Patrick Griffiths (Editor) - Chris Cummins (Editor) - An Introduction To English Semantics and Pragmatics-Edinburgh University Press (2016)

An Introduction to English Semantics and
Pragmatics
Edinburgh Textbooks on the English Language
General Editor
Heinz Giegerich, Professor Emeritus of English Linguistics, University of Edinburgh
Editorial Board
Heinz Giegerich, University of Edinburgh – General Editor
Laurie Bauer (University of Wellington)
Olga Fischer (University of Amsterdam)
Willem Hollmann (Lancaster University)
Marianne Hundt (University of Zurich)
Rochelle Lieber (University of New Hampshire)
Bettelou Los (University of Edinburgh)
Robert McColl Millar (University of Aberdeen)
Donka Minkova (UCLA)
Edgar Schneider (University of Regensburg)
Graeme Trousdale (University of Edinburgh)
titles in the series include:

An Introduction to English Syntax, Second Edition
Jim Miller
An Introduction to English Phonology
April McMahon
An Introduction to English Morphology: Words and Their Structure
Andrew Carstairs-McCarthy
An Introduction to International Varieties of English
Laurie Bauer
An Introduction to Middle English
Jeremy Smith and Simon Horobin
An Introduction to Old English, Revised Edition
Richard Hogg, with Rhona Alcorn
An Introduction to Early Modern English
Terttu Nevalainen
An Introduction to English Semantics and Pragmatics
Patrick Griffiths
An Introduction to English Sociolinguistics
Graeme Trousdale
An Introduction to Late Modern English
Ingrid Tieken-Boon van Ostade
An Introduction to Regional Englishes: Dialect Variation in England
Joan Beal
An Introduction to English Semantics and Pragmatics, Second Edition
Patrick Griffiths and Chris Cummins
An Introduction to English Phonetics, Second Edition
Richard Ogden
An Introduction to English Morphology, Second Edition
Andrew Carstairs-McCarthy
An Introduction to English Phonology, Second Edition
April McMahon
An Introduction to English Semantics and Pragmatics, Third Edition
Patrick Griffiths and Chris Cummins
Visit the Edinburgh Textbooks on the English Language website at www

.edinburghuniversitypress.com/series/ETOTEL
An Introduction to English
Semantics and Pragmatics
Third Edition
Patrick Griffiths
Revised by Chris Cummins
Edinburgh University Press is one of the leading university presses in the UK. We
publish academic books and journals in our selected subject areas across the humanities
and social sciences, combining cutting-edge scholarship with high editorial and
production values to produce academic works of lasting importance. For more
information visit our website: edinburghuniversitypress.com
© in the first edition of this work, Literary Estate of Patrick Griffiths, 2006
© in the revisions and additional third edition material, Chris Cummins, 2023
Cover design & illustration: riverdesignbooks.com
Edinburgh University Press Ltd

The Tun – Holyrood Road
12(2f) Jackson’s Entry
Edinburgh EH8 8PJ
Typeset in 10.5/12 Janson by

Cheshire Typesetting Ltd, Cuddington, Cheshire
and printed and bound in Great Britain
A CIP record for this book is available from the British Library
ISBN 978 1 3995 0460 7 (hardback)

ISBN 978 1 3995 0461 4 (paperback)
ISBN 978 1 3995 0462 1 (webready PDF)
ISBN 978 1 3995 0463 8 (epub)
The right of Patrick Griffiths and Chris Cummins to be identified as the authors of this
work has been asserted in accordance with the Copyright, Designs and Patents Act
1988, and the Copyright and Related Rights Regulations 2003 (SI No. 2498).
Contents
List of figures and tablesviii

Preface to the third editionix
1 Studying meaning 1
Overview 1
1.1 Sentences and utterances 4
1.2 Types of meaning 5
1.3 Semantics vs pragmatics 10
Summary 14
Exercises 14
2 Sense relations 16
Overview 16
2.1 Propositions and entailment 16
2.2 Compositionality 19
2.3 Synonymy 21
2.4 Complementarity, antonymy, converseness and
incompatibility23
2.5 Hyponymy 26
Summary 30
Exercises 30
Recommendations for reading 31
3 Nouns 32
Overview 32
3.1 The has-relation32
3.2 Count nouns and mass nouns 40
Summary 42
Exercises 42
vi an introduction to english semantics and pr agmatics
4 Adjectives 44
Overview 44
4.1 Gradability 44
4.2 Combining adjective meanings with noun meanings 47
Summary 52
Exercises 53
5 Verbs 55
Overview 55
5.1 Verb types and arguments 55
5.2 Causative verbs 57
5.3 Thematic relations 61
Summary 64
Exercises 64
6 Tense and aspect 66
Overview 66
6.1 Talking about events in time 66
6.2 Tense 69
6.3 Aspect 73
Summary 79
Exercises 79
7 Modality, scope and quantification 82
Overview 82
7.1 Modality 82
7.2 Semantic scope 87
7.3 Quantification 90
Summary 98
Exercises 98
8 Pragmatic inference 101
Overview 101
8.1 Some ways of conveying additional meanings 103
8.2 The Gricean maxims 106
8.3 Relevance Theory 112
8.4 Presuppositions 115
Summary 119
Exercises 119
contents vii
9 Figurative language 121

Overview 121
9.1 Literal and figurative usage 121
9.2 Metaphor 123
9.3 Metonymy 125
9.4 Simile 127
9.5 Irony 128
9.6 Hyperbole 129
Summary 131
Exercises 131
10 Utterances in context 133
Overview 133
10.1 Tailoring utterances to the audience 133
10.2 Definiteness 134
10.3 Given and new material 136
10.4 The Question Under Discussion 143
Summary 145
Exercises 146
11 Doing things with words 148
Overview 148
11.1 Speech acts 148
11.2 Indicators of speech acts 151
Summary 159
Exercises 159
Suggested answers to the exercises161

References175
Index179
List of figures and tables
Figures
2.1 An example of levels of hyponymy 28
2.2 Hyponym senses get successively more detailed as we go
down the tree 28
2.3 Part of the hyponym hierarchy of English nouns 29
4.1 The simplest case of an adjective modifying a noun is like
the intersection of sets 48
6.1 Time relations among the events described in (6.1). Time
runs from left to right: open-ended boxes indicate events
with no endpoint suggested 69
7.1 Venn diagrams for the meanings of (7.20a) and (7.20c) 92
Tables
3.1 Distinguishing between count and mass nouns 41
5.1 Examples of causative sentences with an entailment from
each59
5.2 Three kinds of one-clause causative with an entailment
from each 61
6.1 Two-part labels for tense–aspect combinations, with
examples69
6.2 The compatibility of some deictic adverbials with past,
present and future time 73
6.3 A range of sentences which all have habitual as a possible
interpretation74
10.1 A selection of indefinite and definite forms in English 137
viii
Preface to the third edition
When Laura Williamson from Edinburgh University Press asked me

