be on the surface of the object, and thus, if he
makes enough observations, he can arrive at quite a detailed
description of the object. If this is the true analogy, psychologists
and linguists are not in such a bad position. .

9
 Chapter 1

and most do, legitimately choose to give an account of a particular

subset of linguistic data without wishing to claim psychological

reality for that account. Such an approach takes the view propounded

for example by Wilks (1977: 72) that predictive power is the final

test of a theory "Their [.. only

- ultimate accountability .1 can
be whether or not they work. " According to one widely quoted

definition, such a theory could claim psychological reality simply by

virtue of its ability to account for the observed data:

"A linguistic concept is psychologically real to the extent

that' it contributes to the explanation of behaviour relative
to linguistic judgments, and nothing more is necessary for
.
this. " (Levelt 1974, vol 3: 70)

(cf also, Kiparsky, 1968: 171)

Because of the "bouncing ball" problem, this claim is in fact the

strongest that can at present be made for a linguistic construct.

Yet this is intuitively unsatisfactory. The problem is

articulated by Wierzbicka in the passage previously quoted: many

linguists would like to have a further criterion for empirical

validity than mere descriptive adequacy. The appearance 1of an entire

book devoted to the topic (Halle, Bresnan 1978)

and Miller, testifies

to the strength of this desire. Most discussions and invocations of

psychological reality as a criterion spring'from, this wish, which is

certainly one I share. Yet, as we have seen, a genuine claim of

psychological reality is empirically impossible at present.

There is, however, a further guiding principle which may be used.

This is what I shall call psychological plausibility. Making a model

psychologically plausible means on the one hand making' it compatible

with any experimental information about psychological aspects of the

phenomenon under investigation, and on the other, making it compatible

with those sections of psychological theory which the investigator

10
 Chapter 1

finds (Note that the conclusions drawn by a psychologist

convincing.

are equally a matter of combining inference with plausible

assumptions, see Kosslyn (1978: 218ff).

It is important to that psychological plausibility in no

realise

way confers empirical validity on a model. I think this may be the

invoke reality; for

mistake made by many people who psychological

example Bresnan (1978: ): "... previous attempts to realize

transformational grammars as models for a speaker or hearer are

valuable and informative. By showing us in what respects grammars may

be psychologicaunrealistic, they can guide us in our efforts to

better theories of grammar. " [my underlining] What

construct

plausibility does do is to make the model more

psychological

intuitively acceptable, and perhaps acceptable to a greater number of

people.

In the present work, for example, the theoretical discussion in

Chapters 7 and 8 is, compatible with the results of research into

cognitive aspects of perception of category membership (in particular

that reported by Rosch and her various associates, see references).

The fact that my, analysis of the vague expressions studied is

supported by Rosch's results renders it more pretheoretically

but it would not support the claim that it is

plausible,

'psychologically real'.

The claims I would make then for the work presented here are the

following:

1 It is psychologically real only in the weakest sense of

Levelt (1970) in that it account of the data,

gives a correct

and

2 the plausible in that they

proposals are psychologically

11
 take account of available psychological data, and cognitive

assumptions currently accepted by some psychologists.

1.4 General Theoretical Approach

The main emphasis of this work is descriptive. This section aims only

therefore to provide information about the general theoretical (and

pre-theoretical) approach which I adopted, and to describe my

theoretical assumptions.

1.4.1 Idealisation

Any scientific investigation must idealise. It is impossible to study

everything to the same depth from the beginning. I therefore set out

to describe vague language use in just one speech variety of English:

that of Standard English as used by educated native-speakers. As

noted. by Milroy (to appear 1983: mimeo) "the notion of a standard is

notoriously complicated and hard to pin down", yet it is reasonable in

a linguistic study to make an overt assumption that such a variety can

be the object of description (cf Lyons, 1977: 587). A sensible

rationalisation of the choice of Standard English is given by Crystal

(1969: 11-12):

This choice was directed, not of course by any concept of a

linguistic betterness within this dialect but by
or accent,
the greater usefulness of research based upon them, as
opposed to any other British regional or class dialect or
accent: this variety is the one upon which most research has
already been done, and is the basis for the majority of
textbooks available, hence its use will facilitate
correlation of my results with already familiar information".

In collecting my data I deliberately set out to minimise

variation arising from social differences, by taking data from what

might plausibly be assumed to be one speech community: that of people

working and studying in higher and further education institutions in

England. The written data is from sources which should broadly

all

12
 Chapter 1

fall within the same category, mostly the 'quality' newspapers. The

test subjects equally can be classified in this category, since one

group were higher education students, and another sixth-form students

from an 11-plus selected grammar school.

An empirically testable prediction arising from this work would-

be that a theoretical framework which successfully accounts for vague

language use in standard English should be applicable in principle to

vagueness in any other language or language variety.

Another approach to vagueness is to view the vague expressions

themselves as sociolinguistic variables. Dines (1980) suggests that

for Australian English the tag approximations are markers of social

class differences. Some subjective comments by test participants

suggest this may also be true between varieties of English; and that

especially is felt by some standard English speakers to be a

stigmatised form (see Chapters 4 and 6). In order to discuss a

particular set" of structures relatable -by their meaning as

sociolinguistically variable, it is first necessary to have` an

adequate description for just one variety. My decision was to provide

that description, in order that further work could subsequently

investigate the sociolinguistic aspects of vague language use.

1.4.2 A Note on "meaning"

"Pending a satisfactory explanation of the notion of meaning,

linguists in semantic fields are in the situation of not
knowing what they are talking about. " (Quine, 1953: 47)

Since I believe that anyone not defining the term 'meaning' adequately

before using it must be viewed, by Quine, as not knowing

as suggested

what they are talking about, there now follows a short section

13
describing the general approach to meaning adopted in this work.

"Meaning' will be used as. a broad cover-term indicating all the

propositions which a hearer can reasonably derive, taking account of

contextual and background knowledge, from the utterance of a given

sentence on a particular occasion. Thus part of the meaning of

(2) Do you always make your own pastry? (Hudson, 1975: 4)

in a given context, might be taken to be "This pastry is really

leathery". Thus (meanings is used to encompass all the subdivisions

which have been proposed for it, such as 'sentence meaning',

'propositional content', 'entailments' and 'implicatures'.

The meaning investigated is hearers' meaning. At a theoretical

level, this is because hearers' meaning is seen as more

conversationally salient in the sense that it is hearers' meaning

which gets acted upon, and which therefore influences the developing

structure of a conversation. The suggestion is that 'speakers'

meaning' is inasmuch
, as communication succeeds, the same as hearers'

meaning because speakers have to take into the

, account mechanisms

they know "hearers will use, in formulating Hudson

their utterances.
(1975: 5), who is reformulating Grice (1967), I think
with, a clearer

example, puts this view forward as follows:

"The work done by the speaker in conducting is

a conversation
closely related to that done by the hearer [... 1: what
the speaker must do is decide what conclusions he wants the
addressee to come to, and then find a way of ensuring that
this happens. This means-that before saying [(2) above] Mrs
Green has to work out how Mrs Brown is likely to take it:
e. g. does she realise that Mrs Green prefers home-made
pastry to shop-bought pastry? Needless to say, speakers
often make mistakes in this kind of calculation, although
they usually put the blame on the addressee: "

14
 Chapter 1

Psycholinguistic work offers confirmation of this point. Harris,

Begg and Upfold (1980) report an experiment in which it was observed

that speakers consistently tailored their messages to anticipate the

hearer's interpretation. Therefore hearers meaning should be of the

most interest to linguists.

In attempting to describe hearers' meaning, I reject the

formulation that what hearers do is to "recover" what speakers intend

The meanings hearers arrive at may be quite other than those intended

by speakers. For example in the following exchange:

(3) [Context: wife to husband at 6.30 pm in Christmas week]

+B: Would you like to have a bath - the water's hot?

C: Why, where are we going tonight? [12.78]

C's reply makes it clear that he had inferred that one

proposition forming part of the meaning (as broadly characterised

herein) of B's utterance is

I think you should have a bath, because we are going out

The subsequent breakdown and repair enabled the participants to see
(amid hilarity) that B had not intended this proposition at all. So

far as its role in determining subsequent conversation was concerned,

the meaning of B's utterance was as understood by C, hearer.

_its

At a methodological level, the reason for to

preferring'

investigate and describe hearers' meanings is that while they may

still be difficult to observe, they are more open to empirical.

observation than information about speakers' intentions. For example,

hearers' meaning may be construed through observation, of hearers'

action resulting from a particular utterance. Thus the investigator

observing the exchange at example (3), may legitimately attribute the

proposition "request/strong suggestion" to the meaning of B's

15
 Chapter 1

utterance, on the basis of C's reply. C has indeed made inferences

about B's intention, and has acted upon them in his reply. What the

investigator may not legitimately do is either to state B's intentions

on the basis of what B says, or to state that C has in some, sense

"recovered" what B intended. The ways in which hearers' meanings are

investigated are described in Section 1.5 Methodology.

1.4.3 Grammar

I assume that the three components traditionally recognised (syntax,

semantics, phonology) can be separated for the purposes of

description. (This leaves aside the question of whether they are

valid for psychological models of speech production. )

Syntax

This work makes no direct contribution to syntactic theory. Therefore

the relevant assumptions about syntax are not well-developed. I

assume that there are syntactic rules which map logical forms into

"surface strings". Where syntactic terminology is it be

used, should

seen merely as a convenient way to refer to a generally acknowledged

phenomenon rather than as indicating adherence to any particular

syntactic theory. For example, I use the term "conjunction reduction"

in reference to semantically-relatable sets of like:

sentences
(4) aI bought some apples and some oranges the
at market

bI bought some apples at the market

cI bought some oranges at the market

but this does not mean that I necessarily believe that (a) results

from a transformation working on (b) and (c). I use the term as. a

convenient way of linking with the relevant existing literature.

Phonology

The area of phonology which is drawn on in this study is intonation.

Intonation is relevant in that it is observed that certain expressions

16
 Chapter 1

from the list at 1.1, which are structurally ambiguous between a vague

and an exact reading, are differentiated by means of intonation

differences. In addition the 'tag' expressions use intonation to mark

the 'scope' the is in [ VP TAG ] where VP

of tag - that an expression

consists of [V NP ], whether the vagueness is relevant to the whole

VP, or just to NP.

Intonation should be understood in the sense of Crystal (1969:

195) as "a complex of features from different prosodic systems. These

vary in their relevance, but the most central are tone, pitch-range,

and loudness, with rhythmicality and tempo closely related. " He has

previously defined prosodic systems as "sets of mutually defining

phonological features which have ahessentially variable relationship

to the words selected"(1969: 5).

The contrasts referred to in the first paragraph of this

sub-section are contrasts of tonality and tonicity. Crystal notes

(: 263) that tonicity may perform function,
either an accentual as in

the case of the choice of nucleus positions possible for a sentence

like

(2) Do you always make your own pastry?

or it may perform what he calls a grammatical function, as in the case

of, for example,

(4) Was she wearing a green dress or a red one? (Crystal,

1969: 263)

where the word red must be nuclear. In the cases I consider, the

tonicity and tonality have a grammatical function.

The transcription system used is that developed in Crystal

(1969). For convenience the are set out in Table 2.

symbols

Semantics and Pragmatics

17
 Chapter 1

I understand semantics and pragmatics to be convenient (or not so

convenient) labels which arise from a hypothetical division of

cognitive skills relating to meaning into those which are specifically

linguistic and those which apply also to other types of behaviour.

The considerable amount of attention paid to the exercise of

boundary-drawing between semantics and pragmatics may be a

pretheoretical indication that drawing such a boundary might turn out

to be spurious.

If the distinction is adopted, it will be axiomatic that

SEMANTICS & PRAGMATICS ---> Pmsb

(cf Wirth's formulation, Section 1.3.1 and Gazdar's "PRAGMATICS

MEANING - TRUTH CONDITIONS"(1979: 2))

where SEMANTICS and PRAGMATICS are sets of hypotheses about the

meaning of linguistic expressions, and Pmsb is any proposition about

the meaning of particular speech behaviours (to the extent that the

empirical difficulties outlined in 1.4.2 can be circumvented to allow

observation).

From observations of speech behaviour have

we at present no

clear-cut theoretical or practical way of deciding which aspects of

meaning are semantic and which are pragmatic (it is again the point

made by Wirth, in an expression with three variables, if only one

value is known, its impossible to- deduce the other two). Wachtel

(1980) in criticising Channell (1980), that it is invalid to

claims

base semantic constructs on evidence from observation of speech

behaviour (specifically, elicitation tests). Lyons (1977: 117) states

that "the distinction of pragmatics and semantics in-relation to the

analysis in language is, [... ) generally

of meaning natural

recognized as controversial". This area of controversy will be taken

18
 Chapter 1

up in Chapter 7.

Representations of Meaning

A problem which has bothered many writers interested in semantics is

what sort of representations should be found at pre-lexical level.

Confusion arises because the most obvious way of talking about the

meaning of words is with other words, with a consequent possible

confusion of levels. Logicians-employ '' to mark the sentences of

the object language, so in, for example, the expression

'snow is white' is true, iff snow is white

the sentence with '' refers to an object language proposition, and

the whole expression is a statement in metalanguage of its meaning.

Generative semanticists used capital letter written words, thus

[WE CAUSE[DRAGONS BECOME DEAD]]

SS

"The terminal elements in this phrase marker are written in capitals

to indicate that they are. not language specific lexical items but are

universal semantic primitives. " (Fodor, 1977: 78) Others have used

square bracketing around prelexical semantic features (eg Katz and

Fodor, 1964; Lehrer, 1972).

Such attempts at describing meaning by the use of a meta-language

which is an adaptation of the object language have been termed as

nothing but translations into "semantic markerese" by Lewis (1972).

That is, he claims that merely using other words to describe meaning,

reveals as few interesting generalisations about the nature of

language as does translation from English to French. His criticism

is, I think, too sweeping. The proposals for a semantic description

of verbs made in Dixon (1971) use a small set of semantic components

to define nuclear verbs, and then define non-nuclear verbs using them.

Evidence from the Dyirbal 'mother-in-law' language demonstrates the

19
 Chapter 1

salience of the two categories, as does evidence about the order in

which children acquire words (nuclear before non-nuclear, as a general

rule). Rosch et al (1976) examined Roger Brown's corpus of

spontaneous child speech of one child (Brown, 1973) and found that at

Stage 1, only basic level (=nuclear) names were used (Rosch et al were

only interested in nouns). So it may well be true that expressing the

semantics of non-nuclear words in terms of nuclear ones does reveal an

interesting generalisation about the nature of linguistic meaning.

My purpose is not to contribute to this debate, but to state how

descriptions will be made here. Despite the circularity and possible

confusion arising from formulations in English, I shall continue to do

this, principally so that the description is comprehensible. J. D.

Fodor's rationalisation of this is to say that such descriptions

"mirror the prelexical structures from which words would be derived

[in a Generative Semantic Grammar]" (1977: 119). Contra the criticism

of Lewis, I hold that the resulting descriptions will embody valid

generalisations about meaning.

Pragmatics

Pragmatics is the area of language study which probably at present

enjoys the smallest degree of agreement as to what it includes and how

to study it. Since the purpose of the present work is practical more

than it is theoretical, I shall not do more here than draw attention

to some of the unresolved considerations, and say how I propose to

proceed.

Given the axiom

SEMANTICS & PRAGMATICS ---> Pmsb

it was inevitable that those with a major interest in semantics would

assign to pragmatics observed aspects of meaning which they did not

20
 Chapter 1

wish to include in their proposals for semantics, and thus, that the

area would acquire the 'waste-basket' air which has been attributed to

it (for remarks in this vein, see eg Chomsky, 1969: 81; Bar-Hillel,

1971; cited in Kasher, 1977). If however, we take as our object of

study a set of observations of the meaning of, particular expressions ,

with a view to providing a description which accounts for that set of

observations, we shall, as suggested by Levinson (1979) be equally

interested in semantics and pragmatics.

It is usually implied, if not made explicit, that hearers faced

with incoming utterances first process these through their grammar,

and then deploy pragmatic inferences to sort out problems (for an

example of this view, see Smith and Wilson, 1979: 197). Contra this

view, some recent work suggests that in accounting for language

understanding, we should possibly be more interested in pragmatics

than in grammar, on the basis of evidence that hearers deploy

non-grammatical inferences of the type - "What could this speaker

possibly, and reasonably, be meaning, in this context? " before they

deploy the full range of relevant (cf
semantic and syntactic rules
Milroy to appear 1983: mimeo; Danell,. 1978: 14).

There are currently no well-developed theories

or-widely-accepted

of pragmatics which can be adopted. Grice's (1968,1969,1975,1978)

proposals provide a possible set of principles for a theory, but do

not constitute in themselves such a theory. Gazdar (1979) is an

attempt to formalise a theory of pragmatics, because such a

formalisation, he contends, will directly the explanatory and

expose

predictive value of that theory (cf Gazdar 1979: 10). His formal

system, it is claimed, "given an utterance, tells us what that

utterance implicates and presupposes" (: 129). Examples are given

21
 Chapter 1

which show that his system makes correct predictions. I think it is

correct to say that most of the examples used are invented. My view

is that such examples cannot provide a real test of the validity of

the proposed system. It must be tested against real utterances

complete with accompanying observations relevant to determining

hearers' meanings.

Given the lack of any generally-accepted or developed theory, I

shall proceed with the two goals of discovering (a) what propositions

relevant to the meaning of vague utterances a putative pragmatic

theory must account for; and (b) what its necessary components are.

The putative set of pragmatic rules relevant to language

understanding has, as noted, an interface with semantics. It also has

an interface with more general rules of social behaviour. In drawing

that other boundary, I follow the suggestion of Levinson (1979),

whereby pragmatic rules make- reference to particular social

relationships, which bear on the communicative effects arising from

utterances, which in turn are codified by sociolinguistics. In the

present case, for example, the pragmatics include

of vagueness could a

rule to the effect that number approximation is often used where a

speaker is/feels inferior in knowledge/ability to his hearer(s); while

the description of the social relationship had as a product the

which

use of a numerical approximation, would be the concern of

sociolinguistics. This is a practical division of labour, and as

Levinson notes "coincides with the kind information that on the one
of

hand one would expect, and on the other not, to find in a

expect

grammar, whether theoretical or practical" (: 218). Incidentally, it

is also useful in that it allows a unitary approach to human

communication, that is, the same kind of rules as will account for the

22
 Chapter 1

appropriateness of number approximation will also account for the

appropriateness of kissing, or bowing.

"Context" and "Pragmatics"

The putative subject matter of pragmatics has been defined as an

account of those meaning propositions not accounted for in semantics.

The widely-held assumption is that propositions not directly

attributable to the semantics are in- large part a result of the

influence of context. Should this be taken to mean that all facts

about any context of utterance were relevant to pragmatics, the

linguist's task would be quite unmanageable. Clearly though, we can

distinguish theoretically between linguistically relevant, and

irrelevant context, as suggested by Lyons (1977: 572):

"... the linguist abstracts from the actual situation and

establishes as contextual all the factors which, by virtue of
their influence upon the participants in the language event,
systematically determine the form, the appropriateness or the
meaning of utterances. "

Yet this is unhelpful, unless accompanied by an indication how to

of
do it. "Here, I believe an ethnomethodological is
perspective useful.

In accounting for stretches of talk, identify

we as pragmatically

relevant those factors of context which to the

emerge as relevant
developing structure of the talk. Thus, for identifying
example, a

perceived social difference of the type mentioned above as relevant,

must arise from evidence in the talk, possibly in the form of

utterances like "you know more about this than I do" (examples of this

approach will be seen in Chapter 5). In this way, all that is context

is not necessarily relevant to an account of meaning, and yet what is

excluded is excluded for motivated reasons.

1.5 Methodology

My fundamental assumption is that the linguistic study of meaning

23
 Chapter 1

should be an empirical study. It must involve accounting for real

occurrences of talk, rather than accounting for invented sentences.

There are reasons why this is particularly true for the study of

meaning. The data relevant to the study of meaning is 'meanings'.

They cannot be counted, measured, recorded and quantified in the way

that sounds, syntactic patterns and words can. Ringen (1980) makes
,

the distinction by calling this data 'Putative Linguistic Facts', in

opposition to 'Linguistic Data'. An example sentence on a page, or a

recording of an utterance, does not itself provide information about

its meaning. This must mean that the meanings investigated by a

linguist are meanings which he has ascribed to an utterance or

sentence by virtue of his knowledge of the language. Such use of

intuitions does not appear very empirically valid (for a similar

observation, see Fodor, 1977: 6-7)[4]. It is, however, greatly

improved when the investigator attributes meaning to utterances

forming part of real conversations. In this his
situation, assignment
of meaning to any utterance can be made on the basis of, and

constrained by, the subsequent development (as

of the talk was

suggested for example (3), Section 1.4.2). No such constraints are

present with invented sentences, the meanings attributed to which can

be any the writer is able to his
persuade readers to accept.

Conversation data, used in the way I am suggesting, lead to

should

accounts of meaning more valid than those which arise from accounting

for invented data. Martinet, as long 1958, out the

ago as point

dangers of abstracting language data from he called "linguistic

what

4 Despite this, most writers do not question the process, eg Gazdar

(1979: 11) "I shall assume throughout this book that invented strings
and certain intuitive judgements about them constitute legitimate data
for linguistic research. "

24
 Chapter 1

reality in the raw".

Naturally it is acknowledged that this type of post-hoc analysis

has problems. The linguist's tendency to allow his theoretical

perspective to influence his intuitions is well-known (see Spencer,

1973, for an empirical study, and also Ringen, 1980: 115 footnote 30).

Wootton (1975: 6 and 64), writing within a Conversation Analysis

framework, gives examples where particular investigator glosses lead

to particular analyses of the conversations being studied.

I, would argue, however, that post-hoc analysis of meanings, if it

is carried out with rigorous attention to seeking justification for

glosses within the structure of the conversation or its situation, is

a reasonable way to study meaning. (For similar discussion, see

Milroy 1983, and references therein. )

In addition, using real data has certain other advantages:

1 All utterances are attested having been in
as produced -a

non-experimental linguistic Their

situation. acceptability can be

assessed in part by the reaction of hearers to them.

2 Examples drawn from real data can be seen in their linguistic

real

context. This avoids the uncontrolled recontextualisation by the

reader which is necessary with ordinary decontextualised examples and

which can drastically change his interpretation of them. It also

greatly reduces the opportunity for the analyst to introspectively

invent contextual or situational details to support his argument.

3 Each example has been spoken naturally by a native speaker. It is

therefore possible to make direct appeal to intonational and other

non-segmental phonological features, rather than imagining what kind

of prosodies would accompany an invented example.

4 Real data reveals characteristics of English which might well not

UNIVtRS[ýL
DF YO
25 EI MR'+
 Chapter 1

arise from introspection.

,
1.5.2 The Data

There are four types of data:

1 Attested Conversation Examples

Most of the examples used are taken from surreptitiously recorded

conversations. The conversations are not a truly principled sample of

all possible conversations. They were used because they provided the

opportunity for recording. The participants did not know of the

recordings until afterwards, when their permission to use the material

was obtained.

The practical and ethical difficulties of collecting naturalistic

data are manifest (as noted by Crystal and Davy 1975: ix), especially

with limited technical and financial resources, and working alone.

The investigator must prearrange the recording materials without the

participants' knowledge, and make sure he is not given away by sudden

mechanical noises in the corner. Many people consider it an

infringement of privacy to be recorded. So the investigator must also

try to preselect participants who might be supposed, because of their

interests and background, to be amenable. The students and resea*-h

students I recorded were chosen with this in mind.

The second type of recorded data was collected from BBC Radio, to

serve as comparison data against the conversation corpus.

Thirdly, Dr Marion Owen kindly supplied me with attested examples

of number approximations from her corpus of surreptitious recordings

made in a variety of everyday situations. Fourthly, some attested

data published in other work has been used. In each case, this is

credited to the investigator involved. In addition, many examples

were observed and noted, but not tape-recorded, as and when I heard

26
 Chapter 1

them.

A different type of attested examples are the written ones,

collected from a random selection of reading matter encounted over the

last four years. These are credited to their source.

2 Elicitation Data

The elicitation tests described in Chapters 3 and 4 aimed to

complement the attested data by providing information about what

meanings subjects acting as hearers would assign to vague expressions.

This also provided a further check on my post-hoc assignments of

meaning. In both tests, attested examples were used as stimulus

items, together with invented items which systematically varied

particular parts of the structures of vague expressions.

3 Introspective Data

After every session of elicitation testing, test discussed

subjects
the test material with me. These discussions were recorded, and the

subjects comments are referred to.

4 Unattested data

Unattested examples occur where I refer by

to examples used other

writers. They also occur where I analyse the behaviour of a

syntactic

particular construction and need to try it in a number of similar

environments. A limitation of real data is that it rarely (even if

the corpus is enormous) produces examples of the same word or

expression in even all of its commonly accepted environments.

The work is limited by the scope, in terms of time and resources,

of a D. Phil. thesis project. Further work using a much larger

database, and considering a. greater range of expressions, should

follow. Further elicitation tests could be designed to investigate in

detail the pragmatic variables identified as relevant to vagueness.

27
 Chapter 1

1.5.3 Transcription of recorded data

Given the principal interest of the work, the conversation data was

transcribed in conventional orthography. Punctuation, which might

impose the investigator's interpretation, is not used. The

transcription schema is. set out in Table 1.

Each example used is prefaced by introductory notes about the

_
situation and participants in the conversation and, in some cases, a

longer extract containing the example is reproduced in Chapter 5. In

all conversations, participant A is me.

1.5.4 Elicitation Tests

I have said that the meaning under investigation is hearers' meaning.

One way of obtaining information on hearers' meanings is by directly

asking hearers (as subjects in tests) what they understand by certain

expressions. As suggested, this also provides a check on the meanings

assigned by the investigator. Leech (1970: 346-7) suggests three

reasons for carrying out semantic tests in order to test introspective

hypotheses about meaning - (1) it will give the resulting analysis

claim to generalisability to a population, (2) it avoids the

fallibility of the investigator's introspections, and (3), it corrects

the bias possibly arising from the investigator's theoretical

perspective.

Clearly, however, there are problems with data obtained in this

way. What subjects presented with a decontextualised linguistic

stimulus do is to invent plausible contexts for that stimulus, and

then attribute meanings to it in those contexts. The ways they

individually recontextualise cannot be directly controlled by the

experimenter. (For similar points, see Mikkel Blakar and Rommetveit,

1975: 6; Lanin, 1977: 292; Channell, 1980; Wachtel, 1981; Greenbaum,

28
 Chapter 1

1977). Set against this is, however, the fact that the informants in

the test reported in Chapter 4 all appeared to recontextualise the

test stimuli in the same way (ie, they imagined the same or similar

contexts), although no contexual information was given. I infer from

this that such a test is valid, and that in addition, the systematic

ways in which recontextualisation must take place are themselves of

interest.

Evidence from controlled tests designed to investigate the

influence of particular variables on subjects' responses, shows that

these may vary systematically according to the subjects' perceptions

of, among other things, the purpose of the test (Greenbaum and Quirk,

1970: 50ff); the number of contextual cues (Oller- and Eilers, 1975),

and the linguistic discourse (Greenbaum, 1977). It has also been

shown that the subject's state of mental self-awareness affects his

responses (Carroll, Bever and Pollack, 1981). Given this evidence, it

could be argued that elicitation data do not provide a valid basis for

making generalisations about meaning. This is the substance of the

criticism of my work made by Wachtel (1981).

It seems to me that what is in question is really the status of

the results. The most any investigator can claim for a test is that

certain observations are true for a set of expressions and a set of

subjects. If (as is the case for my two tests) the observations from

the tests are consistent with quite unrelated, independent,

conversation- data, then it is reasonable to claim that the test

results are valid for a larger population. If test results are

inconsistent with other data, then they are correspondingly less

interesting.

29
 Chapter 1

1.6 Organisation of the Thesis

In Chapter 2, firstly, the general notion of linguistic vagueness is

discussed, through description of the relevant literature. Secondly,

a working classification of types of vagueness is presented, with

examples.

Chapters 3 and 4 are descriptive, each being devoted to one of

the two major groups of vague expressions studied. These chapters

present the results of the elicitation tests and set out in detail the

observations about linguistic vagueness which linguistic theory must

account for.

Chapter 5 takes a different perspective, drawing on real data to

describe the conversational effects which arise from use of vague

expressions and the problems which participants may use them to solve.

In Chapter 6I concentrate on lexical considerations. In

particular, I discuss the relationship between the words found in the

vague expressions, and their other (non-vague) uses. This leads

naturally into consideration of the extent to which the expressions

studied should be analysed as 'idioms'.

Chapter 7 relates the descriptive content to linguistic theory,

and in particular to semantics and pragmatics. It confronts existing

proposals for drawing a distinction between semantics and pragmatics

with the data on vague language, and concludes that no satisfactory

ways of drawing it are forthcoming. Having reached this rather

negative conclusion, in Chapter 8I finish by offering a speculative

sketch of a possible future approach to the study of meaning.

30
 Chapter 2

Chapter 2

Vagueness and Language

[asking for a5x3 index card]

A: + Have you got one that's blank

B: Fairly blank yes.

In a section entitled 'Words with blurred edges' Ullmann (1962) traces

from Plato to Byron man's recurrent feeling of the inadequacy of

language to express thought, particularly because of its lack of

precision. He notes also the converse feeling among poets and

creative writers, that such vagueness is in fact an advantage,

reflected also by Wittgenstein who suggested that words are like

blurred photographs and added "Is it even always an advantage to

replace an indistinct picture by a sharp one? Isn't the indistinct

one often exactly what we need? " (1953).

In this chapter, I look generally at vagueness, and review other

work. Most of this suggests that vagueness is present in a great deal

of language use, and that therefore a complete theory of language must

have vagueness as an integral component. I identify different ways of

"being vague" and suggest that these should be divided into two

general categories - cases of vagueness, and cases of suppression of

reference.

2.1 Vagueness vs. Ambiguity

The standard distinction between vagueness and ambiguity is assumed.

In both cases, speaking pretheoretically, the hearer does not know

exactly what he should understand. Ambiguity has traditionally been

identified where has two or more competing but distinct

a sentence

meanings attaching to it. However, for an analyst working on the real

31
 Chapter 2

data of conversations, questions of ambiguity are not as relevant to

language understanding as they have been claimed to be by theorists

studying sentences in vacuo. Lyons (1981b: 203) describes the attitude

of philosophers and linguists to ambiguity as "a highly prejudiced and

unbalanced view". He continues:

"Not only is it frequently, and erroneously, associated with

the view that all sentences have precise and determinate
meanings; it is based on the equally erroneous assumption
that clarity and the avoidance of vagueness and equivocation
are always desirable, regardless of what language game we are
playing. "

As John Local has pointed out to me (personal communciations)

ambiguity is rarely a factor in real communication, because hearers

generally read off a meaning without even realising that there could

have been another one. Ambiguity becomes interesting and

descriptively relevant only when it can be observed to be being

actually used by conversational participants, for example in punning,

or where a breakdown can be attributed to a wrong reading being given

to an utterance. Take the example:

(1) A: +Tim Bailey plays the trombone

B: No he doesn't, it's something else brass

C: Well in that case, it must be a trumpet [NS, 10.9.79]

The hearer of this (me) chose the wrong reading, paraphraseable by

"given what you have told me, it's a trumpet", as was made clear by

subsequent turns. The other reading can be glossed as "In a case

shaped like that, it must be a trumpet". Here the difficulty arises

from the structural identity of the partial idiom in that case with a

locative prepositional phrase introduced by in.

There is considerable intuitive agreement regarding the

distinction between ambiguity and vagueness but independent evidence

is lacking. Weinreich (1966: 411) that if a lexeme "can be

suggests

32
 Chapter 2

data of conversations, questions of ambiguity are not as relevant to

language understanding as they have been claimed to be by theorists

studying sentences in vacuo. Lyons (1981b: 203) describes the attitude

of philosophers and linguists to ambiguity as "a highly prejudiced and

unbalanced view". He continues:

"Not only is it frequently, and erroneously, associated with

the view that all sentences have precise and determinate
meanings; it is based on the equally erroneous assumption
that clarity and the avoidance of vagueness and equivocation
are always desirable, regardless of what language game we are
playing. "

As John Local has pointed out to me (personal communciations)

ambiguity is rarely a factor in real communication, because hearers

generally read off a meaning without even realising that there could

have been another one. Ambiguity becomes interesting and

descriptively relevant only when it can be observed to be being

actually used by conversational participants, for example in punning,

or where a breakdown can be attributed to a wrong reading being given

to an utterance. Take the example:

(1). A: +Tim Bailey plays the trombone

B: No he doesn't, it's something else brass

C: Well in that case, it must be a trumpet [NS, 10.9.791

The hearer of this (me) chose the wrong reading, paraphraseable by

"given what you have told me, it's a trumpet", as was made clear by

subsequent turns. The other reading can be glossed as "In a case

shaped like that, it must be a trumpet". Here the difficulty arises

from the structural identity of the partial idiom in that case with a

locative prepositional phrase introduced by in.

There is considerable intuitive agreement regarding the

distinction between ambiguity and vagueness but independent evidence

is lacking. Weinreich (1966: 411) suggests that if a lexeme "can be

32
 Chapter 2

understood as ambiguous" in a neutral context, it has two dictionary

entries, if it cannot be understood as ambiguous in a neutral context,

but different meanings seem possible, it is vague. In his examples

eat is vague between the action required to eat soup and to eat bread,

whereas file is ambiguous. - Considering a number approximation, such

as around four o-clock "in a neutral context", we might want to argue

that it can be understood as ambiguous between, say, all the various

times between 3.45 and 4.15. But analogy with Weinreich's example,

eat, however a vague assignment seems preferable, inasmuch as this is

a valid test.

Zeugma also shows lexical ambiguity, such, that:

(2) ?I filed the letters and my nails

should be either zeugmatic, or require bizarre contextualisation. In

Chapter 3I describe how the number approximation n or m (like example

(1) in Chapter 1) is distinguished from the alternative use of or, by

a special intonation pattern. This means that zeugmatic examples

cannot be constructed, because choice of one intonation or the other

will automatically select only one reading. Zeugma does not arise

with other number approximations, either. Comparing:

(3) I saw her duck and I saw Susie bend down as well

= potentially zeugmatic

(3') She came at around 4 o-clock and so did Clive who arrived at

4.15 = zeugma impossible

we see that around 4 o-clock is not ambiguous. I have not succeeded

in constructing potentially zeugmatic examples with or something

either.

33
 Chapter 2

Zwicky and Sadock argue that a sentence "isn't many ways

ambiguous just because we can perceive many distinct classes of

contexts in which it would be appropriate or because we can indicate

many understandings with paraphrases" (1975: 4). For example in the

case of the number approximations described in Chapter 3, these are

appropriate in varying contexts, and- moreover, are understood

differently in different contexts, but this does not mean such

expressions require analysis as being many ways ambiguous.

Vague expressions appear to fail the contradiction test

demonstrated by Zwicky and Sadock. Where there is ambiguity, a second

conjunct denial, using the other sense, is possible, thus

(4) That's a dog, but it isn't a dog

but this is nonsense if done for a vague expression, thus

(5) Sam is about six feet tall but he, isn't about six feet tall

seems not to make sense, and it is certainly impossible to assign a

meaning where Sam is 5' 11" in the first conjunct, and 6' 1" in the

second, although, as we shall see, both of these meanings can be

assigned to the expression about six feet tall.

Lakoff (1970) proposes a test which works for some sorts of

constructions but not for others. It shows that a number

approximation like (1) in the last chapter is vague rather than

ambiguous, since it is compatible with

(6) George has written five or six articles and so have I

that George has written seven articles while I've only written five.

Lakoff's test is used where it is applicable.

Confusing vagueness with ambiguity may lead to false attribution

of polysemy. Bennett (1975) notes that

(7) John has to the

gone study

34
 Chapter 2

is, by Lakoff's test, intrinsically vague as to whether John is now in

the study or outside it. However the possibility of either of these

meanings, Bennett notes, led Lindkvist to set up two senses of

to, one for each meaning. Dictionaries often multiply polysemy

because lexicographers tend to identify vagueness as ambiguity (cf

Weinreich 1966). Bennett's thesis, which describes the use. of spatial

and temporal prepositions in English, identifies considerable

vagueness in the meanings which can be attributed to these

prepositions.

It is important to understand that a sentence or an utterance can

be both ambiguous and vague. Concerning number approximations, Zwicky

and Sadock (: 31) analyse as ambiguous the distinction between a

"literal" and a "hyperbole" reading, as in their example:

(8) [31] There are about a million people in San Antonio and there
are about a million people in my introductory course

where the first is-literal and the second hyperbole. They claim that

this sentence is acceptable, whereas

(9) [32] There are about a million people in my introductory course

and there are about a million people in San Antonio

is not. This is, they say, because a violation of the sincerity

condition causes a perceptual set in the hearer who then expects

another violation and in (9) does not find it. I think that both

these sentences are equally odd - they certainly create zeugma,

deducible from a possible humorous effect, which would indicate that

Zwicky and Sadock are correct in suggesting ambiguity is in play. At

the same time about a million would receive a vague reading, as we

shall see from informant test results in Chapter 3.

35
 Chapter 2

It is perhaps the case that linguists have been misled, by

philosophers' emphasis on ambiguity as an important part of semantics,

into thinking that ambiguity plays a greater part in the act of

meaning than it actually does. From now on, I shall be concentrating

on vagueness, which I do believe, and hope to show, plays a very

important part in the act of meaning.

2.2 Previous Work on Vagueness

Interest in vagueness in language use and meaning has arisen in a

number of disciplines: literary criticism, linguistics, psychology,

philosophy. this survey sketches the lines of interest'that have been

.pursued.

Ullmann's discussion of vagueness has been introduced above. He

notes that:

"If one looks more closely at this vagueness one soon

discovers that the term is itself rather vague and ambiguous:
the condition it refers to is not a uniform feature but has
many aspects and may result from a variety of causes. Some
of these are inherent in the very nature of language, whereas
otherg come into play only in special circumstances. "
(1962: 118)

He attributes vagueness to four factors: (a) generic character of

words; (b) meaning is never homogeneous (ie, it is context-bound); (c)

lack of clear-cut boundaries in the non-linguistic world; and (d)lack

of familiarity with what the words stand for.

Reason (a) yes indeed. As he says, words refer to "not single

items but classes of things or events bound together by some common

element" (1962: 118). This inevitably leads to vagueness which is "in

some ways regrettable, but it is the price we have to pay for having a

means of social communication flexible enough to cope with the

infinite variety of our experiences". (For a similar view that if

language were not vague, it would not permit adequate communication,

36
 Chapter 2

see Daitz, 1956).

Reason (b) - interpretation of meaning is context-bound.

Indisputably so, but Ullmann's implication is that context will permit

an exact interpretation to be put on any word: "only context will

specify which aspect of a person, which phase in his development,

which side of his activities we have in mind"(: 124). That is, he

holds that ultimately there are exact interpretations. I shall

suggest, in reference to specific examples of vague language discussed

in Chapters 3 and 4, that this is not the case.

Reason (c) - the extra-linguistic world is vague. Indeed, in any

case as far as our subjective perception of it goes. Reason (d) -

definitely, as we shall see from conversation examples in Chapter 5.

Ullmann's points are true, but I think he confuses causes and effects.

That is to say, (c) and (d) are facts about the world and people in

it, which in turn are reflected by, even necessitate, the capacity of

language to express vagueness, that is (a) and (b), among other

factors. So linguistic vagueness is not gratuitous - it is caused

(like many other observed characteristics of language) by the world

(in the most general sense) in which language is used.

Schmidt (1974) cites C. S. Peirce as the originator of the

notion of vagueness, although as we have seen Ullmann dates it rather

earlier. Peirce was perhaps the first to try to formulate the notion

in any precise way, as follows:

"A proposition is vague where there are possible states of

things concerning which it is intrinsically uncertain
whether, had they been contemplated by the speaker, he would
have regarded them as excluded or allowed by the proposition.
By intrinsically uncertain we mean not uncertain in
consequence of any ignorance of the interpreter, but because
the speaker's habits of language were indeterminate; so that
one day he would regard the proposition as excluding, another
as admitting, those states of things. Yet this must be
understood to have reference to what might be deduced from a

37
 Chapter 2

perfect knowledge of his state of mind; for it is precisely

because these questions never did, or did not frequently,
present themselves that his habit remained indeterminate. "
(1902: 748)

That is to say, the language system permits speakers to produce

utterances without having decided whether certain facts are "excluded

or allowed by" them. Taking this definition as her basis, Schmidt

says "Language use (ie the use of words) is inconsistent. It is not

only inconsistent among speakers, but within an individual speaker

himself. The assumption that the speaker possesses an unchanging

definition of the words which governs usage is not borne out by actual

performance. " (1974: 620) Schmidt argues for the least specification of

words in competence, to account for their varied uses in performance.

She says "insofar as a word is vague on a particular occasion, it must

be vague when considered in competence as well as in performance".

The best way to account for inconsistent word meaning, she holds, is a

dynamic model in which "the meaning of concepts or words [... ] is

continually being reconstructed from event to event". It seems to me

that allowing the semantic specification of words to change from use

to use would be a highly dangerous proposal from the of

point of view

linguistic model making. In addition it would be difficult to account

for how speakers and hearers actually do understand each other. Where

'I do agree with Schmidt is in holding that vagueness has to be

represented.

Schmidt differs from Ullmann, and from Deese (below) in that for

her vagueness is a phenomenon of language, and not of reality. There

does not seem to be any strong reason why both proposals should not be

the case (as suggested by Ullmann). That the extra-linguistic world

(or rather, perhaps, our perception of it) is vague supports, at least

intuitively, the view that the language system man has evolved to

38
describe that world is correspondingly vague.

Another approach to vagueness is found in the more

psychologically oriented work of James Deese. He (in Deese 1974)

holds that vagueness of communication is inherent in the structure of

our ideas, rather, than in the language system:

"I have been arguing for some years now [... ]. that the
correspondence between the ideas possessed by two individuals
who are in communication on a common topic is rather poor, a
condition which-we ordinarily do not notice because we seldom
make explicit attempts to validate a communicated idea
against the original. When we do, as in the case of giving
directions to someone about how to do something, we are
suddenly made aware of the discrepancy that exists between
'the same' idea in the minds of two different people.
Ordinary situations demand that we place only the loosest of
interpretations upon some linguistic utterance we hear.
(1974: 72, underlining mine)

My view would be that at present we have little more than intuitive or

circumstantial evidence to suggest that -the poor apparent

correspondence of ideas between two people results from their ideas

rather than from the language system. Binnick (1970: 151) argues

similarly that vagueness is not a concept which applies to language,

"but rather to the ideas which language Ambiguity,

expresses". on the

other hand, he argues, is a property of language. I think Binnick was

wrong, because he looked only at lexical vagueness When

vs ambiguity.

one considers (as herein) vague expressions, it becomes clear that

vagueness can get into language via the combination of words involved.

The idea that the structure of ideas is vague in no way precludes the

language system also incorporating vagueness, but in a sense,

discussing the structure of ideas goes beyond the proper province of

the linguist (cf remarks in Chapter 1) in this matter, which is to

explain how it is that vagueness is part of language.

39
 Chapter 2

Vagueness, or as they call it, imprecision, is referred to in a

very different context by Crystal and Davy (1975), in an applied

linguistics textbook which accompanies a series of recordings of

natural conversations. Analysing conversational English from the

point of view of helping the foreign learner,, they ' state that "lack of

precision is_ one of the most important features of the vocabulary of

informal conversation" (: 111). They put forward the view that

vagueness is on a scale related to the formality of the occasion, and

that speakers can, if they choose, be more precise. They give four

reasons for vagueness: (a) memory loss - speaker forgets the correct

word; (b) the language has no suitable exact word, or speaker does not

know it; (c) the subject of the conversation is not such that it

requires precision, and an approximation or characterisation will do;

and, (d) the choice of a vague item is deliberate to maintain the

atmosphere. (We shall see examples of all these in Chapter 5. ) They

note that speakers in their extracts seem to mark vagueness by use of

certain expressions. Among others they cite are something like that,

or something, somewhere, probably, and in a way. Dines (1980) reports

an investigation of some of these, from the point of view of

sociolinguistic variation. This will be discussed in more detail in

Chapter 4.

A useful point which Crystal and Davy make is to draw attention

to the existence, in spoken English, of three types of device for

expressing vagueness: Firstly, a set of nouns "which express total

vagueness", such as thingummy, thingy, thinguammajig, whatsit. They

note that "their spelling is somewhat uncertain, since they are

features of the spoken language only"(: 112). In the same way,

Sankoff, Thibault and Berube (1978) note chose, affaire, de quoi,

histoire, patente, machin and truc for Montreal French. Probably all

40
 Chapter 2

languages have dummy nouns of this sort.

Secondly, Crystal and Davy note a number of generic terms and

collective nouns, such as oodles, bags of, heaps of, umpteen and a

touch of; and thirdly, ways of expressing number approximations: a

class of thirty odd, there were about/around thirty, there were

getting on for thirty. They note also-the existence of prefixes and

suffixes which "are frequently used to express approximation, when

precision is not of primary concern" (: 116), giving as examples

(10) That mountain is rather table-like'

(11) Linguisticswise she's-rather clever

They conclude "Native speakers manipulate their language in this way

all the time in informal speech".

Their 'reasons' for vagueness mentioned informally begin to sound

like informal statements of pragmatic rules relevant to vagueness.