whether I would like to produce a new edition of this textbook, my first
reaction was to be flattered. But, of course, what Laura said was rich in
possible interpretations. I still remember being confused by just this
kind of utterance at primary school, and failing to understand that the
teacher’s “Would you like to sit over there?” was actually an instruction
rather than just a slightly odd question.
With more experience of language under my belt, naturally I under-
stood Laura’s question as an indirect request. But, inevitably, I also won-
dered whether there was a shade of meaning along the lines of ‘we need
a new edition, because there is so much wrong with the existing edition’.
I still don’t know whether that meaning was intended – I didn’t dare
ask. But the opportunity to go back and revise a text like this is always
welcome, and of course I found all sorts of places where what struck me
as crystal clear six years earlier strikes me as a bit murky now. So I’ve
welcomed the chance to try to filter out some of these issues. I’ve also
attempted to plug a few gaps in coverage, while trying not to go on at
unwarranted length.
This book retains much of the structure of Patrick Griffiths’s original
edition, and draws on many of the examples he introduced. Therefore,
much of the credit goes to him, and to the people to whom he gener-
ously gave credit, notably Heinz Giegerich, Anthony Warner, Kenji
Ueda, and Janet and Jane Griffiths. For my part, I’d also like to thank
Ronnie Cann for his detailed comments on the second edition and
Hannah Rohde for her comments on and support with the third edition.
And, of course, thanks go to the team at Edinburgh University Press for
their encouragement, professionalism and, above all, patience.
ix
1 Studying meaning
Overview
This book is about how the English language allows people to convey
meanings. As the title suggests, semantics and pragmatics are the two
main branches of the linguistic study of meaning. Semantics is the
study of the relationship between units of language and their meaning.
Pragmatics is concerned with how we use language in communica-
tion, and so it involves the interaction of semantic knowledge with our
knowledge of the world, including the contexts in which we say things.
Explanations of terms
Numbers in bold print in the index point to the pages where
technical terms, such as semantics and pragmatics in the
paragraph above, are explained.
To illustrate some of the differences between semantics and prag-

matics, let’s consider example (1.1).
(1.1) That’s what I’m talking about!
Language enables us to communicate about the world. This works
because there are conventional links between expressions in language
and aspects of the world (real or imagined) that language users talk and
write about: things, activities, and so on. We sometimes use the word
denote to describe these links: I denotes the speaker, talking denotes
an activity, and so on. An expression is any meaningful language unit
or sequence of meaningful units, such as a sentence, a phrase, a word,
or a meaningful part of a word (such as the s of cats, which denotes
plurality, as opposed to the s of is, which doesn’t have its own separate
meaning).
1
2 an introduction to english semantics and pr agmatics
In (1.1), that denotes something that is assumed to be obvious to the

hearer in the current context of utterance: it might be an object that is
physically present, or it might be a linguistic expression that has just
been uttered. The expression what I’m talking about also denotes some-
thing, specifically the topic that the speaker has been discussing. And
the expression ’s, in (1.1), denotes a relationship of identity between
those two denotations. So one possible interpretation of (1.1) is that
something which is obvious to the hearer at the moment (that) corre-
sponds (’s) to the topic that the speaker was previously addressing (what
I’m talking about). But actually there’s another possible interpretation for
(1.1), because it also functions as a fixed idiomatic expression meaning
something like ‘That’s excellent’ or ‘That’s impressive’ or ‘That’s great
news’. In this case, that would still denote something that is obvious to
the hearer in the moment, what I’m talking about would denote ‘excel-
lent’, and ’s would denote a relationship between these two things which
attaches the quality ‘excellent’ to the denotation of that (often called a
predication).
The important point here is that there is an interplay between seman-
tics and pragmatics. Semantics provides a set of possible meanings, and
pragmatics is concerned with the choice among these possibilities.
Language users can take account of context, and use general knowledge
about the way the world works, to build interpretations on this semantic
foundation. In understanding (1.1), we use the context of the utterance
not only to understand what the speaker means by that, but also to figure
out which of the possible senses of the whole utterance is relevant: the
one in which they mean that that matches a previous topic of discussion,
or the one in which they just mean that that is excellent.
This view fits with a way of thinking about communication that was
pioneered by the philosopher H. P. Grice in the mid-twentieth century
(see Grice 1989 for a collection of relevant work) and which has been
highly influential within linguistics. According to this view, human
communication is not just a matter of encoding a signal, sending it,
and decoding it at the other end. Rather, it requires active collabora-
tion on the part of the addressee or hearer (the person the message
is directed to). The hearer’s task is to try to guess what the speaker
intends to convey, and the message has been communicated when (and
only when) the speaker’s intention has been recognised. (By conven-
tion, I’ll refer to the ‘speaker’ and the ‘hearer’ whether we’re actually
talking about speech or written communication.) The speaker’s task
is to judge what needs to be said (or written) in order for this process
to work: that is, so that it’s possible for the hearer to recognise the
speaker’s intention.
studying me aning 3
This way of looking at communication has several important

consequences:
• The same string of words can convey different messages, because –
depending on the context – there will be different clues to help the
hearer recognise the speaker’s intention.
• Similarly, different strings of words can be used to convey the same
message, because there may be lots of different ways for the speaker
to give clues to their intention.
• Sometimes a great deal can be communicated with very few words,
because of the active participation of the hearer in recovering the
speaker’s meaning.
• Mistakes are possible: the hearer might arrive at the wrong inter-
pretation. If a speaker suspects that this has happened, they may try
to repair the communication by saying more. This is easy in face-
to-face interactions (where the speaker can monitor the hearer’s
reaction for clues that they might have been misunderstood), harder
in live telephone conversations or text chats (because of the lack of
feedback), harder still in emails or letters, and essentially impossible
in books (where the writer may never find out whether the reader
understood them as intended).
The rest of this chapter introduces some other concepts that are
important to the study of linguistic meaning, and indicates which later
chapters develop them further. It will define a few more technical terms,
but (hopefully) just enough to give you a reasonable initial grasp of
semantics and pragmatics and set you up for reading introductory books
in this area.
In everyday life, competent users of a language generally don’t give
a lot of thought to the details of what they’re doing and how it works.
Linguistics researchers do: they generally operate on the assumption
that there are interesting things to be learned from those details. This
can appear a little obsessive at times, like in the discussion of (1.1) – why
should we care what that, on its own, means? And does it make a differ-
ence that ’s does a slightly different thing in the two possible interpreta-
tions? But the general goal of linguistic analysis is often to try to bring
all the unconscious knowledge that we have about language to con-
scious awareness. It turns out that close inspection of bits of language
and instances of usage – even quite ordinary ones – can sometimes give
us new insights into how language really works and, by extension, how
we think.
1.1 Sentences and utterances

Just as we distinguish semantics and pragmatics, we distinguish sentences
and utterances. Utterances are the things that have meaning in our
immediate experience as language users. (1.2) presents three examples.
(1.2) a. I agree with you.
b. The departments.
c. All that glisters is not gold.
Utterances are the raw data for our linguistic analysis. Just as we use
the term “speaker” to mean the sender of a particular message, whether
that’s in speech or writing, we use the term “utterance” to refer to that
message (or a part of it), regardless of whether or not it is actually
spoken out loud. Each utterance is unique, having been produced by a
particular speaker in a particular situation, and can never precisely be
repeated. Even when someone “says the same thing twice”, by repeat-
ing themselves, the two utterances differ in potentially important ways
because the context has changed (it may be relevant that the previous
utterance was made, for instance). It’s easy to see how the content of
what a speaker means by saying (1.2a) depends crucially on the context:
after all, the speaker is more likely to mean that they agree with the
hearer about the particular issue under discussion at the moment than
they are to mean that they always or generally agree with the hearer.
Notation
When it matters, I use “” double quotes for utterances, ‘’ single
quotes for meanings, and italics for sentences and words that are
being considered in the abstract. I also use single quotes when
quoting other authors, and double quotes as “scare quotes” around
expressions that are not strictly accurate but potentially useful
approximations.
The abstract linguistic object on which an utterance is based is a

s entence. For instance, if (1.2b) is uttered in response to the question
“Who is responsible for lecture scheduling?”, we could think of (1.2b) as
being based on the sentence The departments are responsible for lecture sched-
uling. If it had been uttered in response to some other question, it would
likely correspond to a different sentence, and bear a different meaning.
In the case of the utterance of the proverb (1.2c), we could argue that
the underlying sentence is straightforwardly All that glisters is not gold. We
studying me aning 5
don’t need to rely on context to figure this out. However, that sentence
is itself ambiguous: it has two possible meanings, ‘Not everything that
glisters is gold’ or ‘Nothing that glisters is gold’. Sometimes a speaker
will signal which interpretation is correct – in this case, for instance,
emphasising not might hint at the ‘not everything’ interpretation – but
often things are not so clear-cut. (We’ll see more examples like this in
Section 7.3.5.) However, even if the speaker is not giving us a clear indi-
cation of which meaning to choose, context might still help us under-
stand which one is more likely to be appropriate. In the case of this
proverb, the sense that is usually relevant is the weaker ‘not everything’
meaning – the speaker is more likely to want to convey the message
‘things may not be what they appear to be’ than to convey the message
‘things are never what they appear to be’.
1.2 Types of meaning

When we think about meanings, there are several slightly different
ideas that we might want to distinguish, among them speaker meaning,
utterance meaning, and sentence meaning. We can think of speaker
meaning as the meaning that the speaker intends to convey when they
produce an utterance. Assuming that, as hearers, we’re interested in
figuring out what people are trying to tell us – which is probably for
our own good, in general – we are constantly having to make informed
guesses about speaker meaning.
Speaker meaning, construed this way, is a private thing: only the
speaker really knows what they meant to convey. Indeed, a speaker
may not wish to admit that they intended to convey certain mean-
ings. Furthermore, a speaker might not be successful in conveying a
particular meaning – or, to put it another way, the hearer might not be
successful in recovering the meaning that the speaker intended. So we
might want to distinguish speaker meaning from utterance meaning,
which we could think of as the meaning that an utterance could be
understood as conveying when interpreted by people who know the
language, are aware of the context, and have whatever background
knowledge could reasonably be assumed by the speaker. Utterance
meaning is, in this sense, a necessary fiction or idealisation that linguists
doing semantics and pragmatics have to work with. It relies on the
intuitions that we have as language users about what would be likely
to happen, communicatively, as a result of a particular utterance being
made under a particular set of circumstances.
Because utterances are instances of sentences in use, an important
first step towards understanding utterance meaning is understanding
sentence meaning. I’ll take this to be the meaning that people familiar
with the language can agree on for sentences considered in isolation.
This is a good place to start because language users have readily acces-
sible intuitions about sentences. I appealed to a shared intuition about
sentence meaning when I said that (1.2c) was ambiguous. That was
a relatively subtle example, though: lots of sentences are much more
obviously ambiguous, such as (1.3a), which could mean (1.3b) or (1.3c),
or (1.4a), which could mean (1.4b) or (1.4c).
(1.3) a. We arranged to meet yesterday.
b. Yesterday, we arranged that we would meet (on some day I
am not specifying here).
c. We arranged a meeting that was to take place yesterday.
(1.4) a. Jane saw the guy with binoculars.
b. Jane, using binoculars, saw the guy.
c. Jane saw the guy who was carrying binoculars.
Language users’ access to the meanings of individual words – what
we call lexical meaning – is less direct. We can think of the meaning
of a word as the contribution it makes to the meanings of sentences in
which it appears. Of course, people know the meanings of words in their
language in the sense that they know how to use the words, but this
knowledge is not immediately available in the form of reliable intui-
tions. Speakers of English might be willing to agree that strong means the
same as powerful, or that finish means the same as stop; but these judge-
ments would be at least partly wrong, as shown when we compare (1.5a)
with (1.5b), and (1.5c) with (1.5d).
(1.5) a. This cardboard box is strong.
b. ?This cardboard box is powerful.
c. Mavis stopped writing the assignment yesterday, but she
hasn’t finished writing it yet.
d. *Mavis finished writing the assignment yesterday, but she
hasn’t stopped writing it yet.
Notation
Asterisks, like the one at the beginning of (1.5d), are used in
semantics to indicate that an example is seriously problematic as
far as meaning goes – in this case, the sentence is a contradiction.
Question marks, like the one at the beginning of (1.5b), are used
to signal less serious but still noticeable oddness of meaning.
studying me aning 7
From the above examples, we might conclude that finishing is a

special kind of stopping, specifically ‘stopping after the goal has been
reached’, and that strong can mean either ‘durable’ or ‘powerful’ (among
other possibilities), only one of which is applicable to cardboard boxes.
1.2.1 Denotation, sense, reference and deixis

Earlier in the chapter I said that expressions in a language – sentences,
words, and so on – denote aspects of the world. The denotation of an
expression is whatever it denotes. For many words, their denotation is a
large class of things: the noun thing itself would be an extreme example.
The links between language and the world are what makes it so
indispensable – we can use language to talk about things in the world.
That being the case, it is tempting to think that the meaning of an
expression is just its denotation. If I wanted to explain what window
meant, I might be able to do so by uttering the word and pointing to
a window. And in early childhood, our first words are probably learnt
through just such a process of live demonstration and pointing, known
as ostension. However, this is not plausible as a general explanation of
how meaning is acquired, for several reasons including the following:
• After early childhood, we usually acquire word meanings through
the use of language rather than the use of ostension (“A sash window
is a window which you can open by sliding it upwards”).
• Ostension isn’t a very good way of specifying what we’re talking
about. If I just said window while pointing at the window, how would
the hearer know whether I meant to label the window itself, or
something I could see through it, or indeed some other property
that it had (‘made of glass’, ‘transparent’, etc.)? And how would the
hearer know whether I meant to refer to this particular window, or
windows in general? So, in practice, when we do use ostension, it is
often accompanied by explanatory utterances.
• There are all kinds of abstract, non-existent and relational denota-
tions that cannot conveniently be explained by ostension: consider
memory, absence, yeti, or instead of.
If we can’t think of meaning as just a matter of denotation, how can
we think about it? One approach is that taken in formal semantics
(“formal” because it uses systems of formal logic to set out descriptions
of meaning, and theories of how the meanings of expressions relate to
the meanings of their parts; see Lappin 2001). In formal semantics, it
is important to consider what kind of denotation we are dealing with.
Count nouns, such as tree, may be said to denote sets of things; mass
nouns, like honey, denote substances; singular names denote individuals;