The existence of the words and expressions they note, and their

prevalence in their recordings, lend support to the contention that

vagueness is both intrinsic, and important, in the language system of

English.

Another writer approaching vagueness from an applied linguistics

perspective is Brown (1979), who discusses the importance for foreign

leaners of English of 'Learning to be Imprecise'. He says that

learners who do not know how to refer vaguely are on the one hand

often stuck for a way of talking about some item which they don't know

the word for; and on the other, tend to sound "bookish and pedantic,

which is to say, inappropriate". His observations about the English

of foreign speakers make it appear that degree of vagueness is closely

and importantly bound up with conversational appropriateness.

41
 Chapter 2

Crystal and Davy's number approximation examples recall the

observation of Jespersen (1924: 325ff) who, in discussing the meaning

of negation) introduces the idea of uncertainty as defined by Peirce,

when he notes the contrast between

(12) These shoes cost no less than £20 (=exactly £20)[1]

and

(13) These shoes cost not less than £20

which "implies uncertainty with regard to the exact amount".

A quite different slant on the inherent vagueness of language is

given in Lakoff (1972), in which he attacks the tripartite division of

sentences into true, false, and lacking a truth value:

"Clearly any attempt to limit truth' conditions for natural

languages to true, false and"nonsense" will distort the
natural language concepts by portraying them as having
sharply defined rather than fuzzily defined boundaries. "
(: 183)

He draws evidence to support this view of language from the work on

category membership of Rosch (1973). Rosch reported experiments which

showed that perceptual categories such as colour and shape have

internal structure. By internal structure the following is meant:

categories are composed of a "core meaning" which consists of the

"clearest cases" (best examples) of the category, "surrounded" by

other category members of decreasing similarity to that core meaning"

(: 112). She then extended this to see if psychological categories not

showing obvious perceptual scaling, such as vegetables, or birds, were

also internally structured. Her results suggested strongly that they

are and that there was a high degree of agreement about the

"exemplariness" of any item for any category. For example, for birds,

1 There is some evidence to suggest that the form in (12) may also be
vague, since it is used also to predict costs in the future.
 Chapter 2

she found the following judgments held:

typical birds: robins, eagles

chickens, ducks, geese

penguins, pelicans

hardly birds at all: bats

This is reminiscent of Ullmann's attribution of vagueness to "lack of

clear-cut boundaries in the non-linguistic world". From Rosch's

findings Lakoff suggests that sentences about category membership, such

as

(14) A penguin is a bird

are judged by speakers as to degree of truth, (14) being more true

than

(15) A bat is a bird

No classical set theory or logic can cope with this, but Lakoff

deploys Zadeh's fuzzy set theory which can deal with degrees of set

membership.

For Lakoff, further, "some of the most interesting questions are

raised by the study of words whose meaning implicitly involves

fuzziness, words whose job it is to make things fuzzier or less

fuzzy"(: 195). These are Lakoff's hedges - sort of kind of,

,

technically speaking, etc. Lakoff's idea that the function of such

expressions is "to make things fuzzier" is similar to the way Crystal

and Davy suggest particular lexical items exist to make conversation

appropriately imprecise. Lakoff's observations suggest strongly that

vagueness is an essential feature of language, and one which any

theory we devise must take account of shall suggest, in Section

.I

2.4.1, that Lakoff's hedges should be sudivided into different types,

for purposes of analysis.

43
 Chapter 2

Lakoff proposes what he calls "fuzzy semantics", in which,

however, the semantic representations which are arrived at in any

given case (via functions mapping from context) are exact. Sadock

(1977) devotes himself in part to showing that the neat formalism

proposed in vacuo by Lakoff will not, in practice, produce acceptable

semantic representations, because of its extreme complexity. Sadock

(1977: 434):

"If the reader is growing suspicious of this increasingly

inelegant equation, I am not surprised. The more it is made
to fit our impression of what determines the defensibility of
an approximation, the more it diverges from an honest
representation of a purposely, and unabashedly inaccurate
statement, which is what an approximation is. Furthermore,
and more importantly, nothing that I have observed about the
various contingencies that seem to play a role in the
evaluation of the validity of approximations is really true. "

Sadock's own solution to the vagueness of number approximations is to

make the semantics simple and exact. On his view it is the purpose

(seen as part of the context of utterance) of the approximation which

determines its acceptability in any given case. His view is that

approximations are almost "devoid of real semantic content" (p435) and

that the way to state the truth conditions of an approximation is to

make it true in all circumstances, except those where the thing being

approximated does not have the property at all. Sadock's account of

number approximations leads to some wrong predictions, detailed by

Wachtel (1980). His proposals will be evaluated in greater detail in

Chapter 7.

Weinreich's (1963) discussion of vagueness has broadly the same

viewpoint as Sadock:

"Some vagueness is inherent in every sign, and the vagueness

of different signs is not commensurable since vagueness is a
pragmatic factor in denotation and hence beyond the province
of semantics as the study of designation. " (: 143)

He discusses (: 130) "hedges" such as real and so-called "which

44
 Chapter 2

function as instructions for the loose or strict interpretation of

designata". I think his view is that designation is always exact, but

that pragmatic rules can mediate to give vague meanings.

Wachtel's semantics for number approximations (1980,1981)

involves an account in which the truth value of an approximation is

dependent on its containing a number which is an appropriate exemplar

number for the exact number in question. In this account, an

approximative sentence is true if and only if (iff) the exemplar

number in it is an appropriate round number for the actual number

involved. Importantly, the formalisation states, for example, that if

10 is an appropriate approximation for an actual quantity 8 then the

definition would also be satisfied by any real numbers between 8 and

10, and for real numbers above 10 up to the same distance away as 8 is

from 10, ie in this case 12. So his function f associates an 'actual'

number with an interval Iy, where y is the number appearing in the

approximation, ie the exemplar number. The "appropriate round number"

for any number in the interval x,..., z is the number central to the

interval. On Wachtel's system therefore, understanding a number

approximation entails computing an exact interval of acceptable

numbers, and then seeing whether the actual number involved falls

within that interval. This proposal will also be evaluated in Chapter

7.

The common thread running through Lakoff's. Sadock's and

Wachtel's proposals for dealing with this particular bit of vague

language is that the semantics for it will be exact, and trying to

find ways of arriving at exact semantics leads them to the

semantically complicated (Lakoff) and pragmatically complicated

(Sadock and Wachtel) structures they propose.

45
 Chapter 2

For Fine (1975), vagueness is "a semantic notion" defined as

"deficiency of meaning". For him, "any type of expression that is

capable of meaning is also capable of being vague". He cites names,

name-operators, predicates and quantifiers. For example, he says, the

predicate 'bald' is extensionally vague, and intensionally vague in

our particular world. He shows that logical relations hold among what

he calls indefinite sentences (these are vague); for example, if P

'the blob is pink', then P& -P is false even if P is indefinite.

These logical relations are what he calls 'Penumbral Connections'.

His paper is mainly concerned with the theoretical problems posed for

a truth-conditional account of any language by the existence of

vagueness in that language. His examples (: 276)which focus on

NiEIý
English, show that his discussion is chiefly concerned what I call

Categorial Vagueness (see below, 2.4.2) - problems of extension.

The same issue is one of the chief concerns of the concluding

speculative chapter of J. D. Fodor's (1977) book which raises a

number of issues connected with vagueness without explicitly

explaining them in such terms. The causal theory of meaning, which

she discusses, makes specific provision for vagueness in the sense

that it attributes it'to lack of knowledge about which things are of

the same kind:

"Both for proper nouns and common nouns, one's knowledge of

the referent may be extensive or quite fragmentary. One may
even have to rely on more expert members of the language
community to fill in gaps, to show how to distinguish tigers
from leopards or Chomsky from Halle, and perhaps even to back-
up one's belief that they differ at all. But though I may be
hopelessly ignorant about tigers, my word tiger still refers
to tigers, because I acquired it from someone who acquired it
from someone who... acquired it from someone who does know how
to attach it to the world. The chain of communication (some
kind of causal chain) ultimately reaches all the way back to
the original event of "baptism" in which tigers were given
their name. "(: 210)

Such an account fits in well with Ullmann's and Crystal and Davy's

46
 Chapter 2

remarks about lack of knowledge leading to vagueness. Fodor notes

that the fuzziness of categories discussed by Lakoff can be explained

in the same way. Lakoff's claim that the sentence a penguin is a bird

is only partly true "seems to rest on the view that the word bird IN

FACT applies to whatever competent English speakers BELIEVE it applies

to, and that their criteria for 'birdiness' may be many and vague and

can be satisfied to a greater or lesser extent. " (Fodor, 1977: 212)

Fodor's view of the causal theory is that it probably exaggerates

"the extent to which referents of common nouns are genuine natural

kinds". She suggests:

"A realistic view of natural language surely must recognise

that ordinary people often do use the word bird without
intending to include penguins, or use the word fish intending
to include whales. To what extent do we really care, in our
everyday conversation, whether the words we use carve nature
at its joints? "(: 212)

and concludes by making two speculations. Firstly that perhaps all

kind be found to be fuzzy. If then

terms will eventually so, an

explanation of how we use such words will have to depend on

stereotypes. Her second suggestion is that the special fuzzy

properties of kind terms referred to, may perhaps not be so special

after all, but be properties of language in general. This is the view

of Bolinger (1965):

"It is
characteristic of natural language that no word is
even limited to its enumerable senses, but carries within it
the qualification of 'something like'. " (: 567)

Danell (1978), in a wide-ranging but therefore rather general

paper on vagueness and language, discusses whether apparent vagueness

in language is the result of description based on faulty models, or

whether language itself is vague. Drawing on speculative discussion

of how language processing takes place, he concludes that "vagueness

is a property of language that must be included in the models" (: 21).

47
 Chapter 2

2.3 Vagueness and Communication

If words and expressions are vague as often as has been suggested in

the preceding pages, we might reasonably expect communication using

them to be rather inefficient and unsuccessful, as suggested

informally by Deese. A piece of experimental work carried out by

Lehrer (1975) shows indeed that, at least in one limited area of

activity, this is the case. Lehrer was interested in vocabulary used

for talking about wine, from two points of view. "Firstly, assuming

that some of the wine words mean anything at all, even if they are to

be interpreted subjectively and evaluatively what is the structural

analysis of this vocabulary? Second, how do typical (non-expert)

wine-drinkers use these words, and what do they understand when they

hear these words? "(: 901)

Lehrer's subjects all professed some interest in wine and drank

it at least once a week, so they could reasonably be expected to be

quite accustomed to manipulating the kind of vocabulary under

investigation. There were two kinds of experiments; description tasks

and communication tasks. On the description tasks

"the most striking result on all the tasks, [... ] was that
descriptions (not just evaluations), varied enormously
depending on whether or not the taster liked the wine. Wine
F was described by one person as 'sweet, bubbly... flowery,
light fizzy feeling in the mouth' and as 'quite dry, quite
tangy' by another (both liked the wine). The same wine was
described as 'harsh odor, pungent, unpleasant; taste is
bitter, sharp' by one who did not like it. "

This already suggests that use of such terms cannot be regarded as

leading to accurate descriptions. A more conclusive result was

obtained in the communication tests in which the percentage of

successful communications (ie, being able to identify a wine by its

description) was never significantly above chance.

48
 Chapter 2

While, as Lehrer says, these results raise more questions than

they answer, they do demonstrate fairly adequately that communication

between participants on a topic on which they expect to share quite a

lot of knowledge, is in fact very poor. These terms used to describe

wine, despite appearances to the contrary, and protestations to the

contrary by those that use them, are very vague, and their failure to

provide adequate communication in the reported tests shows that this

is so.

In her concluding sections Lehrer raises a number of points about

precision vs vagueness, and appropriateness. Professional experts on

wine need to communicate precisely, whereas "the wine drinker who

comments on the softness of the wine to his companion across the

candlelit table does not need to be so precise", and we might add, it

would be rather inappropriate if he was. Lehrer judges that the

primary function of wine discussions in social settings is "to. share

an experience rather than to convey precise information". This brings

to mind Malinowski's (1923: 313-5) coining of the description "phatic

communion" for communication which has more a social than an informing

function, and again suggests a link between appropriateness' and degree

of precision.

Lehrer notes that wine experts have been carefully trained to

know which wines are suitable for which descriptions, whereas "most of

us have to learn to use these terms, and perhaps most of our language,

in a rather haphazard manner, picking up what we can from conversation

and abstracting properties from objects and contexts". This again is

like the causal theory of meaning, and Ullmann's idea that vagueness

arises from lack of knowledge of the words being used, or the subject

under discussion. Lehrer is suggesting, I think, that a great deal of

49
 Chapter 2

language use is vague.

An important point which she makes about vague and precise use,

and one which will be taken up in Chapter 5 is the following:

"When a need for precision and a scientific use of language

does arise, as among enologists or shippers of wine, the
vocabulary can be sharpened, both in its internal relations
and the application; but this use is derivative - both
historically and functionally. In other words, I am opposed
to a view of language in which the scientific meaning
attached to words is seen as basic and correct and where any
deviations are looked on as performance errors made by sloppy
speakers. " (: 920)

I take Lehrer to be arguing here that vagueness of meaning is inherent

in language and must be accounted for as such.

Her speculative conclusion asks a question to which this 'study

contributes part of the answer:

"My study of wine words has shown that people do not apply
in , Is the domain of wine
words to things the same way.
discussion an unusual one, or it is fairly typical of
speakers' application of words? If the latter is the case,
then speakers probably do not communicate with each other
nearly as well as they think they do, since they usually have
no way of knowing that others apply words differently. But
perhaps they communicate well enough for their purposes, even
without knowing. "

As we shall see, especially in Chapter 5, where language is vague,

speakers take this into account, and manage to communicate apparently

rather successfully.

The similarity of Lehrer's point to that raised quite separately

by Fodor in relation to kind terms, taken together with the other

investigations referred to in this chapter, indicate that the eventual

answer may well be that, as both Lehrer and Fodor suggest, most

linguistic communication is vague rather than precise.

50
 Chapter 2

Considering the disparate origins and purposes of the work

referred to here, it might be quite surprising that everyone makes

such similar points about the nature of and reasons for linguistic

vagueness. It is less surprising if we construe their remarks not as

a random selection of circumstantial and trivial observations about

little understood areas of language, but rather as evidence for

intuitions about the nature of language which will at some future time

prove to be correct.

2.4 Different kinds of vagueness: a working classification

The varied work referred to in the foregoing review has suggested that

there are a number of different ways in which speakers can avoid being

precise or exact. From my observations of vague language, I can

propose four categories, but this is not to say that there may not be

others. They are VAGUE ADDITIVES, VAGUE WORDS, VAGUENESS BY OMISSION,

and VAGUENESS BY IMPLICATURE.

2.4.1 Vague additives

In these cases lexical material is added to what would otherwise be a

precise statement, to result in a vague reading.

One type of addition is to embed a sentence under a predicate which

confers vagueness on the whole assertion, for example:

(16) It seems that he hit her

where the same sentence, without the introductory element, makes a

precise assertion. In general, these allow possibilities other than

that the precise statement is true. Often they are used as

self-defence to cover the speaker against the event of being mistaken.

Similarly, Baker (1975) describes the way remarks which express a

negative viewpoint are often prefaced with self-defence expressions

like that in the title of her paper: "This is just a first

51
 Chapter 2

approximation, but". She claims the ability to use these things is

part of "pragmatic-competence".

Cogen and Herman (1975) describe the use of let's just say to

introduce violations of conversational rules, which will create

implicatures, for example in:

(17) [6] Antony: Who was that man I saw you with last night?

Cleopatra: Let's just say he's a friend

The hearer reads off from the presence of lets just say that the man

is in fact quite other than a friend, but his true status is left to

guesswork. The use of friend is perhaps, in addition, an instance of

paradigmatic vagueness (see below).

2 Another type of vague additives are those found in combination with

numbers, for example:

(18) [article reviewing Habitat catalogue]

+A team of around ten people at Conran Advertising works on the

design and production for most of the year [ST 23.08.81]

Here the precise quantity is made vague by the addition of around.

Approximately, which was studied by Lakoff in his paper on hedges, and

subsequently by Wachtel and Sadock, as mentioned previously, falls

into this category. Chapter 3 gives a detailed account of number

approximations.

3A third type of vague additives are those referred to as TAGS by

Dines ( 1980), and listed at the beginning of chapter 1. Here again,

a precise assertion is made vague to allow for other possibilities:

(19) [tennis commentary - Mark Cox at Wimbledon]

+And we see McEnroe there with his foot - maybe a little bit
of stone or something like that [DS/10]

The structure, use and meaning of these expressions are described in

detail in Chapter 4.

52
 Chapter 2

The set of expressions I am referring to as Vague Additives, as

opposed to Vague Words (cf 2.4.2) and Vagueness by Omission (cf

2.4.3), are, I think the same as those identified in Lakoff(1972) as

HEDGES. He defines hedges as "words whose job it is to make things

fuzzier or less fuzzy"(: 195). I have chosen, however, to subdivide

the major category into different types of hedges, recognising (1)

sentence hedges, (2) number hedges, and (3) tag hedges. Obviously

there are other types. The major part of this thesis is devoted to

investigation of types (2) and (3). I have chosen to focus on these,

firstly because each forms a cohesive and identifiable set, and

secondly because even my quite small sample of data contained plenty

of examples.

2.4.2 Vague Words

1 Vague Substitutes

This category embraces lexical items which are used as substitutes for

exact specifications. Crystal and Davy's set of nouns which express

near vagueness (thingummy, whatsit, etc) would be in this category.

So would vague "pronouns" like someone, something. These items fill

places for expressions which must be present but about whose exact

specification the speaker wishes to remain vague or cannot be more

specific (for example because lexical search has failed to yield an

appropriate word).

2 Vague Quantifiers

Similar to place fillers, but giving a little more referential

information, are what Crystal and Davy categorised as "generic terms"

and "collective nouns", examples being oodles of, or a touch of.

Since I prefer to reserve these terms for the types of words discussed

in 3 below, I have called such expressions VAGUE QUANTIFIERS. Notice

53
 Chapter 2

that using a vague quantifier can fulfil a similar purpose to using a

vague additive with a number, although it will give, usually, very

much less information as to the quantity involved (compare oodles of

mayonnaise with about six tablespoons of mayonnaise).

3 Paradigmatic Vagueness

Another sort of vagueness arises through choice of a superordinate

lexical item when world knowledge and lexical knowledge would make

further precision possible'. For example:

(20) He came into the room leading an animal

We know that a specific type of animal must have been involved, but we

are not told what it was. Similarly in:

(21) A person telephoned

(22) The car has broken down again

An interesting paper by Alan Cruse (1977) draws attention to the

fact that in some cases it is actually inappropriate to give more

precise information, even if it is known. For example, of

(23) (a) I think I shall take the dog for a walk

(b) I think I shall take the alsatian for a walk

(where there is only one dog in the house) (a) is "more neutral, more

matter-of-fact, perhaps also more usual". Cruse shows that the degree

of specificity necessary is dictated by contextual features such as

how much information is already known.

The vagueness produced by these vague words is different to that

produced by the Vague Additives itemised previously. These examples

involve a lack of specificity to a greater or lesser extent, rather

than a vagueness, at a fixed level of specificity, about what is and

is not being 'asserted.

54
 Chapter 2

4 Categorial Vagueness

This is the sort of vagueness to which Fodor drew attention in the

extract quoted earlier - the fact that the terms we use for categories

often do not have a completely agreed extension. Thus, is someone who

asserts I love animals understood as including yeast or mosquitoes?

And when does a hill become a mountain? Putnam (1975) addresses the

question of what it means to 'know the meaning' of a word. He holds

that knowing the meaning may involve being able to identiFj a

stereotype, to which members of a category conform to a greater or

lesser extent. The psychological evidence from the work on category

membership by Rosch and her associates, and by Battig and Montague

(1969), testifies to the psychological reality of this extensional

fuzziness and also provides evidence for the notion of stereotypes.

2.4.3 Vagueness by Omission

A quite different way of being less than precise is by simply not

saying things.

The grammar of English allows this for example in the use of

imperatives. Fowler and Kress (1979: 30ff) explain how imperatives

involve massive deletion of elements which are supplied by the

addressee "from his knowledge of the non-linguistic context in which

the speech act occurs". One of their examples is

(24) a Take particular care of untrained children

[Rule of school swimming club]

from which they say

bI order you that you will

containing Agent, Patient of order and person to whom the imperative

is directed, can be reconstructed. Omitting these, as in (24a),

enables the writer (in this case) to withold from the reader

55
 Chapter 2

information about the identity of the agent.

They go on to discuss what might be called the social effects of

such deletions:

"Tne point at issue is that the uncertainty about agency

spreads a general vagueness through the rules, and a
vagueness precisely in the area of who does what, The readers
of the rules are left in a situation of helpless ignorance:
apparently the knowers know, but seem to keep the ignorant
from knowing. A dissatisfied member can be left very
frustrated by not knowing where to turn for specific action.
Here the process merely confers the power derived from
relatively trivial knowledge on those who have it, and
creates a class of those who do not have such
power/knowledge. In more important contexts it works as a
powerful means of control. Anyone who has even comeýup
against 'faceless bureaucracy' will know what this is about. "
(: 32)

A different example of omission is agentless ("truncated")

passives. Here, the grammar of the language allows us to omit mention

of the agent and thus avoid specifying its identity. Fillmore (1971)

exemplifies this with:

(25) The girls were blamed for the mess

of which he says: "the speaker is merely being indefinite about the

.
identity of the accuser" (: 379). At the same time, knowledge of this

grammatical structure allows hearers to infer that an agent is

involved but has not been mentioned (Freidin, 1975, discusses the

derivation of these structures via transformational rules). They are

particularly well-exemplified in news reporting, where it is

frequently essential that sources are disguised, for example:

(26) +The Independent Peace Group was formed in Moscow in June.

Since then its founder members have been detained for various'
reasons [The World at One, 27.10.82]

It will be clear that agentless passives are often used with the same

power/knowledge versus ignorance/weakness effect, as that described

for imperatives by Fowler and Kress.

56
 Chapter 2

A third type of omission occurs when elements which can be

reconstructed from the semantics of the verb involved are omitted.

Work, separate from this project, in which I have been involved

(reported in part in Channell, Cowie and Jeffries, 1981, and in Willis

and Jeffries 1982) investigated the way in which particular

participants are necessarily present in a situation, for a particular

verb to be appropriately used. Put in a different theoretical

framework, the presence of these participants is part of the truth

conditions for sentences in which the verb appears. To take a

well-worn example (discussed by Fillmore, 1968: 390, and Anderson,

1971, Freidin, 1975), for:

(27) He bought a car

to be felicitously uttered, it is necessary that there have been an

exchange of money, and a vendor. If either of these are absent from

the real life situation, buy cannot be used. So the semantics of b

allows a speaker to suppress mention of both vendor and money, if he

does not wish to detail them, and equally allows a hearer to infer

that they were present in the reported situation. Again, with hit,

Fillmore (1971) suggests that in

(28) He hit the dog [21

"the speaker is merely being indefinite about the implement he used".

The vagueness by omission exemplified by (24) to (28) is again

qualitatively different from additive vagueness in the sense that

these are cases of things whose identity is unspecified in the

sentence concerned. While the ordinary language user may refer to

these three sorts of omissions as "being vague", an analytical

2 Of course (28) is also ambiguous between the reading Fillmore is

focussing on, and the one where he falls out of a tree and onto the
dog.

57
 Chapter 2

approach will do better, on my view, to see this not as vagueness, but

as suppression of reference/mention. Recalling Peirce's definition,

and applying it to for example, (27) as a test, we can see that we

would not want to say that it was intrinsically uncertain whether (27)

applied to a particular vendor or not. It is rather that the identity

of the vendor is simply not given.

2.4.4 Vagueness by Implicature

The fourth way of being vague which I have identified is where an

apparently precise sentence can be used with a vague meaning. This

has been observed to occur frequently in cases involving numbers and

quantities. Wachtel (1980) notes that sentences like:

(29) Sam is six feet tall

(30) Sam has $10,000 in his savings account

(31) Odessa has a population of one million

are on one reading "inherently approximations" (of course there is

also an exact reading), in the sense that (31) would not normally be

considered false if the actual population of Odessa were 1,002,593.

The reason for this is that the numbers in (29) - (31) are seen as

'round' numbers within the base-10 number system. Attested examples

from my data are similar:

(32) [article about Habitat catalogue]

+However, there is less difference between London and the

provinces than one might suppose - 25,000 have bought a wok;
they can't all be Wong and live in Gerrard Street [ST
23.08.811

(33) +I was teaching on a language summer school in UCL once and-

they were working on the roof there and we were on the'top
floor, there were no windows, it was a sort of attic floor
really but there were skylights and every so often you'd
notice a bit of disturbance around the class and there'd be
half a dozen faces all looking, staring in at you (laugh) it
was really creepy [Camb 13A180/9631

In these two examples, hearers infer, I think, from the context

supplied by the speaker, that the quantities given are not to be taken

58
 Chapter 2

as exact.

Percentages are often used in this way. We all remember the

household disinfectant which "kills ninety-ninety percent of all

household germs". Presumably most people took this to mean "nearly

all". Here is another similar example:

(34) [electrical shop assistant explaining why repair be

cannot
carried out, because of lack of spares from factory]

+They've had one fire which has destroyed ninety-five percent

of their total, in fact ninety-five percent of their total
stock [Camb 1B267/p. 7]

The speaker, one infers, did not really know exactly how much stock

was lost, 'he wanted rather to stress that it was most it.
of

In some cases, the implicated-vagueness use is institutionalised

to the extent that it almost completely supplants the exact use. In

North American usage, and similarly but to a lesser extent in Britain,

a couple of means "a few" or "a low number". Bernstein (1971), a

tL prescriptive guide to good Eng/sh, notes that "using a couple of in

the indefinite sense of a few or several is frowned upon by judges of

reputable writing; dictionaries tend to label it' "informal". " (: 27).

Bolinger (1979) describes this use and its constraints in some detail.

Clearly it is sufficiently widespread to make it worth legislating

against. In an example from Cicourel's doctor/patient interviews we

find an example of this use:

(35) [patient sometimes "sees stars"]

+P: Oh I see them occasionally [I: delivery] (pause) yes uh [I: '
okay] sometimes I see them during the day, like I saw (slight
pause) a couple this morning [Cicourel, 1974]

This process is noted by Menninger (1969), who calls it SEMANTIC

FADING, because the exact number meaning "fades", leaving the

signification "a few" or, more often "many". He says:

59
 Chapter 2

"In Greece both and 60

its multiples such as 360 were
frequently used as "round numbers" - that is numbers whose
specific meanings are inflated into the indefinite "many", as
in the expressions "if I've told you once, I've told you 'a
hundred times' that... " The inflation or semantic fading of a
number word is the opposite of the specification of its
meaning and usually involves the three ranks 10,100, and
1000, but it can happen to any other number which for one
reason or another has acquired special significance (as in
seine sieben Sachen packen, 'picked up his seven things'. "
(: 153)

and quotes an example from the Odyssey:

"the number of swine remaining was only 360" [Odyssey 14,

20]

Other common English examples are the sixty-four thousand dollar

I
question (="very important") and Ive got a thousand (and one) things

to do (= "lots").

For the non-institutionalised examples, one question is what

triggers the vague as opposed to the exact reading. One factor is

whether or not the number is round.

(36) Sam is five foot eleven and a half tall

(37) Sam has $9,873 in his savings account

(38) Odessa has a population of 1,002,493

I submit that it is (at least) very hard to get a reading of these

where they are vague, whereas as we have seen (29) to (31) are quite

normally judged as being vague. The reason is that (36) to (38) do

not contain what we think of as round numbers. Clearly the presence

of a round number is a good indication that a vague reading is

intended. Further, Menninger suggests that the numbers which are more

likely to be institutionalised as vague quantities are round numbers.

As we shall see in Chapter 3, round numbers also have a key role in

the formation and interpretation of additive number approximations.

In addition, as has been mentioned, context and world knowledge are

important in deciding whether an exact or a vague quantity is being

60
 Chapter 2

given.

2.5 Summary

We have seen that there are, in English, a number of ways of "being

vague". I have suggested that from the point of view of analysis,

some of these are true instances of vague language, whereas others are

cases of suppression of reference. Vagueness has been observed to

occur widely in language use, and we have noted some investigators

wishing to maintain that all language use is vague in some way. Vague

utterances are defined here, following Peirce, as those for which it

is intrinsically uncertain whether they apply to particular referents

or states of affairs, or not. The major question which arises is how

to deal with vagueness in an account of language understanding. I

shall return to the issue of the theoretical implications of the

existence of vague language, in Chapters 7 and 8.

61
 Chapter 3

Chapter 3

Number Approximations

Neddie Seagoon: How far is it to the valley?

Major Bloodknock: Roughly sixty miles.
NS: I know it's roughly sixty miles but what is it exactly?
MB: Seventy miles.
NS: We'll go roughly, it's ten miles shorter.

[The Goon Show]

3.1

We saw in Chapter 2 that there exist a number of ways of being vague

about quantities in English. In particular, speakers have the option

of either adding something to a precise number or numbers, or of using

a vague quantifier or of omitting information, the necessary

,

presence of which can be inferred, for example from the semantics of

the verb involved. In this Chapter I concentrate on the first of

these types, those where lexical material is added, to result in a

vague quantity reading. Examples of this are:

(1) +You find that you get five or six articles and they're, all very

much the same [11,21.2]

(2) +He's producing about ten pages a week and they're all getting

published [11,21.2]

(3) +It's something around the twenty per cent mark, and it's never

changed [11,21.2]

(4) +This does save you eighty or so pence [W1]

All these are examples of approximations which contain some lexical

material (such as about or approximately) which leads to the

approximation reading - these will be called APPROXIMATORS); one or,

optionally, two numbers, which I shall call EXEMPLAR NUMBERS; and,

also optionally, a measure noun (pounds, feet, etc). Note that this

62
 Chapter 3

use of the term APPROXIMATOR is quite different from and unrelated to

that employed by Quirk et al (1972: 8.29ff) to refer to downtoners to

the force of a verb as in, I almost resigned.

An intuitive analysis of structures like those in (1) - (4)

suggests that they designate not precise numbers or quantities, but

rather intervals whose extent is apparently not exactly specified.

"Interval", throughout the following, should be understood to mean

"continuous sequence of whole numbers and parts of numbers, or real

numbers". The expressions are in wide use, and despite their

vagueness, it is clear that they present no problems of interpretation

to either speakers or hearers:

"There's a lot of occasions where you don't pay any attention

to how exact it is so its like there's fifteen or so
...
people out in the hall its probably likely to be less than
that I would think rather than exactly fifteen or
more" [comment made by test subject]

The three theories of approximations which have been proposed

(Lakoff, 1972; Sadock, 1977; and Wachtel, 1980,1981), which I

referred to in the previous chapter, approach them as posing a problem

for a bivalent truth-conditional semantics, in the sense that it is

apparently impossible to specify the points on the number continuum at

which any particular approximation ceases to be true [1]. These three

accounts direct their attention towards the problem of incorporating

apparently intractable data into an existing semantic theory. As

such, they do not overtly interest themselves in psycholinguistic

aspects of approximations; ie what are speakers, and more

,
particularly, hearers, actually doing when they process

1 Klein (1982) discusses adjectival comparatives, which are similar

to these expressions in being vague, but he notes specifically (note
7: 121) that number approximations are an independent issue.

63
 Chapter 3

approximations? Neither do they interest themselves in attempting to

discover empirically the propositions that are expressed by sentences

containing approximators. They assume that they (and we) know them.

The present study approaches approximations from the opposite point of

view (ie, that of language users) and presents data which complements

and illuminates the theoretical perspective of the studies mentioned.

All four suggest that number approximations, rather than being

phenomena which require a special theoretical corner of their own,

have characteristics which are rather general in language use.

In this Chapter I examine in detail the structure, intonational

characteristics, and meanings of a sample of number approximations,

drawing on other studies, conversation data, and informant work. A.

note on data (3.2) is followed by an account of the elicitation tests

I carried out (3.3), a full description of the structures involved and

their meanings (3.4), together with some general aspects (3.5,3.6 and

3.7). 3.8 is a note on Partial Specifiers, which are shown to be, in

use, almost identical to Approximators. Finally a summary (3.9) of

the propositions relevant to speech behaviour which must be accounted

for by a theory concerned with the meaning of number approximations.

3.2 Conversation examples

The conversation data I had collected at the start of this study

contained a number of examples of what I had tentatively identified as

number approximations. These appear to be used naturally and with no

observable difficulties of comprehension. As we have seen, Crystal

and Davy (1975: 113) assign an approximation reading to these kind of

structures, grouping them in "the wide range of devices that the

colloquial lexicon contains which allows for approximations to be

made". Quirk et al (1972: 13.70) comment similarly on the expressions

64
 Chapter 3

containing or that these are "expressions of approximation". Lakoff,

Sadock, and Wachtel (op cit) also assign vague readings to these

structures. A desire to test and confirm these plausible but

intuitive assignments, was one of the motivations for the elicitation

test I carried out. In Chapter 51 examine some of the conversation

extracts in detail, and discuss the kinds of conversational effects

which number approximations are used to achieve.

3.3 The Elicitation Test

3.3.1 Objectives

The test had two main objectives:

1 to test the hypothesis that number approximations designate

intervals of numbers

2 to find out the length of intervals which different approximations

designate, and the placing of the intervals relative to the exemplar

number(s) present.

3.3.2 Method

A paper and pencil test was used to present 32 examples of putative

number approximations. Some were attested examples from my data, and

the rest were invented sentences. The test items contained four

different approximators and a-selection of exemplar numbers. Subjects

were 26 first-year University of York students who took the test in

two groups on two successive days. They were told that the two

sessions would replace their normal descriptive linguistics seminar

and that there would be a short test, followed by discussion.

The test items were presented in reverse order to alternate

subjects. This was intended to balance out the effects of practice

and fatigue across the items, and also the effect of an item acting as

context for the subsequent one (cf Greenbaum, 1977). It also had the

65
 Chapter 3

advantage that each subject was completing a test paper different from

that of his two immediate neighbours.

Before seeing the test materials, subjects heard a short tape

extract of a conversation in which an approximation was used (it was

example (1)), and they read a transcript of it, in which the

approximation was underlined. This served two purposes: first to

direct them towards the intonation normally used for n or m

approximations (such as that in (1), see Section. 3.4.3 for discussion)

and, second, more importantly, to encourage them to believe that every

test item they read was an attested example from my corpus of data.

For this reason also, each item was written between quotation marks.

I wanted to encourage them to act, as far as possible, as hearers of

the test stimuli; that is, rather than asking themselves "do I say

this and what do I mean when I do? ", to ask themselves "if I heard

this, what would I understand? ". The written instructions each

subject saw are reproduced below:

Example

"You find you get five or six articles and they're all very

much the same"

Someone who thought this could ordinarily refer to anywhere

between 3 and 8 articles (inclusive) would mark their answer

as:

123456789 10

IF YOU FIND THE NUMBERS GIVEN NOT EXACT ENOUGH PLEASE WRITE

IN ANY ADDITIONAL NUMBER YOU NEED

PLEASE WORK ALONE

The list of test items is given in Table 3.1 at the end of this

chapter.

66
 Chapter 3

3.3.3 Results

The results for each item (identified by its approximation) are given

in Table 3.2.

3.3. t1- Discussion

The hypothesis that number approximations are understood to designate

intervals of numbers was clearly supported by the large percentage of

subjects (100 percent in 15/32 items and less than 90 percent for only

five of the items) who marked intervals for each item. The

comparatively low score for Item 10 around ten million is explainable

by faulty test design. In discussion after the test, subjects said

,
that the interval for this did not go as low as nine million, or as

high as eleven, yet the answer sheet had given only whole millions as

possible numbers to mark. Despite the specific instruction to add

additional numbers if they needed to, only one of them had done this.

Commented one:

"It was obvious that it wasn't exactly ten million people and
I didn't think it was as few as nine million or as many as
eleven. I didn't want to sort of go round deciding well its
between ten million two hundred and fifty thousand so and so
but you know sort of it obviously wasn't exactly the ten
...
million"

Results relative to the different approximators are discussed

under separate headings below.

The second finding was. that as a general rule, the length of the

interval increases as a function of the size of the exemplar number,

such that, for example, about £14,000 is judged as designating a much

larger interval than about £500 (see Table 3.2). The results were

equivocal as to demonstrating a consistent proportional relationship

between exemplar number and interval length. As we shall see

subsequently, this is because Interval length (hereinafter I-length)

is determined not only by the E-number (=Exemplar Number), but also by

67
 Chapter 3

several other factors, such as the nature of the item being

approximated, the purpose of the utterance in which it is found, and

which approximator is used.

The third finding supports the general observation made by all

three writers on this topic that the nature of the item(s) being

approximated affects the length of interval for which the

approximation seems appropriate. Taking Sadock's example, he claims

that we understand something different when told that a man is

approximately six feet tall from when we are told that a cockroach is

approximately six feet tall, because of our world knowledge about how

tall men usually are. My own, more specific finding, is that a

/ difference is discernable between discrete items (such as people,

pages, replies) and non-discrete measure nouns (such as pounds, feet,

litres). From Table 3.2, we see, for example, that responses to about

ten pages were different, a longer interval resulting, than those for

10 lbs or so and 10 or so litres. This is the more striking because

it runs contrary to the general rule observed, that or so

approximations designate longer intervals than about/around

approximations. This difference would account for a secondary result

of a series of experiments on category width (Pettigrew, 1958) which

showed a distinct loading on test items involving time and speed

measures (ie non-discrete items) in approximations, as opposed to

another loading towards judgments on discrete items such as numbers of

births, submarines, or churches.

Having noted these three general results of the elicitation test,

we now pass on to a detailed consideration of individual number

approximators, drawing not only on the test data but also on attested

examples and other studies.

68
 Chapter 3

3.4 Number Approximators

3.4.1 about/around/round

About, around and round appear to be interchangeable in most examples,

and test results for about and around indicate that they have the same

effect on the meaning of an approximation. About occurs more

frequently in my spoken examples, and around more frequently in my

written examples, but the samples are not large enough to draw any

conclusions from this. I am treating round as a variant of around.

Distribution

These words appear as modifiers to a number and they always, appear

before the number W. Anywhere that a number can occur, about, round

or around can be added before the number and result in a grammatical

string (except if the number is already modified by something which

X would contradict the approximator, ie *exactly about 10. ) Thus:

(5) +1 wonder what time I've got to go to the dentist. Its

always round four-o-clock [LJ 2,7]

(6) [context: talking about buying Christmas trees]

+I had one some while ago that my mother managed to
keep for about three years [Camb 9B14/63]

(7) [article on climbing Everest]

+He hopes to set up an advance camp at around 20,000 feet
in early July [written, ST 4,5,80]

(5) - (7) could all have appeared without their approximators, and

still been acceptable and meaningful. It is in this sense that I

classify these as lexical additions to sentences which bring vagueness

to propositions which could otherwise be precise. To schematise, an

approximation of this type has the form:

(about )
((a)round) n

where the subsequent material is subject to the general constraints of

what can follow a number.

69
 Chapter 3

Meaning: about/(a)round

At this point I cover those aspects of meaning (construed in the broad

sense in which it was characterised in Chapter 1) which are particular

to about/(a)round. More general features are discussed later. The

effect of "adding" about/(a)round to a sentence containing a number is

to make it vague as to the exact quantity involved. The speaker in

(5) wanted to include appointments at, perhaps, 3.45 and 4.10 in her

reference to her dental appointments. The speaker in (6) was allowing

for the tree to have lasted less than, or more than, or exactly three

years, and so on. -

From the test results for items containing about/(a)round it was

observed that subjects judge these approximations as 'designating

intervals of numbers which are symmetrical about the exemplar number

given. Frequency counts on each item give in all cases a mode which

is the E-number (or two modes, one of which is). Tabulating the

frequency gives in every case a symmetrical figure (Table 3.3). So a

hearer hearing example (2) understands apparently that this person

writing "about ten pages a week" writes between 8.5 and 11.5 pages

each week.

These results explain the judgment reported in Sadock (1977),

that

(8) [26] *John ate around all of the beans

is unacceptable. Clearly if around designates intervals symmetrical

about their exemplar, a constraint on its use is that it cannot be

used with an exemplar such as all which is at the extreme end of a

continuum. Dubious for the same reason are sentences like:

(9) *? About 100 percent of those questioned would vote for the SDP

and *approximately/about/(a)round none is also not possible.

70
 Chapter 3

3.4.2 Approximately

The use of approximately has received the largest amount of attention

in the literature - Lakoff discusses a semantics for approximately

without consideration of other approximators. Sadock only discusses

approximately in any detail, though he does mention a few other

approximators in passing (1977: 438). Wachtel (1980) discusses

approximately and its relationship to sentences not containing a

lexical approximator, but he does not refer to other approximators.

Since Lakoff, Sadock and Wachtel are principally interested in the

theoretical problem of finding a suitable semantics for vague quantity

expressions, they look at only a very few (invented) examples.

It is curious that approximately has received so much attention,

since it does not appear to be the most commonly used approximator.

My sample is perhaps too small for firm conclusions to be drawn from

it, but in the informal conversations I recorded, there were no uses

of approximately. The examples I have suggest that use of

approximately is generally confined to rather specific official,

semi-official and scientific registers. These two, examples appear to

be semi-official situations:

(10) [telephone call to estate agent]

+B: How many houses are there in the street?

C: There are approximately four houses in the street [ATP

DN/ 9]

(11) [Doctor (I) taking preliminary notes from patient (P)]

I: Okay and when was that approximately? [P: hm] This was

after delivery or [interviewer mumbled something]

71
 Chapter 3

P: yeah [slight pause] now this was [pause] the last visit I

[pause] yeah [slight pause] this was approximately a month

ago

[Cicourel, 1974]

Further examples will be discussed in Chapter 5.

Distribution

The distribution of approximately is the same as about/(a)round, that

is, anywhere a number can be used, approximately can be added to it

providing 'a contradiction does not arise. It may well be that

approximately and about/(a)round are in complementary distribution in

different sociolinguistic contexts, but a large corpus study with

control of the relevant sociolinguistic dimensions would be necessary

to demonstrate this.

A constraint on the use of approximately noted by Sadock

(1977: 436) is the following:

(12) [23] *Sam has approximately some money in his savings account

(13) [24] *Sam has written approximately a few/several/many books

That is, approximately is grammatical only when combined with exact

numbers or measures denoting exact quantities. It is apparent that

this constraint applies equally to all the approximators under

consideration, thus:

(14) *Sam has (about ) some () money in his savings

(around) () account
(round )()
(or a bit of)
(or so )

are all unacceptable. This constraint is easily explained by

reference to the test results reported above. According to these,

number approximations are understood by hearers as designating

continuous intervals of numbers, whose position on the number

72
 Chapter 3

continuum is established by reference to the exemplar number given.

An exact exemplar is required to do this (although as I shall show in

3.6.1, some exact exemplars are more favoured than others). It is to

be expected therefore that some, a few or several cannot act as

exemplars because they are not exact.

The requirement of approximators to be followed by something

exact is shown also by what happens in combining approximately with

other things than numbers:

*? approximately pink and *? approximately rabbit-shaped are ruled out

because they are not exact enough, whereas approximately circular is

acceptable because circular indicates something precise.

Meaning: approximately

Lakoff (1972: 222) tries to demonstrate, by means of a series of

invented examples, that the sentence

(15) [3b] Sam had approximately $10,000 in his savings account

is "true no matter what" of a situation where he actually had $9,992,

and progressively less true of one where he had $9,950, $9,500,

$9,200, and so on. He does not say whether he thinks the same
judgments would hold for $10,021, $10,500, $10,750 etc, so his only

claim about the meaning of approximately is that it can be vague to

different degrees in different circumstances.

Much of Sadock's paper is devoted to showing that Lakoff's

proposal for a semantics for sentences containing approximately is

unworkable. This will be discussed in Chapter 7. He makes two

important observations about the meaning of sentences containing

approximately: Firstly, that the judgment of how good the

approximation is, depends partly on the nature of the item being

73
 Chapter 3

approximated. He compares:

(16) [8] Sam is approximately six feet tall

(17) [9] That cockroach is approximately six feet tall

and asserts that if both are actually 5 ft 8, (17) is a better

approximation than (16), because of that we know about humans and

cockroaches. That-is to say, we know that humans are often around six

feet tall, and we know that cockroaches most usually, are not. Second,

he holds that the form of the whole approximation alters the length of

interval over which it is acceptable. By comparing

(18) [11] Odessa has a population of approximately one million

(19) Odessa has a population of approximately 990,000

he suggests that the number of significant figures is relevant. He

also asserts that about a dozen is rougher than about twelve and that

approximately two and a half tons is "not as accurate sounding" as

approximately 2.5 tons. Again here, since it is clear that if these

points are valid, they apply equally to other types of number

approximation ,
discussion of them is deferred to section 3.5 - 3.7.

The sum of Sadock's account of the meaning of approximations is

that they are in some way vague, but that the degree of vagueness is

related to the two observations above. As the aim of this section is

to set out observations about the meaning of approximately, rather

than to discuss how to deal theoretically with that meaning, Sadock's

proposals for the semantics and pragmatics of approximately will be

dealt with in Chapter 7.