property words, like purple, denote sets of things (the things that have
the property in question); spatial relation words, like in, denote pairs of
things that are linked by that spatial relation; simple sentences such as
Amsterdam is in Holland denote facts or falsehoods; and so on. What is of
interest is the fact that the denotation is what it is – an individual, or a
set of things, or a set of pairs of things, and so on.
An alternative approach takes sense to be the central concept: this
is a slightly elusive idea, but we can think of it as the aspects of the
meaning of an expression that give it the denotation it has. Differences
in sense therefore lead to differences in denotation. Ambiguous words –
of which there are many in English – have multiple possible senses, and
consequently might denote different things. Consider (1.6), as part of a
job advertisement.
(1.6) a. The applicant should list his three most recent employers.
b. The applicant should list her three most recent employers.
c. The applicant should list their three most recent employers.
One sense of his in (1.6a) could be loosely expressed as ‘of the most con-
textually salient man’, while an alternative sense could be expressed as
‘of the most contextually salient person’. Given that, in this context, his
is clearly supposed to mean ‘of the applicant’, the former sense results
in the inference that application is restricted to men, whereas the latter
sense does not. Assuming that the advertiser does not wish to suggest
that there might be such a restriction, they would probably be better
advised to use an alternative formulation, such as (1.6c).
The idea of sense as the core of linguistic meaning offers a
helpful route into the study of semantics. This book will focus on
this approach, presenting it in a version that will hopefully also form
a reasonable foundation for anyone who later chooses to learn formal
semantics.
There are different ways in which we might try to write down
“recipes” for the denotations of words. One way of doing this is in
terms of sense relations, which are semantic relationships between
the senses of expressions. The idea here is that, if we can explain the
interconnections between words using well-defined sense relations,
then it is possible for a person who knows the denotations of some
words to develop an understanding of the meanings (senses) in the rest
of the system. To take a trivial example, if I know that I like corian-
der, and I learn that cilantro means the same as coriander, I know that I
like cilantro. This approach meshes well with the observation that we
commonly use language to explain meanings. Chapter 2 will start to
studying me aning 9
explain the sense relations that can hold between words, and phrases,
in a more systematic way.
We can usefully distinguish the idea of sense from that of reference.
Reference is what speakers do when they use expressions – which we
call referring expressions – to pick out particular entities – which
we call referents – for their audience. These referents can be of many
kinds, including, for instance (with sample referring expressions in
parentheses), people (“my students”), things (“the Parthenon Marbles”),
times (“midday”), places (“the city centre”) and events (“her party”).
Reference is strongly reliant on context. Consider (1.7), where the
speaker intends to use Edinburg to refer to the city of that name in
Indiana.
(1.7) We drove to Edinburg today.
The speaker of (1.7) would have to be sure that the hearer knows that
they are in Indiana, if the utterance is not to be misunderstood as refer-
ring to the Edinburg in Illinois, or the one in Texas, or even Edinburgh
in Scotland.
But of course there is more context-dependence in (1.7) than just
this. We in (1.7) refers to the speaker plus (usually) at least one other
person. Similarly, today refers to the day on which the utterance was
made. Thus, in order to understand what (1.7) means, we need to know
who the speaker is and when they are speaking – which will be trivial
face-to-face, but not possible if we just encounter the utterance out of its
original context, such as within the pages of this book. And even in the
face-to-face encounter it may not be obvious precisely who the speaker
means to refer to by we. A similar problem arises with the notice (1.8),
once posted on a course bulletin board.
(1.8) The first tutorial will be held next week.
The notice was posted in week 1 of the academic year, but not dated,
and the lecturer forgot to take it down. Some students read it in week 2
and missed the first tutorial because they quite reasonably interpreted
next week to mean ‘week 3’.
We refer to expressions like this, which have to be interpreted in rela-
tion to their context of utterance, as deictic expressions (or instances of
deixis). Deixis is pervasive in language, presumably because we can
often save effort by appealing to context in this way: it’s often easier to
say she than to specify a person’s name, easier to say here than to give a
clear description of the place of utterance, easier to say tomorrow than
to remember the date, and so on. There are different kinds of deixis,
relating to:
• time: now, soon, recently, ago, tomorrow, next week . . .

• place: here, there, two kilometres away, that side, this way,
come, bring, upstairs . . .
• persons and entities: she, her, hers, he, him, his, they, it, this, that . . .
• discourse itself: this sentence, the next paragraph, that’s what she
said, this is true . . .
Our semantic knowledge of the meanings of deictic expressions
guides us as to how we should interpret them in context. As always,
where context is concerned, these interpretations will be guesses rather
than certainties: perhaps a speaker who points to an object and says this
means that specific object, but perhaps they mean some property of it,
or perhaps they are referring to their hand or the pointing gesture itself.
Much of language is in some sense deictic: tense, which will be
discussed in Chapter 6, can also be thought of this way. Reference in
general is an important topic that will recur in many of the following
chapters, particularly Chapter 10.
1.3 Semantics vs pragmatics

As we’ve seen, the essential difference between sentences and utter-
ances is that sentences are abstract, and not tied to contexts, whereas
utterances are identified by their contexts. This is also an important way
of distinguishing between semantics and pragmatics. If you are dealing
with meaning and there is no context to consider, then you are doing
semantics, but if there is a context to consider, then you are probably
engaged in pragmatics. (As we shall see later in this section, there is
a bit of a grey area as to whether context can sometimes intrude into
semantics.) In the terms we’ve discussed so far, pragmatics is the study
of utterance meaning, while semantics is the study of sentence meaning
and word meaning.
As an example of how semantics and pragmatics relate to one
another, consider how we might interpret (1.9).
(1.9) The next bus goes to Cramond.
Considering this as a sentence, we can think about its sentence meaning,
drawing on the semantic information that we have from our knowledge
of the language. Anyone who knows English well can explain various
features of the meaning of (1.9): it means that a bus goes to Cramond,
and that one bus that does this is, in some sense, the ‘next bus’ (perhaps
the adjacent one physically, but more usually the next one to arrive).
Also, more subtly, goes to Cramond can be understood either to be making
studying me aning 11
a prediction about what the next bus will do, or stating a generalisation
about where that bus habitually goes. These meanings are available
without considering who might be saying or writing the words, or when
or where they are doing so: essentially, no context is involved. Hence,
their study falls within the domain of semantics.
By contrast, if we consider (1.9) as an utterance that takes place in
a particular context, we can derive a richer interpretation that goes
beyond the sentence meaning (which we will sometimes refer to as the
literal meaning). Suppose that a passenger steps onto a bus and asks
the driver whether this bus goes to Cramond, and the driver replies
by uttering (1.9). Assuming that the driver is being cooperative, we
can interpret (1.9) as conveying the answer ‘no’. This understanding
relies on us trying to figure out why, given the contextual and back-
ground information, the driver produced the utterance (1.9). The
literal meaning of (1.9) is obviously relevant to this process, but on this
occasion it doesn’t relate closely to the meaning that we end up deriv-
ing: whether or not the next bus goes to Cramond doesn’t really tell
us anything about whether this bus does so, and the utterance of (1.9)
doesn’t generally convey the meaning ‘no’. So the interpretation of (1.9)
as ‘no’, in this case, is a contextually driven additional meaning that goes
beyond what was literally stated – a type of meaning that we call an
implicature. And because it relies on context, the study of implicature
falls within the domain of pragmatics.
There are also forms of meaning which do not fall very clearly
within the scope of semantics or that of pragmatics. In (1.9), the ques-
tion of which bus actually is the next bus is dependent on the time and
place at which (1.9) is uttered. We have already seen that the resolu-
tion of deictic expressions depends on context in this way. In order to
understand what the speaker of (1.9) is committing to, we therefore
need to figure out a complete interpretation of the utterance, by using
contextual information and world knowledge to work out what is being
referred to (and how to understand potentially ambiguous expressions,
like next in this case). The basic interpretation of a sentence with these
details spelled out is sometimes termed an explicature. We could think
of explicature as part of pragmatics because it is context-dependent, or
part of semantics because it is essential to the meaning of the utterance
in a way that implicature is not. I will try to sidestep that theoretical
debate in this book, although the notion of explicature will be used in
the discussion of figurative language in Chapter 9.
1.3.1 A first outline of semantics