Wachtel's (1980) paper makes a number of points relevant to the

meaning of approximately, the most important of which is that if, in a

given context, an approximation (containing approximately) is true for

a particular actual number which is less than the exemplar number, it

74
 Chapter 3

is also true for any other number between the actual number involved

and the E-number, and for all numbers up to the same distance away

which are more than the E-number. Put another way, this says that an

approximation using approximately is true for an interval of numbers,

the centre point of which is the exemplar number. The exact length of

the interval in any given case is determined, in his analysis, by a

function from C, the set of contexts, into F, the set of rounding

functions which map from actual numbers into the set of real numbers

to select an appropriate approximating (exemplar) number. His account

will be discussed in Chapter 7.

The test results indicate that the meanings of the set of number

approximations I looked at are basically of two types: those which

have intervals symmetrical about their exemplar number, and those

which have skewed intervals. Approximately is then, apparently, a

symmetrical approximator.

3.4.3 n or m

These approximations differ from those so far considered, in having

two exemplar numbers, thus:

(20) [reporting on Christmas trees]

+There's another shop in Blackheath that was selling really

scrawny ones, five or six quid a time [Camb 9B14/32] [2]
3
Faced only with examples of this structure, and in the absence of any

experimental evidence, a possible approach would be to subsume it into

an analysis applying to other uses of or. Two analyses suggest

themselves:

1 the two numbers given are alternatives (like "either... or... "

constructions)

2 This example is discussed further in Chapter 5.

75
 Chapter 3

2 the two numbers are inclusively disjunctive (either one is true, or

both are true simultaneously).

Evidence against these two approaches, but quite independent of that

gained from the elicitation test, is provided by an examination of

intonation patterns for or constructions.

Intonation

Crystal (1969). notes that tonality may have what he calls a

grammatical function of distinguishing between "certain types of idiom

and literal interpretation", among which he lists:

I 'sugar ll
(21) would you like one or two lumps of

compared with
11
1
(22) would youI like one or two lumps of sugar

where the first is an approximation and the second presents two

alternatives. Another example of alternative use of or which he gives

(: 273):
I
(23) would you like gin or whisky) or tea

also shows the same pattern of a separate tone unit for each

alternative. Even where two alternatives are in the same tone unit,

the tonicity is non-optional. In Crystal's example (: 263):

(24) Was she wearing a green dress or a red one?

"the item 'red' must be nuclear because of the grammar of the

co-ordinate construction, and the speaker has no choice as to the

distribution of emphasis when he enters the 'alternative' part of this

construction". Alternative uses of or in my own data show the same

characteristics:

(25) +1 just intuitively feel that being in a department thats

I 11 /\ 1i ^U\ Ij
full of people who are really

further
11 really

it
famous
(
is
either

likely
in their discipline or even outside
n
to be a good ideal [II, 21.2]

76
 Chapter 3

II
(26) +either you take them or you dont take them or if
ll
theyre` good well fair enough youve said once why say

it again
1ý [11,21.2]

In contrast with these alternative constructions, my examples of

approximative or show different tonality and tonicity: [3]

I I II
(27) +yeh but its still around that four or five
11
Tregion [1,13.2]

In alternative constructions the two alternatives must both be

stressed, and at least the second, or both, must be nuclear. In the

above example, fourtor five region is one tone unit and region carries

the nucleus. It is behaving intonationally like any prehead modifier

plus noun group, eg

(28) 1 want some; orange paint

Similarly in:

(29) tarticles
you find that you get five or six and
\I II IN II
theyre very much the same [11,21.21

This shows that n or m constructions are distinct from alternative

uses of or and should not be analysed like them. It also underlines

the necessity of using real data in a study of meaning of this kind,

in order to be able to draw on intonational evidence where it is

relevant.

Distribution

There are some restrictions on the way an n or m approximation can be

put together. Firstly the smaller number must always precede the

larger:

3 Notice that this same intonational contrast occurs also to

differentiate more or less with the reading "nearly" from the reading
"either more, or less".

77
 Chapter 3

(30) *There were twenty or fifteen people there

Secondly there are some constraints on what the two numbers may be:

(31) There were ten or twelve people there

(32) *There were ten or a hundred people there

(33) *There were five hundred and nineteen or five hundred and twenty

people there

(34) *There were (fifteen or twenty-one ) people there

(fourteen or twenty-two )

(sixteen or twenty-three)

If the * examples are considered with the one tone group intonation'

observed in attested examples of approximations (and described above)

they seem rather unacceptable (and even quite difficult to say). Of

course as presentations of alternatives (using the other intonation

pattern) they are quite all right.

These unacceptable examples, plus evidence from my attested

examples, suggest the following pattern for acceptability of n or m

approximations (with other combinations being, in general,

unacceptable):

The m number is: for an n number which is:

adjacent up to 19

2 above even, 6- 18

5 above a multiple of 5,10 - 95

10 above a multiple of 10,10 - 190

20 above a multiple-of 20,60 - 140

50 apart a multiple of 50,150 - 950

Hundreds, thousands and millions follow the same rules applied to the

numerals preceding them, eg:

two or three hundred

eight or ten thousand

78
 Chapter 3

fifteen or sixteen thousand

A general constraint is that the higher level round number (ie 200 vs

150) is preferred in the m position, that is to say, everyone accepts

950 or a thousand, but some informants find 900 or 950 awkward. The

top ends of the intervals of possibilities for n numbers are also

i variable across speakers, but not in a way whch affects the general

pattern. So, for example, no speakers I have found can make an

x approximation out of (33).

From this patterning it appears that for an acceptable

approximation to be made, the distance apart of the two exemplar

numbers increases with the size of the numbers involved.

The above observations highlight, I think, important facts about

the way number concepts are internalised, and their relationship to

the number names of English. Numbers which we think of as "round"

have special functions. It is clear that unity, duality, base 5, base

10 and base 20 play a vital role in the way speakers of English

manipulate these approximations in particular, and their base 10

number system in general.

Meaning: n or m

Quirk et al (1972: 13.70) suggest that these structures designate

continuous intervals of which the two exemplar numbers are the extreme

ends. They note that such expressions "will not normally allow the

actual time to go beyond the specified range though it could err

slightly upwards". They do not unfortunately discuss n or m where it

refers to other things than time. Their suggestion about time

expressions is partly supported by the result obtained for the one

time expression I tested - six or seven hours which produced a mean

interval centre of 6h 42 mins and a mode of 7h, with 80 percent of

79
 Chapter 3

informants marking times above 7h, compared to 42 percent marking

times below 6h.

It is. clear that the two numbers given are highly relevant to the

interval designated. Over the ten examples tested, the mean

percentage of informants marking both given numbers as part of the

interval was 79.0. Scores for the individual items are given in Table

3.4.

Table 3.4 shows why Quirk et al's ascription of meaning is not

Y

viable. Over the ten items, the percentage of informants marking

numbers outside the interval bounded by the two exemplar numbers was

75.0, nearly as high as the number marking the two exemplars. The

majority of informants clearly did not feel that the two exemplar

numbers marked the ends of the possible interval. As to Quirk et al's

other suggestion, we can see from Table 3.4 that the overall tendency

for the n or m expressions I tested was to allow an interval running

higher above the higher exemplar than it did below the lower one.

A further observation concerned whether n or m was combined with

a quantity measure or with a discrete item. It was observed that this

had no effect on the percentage of subjects who judged the interval to

extend outside that bounded by the two E-numbers. The difference was

that where a quantity measure was used, fewer subjects judged the

interval to extend below the n number: 41.5 percent, compared to 56.0

percent of subjects for n or m with discrete items. As far as the

upper limit a marginally larger proportion of subjects (66.1

went,

percent) thought the interval could stray upwards where there was a

quantity than where there was a discrete item (60.8 percent).

measure,

The ten n or m items were ranked according to the percentage of

informants who judged both E-numbers to be included in the interval.

80
 Of the five items in the lower half of the ranking, four out of five

used another quantity measure.

These two results demonstrate that, as Sadock and Wachtel

suggested, the item being approximated is important in determining the

length of the interval understood.

The test results contribute the following factors relevant to the

meaning of n or in approximators. They designate intervals, which

contain the two exemplar numbers. The most frequently occurring score

(mode) is most often the point half way between the two exemplar'

numbers. That is, here is a symmetrical approximator, symmetrical

about the mid-point between its two exemplar numbers.

A notable contrast between n or m and the other symmetrical

approximators is in the general length of intervals it triggers.

Expressing interval lengths as a percentage of the exemplar number

given, (or for n or m, of the mid-point) shows that about/around

approximations had mean interval lengths varying from 7.6 to 32.3

percent of the exemplar. n or m approximations had interval lengths

of 22.9 to 66.6 percent of the mid-point. Thus an observed effect of

using an n or m approximator is to increase the size of the interval

designated, or, to put this another way, to be more vague.

Here I have treated n or m holistically, having rejected its

analysis as a use of alternative or. While still maintaining that it

should be analysed holistically, in Chapter 6I examine the

.

relationship between its observed meaning, and the meanings of its

constituent parts.

81
 Chapter 3

All the approximators discussed so far appear to designate

symmetrical intervals. We now come to consideration of an apparently

different phenomenon - approximations' whose intervals are skewed in

relation to their E-numbers.

3.4.4 n or so

Distribution

For some speakers, the following judgments hold:

(35) Six or so books

(36) *six books or so

(37) ten pounds or so [lbs]

.
(38) ?ten or so pounds

With regard to the unacceptability of (36) Quirk et al state: The

items preceding the or so approximation must be units " of measurement

or items contextually rendered units of measurement". For certain

speakers an arrangement like (38) is also unacceptable. These people

appear to have simplified their grammar such that they have one rule

stating that Ns designating units of measurement precede or so and Ns

designating discrete items follow it. A further factor is that an

arrangement like (36) becomes more acceptable (less awkward) where the

number involved is larger, eg:

(39) three thousand students or so

Quirk et al's observation is born out by all the examples I

collected, thus

(40) [letter to York Weekly Advertiser]

York is a wonderful place. All through the year we have

holes in the road, we have the tourists and for six months or
so we have the stink from the sugar beet factory [DN/7]

whereas if the N is not a unit of measurement, it follows:

(41) [questionnaire on student housing]

82
 Chapter 3

+The response was again poor - only fifty replies from the
1,500 or so questionnaires which were
distributed [written SO 4.12.81]

Configurations of the type in (38) also occurred ("eighty or so

pence", "ten or so years ago"). No obvious explanation for Quirk et

al's rule suggests itself. The unacceptability of (36) as compared to

(35) may arise- from what can be observed of their meanings. Or so

triggers -a vague reading which is understood to attach vagueness to

the item immediately preceding the or so. Where a number immediately

precedes, this means that parts of whole numbers (ie real numbers

between the whole numbers) may be involved in the'interval. Something

hearers know about the number system is that there are units smaller

than whole numbers. Equally where a measure noun is immediately to

X the left of or so as in (37), it is this which is made vague, but

quite acceptably, since language users know that the units designated

by measure nouns are susceptible to division either into smaller units

or into parts of units. But where the item preceding or so cannot be

so subdivided, as with books or people, the approximation seems

contradictory, and gives rise to a judgment of unacceptability. This

_
would leave the contrast between (36) and (39) still unexplained. We

know from the test that, pragmatic considerations aside, larger

E-numbers produce longer intervals; so the interval in (39) may be

something like 500 (cf Table 3.2, item 21, the result for 3,000 or so

students), whereas that for (36) might be 2.5 (cf Table 3.2, item 15).

One possibility is an explanation connected with the way people

perceive numbers. The quantities involved with low numbers can" be

conceived of at a practical level, in a way that. the high numbers

cannot. "Six people" are imagineable, and picturable, in the mind, in

a way that 3,000 are not. From the point of view of perception 3,000

is just "very many". In this connection we may note the many examples

83
 Chapter 3

cited by Menninger (1969) of 'primitive' peoples having only lexical

items for a few low numbers, and a word for "many". It can be said

perhaps that we "know the meaning" of the high numbers intellectually,

but not experientially. So, a hearer interpreting an utterance

containing "3,000" does not "picture" 3,000 students; he notes rather

that a high number which he knows of is being used. Pragmatic

knowledge about the possibility of dividing up students (or not) is

perhaps not called on in this juxtaposition so no clash results.

Whereas in the case of six books or so the ability to "picture" six

whole entities conflicts with the vagueness arising from or so. This

is not a very well-developed or well-motivated explanation, but since

I believe that the work necessary to improve it lies in the area of

perception of number, I feel justified in excluding further pursuit of

it from the current work.

Meaning: or so

Two main characteristics differentiate or so approximations from those

mentioned so far. Firstly, a comparison of interval lengths shows

that an or so approximation tends to designate a longer interval than

an about/(a)round one. It shares that tendency with n or m. On

about/(a)round approximations the sizes ranged from 7.6 to 32.3

percent of the exemplar numbers, whereas for or so the percentages are

from 15.5 to 56.8.

The intervals judged in the test as being designated by

approximations containing or so, are skewed upwards from their

exemplar numbers in the manner shown by the histograms in Table 3.5.

This result confirms the observation of Quirk et al (1972: 13.70) to

this effect.

84
 Chapter 3

Again I discuss or so in Chapter 6, from the point of view of its

lexical composition.

3.5 Combinations of Approximators

From the data, it appears that some approximators may be combined, and

others not. Examples of combinations are:

(42) [tutorial discussion: length of sequence of numbers you can

hold in short term memory].

Yeh but its still around that four or five

region [I, 13.2]

(43) B: What's the time now

C: It's about-sevenish or a bit later [Carob 9B14/70]

-ish when added to numbers representing times appears to mean "about".

Quirk et al (App 1.29) say "with ages, it has the meaning

'approximately'. They do not discuss its use with time expressions.

And:

(44) [paper reporting informant work at LAGB]

+We've got about five or six of them but I'm only going to
talk about three of them today [LAGB 9,811

Sadock notes in his paper (: 436) certain combinations which he judges

unacceptable:

(45) [21] *Sam is about approximately six feet tall

As we saw previously combinations like approximately some are not

possible, and I suggested that this is because a symmetrical

approximator like approximately requires an exact number. The

unacceptability of example (45) would appear to be for the same

reason. But, as we can see, -this explanation is challenged by'(42) -

(44) which according to my explanation, should not be possible. What

alternative account can be given?

85
i
 Chapter 3

The first point to note is that (42) to (44) all contain a use of

or". Combination possibilities appear to be restricted to:

approximately
( about )n (or m) (or so)
(a)round
AB

with no combination of the items in group A being allowed[41. This

restriction of combinations appears at first arbitrary - why some

combinations of approximators, but not others?

Firstly, B contains approximators with or, whereas A does not. I

have described the meaning differences observable between the

approximators in Group A, whose meaning is close; and those in B.

Clearly, then, those in Group A are in a simple paradigmatic

opposition which disallows combination (as in *a the book).

By analogy, then, the two approximators-in Group B should not

combine, but they do. - This must be because, being partial idioms (as

I contend in Chapter 6), they retain some syntagmatic properties of

alternative uses of or, which permits several disjuncts, as in:

(46) Mary or Bill or I will meet you at the station

As we have seen, the effect of adding an approximator is to give a

vague reading, so what happens when you add two or three, is that the

reading is even vaguer, cf

4 There
is another approximator (not considered here because of lack
of data),seen in
He's coming over round about seven
This is not just the result of combination, since (a) *about (a)round
seven is not an acceptable substitute; and (b) a fixed intonation
pattern has to be applied. Hence it should be seen as a partial
idiom.

86
 Chapter 3

(47)

a It's seven

b It's sevenish

c It's about sevenish

d It's about sevenish or a bit later

which seem to me to allow a progressively greater latitude as to the

actual time. Had I been aware at an earlier point in this study of

the interest of these double and triple approximation cases, I would

have liked to offer informant data to show this. I would predict

quite confidently a result supporting my intuition.

These data, incidentally, demolish one of Sadock's arguments for

proposing that approximations are more or less semantically empty. If

approximators have semantic content, he argues, you should be able to

pile them up. Look at (45), you cannot. All the approximator does is

to trivialise the semantics so "double approximations would therefore

be ruled out since a single approximator does as much semantic

trivialising as is possible". This falls, since'(a) double and even

triple approximation is allowed (42) - (44) and (b) examples like (47)

seem to show that combining does indeed give vaguer and vaguer

readings.

3.6 Numbers in Approximations

3.6.1 Certain Numbers are Favoured

In theory, it is possible to combine any number approximator with any

number, whole or real, so about 10 and about 9.568 are equally

possible. In practise however, approximations are very much more

likely to occur with what we think of as "round" numbers, such as

multiples of 5 and 10, hundreds or thousands. I have very few

examples of approximations containing non-"round" numbers. We shall

87

i
 Chapter 3

see below that using a non-"round" number in an approximation may

produce several particular effects not observable with "round"

numbers. Earlier in this Chapter, I described how "round" numbers

appear to be important in dictating the acceptable combinations of n

or m approximations. Given their importance for both the form and the

meaning of approximations, it is. necessary to establish more

definitely what exactly the notion "round number" means to language

users.

3.6.2 "Round" Numbers Reference Point Numbers

In a paper entitled 'Cognitive Reference Points', Rosch (1975b)

investigated the composition of what she called "natural categories":

ie sets of items perceived as belonging to the same set, eg 'numbers,

fruits, straight lines. She tested the hypothesis that such

categories, rather than being undifferentiated bundles of items, are

in fact structured internally, in the sense that some members of the

set serve as reference points for positioning the other members in

relation to the whole set. She has previously shown (Rosch, 1973) that

certain members of such sets were judged as 'more typical' than others

or even as 'best examples' of their set. For example, her informants

judged an orange to be a more typical fruit than a tomato. In her

1975 experiment, subjects were required, among other tasks, to place

given numbers in particular 'hedge' sentence frames such as "..... is

roughly " or "..... is basically..... ". The numbers used are

.....

shown in the Figure 1. The numbers in the Stimulus 1 column for

experimental pairs were those expected to be taken as reference

points.

88
 Chapter 3

STIMULUS PAIRS USED IN EXPERIMENTS I AND II

Stimulustype
Experimentalpairs Control pairs

Pair Stimulus I Stimulus2: Stimulus I Stimulus 2

Numbers
Set 1-Variation "greater than"
1 10 11 17 18
2 50 52 36 38
3 100 103 164 167
4 1000 1004 1027 1031
Set 2-Variation "less than"
1 10 9 17 16
2 50 48 36 34
3 100 97 164 161
4 1000.996 1027 1023

Figure 1 (adapted from Rosch, 1975b)

The following figure shows the number of subjects (out of a total of

240) who used stimulus 1 as the reference point (ie, they placed it in

the right hand blank in the sentence frames, producing a result like

'eleven is roughly ten':

Set 1 Set 2

Experimental Control Experimental Control

166 109 220 151

<. 001
p-

Figure 2 (adapted from Rosch, 1975b)

Rosch reports her results as "highly confirmatory of the basic

hypothesis. Subjects consistently judged supposed reference stimuli

in the domain of numbers in the decimal system to belong in

...

the reference position in sentence frames containing hedges. " The

substantive implication of this result, she suggests, is that in a

natural category such, as numbers, not all members are equivalent.

Certain members of'the category serve as reference points to which

other members are related. The relevance of this finding to the

present study is that Rosch's number stimuli (that is those judged to

89
 Chapter 3

be reference points) were all multiples of ten. I have already

suggested that multiples of ten are among those numbers available for

making n or m approximations. They are are also frequently found in

other approximations. These observations allow us to go beyond an

intuitive judgment that certain numbers are "round", to the point

where we may suggest that the expression "round number" designates, as

far as ordinary English is concerned just those numbers which, in

Rosch's terms, are reference points in the base ten number system. To

take this (speculatively) further, if we imagine a mental lexicon

which utilizes these reference points in its organisation, we might

expect the reference point numbers (RPNs) to be more "easily"

accessible, and therefore perhaps faster to interpret. If language

processing uses what Yorick Wilkes (personal communication) has called

'the laziness principle', RPNs might require less processing than

other numbers, and therefore be suitable for use in approximations.

Coincidentally, when I asked some test informants why speakers would

use an approximation. when they in fact knew the exact number in

question, they said that it was "easier".

It may be seen that those numbers which emerge as RPNs, do so in

part as a result of the structure of the base-10 number system. The

whole system is generated by a small set of recursive rules. Thus

Hurford (1975) was able to propose grammars which would generate all

the number words from a fixed set of recursive rules, plus some

phonological exception rules.

Henninger (1969: 10) suggests that the structure of the number

system has perceptual effects on how we use it. He thinks we access

RPNs by grouping, and other numbers by counting on from an RPN; so

that 1000 is "found" sooner than 543. He writes:

90
 Chapter 3

"Even today, if we contemplate the number sequence in our

minds, 1000 seems clearer, more "available" to us than 543, a
number of which we can really say only that we must certainly
arrive at it if we keep counting long enough. We can
visualize it directly only with difficulty, if at all. " (: 46)

If this is right, it accounts for the test subjects believing that

approximations containing round numbers are "easier" to use than exact

numbers.

But it is the case that the numbers which function as reference

points may be determined by other factors than the structure of the,

number system. Menninger notes the importance of 12 in Western

European Cultures. 56 lbs (a half a hundredweight) is an RPN

quantity while we still use non-metric measures, as is 8ozs. Again,

six feet is an RPN quantity in Western Europe and North America, when

talking about people's heights (but as Sadock points out, it is not

perceived as such if we are talking about the height of cockroaches).

A different kind of evidence for the importance of RPNs in

communication, and a good indication of which numbers they are, 'comes

from word frequency counts. Johansson (1980) looked at the frequency

of the words for numbers one to twenty in four different corpora. His

findings are reproduced in Figure 3.

91
 Chapter 3

The frequency of the forms one to twenty

LOB Brown Carroll Jones and

et al Sinclair

one 3,203 3,439 19,976 775

two 1,652 1,516 10,085 251
three 787 704 4,413,157
four 442 417 2,357 87
five 334 385 1,725 127
six 269 247 1,229 92
seven 140 132 687 51
eight 121 123 651 45
nine 100 103 417 31
ten 203 185 1,225 100
eleven 49 40 146 17
twelve 58 49 339 32
thirteen 11 11 116 0
fourteen 22 34 93 0
fifteen 48 61 23 23
sixteen 27 21 104 17
seventeen 21 27 59 10
eighteen 24 18 68 13
nineteen 14 21 40 17
twenty 189 174 387 66

Figure 3(from Johansson, 1980)

He notes that there is a pattern of decreasing frequency from low to

high numbers, except that:

"The consistently decreasing frequency with higher numbers

is, however, broken by peaks for 'round' numbers (10,15,20
etc). There are also one or two other deviations, in
particular a peak for twelve and a slight drop for 13. " (: 69)

He notes that in Jones and Sinclair

"there is a peak for five, which seems to count as a 'round'

number in spoken English (: 71)

The RPNs are very apparent, and include 12.

RPNs are also important for approximation because they can be

used to approximate on their own, without a lexical approximator, as

we saw in Chapter 2. Wachtel (1980) has some discussion of these

uses.

3.6.3 Reference Point vs Non-Reference Point Approximations

Number approximations often contain RPNs. What happens when they do

92
 Chapter 3

not? Some examples:

(48) [doctor/patient interview]

+I: When were you, how old were you uh (pause) when you were
in Japan? When you were in Japan at that time?

P: (mumbling and unclear initially) I'd say three years ago

I: Okay so you were about age 31 [Cicourel, 1974]

(49) [article in The Guardian on the history of the death penalty]

+The records, though far from complete, show that about 61

people died on the block

Now in (48) the doctor's use of about reflects the patient's

uncertainty (indicated by "mumbling" and "I'd say"). Neither of them

know exactly how old she was (cf Chapter 5, Section 5.4.4). But

supposing it was three years ago, 31 would be-right rather than the

nearest RPN, 30. (Although years can be subdivided, we do not usually

divide them when talking about people's ages, so the effect I describe

below relevant to (53) and (54) is not present here. Of course we can

imagine a situation where years would be seen as divisible, for

example a discussion of children's ages for examinations. ) Notice now

that one would tend as a hearer to understand a different (longer)

interval in:

(48b) Ok so you were about age 30

than in (48) as it was actually said. (48) is a narrower

approximation. This is shown also by considering the invented

(50) It's going to cost about £25 to have the exhaust done

(50') It's going to cost about £26 to have the exhaust done

where in the same situation, the interval for "about £26" would appear

to be understood as smaller than that for "about £25", an RPN. I am

arguing that whether an approximation contains an RPN or not affects

the length of interval understood. A different account, proposed by

Sadock, is that the number of significant figures in the E-number

93
 Chapter 3

alters the meaning, such that

(18) [111 Odessa has a population of approximately one million

(19) Odessa has a population of approximately 990,000

are judged as designating a wider and a narrower interval,

respectively. He further suggests that the relevant interval can be

obtained by subtracting fifty percent from, and adding fifty percent

to, the last significant figure (: 433). This would give an interval

of half to one and a half million for (18) but only 945,000-1,035,000

for (19). This appears to be quite a good explanation. The entry

referring to Mount Everest in the Encyclopaedia Britannica says that

it has a height of "29,028 feet". The presence here of five

significant figures has the'function of telling the reader that this

height is being given to the nearest foot. Compare the heights given

in the accounts of particular ascents, for example:

(51)+ Norton went on to 28,300 ft, a height unsurpassed until 1953

(52)+ Finch and Bruce reached 27,300 ft

which I think have the interpretation "approximations to the nearest

hundred feet".

However, Sadock's account will not deal with (50) vs (50').

Neither will it explain an effect which I think is present in (48) -

X (49). In each of these, hearers receive, I think, 'a strong impression

that the figures given are the exact figures involved to the best

knowledge of the speaker or writer. The tendency is to suspend the

conventional approximation reading, and understand rather a self-

defence ploy : "this is the figure, but just in case it is slightly

different, I'm using an approximation, so that I'm not wrong". I

discuss. this effect further in Chapter 5, Section 5.4.5. The point

here is, why does the effect arise in these examples, but not

apparently with RPNs? I suggest that hearers have a very strong

94
 Chapter 3

predisposition to expect RPNs to follow approximators. When they do

not find them, they judge that there is a special reason for it, and
,

read off particular implicatures (ie, there is a violation of the

Maxim of Quantity). If it was not the case that RPNs have special

status and that approximations usually use RPNs, this effect could not

be accounted for.

There is another effect which arises from using anon-RPN. This

is described in Wachtel (1980: 204). Compare

(53) Sam has $10,000 in his savings account

(54) Sam has $ 9873 in his savings account

of which Wachtel says: "10,000 is a round number and 9,873 is not, if

we are measuring Sam's wealth to the nearest thousand dollars, but

9,873 is a round number if we are measuring his wealth to the nearest

dollar, his account may actually contain exactly $ 9,872.91. "

[It must be remembered that for Wachtel "round number" appropriate

-

exemplar number, in my terms; it has nothing to do with RPNs]. Now,

what triggers the reading of 9873 as expressing a measure to the

nearest dollar, is that it is a non-RPN. This reading of a non-RPN is

different from the "self defence" one suggested for the earlier

examples. I think this one arises in this case because dollars are

perceived as things which can be subdivided into smaller conceptual

iý -

units (=cents). If a non-RPN is used of something which cannot be so

subdivided (eg people, as above) the self-defence reading is favoured. I

Wachtel's paper takes no account of these differences (cf Channell,

1980, for discussion in the same vein).

3.7 Implicatures and Entailments of approxitnator +n structures

In considering several different approximators, we have seen that the

effect of using an approximator is to designate an interval of

95
 Chapter 3

numbers, any of which could be the exact number in question. In his

1980 paper, Tom Wachtel makes some interesting observations about the

likely implicatures arising. from, and also the entailments of,

sentences containing approximators.

Firstly, a sentence containing an approximator has the

im-plicature (Gazdar's term for a potential implicature not

necessarily realised in any given case; Gazdar, 1979: 55), that the

exemplar number is not the number in question. For example, the

assertion of example (41), repeated below:

(41) [questionnaire on student housing]

+The response was again poor - only fifty replies from the
1,500 or so questionnaires which were
distributed [written SO 4.12.81]

implicates that not exactly 1,500 questionnaires were distributed.

Furthermore, as Wachtel notes, this im-plicature can be explicitly

cancelled (Grace 1975), or reinforced (Sadock 1978), cf

(55) 1,500 or so questionnaires were distributed - in fact

exactly 1,500

(56) 1,500 or so questionnaires were distributed, but not exactly

1,500

(cf Wachtel's examples, 203-204)

Another im-plicature is "speaker does not know how many questionnaires A

were distributed", and this can also be cancelled by (55) or (56). '

Secondly, an exact sentence always entails its equivalent with an

approximator; that is

(57) Exactly 1,500 questionnaires were distributed

entails

(58) 1,500 or so questionnaires were distributed

A third point, not noted by Wachtel, but which follows quite naturally

from this entailment, is that an approximation like (41) cannot ever

96
 Chapter 3

be used to assert (57). An approximation of this type can never,

whatever the context, be interpreted as asserting that exactly and

only its exemplar number is the quantity in question (although a

hearer may infer this, as we saw above in relation to approximations

with non-RPN numbers).

Notice that this is not the same thing as saying that the fact

reported in (57) is not compatible with (41). Precisely as we have

seen, from the test results and other evidence, the quantity 1,500 is i
among those for which (41) would be Appropriately used.

3.8 A Note on Partial Specifiers

Not less than is one of a set of expressions which specify upper or

lower limits (Wachtel's term is PARTIAL SPECIFIERS) for quantities on

the number continuum. Others are at least, at most, less than, (not)

more than, under, over. These contrast with the approximators under

investigation here in that they do not apparently involve the cut-off

point problem which is the focus of interest of the expressions

discussed so far. As noted by Wachtel (1981: 320), in terms of

truth-conditional semantics,
i
(59) She was wearing a dress costing not less that £500

[test item 11]

"is not false even if the dress cost £900 or £3,500".

I had included two examples with not less than in my test to

provide a deliberate contrast to the other test items. I thought that

people would treat these expressions differently. Contrary to my

expectations, however, many subjects indicated that not less than

expressions designate intervals which extend below their exemplar

numbers, 27 and 24 percent respectively for the two items tested.

97
 Chapter 3

Furthermore, contra the contention that there is no upper cut-off

point for these expressions, subjects gave consistent upper limits.

The size of the interval designated by not less than appeared to

be the same as that for about. For example "about £500" produced a

mean interval length of 51.92, and "not less than £500", a mean

interval length of 54.62. The important difference is that for not

less than the mode in each case is higher than the exemplar number

rather than the same. That the majority of subjects place their

interval for not less than above the exemplar number is shown by the

resulting positively skewed histograms (see Table 3.6), which can be

revealingly compared with the symmetrical ones for about/(a)round

(Table 3.3).

Apart from having failed to show a result contrastive with that

for the approximators, the effect noted raises a distinct question

mark over the suggested analysis of these partial specifiers, ie that

they have a specified limit in one direction and none in the other.

Could it be that these should be dealt with in the manner of the other

approximators? While partial specifiers (if such they are) have not

been singled out for close attention herein, a short digression on the j' t

above question is justified, inasmuch as the relevant arguments are

closely related to (or the same as) those pertaining to the

established set of approximators. In addition, an answer to the

explicit criticism'of my approach made by Wachtel (1981) is required.

So there are two problems: (1) subjects judged that the intervals

could run below the point marked by not less than, and (2) subjects

judged that there was an upper limit. I shall first consider these

separately.

98
 Chapter 3

One possibility is that the test result is spurious. Wachtel

(1981: 320) suggests this, claiming that subjects were misled into

judgment (1) by the use (in both items) of natural round numbers,

which are habitually used alone to imply symmetrical approximation (cf

discussion in Wachtel, 1980, Channell, 1980, and 3.6 above).

Furthermore, the experimental set across items could have led them to

erroneously judge these two items as more like the others than they

really are in non-test uses. Wachtel takes this result as evidence

that one should not rely on the results of elicitation tests to give

indications about semantics.

Let us now consider a non-test use of not less than n.

Unfortunately this was not collected verbatim, and so is to a certain

extent reconstructed. I telephoned the garage to book my car in for a

rear brake overhaul. To help myself come to terms with the eventual

bill, I asked for a verbal estimate of the likely cost and was told

"not less than £60, more if we find anything else wrong". Now,

suppose that when I collected the car, I was told that the cylinder

seals had not needed replacing, so the bill was only £58.50. I think

I would be unlikely to challenge the garageman with the falsity of his

estimate for the job. That is to say, I would be making exactly the

same kind of judgment as we have seen applying to classical

approximations - not less than £60 would not normally be considered

false, say, for amounts between £57 and £60. cf the formulation of

Wachtel (1980: 204): "(14) would not normally be considered false, or

even misleading, if the actual population of Odessa is, say,

1,002,493".

99
 Chapter 3

Another example shows that the point "specified" by the partial

specifier does not have to be exact. In this case, the point is made

vague by an approximator. If the point mentioned by a partial

specifier was perceived as a fixed limit, (60) would be some sort of

contradiction, or at least should sound odd:

(60) +She's not old enough, rubbish. Well she's only about 45
at the very most [Camb 1B267/p3]

If these examples are typical of the use of partial specifiers,

then the result obtained in the test may be seen as a true reflection

of the meaning attaching to them.

It appears therefore that in this respect the evidence which will

constrain our treatment of the meaning of not less than n is exactly

the same as that which will constrain our treatment of the

approximators. Note also that the same sort of pragmatic factors as

have been mentioned previously also influence these meanings. Not

less than n will receive a very different reading depending on whether

it is being used in an engineering shop, or to talk about dress

prices.

The second point was that subjects judged upper limits to apply,
.
in direct contradiction to the usual view that no limit is involved in

these cases. As pointed out by Wachtel, what happens is that hearers

draw on such information as the purpose of the utterance, and

assumptions about the speaker's knowledge, to determine the extent of

the interval of possible numbers in a given situation. Examples show

these factors at work.

(61) [Customer telephoning builder]

+B: Could you give me some idea of how long it would take?

C: Well the quote might be done within three or four days

but the job won't be done for at least five weeks [Carob
111B1058]

100
 Chapter 3

I think it is likely that the customer in this example would decide

that a false or very misleading estimate had been given if no work had

been done after twelve weeks. One normally reads at least five weeks

as "in an interval including five, but more likely over than under".

Here, the situation appears entirely parallel to that for the

approximators, and just the same kind of pragmatic information is

involved in deciding where the upper limit comes.

My conclusion (contra that reached by Wachtel; 1981: 320) is that

partial specifiers are actually used to approximate in very much the

same way as the approximators so far discussed.

3.9 Summary

In this chapter I have presented a considerable quantity of different

data concerned with number approximations of the approximator +n

type. I have been especially concerned with describing the meanings

which hearers attribute to these expressions when they are used.

Before turning to look at data concerned with another set of

vague expressions, I summarise below the Propositions about Speech

Behaviour relevant to the meaning of number approximations, which a

theoretical analysis must account for.

1 There is a set of expressions whose effect is to bring a vague

reading (= an approximation) to an utterance containing a number

2 The resulting approximations are understood as designating

continuous intervals of numbers

3 Different approximators change the interval designated. Given an

Exemplar-number x, (and setting aside the effect of other factors such

as situation, and the form and nature of the E-number), the effect of

the different approximators (as observed from the test results) can be

101
 Chapter 3

seen schematically as follows:

about/(a)round [--7-- 7.62

32.30
----------

n or so 15.46
------- ------ -------56.69

norm 22.86
---------------- ý--- - -- -- ------------ 66.60
= minimum

-------= maximum

Figure 4

4 Although there is a high degree of agreement among that

speakers

numbers near the exemplar number are members of the interval, there is

variation about the extent of the interval in any given case

5 The size and form_of the exemplar number both affect the length of

the interval

6 Whether the E-number is a round number (- Reference Point Number)

or not affects the length of the interval ("about 31" vs "about 30")

7 The nature of the item being approximated (discrete vs

non-discrete, man vs cockroach) affects the interval

8 The conversational setting in which an approximation occurs affects

how it is understood

9 Sentences containing approximators characteristically have the

entailments and implicatures described in 3.7.

102
 Table 3.1

List of items used in number approximations elicitation test

1 "After the girl had rung up everything I'd bought, I suddenly

realised I only had about £2.00 on me"
2 "You'll need about 4lbs of oranges"
3 "We should be there around 6"
4 "He's producing about ten pages a week and they're all
getting published"
5 "There were about 15 people there"
6 "... it's something around the 20 percent mark and it's
never changed"
7 "We sent out two hundred questionnaires and had about 40
replies"
8 "It's going to cost about £500 to fly there and back"
9 "I want to spend about £14,000"
10 "In a country with a population of around ten million, like
Belgium, proportional representation makes much more sense"
11 "She was wearing a dress costing not less than £150"
12 "The repair bill certainly won't be less than £500"
13 "It was a good evening, we must have drunk a bottle or so
of wine each"
14 "How much flour shall I put? " "Two spoonfuls or so"
15 "Six or so books will be enough for a week's reading"
16 "There are ten pounds or so of butter in the freezer"
17 "Ten or so litres of wine should be enough for the party"
18 "They hired
the de Grey rooms and invited 200 or so people
to a champagne lunch"
19 "It's okay, I've got £500 or so in my account the
at moment"
20 "It'll cost two thousand or so pounds to do this place up
reasonably"
21. "There are 3000 or so students at York"
22 "The Tower of London gets 30,000 or visitors a year"
23 "You'll need three or four metres of rope"
24 "Yes, but it's still around that four or five region"
25 "It takes six or seven hours to drive from Paris to the
Midi"
26 "The garden extends eight or nine feet beyond the true
boundary of the property"
27 "Eight or ten students were waiting in the entrance"
28 "How many people will turn up for the meeting? "
"We usually get fifteen or twenty"
29 "They had seventy or eighty people with broken bones over
just one weekend"
30 "He's bought a stereo costing three or four hundred pounds"
31 "Two or three thousand people turned up to hear him speak"
32 "A burst in the Hull Road flooded
water main neighbouring streets
with two or three thousand gallons of water"
 \x Irt x e\° xxXxx
.C c1 O N1 N %IG O
41 0 .t
00 ý .......
Oý N O tý ý11
tr1 tIý .ý . -1 ýt
4) 14
w
Cd
,-4 o r4

Cd 8 aý
>ýnd 0
ý+ cý k 0
oý m
,. ý 0
41
H
w
fA
d
CO
0
0 to
iJ
um
. F4 r.
0000 o .ýOo". 4 " 4:
0 0000 r-4 oýOo 0 r+
"rl Qu
b OOO {d
N 0 d
.00 ýi ý0 ý N O 41 C!
x r, ýt
tr1 .10 ".4 a
ri
4+
.cV0
Cd 44
d
b
co G) }I C+
41 0 t0
ý w 0 44
0 w
ý N
aý d t- co 0 %D U'% 0M0 KM '17
tn u c% OO c% O o .-1
%. 0 %o 0
.........
O
o,w "ý
ý N
Co CVI N
.-ý ýt OM oa
cý ca u1 O
d w Oi
ýa
aý ý oý
QN
u
ri
aý
cti
"ý ý
NI ;; I 4 0
41 0 dý
e+1 m on .,., ",4
dl 9-4
Hý p CO M tf1 r- N0
d
aý .-1
N N
%D
O% rn M .,4
r4
r4 c+1 %D u1 %D
.,ý uý
d pl
......
c"1 U1 tl% 4)
t0 0 9: .t .t 0-4
tf1 %D u1
a
H ", I co mu
41 aý 0 0
ý 14 ra
8
0
k .+
O .0
1+ m
aý u "ý,
a Uý
ONN 0N
Nu
o0cb O N t7 %D cý d
dl M1
d . -I
"n p7 ...
"P4 O crf ýD cv ODD ýt O oý
d .o
ax d Oý rn oN O O. CO O ýD 0
.n
ap W Lj . -i .-a C7. 'b 3
Coa A 4+
s x0", + co
z LA
0
Aj
w ww
0
0
j
ý o0
Q) 07
.,.
Co rt
.,a ý m "ý+
0 o ý ßb
a) a
OQ 0 N r4
ao .a
d o c0 a) do 0 ý ý"
o ý od a 0 Ho r4 0
V-4 to O 11
a+ .C 0 0
,oa 0 41 C u1 u al
",ý tV . -4 %C Uli . -ý .. 4
Ai tat e .4 .-t .*u: tat Wu

i
'O b b V-4 Co
" ,4 to . -+
d
OOOO0 0000 0 fA ,nP.

i
waan LJ "O
. L1
H Gw .Gp .+
Co ta to ca Co cu Co ' Co 0 cd ". + i1 c0
q co p
D 'til "ýc
r. U
"" O) c0 d
tA C 41 'L!
0) H cn
Ai

i
N trl .o r" 00 a. o O
-o -4 -7 ºn zýN
z
 KK
.+v..
%0 o
V-4
KKKKKKKKKK
ap

1;
ý ri

Oý
! "f
Oý
MN
Vº

r1
N

1;
"+
ri

11
N
Y1

W,
. -.
tf1

r
N
-+ "+
ii K"
e

.ý
cý1
.o

.c
%O
>e
ch

c7
N
W4 W,

O p 0 tf1
in N
u,
ppp .ý
00 . -1 N0
00 ýýý
ýD N
ý -I pppp 0
ýh p an Np 0
N t+1
04 0
.
0
M

.? O Co O
.. Co cm .O %D OOO
O rh O
eOO e"1 N vº .Dd o% .ofl-
ý
O Oý OO tn rý a .o
N
%D
c.
N
.. N
ri ra O ND 0 .1N 1,
ýN N v'1 r"1 1 rr
N rn 0
coi

Co Co
N ,p Co
v, an N u1 O o .co 00 o "4
u'º
SS
v, 1ý e7 "-+ %D "+ cý oýo r+

O co"_ cn r4
41 t1NN.. 4 .o r- .c ,+
04 u1 C. r ... .. ý
NNý

-ýr

O0
NOO
OO
NQ NNO V1 N1 OýON
."........ 0
0
%D 01 fD %D 0 Co t4 Co %D %D .c00
O% 0 C% CY% OW C% O aa v v. v. 0% .00
-1 r-1
r1 "4 +-1
-4

to
Co 14
aý O
00 O a1 "ý
vº o in 41 'd 0
.. 0 rl
O ./
b ww
al w +º 9
m O 4.
ý ae eOaN0ý
d00 o) 0m
C ac O+ý ý "ý ý
M N to 0 o-"+ LL mO ro
tx a+ oo ý
V 41 Aj
o .+0.. 0m 1+ F+ 0) 0 0)
C " 40 in l+
-f% 0 Ei w .
O edw-O0m0º.
U c ee.. C0 10 ÖO0O
ý {I O 41 000-^ 1a e u, .o
N sa+ 41 00 s+ 0-vý ý--
-- wN $4
c. s,. .doo cý ö %.
0 000
ý"t u ne000
0 a1 "" OOO V1 N0
OO O)
m
e Cße ý0 N^ýD "-ý "ý N ut W a1 t+1 cm t .c
v
. -a
1] .. 1$a L+
41
0
V
E+ 0 ! +. I. n .0t. 00 0.0 "-+ N M V\
01 -N q -.
-1 4 NNN 0 NN
Cý;, rl .ý... rl
ý ZI -+ -. . "ý
 Kýý KKKK
er .. + O o. ao .t*+
.
.oo, nN in Oý
NdJNN v1 41
-?
C
.. ý
0

0
.. ý
Co
a
8
ao 0
0.4
ldl% V
r
1-%.
n d
t"
a. Co "+ .n -ý 00 aý
o
MN
Y, o
tn
ýp
.i N1 NN
ý

d
x
Co
iJ

0
ýe
C
.. fý- ., Sn . Q'.
.c ao .o o\ -+ l"! O

....
ao az t: ca
A .. .a
m
WNs oý
., r- cn U-1
NN I^ Ci

s+
ý
a
d
x
e'7 N e°1 u1 ro OO
d
N ýt N w4 S %O W.4
".... 0
N. t a0 n00ý
"-+ aý .oý
Ný
.i rl d
x1
,Z
OA-4
0
8
r.
Co ý
NOOOOOO d 0
....... 8 u
.0000000 . ý.
000000
o. V4 ar
r. . ". V-4 r. .. .r-
Ai
cc
m
vO .a
m . -+ o
va .+ .4
a0 "+ Co
zv ý
Oa eo w
d
ý+ e o. .. a
'o umm -o v x
mC veF 0 H
ýmaama .+.. at
evOO
Mu0m w" e
m'C 7ý
Co
++ oup. ß. G0 O Co
Gme7S
,ý
0WOOcy
uON ao
s ý
N d
ý?º . -ý ý t'1 e"1
N Lý N 0 x
OO 1+ I+ 1a ý
0000 O
vº O w
mW ap .r fý N1 NN
.. ý. 0
m0 w
dW
u
H aD n CO a p- eV 0
vNNNN l" 1 99 ! '1
 Table 3.3

Frequency Distribution of Informant Responses-for ABOUT/AROUNDApproximations

Note:

The score at the mode is often 25/26 informants. This is almost

always due to the divergent performance of just one informant who
consistently gave a low interval relative to each exemplar, or
no interval at all (cf note to item 19). He apparently had his own
theory of approximations which he was putting into practice. In
discussion after the test he said that he thought approximations
were always used for exaggeration. A second factor which could
account for his different responses is that all the informants
except him were in the age range 18-25, while his age was more
like 50.
 10
N

..
a

ý,
N

0
--I ::

ýo
A2

ý, -
Y
OD
d
IA

O"
a

c
G .nO

v
H-
[
ý O
0
ý n" ý,
JO"Oý
-' p .np .n NN .+ r'
a m
° d
m
. m
a
:
t

ri
ai
,O
cu
F

s
-0
ý r

0
0
ý C

C.