Semantics, the study of word meaning and sentence meaning abstracted
away from contexts of use, is primarily a descriptive subject. It is an
attempt to describe and understand the nature of the knowledge about
meaning that people have as a consequence of knowing a language. It
is not, however, prescriptive: that is to say, it is not about defining what
words “ought to mean” or pressuring speakers into abandoning some
meanings and adopting others. Nor is it about etymology, because
you can know a language perfectly well without knowing its history.
It may be fascinating to explore how different meanings have become
associated with historically related forms – arms, armour, army, armada,
armadillo, etc. – but that knowledge is not essential for understanding
or using present-day English, so it is not covered in this book. Nor do
we focus on semantic and pragmatic change, for the same reason – it is
interesting that the meanings of words change over time, and that, for
instance, silly originally meant ‘blessed’ and subsequently ‘innocent’, but
you do not need to know this in order to understand what it means now.
Having said all that, studying semantics and pragmatics is helpful
in understanding some of the processes involved in historical meaning
change: for instance, meanings that were once context-dependent can
end up becoming part of the semantics of a word. (A case in point:
armada in Spanish just meant ‘an armed force’, and its use in English
to mean specifically ‘a fleet of warships’ arose because Spanish Armada
denoted an armed force from Spain that happened to comprise a fleet
of warships. We might want to appeal to the notion of compositionality,
introduced in Section 2.2, and ideas about adjective meanings, discussed
in Chapter 4, to explain how this happened.)
The process of giving a semantic description of language knowledge
is also different from the encyclopedia-writer’s task of cataloguing
general knowledge. The words tangerine and clementine illustrate a dis-
tinction that may not be part of our knowledge of English: although an
expert will be able to tell these apart, most users of English will not. But
our focus in this book will be on the more abstract kinds of semantic and
pragmatic knowledge that underpin our ability to use language.
1.3.2 A first outline of pragmatics

Pragmatics is concerned with characterising how we go beyond what
was literally said, both in terms of what additional meanings get con-
veyed by a speaker and in terms of how the speaker encodes them and
the hearer recovers them. A crucial basis for making pragmatic infer-
ences is the contrast between what might have been uttered and what
actually was uttered. Consider (1.10), from the information provided to
guests at a B&B.
(1.10) Food and Drink. Breakfast is served from 7:30am to 9:00am.
There is a fridge in the hallway with soft drinks and snacks.
Payment for these is on an honesty basis.
As no further information is provided under this heading, we are invited
to infer that the establishment does not serve dinner and does not
provide alcoholic drinks. However, we cannot be entirely certain about
this: perhaps the proprietors simply didn’t want to promise that dinner
or alcoholic drinks would be available. These pragmatic inferences do
not have the same status as the information that is explicitly asserted: a
guest who read (1.10) would have grounds to complain if the fridge in
the hallway contained no soft drinks or no snacks, but they would not
(usually) have grounds to complain if the fridge also contained beer.
Another widely available pragmatic inference, often called a scalar
implicature, arises when words can be ordered on a semantic scale, as
for example the value-judgements excellent > good > OK.
(1.11) A: What was the accommodation like at the camp?
B: It was OK.
Speaker A can draw an inference from B’s response, because if the
accommodation had been better than merely OK, B could have used
the word good or indeed excellent to describe it. As B does not do so, A
can infer that the accommodation was no better than satisfactory. But, as
in (1.10), B is not as committed to the accommodation being ‘no better
than OK’ than they are to it being ‘at least OK’. Indeed, if B sounds sur-
prised, the inference ‘no better than OK’ may be less readily available:
B then might be assumed to mean something like ‘contrary to expecta-
tions, it was acceptable, and maybe even better than that’. We’ll revisit
this kind of inference in Section 8.2.2.
Pragmatic inferences of this kind occur all the time in communica-
tion: even though they are really just informed guesses, they are crucial
to the smooth functioning of much of our communication. Because they
are informed guesses, it is one of their defining features that they can be
“cancelled”, in the sense that the speaker can typically deny a pragmatic
inference without seeming to contradict themselves. In (1.11), B could
continue “In fact, it was pretty good” without being self-contradictory,
because what is ‘pretty good’ in this case is also ‘OK’. Pragmatics is the
focus of the later chapters of this book, but will also figure in sections of
most of the other chapters.
Summary
Hearers (including readers) have the task of guessing what speakers
(including writers) intend to communicate when they produce utter-
ances. If the guess is correct, the speaker has succeeded in conveying
the meaning. Pragmatics is about how we interpret utterances – and
produce interpretable utterances – taking account of context and back-
ground knowledge. Semantics is the study of the context-independent
knowledge that users of a language have about the meanings of expres-
sions, such as words and sentences. Crucially, expressions of language
relate to the world outside of language. In this book, we will explore this
idea through the notion of sense and of the meaning relations that hold
within a language, in ways that later chapters will make clearer.
Exercises
1. Here are two sets of words: {arrive, be at/in, leave} and {learn, know,
forget}. There is a similarity between these two sets, in how the words
relate to one another. Can you see it? Here is a start: someone who
is not at a place gets to be there by arriving; what if the person then
leaves? Once you have found the similarities between the two sets,
answer this follow-up question: was this a semantic or a pragmatic
task?
2. A student says to the tutor “How did I do in the exam?” and the
tutor replies “You didn’t fail.” What the tutor opted to say allows the
student to guess at the sort of grade achieved. Do you think the grade
was high or low? How confident are you about this? Briefly justify
your answer. In doing this, were you doing semantics or pragmatics?
3. Pick the right lock is an ambiguous sentence. State at least two mean-
ings it can have. How many different propositions could be involved?
4. A common myth about the word kangaroo is that it comes from an
expression in an Australian language (specifically Guugu Yimithirr)
meaning ‘I don’t understand you’. This was supposedly because
explorers in the eighteenth century pointed to a kangaroo and asked
what it was, and a local replied kangaroo. What does this story tell us
about the limits of ostension? And how would we disprove the claim
that this is what kangaroo originally meant?
5. An old joke concerns someone reading a sign saying Dogs must be
carried on this escalator and having to wait ages for a dog to appear so
that they could use the escalator. If that sign really caused people any
problems, how could you add a deictic term to it and thus resolve the
ambiguity?
6. Example (1.6a) is potentially problematic as an utterance because,

on one interpretation of his, it suggests that only men are welcome
to apply for the job. Does (1.6b) run into a similar problem? What
about (1.6c)?
2 Sense relations
Overview
As mentioned in Chapter 1, we generally learn the meanings of our first
few words through close encounters with the world, carefully mediated
by our caregivers. But once we have a start in language, we learn the
meanings of most other words through language itself. This might be
through having meanings deliberately explained to us – for instance, we
might be told that “tiny means ‘very small’”. It might also be because we
draw inferences about meanings based on our knowledge of language.
For instance, if we see the title of Gerald Durrell’s book My Family and
Other Animals, we might infer from this that people can potentially be
classified as a type of animal.
In both of these cases, we are relying on the existence of relation-
ships between the senses of expressions (words and phrases) within the
language that we speak. In the first case, we rely on the fact that the
meaning of tiny can be expressed in terms of the meanings of very and
small. In the second case, we use our understanding of the meaning of
other to spot the existence of a relationship between the meanings of
family and animals. One task for semantics is to describe these relation-
ships systematically, with the ultimate ambition of identifying exactly
what it is that a competent speaker of a language – in this case, English
– has to know about the meanings of sentences. Putting this together
with the speaker’s knowledge about pragmatics – how we use sentences
– we would then have a picture of what a speaker knows about linguistic
meaning altogether. In this chapter, we make a start on this task, begin-
ning in the following subsection with some technical preliminaries.
2.1 Propositions and entailment

We need to account for sentence meaning in order to develop explana-
tions of utterance meaning, because utterances are sentences put to use.
16
sense rel ations 17
The number of sentences in a human language is potentially infinite, so

we cannot analyse sentence meaning just by listing every possible sen-
tence and its interpretation (and nor do we learn sentence meanings this
way, with the exception of sentences that are fixed idiomatic expres-
sions). Both as linguistic scholars and as language learners, we have to
generalise in order to discover the principles of sentence meaning.
One important generalisation is that different sentences can carry the
same meaning, as in (2.1a–c).
(2.1) a. Sharks hunt seals.
b. Seals are hunted by sharks.
c. Seals are prey to sharks.
Proposition is a term for the core meaning of a sentence: we could say
that (2.1a–c) express the same proposition. Propositions are therefore
not tied to particular words or sentences. They have the property that
they are either true or false. That is not to say that the speaker or hearer
(or anyone else) has to know whether the proposition actually happens
to be true or false, as far as the real world is concerned – just that it must,
in principle, be either true or false.
The sentences in (2.1) are declaratives: that is to say, they follow
the sentence pattern on which statements (utterances that explicitly
convey factual information) are typically based. It is easy to see that
they express propositions, because it is possible to affirm or deny or
query the truth of these statements (“Yes, that’s true”, “No, that’s a lie”,
“Is that really true?”). By contrast, such responses are not appropriate to
utterances such as (2.2a–b), which are based on other sentence patterns.
(2.2) a. What’s your name?
b. Please help me.
Although (2.2a–b) don’t actually express propositions, we could still try
to analyse their meaning with reference to propositions. We could say
that (2.2a) relates to a proposition of the form ‘the hearer’s name is ___’,
and invites the hearer to supply their name to fill in the blank. Similarly,
we could say that (2.2b) presents the proposition ‘the hearer will help the
speaker’ and indicates that the speaker wants that proposition to be true.
But it might be more useful to think about utterances like this in terms of
the social actions that speakers are trying to perform when they produce
them, as we shall see in Chapter 11 when discussing speech acts.
An important relation that holds between propositions is entailment,
which we can write as ⇒. We say that one proposition entails another
(p ⇒ q, “p entails q”) if the first proposition being true guarantees that
the second is also true. Entailment is also called “logical consequence”
and is one of the most fundamental concepts in logic: arguments are

logically valid if and only if the starting points (premises) entail the
conclusions. Although entailment holds between propositions, we often
talk about it holding between sentences – which is fine as long as we
make sure that we are considering each sentence with one specific
propositional meaning.
The examples in (2.3) illustrate some of these points.
(2.3) a. Moira has arrived in Edinburgh.
b. Moira is in Edinburgh.
c. Moira has arrived in Edinburgh ⇒ Moira is in Edinburgh
d. *Moira has arrived in Edinburgh and she is not in Edinburgh.
When (2.3a) is true we can be sure that (2.3b) is also true – assuming,
that is, that we are talking about the same person called Moira and the
same place called Edinburgh in both sentences, as otherwise the propo-
sitional meanings won’t be related. If the truth of (2.3a) guarantees the
truth of (2.3b), that means that the entailment shown in (2.3c) holds.
Another way of stating this is to point out that (2.3d) is a contradiction:
given that (2.3a) entails (2.3b), we cannot sensibly affirm (2.3a) and the
negation of (2.3b) at the same time. To put this in more general terms,
if we have propositions p and q such that p ⇒ q, we know that the
complex proposition ‘p & not-q’ is necessarily false.
(2.3a) has other entailments, as shown in (2.4).
(2.4) a. Moira has arrived in Edinburgh ⇒ Moira is not in
Birmingham
b. Moira has arrived in Edinburgh ⇒ Moira was travelling to
Edinburgh
The word arrived is an important contributor to (2.3a) having the entail-
ments shown. If lived or been were substituted for arrived, the entailments
would be different. If someone were to ask what arrive meant, a sen-
tence like (2.3a) could be given as an example, explaining that it means
that Moira travelled from somewhere else and is now in Edinburgh.
However, the entailments from a sentence depend not only upon the
content words in the sentence, like arrive, but also upon their grammati-
cal organisation. For instance, the grammatical construction with has,
sometimes called the “present perfect” construction, is crucial to the
entailment in (2.3c) – if we replaced has with had, this entailment would
no longer be valid, as Moira may subsequently have left Edinburgh
again. There is more detailed discussion of the present perfect construc-
tion, alongside other ways of expressing tense and aspect in English, in
Chapter 6.
sense rel ations 19
If (2.3a) is understood and accepted as true, then none of its entail-

ments need to be put into words. They follow automatically, and can
be inferred from (2.3a) by the hearer. It is obviously crucial to success-
ful language use for speakers to make sure that the sentences that they
utter have the correct entailments. In Chapter 1 I introduced the idea
of sense as the aspects of the meaning of an expression that give it the
denotation it has. We can think of the sense of a word in terms of the
particular entailments that a sentence has as a result of containing that
word. Whichever aspects of the word’s meaning are responsible for
the sentences having those entailments are its sense. We can think of
the entailment in (2.3c) as part of the sense of arrived, for instance: it is
crucial to the meaning of arrived that if someone “has arrived” in a place,
it follows that they are now “in” that place.
2.2 Compositionality
Given the potentially infinite supply of distinct sentences in a lan-
guage, semanticists aim to develop a compositional theory of meaning.
The principle of compositionality is the idea that the meaning of a
complex expression is determined by the meanings of its parts and how
those parts are put together. The idea that human language is compo-
sitional in this sense has a very long history, and is of great importance
to linguistics: among other things, it offers a partial explanation of how
we can comprehend the meanings of infinitely many different poten-
tial sentences, just by knowing the meanings of finitely many different
sentence parts (and their combining rules). The meaningful parts of a
sentence are clauses, phrases and words, and the meaningful parts of
words are morphemes.
The idea that the meaning of a complex expression depends on the
meaning of the parts and how they are put together is hopefully a rea-
sonably intuitive one. It’s comparable to what happens in arithmetic.
Several things affect the result of an arithmetical operation, as we see
in (2.5): it makes a difference what numbers are involved (2.5a), what
operations are performed (2.5b), and – where there are multiple opera-
tions – also the order in which the operations take place (2.5c).
(2.5) a. 3 + 2 = 5 but 3 + 4 = 7
b. 3 + 2 = 5 but 3 × 2 = 6
c. 3 × (2 + 4) = 18 but (3 × 2) + 4 = 10
The linguistic examples in (2.6) show something similar to what we
see in (2.5c). Here we are considering a word that consists of three
morphemes, un, lock and able. Our operations are not addition and
ultiplication, but negation (or reversal) performed by the prefix un-