O^
C ýO
0
ti O 1

.a
N

r.
ý
q C^
.7
0

C I^ ýfN

rv hýý
 c1
ý

a
ý
b
a
ýs

-__- 1,,
ý
__- a
.3

ý
ä

a
r N
G

m N

.O
N
.

9 lý
e
_4 -
ý
In
en
ý
ý In
O
X a
n_
29
ýý ý
nO'O^
NN^^
CM
V.,
ý

0,
04 cn
tt
a,
N ý
ý
N
-4 N
.n
ro
N
N
H
N
b'ý

ý
O N
N
ý

O
OP N

0
ti

r O`
.+
ýY
a
pY
---1 ~+
ý0

co
N
-
L m
d
a U,

nr
V .. c
0
C
V .)
G

`
Y
71
-m
ý.
-t
V
ýý .. 19
°c
N
0 ti

ý
Ic ý
ý rn ý
0
CO Oý te
.. J N
NN . -.
. /ý
 ýý
N

ý -2

- _. y O

ol
o ---_;
-10
ý

aÖ

v f
co
0
16,000

N fV rC 15,750 ý

.. +
15.500
.,. ý
15. °_50
C)
15, a0J
.ý
M

M
14 '5p
d

_- .. .5D0 .0
co
Hi
14,250

14,000

13,750

13,500

13,250

13,000

12,750

12,500

ý 12,250

8
N Y. 1
12,000
O
7

C
O .0
" 11,750

CýCý

0 .n 0 V.' P ý
r. N rr .r

m r« N
 Chapter 3
Tables

Table 3.4

Special results for n or m approximations

subjects % subjects
Item specifying specifying % subjects % subjects
no. interval cont. int. bounded specifying specifying
both e-numbers by e-numbers nos below nos above

23 57.7 76.9 30.8 73.1

24 100.00 57.6 50.00 46.1

25 69.4 88.5 42.3 80.8

26 88.5 73.1 46.2 65.4

27 96.2 77.0 61.5 57.7

28 92.3 77.0 65.0 69.2

29 80.8 88.5 57.7 65.4

30 69.2 61.5 30.8 53.8

31 69.2 76.9 41.1 65.4

32 73.1 76.9 57.7 57.7

Mean over
79.6% 75.4% 48.8% 63.5%
10 items:
 N.

ý
.,
fV 4b
rti

ý 2
E0
..
o' ý
..
N

..
-1
ai
r1
0
O ",a
:S
r+ 4-1
Y cli
O
`o E
O
.0
.
C7, ., ý
O
O
I. E 0
ý4
Ö
ý
Co

M
x M
O
.+ ý
ýIl 0
W
O O vý ýo
ý
.r
ý+
tV fV
ýn
.-. ...
N
0O ry .r In x
0
z
ý4
0
w
y
a,
0
Ln
c+1
ý

4-1
JD G
ct
E
w
0
w
G
H

44
0
G
0
N
"rl

"rl

1J
Cl)
"rl
iý
s

r-= ý

F U

.ý
M a

---r M
ý
O
U
i W
~

C
0
r,.
A
" m
a ý
L
O M
ý C
O
ýý C
1ý0
ýý .+f-
6
e
N

d
O ý O II
3
G
I

'1' o 2 0 ýý
nn0 "n 0r s
NN ^ NN
'+
 0
rn
fV
0
0
co
ev

0
n
N

ýC u
a
N
a
O7
ýd
C 0
M

_0
C ý
0
N

0
8
N

0
a
I

0
,°'
-1

ý r=
a
0
u 0
a
0 O
m ,n
M 4n
O

0
8 0
y
tn
'iý
__ii .. O
ý
In
co Ln 0
N
ý. Q ý
ev
.,.. +
0
N
ý

0
r

80 Lr)
ý U')
OW
cn
rn
Tý
CD ,_.a 1-4
Co r. A
JON
F'
0
vN

o
ýo
Ei

sU
0)
0
ä
'ý
:i
O fA
ý
1J

oQ Co
f4
v. H
ý
oN
ä 'b

8 aý
v u
0 0
z
A J_ n
ý
ý
N 0
N 2 0 -1
m
ý
u
I.
ý
.H

O
C

I.
0

2
,. y ý
N
0
N ý ý m
v
N

.ý
 n
n

fV
R1

.n
ý
_D
.t
en
ý
N

L
O
M
M '°
N

O
N II"
N
m
ý
Ln

ýN
ýý
_ý 1-1-_L-
CN V1 0 0ýNNN
. -ý-i .dý
ý
co
a
N
§ H

Co
r4

I
ý
N

0
eV

U%
N

0
In
er
a

M
ý ý
eV

s
N ý
N
ý
ý O
O
~
lV h
n
--H 0
pew -a+ ----ý in ý,
er

oý v ----- -
.-. ý -
--
°
co o -----_ý - '^
-- --
... - --- -ý
..
Oý "g
.aQN

0OO "
P1 N

a
In V1 OýO^
8
4. d
..
t, l rN-
 0 O []
n {-y
er ol -
0
Y

m
d
ý
c
Y.,
cC
u

ry0 i
CD
oy '+ m
Oýw
§

0)
ýn Oa
mý

Io1
ýoö
ýn
ý.y- CCýýII ý
Y1 ..L +
ý
C
v 10
C N ýI'y y
] I
O
U1 ý
C. ". ý

v
T t ý1 0 1R

W cU
oýoc
w oýam
C
C O pý
C

'O
u, w a.
u
V
O oyo a
C .. jr.
]
t d
>-
N "ý -o
ý
a ý+
C .ýü, c
C pý --
ýt$
ý ý

C m
m mO
L
rýa 1

.+
O
C ýO Cý
ýý .yýI1I I__L__ J

W N vl O ýn Oý y+
ý
N
C .,..

ýN
II

G
 Chapter 4

Chapter 4'

Vague Category Identifiers

The second type of vague expressions which I have chosen to examine in

detail 23, exemplified in the following: I

(1) +There was no kind of social contact - there was no

coffee room or anything (11/21.2)

(2) +But what about things like when you read sentences or something
and then you're asked to reproduce them or something (1/13.2)

(3) +You can remember four lots of four fairly easily say in the form
of dates or something, 1972 or something like that (1/13.2)
(4) +Lady Arran, who pays for her hobby by 'selling off silver and
things' set a world record at Windemere last October (DS/1T

(5) +1 hope we didn't have lots of horrible conversations when you

went out of the room in tutorials and things like that
(11/14.3)

The structures in the underlined parts of these examples will be

referred to, following Dines (1980), as TAGS, because they can be

"tagged onto" the end of otherwise more precise statements. They are

in the category of vague additives, similar in effect to the number

approximators just looked at. The structures which I shall consider

here are the following:

or something or anything

or something like that or anything like that

and that

and things

and things like that

What are these tags, and how are they used?

In this Chapter, I look in detail at their distribution and

structure (4.4 and 4.5), I again draw on the results of an informant

test designed to investigate their meaning (4.2 and 4.3 ). Intonation

is important here in showing how much of the preceding string the Tag

103
 Chapter 4

is operating on, and this is described in 4.6. In Section 4.7 I

describe some co-occurrence restrictions which I have observed.

Finally, as for the number approximations, I summarise (in 4.8) the

propositions relevant to their meaning which a theory of vagueness

must account for.

One suggestion for the function of these tags is that they are

purely performance 'fillers', introduced to give both speaker, and

hearer, additional time for processing. While this may be the case

for some occurrences, it is obviously not always so, since removing

them makes a difference in meaning :

(6) You can remember four lots of four fairly easily say

in the form of dates 1972

,

This sentence has lost something which (3) had, and needs additional

paraphrase to recover it, cf

(7) say in the form of dates, like 1972

...
More convincing than juggling around with examples is the following

extract which neatly shows that or something is far from empty:

(8) [Context: A is going to the shops. B asks A to buy him

some bread. Just as she leaves]

A: So you'd like some bread?

B: Or something. Anything edible will do. (DS/8/11.9)

If or something were a filler, B would hardly use it to modify what A

had said. He is giving additional information. He then clarifies

with "anything edible will do" which I take to be a gloss for bread or

something. Further evidence that these tags are not empty is

presented in Dines (1980) who cites an investigation by Brotherton

(1976) in which "the majority of her lay judges classified all the

terminal tags as qualifying elements". Dines says "This data suggests

that terminal tags out of context are not considered redundant by

104
 Chapter 4

middle-class speakers but are salient as elements which qualify the

content of the utterance" (: 19).

What then do these tags mean? My intuitive analysis, which was

the same as the suggestion made independently by Dines (1980) was. that

these tags "cue the listener to interpret the preceding element as an

illustrative example of some more general case" (: 22). That is to

say, the preceding element, which I shall call an exemplar, directs

the hearer to access a set, of which the exemplar is a member whose

characteristics will enable the hearer to identify a set. We may note

in support of this that speaker B in example (8) goes on to gloss

"bread or something" as "anything edible", which suggests that "bread

or something" is to be understood in this context as referring to the

set of edible things. Thus an exemplar + tag construction fuctions as

a (vague) CATEGORY IDENTIFIER, and this is how I shall now refer to

them.

Duncan and Fiske (1977) classify or something as a "sociocentric

sequence", in which category they also include but uh and you know.

They cite Bernstein (1962) as the source of this attribution, but

although Bernstein does use the term in his paper, it is nowhere

defined, nor are examples of what it covers given. Duncan and Fiske

say that these are "stereotyped expressions" (: 185) which "when used,

[. typically follow a more substantive statement" (: 171). I am

. .)

worried by the glibness of these assignments of value (what counts as

"stereotyped" in language? Do they mean 'idiomatic', or 'frequent'?

What would a substantive statement be? ). If by substantive statement,

they mean 'some language string which does not have explicit

vagueness', then it is certainly true for the vague Category

Identifiers. Their exemplars are not vague. But Duncan and Fiske

105
 Chapter 4

should I think have looked further at the meanings of the various

expressions they consigned to the category "sociocentric sequence".

As far as or something goes, this will now be done, and it will become

apparent that it is rather different from but uh and you know.

4.2 The Elicitation Test

4.2.1 Objectives

The test had two main objectives:

1. To find out what the Tags mean, and possibly, to confirm

the hypothesis outlined above

2. to get some indication of the effect of different exemplars

preceding the Tag

4.2.2 Method

A paper and pencil test was used to present 31 examples of, tags (one

item contained two different ones). These consisted of 15 attested

items from Rosch's (1975a)

examples, and 15 constructed examples using

work on prototypicality within categories. The reasons for

constructing examples from Rosch's data will be explained below in the

general discussion of the results. A pilot study was first carried

out using four items, in order to see whether subjects could perform

the test, and as a result of this, certain modifications were made,

for example, providing longer extracts in some cases. Table 4.1 gives

the complete list of test items with their reference numbers, showing

in the item that were The test

each case the parts of underlined.

contained three attested examples of structures which were not

eventually selected for detailed analysis herein: no. 5 or anything

else, and nos. 10 and 14, or whatever. They are however of interest,

in that they appear to be interpreted similarly to the tags under

106
 Chapter 4

investigation.

The subjects first read the following instructions:

This is a very simple experiment to investigate the meanings

of utterances which occur in ordinary speech. For
convenience they are presented in written form, but you
should try to think of them as spoken.

Now read what you have to do. DO NOT START UNTIL I TELL YOU
TO.

On each page in the answer books is one example of a spoken

utterance. In each case, please list at least two items
which the speaker could reasonably have been thinking of when
he/she said the underlined part of the sentence. Imagine you
are trying to understand the speaker.

When I say 'turn

to the next page', read the sentence and
write your answer as fast as you can until I tell you to turn
to the next page again.

Now we're going to do a trial answer, Please do not turn, to

the next page until I tell you.

The subjects did not see any example answers, and so were free to

interpret the test as they chose. I was anxious to avoid any

suggestion as to the kind of responses I expected. The test items

were presented in reverse order to half of the subjects in each group,

to compensate for any practice or fatigue effect across items.

Subjects had a timed 60 seconds for each answer, and they had a

resting time between each answer book of 10 answers.

4.2.3 Subjects

Subjects were drawn from three groups:

1. first-year language students at University of York 18

2. arts students at the College of Ripon and York, St John 6.

3. sixth-formers (age range 16-18) at Nunthorpe Grammar School, York

and their teacher (age 53) 30

Since, for timetable reasons, it wasn't possible to test the Nunthorpe

107
 Chapter 4

subjects on the complete 30 items, they were tested on alternate

halves of the test, and their results matched in pairs to give the

equivalent of 15 subjects, and thus, a total of 39 subjects' answers

was analysed.

4.2.4 General Approach to the Data

The following terminology will be adopted: 'answer' will refer to the

whole of what any subject wrote in response to any test item;

'response' will refer to discrete parts of a subject's answer. There

were two distinct types of response:

1 single-word-responses (for example "classes", "lessons")

L pnrases 'a pnrase was taxen to De any part or a subject s answer

where two or more contiguous words could be syntactically related; for

example "formal teaching methods", "anything formal obviously

teacher/pupil")

Analysis of the content of these two types of response brought

further coding problems. Broadly there were three types of content

found in both the two categories:

Type 1: the further example type, that is, naming items

belonging to the same category as the exemplar

Type 2: the category identifying type, that is, describing or

identifying categories which could reasonably be held to

include the exemplar, either by a gloss, or a category name

Type 3: the non-compliant type (in the sense of Greenbaum and

Quirk, 1970) where subjects did other things than what they

were asked, such as describing the situation where somebody

would say the stimulus, or describing the speaker's attitude

These three types of responses, and the problems inherent in

categorising them, may be exemplified by examining the responses to

one particular stimulus in some detail.

108
 Chapter 4

Item 18

Stimulus: A: But when you were an undergraduate, didn't you ever

feel you were being taught by people who weren't
actually that good?

B: No it didn't seem -I mean, well - we had a very

funny undergraduate - there were no lectures
or anything like that, it was all sort of
seminars. (1/21.2)

Total of responses: 84

Type 1 46.4 percent

Type 2 45.2

Type 3 8.3

Rank Order of Type 1 responses

6
classes
lessons 6
films 3
tutorials 3
discussions 3
talks 2
conferences 2
assessment
debates
dictations
essays
examinations
exhibitions
grading
groups
marks
putting in order, ranking
practicals
teachers
teach-ins

Type 2 Responses

formal classes 2
formal talks 2
formal teaching sessions 2
anything formal obviously teacher-pupil
anything that there should have been
any kind of tuition where you can take notes and are talked
to entirely by a professor
big groups
big lessons
discussions apart from seminar
formal conventional methods
formal dictation
formal lessons

109
 Chapter 4

formal situations professor makes a speech

formal teaching methods
hard work
informative talks
instruction
large formal teaching groups
large groups, in lessons
lessons where no participation expected
little teacher participation
nothing easy to cope with
one-many talks
teaching in (illegible) intimidating
professors
proper teaching
serious classes
set curriculum
set talks
talks apart from seminars
talks compulsory attendance
talks to lots of people
teaching
teaching without feedback
traditional teaching
usual expected methods

Type 3 (other) Responses

at-all hardly, means nothing, redundant, to speak ofj no

formal lectures, no teaching as such

Firstly, that the type 1 responses clearly share some

we may agree

Yet linguistic theory at the present time lacks

aspects of meaning.

any tried and tested, and more important, any non-intuitive method of

classifying their differences and similarities. Pre-establishing such

a method would surely fall outside the scope of the present 'study,

is to describe an aspect of vague language use, "rather

whose objective

than to do lexical semantics. So a 'working solution' must be sought.

A of discussing the type 1 responses is in terms of the

useful way

'gamelike similarity' invented by Wittgenstein. He gives

concept of

the example of games which all belong'in some sense to a category.

Yet be hard put to it to actually state what similarity it

we should

is that they all share. Perhaps there is none, yet they share a

'category' by virtue of each having something in common with some of

Rosch Mervis's (1975) study of 'family

the other members. and

110
 Chapter 4

resemblances' demonstrated experimentally that subjects could work

with this kind of similarity. I suggest that the type 1 responses for

each question in this test should be analysed as having gamelike

similarity to each other. Clearly, though, decisions about gamelike

similarity will largely be intuitive ones. There is however, other

evidence available which suggests that the type 1 responses fall into

coherent categories, and this is described in 4.3.1.

Secondly, for type 2 responses, which give what I have suggested

are category identifying glosses, or superordinate terms, the only way

to use these is to subjectively interpret them. We may agree that

s03 Lessons where no participation expected

s04 Any kind of tuition where you can take notes and are

talked to entirely by a professor

s16 formal situations: professor makes a speech, no

interraction between students and teachers

for example, are all describing very much the same thing, but it would

quite difficult to say exactly how, since they are in general not

using the same words. A possible way to achieve greater objectivity

in interpretation of the type 2 responses might be a rank order

frequency count of different words used in the responses. From this

we could learn that here, for example, 'formal' was used 5 times, and

'talk' 4. Yet I do not believe that this would be a helpful or

revealing piece of data, because such an analysis would lack all the

semantic import deriving from precisely the combination of lexical

items in the phrase responses. Accordingly, analysis was performed as

follows: the total number of responses for each item was counted.

Where the same word appeared in the singular and the plural these were

totalled. Three separate counts were made of type 1,2 and 3

responses and these are given in Tables 4.4,4.5 and 4.6. Then,

111
 Chapter 4

percentages of each type of response were calculated for each item and

these appear in Table 4.2. Assigning responses to type 1 or type 2

was in some cases contentious, and readers may choose to disagree with

my decisions for certain responses. Alterations to the categorisation

of the few contentious responses would not however alter the general

effects which I describe for the test results.

Null Answers

A null answer was scored for any item where a subject wrote nothing,

or wrote something illegible. 39 subjects responded to 31 stimuli,

giving a total of 1209 possible answers. Out of these, there were 58

null answers - 4.83 percent of the total.

Null answers followed a pattern, with certain items attracting

significantly more than others. The test items ranked in order of

numbers of null answers appear in Table 4.3. No Rosch-type item

scored more than 2 null answers (out of a possible total of 39) and

most scored none. The three items attracting a lot of null answers

were 28,10, and 19. In addition these items received in total very

many fewer responses (46,56 and 71, respectively) as against, say,

191 responses for item 16 and 200 for item 27. (note also that item

10 contained or whatever, and so might be expected to be different. )

When questioned afterwards, many subjects said that they just did not

understand what was being talked about in these three items. It is

fairly obvious that some specialised knowledge would be required.

This probably demonstrates a disadvantage of using real examples in a

test - they can be rather complicated, and almost definitely more

complicated than invented ones. Thus in this case they detracted from

the test results by confusing the subjects.

112
 Chapter 4

The second point is that null answers were not distributed evenly

across subjects. They were much more frequent among the sixth-former

subjects -a mean of 3 per 'subject' as against only 0.5 per subject

for the other two groups. This should probably be attributed to two

factors. Firstly, the sixth form pupils would have less experience

and less world knowledge, which might have led to comprehension

difficulties, especially on the three items mentioned above. Item 28

demonstrates especially well the comprehension failure resulting from

lack of necessary world knowledge. Supervisors do not form part of

school life, so the school subjects did not, presumably, know about

them. Thus 12 out of 15 gave a null answer. Lack of knowledge might

in turn have given them less confidence, and a fear of giving the

wrong answers (even though I had specifically told them that there

were no right answers). Secondly the school subjects may have been

less motivated to fulfil the task well. The other two groups were

volunteers, whereas the Nunthorpe boys had been volunteered by their

teacher.

4.3 Results and Discussion

The results show clearly that in the majority of cases subjects judged

that the tag was an instruction to think of a category consisting of

items similar in meaning to the exemplar. We remember that subjects

had no example to suggest a particular answering strategy, yet between

X 87 and 100 per of their responses on every item either

cent named

category members, or gave category identifying glosses or category

names.

4.3.1 "Rosch" type test items

These used as exemplars items taken from Rosch's (1975 work on

category membership and prototypicality. I used these in order to

have an independently established measure of category membership with

113
 Chapter 4

which to compare my results. Rosch used constructed lists of category

members. In her work, categories were initially adopted if at least

five items from them in the Kucera and Francis (1967)

appeared

Frequency List with a frequency of 10 or greater. All the categories

she used were also in Battig and Montague's (1969) tabulation of

production frequencies for responses to a particular category name.

Each of Rosch's categories had 50 or 60 items in it, and her subjects

were asked to judge each item for how good an example of the category

it was. Drawing on her results, I'used some items having a high

degree of prototypicality and others have_a low degree, in order to

see the effect of the difference. We shall now examine the type 1

responses to these "Rosch" test stimuli in some detail.

Rosch's category lists are compared with my subjects' type 1

responses, by obtaining the intersection-union ratio of the two sets.

Thus, for item 1 union yields 43 items of which 15 intersected. This

gives a percentage overlap of 34.9 percent. The results of doing this

for each 'Rosch' test item are presented in Table 4.4. In making the

comparison I have felt justified in omitting from Rosch's lists those

items which (a) do not exist in Britain, or (b) have different names

here, since my subjects could not be expected to mention them. Where

translation equivalence exists (eg ' US slingshot - Br catapult) I

treated the equivalents as naming the same item.

The Rosch-type test items fall into two categories. Firstly

those with a high percentage overlap with the original Rosch category

(upwards 34 percent), that is, items 1,3,6,14,16,22,27, and

of

30. In every case the exemplar for these was one having a high degree

for its category. Secondly, those having a

of prototypicality

low overlap (below 13.4 percent), that is, items

markedly percentage

114
 Chapter 4

4,9,17,20 and 24. In every case (with the exception of item 24,

discussed below) the exemplars for these were items having a low

degree of prototypicality for the category involved. Of the two

remaining items, 11, with a percentage overlap of 27.6, was a bad

example, and 25, with 28.2, was a good example.

These results show quite clearly I think, that successful

category identification using a tag-necessitates giving a prototypical

example of the intended category. Giving a 'bad example' as an

exemplar results in hearers identifying a category which has little

correspondence with the intended category, as, shown by the low

percentage overlap items listed above.

The idea that hearers take the exemplar as a good example is

additionally supported by the results for some of the 'bad examples'.

I suggest that where a Rosch 'worst example' may be the worst example

for the category she (and her subjects) had in mind, it is, as well, a

best example for another category. Taking item 4, necklace was judged

by Rosch's subjects as the worst example of the category CLOTHING.

The category members given by my subjects in response to this seem

fairly obviously to belong to the category JEWELLERY, of which

presumably necklace and bracelet are good examples. That is, they

took it to be a good example. This is shown also by item 17 where a

poor example of the intended category WEAPON was taken by subjects as

a good example of the category TOOL (cf their type 2 responses in

Table 4.5).

A general question arising from these results is whether we

should consider the tag as an instruction to access a semantic

category, or to access a non linguistic associational or perceptual

category. By 'semantic' I mean one whose members are related by

115
 criteria which are defineable independently of context, for example by

entailment: "I bought apples" --> "I bought fruit" and "I bought

oranges" --> "I bought fruit". It is of course often the case that

semantic and associational categories can coincide, as with the

category FRUIT, but by no mean always, since things whose names are

not at all semantically close, may be strongly associationally

related.

Initially, the results for the 'best examples' items which we

have just looked at, suggest that semantic categories are involved.

But the results for other items suggest rather the opposite, that in

fact it is rather pragmatically established associational categories

that are involved. For example, responses to item 9 showed subjects

focussing on culturally defined categories:

s05 "vegetarian context" rice beans dried-vegetables

"chinese context" rice chicken green vegetables sweet and sour

"greek context" stuffed pepper pilaf

s16 chinese food exotic vegetables (superordinate term) farinaceous

food potatoes polenta pasta

a27 Eastern foods curry chinese take-away food

The responses occurring more than twice in the rank order of responses

of both types are mostly not semantically relatable to rice, the

exemplar, but associationally, in terms of what people know about food

in the world, they clearly are.

curry(ies) 12
vegetables 7
pasta 4
bread 3
chinese food 3
corn 3
macaroni 3
maize 3
potatoes 3
bamboo shoots 2
beans 2
carbohydrates 2

116
 Chapter 4

fish 2
foreign food 2
health foods 2
Indian food 2
oats 2
rice 2
spaghetti 2
spices 2
tapioca 2
water 2

Again for item 24, most subjects appeared to put car in the

associational category "possessions of affluent persons", rather than

in the semantic category "vehicles". Hence the low percentage overlap

recorded. Their most frequent type 1 responses were:

caravan 10
washing-machine 8
tv 7
colour tv 6
motor bike 6
boat 5
dishwasher 5
house 5
trailer 5
freezer 4
big house 3
nice house 3
record player 3
scooter 3
bicycle 2
bike 2
f ridge 2
mortgage 2
stereo 2

Subject 11's response is especially telling:

washing machine, dishwasher, 3-bedroomed semidetached, 2 children,

dog, cat, mortgage, steady income

Their type 2 responses also show the same idea, for example:

s02 expensive things

s24 everything you could wish for

b31 any material goods of a middle class nature

(these are given in full in Table 4.5)

117
 Chapter 4

Contextualisation

Another effect which is important is the effect on the responses of

the contexts which subjects attributed to the different items. This

is especially clear where a word occurred in two different test items.

For example, car appeared as an exemplar in both 16 and 24. From the

type 1 responses listed in Table 4.4 we can see that subjects

interpreted the two items very differently, because of the context

imposed by the surrounding language. In one case(16), they judged car

to be an exemplar of the set of wheeled vehicles which go on roads,

and in the other (24), as a member of the set of desirable material

goods.

4.3.2 Attested Test Items

The Rosch-type test items were included in the test, as I said, to

provide an independent measure of categorial similarity. We have seen

that where the exemplar was prototypical of the intended category,

there was a very marked similarity between Rosch's ranked category,

and the rank order frequency count of my subjects' responses. This

was taken to demonstrate that what subjects did for those items was to

access a category consisting of items they judged to be similar to the

exemplar given, and a category for which the exemplar was

prototypical.

When it comes to interpreting the results for the attested items,

no independent evidence fox the categories is available. It seems

more than reasonable, however to assume that what the subjects did for

the attested items was the same as what they did for the Rosch items,

and that therefore they again provided answers which either gave

category members, or category names.

118
 Chapter 4

It is, -, of course, only possible to evaluate these data

subjectively, but it is the case for each attested item that the

responses-show quite convincingly a majority of these two types of

response. The full rank order frequency count of responses for these

15 items is given in Table 4.6. In' addition, the answers to

individual items show particular factors at work, as follows:

Item 2: "One
of the secretaries was saying there was a film
or something. ' (B: -really? ) Last Thursday we were all down in
2017 and the lecture was actually in Vanbrugh so I went to
the secretary and, said where is everyone, and she got out
this book and said its in Vanbrugh but some weeks there's a
film can you have films in linguistics? "

The responses to-this item-demonstrate particularly well the influence

of given linguistic context on subjects' interpretation of the

exemplar +-tag. The seven most frequent responses, in rank order,

were lecture, slides, talk, video, play, show, discussion. Now I do

not think that out of context lecture would normally be considered to

be most similar to film - T. V. and video are much closer. What

brings lecture, talk and discussion into prominence is the context

provided which suggests that academic teaching is what is being talked

about. In this particular context, lecture, talk etc are likely

members of the category understood.

In the pilot study for this test, item 2 appeared with only the

speaker's first remark (up to B's interjection), and subjects

responded with slides and video. These pilot informants suggested

longer extracts should be given, since they found the short ones

difficult to interpret.

Item 19: "I'm talking about acceptable middle class language and
sort of working class language - the thing that Bernstein,
you know, sort of - elaborated code and things like that"

This shows I think how important it is in comprehension to understand

the topic that is being spoken about. Many subjects commented

119
 Chapter 4

that this was among the items which were difficult 'to
afterwards

answer. This is reflected firstly in the high number of null answers

(viz 7 out of 39). Secondly it is reflected in the low degree of

responses. Only 5 response tokens out of a total of 72

sharedness of

were shared at all, and then only by a maximum of 3 subjects.

The smaller quantity of responses, and greater frequency of null

answers for the items where subjects, on their own admission, did not

understand what was being talked about, shows, I would judge, the same

effect as that demonstrated experimentally by Bransford and Johnson

(1972). This work'showed that their subjects experienced considerable

difficulty in assigning meaning to, and recalling, stimulus material

which had been artificially contrived to omit suitable contextual

cues.

Item 23: "You can remember four lots of four fairly easily - say
in the form of dates or something - 1972 or something like
that"

Because the exemplar consisted of a number, these answers show

the difference between type 1(further example) and

especially well

type 2 (category identifier) responses. By subject, they were as

follows:

sOl ymca ucca

1974 1979 a recent date, a used date, a reoccurring date
s02
s03 [null]
s04 some type of group of numbers/figures, various other dates
s05 16 4x4x4x4 etc significant year names eg 2000 1984 1000
s06 four-syllable words
SO sequence, year
s08 1970 1971 2000 4000 1000 6000 9000 7000 1001
s09 other similar dates, other relevant dates
s10 1843 1652 1004 5302 6392
sll 1875 1066
s12 1973 2001 8.8.60 any year
1066 and all that, 1945, numbers of relative importance for
s13
historic or personal reasons
s14 phone numbers, mnemonics, paradigms
s15 1968 1964 the olympics
s16 don't know [null]
for example, that's a date isn't it? the full date I mean, 1948
s17
1976 etc
s18 another year

120
 Chapter 4

s19 1810 2001

s20 1843 1764
s21 1971 1973 1974
s22 1973 1974 1975 1976 1977 1978 1979 1800 1801 1802 1926 1923
1921 1851 1854 1960 1901 1700
s23 1066 1666 1665 1969
s24 telephone numbers, birthdays
b25 mnemonic, catchphrase, cliche
b26 [null]
b27 other easily remembered numbers
b28 1971 1970 1969 1968 1967
b29 1976 1980 1984 1988
b30 early 70s 62 72 82 years of the Olympic games, years of the
European soccer championship
b31 any other relative date, any other mnemonic
b32 1968
b33 [null]
b34 [null].
b35 1976
b36 Boycott's. run total for season
b37 1973 1977
b38 16
b39 ABCD

The type 2 responses show subjects imposing different

recontextualisations as they try to identify likely categories. Many

of them concentrate on the remembering aspect, eg subject 13,14,24,

b27 and b31. Several subjects saw 1972 as an exemplar of a category

consisting of years when particular events took place: s15 the

Olympics, suggested again by s30, who also tried "years of the

European Soccer Championship". Subject 6's response is unexpected,

given that 1972 is not a four syllable word. I wonder whether subject

b36 was being non-compliant, or whether he sees the world entirely in

terms of cricket?

Item 26: "Could we, when you give us our essays back, and give us
titles, could we sort of meet or something because, I mean,
there might be things we want to ask"

This item shows the influence of the subjects' perception of purpose -

in this case the purpose of the event proposed by the speaker.

Equally ranked in first place were both get together, and discuss,

with 11 occurrences each. There were many responses containing

phrasal equivalents to discuss (see Table 4.6). Discuss is not

121
 Chapter 4

however semantically close to net and not always associationally

close. One can discuss without meeting, and meet without discussing.

Yet here the subjects judged that the purpose of the meeting, which

they interpreted as. "to discuss", was highly salient. It is of course

specified by the speaker: "there might be things we want to ask".

To sum up, then, the answers given for the attested test stimuli

show (a) subjects interpreting the tags as instructions to access

categories, and (b) that the characteristics of the categories they

fixed on are determined not only by the given exemplar, but also by

linguistic context, situation and pragmatic knowledge. I asked some

of the university test subjects what they thought the function of the

exemplar word was. Two of them suggested:

B: It's a member of the class of objects

C: and it's got - it calls up associations with it...

4.3.3 General Conclusions from the Results

The observations above lead me to propose that the correct way to

interpret these data is as follows. The exemplar + tag construction

is understood as an instruction to access an associational category,

whose characteristics are defined for the hearer by the exemplar

provided, taken in conjunction with relevant pragmatic information.

It is for this reason that I have called the exemplar + tag

construction a (vague) CATEGORY IDENTIFIER. The associational

category may be coterminous with a semantic set to which the

exemplar-word belongs (if it is a word). This is expected, since

words which are semantically related often refer to entities which are

associationally related, at least in some contexts. It appears, as we

have seen, to be the associational relationships which are most

important in determining the categories hearers think of.

122
 Chapter 4

I think the categories hearers use are probably those identified

by Klatsky and Stoy (1978) as 'higher order categories'. In their

.

experiments, they found that knowledge of category membership had an

effect on- subjects' performance in a simple visual task of comparing

pictures. Klatsky and Stoy state that their results indicate subjects

using a higher order category (not necessarity a namable one) to

compare instances.

The general meaning effect described above is the one I shall

work with in subsequent chapters.

4.4 The Structure of the Tags

These tags contain either and or or:

or something/anything (like that)

{things
and {things like that
{that

The test results and data examples both indicate that there is no

difference in the way that the tags containing like that and those

without are. understood. It is reasonable to see the shorter tags as

ellipted or shortened variations of the full ones.

Secondly or something (like that) and or anything (like that) are

found in complementary distribution as between assertive and

non-assertive contexts, thus:

(9) There was no coffee room or anything (11/21.2)

 Chapter 4

i (10) *There was no coffee room or something [11

(11) One of the secretaries was saying there was a film or something

(12) *One of the, secretaries was saying there was a film or anything

In this they follow a rather general rule of special marking in non-

assertive contexts, (cf Quirk et al, 1972, pp 376ff, for discussion).

Thus there are really just three tags under consideration:

or {something} like that

{anything }

and things (like that)

and that

The relationship between the meaning of the words which constitute the

tags, and the meanings of the tags in use, is discussed in Chapter 6.

4.5 Structure of Exemplar + Tag Constructions

The exemplar always precedes the tag. In my data I have examples of

the following syntactic types of exemplars, shown here with the tags

they occurred with:

NP: and things like that

and things
and that
or something
or something like that
or anything like that

VP: or something
and things like that

Adverbial phrase: or something

or somewhere
or something like that

In each case, intonation makes it clear which constituent is being

tagged, as I describe in Section 4.6.

1 (10) is not asteriskable if the tag applies to the whole sentence,

rather than just to coffee room

124
 Chapter 4

4.5.1 NP + Tag

In the category NP + Tag, the three possibilities countable singular,

countable plural,. and uncountable N all occur, thus:

(13) [tutorial discussion]

+I'm talking about acceptable middle class language and sort

of working class language the thing that Bernstein you know
sort-of elaborated code and things like that [II, 8/3)

(14) [title of children's book]

+All about cuckoos and robins and things

(15) +Lady Arran, who pays for her hobby by 'selling off silver
and things', set a world record of 96 mph at Windemere last
October [DM, 17.5.80]

But there are far more examples of countables. I cannot see any

immediate reason for this. The examples I have with uncountables

sound quite normal, and I would have expected as many of these as of

countables. Possibly they do not occur because many uncountables are

also abstract. An uncountable abstract exemplar is perhaps not felt

to be able to exemplify a category. Made-up examples seem acceptable,

though:

(16) He's studying syntax for something

{and things like that

(17) I don't like Peregrine, I think its his

bumptiousness {or something
{*and things

NPs with tags tend not to be subjects (I have no examples). Some

made-up examples seem acceptable:

(18) A pub or something in the next few miles would be most welcome

In this case, however, the exemplar is grammatical subject, but

semantic object (ie 'we' would welcome a pub). More importantly, it

is the 'new' information in the utterance (as opposed to given). I

suspect that tags on 'given' information are virtually ruled out by

the fact that speakers cannot be uncertain about something which is

125
 Chapter 4

being considered as given in a conversation. Since grammatical

subjects often embody the 'given' information in an utterance, tags

are not found with them.

4.5.2 VP + Tag,

(19) +Could we, 'when you give us our essays back, and give us
titles could we sort of meet or something ...
(20) +"that most people" they're three words that fit together
and could fit into the sentence and things like
that [I, 13.2

(21) [J. MacEnroe's father in TV interview]

+He's a bit impish or scampish or something like that [DS/9]

(21) looks as if the tag might be attached to the adjective, but it is

not, since it is not possible to move in into attributive position:

(21') *He's an impish or scampish or something tennis player

There don't appear to be any restrictions structurally on what VPs can

be tagged, although very long ones are avoided because they give rise

to problems for the speaker when he tries to provide the appropriate

intonation (see 4.6 below, test item 21).

4.5.3 Adverbial + Tag

My examples are adverbial strings:

(22) [magazine]

+It comes out every four months or something like

that [DS/7]

(23) [eye pain]

+I happen to have that when my - when I menstruate or just

before my period or something [Cicourel, 1974]

4.6 Intonation and Vague Tagging

As noted in the previous section, these tags may be adjoined to an NP

or to a VP. In the case that a VP has the structure

[V NP]
VP

there may be a structural ambiguity as to whether the tag applies to

126
 Chapter 4

the NP, or to the whole VP. The recorded examples show that this

ambiguity is not present in speech because it is avoided by use of

intonation...

The material to which the tag applies (ie, the exemplar), and the

tag, always appear in the same tone group (except obviously where the

tag has been added by another speaker). Usually the tone group

beginning shows where the material to which the tag applies begins.

The nucleus is carried by the principal item of the material to which

the tag applies, and the tag appears as a tail. This tail has either

a level intonation or in some cases a slight rise. A straightforward

example is:

You can remember four lots o four fairly easily i'""' in the form
Idates IN
I Ii
of or something nineteen seventy two or something like thatll

where or something tags dates, with the nucleus, a fall, on dates

and or something with a level pitch. Nineteen seventy two or

something like that shows the other pattern observed, with a very

slight rise on something. The usual pattern ism "'"'"'/ with a

,

fall on the tagged material paired with a rise on the tag.

In the case of a VP the same thing occurs:

e\1
and then youre asked to sort of reproduce them or something

and[2]

with the nucleus -a fall - on reproduce and a level pitch on the tag.

2 apostrophes have been omitted in this section, in order not to

interfere with the transcription of intonation

127
 Chapter 4

A structurally ambiguous case is:

(24) What about things like when you read sentences or something

where the tag could apply to the full NP things...., or to the VP

read..., or to sentences. In speech, however, it is clear that the VP

read sentences is being tagged.

I
But what about things like when yourewhen you read sentences or
II Ithen, a\
0I
something and youre asked to sort of reproduce them or
u
something) and you reproduce them in the way that it struck you

most

Another structurally ambiguous example (from the same speaker, but on

a different occasion), is:

It was really difficult (. ) because theyd read these words in a

)come
book or somethinglf and theyd to me and say what does this

mean

where the tag applies to the whole VP read....

Speakers may use other very clearly observable prosodic features

to delimit the exemplar + tag structure. For example in:

I had to give them all these different sort of meanings) like

i
(. )I sentences wi with it in and things like that 1 and it was
low
1%
really difficult for them to grasp itl)

to mark the beginning, there is a tone group boundary after meanings,

then a very discernible pause after like, and a clearly stressed onset

on the first syllable of sentences. To mark the end, as well as the

128
 Chapter 4

tone group boundary, there is also a marked drop to low pitch, plus

acceleration on and it was. It is apparent that speakers go to quite

x some lengths, prosodically, to ensure that the reading NP + tag, in

the situation. like this where the NP is in a VP, cannot be

misunderstood as a VP + tag reading, seen so clearly in the previous

two examples.
,

This example is additionally interesting because it appears to

show the speaker getting into slight difficulties, due I think to the

NP sentences with it in being a bit long or "heavy" and thus making

its tag (with its rise) rather uncomfortably far away from the nuclear

fall which'it pairs with. Having placed a fall in the expected

position on sentences, the repetition wi with shows something going

wrong. The repair seems to be to'copy the fall on sentences exactly

onto in and then bring the tag in normally with a rise on things.

Given this differentiation by intonation alone, when these

structures are written, ambiguity can arise. As we might expect, this

showed up in some of the responses to the test stimuli, which were

presented in written form. This occurred with Test item 21:

+I hope we didn't have lots of horrible conversations when

you went out of the room in the tutorials and things like
that
[II, 8.3]

In written form, the tag could apply either to tutorials, or to the

adverbial phrase when you went out of the room or to the VP part of

it, or even the NP the room or to horrible conversations. Test

answers among the 39 subjects divided as follows:

129
 Chapter 4

conversations 12

went out of the room 3

tutorials 21

other (null, deviant etc) 4

40*

*because one subject spotted the ambiguity and gave two sorts

of content answers

Although tutorials was favoured, there are enough answers choosing the

other options to show that the ambiguity is present. The fact that

the test stimuli were presented written is thus an advantage and a

disadvantage. The division of answers into broadly two content types

is useful evidence for the structual ambiguity I am claiming.

Secondly (by their absence in the test), it is shown that phonological

features muse be responsible for clarifying the ambiguity. The

disadvantage is, of course, to withold from the test subjects an

important piece of information which they would have had as hearers.

The prosodies for this example actually were:

Ilhope we didnt have lots of horrible conversations when you went

\a1
out of the room in the t torials and things like that

This shows again the problem of interposed material. In the tutorials

is prosodically subordinated, so as not to interfere with the

fall-rise pattern on room and things. This rules out tutorials as the

exemplar for the tag.

130
 Chapter 4

Something noticeable about this utterance is that the whole thing

is one quite unusually long tone group. Given the syntactic

arrangement, this is actually non-optional, if 'the speaker wanted to

link the tag to the whole VP

[have lots of horrible conversations when you went out of the room]
VP

A boundary after conversations would have tagged only the adverbial

phrase. I'd argue that the former account is the correct

intdpretätion of the prosodic evidence on two grounds. Firstly,

situational factors; given the context, there is not any set of events

that could reasonably be intended by the exemplar when you went out of

the room. The tutor is either in, or out, and that's it. Secondly,

the way this utterance is understood by subsequent speakers. They

focus on thinking of whether or not A did actually leave the room

during any of the tutorials or not, and she states that she did not.

Thus when you went out of the room is seen by them not as vague, but

as precise.

Among the test subjects, the favoured exemplar was tutorials. It

might be thought that this was purely because of its contiguity to. the

tag. However another test item, 13, shows that contiguity cannot be

the only factor in play.

Test Item 13

+But lots of big sort of important numbers that you have to

remember and things, I'm sure they're made into a pattern
that you can remember them by...

In this case, the relative clause on numbers creates several possible

scopes for the tag, as follows:

NP + tag: [big sort of important numbers and things]

NP [that you have to remember]
S

VP + tag: [have to [remember] and things]

VP S

131
 Chapter 4

VP + tag: [remember and things]

VP

The spoken example had the the following intonation:

\ ýButl 11
lots of big sort of (. ) important numbers thatI that you have
IN ol I
11
Im sure
s,\
to ring and things and they theyre made into a kind of
11
patter that you can remember them byfl [3]

The tag thus applies unambiguously to ring.

Without knowing the prosody, the 39 subjects again divided

between the different possiblities, as follows:

numbers 26

have to remember 1

remember (_ "ring") 9

null 4

The favoured interpretation is numbers, a long way back from the tag,

thus suggesting that contiguity is not guiding the subjects'

interpretation. Probably the informational importance of numbers as

the head of the relative clause, led them to focus on it.

4.7 Co-occurrence restrictions

Apart from the syntactic distributional constraints described in 4.5,

there are also co-occurrence constraints related to meaning. The

constraints I have observed are however explained quite naturally by

the meaning specification for tags which has been provided in 4.3.

Here are two instances reported to me as violations of proper

vague tagging[4]:

3 Reexamination of this example showed that it had originally been

transcribed wrong, with remember for ring. But this fact does not
alter the import of the test result.

132
 Chapter 4

(25) [Context: a baby has just been born, someone is asked what sex
it is]
+It's a ,boy or a girl or something

(26) [Context: Chairman of a Linguistics Department is looking for

part-time tutorial staff; what will they have to teach? )

+Syntax and phonology and things like that

The explanation-for these strange sounding combinations lies in what

has been observed of the way people understand exemplar + tag

constructions. We have seen that they are understood as designating

an, associational category whose characteristics are defined in part by

the situation involved. The exemplar(s)(a) are taken as good

example(s) of the set, and (b) must be capable of being members of

some reasonable set. Example (25) names both members of the set of

possible human offspring, so there are no other members which the or

something can be standing for. Therefore it sounds wrong.

The second example is a bit more complicated. It is not so much

a violation of vagueness tagging, as a failure to give expected

information. I take syntax and phonology and things like that to be a

direction to the set whose cover-term is Linguistics. Given that here

we have two 'linguistics Professors in conversation, the questioner

knows that teaching linguistics is involved, he wanted to know, as he

reported, which branches of linguistics were required. He gave the

reply the reading I've suggested, and thus found it less than

informative. If we assume the speaker intended to be informative,

another permissable reading of (26) is "the core areas of linguistics"

(as opposed to sociolinguistics, stylistics, textlinguistics etc). '

4I am grateful to Patrick Griffiths and Bob Le Page respectively for

these examples.

133
 Chapter 4

Here then we see two pragmatically determined conditions on

successful vague tagging:

1 There must be other members of the set which the tag can

stand for

2 The hearer must make the relevant category boundaries at

least similarly to the speaker

1 is a powerful influence on interpretation of these structures, such

that if an exemplar which is a member of a category of only 1 is used,

the interpretation preferred will be that the reference is to a larger

category of which the exemplar is also a "good example" member.

Imagine, for example:

(27) She wants to become Prime Minister or something

There is away in which prime minister belongs to a category with only

one member: the category of current Prime Ministers of the U. K. If

one is trying to refer to that particular category, (27) will not be

acceptable. The interpretation given to (27) is rather that it refers

to a category such as "important people in Government", or "the

Cabinet".

A different type of constraint is on which kinds of exemplars can

be, used. The following sound odd, although I would not want to

suggest that there might not be contexts in which they would be

suitable. -
(28) It was a freestone peach or something
(29) It was a metallic finish 1975 Renault 5TL or something

Regrettably I have no examples of violations of this type, and so was

forced to invent. If (28) is intended to designate members of the

category of fruits (apple, orange etc), or (29) the category of small,

fuel economic cars, then it is apparent that too much information is

being given. These are simple violations of the conversational rule

 Chapter 4

described by Grice as the Maxim of Quantity. This indicates that one

rule relevant to category reference by tags is that only sufficient

information to identify the exemplar as a member of the category need

.
be given. The over-specifications in (28) and (29) leave the hearer I

think searching for a smaller, more specific category, of which the

exemplar could be a good example.

At the other end of the scale, supero inates do not make good

exemplars either, so:

(30) ?*She went by plane or something

is odd. The only category which plane can plausibly be an exemplar

for is METHOD OF TRANSPORT. Since a hearer knows from she went, that

transport is involved, plane or something fails to be informative,

w

again violating the Maxim of Quantity. Contrast:

(31) She went by 747 or something

which acceptably suggests, perhaps, the set of large modern planes[5].

The constraints on what can be suitably tagged, then, interact

with the principles of categorization outlined in the work of Rosch

(see Rosch 1978: 28-49). Rosch observed that given the existence of

taxonomic relationships of the type:

superordinate basic level subordinate

furniture chair kitchen chair

the basic level is the one usually used in referring to objects which

could be included in the taxonomy, instead of either the subordinate,

It looks basic level is most

or the superordinate. as though the

likely also to be used with tags.

51 am indebted to Michael Lumsden for this point

135
 Chapter 4

Of course, Rosch was only working on taxonomies of concrete

objects. We do not yet know whether the principles of categorisation

she proposed are applicable to terms designating non-concrete things.

A study carried out by Pulman (1983) indicated that subjects could

judge prototypicality within given sets of verbs, but it is not known

whether a 'basic level' can be identified for referring to actions and

abstracts.

In support of her claim for the salience of the basic level in

communication, Rosch cites what she claims are two different types of

deviations from it. Firstly the experiments conducted by Bransford

and Johnson (1972), referred to above, in which they omitted context

cues. Rosch suggests that context cues are actually "basic level

events". The effect of missing them out of a text is to impede or

block comprehension. Secondly, "substitution of subordinate terms for

basic-level object names in scripts gives the effect of satire or

snobbery" (1978: 45). She quotes a book review:

"And so, after putting away my 10-year old Royal 479 manual
and lining up my Mongol number 3 pencils on my Goldsmiths
Brothers Formica imitation-wood desk, I slide into my
oversize squirrel-skin L. L. Bean slippers and shuffle off to
the kitchen. There, holding Decades in my trembling hand, I
drop it, plunk into my new Sears 20-gallon celadon-green
Permanex trash-can. "

in which the effects are, I would judge, similar to those in (28) and
(29).

As far as the tag constructions are concerned; the examples show,

I would hold, that you can tag at any of Rosch's levels, but that the

general rule is the one given with a picturesque analogy by Miller and

Johnson-Laird (1976: 259):

"treat your hearer like the tax-man, give him no more than
you have to. "

136
 Chapter 4

4.8 Summary

In this chapter I have described a second type of vagueness arising'

from the use of vagueness-bringing additives. I have called this set

of additives TAGS, and the resulting exemplar and tag combinations,

(vague) CATEGORY IDENTIFIERS. We have seen that the tags are combined

freely by speakers with a variety of grammatical categories. In

particular we have looked at evidence of how hearers interpret these

tags. The set of propositions relevant to their meaning, which the

theory of vagueness must account for, includes the following:

1 Vague tags are understood to designate categories, either

conjunctively or disjunctively, consisting of either concrete

("bread", "silver") or abstract ("meet", " elaborated code")entities.

2 These categories are associationally rather than semantically

defined

3 Several factors establish for the hearer the characteristics

determining membership of the category. They are (a) the exemplar,

which is understood to be a "good example" of the intended category;

(b) the surrounding linguistic context; (c) the purpose of the

conversation; and (d) the hearer's world knowledge which he brings to

bear as relevant to the linguistic context and conversational setting.

4 There is some evidence to suggest that these expressions cause

particular comprehension problems for hearers who lack specific world

knowledge of the conversational topic. That is to say, more problems

than non-vague expressions in the same situation.

In the next chapter, I shall look at attested examples of these

vague tags (with other examples of vagueness) in order to ascertain

the conversational purposes which speakers use them to achieve. Then

in Chapters 1 and 8I discuss theoretical considerations relevant to

accounting for the observations in 1,2,3 and 4 above.

137
 Chapter 4
Tables

Table 4.1

Stimulus Items used in Tag Informant Test, in Order A

Notes:

(a) half the subjects responded to these in reverse order (B)

(b) 'R' denotes an invented 'Rosch' item

1. "Could you get me some oranges or something at the market? "[R]

2. A: One of the secretaries was saying there was a film or

something. (B: really? ) Last thursday we were all down in Z017
and the lecture was actually in Vanbrugh so I went to the
secretary and said where is everyone, and she got out this book
and said its in Vanbrugh but some weeks there's a film can you
..
have films in linguistics?

3. "I love peas and things like that"[R]

4. "... and when you think of necklaces and things like that, the
possibilities are endless"[R]

5. "I think its just sort of learning ability that's sort of there,
you've got sort of sponge waiting to soak things up and whether it
be language or anything else -"

6. "What about if you carry a gun or something like that"[R]

7. "... and they're much easier to remember than say your friend's
telephone number because they give you something like 23578 or
something like that which is much easier to remember"

8. "and I just could not translate it and I had to give them all
these different sort of meanings like sentences with it it and
things like that and it was really difficult for them to grasp it"

9. "She said they eat rice and that, didn't she"[R]

10. "Maybe its just because you're so its not exactly innate but
its just because at the start you realise you're so dependent or
whatever on ... I mean human babies particlarly are so dependent
that they need communication so much that they have to somehow get
it over to someone that what they want is what they need or
whatever"

11. "She said it was olives or something"[R]

12. "Well she felt a bit out of it all you know she saw her
...
supervisor during the week whenever it was and they had these sort
of lectures and that, and that was it. There was no kind of
social contact ... there was no coffee room or anything"

13. "But lots

of big sort of important numbers that you have to
remember and things, I'm sure they're made into a pattern that you
can remember them by... "
I
 Chapter 4
Tables

14. "okay so you take the trousers or whatever and you spray them with
this spray... "[R]

15. "But what about things like when you read sentences or something
and then you're asked to sort of reproduce them or something, you
,
reproduce them in the way that it struck you most - its the
meaning that's the most important in reproducing what's there"

16. "I stopped my bike by the" verge, then a car or something came.
along and splashed me all over with water"[R]

17. "He used a screwdriver or something"[R]

18. A: "but when you were an undergraduate, didn't you ever feel you
were being taught by people who weren't actually that good? "
B: "No, it didn't seem -I mean, well - we had a very funny
undergraduate - there were no lectures or anything like that, it
was all sort of seminars"

19. "I'm talking about acceptable middle class language and sort of
working class language - the thing that Bernstein, you know, sort
of - elaborated code and things like that"

20. "Have you got a wheelbarrow or anything like that which we could
borrow? "[R]

21. "I hope we didn't have lots of horrible conversations when you
went out of the room in tutorials and things like that... "

22. "She's mad, she has no money - but she bought some new trousers or
only yesterday... "[R]
something

23. You can remember four lots of four fairly easily - say in the form
of dates or something - 1972 or something like that"

24. "They've got a car and that"[R]

25. "You had to hand in guns and things like that"[R]

26. "Could we, when you give us our essays back, and give us titles,
could we sort of meet or something because, I mean, there might be
things we want to ask"

27. "... they don't need a dessert, they can eat oranges and things
like that"[R]

28. A: "What else do we want to talk about on friday? "

B: "Well, wait a minute, what are we doing, we're having, second
supervisors -"
A: "- or something -"
B: "subsidiary -"

29. "When I was trying to teach them certain words it was really
difficult because they'd read these words in a book or something
and they'd come to me and say what does this mean"

30. "They never give us peas or anything"[R]

 Chapter 4
Table 4.2

Table 4.2

Tag Informant Test: Summary of Results

Type 1 Type 2
Item Total Type 3
Title (further (category
Responses (other)
no. example) identifier)
%%%

or something/anything:

1 165 77.0 21.8 1.2

oranges
film 86 86.0 9.3 4.7
2
11 106 79.2 17.9 2.8
olives
12b 75 64.0 30.7 5.3
coffee room
133 94.0 3.0 3.0
15 read sentences
191 86.4 7.9 5.8
16 car
128 83.6 11.7 4.7
17 screwdriver
189 81.5 14.8 3.7
22 new trousers
99 85.9 8.1 6.1
26 meet
46 60.9 26.1 13.0
28 supervisors
29 book 129 96.1 3.9 0.0

30 161 82.0 17.4 0.6

peas

something like that/anything like that:

or

6 153 75.2 20.3 4.6

gun
7 23578 119 61.4 37.8 0.8

18 lectures 84 46.4 45.2 8.3

20 110 85.5 11.8 2.7

wheelbarrow
23 1972 109 71.6 24.8 3.7

or anything else:
5 language 127 85.0 13.4 1.6

or whatever:
10 56 50.0 35.7 14.3
need
14 trousers 167 80.8 15.0 4.2

and that:
9 rice 123 50.4 45.5 7.4
12a' lectures 76 80.3 17.1 2.6
24 car 144 84.0 8.3 7.6
 Chapter 4
Table 4.2

and things like that:

3 peas 164 73.8 25.6 0.6
4 necklaces 166 80.7 19.3 0.0
8 sentences 78 76.9 14.1 9.0
19 elaborated code 71 57.7 33.8 8.5
21 horrible conversations 92 84.8 7.6 7.6
25 guns 176 73.9 24.4 1.7
27 oranges 200 85.0 14.5 0.5

and things:
13 numbers 91 85.7 12.1 2.2
 Table 4.3

Test Items arranged according to frequency of Null Responses

Item Title No. of Total

No. Nulls Responses

28 supervisors 14 46
10 need or whatv 10 56
19 elaborated code 7 71
12a lectures 5 76
13 numbers you 4 91
23 dates 4 109
21 3 92
cony/tutorials
2 film 2 86
8 2 78
sentences
9 R2 123
rice
11 R2 106
olives
14 trousers R2 167
20 R2 110
wheelbarrow
24 R2 144
car
12b 1 75
coffee-room
15 read sentences 1 133
1 oranges R. 0 165
3 peas 'R0 164
4 necklaces R0 166
5 language 0 127
6 gun R0 153
7 23578 0 119
16 car R0 191
17 screwdriver R0 128
18 lectures 0 84
22 new trousers R0 189
25 guns and things R0 176
26 meet"or somethg 0 99
27 oranges R0 200
29 book or sth 0 129
30 peas R0 161
 Table 4.4

Rosch-type test items:

Type 1 responses compared to Rosch's category ranking

1 "Could you get me some oranges or something the "

at market?

Responses Rosch's ranking

category

[orange] date
apples 27 apple raisin
pears 18 banana coconut
bananas 12 peach avocado
grapes 8 pear tomato
tangerines 8 apricot nut
lemon(s) 7 tangerine olive
satsumas 6 plum pickle
grapefruit(s) 6 grapes
mandarins 5 nectarine
peaches 5 strawberry
plums 4 grapefruit
cabbages 2 cherry
carrots 2 pineapple
oranges 2 blackberry
potatoes 2 melon
(water)melon 2 raspberry
tomatoes 2 lemon
pineapples 2 lime
clementines fig
greengages mango
lime pomegranate
mushrooms cranberry
nuts prunes
quince gooseberry
Xmas trees

intersection: 15 Union: 43

percentage overlap: 34.9

3 "I love peas and things like that"

Responses Rosch

beans 24 [peal
sprouts 20 carrot
carrots 15 green beans parsley
cabbage 14 string beans mushroom
cauliflower 7 spinach. avocado
broccoli 5 broccoli rhubarb
french beans 3 asparagus kale
greens 3
green beans 3 corn pickles
lettuce 3 cauliflower baked beans
potatoes 3 brussels sprouts pumpkin
broad beans 2 lettuce seaweed
sweetcorn 2 celery garlic
turnips 2 cucumber dandelion
beetroot beets rice
celery greens
f inocchio tomato
lentils artichokes
meat
mushy peas turnip
onions eggplant (- aubergine)
peppers peppers
pulses radishes
pumpkin onions
runner beans bean
silver beet potato
spinach parsnip
tomatoes watercress
zucchini leek
sweet potato

Intersection: 21 Union: 47

percentage overlap: 44,. 7

 Chapter 4
Table 4.4

4 "... and when you think of necklaces and things like that, the
possibilities are endless"

Responses Rosch

bracelets 27' pants (= trousers, Br. E)

rings. 24 shirt stockings
earrings 16 dress vest (- jacket)
brooches 11 skirt nylons
pendants 10 blouse cape
watches 6 suit boots
chains 5 slacks sandals
anklets 4
beads 3 jacket tie
pearls 3 coat girdle
bangles 2 sweater belt
charms 2 sweatshirt scarf
perfumes 2 underpants mittens
tiaras 2 sports jacket slippers
trinkets 2 jumper hat
beetlecrushers panties gloves
choker socks apron
cufflinks parka earmuffs
diamonds pajamas handkerchief
diamond rings (a vest) purse (- handbag)
undershirt
drainpipes overcoat hairband
hairslides nightgown ring
hair decorations raincoat earrings
head bands bathing suit watch
home made pottery bathrobe cufflinks
mugs
lockets [necklace]
slip
bra bracelet
papier mache
f lowers
pieces of string shoes cane
rosaries
scarves

Intersection: 7 Union: 76

percentage overlap: 9.2

 Chapter 4
Table 4.4

6 "What about if you carry a gun or something like that? "

Responses Rosch

knife(s) 27 [gun] ice pick

pistol 10 pistol hatchet
dagger 6 revolver slingshot (s catapult)
sword 6 machine gun fists
revolver 5 rifle axe
rifle 5 knife bow
axe 4 dagger razor
club 4 shotgun razor blade
truncheon 4 sword rocket
bow and arrow 3 bomb judo
cosh 3 hand grenade stick
spear 3 A-bomb poison
chain 2 bayonet rock
crossbow 2 spear stone
dart 2 bazooka gas
razor 2 cannon chain
addresses of terrorists bow and arrow scissors
aerosol club bricks
baton lance pitchfork
bludgeon brass knuckles hammer
bottle bullet words
brick mortar hand
catapult arrow pipe
false passport tank rope
hand grenade teargas airplane
hat pin missile foot
hatchet whip car
knuckleduster screwdriver
lead-filled cosh glass
lead pipe shoes
light sabre
mortar "
phial of poison gas
pickaxe handle
poison
rope
sharp stick
shield
staff
stake
stick
weight
whip

Intersection: 27 Union: 77

percentage overlap: 35.1

 Chapter 4
Table 4.4

9 "She said they eat rice and that, didn't she? "

Responses Rosch

curry(ies) 12 pea potato

bread 3 carrot parsnip
corn 3 green beans watercress
macaroni 3 string beans leek
maize 3 spinach sweet potato
potatoes 3 broccoli parsley
bamboo shoots 2 asparagus mushroom
beans 2 corn avocado
fish 2 cauliflower rhubarb
oats 2 brussels sprouts kale
rice 2 lettuce pickles
spaghetti 2 celery baked beans
tapioca 2 cucumber pumpkin
water 2 beets seaweed
barley greens garlic
bran tomato dandelion
artichokes [rice]
chicken
chow mein turnip
dogs eggplant (- aubergine)
dried fish peppers
flies radishes
lentils onions
mangoes bean
meal
pepper
pilaf
polenta
semolina
soya beans
stuffed pepper
sweet and sour
sweet potatoes
wheat

Intersection: 7 Union: 66

percentage overlap: 10.6

11 "She said it was olives or something"

Responses Rosch

cherries 7 orange
gherkins 7 apple
grapes 7 banana
figs. 6 peach
pickled onions 4 pear
plums 4 apricot
anchovies 3 tangerine
dates 3 plum
garlic 3 grapes
onions 3 nectarine
peppers 3 strawberry,
chives 2 grapefruit
pickle(s) 2
prunes 2 cherry
stuffed olives 2 pineapple
almonds blackberry
aubergines melon
biscuits raspberry
broccoli lemon
brown olives lime
carrots fig
cashew nuts mango
courgettes pomegranate
crisps cranberry
cucumbers prunes
gherkin gooseberry
gooseberries date
green olives raisin
lemons coconut
nuts avocado
okra tomato
peanuts nut
[olive]
pickled cucumbers
pimentoes pickle
pineapple
pomegranates
sheeps eyes
stuffed baby egg plants
sultanas
tomatoes
vol-au-vent

Intersection: 16 Union: 58

percentage overlap: 27.6

 Chapter 4
Table 4.4

14 "Okay so you take the trousers or whatever and you spray

them with this spray"

Responses Rosch

shirt(s) 17 [pants (a trousers)] handkerchief

coat(s) 13 shirt purse (= handbag)
jacket(s) 13 'dress hairband
skirt(s) 13 skirt ring
shorts 8 blouse earrings
jumper(s) 8 suit watch
jeans 5 slacks cuff links
socks 5 jacket necklace
dress(es) 5 coat bracelet
blouse(s) 2 sweatshirt
boots 2 underpants
hat 2 sports jacket
pants 2 jumper
pullover 2 panties
shoes 2 socks
suit 2 parka
underpants 2 pajamas
waistcoat 2 undershirt (e vest)
vest(s) 2 overcoat
cagoule nightgown
cloth bag raincoat
cloth bathing suit
costume bathrobe
culottes slip
drain-pipes bra
dungarees shoes
gloves stockings
handbag vest (- jacket, Br)
mack nylons
material cape
overalls boots
rucksack sandals
sacks tie
short trousers girdle
sleeping bag belt
suede shoes scarf
swimming-trunks mittens
tent slippers
tie hat
towel gloves
trousers apron
trunks earmuffs

Intersection: 26 Union: 66

percentage overlap: 39.4

 Chapter 4
Table 4.4

16 "... a car or something came by... "

Responses Rosch

lorry 31 automobile trailer

van 29 station wagon cart
bus 23 truck (a lorry) wheelchair
motor-bike 17 [car] yacht
truck 9 bus tank
coach taxi go-cart
,7 jeep
bicycle 5 rowboat
taxi 3 ambulance dogsled
wagon 3 van tricycle
articulated lorry 2 Honda canoe
dustbin lorry 2 cable car raft
juggernaut 2 train submarine
milk cart 2 trolley (car) sled
moped 2 bicycle horse
small van 2 carriage rocket
ambulance airplane skates
armoured car bike camel
bike with attached boat feet
side-car
boy jet skis
car ship skateboard
caravan scooter wheelbarrow
cattletruck tractor surfboard
cement mixer wagon
donkey
dormobile
fire engine
girl
heavy goods vehicle
horse
invalid-car
man
motorcycle with sidecar
police-van
racing car
small truck
station wagon
steam-roller
three-wheel-car
traction engine
tractor
woman

Intersection: 24 Union: 61

percentage overlap: 39.3

 Chapter 4
Table 4.4

"17 "He used a screwdriver or something"

Responses Rosch

spanner 12 gun gas

knife 11 pistol chain
hammer 10 revolver scissors
9 rifle bricks
chisel
nailfile 9 knife pitchfork
wrench 4 dagger hammer
bradawl 3 shotgun words
pliers 3 sword hand
hairclip. 3
crowbar 3 bomb pipe
drill 2 hand grenade rope
file 2 A-bomb airplane
pen 2 bayonet foot
pencil 2 spear car
2 bazooka [screwdriver]
penknife'
rod 2 cannon glass
stick 2 bow and arrow shoes
allen key club
awl lance
crow bar brass knuckles
edge of a coin bullet
finger mortar
gimlet arrow
high-power drill tank
iron rod teargas
jack missile
knife blade whip
lever ice pick
hatchet
mallet
slingshot (- catapult
metal rod
fists
nail
penknife blade axe
ratchet bow
scissors razor
screwdriver razor blade
screwed rocket
sixpence judo
small file stick
strong finger nail poison
toe tap rock
turned stone
twig
vice

Intersection: 7 Union: 90-

percentage overlap: 7.8

20 "Have you got a wheelbarrow or anything like that which
we could borrow? "

Responses Rosch

cart " 14 automobile

trolley 13 station wagon
trailer 7 truck
bucket 5 car
pram 3 bus
spade 3 taxi
t ruck 3 jeep
old pram 2 ambulance
rake 2 motorcycle
sack 2 van
shovel(s) 2 Honda
van 2 cable car
box(es) 2 train
articulated lorry trolley (car)
barrel bicycle
barrow carriage
basket airplane
bogey bike
box cart boat
box on wheels jet
carrier ship
carrier-bag scooter
car, boot of which can be used tractor
dumper wagon
fork trailer
hand-truck cart
hand cart wheelchair
hand trailer yacht
hob tank
hod go-cart
hoe rowboat
large box dogsled
lawn mower tricycle
little truck canoe
push chair raft
scythe submarine
shopping jeep sled
trolley horse
shopping
skip if it is big rubbish rocket
sled skates
tractor camel
tractorlette with wagon feet
transporter skis
tray for carrying rubbish skateboard
trowel [wheelbarrow]
wheelbarrow surfboard
wheels on a base

Intersection: 11 Union: 82

percentage overlap: 13.4

 Chapter 4
Table 4.4

22 "... some new trousers or something... "

Responses Rosch

skirt(s) 20 [pants (- trousers)] hat

dress 15 shirt gloves
shoes 12 dress apron
coat 11 skirt earmuffs
blouse(s) 11 blouse handkerchief
jumper(s) 11 suit purse (- handbag)
jeans 9 slacks hairband
shirt 9 jacket ring
jacket 6 coat earrings
hat 4 sweater watch
socks 4 sweatshirt cuff links
boots 3 underpants necklace
dungarees- 3 sports jacket bracelet
slacks' 3 jumper cane
tights 3 panties
bra 2 socks
cardigan 2 parka
culottes 2 pajamas
new shoes 2 undershirt (" vest)
stockings 2 overcoat
vest 2 nightgown
corset raincoat
handbag bathing suit
jersey bathrobe
jump suit slip
new blouse bra
new dress shoes
pants stockings
record vest (= jacket)
scarf nylons
shorts cape
ski outfit boots
suit sandals
suspenders tie
suspender belt girdle
tee-shirts belt
tie scarf
track suit mittens
trouser suit slippers

Intersection: 25 Union: 65

percentage overlap: 38.5

 Table 4.4
1
24 "They've got a car and that"
Responses Rosch

caravan 10 automobile
washing-machine 8 station wagon
tv 7 truck
tv 6 [car]
colour
motor-bike 6 bus
boat 5 taxi
dishwasher 5 jeep
house 5 ambulance
trailer 5 motorcycle
freezer (deep-freeze) 4
nice house 4 van
big house 3 Honda
record player 3 cable car
scooter 3 train
bicycle 2 trolley (car)
bike 2 bicycle
f ridge 2 carriage
mortgage 2 airplane
stereo 2 bike
annual holidays boat
automatic washer jet
bus ship
canoe scooter
carpets tractor
car accessories wagon
cassette player trailer
cat cart
decent looking garden wheelchair
dog yacht
expensive house tank
fitted , go-cart
carpets
garage rowboat
good job dogsled
holiday bungalow tricycle
income canoe
money raft
moped submarine
two children sled
music centre horse
outboard motor rocket
posh car skates
pots of money camel
private house feet
.
roof rack skis
servants skateboard
steady income wheelbarrow
swimming-pool surfboard
telephone
tent
three bedrooms
transport
two toilets
vacuum
washing up machine
water skis
3- bedroomed semidetached

Intersection: 9 Union: 93
percentage overlap: 9.7
 Chapter 4
Table 4.4

25 "... guns and things like that"

Responses Rosch

knives' 27 [gun] rope

pistols 10 pistol airplane
9 revolver foot
rifles
ammunition 5 machine gun car
chains 5 rifle screwdriver
hand grenades 5 knife glass
bullets dagger shoes
.4 4
revolvers shotgun
swords 4 sword
clubs 3 bomb
coshes 3 hand grenade
daggers 3 A-bomb
grenades 3 bayonet
bombs 2 spear
bottles 2 bazooka
cameras" 2 cannon
bows and arrows 2 bow and arrow
gun powder 2 club
tanks 2 lance
pen knives, 2
airguns brass knuckles
atom-bombs bullet
big sharpened sticks mortar
bomb-making equipment arrow
Boyes anti-tank rifles tank
cans teargas
cassettes missile
contraband whip
cut throat razors ice pick
dangerous drugs hatchet
explosives slingshot catapult)
flamethrowers fists
gelignite axe
hatchets bow
inflammable liquids razor
keys razor blade
knitting needles rocket
knuckle dusters judo
mortars stick
nuclear weapons poison
razors rock
razor blades stone
rockets gas
sawn-off shotguns chain
sharpened scythes scissors
shells bricks
stakes pitchfork
sticks hammer
taperecorders words
tapes hand
truncheons pipe

Intersection: 24 Union: 85

percentage overlap: 28.2

27 "oranges and things like that"
Responses Rosch

apples 31 [orange]
bananas 23 apple
pears 23 banana
grapes 17 peach
peaches 11 pear
tangerines 6 apricot
plums 6 tangerine
cheese and biscuits 5 plum
melons 5 grapes
apricots 4 nectarine
,
pineapples 4 strawberry
satsumas 4 grapefruit
cheese 3 cherry
lemons 3 pineapple
mandarins 3 blackberry
nuts 3 melon
.
pomegranates 3 raspberry
sweets 3 lemon
biscuits 2 lime
grapefruit 2 fig
cake(s) mango
.2
avocado pear cranberry
citrus fruits prunes
damsons gooseberry
dates date
fresh fruit salad raisin
mars bars coconut
strawberries avocado
tomato
nut
olive
pickle

Intersection: 17 Union: 43

percentage overlap: 39.5

 Chapter 4
Table 4.4

30 "They never give us peas or anything"

Responses Rosch

beans 21 [peal baked beans

carrots 21 carrot pumpkin
cabbage 19 green beans seaweed
sprouts 14 string beans garlic
8 spinach dandelion
cauliflower
potatoes 8 broccoli rice
broccoli 6 asparagus
greens 3 corn
spinach 3 cauliflower
turnip(s) 3 brussels sprouts
green beans 2 lettuce
tomatoes 2 celery
asparagus cucumber
aubergine beets
baked beans greens
beetroot tomato
beets artichokes
bread turnip
broad beans eggplant (- aubergine)
corn peppers
cucumber radishes
finocchio onions
jam tarts bean
kale potato
lentils parsnip
marrow watercress
onions leek
paprika sweet potato
peppers parsley
runner beans mushroom
silver beet avocado
swede rhubarb
sweetcorn kale
water pickles

Intersection: 24 Union: 49

percentage overlap: 49.0

 Chapter 4
Tables

Table 4.5

Roach-type test items:

Type 2 (Category Identifying) Responses in Rank Order

1 "Could you get me some oranges or something at the market? "

fruit 5
any fruit 4
food 3
another type of fruit
anything juicy
anything round
any other fruit
a food containing vitamin C
citrus fruits
food of some kind
fresh fruit
fruit containing vitamin C
grub
oranges if you can anything if not
oranges or some other fruit
or something to have for a certain meal
similar fruit
something else to eat
something for colds
something for dessert
something similar
some fruit similar to oranges
some other thing which will give pleasure
some other types of fruit
some other type of fruit
some table fruit
vegetables
 Chapter 4
Tables

3 "I love peas and things. like that"

No. of subjects

vegetables 9
green vegetables 3
small vegetables 2
and all vegetables
and other green foods
and other green vegetables
and other round vegetables
any green vegetable
any vegetable (s) 3
brightly coloured food
dishes made from, or with peas
fresh goods in general
green cooked vegetables
green vegetables in general, excluding that is turnips
most vegetables
natural goods
non-tastable foods
other vegetables 2
other vegetables like peas, beans etc
pulses
small foods
small round objects
small spherical objects
some vegetables
sweet things
tasty things
things coming in pods
vegetables like that

Note: number of mentions of word "vegetable(s)" 28

4 "... necklaces and things like that... "

jewelry 13
other jewelry 2
any jewelry
baubles
body adornments
decoration
even items other than jewelry
expensive
hand-made decorations
jewelry in general
necklaces and other expensive items
neckwear
only necklaces
other jewels
precious objects
similar jewelry
valuables
valuable items
valuable objects
 Chapter 4
Tables

6 "What about if you carry a gun or something like that? "

I
" No. of Subjects

weapon 12
firearm 3
another weapon 2
dangerous weapon 2
another firearm
another metal object
any implement which could be put to violent (illegible)
any weapon
dangerous implement
defence
something else dangerous
something else illegal
something heavy and or lethal
something-metal
something to frighten assailants
sup term: deadly weapon

9 "She said they eat rice and that, didn't she? "

vegetables 6
pasta 4
chinese food 3
carbohydrates 2
foreign food 2
health foods 2
Indian food 2
spices 2
and all the other things that group is known to eat
= dal, chapattis etc
carbohydrates common to those people being spoken about
cereal
cereals
Chinese take-way food
dried vegetables
Eastern foods
exotic vegetables
foodstuff
food grown in the area
green vegetables starchy stuff
little meat third world diet
other carbohydrates vegetables only
other Chinese food vegetarian food
other dry crops grown sup term: farinaceous food
other things like rice
polysaccarides
poorer grain
pulses
rice dishes
roots
simple food
small hard starchy food
some vegetables
spaghetti, pasta. generally
spicy foods
staples like rice
starch
 Chapter 4
Tables

11 "She said it was olives or something"

No. of subjects

something similar 2
some other food 2
drink situation goodies
foreign fruit
fruit
hors d'oeuvre
other sharp fruit
small savoury objects: vegetables
something edible
something equally exotic, strange, rare etc
something irrelevant
something like olives
something salty
some other food similar to olives
some type of exotic foodstuff
spices
uncommon fruit with strange taste

14 "Okay so you take the trousers or whatever and you spray them
with this spray"

clothes 3
clothing 3
garment(s) 3
anything you want waterproofing
any material
any similar clothing to trousers
any soiled (emph) garment
article needing to be ironed
article of clothing
article of outside clothing
material
other clothes
other clothing
underwear
whatever else you fancy spraying
whatever modesty forbids my particularizing
whatever they're wearing
whatever you're using the material from
(gloss) other garment

16 "... a car or something came by... "

vehicle 6
some other vehicle 2
another vehicle
any other road-using motorised vehicle
any relatively fast vehicle
motor vehicle
not bus not motor-bike, wants to indicate nots
something moving fast and close to me
some moving object
 Chapter 4
Tables

17 "He used a screwdriver or something"

No. of subjects

tool 5
another tool
implement
instrument
other tool same shape
something long and thin
something of the same shaped end as a screwdriver ie could
turn screws equally well
something that undoes screws similar job as screwdriver
some kind of tool
some long thin object

20 "Have you got a wheelbarrow or anything like that which we could

borrow? "

anything which I could move something in like the loan of you car
or perhaps yourself
any garden equipment (eg for an exhibition)
a large 'container'
a large container which has wheels and can be pushed or drawn
small garden vehicle with room to put things
something for transporting things
something I can carry things in
something mobile
something that can carry heavy bricks
something to carry things in
something very large
something with wheels
useful receptacle

22 "some new trousers or something... "

clothes 5
clothing 2
anything new
anything not necessarily clothing
any small article which costs money
article of clothing
clothing that she didn't really need
expensive items of clothing
extravagant impractical garment
fashionable garment
'has to include trouser element'
items of clothing of any size
luxuries
Piece of clothing
other article of clothing
other items of clothing as well
other purchases
something expensive
something new
some article of clothing
some new clothes
some weird stype of trousers
things from shop
 Chapter 4
Tables

24 "They've got a car and that"

No. of subjects

other luxuries 2
any material goods of a middle-class nature
children at good schools
double garage and semi-detachesd house
everything you could wish for
expensive things
good jobs or professions
large house and garage
other modern assets eg fridge, tv, spindryer, phone, etc
other things in same category
trappings of the bourgeoisie

25 " guns and things like that"

...

weapons 11
dangerous weapons 3
offensive weapons 3
any weapons 2
arms 2
dangerous objects 2
metal objects 2
sharp objects
all metal objects
all other army gear
all weapons
any dangerous weapon
dangerous things
explosives etc
firearms
forbidden articles: weapons, firearms etc
heavy things
implements
objects
only firearms
other offensive weapons
other service and military equipment
other weapons
other weapons any kind
sharp implements

27 "... oranges and things like that"

fruit 9
all other fruit, ie mixed bowl of fruit
any fruit
any{jnd of fresh fruit
citrus fruits
different kinds of fruit
food from home
fresh fruit
ie fresh fruit
larger fruits in general
other fresh fruit
other fruit
other fruit in season
pudding
 Chapter 4
Tables

(27 continued)

simple dessert without preparation

something uncooked
sup term: fruit
things that are good for you
things that don't need cooking
things we've got lying around anyway
unprepared dessert

30 "They never give us peas- or anything"

No. of subjects

vegetables 7
all green vegetables
another vegetable
anything 'colourful' to eat
anything good for us
anything to supplement a meal
anything with vitamins
any food at all
any kind of vegetables
any other green vegetable
any other vegetable
any other vegetables
cooked vegetables to supplement the meal
good food
green vegetables
interesting vegetables
never give us green vegetables
not potatoes or farinaceous vegetables
salad
school meals vegetables
something besides meat and potatoes
something else they like as well as peas
 Chapter 4
Table 4.6

Table 4.6

Attested Test Items: Results

"
2 "One of the secretaries was saying there was a film or something...

Type 1 Responses No of subjects

lecture 15
(inc. show) 9
slides slide
6
talk
(video showing, video tape) 6
video programme, video
4
play
4
show
discussion 3
2
seminar
2
tape recording
TV programme 2
audio-visual tape
cabaret
cartoon
concert
dance
demonstration
disco
display
documentary
film of experimental programme
film show
illustrated lecture
illustrated talk
open class
photograph slides
production
radio broadcast
slides and commentary
slides and recorded speech
strip
tape session

Type 2 Responses '

agreeable pastime relevant to course

entertainment
film or something like a film
night entertainment
not a lecture, informal situation
other entertainment
something to see, watch
something unusual
 Chapter 4
Table 4.6

7 "... 23578 or something like that"

22 subjects (out of 39) gave at least one 5-figure combination in

their answer

The other responses were all type 2 (except for one type 3),
as follows:
numbers 3
sequences 2
in sequence of 4
in sequence of 3
in sequence of 2
in sequence of 1

abbreviations
another easily remembered sequence
any five figure number
any number beginning and ending with two consecutives
any unknown previously number
an easy combination
any easy string of numbers
ascending numbers
a logical sequence of numbers
a simple sequence
calculation
consecutive numbers
21488 etc, rhythm
letters
numbers related to make them easier to remember

numbers in order
numbers in the right order
number in order
number plate
only five numbers
ordered according to size
other numbers with five digits
pattern
predictability
relationship - measurable
series of numbers
set patterns
short numbers
similar number
something similar to example
something which sticks in the mind
some range of figures
some sort of aid to memory
some sort of code
5 numbers the same
 Chapter 4
Table 4.6

5 "... whether it be language or anything else... "

Type 1

maths 13
science(s) (etc) 10
history 8
geography 8
physics 6
ideas 4
riding a bike 3
art 2
arts 2
biology 2
chemistry 2
English 2
information 2
literature 2
philosophy 2
skills 2
sport 2
walking 2
crafts 2
Arabic
arithmetical knowledge
art appreciation
basket weaving
black hole
Celtic studies
codes
concepts of space
concepts of time
cultural attitudes
dancing
experience
French
how to get home from school
job
knowledge
learning
life
living
music history
music
psychology
R. E.
skill at sports
skill
social learning
sociology
technique
theology
thinking
tissue
who is related to who
 Chapter 4
Table 4.6

5 continued

Type 2

another subject
anything humans do
any other activity
any other kind of learning
any subject
any subjects requiring learning of facts
educational subjects
facts and figures
general cognitive requirements
general knowledge
humanities
material picked up from environment
other subject
recognition of people and places
scientific knowledge
something else we don't know about
subjects

8 "... sentences with it and things like that... "

Type 1

phrases 12
examples 8
clauses 3
paragraphs 3
explanations 2
words 2
approximations
circumlocutions
clarifying examples
complicated explanations
context
contrasting words
current usage
demonstrations
demonstrations with mime
diagrams
drawing in the air
expressions
generalisations
hand waving and gestures
its meaning
lines
long-winded complex explanations
miming
paraphrases
passages
physical examples
physics
pronouns
roundabout definitions
situations
 Chapter 4
Table 4.6

8 continued

sketches
stories
this booklet
this word
usages

Type 2

and all that stuff

anywhere this thing cropped up
different words to explain it
examples of how it's used
other examples
other exercises of the same kind
other types of speech
other words
phrases with it in
ways of saying it
roundabout
words without visible objects

10 "... what they want is what they need or whatever"

Type 1

desire 6
food 3
require 3
feel 2
lack 2
attention
comfort
drink
intelligent
meaning
milk
response
satisfaction of curiosity
socialisation
their desires
to have
wrong

Type 2

anything 2
cannot do without
have to have
or seem to need
perhaps something else
stimulating change of environment
that they need it
whatever they want
what they'd like
what they desire
what they dislike
what they think they need
 Chapter 4
Table 4.6

10 continued

what they want

what they want is to be fed
what they want is to be held
what they want is to be noticed
will need
wish to get rid of
would like to make them comfortable

12a "... they had these sort of lectures and that... "

Type 1

seminars 13
tutorials 9
talks 7
discussions 6
lessons 5
5
meetings
4
classes
chats
confrontations
films
information
little talks
periods
sermons
study
teachers
tete-a-tete
work

Type 2

formal teaching 2
contact
etc
formal classes, distance maintained between them
formal talks
impersonal teaching methods
other forms of teaching
similar lessons
stern chats
things like lectures
things one is obliged to attend
whatever
 Chapter 4
Table 4.6

12b "... there was no coffee room or anything"

Type 1 No. of Subjects

common room 12
bar 9
canteen 4
lounge 4
games room 2
snack bar 2
cafe 2
rest room 2
JCR
quiet room
reading room
recreation room
refectory
staffroom
tea room
toilet
tv room-
union bar

Type 2

meeting place 4
social room (social area, social base) 4
anything which promoted contact
any communal life (or possibility for)
any communal room
any other facility
discussion
facilities
fellow students
informal contact
informal contact, not friends
nothing comfortable
no food
no place to sit
place to meet
relaxation
room for relaxation

I
 Chapter 4
Table 4.6

13 "But lots of big sort of important numbers that you have to

remember and things, I'm sure... "

Type 1 No. of Subjects

(where set exemplar appears to be important numbers)

dates 7
formulas 7
figures 4
letters 3
names 3
words 3
facts
amounts
car number plates
chemical symbols
data
dates of births
demographic calculations
diagrams
equasions
exams
grammatical rules
historical events
important dates
important words
lists
long quotations
mathematical data
oil tanker crashes
percentages of the population
phone numbers
places
series
shapes
statistics
tables
telephone numbers
dates to learn

Type 2

(where set exemplar appears to be important numbers)

information
anything hard to remember
numbers to be used without reference
other facts
other important things
other numbers 2
other things (facts) to be remembered
patterns
things containing the important number
which are important to you
 Chapter 4
Table 4.6

13 continued

Type 1

(where set exemplar appears to be remember)

recall 2
repeat 2
understand 2
learn
and do things with
and have at hand
and learn
call to mind at will
deduce
differentiate between
distinguish from one another
identify
know
manipulate
refer to
state
store in memory
to use in calculations
to work with
use
utilize

Type 2

(where set exemplar appears. to be remember)

none
 Chapter 4
Table 4.6

15 "But what about things like when you read sentences or something... "

Type 1

(where set exemplar appears to be sentences)

paragraphs 25
phrases 24
passages 8
words 7
articles 5
chapters 4
books 4
clauses 2
expressions 2
poems 2
quotations 2
addresses
argumentative articles
descriptions
digest
document
essays 2
experiment
extracts
given piece of reading material
groups of words
half sentences
histories
holophrases
jokes
lines of verse
lines
lists
meaning
meaningful groups of words
notes
pages
paraphrasing
quotes
reports
sections or units of a dialogue
sentences
statements
stories
summaries
syllables
telephone numbers
texts

Type 2

(where set exemplar appears to be sentences)

anything written
do similar comprehension tests
instructions ie road signs
some written work
 Chapter 4
Table 4.6

15 continued

Type 1

(where set appeared to be read... )

exemplar

learn
listen to sentences
look at something
read anything-
read unrelated words
say
sort passages

18 "... there were no lectures or anything like that... "

Type 1

classes 6
lessons 6
films 3
tutorials 3
discussions 3
talks 2
conferences 2
assessment
debates
dictations
essays
examinations
exhibitions
grading
groups
marks
practicals
putting in order, ranking
teachers
teach-ins

Type 2

formal classes 2
formal talks 2
formal teaching sessions 2
anything formal, obviously teacher-pupil
anything that there should have been
any kind of tuition where you can take notes and are
talked to entirely by a professor
big groups
big lessons
discussions apart from seminar
formal conventional methods
formal dictation
formal lessons
formal situations, professor makes a speech
formal teaching methods
hard work
informative talks
 Chapter 4
Table 4.6

18 continued

instruction
large formal teaching groups
large groups in lessons
lessons where no participation expected
little teacher participation
nothing easy to cope with
one-many talks
professors teaching in (illegible) intimidating
proper teaching
serious classes
set curriculum
set talks
talks apart fromseminars
talks compulsory attendance
talks to lots of people
teaching
teaching without feedback
traditional teaching
usual expected methods

19 "elaborated code and things like that"

Type 1

slang 3
accent(s) 3
class dialect 2
jargon 2
regional dialect 2
academic language
articulate speech
believed
body language
class accent
"clean" language
clear speech
colour dialect
customs
cycles
dialect
diglossia
grammar
identify
individuality
jargons of particular jobs
language of a clique
local language
localphrases
methods
morse
new language
punctuation
regional accent
semaphore
sign
superior dialect
theories
19 continued

updated slang

Type 2

language
and all those other indefinable linguistic terms
black English not applicable
complex sentence structures
conforming language
correct English
elegant speech
exclusive vocabularies
lingustic attitudes
or something
other experiments
other kinds of code
own language
pedantic language
posh words
secret verbal communication systems
similar things
sociolinguistic patterns
speech forms
types of language
usual writing language
various formulas
vocabulary
ways of speaking
 Chapter 4
Table 4.6

21 "I hope we didn't have lots of horrible conversations when you went
the room in tutorials and things like that... "
out of

Type 1 No. of subjects

seminars 13
lectures 9
lessons 5
gossip 4
discussions 3
meetings 3
arguments 2
calumny
chats
class discussion
courses
damning conclusions
dentist
difference of opinion
English
fights
form period
geography
get togethers
had fights
history
jeering talk
jokes
library monitors' duty
maths
mutterings
personality destruction
played silly games
prefect duty
private meetings
private study
RE
rude jokes
rumours
scandal
small lectures
small seminars
sneaky remarks
snide remarks
swearing
TD
when out of hearing range
when we were in a group with you
when we were talking all together
whispering

Type 2

classes 2
anything like a horrible conversation
informative classes
in classes
meetings where both staff and students were present
other conversations
 Chapter 4
Table 4.6

23 "... 1972 or something like that"

Type 1

68 responses were four-figure year dates

Other type 1 responses:

ABCD
16
8.8.60
a recent date
a reoccurring date
a used date
catchphrase
cliche
UCCA
YMCA

Type 2
mnemonics 2
4x4x4x4 etc
1948,1976 etc
another year
any other mnemonic
any other relative date
any year
birthdays
early 70's
figures
four syllable words
numbers of relative importance for historic or personal
reasons
other easily remembered numbers
other relevant dates
other similar dates
paradigms
phone numbers-
sequence
significant year names, eg 2000,1984,1p00-
some type of group of numbers
telephone numbers
the Olympics
various other dates
years of the European soccer championship
years of the Olympic Games
1066 and all that
 Chapter 4
Table 4.6

26 "... could we sort of meet or something... "

Type 1 No. of Subjects

get together 11
discuss (discussions) 11
talk 7
have a chat 4
have a discussion 3
chat 2
confer 2
have a class 2
have a tutorial 2
rendezvous 2
see each other 2
and talk
arrange a rendezvous
arrange a time to discuss them
a revision period
be given instructions
collect ata certain place
come and see you
congregate
could I visit you
debate
discuss privately
extra classes
gather
gather together
gossip .
go round to teacher's house
go through them
have a chance to talk them over
have a class discussion
have a conversation
have a general meeting to see about and problems arising
from the essay
have a talk
have individual meetings
hold a discussion
join together
liaise
phone each other up
private appointment between teacher and student
private discussion
see you if we have any problem
seminars
speak at the end of the lesson
talk about it
talk informally about the essay
talk to you about them
write a letter
write a note

Type2

arrange somehow to see each other

be in contact
find time to discuss them
get a message round
get in touch by phone or letter
26 continued

give me a chance to ask you what you meant

informal discussion
or get together some time

28 "... we're having second supervisors"

10 or something"

Type 1

inspectors
organisers
tutors
advisers
assistants
bosses
deputy supervisors
elected tutors
essays
examinations
extra teacher
fifth supervisors
fourth supervisors
helpers
obligatory classes
observation
observers
optional classes
or third supervisors
porters
somebody
student teacher
student help
undergraduates

Tvpe 2

a different professor again

equivalent of second supervisors
other people in the same category
other supervisors
overlookers
something like that
some change in normal routine
some other person in authority
some people
stand-ins
supervisors
teachers
 Chapter 4
Table 4.6

29 "... they'd read these words in a book or something"

Type 1 No. of Subjects

magazine 34
newspaper 24
comic 19
television 5
article 3
advertisement 2
journal 2
letter 2
poster 3
pamphlet 2
paper 2
story 2
be shown these words
book in a foreign language
brand names on packages
brochure
brother or sister's school report
description of scientific experiment
dictionary
essay
exercise (grammar)
heard or seen on tv or radio
heard them in conversation
history book
hoarding
in the street
minutes of the AGM
newspaper article
on a card
on the radio
public sign
record
shop window
signpost
they'd hear them
toilet door

Type 2

came across them in reading

other methods of communication
perhaps they'd hear the words in speech
seen these words written down somewhere
magazine etc
 Chapter 5

Chapter 5

Being Vague

5.1 Introduction

In previous chapters I have introduced the idea of vague language use,

and -have looked in detail at the structure and meaning of two

particular types of vague expressions from the set of those I have

called vague additives. Having established for both types of

expressions a set of observations about what they mean, I now go on to

describe and discuss, in this chapter, the way that speakers use these

expressions in conversations. In particular what conversational

effects arise from using such vague expressions, and what goals do

speakers use them to achieve?