m
and the formation of “capability” adjectives by the suffix -able.
(2.6) a. un(lockable) ‘not able to be locked’
b. (unlock)able ‘able to be unlocked’
As in (2.5c), the brackets in (2.6) indicate the scope of the operations:
that is to say, which parts of the representation un- and -able operate
on. In (2.6a), un- operates on lockable, whereas ‑able operates only on
lock. In (2.6b), un- operates only on lock and -able operates on unlock.
However, unlike in arithmetic, there isn’t a default order of precedence,
and we don’t use brackets like this in ordinary writing or speech. In
practice, unlockable is just ambiguous – and the same goes for various
other expressions which take un- as a prefix and -able as a suffix, such as
unbendable, undoable and unstickable. Of course, sometimes only one inter-
pretation makes sense: as it is not usually possible to unbreak something,
unbreakable must be understood as ‘not able to be broken’.
In syntax too there can be differences in meaning depending on the
order in which operations apply. We saw an example of this in the first
chapter, repeated here as (2.7); (2.8) is another straightforward example.
(2.7) Jane saw the guy with binoculars.
(2.8) I didn’t sleep for two days.
Both of these are ambiguous for essentially the same reason as unlockable:
we can think of them as involving two possible “bracketings”. On one
reading of (2.7), with binoculars can be bracketed together with the guy as
a constituent of the sentence: on the other reading, with binoculars modi-
fies saw. On one reading of (2.8), for two days is an adjunct to sleep and the
sentence expresses the meaning that ‘For two days, the speaker did not
sleep’; on the other reading, for two days is a complement to sleep and can
be bracketed with it, and the sentence expresses the meaning ‘It is not
the case that the speaker slept for two days’.
As with unlockable, the ambiguity is not a one-off fact about these
particular sentences, but occurs systematically with similar sentences.
(2.9a–c) are additional examples in which, like in (2.8), a prepositional
phrase could be an adjunct or a complement, giving rise to different
meanings. Our account of how compositionality works is going to have
to allow for the fact that, in cases like this, sentences systematically end
up with multiple possible interpretations.
(2.9) a. I won’t be in town until 4 o’clock.
b. I refuse to see him twice a day.
c. I agreed to contact her during the committee meeting.
sense rel ations 21
At the same time, our language also contains expressions which

appear to be compositional but in fact are not. These are expressions
for which the meanings cannot be worked out by considering the mean-
ings of the parts and how those parts fit together. We call them idioms.
Classic examples in English are expressions such as head over heels, see
eye to eye or kick the bucket. These simply have to be learned as wholes
(see Grant and Bauer 2004 for more discussion). In this respect, idioms
behave like individual words that consist of one morpheme: all we can
do is to learn the association between the form and its meaning. As we
have seen, for some words that consist of multiple morphemes, like
unlockable, we can work out the meaning(s) of the word from the mean-
ings of the individual morphemes and how they are combined. But
there are also words that appear to consist of multiple morphemes but
are idiomatic in their overall meaning: we can’t predict the meaning of
greengrocer just from knowing the meanings of green and grocer, or that of
greenhouse from knowing the meanings of green and house.
2.3 Synonymy
Synonymy is equivalence of sense. The nouns mother, mom and mum are
synonyms of each other. When a single word in a sentence is replaced
by a synonym – a word equivalent in sense – then the literal meaning
of the sentence is not changed: My mother/mom/mum is from London.
Sociolinguistic differences – things like the fact that mum and mom are
informal, and that they are used respectively in British English and
North American English – are not relevant to the sense of the word
here, because they do not affect the propositional content of the sen-
tence (although sociolinguistics is a huge and important area of study in
its own right, which we won’t attempt to get into in this book).
Sentences with the same meaning are called paraphrases of one
another. (2.10a) and (2.10b) are paraphrases, differing only by substitu-
tion of the synonyms impudent and cheeky.
(2.10) a. Andy is impudent.
b. Andy is cheeky.
c. Andy is impudent ⇒ Andy is cheeky
d. Andy is cheeky ⇒ Andy is impudent
e. *Andy is impudent but he isn’t cheeky.
f. *Andy is cheeky but he isn’t impudent.
As indicated by (2.10c), (2.10a) entails (2.10b) – again, assuming that
we are talking about the same person called Andy at the same point in
time. Similarly, (2.10b) entails (2.10a). If one of these entailments failed,
(2.10a) and (2.10b) would not be paraphrases of each other. Also, both
(2.10e) and (2.10f) are contradictions: it’s not possible for someone to be
cheeky if they are not impudent, or vice versa. This also follows from
(2.10a) and (2.10b) being paraphrases.
In order to produce a semantic description of English, we would
typically start from judgements about sentences and try to use those to
establish sense relations between words. If we find a pair of sentences
such as (2.10a) and (2.10b) which contain a different adjective but are
otherwise identical, and it transpires that each entails the other (that is,
they are paraphrases), then that would be evidence that the two adjec-
tives are synonyms. Similarly, if we find a pair of sentences such as
(2.10e) and (2.10f) that are both contradictions, that would be evidence
that the two adjectives contained in those sentences are synonyms: if
they were not perfectly synonymous, at least one of these sentences
could at least potentially be true.
The relation of paraphrase depends upon entailment: paraphrase
means that there is a two-way entailment between the sentences. We
can think of this as entailments indicating the sense relations between
words, and sense relations indicating the entailment potentials of words.
How can we find paraphrases? We might do this by observing lan-
guage in use; but we might also invent test sentences and see whether or
not particular entailments are present, according to our own judgement
as language users (or the judgement of other proficient speakers of the
language). To make this task easier, we might be interested in develop-
ing examples about which it is easy to have intuitions, such as (2.11).
(2.11) a. You said Andy is cheeky, so that means he is impudent.
b. You said Andy is impudent, so that means he is cheeky.
So generally signals that an inference is being made. If both (2.11a) and
(2.11b) are judged to be true, it appears that the entailments (2.10c) and
(2.10d) both hold, and hence cheeky and impudent are synonyms. Note
that we are interested in accessing our knowledge about the general
pattern of entailment, not about the likely character of any specific
person named Andy. We don’t want to know that if Andy is impudent
then he is probably cheeky, or vice versa: we want to know that, if a
speaker is committed to the idea that Andy is impudent, they are auto-
matically committed to the idea that he is cheeky, and vice versa. If
people accept (2.11a) and (2.11b) as reasonable arguments, they must
agree with this, and we can conclude that the adjectives are synonyms.
Having said that, we might need to allow for the possibility that
someone will reject (2.11a) or (2.11b) – or both – on the basis that there
is a difference in register between cheeky and impudent. That is to say,
sense rel ations 23
the circumstances under which you would use cheeky, as a speaker, are
not the same as the circumstances under which you would use impudent.
Like the contrast between mum and mother, this is part of our knowledge
of language, but part of our sociolinguistic knowledge rather than our
semantic knowledge. So one of the challenges we have to confront
when we are trying to have intuitions about entailment is whether we,
as judges, are relying only on the kinds of knowledge that we are inter-
ested in as analysts.
Other pairs of synonymous adjectives include silent and noiseless,
brave and courageous, and polite and courteous. But there are also many
pairs of adjectives that have similar meanings without being synonyms.
Consider (2.12).
(2.12) a. The building is enormous.
b. The building is big.
Here, (2.12a) entails (2.12b), but the reverse is not true: the build-
ing could be big without qualifying as enormous. Hence, the relation
between big and enormous is not one of synonymy.
Synonymy not only applies to nouns and adjectives: it is also present
in other word classes. The adverbs quickly and fast are synonyms, and,
in Scottish English, so are the prepositions outside and outwith. And, as
shown in the mother/mom/mum example, synonymy is not restricted to
pairs of words: as another example, the triplet sofa, settee and couch are
all synonymous. We can tell this because each pair of words within the
triplet exhibits synonymy. In fact, because of the way entailment works,
synonymy is a transitive relation: that is to say, if a is a synonym of b and
b is a synonym of c, then a must also be a synonym of c.
2.4 Complementarity, antonymy, converseness and incompatibility