5.2 Analytical Approach

The approach taken is broadly in the spirit of the work on

conversational interaction known as CONVERSATION ANALYSIS (hereinafter

CA), as summarised for example in Wootton, 1981:

"In approaching interaction, then, it is the problems

confronting participants which are of interest in CA, and the
systematic procedures and designs through which such problems
are displayed and resolved. There is no prior analytic
'theory' of interaction being applied. "
(: 103)

Thus I would share the analytic predilection sketched by Schenkein

(1978: 2) - that is to "take[ing] seriously the details of the natural

interactions themselves". and I. try to be as true to the data, as

possible by validating the analyses and categories I propose solely by

reference to that is observable within the conversational extracts[1].

138
 Chapter 5

This general approach means that the material in this Chapter has

the appearance of being somewhat random, in the sense that I just set

out to describe the observations I have made of the data I have. It

is only at the end that I try to link'these observations together in

any schematic way. In addition, many examples show either a number of

different conversational effects, 'or else it is impossible to sort out

quite what the effect is. The fact that hearers may themselves quite

often be in this position is itself of interest.

One difference between my work and CA is that I look at written

examples as well as spoken, especially written examples which seek,

for particular effects, to imitate or be associated with- spoken

discourse. These are useful because they can show up those aspects of

spoken discourse which language users judge to particularly identify

its "spokenness".

I begin by looking in some detail at three different kinds of

extracts. These serve to introduce both the complexity of the effects

under observation, and the kinds of categories I want to suggest.

5.3 Three scenarios for vagueness

5.3.1 "Eighty or so pence" - S2si_ t ýj

In this example taken from an advertisement, it is'not really possible

to observe what communicative effects are understood by its readers.

But it is possible, as a reader oneself, to list what they might be.

1 This is not unfortunately always possible. For examples where the

data is incomplete (eg no recording, only a short extract),
assumptions based on analogy have had to be made.

139
BEST COPY
AVAILABLE

Variable print
quality
 Ijie important
i

an ski
ichye
Vichy'. dedicatethemselves.solely' to the care the skin.
-of
For-aclean=healthy skin is Where beauty begins.
It is e sentialthat you cleanseyour facenight andinoi nm -.
but equallyimportant that you cleanseit gently
Thafs the laboratories at Vichy have produced Vichy
CleansingMilk-why
and Vichy Tonic-Lotion. c'
Applied with the fingertipsand lightly massaged into your
skin, the mild ingredients in Vichy Cleansing Milk
"gently absorb impurities.These can then be, -
.
removed with cotton wool or tissue.
After"this, anýipplicationof Vichy'
Tonic Lotion seeksout andremovesthe
last tracesof the CleansingMill fl
leavingyour skin clean,soft
and refreshed.
More clean,soft and
refreshed,perhaps, than it has
ever been.
As your introduction to
their way of skin care,Vichy ..
would like you to receivethis pack
of their CleansingMilk andTonic
Lotion for £3.
This does saveyou eighty or
so pence.But it is asnothing to the
kindness you will be showing
your skin.
nOl'mal
ýýý1 ýý ýý. a2SII1

v
L_1_. ý f it/i
11
iZ.:

Lait ct Lotion. Fornr. aatcd for ý mar kin

17 c
,v \ r,
1ýý L. t apc
Jý1. 11 W
ýýý.

.,
 I
Chapter 5

Now, presumably Vichy know the retail prices of the products

referred to. Brands of this up-market type do not normally allow

shops to vary the prices of their products. Therefore they know

exactly how much "you" are saving. The copywriter must have

deliberately included an approximation to achieve one, or several,

particular effects.

1 The actual number is lower, say 77p. Rounding up to 80, and

approximating, implicates to' the naive, or hasty, a greater saving

than is really the case. (This is similar to the price tag ploy:

"only 14.99". )

2 For the purpose of the advertisement, the actual saving is not very

important. This approximation gives as much information as is

necessary, and knowing the exact figure won't give the reader

information he needs to know, it will be redundant. The relative

unimportance of the sum of money involved is confirmed by the content

of the next sentence.

3 It is for self-protection. Retail prices do vary, so savings will

be different in different shops. The advertiser ensures he is telling

the truth ("legal, honest, decent and truthful") by using a hedge. [2]

4 It sets a tone of chatty informality - phatic communion between

advertiser and reader. Test informants who were asked about use of

vagueness (in semi-formal group discussions conducted at the

conclusion of the tests described in Chapters 3 and 4) suggested that

tags are not used in formal types of writing, or in more formal

2I once wrote to the Advertising Standards Authority suggesting

that they should employ a person trained in linguistics to uncover all
the misleading statements which are systematically perpetrated by
implicatures in advertisements. Regrettably, but unsurprisingly, they
weren't interested.

140
 Chapter 5

conversations (the sixth-formers' example was "not in an interview").

The copywriter is trying here to render the effect of informal

conversation.

Any or maybe all of these effects could be understood by readers of

this advertisement.

5.3.2 "Elaborated code and things like that"

11,8.3

Extract 1[3]

[Tutorial discussion on innateness of language capacity in humans]

note: A, the tutor, is me

A: What about this business about

E: (laughs)

it doesn't matter how intelligent the

individual is - they all manage to achieve the same "level of
linguistic competence

E: do they

D: "a level anyway

A: well everyone achieves intelligibility

E: yeah yeah okay - within a set

group
"'""-"'"'"'"'-"'I mean you

A: tthere's been a lot of work done to show that there's

no difference in sophistication between one language variety
and another

E: I'm talking about you know sort of acceptable middle class

language and (. ) sort of working class language you know Bernstein and
mm you know sort of elaborated code and things like that

C: well that's more environment is'nt it

E: yeah

3 Since almost all the extracts in this Chapter are attested, I have
not thought it necessary to mark them with a '+'.

141
 Chapter 5

C: it's the way you

you're -------
I
D: most people achieve competence

E: yeah yeah a lot of the time (. ) a lot of the time it's linked to
intelligence as well

A: but is it
E: at least it is in schools - if you can "sound more intelligent by
the way you speak you are categorised as being intelligent

A: well yes but never mind about "categorised as being more

intelligent - you "aren't more intelligent are, you

B: but what do you mean by intelligent

C: its a very good question

E: yeah by middle class based intelligence tests

A: because if you use "other intelligence tests

E: which is what the

elaborated code proved

[continues]

Speaker E precedes her vague exemplar and tag ("elaborated code and

things like that") with considerable hesitation -a pause, two sort

of s, two you knows and an "mm". These are indications that (among

other things) she is not sure what she's talking about. The

subsequent turns all feel able to treat what she's said as wrong, and

they are permitted to do this by her unsureness. Speakers D and A,

and B, all disagree in some way with her. Duncan and Fiske (1977) in

their work on turn-taking, think that expressions like or something

facilitate hearer interruption, but they do not show enough data to

test this assignment. I would say that my data do not show this, what

they do show in some instances is increased possiblity for hearer

disagreement/criticism, as, for example, I discuss below for this

extract.

142
 Chapter 5

Later in the same discussion, E makes explict that (a) she

doesn't know or understand the subject matter of the talk very well,

and (b) that she does not command with any certainty the language

necessary for the topic of discussion, as we can see from the

following extract.

Extract 2

[later in-same discussion, talking about an article by Putnam]

E: I'll tell you what I found difficult in this - all these different
symbols -I didn't know what they meant

A: where

E: in this Putnam thing

C: --__.. »--_.... ------

I
B: .. ----------------

E: that I class as being sigma - the sum of

A: [reading] a highly restrictive sigma class of grammars -

jI don't think you need to worry about that
,
E: but I did -I mean I just sort of worried about it

[... continues with details of losing first page, worrying etc]

A: you mean this bit at the beginning where he says [reading] we

should assume that the speaker has a built-in function which assigns
weights to grammars G1 G2 and G3 /E: mm hhh/ to a certain class sigma
of transformational grammars

E: and immediately I thought [loud] don't understand this put it

down (laugh)

A: I hope you went on reading that because it becomes "much clearer

after that

E: no -I get terrified

[continues with details of nervousness amid general laughter]

[conversation continues]

On this basis, I suggest that one use of vague additives is to enable

a speaker in fact to talk about a subject he is not very knowledgeable

143
 Chapter 5

about, or a subject where he does not know the necessary vocabulary.

On this last point, elaborated code and things like that directs

hearers to access a category. It may well be that there is no clear

term for this category. - Certainly there was no

-superordinate

particular sign of agreement as to what it would be, among the test

subjects (cf results, Table 4.6, Test Item 19). If there is one, E

didn't know it, so her vagueness may show her finding a way of

actually talking about something she does not quite have the

vocabulary to express.

There is a way in which E works around what she means to say,

using items of vocabulary relevant to her point. It is interesting

that speaker C does not appear to be bothered by the multiple

uncertainties in what E has said. She refers back to it with a

definite that. The subsequent turns show I think that all the

speakers decide that what is being talked about is different language

varieties and their relationship to measures of intelligence.

There is another point about what E says. She uses, as noted, a

lot of vagueness. Yet there is evidence that she does not feel that

uncertain about what she is saying. Notice her refusing to give up

the turn to either C or D. As she goes on there is a noticeable

absence of vagueness. Leaving aside the-sort ofs for a moment, it

seems she could have rather successfully defined a category (? or two)

- middle class language and working class language, without the

vagueness, and without her very unclear "elaborated code and things

like that". It is also the case, as I noted, that her hearers react

to. what she says as definite. So perhaps this vagueness is for

something else.

144
 Chapter 5

Notice that the purpose of Es turn in which the vague tag occurs

is to disagree with what A has said about everyone achieving

competence. We can see that A (who is the tutor) takes an assertive

role in changing the direction of the conversation (by her question on

a new topic, and in asking a direct question to the tutorial group

(and, as usual with teachers, one to which she already knows several

answers; cf C6ulthard, 1977: 104 for this observation)). Given A's

assertive behaviour, E's vagueness may be a marker of deference to

someone established as superior in the context of this discussion, (cf

Schenkein, 1978, 'Identity Negotiations in Conversation'). E is

clearly quite determined to get her point out. She resists all

interruptions, including A's very firm attempt to cut her off

("there's been a lot of work done... ") and other interruptions. Yet

she must continue to mark deference to A, even while disagreeing with

her.

A general aspect of the two extracts is that they show speakers

under some kind of stress, arising from, perhaps, uncertainty of the

subject, lack of knowledge, and the unequal relationship of the tutees

to the tutor.

5.3.3 Humorous Effects

The 'punchline' of cartoon strip no. 1 is provided by a vague tag.

By analysing the source of the humorous effect, it is possible to

deduce how the vague tag itself works. First a 'gloss' of the story:

Boopsie has been to Graceland and loaded herself up with what looks

like a pile of varied Elvis memorabilia. BD asks her if she has

bought any souvenirs. We see that she has, in fact she has bought a

ludicrous quantity of them. "A postcard or something" would be a bad

description of what she has bought. Hence the humour. The joke

provides independent evidence for the assignment of meaning to

145
 '!

ýý
 Chapter 5

exemplar + tag constructions given in Chapter 4- that they designate

a category of which the exemplar is understood as a good example. If

that wasn't the meaning, the joke would not work. Boopsie uses just

"or something" as her (agreeing) reply. That works, I think, to tell

us that although the collection of Elvis tat that she has collected

could be described by "a postcard or something", what she has got is

anything büt a postcard, and indeed has rather different

characteristics from a postcard.

Conversationally we may understand the punchline as a form of

witholding - Boopsie doesn't want to confess to BD exactly what she

has got (cf his reaction in the next strip, no. 2).

What makes no. 3 funny? Probably that Riley shows by his use of

vagueness that he doesn't know what he's talking about in the field of

medical care (neither does Duke, as fans of the strip will know, but

that isn't evidenced here). "Boil some water" is additionally funny,

because it occurs classically in 'anxious father at emergency

delivery' jokes. The associational category identified by "boil some

water or something" would include, on my reading, other items of folk

medical knowledge. Notice that this "or something" could also be

deference behaviour, followed as it is by "sir".

One thing a cartoonist is anxious to do is to create a

naturalistic conversation effect. Two tags on one page suggests that

he judges them as salient as markers of informal conversation, in the

same way as the advertising copywriter did.

In these three extracts, a diversity of conversational effects

has been observed. I'd like to now firm up these observations by

looking at further examples which will serve to substantiate the

146
 Chapter 5

categories suggested, and also to add some others.

5.4.1 Giving the right amount of information

We saw in the Vichy example that one possible use of vague additives

is to tailor an utterance such that the right amount of information is

given.

Grice (1975) noticed that speakers appear to tailor their

contributions in particular ways, and he suggested that two important

rules of conversation are what he termed the two Maxims of Quantity:

1 Make your contribution as informative as is required (for the

current purposes of the exchange)

2 Do not, make your- contribution more informative than is

required, (: 45)

The examples in this section show these rules being used, and show that

vague additives are a device which speakers use to tailor their

contributions such that they give the right amount of information for

the purpose of the conversation.

Grice notes that his second maxim is disputable - "it might be

said that to be overinformative is not a transgression of the CP but

merely a waste of time". I think the evidence from data is that the

second maxim of quantity is indeed a rule of conversation, the

transgressing of which produces, as predicted by the Grice account,

particular implicatures. We, saw in the previous, chapter the text

cited by Rosch, containing many overspecifications and producing a

humorous or ironic effect. We saw also that my made-up examples:

(28) It was a freestone peach or something

(29) It was a metallic finish 1975 Renault 5TL or something

would probably also produce bizarre effects in many circumstances.

147
 Chapter 5

Some informants recognise this consciously. When I asked the

informants in the number approximation test (reported in Chapter 3)

about the use of approximations, one of them said:

"I think it's more used because it doesn't convey more

information. than if you knew the exact figure - if you say to
someone 'it cost five hundred or so pounds' - if. they know it
cost five hundred and thirty nine it's not going to get them
any further. "

Clearly, though, we must add 'for the purpose of that conversation',

because 4539
there are obviously occasions when knowing it was exactly

would be necessary, for example in a banking or accounting context.

Sadock and Wachtel in their writing on number approximations both

notice that approximations can be used to vary the amount of

information given. Crystal and Davy (1975: 111-114) observe that

vagueness is quite appropriate in some conversations. I asked one of

the test subjects why a speaker would not give the exact amount, even

if he knew it; he said "because in casual conversation like that you

don't go into details like that". We have seen this as one possible

effect involved in the advertising copy "eighty or so pence".

A less complex example is

[geraniums]
Theres a room downstairs you see which is only one floor and
gets really cold and I. lost two or three with the
frost [Camb 13A180/36]

In this example so few geraniums are in question that a hearer may

,

well conclude that S must know how many he/she-lost. Notice that

being told the exact number will not contribute anything of useful

interest to H: S could equally have said some or a few. S has avoided

being precise, in view of the purpose of the conversation, and its

possible informal setting.

[paper reporting informant work at LAGB]

Weve got about five or six of them but I'm only going to talk
about three of them today [LAGB 9,81]

148
 Chapter 5

Clearly a linguist engaged in phonetics research must know how many

informants he has. But it is not relevant to the audience to know

this for the purpose of the paper he is about to give. The

information which is important to them is that they are going to be

told about three informants. Tailoring the amount of information by

using an approximation in direct contrast with an exact number

("three") may have the effect of focussing attention towards, or

foregrounding, that is considered most important in the utterance.

Using an approximation here communicates something like "don't pay too

much attention to this, it's not very important".

In order to observe the converse, use of precision in contexts

which demand it, I looked at data from three BBC Radio 4 programmes on

financial topics: Money Box, The Financial World Tonight, and It's a

Bargain. There were overall very few examples of any kind of

vagueness. There was a noticeable absence of vague tags - one

occurrence of and the like in an interview. It might be thought that

this was because vagueness is unacceptable altogether on Radio 4.

However, this is not the case, since vague expressions did in fact

occur in a particular context: that of making predictions about the

future, and I shall discuss some examples in the section below on

"Talking about things you aren't sure about ".

Two extracts will show the salience of precision for the purposes

of these sorts of programmes. The first is from The Financial World

Tonight, stock market report:

and on the immediate issue Dalgety gained six to two eighty

after those figures (. ) otherwise "firm best describes the
way the market traded with the FT index up three point eight
at four hundred and eighty nine point three - gains among
leading shares ranged from two to four pence (. ) like Beecham
up three at one hundred and eighty - blue circle up four at
three hundred and eighty six and Grand Met up four at one
hundred and sixty six

149
 Chapter 5

Here the changes in share prices must be provided exactly for those

who listen to them.

The second extract is from It's a Bargain, a programme with a

very different purpose from that of The Financial World Tonight, since

it is to tell listeners the usual prices of consumer goods at the

bottom end of the market, and perhaps thus also with, socially, a very

different target audience.

The BBC guide price works out at one hundred and ninety nine
pounds (. ) cheaper than guide price we discovered two models
at a hundred and seventy five pounds [... ] cheaper still we
found a cooker with a small oven and push button ignition to
the hot plate only (. ) and this was the "Valor Corvette (. )
it's on special offer from North Thames Gas at a hundred' and
sixty two pounds and seven pence - but from the Comet
Discount Shops it's a hundred and fifty four pounds ninety
"excluding their delivery charge which is two pounds seventy
five

In this context precise prices are given.

In these two cases, it certainly does "get you further" to be

told the exact numbers. These comparisons show (a) that the amount of

information given is dictated by the, perceived purposes of the

interaction and, (b) that vague additives are used where less

precision is judged to be required.

5.4.2 Withholding

Many (non-linguist) English speakers who have talked to me about vague

language use have emphasised the extent to which vagueness is deployed

deliberately. They suggest that they often use it to withhold

information which in some sense might be'expected by their hearers in

a given situation. We saw an example which I suggested might show

this, where Boopsie in the Doonesbury cartoon gives only an inexplicit

indication of what she has bought.

150
 Chapter 5

Another example which might be heard as withholding is the time

estimate example from Chapter 3.

B: could you give me some idea of how long it would take

C: well the quote might be done'within three or four days but

the job won't be done for at least five weeks [Carob
11B1058]

This three or four days can be seen as unhelpful and insufficiently

informative. On the other hand, it could be that the speaker really

does not want to commit himself on when the estimate will be done, so

this would be a defensive tactic, like those described in section

5.4.5.

An example reported to me was the following:

She's about 29 years old

This appears to be withholding of the exact age, since when questioned

the speaker claimed she had used this form becaus'q en don't like to

be said to be 30.

5.4.3 Saying what you don't know how to say

One conversational goal which a speaker uses vagueness to achieve is

to get across a meaning where he does not have at his disposal the
.
necessary words or expressions which he needs to associate with the

concepts he is forming. Notice that to be convinced by this account,

you have to accept that there is a level of cognitive

activity/representation which precedes words and is independent of

them (for arguments for this view, see eg, Fodor, 1976 The Language of

Thought). I would think that the data in this chapter offer

substantial evidence that people can and often do think about concepts

which they can't really talk about, ie that there is a

pre-lexical/pre-language cognitive level.

151
 Chapter 5

A clear example of a speaker with a word-finding problem is the

following, taken from my transcription of the students in Extracts 1

and 2, and myself) in a tutorial earlier in the same term:

B: that it is very hard isreally interesting where you've

got virtually every word you've got somebody stopping
somewhere

E: cos it doesn't flow /B: yeah/ (. ) its difficult to

sort of say it in sounds /A: uh hu/ its sort of
difficult to sort of

A: on the other hand its still very few people

[1,13.2]

It is plausible (at least) to suggest that the word the speaker would

have been happy to use was articulate. She either did not know this

word, or under the strain of trying to express ideas about unfamiliar

topics in a tutorial, she had forgotten it. Such examples arise both

where the speaker does not know the necessary word, and where he has

forgotten it, since in both cases, for the purpose of the utterance in

hand, the speaker lacks knowledge of the word.

Another example, later in the same tutorial is:

A:. if
we're trying to find out how they take it in
. .
which is what as you say this thing was trying to do then I
think that shows that the syntactic element is important in
structuring - in helping you to decode what you've got more
important than the semantic - well the thing is they're
interlinked - you can't separate them out

C: but there's also things like when you're talking and you
take information in when you - when you're talking its just
words that you pick up - its not whole strings of sentences
and verbs and things - its just the sort of main meaning

E: yeah right

C: because people don't, talk in sentences

[
E: if somebody
tells you a. full story then you don't remember every word
they said but you do remember the general gist of it /mm/

Here, the expression which the participants can't use, but need, I

152
 Chapter 5

think, is syntactic structure. A had introduced this earlier, and had

used "syntactic element" in her turn preceding Cs, but it is clear

that at this stage neither of the tutees has mastered the use of the

word syntax.

A related but different situation in which use of vagueness

arises, is where the language in question does not have the lexical

items necessary for precise expression. An example from later in the

above tutorial discussion is:

C: but lots
of sort big
of important numbers that you have to
ring and things I'm sure they're made into a kind of pattern
that you can remember them by - say your friend's telephone
number because they give you something like two three five
seven eight or something like that which is much easier to
remember than - and -I know - I'm sure they do it into
patterns so that you can remember them

Here, firstly, "ring and things" does not have any obvious lexical

superordinate. C succeeds in referring to her set of 'ringing

activities', by using a vague expression. Her subsequent use is the

same: "something like two three five seven eight or something like

that". There isn't any readily available superordinate to refer to

the category. She clarifies in her final point by using patterns, and

this is taken up by subsequent speakers who refer to: "patterns",

"next number", "double four double two double four which is a sort of

pattern".

Again, some test subjects were aware of this. In discussion

after the test one said that vague tags were used "if you're thinking

of something and you don't know the words for it". Others said:

You might not have had the time to think of the exact thing
you wanted to say (. ) if you say something like it or
like that then people /mm. mm/
say or something

B: and sometimes you don't know (. ) like that Bernstein one

I mean (. ) I don't know (. ) I would have said something that
I knew about him and then

153
 Chapter 5

C: get yourself out of it

B: get myself out of it by saying...

Finally, they suggested that such vague tags would not be used in

writing because,

you've got more time to think, so you can pick one that (. ) a
word that really does represent the things you want to say

This supports my observation that vagueness is a ploy speakers use

when they can't find the words they need.

The two types, word-finding difficulty, and lexical lack in the

language, are seen here as different. However, they are, from the

point of view of the speaker producing an utterance, 'the same, in the

sense that either way, he does not have the necessary word(s) at his

disposal. More importantly, from the point of view of the hearer's

understanding, they are probably the same. since he must go through

the inferential procedure of identifying an appropriate associational

category from the exemplar given.

However, at the level of conversational effect, they may be

different. In the situation of unequal knowledge of the topic of

discussion, for example, that which is usually found in a tutorial,

the tutor takes the tutees' use of vagueness partly as a confirmation

that they don't know as much about the topic or know its vocabulary as f'ý-
t ý-
ýi.
well as she does. Hence her frequent use of certain pertinent

vocabulary items ("syntax", "syntactic structure", "semantic"),

I

perhaps to try to get their meaning over. : ýý

.ßý

5.4.4 Talking about things you aren't sure about

As I mentioned in relation to Extract 1, an observable ploy for a

speaker who lacks specific knowledge is to use vagueness. A clear

154
 Chapter 5

example of this is:

[percentage of university students coming from working class

backgrounds]

B: I can't remember what the figures are but its something

around the twenty per cent mark and it's never
changed (11,21.2)

where B's admission of memory failure ("I can't remember") provides

clear evidence that his vagueness is obligatory - he simply does not

know the exact figure.

Examples like these illustrate the working of what Grice (1975)

formulated as the Maxim of Quality - part of which is the rule:

"2 Do not say that for which you lack sufficient evidence" (: 46)

In another similar example, a speaker giving a talk referred to a

visit to a country abroad as having taken place "about ten or so years

ago". When questioned afterwards, he said it was "about 1969, t would

have to look it up", thus confirming that he actually did not know

when it had occurred.

In a written report on student accommodation in York, the

following appeared:

I have assembled a list of student addresses in York from the

registration cards in the Undergraduate Office. The
Students' Union did not collect housing information in the
usual way this year (NUS card returns) and although the
registration questionnaire was designed to fill the gap, the
poor response rather nullified the attempt. Around 600
students found private rented accommodation in York and
district this October -a loss of around 45 units from last
year. In reality this loss was a little less. There are
always some addresses which escape listing at the first
attempt. "

The writer makes explicit that his data collection was imperfect, and

thus uses vague quantities so as not to make exact statements which he

does not have evidence for. There is about this an element of

self-protection - of wanting to avoid later being shown to have said

155
 Chapter 5

or written something which isn't true. We shall look at other

examples of this in 5.4.5.

Displacement

Speakers use vague expressions to express their degree of certainty.

Two situations in which speakers are often uncertain is where they are

talking about the past, or the future. In the case of the past,

complete evidence may be lacking. An example which shows this is

,
taken from The Guardian:

[article in The Guardian (29.8.81) about the death penalty in

the 18th century]

The records, though far from complete, showed that about 61

people died on the block

This is curious. My reading would be that the records stated that 61

-
people died. But the writer knows (and says) that they are

incomplete, so he uses an approximator to take account of the

unrecorded deaths. He also, I think, wants to 'cover' himself against

the possibility of being wrong.

Vagueness is often used when making predictions about the future.

This can be seen in the pre-budget edition of the programme Money Box

in which possible changes in the tax system were discussed:

[The Stock Exchange propose abolishing Contract Stamp Duty]

that would cost the Chancellor about two million pounds a

year or as they put it - seventeen and a half minutes of
government expenditure

[it also proposes altering the 2% Transfer Stamp Duty; Deputy

Chairman of the Stock Exchange: ]

rather than have a duty which will become more and more
evaded or avoided in the future because of the ways round it
it is better to make it a bearable amount and therefore we're
suggesting one per cent for everybody

Interviewer: And Peter Wills reckons that change will cost

around a hundred and fifty million pounds

Notice the necessity for the exact figure "one per cent" in contrast

156
 Chapter 5

to the approximation for the uncertain future consequences of the

reduction.

A final example from the same programme:

[building societies propose abolishing stamp duty on house

purchase]

Interviewer: Any idea how much that would cost

Spokesman: the cost in this year would be about two hundred

and twenty million pounds,

Again, test subjects provided independent evidence that lack of

knowledge is a reason for using vagueness. On number approximations,

one said:

Usually if you have a figure like that it's because you've

heard it somewhere or seen it somewhere and it's a very
precise figure to start off with but you're just repeating it
without the exactitude cos you can't remember all of it.

Discussing the tag test, the university student subjects told me:

its when you're trying to express yourself (. ) you're really

thinking about what you're trying to say (. ) you say or
something like that

you know (. ) I think its when you (. ) sometimes you don't

know really what you're talking about /you use them/ but if
you know what you're talking about, you know exactly what you
want to say and you won't use them

This is borne out by the low frequency of vague additives in the

three financial programmes recorded. People taking part in such

programes do so because they do know what they are talking about, and

therefore they don't use vague expressions.

5.4.5 Self-Protection

I suggested in relation to the Vichy advertisement that vagueness may

be used as a safeguard against being later shown to be wrong. This

was seen also in the report on student accommodation quoted in 5.4.4.

157
 Chapter 5

Examples of this strike a very odd chord when it is fairly clear

that the speaker knows the information exactly and it is appropriate

for the purpose of the conversation to give it exactly. Here are some

of these.

[House-hunting telephone call to Estate Agent]

B: How many houses are there in the street?

C: There are [reads from printed details] approximately four

houses in the street

In this case the estate agent will have made a survey of the area and

will know how many houses there are, ie four. This approximator

carries a message something like " we've counted four houses, but if

you go along and see an extra one, or think that number 24,

technically in the next street, is really in this one, then we are not

wrong, because we said 'approximately"'.

[BBC Radio 4 news: Police spokesman making statement about

hijackers at Stanstead]

I can tell you that approximately eleven people are helping

us with our enquiries

The non-round number suggests strongly that this is not really, an

approximation (cf discussion in Chapter 3). The speaker knows that

eleven people are involved. His official position inclines him,

however, to extreme caution, hence the approximator.

The third I am including here ý- .

example probably shows a number of
ý
factors at work.
[Chemist interviewed on You and Yours, BBC Radio 4,19.2.81,
i
about not charging the NHS prescription charge for an item if
the retail price is lower]

During that day in fact we had three prescriptions for throat

lozenges which I think the retail price was about 41 pence
and a tube of cream for arthritis which was about 58 pence
and on all three occasions I said to the patient concerned -
there's no point in your paying a pound -I will charge you
the correct retail price.

158
 Chapter 5

I believe this speaker remembered these retail prices exactly. But he

wanted to guard against the possibility of a faulty memory. In

addition, he was probably affected by the stress of being interviewed

for the radio which made him, perhaps, more uncertain, and this may

also have led to his using vague expressions.

My last example of self-protection is from the university

tutorial group:

C: One of the secretaries was saying there was a film or

something.

A: oh yeah

D: really

C: last thursday we were all down in Z017 and the lecture

was actually in Vanbrugh so I went down to the secretary and
said where is everyone and she got out this book and looked
at it and said its in Vanbrugh but some weeks there's a film
being shown and I said well I do psychology and linguistics
is just an elective - but I was intrigued by the film - can
you have films in linguistics

A: oh yes we have lots of films

C: "really [disbelief]

The speaker knows quite well that the secretary said there was a film.

He makes it definite in the preface to his question. Why tag the

first occurrence of film? He is seeking information, he is not sure,

and he doesn't want to be wrong. Also, however, account must be taken

of the unequal tutor/tutee relationship. There may be some deference

here. I discuss this in the next section.

5.4.6 Deference

In discussing Extract 1 in section 5.3.2, I suggested that one reason

for use of vagueness was the speaker expressing deference to the tutor

at the same time as disagreeing with her. Vague additives are used

for these kinds of social reasons.

159
 Weiser (1974) noticed that utterances may be constructed so as to

be deliberately ambiguous between at least two speech acts, so as to

leave a hearer the option of taking up one speech act or another. One

of her examples is:

I'm curious to know what went on at the hearing

which can legitimately be treated by its hearer, she says, as a

request for information, or as a statement. As such it leaves the

speaker a bolt-hole: "I wasn't demanding that you divulge confidential

information, I was just expressing curiosity". I think vague

additives are used for the same sort of politeness reasons, for

example, in this offer, couched as a question:

[A has given up alcohol, speaker knows this]

B: Would you like a drink - an orange juice or

something [15.9.801

The speaker is understood as presupposing that A will want a

non-alcoholic drink - he deliberately cancels the preferred reading of

"would you like a drink? " with his exemplar + tag. This is understood

as referring to the category of non-alcoholic social drinks (orange

juice being a good example) and he politely offers the addressee

options within that category.

Another example, this time of a request, is taken from one of the

tutorial discussions:

Could we, when you give us our essays back - and give us
titles - could we sort of meet or something - because (. ) I
mean - there might be things we want to ask

We have already seen a possible example of deference between tutee and

tutor. We have also seen the tutees' tendency to use a lot of

uncertainty markers. Here the tutee makes a direct request to the

tutor, but is heard as mitigating its directness by leaving the tutor

other options (cf the test informants' interpretations, Table 4.6,

160
 Chapter 5

Item 26).

An example I gave in Chapter 4 is another example of deference.

It is a recapitulation of a direct request for action (unfortunately

not recorded; it would have been useful to have the first formulation

of the request; I suspect it was without vagueness).

[B has asked A to buy him some bread]

A: so you'd like some bread

B: or something - anything edible will do

B mitigates the force of the request by leaving A much wider options

than his original request.

In the next example, the speaker says something which is open to

interpretation as being rather critical, so he mitigates it by using a

vague tag : i

B: I'm a third year physicist

C: I don't know anything about physics at all

B: well neither do I

C: I never even did them at school

B: not at all -ever

C: well I

B: general science

C: I did them for a term - so I know something about them - 1ý

I did chemistry - did biology

B: mmm do you feel that this is a vast hole in your

education or anything

C: no

[Clark, 1981]

These few examples show how vague additives are used to make

conversational turns convey politeness in appropriate ways.

161
 Chapter 5

5.4.7 Informality and atmosphere

I
I noted with reference to the Vichy advertisement that vague- language

is associated with informal conversational settings. Some of the test

subjects told me the same thing. One said:

If you're just having a conversation - with one's friends

you'd probably use them - but if you're in the classroom -
you wouldn't

and another suggested:

you change according to who you're talking to

The following is an extract from a conversation between two close

friends (both men):

[buying Christmas trees]

B: oh uh I was up at Blackheath in the morning to buy a

Christmas tree

C: you/did you actually buy one

B: yes - got one

C: well has it got roots

B: - no it hasn't - no
C: how much was it

B: erm quid fifty - about five foot tall

C: that's not bad - we went down to Henry's and had a look

at them but er ummm not particularly impressive - none of
them had roots
B: --------------------more at this place I went to at
Blackheath was very good

C: whereabouts is that

B: er just opposite the station

C: oh

B: just across the road

C: a cross the road

B: across the road from the station - yeah

C: it's it's normally a plant shop is it

B: well call it green grocery and things yeah

162
 Chapter 5

C: oh yes yeah

B: there there's another shop in Blackheath village that was

selling really scrawny ones five or six quid a time

yeah well Blackheath I can believe it [Camb 9B14/18-33]

Incomplete utterances, yeahs and ums signal this as informal. Three

vague expressions occur in rapid succession. Clearly the presence of

these vague expressions is dictated also by the Conversational Maxims:

Quantity - it is not necessary to know exactly how tall the tree was,

or what the shop sold, and quality - the speakers probably did not

measure the height of the trees, perhaps do not remember the price of

the trees and do not know for sure what kind of shop it was.

Obviously informality and giving the right amount of information are

closely related.

My second example shows calculated informality, similar to that

noted for the Vichy advertisement:

[children's book title]

All about Cuckoos and Robins and Things

The contents of the book make clear that what is intended is not the

whole category BIRDS, but a subset: those that live in gardens and

parklands. We know from Rosch's experimental work on this category

(Rosch, 1973) that the robin is considered to be prototypical of the

category BIRD. Hence the title is not very informative in identifying

the intended category. Given, however, that it is aimed at children,

it looks like an attempt to communicate in their register which is

informal in a special way.

5.4.8 Women's language

The test subjects thought that women use'more vague expressions' than

men do. I did not control for this variable in looking for examples,

but I have no evidence to suggest that it is the case.

163
 Chapter 5

The subjective impression that women use more vagueness could

arise from several factors. Firstly, as we have, seen, vagueness is

associated with deference. If, in turn, women are seen to be or are

required to be, deferent more often than men, then women will be

expected to use vagueness. That is to say, if it is true that women

use more vagueness, they do so not because they are women, but because

they are deferent or uncertain.

Secondly, there is a view that women have their own language

varieties for speaking about matters of concern to women. Angela

Carter (1980) suggests that these varieties incorporate a "high degree

of unspecificity". She writes:

"Any daughter sent out by her mother for a 'piece of cheese'

will return with roughly the same amount (about half a
pound). In fact, we all know what we mean by virtue of
praxis, pure and simple; any fool ought to be able to work
out how much butter, how many eggs, given a knowledge of the
number of mouths there are to feed. (Men always insist on
shopping lists itemized down to the last microgram and
milliliter. )"

Language users recognise that women engage in this sort of time-saving

brevity, hence the test subjects' judgment. Of course it is quite

erroneous to suggest that only women do it. Any social group sharing

interests and knowledge, employs non-specificity in talking about

their shared interest. For example, people who repair their own cars

might say the following:

Could you get me some sparking plugs?

Non-specialists asks 'how many? '. Specialist knows Renault has four,

some Citroens only 2, and you always change them together.

I'll get some spares for our holiday - belts and things

(= small easily changeable engine parts which might need replacing on

the road (eg points, plugs, condenser)). Non- specialist has no idea

what they are and so cannot identify the category referred to.

164
 Chapter 5

5.5 Conclusions

In this Chapter, I have looked at a wide range of conversational uses

of vague additives (however, I have certainly not looked at all

possible uses). These are varied both in the types of discourses from

which they come, and in the effects which could be observed. The

widespread use of vagueness for varied purposes demonstrates what an

important aspect it is of the language user's knowledge of his

language.

I have said that the discourse settings, and the effects

At time, is
observed, are varied. the same there an important

unifying feature, which is that in every case an element of

uncertainty is introduced for at least some participant in the

conversation. Where vague language is used to tailor the amount of

information given, hearers are uncertain by virtue of being in receipt

of less than the full facts. If the hearer of the Christmas tree

height were asked how tall the tree was, which his friend had bought,

he would have to say that he did not know exactly. Of course, hearers

are often not aware of the lack of precise information, because, as I

have argued, it is sufficient for the purposes of the particular

conversation-in which they are taking part. If a speaker is unsure of

his subject, or cannot find the right words, his uncertainty is

present, and is communicated to his hearers. Hearers in turn are

presented not with something precise, but with a series of options.

In the case that the speaker does not know, or does not use, the right

word or expression to name a category, but replaces it with a vague

Category Identifier, his hearer must be uncertain to some degree of

the extension of the category the speaker intends. The same applies

to the use of an approximation for a quantity. In self- protection

uses, the speaker is perceived as uncertain of his authority, and this

165
 Chapter 5

is apparent to hearers, who pick up on such cues as presence of a

non-Round Number (if it is an approximation). In the deference and

politeness examples, the hearer is provided with the opportunity to

act upon the speaker's utterance in different ways, and so has

uncertainty passed on to him, at least until he has made, a choice.

166
 Chapter 6

Chapter 6

Lexical Aspects

6.1 Introduction

In the discussion of vague additives thus far have treated the

,I

expressions under consideration holistically. However, it is the case

that all the expressions are made up of two or more lexical items,

each of which has uses other than those discussed here. So there are

two tasks to be undertaken in this short chapter, firstly to consider

the extent to which the meanings which can be established for other

uses of these words contribute to the meaning of the vague expressions

in which they appear. Secondly, on the basis of this, to discuss,

,
somewhat speculatively, whether these vague additives should be

described as idioms or partial idioms in an eventual lexicon of

English.

6.2 Lexical Descriptions

6.2.1 Or

Or occurs in both the number approximations (n or m, n or so) and in

the tag set or something (like that).

One difficulty for me is that there is by no means any agreement

as to the analysis of what are understood to be "normal" uses of or

(ie, leaving aside those above). There are three areas of doubt; (a)

is there one English or, or two (exclusive and inclusive), or more;

(b) to what extent can English or be equated with the logical

connectives for inclusive or exclusive disjunction; and (c) what are

the conditions on acceptable use of or, and how should they be stated?

Some of the literature on (a) is in Hurford (1974) and for an opposite

view, Gazdar (1979), who summarizes much other work. Fillenbaum

167
 Chapter 6

(1978) and Wexler (1978) tackle this question for both or and and from

the point of view of psychological theory. (b) is of course closely

connected with (a) because if there is only one or, then it is

potentially more like the logical operator. Gazdar (1979) tackles

this issue. Fillenbaum (1978) and Wason and Johnson-Laird (1972)

report psycholingustic work, designed to test whether people employ

the natural language connectives similarly to their putative logical

equivalents. (c) is discussed by, among others, Lakoff (1971), Quirk

et al (1972) and Wason and Johnson-Laird (1972). It is not the place

of this thesis to become involved in these debates. I shall,

therefore, extract from the references cited, what appear to be the

generally agreed observations pertaining to or, and then consider

separately how much the special uses of or which I am looking at are

like them.

Firstly, there is the 'common topic' condition, noted by Lakoff

(1971: 148) and Wason and Johnson-Laird (1972: 92). This is to say that

part utterances connected by or must be perceived as having something

to do with each other. Thus:

(1) Could you get or some oranges at Sainsbury's[l]

me some apples

is fine but

(2) Could you get me some apples or a washing up brush

is odd. Some informants who I questioned informally about (2) said

that they felt impelled to try to find a common feature shared by

apples and washing-up brushes, say price, to get an appropriate

reading. Fillenbaum (1974,1978) demonstrated experimentally that

speakers are sensitive to the 'common topic' condition. In his

1 In this chapter I revert to the earlier convention of marking

attested examples with '+'.

168
 Chapter 6

experiments subjects were asked, among other tasks, to make

appropriateness judgments on sentences containing violations of the

'common topic' condition. Where these were simple questions, for

example, "96 per cent of the judgments characterised them as strange

or weird" (1974: 916).

Secondly, it has been observed that certain constructions

containing or, which are disjunctive, can be thought to have arisen

from ellipsis (Quirk et al: 9.34 and 9.61ff), or what is known as

conjunction reduction (eg Harries 1973). Thus, for example:

(3) We can take John's car or Bill's

can be described as being derived from

(3a) [we can take John's car] OR [we can take Bill's car]
SS

Thirdly, as Quirk et al (and others) note, English disjunctive or

is almost always understood as exclusive, to the extent that, it is

quite hard to construct examples which force an inclusive reading,

unless one adds or both. Their example is:

(4) You can boil yourself an egg or you can make some cheese

sandwiches, or you can do both

Hurford's example of

(5) Inmates may smoke or drink

which we understand as inclusive, tends to show that inferences from

pragmatic knowledge are what decide whether instances are understood

as inclusive or exclusive. Hurford's conclusion, however, is that

"some instances of English or are inclusive, while others are

exclusive" (: 411). Gazdar deals with this by making all instances

truth-conditionally inclusive, but with a generalised conversational

im-plicature of exclusivity, which can be expressly cancelled. The

precise account which should be given of disjunctive or is not

169
 Chapter 6

important here.

A fourth point about disjunctive or is that noted by Hurford,

that

"the joining of two sentences by or is unacceptable if one

sentence entails the other; otherwise the use of or is
acceptable. "

Thus:

(6) *John is an American or a Californian

Gazdar argues that Hurford's analysis is incorrect, depending as it

does on a sentence, (5), which contains a modal. He proposes that the

correct constraint is that sentences entailing one another may be

conjoined if the entailed sentence potentially implicates the negation

of the entailing sentence, otherwise they may not. This will still

account for (6).

6.2.2 And.

The treatment of the connective and is in many ways similar to that of

or. Its equivalence or not to the logical connective has been

discussed (among others by Cohen (1971), Schmerling (1978), Gazdar,

1979). The question of whether there are two ands, one symmetrical

and one asymmetrical, is discussed by Lakoff (1971). Lakoff, and

Fillenbaum (1978), again discuss its various uses and their

constraints, and again sentences containing constituents joined by and

have been analysed as having undergone reduction or ellipsis (Harries,

1973).

The 'common topic' condition applies to and in the same way as to

or and as Lakoff demonstrated, the perceived similarity may be at the

level of presupposition or inference.

170
 Chapter 6

What is described by Lakoff as asymmetric and is the one where

the conjuncts are ordered with respect to time, eg:

(7) *The Lone Ranger rode, off into the sunset and mounted his horse

But Kempson (1975: 56) shows that even if these were two separate

sentences, the ordering conditions would still apply, so this has

nothing to do with the meaning of and as such. Lakoff concludes that

such uses are special cases of symmetrical and.

For the reduction/ellipsis account, there is the problem that

some well-known instances of and cannot be derived by ellipsis, eg

(8) John and Mark collided in the corridor

(9) John and Mary don't like each other

I notice that a contrast between and in a construction arrived at by

ellipsis, and in a construction not deriving from ellipsis, is the

possibility of the reduced version, or not. In an' ellipted

construction, and can be full and stressed, or reduced:

(10) John (/and/ ) Mary came into the room

(/ and/ )
(/an/ )

(/n/ )
where the bottom three have a meaning contrast with the first. Ands

not in a putative ellipted construction cannot be full, thus:

(11) *John /and/ Mark collided in the corridor

6.2.3 So

So has, a number of uses, some of which can be excluded from

consideration on distributional grounds. Thus so as an emphasiser

(Quirk et al: 5.51) can be left out, as can so as a conjunct (eg "I

didn't feel well so I didn't go"). These uses cannot occur

clause-finally. The uses of so which are of interest are those which

171
 Chapter 6

Quirk et al call PROFORMS, very commonly in a construction with the

pro-verb do, thus

(12) She hoped that he would search the room carefully before her

arrival but he didn't do so (2.17)

for)
c part so stands

They note that it can be a proform for a clause (10.62), for an object

(14.16), or for a predication (9.80).

Another use which may be related to the number approximation one

is that with and: and so on, and and so forth which the Longman

Dictionary of Contemporary English glosses as "and other things of

this kind, and continuing".

6.2.4 Something

As seen in Chapter 4, something alternates with anything, in tags, as

between assertive and non-assertive contexts. In this way it

parallels the usual uses of something described by Quirk et al

(4.127). Something acts as a proform NP for (the Longman Dictionary

of Contemporary English) "some unstated or unknown thing", as opposed

to person ("someone") or place ("somewhere").

Mittwoch (1981) notes that "for the purposes of concord,

something functions as a singular NP" (: 119). She notes that there

are two somes: somel being that found in "some bread", and the other,

always stressed, in "some book or other". Mittwoch claims that the

some in something is some2, in which case it has the features

[+delimited quantity] [+specified quantity] [+count] "[+singular]. At

the same time, something can be used to report a plural, for example

when "I ate something" is used to report that I ate three sandwiches.

6.2.5 Like

Prepositional uses of like are said by the Oxford Advanced Learners'

172
 Chapter 6

Dictionary to be "often governing a pronoun, noun or gerund". Given

its position, the tag use of like must be a preposition, and it is

followed by that which, I suggest below, is here functioning as as

pronoun. This like is glossed as "such as, resembling". Weinreich

(1963: 130) calls like "the most powerful extrapolator of all" in the

class of operators which "function as instructions for the loose or

strict interpretation of designata".

6.2.6 Things

Thing is a useful portmanteau noun which may be used to refer to any

material object, and to non-material concepts as well. It also occurs

in some idiomatic expressions such as quite the thing and do one's own

thing. The Oxford Advanced Learners' Dictionary notes a sense which

occurs only in the plural and is glossed as "belongings, articles of

which the nature is clear (or thought to be clear) from the context",

one of their examples being, "Bring your swimming things".

6.2.7 That

That occurs in or something like that, and things like that, and and

that. As far as I have observed, it never occurs in reduced form,

which, together with its position, suggests it is the pronoun use of

that which is of interest here. That is the singular demonstrative

pronoun "used to make a thing specific". Many uses are deictic (as

defined in Lyons, 1977: 637), in that they point the location of a

thing being referred to as "non-proximal" to the speaker. Another

function is in textual deixis (Lyons: 667), where it is used to refer

back to linguistic material previous in the discourse.

6.3 Idiomaticity

"There is no generally-accepted criterion which would enable

us to draw a sharp distinction between phrasal lexemes, on
the one hand, and cliches or fixed collocations on the other"
(Lyons, 1977: 146)

173
 Chapter 6

As the above extract suggests, there is little agreement about how to

treat multi-word phrases which in use seem to have some integrity as

units. Weinreich (1969) summarizes varying approaches in American

work on idioms up to that date, which he describes as having "a

distinct hit and run quality". Becker (1975) talks similarly of

idiom-like structures being "swept under the rug". In addition, as

noted by Cowie (1981), there is a lack of generally agreed terminology

for the various categories of idiomatic expression.

is has a literal
For Weinreich, an idiom an expression which

counterpart, and thus involves ambiguity. He would have an idiom

list, and process idioms by a comparison process with their literal

counterparts. Expressions not having literal counterparts (because

for Weinreich they are "ill-formed") are classed as "complex

dictionary entries", for example by and large. Searle (1969) proposed

the same kind of schema.

Fraser (1970) defines an idiom as

"a constituent or series of constituents for which the

semantic interpretation is not a compositional function of
the formatives of which it is composed" (: 22)

He classified idioms into different levels of "frozenness".

Becker, like Lyons, suggests that the divisions are not so clear

cut. He distinguishes seven classes of expressions from "Polywords",

which are invariable, such as forever, to "Verbatim Texts", whole

texts which we memorize, eg "How ya gonna keep 'em down on the farm".

Becker's view of language production is that it is more or less a

haphazard exercise:

"We start with the information we wish to convey and the

attitudes toward that information that we wish to express or
evoke, and we haul out of our phrasal lexicon some, patterns
that can provide the major elements of this expression. Then
the problem is to stitch these phrases together into
something roughly grammatical, to fill in the-blanks with the

174
 Chapter b

particulars of the case at hand, to modify the phrases if

need be, and if all else fails to generate phrases from
scratch to smooth over the transitions or fill in any
remaining conceptual holes. " (: 72)

Bolinger (1976) takes a similar view, saying that although language is

a system ou tout-se tient, it is heterogeneous as well as homogeneous.

Arguing from examples of collocations, lexemes like else and ago with

an idiosyncratically limited distribution, and idiomatic uses such g3

those of the verb bear, Bolinger's metaphor for language is of "a

jerry-built structure":

"our language does not expect us to build everything starting

with lumber, nails, and blueprint, but provides us with an
incredibly large number of prefabs, which have the magical
property of persisting even when we knock some of them apart
and put them together in unpredictable ways. " (: 1)

He argues therefore in-favour of models of language which reflect this

heterogeneity, rather than those which try to reduce everything to

constituent parts.

Psycholinguistic evidence on how idiomatic structures are

processed tends to argue for the Bolinger view, and against the sort

of schema proposed by Weinreich and Searle, in which the "literal"

meaning is checked against context, and if found to be inconsistent, a

"figurative" meaning is sought. If comprehending idioms involved the

number of steps of back checking etc suggested in the comparison

approach, we would expect idioms to take longer to process. However,

Ortony, Schallert, Reynolds and Antos (1978) found that subjects could

comprehend idioms as fast as literal interpretations, and possibly

even faster, in situations where context was provided. In Swinney and

Cutler's (1979) experiments, subjects saw contextless idiomatic and

non-idiomatic sentences and consistently comprehended the idioms

faster. Swinney and Cutler argue that their results favour a lexical

representation of the idioms they looked at.

175
 Chapter 6

Many idiomatic uses are distinguished from their "literal"

counterparts by intonation. Informant testing conducted by Jeffries

(1981) indicated that hearers accurately recognise recordings of

idiomatic uses, even when they are completely decontextualised. This

also suggests that the Weinreich/Searle type account is not the right

one. It might also indicate that such idioms are listed directly in

the lexicon, complete with their intonation, since they can be looked

up as such.

Ladefoged (1972), speculating on the mental organisation of

language, writes:

"The indications from neurophysiology and psychology are

that, instead of storing a small number of primitives and
organizing them in terms of a (relatively) large number of
rules, we store a large number of complex items which we
manipulate with comparatively simple operations. The central
nervous system is like a special kind of computer which has
rapid access to items in a very large memory, but
comparatively little ability to process these items when they
have been taken out of memory" (: 282)

To summarise then, a string which is a candidate for listing in

the lexicon as a string (rather than being looked up under its

constituent parts), should have some of the following characteristics:

1 meaning not accessible from the sum of its parts

2 some degree of "frozenness" to syntactic manipulation'

3 some degree of constraint on replaceability of items within it

4 some degree of prosodic "frozenness"

Of course this is not to say that some perfectly well-formed

apparently compositional strings might not also be stored as wholes in

memory. From the point of view of linguistic model making, however,

they are not a problem.

176
 Chapter 6

In the next section, I shall discuss each of the phrasal vague

additives, in relation to its constituents, and consider whether it

requires'a lexical listing or not.

6.4 Lexical Meaning and Phrase Meaning

6.4.1 nor m

As far as the 'common topic' condition is concerned, n or m

expressions keep to it, in the sense that n and m must both be

numbers. As we have seen also, they can only be certain numbers, and

being a Reference Point Number is a property common to both n and m.

Secondly, the question of ellipsis, or reduction from two

sentences. Is it feasible to consider deriving

(13a) + He's written five or six articles

from

(13b) [He's written five articles] OR [He's written six articles]

SS

Clearly not, since the alternative meaning assigned to (13b) is quite

different from the interval meaning associated with (13a). Quirk et

al (: 13.70) note that what they call co-ordinate expressions of

approximation with or are very difficult to explain in terms of

ellipsis. In this way then, the vague use of or is unlike its

alternative use.

My third point was that alternative or is almost always

understood as exclusive. Although we have gone some way to

establishing that or in an approximation is not understood as

presenting two alternatives, it is the case that the whole vague

expression is understood as disjunctive between many alternatives (ie,

all the numbers in the interval).

177
 Chapter 6

The final point on or was the entailment relationship described

by Hurford and modified by Gazdar. Here, approximative uses of or fit

in with Gazdar's constraint. (a) the m number always entails the n

number and at the same time, (b) the n number potentially implicates

the negation of the m number, since it is a property of the use of

numbers for quantities that they potentially implicate "at most".

See, for example, discussion of "Mary has five children" in Wachtel

(1981).

I have described in Chapter 3 how the prosodies observed in vague

uses of n or m are quite dissimilar from disjunctive uses. This

prosodic marking suggests that this use of or should not be considered

as disjunction. Secondly, we have seen other dissimilarities.

Thirdly as we have seen, the interval-of-numbers meanings, which

hearers attribute to these structures, have no direct connection with

orthodox disjunction which presents just two (exclusive or inclusive)

possibilities.

Yet this use of or is indirectly related to the disjunctive use.

Firstly, disjunction expresses uncertainty: or connects possibilities,

one or both of which may be true. Similarly in n or m expressions,

there is uncertainty, and a whole interval of numbers is given as

possibilities. Secondly, the fact that the two exemplar numbers each

side of or, are always considered to be members of the interval of

possible numbers, cannot be ignored. An inclusive disjunction reading

accounts for that, what it does not account for is all the other

numbers in the acceptable interval.

Finally, the question of n or m as a corvsAVMC %öv lb- As far as

..

the characteristics of phrasal lexemes were concerned, it is true that

the meaning attributed by hearers to n or m cannot be deduced from the

178
 Chapter 6

sum of the meanings of its parts. Neither can its systematic

difference in meaning from the other number approximations, such that

it is always understood as allowing more possibilities (being more

vague) than about/(a)round. If the approximation meaning is to be

read off without a special entry, this will have to be by

means of an implicature. It will be no good however, having this

implicature attaching whenever two numbers are conjoined by or, since

it is not present when the numbers are genuine alternatives. If it is

present only when the correct intonation for a vague use has been

placed, the implicature will have to be determined by the intonation.

But the intonation. will have to be determined by the vague meaning

(which is the implicature) so this will not work. In addition, the

implicature would be non-cancellable, which implicatures classically

should not be. This is not the way to deal with this., Given the

relative (for a language system) rigidity of the constraints on

selection and ordering of numbers, and on intonation, a separate

iype, entry seems far preferable.

6.4.2 n or so

One of the most interesting points made by Wachtel (1981) in his reply

which commented on a paper (Channell, 1980) in which appeared part of

the material on number approximation presented in Chapter 3, is that n

or so should be regarded as a variant of n or m with so being treated

as a pro-numeral (: 316). This is also the approach of Quirk et al

(: 13.70) who refer to the numeral following or being replaced. by so.

As Wachtel notes, this analysis is consistent with the results I

obtained for the meanings of n or so since the intervals invariably

were skewed upwards from n. If so is being understood as in, and since

m is always larger than n, this is what we would expect. Furthermore

the magnitude of m is, as I have described, determined by that of n,

179
 Chapter 6

so a hearer may compute a likely m for any instance of so, from the n

that he hears.

That so is understood as a proform for an m number, is also

consistent with its many others uses as a proform, which I outlined in

6.2.2. There is also some meaning relationship between this so and

that found in and so on, which we recall was glossed as "and other

things of this kind". Hence this so can be glossed as "another thing

of the kind we just had" viz, a number.

As we would expect, the or in n or so is related and not related

to disjunctive or in just the same ways as the one in n or in. In this

case, there may be some arguments for not creating ä separate lexical

entry for n or so, but for seeing it as a variant of n or in, created

ý by use of the proform so which must have its own (rather abstract)

lexical entry. The constraints noted on n or m will apply in the same

way.

6.4.3 Or Something

Under this heading are encompassed the variants

or (something ) (like that)

(anything )

Here we have another occurrence of or, and we must consider whether

this is a disjunctive use or not.

The 'common topic' condition applies, in the sense that, as we

have seen, the major aspect of meaning which has been observed is that

the tag is understood as directing the hearer to a' category which

shares features with the preceding exemplar (Chapter 4). There were

no (compliant) responses in the tests, where informants interpreted

the tag meaning as completely unrelated to the exemplar. In this

180
 Chapter

respect it is clear that what hearers know of the usual use of or

contributes importantly to their understanding of or something.

Secondly, the question of ellipsis/reduction. This sometimes

appears to work, such that:

[one of the secretaries was saying there was a film] OR

S [one of the secretaries was saying there was
S something like that]

can be read, I think, as having the meaning associated with the

derived sentence. To make this plausible, it seems necessary to see

like that as having been present, but deleted in all instances of

these tags.

Where or something is conjoined to a VP, reduction/ellipsis is

not so neat, thus:

[could we sort of meet] OR [could we sort of do something like that]

SS

since an invariably deletable dummy do is necessary in the second S.

This would make these VP + tag constructions not accountable by the

regular ellipsis rules which could be formulated to apply to

disjunctive or.

Another possible argument against seeing or something as the

product of ellipsis is prosodic. In the case where two alternatives

are presented, with or, the second one is usually a separate tone

unit, and the or may even be a separate tone unit and receive a heavy

contrastive stress, thus:

(14) You can come with us or. yon can stay at home

As I reported in Chapter 4, all the instances of or something that I

have observed, have a one tone group intonation, and I do not think

contrastive stress on the or is possible:

\\
(15) ?*there was a filmllor11something

181
 Chapter 6

It's rather difficult to assign a meaning to (15).

Then there is the question of exclusive vs. inclusive readings.

The evidence is, I think, that or something tags behave just like

other or-disjunctions in that pragmatic considerations dictate whether

hearers interpret these as exclusive or inclusive; thus:

(16) +Could we sort of meet or something

is heard as exclusive (as are most tags, in the same way as most

disjunctions), but maybe something like.

(17) Would you like some dessert - trifle or something?

could be heard as inclusive.

Of course, if this or is not like alternative or, the question of

inclusive versus exclusive may be irrelevant. One reason for

suspecting this, is that speakers do not appear to have options of

forcing the exclusive and inclusive readings by using either and or

both, respectively. I'm not sure-one could say:

(18) *One of the said there'was either a film or something

secretaries

and still less

(19) *One of the secretaries said there was a film or something or

both

and I have no examples of anything like this. These uses of either

and or both are quite normal with alternative or. This then, is a

very clear difference from alternative or.

The entailment situation is somewhat blurred, but subjects'

responses indicated that they used the Gazdar-formulated potential

implicature constraint, such that for example, for oranges or

something, if SOMETHING = any member of-the category of orange-like

things, then "something" implicates NOT "oranges".

182
 Chapter 6

Something is also a constituent of this tag. Its gloss of "some

unstated or unknown thing" fits in with what we can observe of its

meaning here, since its unknownness is qualified by the like that

which attaches to it. The fact that it may appear with plural

exemplars, is also quite in accord with Mittwoch's observation that

other uses of something can be used to report plurals.

Like indicates "such as, resembling". As we have seen, tags are

always taken to indicate items which "resemble" in some sense their

exemplar. It seems obvious that the regular meaning of like

.
contributes importantly to this meaning. In addition, we have the

presence of that, which, given the informants' responses, must be

taken I think as a deictic pointer back to the exemplar.

Thus the meanings hearers attribute to or something tags are very

closely related to the constituent parts of the tags. The only

problem for an account which rests-on combining the constituents each

time, is the behaviour of or, which differs from its "normal"

counterpart, in particular as far as intonational characteristics are

concerned. If there is a cline with at one end cliches and fixed

collocations, and at the other, frozen idioms, then it looks as if

this expression may be closer to the fixed collocations end.

6.4.4 And things (like that)

I gave in Chapter 4 the reasons why I consider like that to have been

ellipted from occurrences of and things.

The 'common topic' condition on and, is seen to apply to these

tags for the same reasons as I gave in more detail for or something,

in the previous section, and to avoid repetitious discussion, I will

not go through them again.

183
 Chapter 6

And in this tag, is understood as a conjunction, in the sense

that both the exemplar and the other members of the associational

category are understood as being indicated. Thus it contrasts in the

usual way with or.

On the question of ellipsis/reduction, again here, some sentences

look quite convincing:

(20) +1 had to give them sentences with it in and things like that

[I had to give them sentences with it in] AND

S [I had to give them things like that]
S

The strongest reason for not seeing this and as the result of ellipsis

is prosodic. The and which occurs in tags, never (at least in all
-is

the examples I have) full and stressed, whereas, as I noted in 6.2.2,

the and of putative ellipted constructions may be full and stressed.

Now, things, as noted, is more or less a proform for anything,

and as we saw, there is another instance of its use where contextual

factors of some kind determine, as here, how it should be "filled"

("swimming things"). Another link with that use is that it is

invariably plural, just as it is in this tag.

Clearly here, just as with or something, the (sometimes tacit)

presence of like that contributes importantly to the meaning of the

whole tag.

So the situation parallels almost exactly that for or something.

Everything can be accounted for from the constituents, except the

special aspects of and, and the particular and rigid intonation

structure.

6.4.5 And That

In Chapter 4,1 considered and that independently of the other tags.

184
 Chapter 6

As we saw at that point, it appears to have the same meaning, in the

sense that something like car and that is understood to designate an

associational category of' which car is a good example. And in and

that is in the same relationship with its "normal" use as that in and

things like that. That also appears to be the same.

The difficulty is that there is nothing lexical in this tag from

which to read off the meaning "some other member of the same

associational category", since and will only give conjunction and

perceived similarity, and deictic that, as a deictic pronoun referring

to the exemplar, gives only a repetition of it. One possiblity then

is to see this as much more idiomatic than the other tags, in the

sense that the meanings observable for it are definitely not available

from the sum of its parts.

The other possibility is that again here, there is ellipsis, from

and things like that, to and that. Then the full version contains all

that is necessary to provide the meaning. I do not however, beyond

the plausibility of this analysis, have any evidence for or against

it. I don't have, as I do for example for or something, evidence that

the two variants are in variation for the same speaker, in similar

contexts and with similar meanings understood by hearers.

It is therefore possible only to conclude that (a) the

constituents of and that contribute to its meaning, but that (b) it

may well need a special lexical entry.

6.5 The Single Word Number Approximators

about/around/round/approximately n

These are different from the expressions so far considered. They are

rather less likely candidates for being seen as idioms, and rather

185
 Chapter 6

more likely to be seen as fixed collocations when conjoined with

numbers. Each of the words has other uses than with numbers, as

follows:

6.5.1 About

About occurs otherwise than with numbers as a spatial preposition, eg

(the Longman Dictionary of Contemporary English) "They walked about

the streets" and as an adverb, (the Longman Dictionary of

,
Contemporary English), "Is there anybody about? ", which Quirk et al

(6.10), suggest is an ellipsis from a prepositional phrase such as

"about the house/ building etc". It also occurs as a non-spatial

preposition: (the Longman Dictionary of Contemporary English) "Have

you a book about the stars". This non-spatial use has an identifiable

relationship with the spatial uses. So about is considered as meaning

"somewhere within the general area of" both literally and

figuratively.

6.5.2 around/round

These are similar to about, in spatial uses. Bennett (1975: 86-88)

analyses the locative use of (a)round as [+locative] [+surround]. He

notes that around and round are very often interchangeable and

concludes: "It is unclear to me whether any correlation can be made

between the semantic structure in a given instance and the likelihood

that one or other of the two forms will occur. "(: 88) My observation is

that in approximation uses around is more common, whichisthe opposite

of Bennett's observation concerning spatial uses, so there may be a

tendency to move towards a rule. Around and round have figurative

uses, too, eg "They talked round/around the subject for a while".

6.5.3 Approximately

Approximately is different, in that the Longman dictionary claims that

it is only used with numbers. It seems to me however that it can be

186
used like more or less, as in:

(21) ?The work is approximately complete

(22) The card is approximately circular

What is clear is that approximately does not have the spatial

prepositional uses which can be observed for the other three

approximators.

6.5.4 Discussion

The first reason for not seeing these as idioms is the degree of

substitutability possible, any of them can be substituted for each

other, and for a number of other prepositions (see below).

Secondly, there are no particular intonational patterns found

with the number uses.. There is not in principle any technical

difficulty about constructing these expressions afresh from the

lexicon each time they are required.

The other general question on these is whether the number uses

are different senses (ie is there polysemy [2] in the sense of Lyons,

1977: 550ff), or will just one sense suffice to account for the

different observable uses? This is not a thesis on lexical semantics,

and again here, it must be acknowledged that there is really little

agreement among those in the field about how to establish different

senses, or how to deal with them when you have. Cruse (1982) surveys

the various discussions and other references are Nunberg (1979),

Kempson (1979) and Lyons (1981a: 146ff). It seems best,

therefore, not to get involved in this, but to offer the observations

on the number approximation uses of these words as a. contribution to

2 Homonymy is ruled out because some sharedness of meaning can be

identified.

187
 Chapter 6

their future lexical analysis.

As far as the number use of about/(a)round is concerned, we could

see this as an extension from the spatial use in the sense that the

language user conceptualises the number system as a long, narrow

space, to points in which he can refer with spatial prepositions, just

as he can to points in a street. Since many of the spatial

prepositions can be used with numbers (under, and over, beneath, near,

etc) this will be an instance of what Lehrer (1978) has analysed as

semantic transfer. Her observation is that when one member of a

semantic field develops a new sense, others will also develop related

senses. In the case of approximately, however, we may prefer to see

the development going in the other direction.

6.6 Conclusions

In this section, I have considered the constituent parts of vague

expressions in relation to their total meaning. What is most

interesting about them, I think, is that they are a good example of

expressions which have a partly systematic and partly idiosyncratic

relationship with their constituent parts. This tends to support the

view of language as something with "more patching and gluing about it

than it has architectonics" (Bolinger, 1976: 1), and hence the goal of

a model of language which reflects just these heteregeneous

characteristics.

188
 Chapter 7

Chapter 7

Semantics, Pragmatics and Vague Meanings

7.1 Approach

As I explained in Chapter 1, one reason for undertaking an examination

of vague language was that there were already suggestions that it

would be a hard testing ground for currently accepted semantic

theories, and in particular for the necessary exercise of boundary

drawing between semantics and pragmatics. In Chapter 1, I also

mentioned a paper by Allwood (1981, hereinafter, Allwood), in which he

reviews various theoretical approaches to meaning, and finds them

wanting, being led thence to the general conclusion that:

"... so far we have no theoretically interesting and

consistent way of separating semantics from pragmatics and
that perhaps the distinction is more of a hindrance than an
aid to clarity in the study of meaning in natural languages. "

In this theoretical chapter, I consider some possible ways to

separate semantic and pragmatic aspects of vague language, and offer

some programmatic sketches of how each approach could work, drawing on

Allwood's discussion. I begin by summarising, in the next section,

the vague meanings to be accounted for, and I argue that the two types

of vague expressions examined should in fact be analysed as being the

result of the same process.

7.2 Vague Meanings

I begin by reproducing from Chapters 3 and 4 the observations of

meaning for the Number Approximations, and for the Vague Category

Identifiers.

Observed meanings associated with Number Approximations

1 There is a set of expressions whose effect is to bring a vague

reading (= an approximation) to an utterance containing a number

189
 Chapter 7

2 The resulting approximations are understood as designating

continuous intervals of numbers

3 Different approximators change the interval designated. Given an

Exemplar-number x, (and setting aside the effect of other factors such

as situation, and the form and nature of the E-number), the effect of

the different approximators (as observed from the test results) can be

seen schematically as follows:

i
about/(a)round 7.62
32.30
----------9 ____

n or so 15.46
56.69
___

norm 22.86
66.60

= minimum

-- -- = maximum

Figure 4

4 Although there is a high degree of agreement among speakers that

numbers near the exemplar number are members of the interval, there

appears to be variation about the extent of the interval in any given

case

5 The size and form of the exemplar number both affect the length of

the interval

6 Whether the E-number is a round number (- Reference Point Number)

or not affects the length of the interval ("about 31" vs "about 3Q")

7 The nature of the item being approximated (discrete vs

non-discrete, man vs cockroach) affects the interval

8 The conversational setting in which an approximation occurs affects

how it is understood

190
 Chapter 7

9 Sentences containing approximators characteristically have the

entailments and implicatures described in 3.7.

Observed meanings associated with Tags

1 Vague tags are understood to designate categories, either

conjunctively or disjunctively, consisting of either concrete

("bread", "silver") or abstract ("meet", "elaborated code") entities.

2 These categories a; e associationally rather than semantically

defined.

3 Several factors establish for the hearer the characteristics

determining membership of the category. They are (a) the exemplar,

which is understood to be a "good example" of the intended category;

(b) the surrounding linguistic context; (c) the purpose of the

conversation; and (d) the hearer's world knowledge which he brings to

bear as relevant to the linguistic context and conversational setting.

4 There is some evidence to suggest that these expressions cause

particular comprehension problems for hearers who lack specific world

knowledge of the conversational topic. That is to say, more problems

than non-vague expressions in the same situation.

7.2.2 A Unified Approach

Now, what is common to the meanings of both types of expressions

considered is that from an exemplar of some kind, a set consisting of

a selection of possibilities is inferred. For the number

approximations, this set is a subset of the set of Real Numbers, of

which the exemplar number(s) is/are (a) member(s). For the tags, it

is a set of abstract or concrete entities, of which the exemplar. is a

member.

An important observed characteristic of the sets associated with

these vague expressions is that features of linguistic context, and

situation, and world knowledge, all play a crucial role in

191
 Chapter 7

establishing, for any particular utterance of any vague expression,

what the set consists of.

We saw in Chapter 3, in relation to number approximations, how

factors such as the perceived purpose of the approximation, and the

item being approximated, appeared to affect the interval of numbers

which were ' thought of as possibilities. Similarly in Chapter 41 for

the tags, I described how the different results observed for the same

exemplar demonstrated the influence of world-knowledge and imagined

context on the sets which the subjects produced as their answers.

Here then, we are able to observe in operation a clear instance

of hearers "going beyond the information given" (in the sense of

Bruner, 1973: 218), by use of particular inferential procedures. Such

a view of language understanding is not new, cf Fillmore (1977) who

says that interpreting a text:

"... involves. much more than the processing of

meanings directly provided by the text, but also memories,
knowledge, and current perceptions of the interpreter, as
well as the application of a set of procedures for
determining the basis of the coherence of the text. " (: 86)

On the more recent view of Milroy (1983), it is the inference

procedures which have primacy in decoding utterances on the lines of

'what kinds of things could this person possibly be meaning, given

this situation, and what I know', and she suggests, it is only if this

fails that recourse is made to the details of syntactic arrangement

and lexical meaning, rather than, as more usually suggested, the other

way round. She writes that "the relation between internal grammars

and comprehension is indirect". Danell (1978: 14) writing specifically

about vagueness, makes the same point. Another instance of going

beyond the information given is the inferential procedures described

by Allan (1981) for the understanding of animal nouns, like lamb,

192
 Chapter 7

which are understood as denoting meat or skins, or other aspects of

lamb, depending on context. He sketches the sort of inferential rules

which would be involved. Clark (1978) also gives a programmatic

account of inference steps. (cf also the various approaches to shared

knowledge and inferences in Smith, 1982. )

In the present case, we may deduce that hearers go beyond the

information given, in rather closely circumscribed ways, such that

they agree strikingly as to the results they come up with. How do

they do this? Bruner:

"... when one goes beyond the information given, one does so
by virtue of being able to place the present given in a more
generic coding system and one essentially reads off from the
coding system additional information either on the basis of
learned contingent probabilities or learned principles of
relating material. " (: 224)

This brings me to considering a further point about these vague sets.

The sets of possibilities which hearers understand are sets which are

well-defined in the sense of forming a recognisable category (cf (Tag)

T-test informants' gloss responses: Table 4.5 and 4.6) (clearly in a

rather more circumscribed way for the number sub-sets). This is to

say that in going beyond the information given, hearers must be using

the principles of categorisation which have been observed as a

fundamental aspect of human cognition ( Palmer, 1975, for visual

perception, Rosch and Lloyd, 1978, for psychological studies of the

structure of categories, and the processes by which they are formed).

Very generally, what this work shows is a cognitive predilection to

organise items into categories, and to relate incoming stimuli to

appropriate categories.

Most of the work conducted in psychology has up to-now been done

on categories of concrete objects. We have seen that where the vague

tags contained concrete objects whose membership of particular

193
 Chapter 7

categories has already been demonstrated experimentally, the sets

associated with them often bore a strong similarity to the

experimentally determined categories (4.3.1). The situation for

categories of non-concrete entities is quite otherwise since there is

little work on them. Thus the analysis I am suggesting here is making

an inferential leap in claiming that the non-concrete vague categories

are equally the results of cognitive principles of categorization. It

is, however, a leap which I think is warranted.

I have said the sets are discernibly circumscribable. But it

appears that their membership "boundaries" are not fixed. Informants'

comments, observations of use, and the results of the two tests, all

suggest strongly that these sets are woolly-edged, - "fuzzy" just in

the sense of Zadeh (1975):

"We have beenslow in coming to the realisation that much

perhaps most, of human cognition and interaction with the
outside world involves constructs which are not sets in the
classical sense, but rather "fuzzy sets" (or subsets), that
is, classes with unsharp boundaries in which the transition
from membership to nonmembership is gradual rather than
abrupt. Indeed, it may be argued that the logic of human
reasoning is not the classical two-valued or even multivalued
logic but a logic with fuzzy truths, fuzzy connectives, and
fuzzy rules of inference. " (: ix)

In keeping with what has been observed in studies of

categorisation, the sets appear to be internally structured in the way

suggested by Lakoff (1972) and Rosch (1975a), such that, for example,

all subjects agree that 15 is one of the possibilities allowed by

"about 15" (N-Test item 5: Table 3.3), and all subjects (who gave item

answers) agree that apples is a member of the set of possibilities

allowed by "oranges or something" (T-Test item 1: Table 4.4). Whereas

at what we may see as the "outer edges" of'the sets, agreement becomes

less and less.

194
 Chapter 7

This brings me to my final general observation about how vague

expressions are understood. We saw in Chapter 4 that the exemplar in

vague Category Identifiers had to be regarded in some sense as a good

example of the set it was to identify, and I described how the

exemplars which were "good examples" from Rosch's (1975a) categories,

produced responses which bore in some cases a very close resemblance

to the membership of the relevant Rosch category. From this data, it

seems reasonable to infer that understanding vague additives involves

using another well-documented aspect of human cognition, the

Prototype. Rosch (1978: 40-41) explains carefully that to speak of a

Prototype is "a convenient grammatical fiction; what is really

referred to are judgments of degree of prototypicality. Only in some

artificial categories is there by definition a literal single

prototype [. For natural language categories, to speak of a

.]

single entity that is'a prototype is either a gross misunderstanding

of the empirical data, or a covert theory of mental representation".

In this sense, what are understood to be used as exemplars are items

which have a high degree of prototypicality for the understood set.

Thus an exemplar with a low degree of prototypicality - necklace, for

the category clothes; ( T-Test item 4: Table 4.4) - does not direct

its hearers to the set of clothes, but to another, different, set.

This principle applies equally to the number'approximations. We

saw in Chapter 3 that these are highly likely to contain Round

Numbers, and we saw also that Rosch's work on the internal structure

of the category of number names showed that Round Numbers were

understood as reference points, or in some way prototypical exemplars,

in the structure of the number system. At the same time, I must

recognize specifically that the number system, as used in

approximations, is rather different, in the sense of being much more

195
 Chapter 7

highly structured than the vague conceptual categories which are

identified by the vague Category Identifiers.

Nevertheless I believe it is fair to say that the same general

principles of manipulation are at work in both cases.

7.3 Semantics and Pragmatics

In this section, I look at different approaches to drawing the

boundary between semantics and pragmatics. As noted by Allwood, this

often takes the form of considering what is semantic to be in some way

situation independent, whereas what is pragmatic will be closely

related to actual contexts.

7.3.1 Literal Meaning

"This meaning is either seen as present in all situations of

use, due to lingtstic conventions dependent on association
between form and content only, or as a basic meaning from
which the meaning of the linguistic expression in all
situations of use can be derived. " (Allwood: 183)

Literal meaning is thought to be closely connected with the meaning

one can understand from decontextualised sentences. The first thing

to be said on this is that it seems doubtful that what one gets from

reading a decontextualised sentence is a decontextualised meaning.

What we do when confronted with decontextualised sentences is to

invent plausible contexts which enable us to understand them (Wachtel,

1981; Mikkel Blakar and Rommetveit, 1975; Lanin, 1977). The necessity

of recontextualising in order to understand is demonstrated partly by

what happens when we cannot do it. As I mentioned in Chapter 4,

subjects in Bransford and Johnson's (1972) experiments who heard texts

which had been artificially contrived. to omit suitable contextual

cues, performed much below normal on comprehension and recall tasks.

The kinds of results Iýobtained in my test point to very much the same

196
 Chapter 7

strategy of 'recontextualisation on the part of my informants. This

showed up in particular in the T-test, where the same exemplar was

recontextualised in two items in two quite different ways (4.3.1 and

Tables 4.4 and 4.5).

So perhaps there is no such thing, psycholinguistically, as a

decontextualised sentence and decontextualised meaning. But this does

not mean there cannot be a valid notion of decontextualised meaning

for descriptive purposes, if we can find a valid procedure for

isolating it.

Allwood considers three approaches to achieving it. Firstly the

intersection or common denominator approach.,

According to this approach, those aspects of meaning which are

semantic are those which are common to every occurrence of a given

linguistic expression. In order to simplify the discussion in what

follows I shall use just a few examples which will be representative

of the two kinds of vague expressions described.

In the case of the number approximations, we will want to

attribute a literal meaning to five or six which reflects what is

common to each of the three occurrences in examples (1) to (3), each

of which was examined in Chapter 3 or Chapter 5:

(1) [academic publishing]

+You find that you get five or six articles and they're all
very much the same [11,21.2]

(2) [buying Christmas trees]

+There's another shop in Blackheath that was selling really

scrawny ones - five or six quid a time [Camb 9b14/32]

(3) [informant work]

+We've got about five or six of them but I'm only going to
talk about three of them today [LAGB, 9,81]

197
 Chapter 7

Now, drawing on the results of the informant test (in which, as it

happened, none of these three examples was used, although (1) was

discussed by the subjects), we know that for each case a vague

interval reading of possibilities will result. We also know that the

different items being approximated will give different intervals:

articles (1) and informants (3) cannot be subdivided (at least not in

the contexts in which they were used in these examples), whereas

pounds (2) can - this will probably give a narrower interval for (2)

than for (1) and (3). Next, there is the influence of the perceived

purpose of the utterance. The test subjects who heard (1), commented

that:

"He's using it for emphasis - because the whole sort of drift

of the argument is that there are too many papers being
produced (. ) that's a signal that it can just possibly be a
sort of vague number saying there are lots of papers being
produced and there shouldn't be that number"

They felt that S did not have exact knowledge. In fact he could not

have have done, since he wished to make his utterance fit a set of

several situations in the world.

(2) is, I would judge, another case of S making his utterance

general for several cases. He many or may not have remembered some

exact prices he saw. In addition, on the basis of the fuller extract,

I suggested that informality played a role here. In (3), as I

described in Chapter 5, hearers probably understand that S does know

exactly how many informants he has. In this case his approximation

draws attention away from this part of what he says, and towards what

is important, that three will be spoken about. In addition, Ss use of

but tells hearers that '3' is not part of the interval intended by

five or six, since it is contrastive.

198
 Chapter 7

What we may extract as general is that an interval (a subset of

Real Numbers) is understood, of which 5 and 6 must be members. Also

this interval is fuzzy, in not having its end points definite in any

given case. All other aspects of the meaning must then be accounted

for by pragmatic rules which reflect the inferential processes hearers

use. For example, for five or six informants:

Inference 1: S knows how many he has (from premise, S is giving

a paper about a subject he can be presumed to
know well)

Inference 2: the formal setting (in comparison with, for example

that of example (2)) rules out the choice of
approximation for informality

Inference 3: S is focussing attention (by elimination of other

possibilities)

Inference 1: persons are wholes (from situation)

Inference 2: interval is probably 4-7 (from the exemplars, 5

and 6, and the number 3 being excluded by but)

For the tag approximations, I shall look at two tags which had

the same exemplar, in the informant test (results are in Tables 4.4

and 4.5):

(4) I
stopped myLiIV- by the, verge, then a car or something
came along and splashed me all over with water (T-test 16)

(5) They've got a car and that (T-test 24)

I leave aside the disjunctive/conjunctive contrast, which can be

accounted for without difficulty either from the semantics of and vs

or, or by stipulation in complex lexical entries, as we saw in Chapter

6. Again here, sets are involved in understanding both examples, and

sets of which car, the exemplar, is both a member, and a central

member or good example.

199
 Chapter 7

In the two cases, different supposed context supplied by the test

informants led them to produce two different sets, which can be

loosely characterised as "the set of wheeled road vehicles" (4), and

"the set of desirable items associated with relative wealth" (5).

What is common to both occurrences is that a set of associated

items is involved, of which the exemplar is a member. Only this then,

can be analysed as semantic. All the mechanism of using the exemplar

as a prototype, taking account of context and purpose, etc, in order

to read off the relevant set, is to be dealt with by pragmatic rules

which again would embody the likely inferential processes, on the

lines, of, for example for (4):

Inference 1: linguistic context gives public highway as likely

setting

Inference 2: linguistic context specifies things which can splash

Inference 3: vehicles are most likely candidates

Two things are noticeable about this sort of approach. Firstly, the

proposed semantics is rather uninteresting, in the sense of making a

quite small relative contribution to the observed meaning of the

expressions in use. (cf Allwood making this same point: 184). Many

linguists, notably Lyons in his recent writings (1977,1981a, 1981b),

have a pretheoretical predilection to keep more of meaning in

semantics than would happen with this approach. Secondly, the

semantic representations would have to be vague. Many people are not

prepared to allow vagueness in semantics (eg, Wachtel and Sadock,

discussed below). (But see Klein (1982) for proposals to allow vague

expressions in the formal language needed for the semantics of

comparatives. )

200
 Chapter 7

The second approach to literal meaning is the idea of an essence

- abstracting 'basic' or 'ideal' aspects from instances of As

use.

noted by Allwood, what is difficult in this case is to recognise what

the essence of a linguistic expression is. In the case of the vague

expressions, the obvious candidate would be that they are vague - this

appears to be the essence of all the actual instances of use. On this

account, perhaps, the semantic representations for English would

contains a subset of vague representations for these kinds of

expressions. Pragmatics would then take care of all the rest of the

meaning effects which were observed. Again this entails semantic

representations which do very little of the work of explaining the

meanings of vague language, and semantic representations which are

vague. In addition it is difficult to see what empirical procedures

could be used to determine basic meanings.

The third approach to literal meaning which Allwood considers is

to say that it consists of the union of all uses of an expression.

This has the effect of "connecting any information that has been

conveyed by a linguistic expression with its literal meaning".

Clearly for the expressions I am considering, this would be nothing

short of disastrous, since, say for car or something, all the

conceivable sets of which it could be thought to be a good example in

all situations of use, would be involved in the literal meaning of the

string. Included also would be all the conversational effects which

we saw in Chapter 5 which can sometimes attach to uses of vague

expressions.

The next solution Allwood considers is that which, says that only

some linguistic expressions have situation independent meaning - the

syncategorematic expressions. Categorematic or contentful expressions

201
 Chapter 7

depend on context for specification. The difficulty here for the

analysis of vague tags is that they contain a mixture of what are

usually thought of as syncategorematic items (and, or, perhaps the

numbers) and categorematic ones like that and things. If we choose,

as I argued we should in the preceding chapter, to regard the

expressions holistically, then they would be categorematic. But the

idea of having syncategorematic items mixed up inside categorematic

"lexical items" is not theoretically attractive.

On the matter of the relationship between literal/conventional

meaning and referential meaning, Allwood cites Kempson (1977) as an

example of someone who takes the two notions to be the same. His

point is that they cannot be equated, - because intuitively,

conventional meaning is much wider than referential meaning. In the

case of car or something, a description of its conventional meaning

(vague, indicates a 'set, etc), will be very different from its

referential meaning, which only comes into existence at the moment of

its use on a given occasion.

7.3.2 Bivalent Truth-Conditional Semantics

A particular version of the common-denominator approach is truth

-conditional semantics, in which a statement of the conditions which

must pertain in any situation for a sentence to be true, are seen as

all and only what is necessary to state its semantics. In a bivalent

semantics, only two values are possible -a sentence is either true,

or false, of a given situation. Pragmatics, in this model, concerns

itself with all those aspects of meaning which are not amenable to

treatment by truth-conditions (cf Gazdar 1979: 2).

202
 Chapter 7

A considerable literature exists which argues that many aspects

of meaning general to all occurrences of a sentence, are quite

unamenable to treatment by truth-conditions, for example Kempson,

1975, Section 2. Lyons (1981b) brings to bear criticism of

truth-conditional semantics on account of "its restriction to

propositional content and its inability to handle the phenomenon of

subjectivity" (: 240), and in Lyons (1982), he describes how it cannot

easily handle such things as modals, and the French 'conditionnel de

citation' which he argues are manifestations of subjectivity. Allwood

rehearses many of the standard examples.

The vague language under study here provides another example of

utterances which are not easily amenable to treatment by truth

conditions. I shall look at two attempts which have been made to

account for number approximations. As far as I know, no-one has

proposed a bivalent truth-conditional account of the vague Category

Identifiers, but given their manifest similarities of meaning to the

number expressions, I can show how they might be dealt with within the

two approaches to be discussed.

The problem for a truth-conditional account is that for any use

of a number approximation, it is apparently impossible to specify the

point at which it ceases to be true, so for example, (2) is probably

judged true of trees actually costing ßf7, but probably untrue of trees

J14.
costing

Similarly the tags exhibit the same cut-off point problem. (4)

would probably be judged true of a situation where the thing in

question was a bus, marginally true if it was a horse, and probably

false if it was an elephant.

203
 Chapter 7

Sadock's (1977) account offered a semantics for sentences

containing approximately. As we saw in Chapter 3, the meanings

associated with approximately are very similar to those of other

number approximations, so we may safely assume that his account could

be applied to them as well. It could also be applied to the tags, as

I shall show below.

Sadock's approach is to argue that because of the multiplicity of

factors affecting appropriate approximation, the only possible

truth-conditional account is "a fairly trivial one", in which a number

approximation is always true, provided the thing being approximated

can have the property of being measurable. All the relevant

constraints on what makes approximations acceptable in given

circumstances will be dealt with by pragmatic rules, in the same way,

Sadock proposes, as Grice's (1975: 70) proposals for tautologies, which

as noted also by Allwood, are referentially- all identical, but

different in their sense.

Sadock's account would extend quite naturally, I think to the tag

approximations, in the sense that we could analyse the category

identifier in (4) as being true in all circumstances, but only

appropriate in certain ones, determined by the sort of contextual

factors I noted, which would'be dealt with by pragmatics.

Sadock makes a number of useful observations on the use of

approximations, which I described in Chapter 3. This is an example

perhaps of the usefulness Lyons (1981b) attributes to attempts to deal

with unamenable data:

"The failure but inadequate, N.

of a precise, account often
points the way to the construction of an equally precise, but
more comprehensive, theory of the same phenomena. And even
when it does not do this, it may throw some light, obliquely
and by reflection, upon the data that it does not fully
illuminate. " (: 143)

204
 Chapter 7

Sadock's account, however, is untenable, even on its own terms.

A truth-conditional semantic account must be able to deal with the

entailments of the sentences it purports to account for, as well as

the effect on them of the putative logical operators such as NOT.

Wachtel (1980: 203) shows that Sadock's account fails, in making an

incorrect prediction about the meaning of:

(6) Sam is not approximately six feet tall

which should entail the proposition that Sam has no height- at all,

which it arguably does not. However the defence to this is the one

used by Wachtel to get out of a counter-example to his own proposal,

that (6) is an example of what Sperber and Wilson (1981) call"mention"

rather that "use", so is not a denial of the content of the

approximation. Fillmore (1971: 122) calls this semi-quotation, one of

his examples being:

(7) 1 didn't "escape" from the prison; they released me

In addition, Sadock's account predicts that

(8) Sam is exactly six feet tall

entails

(9) Sam is approximately three feet tall

which Wachtel argues is "counterintuitive".

Wachtel's own proposal is also for a bivalent truth-conditional

semantics. He assumes that for any particular occasion of utterance

of a number approximation, an exact cut-off point for the interval of

possibilities can be stated, drawing on relevant contextual factors

(as he defines them,, see below). There is thence a notion of

"appropriate approximation" in a given context. Thus a particular

occasion of use of a number approximation will be amenable to a

bivalent treatment. His proposal consists of stipulating a set of

205
 Chapter 7

rounding functions F, which map from exact numbers to exemplar numbers

(from Real Numbers to Real Numbers). In addition there is stipulated

R, a one to one function from C, the set of contexts, into F. "We can

thus define the notion of some number being an appropriate round

number for some other number in some context. "(: 206). In this way the

statement of the- truth conditions for utterance of any particular

number approximation will be determined by its particular context:

"I shall talk [. in terms of a number being an

. .)
appropriate round number in context c, where the notion of
context is intended to cover all the factors that determine
whether a given number is a suitable round number or not,
including questions of culture, beliefs, and knowledge of the
world, as well as the most important factor of the purpose of
the estimate"

I am not certain that stipulating semantic functions really fulfils

the claim of defining, except within a truth-conditional theory which

is in turn defined by stipulation. Allwood rehearses in some detail

how it is that a theoretical account where anything can be stipulated

is empirically not enlightening.

Wachtel's account works well, in its own terms, and he showed in

Wachtel (1981) that his account could be extended to deal with other

types of number approximations, notably n or m and n or so.

I think also that an account similar to his could be devised for

the vague tags, which show the same context-sensitive acceptability.

In their case we may apply the notion of "appropriate exemplification"

to the relationship between the stated exemplar and the set of

possibilities it can be used, with its tag, to identify. Thus,

discussing (4), car is an appropriate exemplar for bus in the

particular context most likely to have been supplied by the test

informants.

206
 Chapter 7

To formalise this we will need a set of exemplification functions

E, which map from entity to entity within the universal set. So

E {e: (x, y) [e(x) [3S: xeS A

sy --> yeS)j}

where x and y are variables and S is any proper subset of the

Universal set. Thus, in a particular context, e(bus) - car

As I have said, vague tags share with number approximations the

property of being sensitive to different contexts for their meanings.

It is possible therefore, along the lines proposed by Wachtel, to let

again be the set of contexts, and stipulate, a one to one function

.C
from C into E (the set of exemplification functions). I shall call

this C. On these lines, somewhat crudely, car is an appropriate

exemplar for an actual bus in a context c iff G(c)(bus) - car. Note

that packed up into function G will be all we have observed about the

structure of categories being relevant to what is considered a good

exemplar, and all we have observed about contextual factors like

intentions and relationships between speaker and hearer.

This account of vague tags predicts that:

(10) It wasn't a car or something that splashed you, it was a car

and

(10a) They haven't got a car and that, they've got a car

will be contradictions, which indeed they are (cf Wachtel 1980: 207).

The important thing to notice about this general approach, and

what I shall argue is one of its serious weaknesses, is that it. says

that on a particular occasion of utterance of one of these vague

expressions, exact cut-off points for possibilities are present. This

is necessary in order to render the apparent vagueness amenable to

two-valued truth conditions.

207
 Chapter 7

Returning to the data on vague language use which I have

presented in Chapters 3,4 and 5, I am persuaded that there is really

no empirical basis for the contention that the vagueness of these

expressions is resolved in particular contexts. On the contrary, the

evidence rather points the other way. To take Sadock's formulation,

these vague expressions are "purposely and unabashedly inaccurate

statements" (: 434). Informants say that using vague expressions

arises from not knowing exactly, or not wanting to say exactly.

Hearers react to vague expressions in particular ways, as we saw in

Chapter S.

To give an account of vagueness via bivalent truth-conditional

semantics is to make the fundamental theoretical error eloquently

described by Zadeh (1975):

"In
our quest for precision, we have attempted to fit the
real world to mathematical models that make no-provision for
fuzziness. We have tried to describe the laws governing the
behaviour of humans, both singly and in groups, in
mathematical terms similar to those employed in the analysis
of inanimate systems. This, in my view, has been and will
continue to be a misdirected effort, comparable to our
long-forgotten searches for the perpetuum mobile and the
philosopher's stone. "

The second problem with this account is its extreme generality

which leads to counter-intuitive predictions. For the number

approximations, because it treats all numbers. as equally available for

use in approximations, equivalence is predicted:

(11) This is approximately one inch long

This is approximately 2.067 cms long

(1 inch = 2.067 ems)

which is clearly wrong, given all we know about differences between

round and non-round numbers. For the vague category identifiers, the

formulation given allows that anything can be an exemplar for

208
 Chapter 7

anything, so, in theory,

e(Taj Mahal) - 53

(although this would be unlikely to find an appropriate context)[1].

There is another reason why I am unhappy with these accounts of

vagueness. This concerns whether they are empirically testable.

Notice that the relationship between the semantics proposed for

approximations, and for vague tags, and actual observations of their

use, is embodied in the functions R and G respectively. Thus in

testing the semantics for number approximations, if we get a mismatch

between the semantics and appropriateness judgments in a given case,

this will be attributable to a failure to correctly specify the

attributes of context c. If it is in an informant test, it will be

because informants have supplied context in unpredictable ways. In

Channell (1981), I drew on the N-test results to criticise some of the

proposals in Wachtel (1980). Wachtel's general response to these (in

Wachtel, 1981) was as follows:

"I think a general point that should be made is that

elicitation tests can never provide conclusive evidence of
the semantics of the sentences involved. They can provide,
at most, indications. What the subject is asked to do is to
invent a plausible specific context for a given sentence.
That is, he or she must find an interpretation for the
sentence, and then specify a context in which the sentence
can have that interpretation. It is clear that preferred
interpretations will, in general, be selected over the other
interpretations. If semantics is considered to be the study
of preferred interpretations, then there is no problem. If,
however, semantics is considered to be the study of the
meaning of the sentence, and if "meaning" includes at least
preferred and non-preferred interpretations, then elicitation
tests of this type do not provide evidence of the required
type. "

This is worrying, because taken to its logical conclusion, it says

that semantic constructs do not necessarily have to be in accord with

observations of language behaviour. This appears to make them quite

1I am grateful to Thomas Baldwin for these observations.

209
 Chapter 7

unassailable, and therefore, on my view, unrevealing as hypotheses

about language behaviour. Allwood makes a similar point, as follows:

"Since extension and intension for natural language

expressions are mostly not stipulatively given, we face an
empirical problem. We have to discover the intensions and
extensions of natural language expressions in order to give
an account of the truth'conditions of sentences in natural
language. This is clear, since we are not dealing with
stipulation. The meanings of expressions must then be
determined empirically. However, once we have empirically
discovered what the meanings are, can we then not sit back in
our chairs and continue as in formal semantics with situation
independent meanings? I do not think this will be
...
possible" (: 179)

7.3.3 Many-Valued Truth-Conditional Semantics

This version of truth-conditional semantics also separates what is

semantic from what is pragmatic along the lines of a statement of the

conditions which must pertain for a given sentence to be truly

uttered. Lakoff (1972) offered an attempt to apply a many-valued

("fuzzy") logic to a range of natural language sentences which he felt

to be unamenable to treatment by bivalent truth conditions. One group

of sentences he looked at contained approximately (: 222):

(12) a Sam was surprised that he had approximately $10,000 in his

savings account

b Sam had approximtely $10,000 in his savings account

c Sam had $9,992 in his savings account

d Sam had $9,950 in his savings account

e Sam had $9,500 in his savings account

f Sam had $9,200 in his savings account

of which he says:

"a presupposes b. b is
But fuzzy - it depends on what counts
as an approximation to having $10,000 in one's savings
account. Suppose c were the case. Then I think b would be
true no matter what, and a would make perfect sense. If d
were the case, I think most people in most situations would
still want to say that b was true and that a made sense. If
e were true, the truth of b would become questionable. In
many situations b would have a high degree of truth given the
truth of e, and a would pretty much make sense. When we get

210
 Chapter 7

down to f, however, the degree of truth of b gets lower, and

it makes less sense to say a. And so on. "

Notice again that this account, although. it allows for degrees of

truth, will still, for any given occasion of utterance of a number

approximation, compute an exact truth value, so that again, here,

vagueness has been banished from the semantics.

It would in principle be possible with this approach too, t6

extend it to vague tags. It is rather clear that (4) would have

rather a low truth value if the "vehicle" in question is a camel, but

a high one if it is a bus, in the sorts of contexts deployed by the

T-test subjects.

Sadock (1977) devotes a considerable amount of his paper to

demonstrating that Lakoff's informally expressed proposals are

actually impossible to put into practice. It is not necessary to

reproduce the discussion here - effectively he shows that the large

number and variety of contextual factors which must be taken into

consideration renders the proposed semantics unattainable. He argues

also that its sheer complexity is counterintuitive to .the language

user's impression of what an approxiation is.

I think there are two grounds for criticising the fuzzy truth

conditions approach.

The first is that it fails on its own terms in coming up with the

wrong predictions about entailments and the effect of negation.

(13) Sam is not approximately six feet tall - he's exactly six feet

tall

As far as I can judge, this is a problem, because only two values can

be involved - negation is not fuzzy. The non-negated version of the

first part will have a high truth value of 1, because Sam actually is

211
 Chapter 7

six feet tall. Negating it makes it false to say that Sam is

approximately six feet tall, but since he is six feet tall, this must

be wrong, and make (13) a contradiction, which it is not. The

'escape' from this is the same one described in the previous section,

of seeing this as 'mention' rather that 'use'.

The other difficulty is the entailment relationship between

(14) Sam is exactly six feet tall

and

(15) Sam is approximately six feet tall

On a Lakoff account, the first sentence will get t-value 1 iff Sam is

six feet tall. So will the second. This gives them the same truth

conditions, or makes them semantically equivalent. Patently, they are

not equivalent, so this is again "counterintuitive" and quite in

disagreement with the N-test results.

Apart from these failures within its own theoretical framework, I

want to argue that this is in general a mistaken approach. However, I

must make clear that I see much of what Lakoff achieved in his work as

highly valuable. The fuzzy approach has a lot of intuitive appeal, it

seems to reflect the kinds of judgments language users can make,

especially about the 'hedge' phenomena which Lakoff draws attention

to. Where it goes wrong is in insisting on a clear division into what

is semantic and what is pragmatic, which in turn leads firstly to an

empirically unfalsifiable semantics, and secondly to a situation where

most of what is interesting about approximation is bundled aside into

a vaguely defined pragmatics where it can be, in Allwood's phrase

"safely left for another day".

212
 Chapter 7

7.3.4 Cognitive/referential vs emotive/non-referential meaning

Another possibility mentioned by Allwood (: 188) is to take the view

that cognitive = referential = semantic, and emotive - non-referential

= pragmatic. This would give a good account of, for example, verb

pairs like put in/insert or enter/go in which are referentially

identical but exhibit differences of contextual suitability (ie

non-referential differences). Applying this approach to vague

expressions is problematical in the same way that it is for some other

examples discussed by Allwood. It is not clear which aspects are

cognitive/referential and which are not. In addition, as we have

seen, vague expressions refer differently on different occasions of

use, so this might be like taking the 'union of all uses' approach,

which has already been found wanting.

7.3.5 Normative vs Descriptive

The last way of distinguishing semantics and pragmatics which Allwood

considers, and also the last one I shall consider, is the idea of

making the distinction in the way Carnap originally formulated it:

"Semantics would then be concerned with the normative

analysis and explication of concepts that are used in many
different and partly inconsistent ways in ordinary language.
Pragmatics on the other hand would be concerned with the full
richness of empirical data. "

In many ways this appears to be close to the distinction argued for by

Wachtel. He suggests (personal communication) that investigating

meaning as I have done herein, is to do "psychopragmatics" rather than

to do "semantics". Allwood explains that the problem with the above

approach is that it is too close to the usual distinction between

theory and data. Theory arises from normative idealisations of data -

these would appear to be semantics. Thus pragmatics would be a

non-theoretical enterprise, orienting towards the normative

213
 Chapter 7

idealisations called semantics.

'7.4 Meaning, Semantics and Pragmatics

The most natural conclusion which arises from the foregoing discussion
,
of the data on vagueness is that no valid ways of seeking a

distinction between semantics and pragmatics are forthcoming in the

analysis of vague meaning. At the same time, the discussion provides

further evidence that there are no valid empirically-based reasons for

seeking to make the distinction. This leads to the same conclusion as

that of Allwood, quoted at the beginning of this Chapter.

If, on the other hand, a semantics /pragmatics distinction is held

to, then the consequence is that the semantic account will be forced

to admit of vagueness, rather than shunting it off into pragmatics,

since there appear to be no empirical reasons but only a theoretical

preference, for making semantics exact. In this connection, Danell

(1978: 16-17) concludes that because vagueness is inherent in the way

language is processed, the goal of precise semantic representations is

"a forlorn one".

214
 Chapter 8

Chapter 8

Implications for the Study of Meaning

It is by now apparent that this study of vague language has raised far

more questions than it has been able to propose answers to. This is,

however, exactly what we would expect of an investigation which takes

on a relatively unworked area, and which does, so with a methodological

insistence that attested data, with all their idiosyncracies and

untidinesses, should be used and given priority wherever possible.

Yet certain answers have been arrived at. I shall first summarise the

main ones, and then make some concluding remarks on their future

implications.

8.1 Summary of Findings

1 Vague language forms a considerable part of conversations. Only a

I

small sample of data -a few hours of conversations - produces many

examples. This means we cannot, in any theory of language, treat it

as the exception rather than the rule.

2 For the vague additives which I have described, the informant tests

show that hearers assign meanings to them which are fuzzy sets,

identified by the exemplar given. It is likely that this fuzzy set

meaning is also assigned to very many other vague expressions. I give

in Appendix 1a list of attested expressions which I think have a

similar fuzzy set meaning.

3 From the data and analysis in Chapter 5, we know that certain vague

expressions are used for particular and diverse conversational

reasons, all of which are, however, unified by the common feature of

uncertainty for at least some participant in the discourse (taken in

its broadest sense).

215
 Chapter 8

4 The relationship between constituent lexical items, and meanings

attaching to complete vague expressions, reinforces a view of language

in which lexical units of varying degrees of cohesion must be

recognised.

5 For these. expressions, it is apparently impossible to describe

their meanings independently of consideration of context and

inference.

8.2 Psycholinguistic/Psychological Implications

Some language is vague, and we understand it as vague and assign vague

meanings to it. I cited Zadeh, in Chapter 7, arguing that behavioural

phenomena are better dealt with by a model which incorporates

vagueness. Here Guilbaud, a mathematician, argues the necessity of

vagueness for communication, even for life:

". take a newspaper. They are full of figures today:

..
most of the figures are given implicitly or explicitly, with
a qualification as to the degree of certainty. The
population of a town, an industrial output, a distance, a
temperature, a duration, a speed, a percentage, etc. How are
we to find our way about? An approximate value, as common
sense says, is that which is not exact. Is it a lie then? I
would not deny that newspapers sometimes contain lies. But
there is not always so much malice. Not Talking and
thinking by means of 'about', 'nearly' is anecessity. "
(1977: 126)

Vagueness in the relationship between language and the situations

it relates to is endemic. As Guilbaud points out, it is not absent

even from contexts where we might expect exactitude: science,

economics, medicine.

Yet this should be just as we would expect, if we consider the

nature of the relationship between language and the world it describes

in its most general aspects. The "direction of fit" of language and

the world is most often that language is required to fit the world.

And the world makes demands on language in two different ways. First

216
 Chapter 8

it is, as we know, constantly changing and developing, making quite

new demands on language which is to reflect it. For language to be

fully useful therefore, in the sense of being able to describe all of

man's experience, it must incorporate built-in flexibility. This

flexibility resides in part in its capacity for vagueness. Secondly,

as noted by commentators whom I mentioned in Chapter 2, many things in

the world which language reflects are themselves vague, or perceived

by us as vague, subject as our perception is to physiological and

neurological constraints. Given, therefore, the direction of fit

mentioned above, language cannot help but incorporate vagueness.

8.3 Consequences for Linguistic Theory

5 above calls into question the notion, fundamental to linguistic

semantics, of decontextualised meaning, and supports the point put

thus by Lyons:

"the notion of sentence meaning is arguably dependent, both

logically and methodologically, upon the notion of utterance
meaning, so that one cannot give a full account of
sentence-meaning without relating sentences, in principle, to
their possible, contexts of utterance" (1981a: 140)

I would want to go further, and argue that if the above is true, then

the goal of describing sentence meaning is no goal at all, since

sentences and sentence-meaning cannot in principle be isolated as

objects of description.
8.4 An Approach to Meaning

An appropriate way to end a piece of work which has tended to be

negative, at least in its implications for existing linguistic

approaches to meaning, is to offer some ways forward. I conclude

therefore by giving a programmatic sketch of a possible fresh approach

to the theory and methodology of the study of meaning.

217
 Chapter 8

The clearest point which emerges from this study is the

inextricable relationship between contexts and understanding. I would

therefore see the methodology of the study of meaning as one that

works back from pairings of contexts with (hearer) assigned meanings.

The model I see would have a lexical list - but a lexical list

containing rather more multiword units than are usually envisioned.

Each entry would need a multiplicity of network relations

(connections). These would be, on the one hand, to other lexical

entries - paradigmatic relationships like apple/orange; class

inclusion relationships, parent/mother, and syntagmatic relationships,

university/student. On the other hand, many relationships would be to

knowledge constructs known by a hearer to connect to lexical entries.

Here would be incorporated all the information about use. Evens et al

(1978) proposed a network lexicon for their computer model for

understanding story texts. This would be an extension of their

proposal.

It may be objected that this model would be a ridiculous goal

since it necessitates taking account of all idiosyncratic connections

a hearer might just happen to make. Clark (1978) shows why this is

not so. Although he is writing about psychological models of

comprehension, his point is equally valid for a linguistic model of

language understanding. His argument runs like this. Although a

hearer may, on hearing any utterance, (his example is he's crazy) make

a bizarre association, such as having a mad Uncle Harry, the hearer

would also set aside that association, as not relevant. The reason

for doing this is that his strategy in comprehension is to make the

connections he thinks the speaker intends him to make. Hearers'

judgments of speakers' intentions are thus the base line of this

218
 Chapter 8

theory. The consequence for the linguistic description of the

utterance of "he's crazy" is that it just takes account of the

relevant relationships for crazy.

In this model then, there would be a relational network lexicon,

linked directly with non -lexical knowledge where appropriate, and

rules of comprehension. A representation of the meaning of "We've got

five or, six of them" (example 3 in the last chapter) would thence draw

on the lexical listing of n or m which might have a network relation

"approximation device" liking it to all the other approximation

devices. These in turn would have links to the knowledge that

approximation suggested not knowing or withholding. Note that these

links would often embody information described as implicatures in

other theories of meaning.

The network links associated with a linguistic expression have

the advantage of being empirically discoverable. Yorick Wilkes

(personal communication) has suggested to me one possible test, where

overt context is supplied and hearers asked to react within it. For

example:

1A paint shop manager instructs his stock controller to

reorder white paint when there are about 50 x1 litre cans
left on the shelf. At a Monday stocktake, there are 60 cans
left. Should he reorder, or not?

2A Guardian report on the South Sandwich Group, states that

it consists of "about 13 islands". If you subsequently went
there, how many islands would you expect to find?

Notice that a test like this would test my contention that such

quantities are always understood as vague. If all test subjects gave

the same precise answers, the cut-off point semantics proposed by

Wachtel would be in a stronger position.

219
 Chapter 8

A model like this will give a good account of the observed

phenomenon of hearer recontextualisation in tests; ie it makes

explicit the machinery they draw on to do it.

Any account on these lines is then empirically testable (a)

against hearers' reactions in experimentally restricted contexts, and

(b) against observations of use, but importantly, not against invented

examples.

The approach I have sketched here has one large advantage, as I

see it, counterbalanced by one large disadvantage. The advantage is

that it strengthens the linguistic investigation of meaning by

bringing it much closer to empirically based scientific procedure, and

taking it away from the weaker position of a stipulatively defined

theoretical model.

Its great disadvantage is that it makes what is seen as the scope

of the linguist's investigation of meaning very much wider than it is

at present, when his principal goal is the description of

decontextualised meaning. The data-base for any study would have to

be large, and would be, in sheer practical terms, difficult to

assemble (cf my points in Chapter 1). It would no longer be possible

to do armchair semantics on invented sentences.

Despite the obvious difficulties, I very much hope to see the

linguistic investigation of meaning moving in the direction I have

outlined, and I offer the work in this thesis as evidence for the

desirability of that approach.

220
 Appendix

Annendix: Othe lions

Number approximations:

The actual things were like two and six (Stop the Week, 14.2.81)

ranged from two LQ four pence...

....
the best thing to do I think is to start by getting a piece which
...
is longer than the length you need and measuring that, and put a
couple of bends, a couple of kinks in the wires at the two distances
that you need, maybe metre, ok, that way then you can position...
(physics lesson, Camb 2lÄ 60)
There are adjustments which have to be made for national tastes. Though
roughly eighty per cent is shared by all countries, the differences
born of experience are growing (article on Habitat catalogue, ST, 23.8.81)
Hear is the favourite round about evens, Peter Shore 2k to one, te
or thereabouts and Silkin 3/ to one and Michael Foot 4A or 5 but he
won't be that for very long because I expect to get a lot of money on
him (labour MP interviewed about Labour leadership race, 15.10.80)

... some thirty students

Some six years ago about twenty women came together to form what looked
like the beginnings of a Jewish women's liberation movement (ST, 23.8.81)
barely thirty
hardly thirty

are closer to fourteen percent

nearly n

almost eight million

Outside the plant contamination is virtually zero (BBC interview with
nuclear physicist on radiation leak at Windscale, 30.3.79)

Pretax profits were less than a quarter of a million (The Financial

World Tonight)

Just over a billion pounds last year (" )

More than a million better ( )
pounds
Over more than ten years' production, its style has changed, grown
up and diversified (Habitat catalogue, ST 23.8.81)

Or so. An extension of the n or so use is or an with month names:

It was in February or March or so (describing when he received a letter)

Last January or so, or December maybe (wife went for an interview)

All but. When used with an adjective, this is a vague additive, but

when it is with a number, it is not an approximation, eg all but three

of the crew, vs, all but ready

221
 Appendix

Vague category identifiers:

It's things like throat lozenges pain killers, creams for arthritis,
.U
things like that (You and Yours, 19.2.81)

affected by 1+ngs like back pay and late returns...

...
x and the like

x and so on
He argues that the kid has worked out ways of referring to things
like that and that it is a language... (tutorial on child
and stuff
language)
A doctor noting that his patient has a temperature and so forth is
said to diagnose his disease as influenza (Ogden and Richardq, 1923: 21)

your stuffed fish or whatever, you're making a fetish of originality

...
(Stop the Week, 14.2.81)
("baned") Bound was what most men would have said but. this one came
from Oxfordshire or somewhere (Amis, K, 1978 Jakes Thing: 104)

hereabouts

whereabouts

Sort of and kind of:

A thing like that, like a back-up supervisor will hardly ever be used
but it might help some of the problem cases, that's what I mean, its
formalising (I1,21.2)
a kind of something without restricting anything
Also I've noticed sort of in some of my friends that the ones who had
older brothers and sisters didn't learn to speak as quickly as their
elder brothers and sisters (11,8.3)

You can do all sorts of things on this TV its pretty good actually how
if they connect up sort of the local bookmaker and things you know its
its quite good idea well its not a bad system really quite uh its a
bit sort of ferry built uh (Dicks, 1974)
("What happens
to the other people? ") Well they're making general
sort of sounds as though they're in agreement (N-test, experiment
discussion, 11.6.79)
I did, I mean, I just sort of worried about it (see Extract 2, Chapter 5)
A: Sue is a geneticist
Sue: Of sorts, yes (I think this works like sort of)

222
 REFERENCES

Allan K. (1981) 'Interpreting from context', Lingua, 53,2/3:

151-174.

Allwood J. (1981) 'On the distinction between Semantics and

, in Klein W. and W. Levelt
Pragmatics' : 177.189,
Crossing the Boundaries of Linguistics,
Dordrecht : Reidel.

Anderson M. (1971) The Grammar of Case Cambridge : C. U. P.

,

Baker C. (1975) 'This is just a first approximation, but',

Chicago Linguistic Society, Papers, 11 : 37-47.

Bar-Hillel Y. (1971) 'Out of the Pragmatic Wastebasket', Linguistic

Inquiry, 2 : 401-7.

Becker J. D. (1975) 'The Phrasal Lexicon' in Nash-Webber B. and

R. Schank (eds), Theoretical Issues in Natural
Language. Processing, Cambridgeb Ma. :
Bolt Beranek Newman.

Bennett D. C. (1975) Spatial. and Temporal Uses of English

Prepositions London : Longman.
,

Bernstein T. M. (1971) Miss Thistlebottom's Hobgoblins New York:

,
Farrar, Straus and Giroux.

Binnick R. I. (1970) 'Ambiguity and Vagueness', Chicago Linguistic

Society, Papers, 6 : 147-153.
.
Bolinger D. (1965) 'Atomisation of Meaning', Language, 41,4 : 555-572.

Bolinger D. (1976) 'Meaning. and Memory', Forum Linguisticum 1,

1: 1-14.

Bolinger D. (1977) Meaning and Form London : Longman.

,

Bolinger D. (1979) 'Couple: An English Dual' in Greenbaum S.,

G. Leech and J. Svartvick : Studies in English
Linguistics: for Randolph Quirk, London : Longman.

Bransford J. D. 'Contextual Prerequisites for Understanding:

(1972) Some investigations of Comprehension and Recall 't
and,
M. K. Johnson Journal of Verbal Learning and Verbal Behavior,
11 : 717-26.

Bresnan J. (1978) 'A realistic transformational grammar' in Halle,

Bresnan and Miller

Brotherton P. (1976) 'Aspects of the Relationship between Speech

Production, Hesitation behaviour and Social Class'
unpublished Doctoral Thesis, University of
Melbourne.

223
Brown J. (1979) 'Vocabulary : learning to be imprecise', Modern
English Teacher, 7 : 1,25-27.

Brown R. (1973) A First Language Cambridge, Ma., Harvard

,
University Press.

Bruner, (1973) Beyond the Information Given: Studies in the

Jerome S. Psychology of Knowing New York : Norton and
,
Company.

Carroll J. M., (1981) 'The non-uniqueness of Linguistic Intuitions'

T. G. Bever and Language, 57,2 : 368-382.
C. R. Pollack

Carter A. (1980) 'The Language of Sisterhood' in Michaels L. and

C. Ricks, The State of the Language, London
University of California Press.

Channell J. (1980) 'More on Approximations', Journal of Pragmatics,

4: 461-476.

Channell J., (1981) 'Participant'Roles and Lexical Relations',

A. P. Cowie and Paper read at L. A. G. B. Autumn Meeting.
L. Jeffries

Chomsky N. (1965) Aspects of the Theory of Syntax Cambridge, Ma.,

,
M. I. T. Press.

Chomsky N. (1969) 'Form and meaning in natural language' in John

D. Roslansky (ed), Communication, Amsterdam :
North Holland.

Cicourel A. (1974) 'Interviewing and Memory' in Cherry C. (ed),

Pragmatic Aspects of Human Communication,
Dordrecht: D. Reidel.

Clark D. D. (1981) 'Orders of Approximation to English dialogue',

Language and Communication 1,2/3 : 207-236.

Clark H. H. (1978) 'Inferring what is meant', in W. J. M. Levelt and

G. B. R. d'Arcais, Studies in the perception of
language, Chichester : Wiley.

Cogen C.. and (1975) 'Interactions of the Expression "Let's just say"
L. Herrman with the Gricean Maxims of conversation',
Berkeley Linguistic Society, 1: 60-67.

Cohen B. (1971) 'The Logical Particles of Natural Language' in

Bar-Hillel Y. (ed), Pragmatics of Natural
Language, Dordrecht : Reidel.

Corsaro W.A. (1981) 'Communicative Process in studies of social

organisation', Text, 1 : 5-63.

Cotton J. W. and (1978) Semantic Factors in Cognition, Hillsdale

R. L. Klatzky Erlbaum.

224
Coulthard M. (1977) An Introduction to Discourse Analysis, London:
Longman.

Cowie A. P. (1981) 'The treatment of Collocations and Idioms in

Learners' Dictionaries', Applied Linguistics,
2,3 : 112-124.

Cruse D. A. (1977) 'The Pragmatics of Lexical Specificity', Journal

of Linguistics, 13 : 153-164.

Cruse D. A. (1982) 'On Lexical Ambiguity', Nottingham Linguistics

Circle.

Crystal D. (1969) Prosodic Systems and Intonation in English,

Cambridge : C. U. P.

Crystal D. and (1975) Advanced Conversational English, London:

D. Davy Longman.

Daitz E. (1956) *'The picture theory of meaning' in Flew A. (ed)

Essays-in Conceptual Analysis, London : Macmillan.

Danell K. S. (1978) 'The concept-of vagueness in linguistics. Some

methodological reflections of a non"apeciali'st',
Studia Neophilologica, 50 : 3.24.

Deese J. (1965) Structure of Associations in Language and

Thought, Baltimore : John Hopkins Press.

Deese J. (1974) 'Towards a Psychological Theory of the Meaning of

Sentences' in Silverstein A. (ed), Human
Communication : Theoretical Explorations, 'Lawrence
Erlbaum Assoc. Hillsdale : New Jersey.

Dines E. R. (1980) 'Variation in Discourse "and stuff like that" ',

-
Language in Society, 9 : 13-31.

Dixon R. M. W. (1971) 'A method of semantic description' in Steinberg

and Jakobovits (eds).

Duncan S. and (1977) Face to Face Interaction, New York : Lawrence

D. W. Fiske Erlbaum.

Evens M. W. and (1978) 'A lexicon for a computer question-answering

R. N. Smith -system', American Journal of Computational
Linguistics, 15,4 : 1-93.

Fauconnier G. (1976) 'Remarque sur la theorie des phenomenes

scalaires', Semantikos I (3) : 13-36.

Fillenbaum S. (1974) 'Or: some uses', Journal of Experimental

Psychology, 103 : 913-921.

Fillenbaum S. (1978) 'How to do some things with IF' in Cotton and

Klatzky.

Fillmore C. J. (1968) 'Lexical Entries for Verbs', Foundations of

Language, 4 : 373-393.

225
Fillmore C. J. (1971) 'Types of Lexical Information' in Steinberg and
Jakobovits (eds).

Fillmore C. J. (1977) 'Topics in Lexical Semantics' in. Cole R. W. (ed)

Current Issues in Linguistic Theory, Bloomington:
Indiana University Press.

Fine K. (1975) 'Vagueness, Truth and Logic', Synthese, 30 :

265-300.

Fodor J. A. (1976) The Language of Thought Hassocks : Harvester.

,

Fodor J. D. (1977) Semantics: Theories of meaning in Generative

Grammar Hassocks : Harvester.
,

Fowler R. and (1979) 'Rules and Regulations' in Fowler R., B. iiodge,

G. Kress G. Kress and T. Trew, Language and Control,
London : R. K. P.

Fraser B. (1970) 'Idioms within a transformational Grammar',

Foundations. of Language, 6 : 22-42.

Freidin R. (1975) 'The Analysis of Passives', Language, 51,2

384-405.

Gazdar G. (1979) Pragmatics: Implicature, Presupposition and

Logical Form London : Academic Press.
,

Greenbaum S. (1977) 'Contextual Influence on acceptability Judgments',

Linguistics, -187 : 5-11.

Greenbaum S. (1970) Elicitation Experiments in English'j London:

and R. Quirk Longman.

Grice H. P. (1967) 'Logic and Conversation' printed 1975 in Cole P.

and J. L. Morgan (eds) Syntax and Semantics,
vol 3: Speech Acts, New York : Academic Press.

trice H. P. (1968) 'Utterer's meaning,. sentence meaning and word

meaning', Foundations of Language, 4,1-18
reprinted in Searle J. R. (1971) Philosophy of
Language.

Grice H. P. (1969) 'Utterer's Meaning and Intentions' Philosophical

Review, 78 : 147-177

Grice H. P. (1978) 'Further Notes on Logic and Conversation' in

Cole P. (ed) Syntax and Semantics, vol 9
Pragmatics, New York: Academic Press.
vt,
ýilbaud G. Th. (1977) 'Mathematics. and Approximation' in Asher H. and
H. Kunle (eds) Proceedings Third International
Congress on Mathematics Education.

Halle M., (1978) Linguistic Theory and Psychological Reality

J. Bresnan and Cambridge, Ma., M. I. T. Press.
G. A. Miller

226
Harris G., (1980) 'On the Role of the Speaker's Expectations in
Ian Begg and Interpersonal Communication', Journal of Verbal
D. Upfold Learning and Verbal Behavior, 13,5 : 597-607.

Harries, Helga (1973) 'Coordination reduction', Working Papers on

Language Universals, 11 : 139-209, Stanford.

Hudson R. A. (1975) 'The Meaning of Questions', Language, 51,1 : 1-31.

Hurford J. R. (1974) 'Exclusive or inclusive Disjunction', Foundations

of Language, 11 : 409-411.

Hurford J. R. (1975) The Linguistic Theory of Numerals Cambridge :

,
C. U. P.

Jeffries L. (1981) 'Intonation of Idioms' paper given at Leeds

International Conference on Spoken English.

Jespersen 0. (1924) Philosophy-of Grammar $ London : Allan and Unwin.

Johansson S. (1980) 'Word Frequencies in British and American English

Some Preliminary Observations' in Allwood J. and
M. L. Jung (eds) ALVAR : Stockholm Papers in
Language and Literature.

Kasher A. (1977) 'What is a theory of use? ', Journal of

Prarmatics, 1 : 105.120.

Katz J. J. and (1964) 'The structure of a semantic theory' in Fodor J. A.

J. A. Fodor and J. Katz, The Structure of Language, Readings
in the Philosophy of Language, Englewood L s,
New Jersey : Prentice-Hall.

Kempson R. M. (1975) Presupposition and the delimitation of

semantics, Cambridge : C. U. P.

Kempson R. M. (1977) Semantic Theory Cambridge : C. U. P.

,
Kempson R. M. (1979) 'Ambiguity and Word Meaning' in Greenbaum S.,
G. Leech and J. Svartvik, Studies in English
Linguistics : for Randolph Quirk. London : Longman.

Kiparsky P. (1968) 'Linguistic Universale and Linguistic Change' in

Bach E. and R. T. Harms (eds) Universals in
Linguistic Theory, New York: Holt, Rhinehart and
Winston.

Klatsky R. L. (1978) 'Semantic Information and Visual Information

and. A. M. Stoy processing' in Cotton and Klatsky.

Klein E. H. (1980) 'A Semantics for Positive and Comparative

Adjectives', Linguistics and Philosophy, 4 : 1-45.

Klein E. H. (1982) 'The Interpretation of Adjectival Comparatives',

Journal of Linguistics, 18 : 113-136.

Kosslyn S. M. (1978) 'Imagery and internal representation' in Roach

and Lloyd.

227
Ladefoged P. (1972) 'Phonetic Prerequisites for a distinctive feature
theory' in Valdman A. (ed) Papers in Linguistics
and Phonetics to the Memory of Pierre Delattre,
The Hague : Mouton.

Lakoff G. (1970) 'A note on vagueness and ambiguity', Linguistic

Inquiry, 1: 357-359.

Lakoff G. (1971) 'On Generative Semantics' in Steinberg and

Jakobovits.

Lakoff'G. (1972) 'Hedges and Logic of fuzzy concepts', Chicago

Linguistic Society, Papers, 8 : 183-228.
.
Lakoff R. (1971) "Ifs, and's and büt's about conjunction' in
Fillmore C. J. and Langendoen (eds) Studies in
Linguistic. Semantics, New York: Holt, Rh niheart
and Winston.

Lanin I. (1977) 'You can take the sentence out of the discourse
but you can't take the discourse out of the mind
of the speaker', Chicago Linguistic Society, 13 :
288-301.

Leech G. (1970) 'On the theory and practise of semantic testing'.

Lingua, 24 : 343-364.

Lehrer A. (1974) Semantic Fields and Lexical Structure, London

North Holland.

Lehrer A. (1975) 'Talking about Wine', Language, 51,4 : 901-923.

Lehrer A. (1978) 'Structures of the Lexicon and Transfer of

Meaning', Lingua, 45 : 95-123.

Levelt W.J. M. (1970) 'Introduction to Psychological Studies of

Grammar' in Levelt W.J. M. and G. D. F. d'Arcais
(eds), Advances in Psycholinguistics,
Amsterdam : North Holland.

Levelt W. J. M. (1974) Formal Grammara and Psycholinguistics, 3 vole,

The Hague
: Mouton.

Levinson S. (1979) 'Pragmatics and Social Deixis: reclaiming the

notion of conventional implicature', Berkeley
Linguistics Society, Papers, 5 : 206-223.

(1978) Longman Dictionary of Contemporary English,

London : Longman.

Lyons J. (1977) Semantics Cambridge : C. U. P.

,
Lyons J. (1981a) Language and Linguistics Cambridge : C. U. P.
,

Lyons J. (1981b) Language, Meaning and Context London : Fontana.

,

Lyons J. (1982) 'Deixis and Subjectivity: loquor, ergo sum? ' in

Jarvella R. J. and W. Klein (eds), Speech, Place
and Action, London : Wiley.

228
 Malinowski B. (1923) 'The problem of meaning in primitive languages'
in Ogden C. K. and I. A. Richards The Meaning of
Meaning, London : R. K. P.

Menninger K. (1969) (Translated from the German edition of 1958)

Number Words and Number Symbols, Cambridge, Ma: M. I. T.

Mikkel Blakar R. (1975) 'Utterances in vacuo and in context',

/
R. Rommetveit Linguistics, 153 : 5-32.

Miller G. and (1976) Language and Perception Cambridge, Ma. :

,
P. N. Johnson-Laird Belknap at Harvard-University Press.

Milroy L. (To appear 1983) 'Comprehension and Context;

Successful Communication and Communicative
Breakdown' in P. Trudgill (ed), Applications of
Sociolinguistics, London : Academic Press.

Mittwoch A. (1982) 'On the Difference between Eating and Eating

Something', Linguistic Inquiry, 13,1 : 113-121.

G. (1979) 'The non-uniqueness of semantic solutions:

-Nunberg
polysemy', Linguistics and Philosophy, 3 : 143-184.

Oiler D. K. and (1975) 'Phonetic expectation and transcription validity',

R. E. Eilers Phonetica, 31 : 288-304.

Ortony A., (1978) 'Interpreting Metaphors and Idioms: Some Effects

D. L. Schallert, of Context on Comprehension', Journal of Verbal
R. E. Reynolds Learning and Verbal Behavior, 17 : 465-477.
and S. J. Antos

(1974) Oxford Advanced Learners' Dictionary, Oxford :

O. U. P.

Palmer E. (1975) 'Visual Perception and World Knowledge: Notes on

a Model of Sensory-Cognitive Interaction' in
Norman D. A. and D. E. Rumelhart (eds) Explorations
in Cognition, San Francisco : W. H. Freeman and Co.

Peirce C. S. (1902) 'Vagueness' in Baldwin M. (ed), Dictionary of

Philosophy and Psychology II, London: Macmillan.

Pettigrew T. F. (1958) 'The-measurement and correlates, of category width

as a cognitive variable', Journal of Personality,
26 : 532-544.

Pulman S. G. (To appear 1983) Word Meaning and Belief

London : Croom Helm.

Putnam H. (1975) Mind Language and Reality Philosophical Papers,

,
2, Cambridge : C. U. P.

Quine W. (1961) From a Logical Point of View Cambridge, Ma.:

,
Harvard University Press.

229
 Quirk R., (1972) A Grammar of Contemporary English,
S. Greenbaum, London : Longman.
G. Leech and
J. Svartvik

Ringen J. D. (1980) 'Linguistic facts:

a study of the empirical
scientific status of TGGs', in Perry T. A. (ed),
Evidence and Argumentation-in Linguistics,
Berlin : de Gruyter.

Rosch. E. (1973) 'On the internal structure of perceptual and

semantic categories' in T. E. Moore (ed)
Cognitive Development and the Acquisition of
,
Language, New York : Academic Press.

Rosch E. (1975a) 'Cognitive Representations of Semantic

Categories', Journal of Experimental Psychology,
104 : 192-233.

Rosch E. (1975b) 'Cognitive Reference Points', Cognitive

,ý- Psychology, 7 : 532-547.

Roach E. ' (1978) 'Principles of Categorisation' in Roach and Lloyd.

Roach E. and (1978) Cognition and Categorization, New York

B. Lloyd Lawrence Erlbaum.

Roach E. and (1975) 'Family Resemblances: Studies of the Internal

1-1 C. B. Mervis Structure of Categories', Cognitive Psychology,,
7: 573-605.

Roach E., (1976) 'Basic objects in natural categories', Cognitive

C. B. Mervis, Psychology, 8 : 382-439.
D. M. Johnson and
P. Boyes-Braem

Sadock J. M. (1977) 'Truth and approximations', Berkeley Linguistic

Society, Papers, 3 : 430.439.

Sadock J. M. (1978) 'On testing for conversational implicature' in

Cole P. (ed), Syntax and Semantics, vol 9
Pragmatics, New York : Academic Press.

Sankoff D., (1978) 'Semantic field variability' in Sankoff D. (ed)

P. Thibault Linguistic Variation: Models and Methods,
and H. Berube New York : Academic Press.

Schank R. and (1974) 'The Goals of Linguistic Theory Revisited',

Y. Wilke Lingua, 34 : 301-326.

Schenkein J. D. (1978) 'Identity negotiations in conversation' in

Schenkein. (1978).

Schenkein J. D. (1978) 'Sketch of an analytic mentality for the study of

interaction' in Schenkein (1978).
conversational
Schenkein J. D. (1978) Studies in the organisation of conversational
(ed) interaction, New York : Academic Press.

230
Schmerling S. (1978) 'Asymmetric conjunction and rules of conversation'
in Cole P. and J. L. Morgan (eds) Syntax and
Semantics, vol 3 : Speech Acts, New York
Academic Press.

Schmidt, (1974) 'The relevance to semantic theory of a study of

Christine vagueness', Chicago Linguistic Society, Papers,
10 : 617-630.

Searle J. M. (1969) 'Metaphor' in Ortony A. (ed), Metaphor and

thought', Cambridge : C. U. P.

Smith N. V. - (1982) Mutual Knowledge London : Academic Press.

,

Smith N. V. and (1979) Modern Linguistics Harmondsworth : Penguin.

,
D. Wilson

Spencer N. J. (1973) 'Differences between linguists and non-linguists

in intuitions of grammatical acceptability',
Journal of Psycholinguistic! Research, 2 : 83-98.

Sperber D. and (1981) 'Irony and the use-mention distinction' in

D. Wilson P. Cole (ed), Radical Pragmatics, New York
Academic Press.

Steinberg D. and (1971) Semantics ,

Cambridge : C. U. P.
L. Jakobovits
(eds)

Swinney D. and (1979) 'The Access and Processing of Idiomatic

A. Cutler Expressions', Journal of Verbal Learning and
Verbal Behavior, 18(5) : 523-534.

Teeter K. V. (1964) 'Descriptive Linguistics in America: Triviality

versus Irrelevance', Words, 20: 197-206.

Ullman S. (1962) Semantics Oxford : Blackwell.

,

Wachtel T. (1980) 'Pragmatic Approximations', Journal of

Pragmatics, 4 : 201.211.

Wachtel T. (1981) 'Distinguishing between Approximations', Journal

of Pragmatics, 5 : 311-322.

Wason P. C. and (1972) Psychology of Reasoning

,
London : Batsford.
P. N. Johnson-Laird

Weinreich U. (1963) 'On the semantic structure of language' in

Greenberg J. H., Universals of Language,
Cambridge:

Weinreich U. (1966) 'Explorations in Semantic Theory' in

T. A. Sebeok (ed), Current Trends in Linguistics,
vol 3, The Hague: Moutan.

Weinreich U. (1969) 'Problems in the Analysis of Idioms' in J. Puhvel

(ed), Substance and Structure of Language,
Berkeley : University of California Press.

231
Weiser Ann (1974) 'Deliberate Ambiguity', Chicago Linguistic
Society, Papers, 10 : 723-731.

K. (1978) 'Comments on Fillenbaum's Paper' in Cotton and

Wexler
Klatzky.

Wierzbicka A. (1978) 'Syntax versus Semantics', Theoretical

Linguistics, 1: 115-132.

Wilks Y. (1977) 'Methodological questions about approaches to

Understanding Natural Language', Journal of
Pragmatics, -1 : 69-84.

Willis P. (1982) 'Participant Roles and Lexical Analysis',

and
L. Jeffries Nottingham Linguistic Circle.

Wilson, Deirdre (1975) Presupposition and Non-Truth-Conditional

Semantics, London : Academic Press.

Wilson D. and (1982). 'Mutual Knowledge and Relevance in Theories of

Comprehension' in N. V. Smith (ed).
D. Sperber

Wirth (1975) 'Logical considerations in the testing of

R. linguistic hypotheses' in D. Cohen and J. R. Wirth
Jessica
(eds) Testing Linguistic Hypotheses, Washington :
Hemisphere.

Wittgenstein L. (1958) Philosophical Investigations, Oxford : Basil

Blackwell.

Wootton A. (1975) Dilemmas'of Discourse, London : Allen and Unwin.

Wootton-A. (1981) 'Conversation Analysis'in P. French and

M. MacLure (eds) Adult-Child Conversation,
London: Croom Helm.

Zadeh L. A. (1975) Foreword to Kaufmann A. Introduction to the

Theory of Fuzzy Subsets, New York : Academic Press.

Zwicky A. and* (1975) 'Ambiguity Tests and how to fail them' in

J. Sadock J. Kimball (ed) Syntax and Semantics, 4,
New York : Academic Press.

232