Some pairs of adjectives, such as moving and stationary, not only apply
to a broad class of objects but also divide that whole class of objects
into two non-overlapping sets. Everything that is capable of moving or
being stationary – that is to say, any physical object – is, at a given point
in time, either moving or stationary. Some other adjectives that divide
their relevant domains in this way are listed in (2.13).
(2.13) same different
right wrong
true false
intact damaged
connected disconnected
These pairs are complementary terms – so called because the comple-

ment of things that are not described by one term are described by the
other. That is to say, they give rise to the pattern of entailments illus-
trated in (2.14).
(2.14) a. Maude’s is the same as yours.
b. Maude’s is different from yours.
c. Maude’s is the same as yours ⇒ Maude’s is not different
from yours
d. Maude’s is different from yours ⇒ Maude’s is not the same
as yours
e. Maude’s is not the same as yours ⇒ Maude’s is different
from yours
f. Maude’s is not different from yours ⇒ Maude’s is the same
as yours
In other words – assuming that Maude’s and yours denote the same thing
in (2.14a) and (2.14b) – when (2.14a) is true, (2.14b) is false, and vice
versa. This is a way of expressing the obvious idea that if two things are
the same, they are not different; and that if two things are different, they
are not the same.
Another, more formal, way of putting this is that (2.14a) entails and is
entailed by the negation of (2.14b), and (2.14b) entails and is entailed by
the negation of (2.14a). Hence, (2.14a) is a paraphrase of the negation of
(2.14b), and (2.14b) is a paraphrase of the negation of (2.14a). So we can
think of complementaries as negative synonyms. Note that we have had
to change the sentence frame slightly between (2.14a) and (2.14b), so it
might be more precise to say that there’s complementarity between the
same as and different from (rather than just same and different).
A similar sense relation, but slightly weaker than complementarity,
is that of antonymy. Sometimes this term is used just to mean any kind
of “oppositeness”, but most semanticists use it to apply to one particular
kind of oppositeness, exemplified by the adjectives noisy and silent, as
in (2.15). Note that we use the symbol ⇏ to mean that an entailment is
not valid: that is, if the sentence before ⇏ is true, it doesn’t necessarily
follow that the sentence after ⇏ is true.
(2.15) a. The street was noisy.
b. The street was silent.
c. The street was noisy ⇒ The street was not silent
d. The street was silent ⇒ The street was not noisy
e. The street was not noisy ⇏ The street was silent
f. The street was not silent ⇏ The street was noisy
sense rel ations 25
Again assuming that we are referring to the same thing by the street in
(2.15a) and (2.15b), we get the pattern of entailments shown above,
which is different to the complementary case. If we know that (2.15a)
is true then we know that (2.15b) is false, and if we know that (2.15b)
is true then we know that (2.15a) is false. However, unlike the comple-
mentary case, knowing that (2.15a) is false doesn’t tell us whether or not
(2.15b) is true, and knowing that (2.15b) is false doesn’t tell us whether
or not (2.15a) is true. This is because there is middle ground between
silent and noisy, whereas there is no middle ground between the same as
and different from: to say that something is not noisy is not to say that it is
silent, and to say that it is not silent is not to say that it is noisy.
To put it another way, pairs of antonyms typically tap into mean-
ings that are at opposite extremes, but unlike complementaries, they
leave gaps in the middle. Under this definition, there are many pairs
of antonyms: happy and sad, full and empty, early and late, and so on. It
is not a coincidence that it is easier to find pairs of antonyms than it is
to find synonyms or complementaries. Synonyms can be thought of as
something of a luxury: given that two synonyms (such as courteous and
polite) give rise to the same entailments, we could really do without one
of them in the language, and still manage to convey all the information
that we need to. Having an additional term might enable us to commu-
nicate in a more expressive or sociolinguistically richer style. Having
words for both members of a complementary pair is arguably some-
thing of a luxury too: we could get away with having just one, and use
negation to convey the other (instead of false we could just say not true).
However, this will not work with antonyms: to say that something is full
is more than just saying that it is not empty. We need both full and empty in
the language in order to talk about quantity in this way.
A general feature of the adjectives that form antonym pairs is
that there is also a sense relation between their comparative forms.
Comparatives are formed by the suffix - er for some adjectives (thicker,
poorer, humbler) or more generally by the construction “more + adjective”
(more patient, more obstinate). The comparative forms of an antonym pair
of adjectives exhibit a sense relation called converseness, illustrated in
(2.16).
(2.16) a. France is bigger than Germany.

b. Germany is smaller than France.
c. France is bigger than Germany ⇒ Germany is smaller than
France
d. Germany is smaller than France ⇒ France is bigger than
Germany
Starting with (2.16a), if we replace bigger with smaller and swap the posi-
tion of the noun phrases France and Germany, we obtain a paraphrase,
(2.16b). Thus we say that bigger and smaller are converses. We can think
of converseness as something like a version of synonymy that also
requires the reordering of noun phrases.
Converseness is also present in other word classes, including nouns
(such as parent (of) and child (of)), verbs (such as precede and follow) and
prepositions (such as above and below). In each of these cases, if we have
two entities X and Y, and X stands in one of these relations to Y, it must
be the case that Y stands in the converse relationship to X (for instance,
if X is above Y, then Y is below X, and vice versa).
Just as synonymy is not restricted to pairs of items, neither is
antonymy. We can often identify sets of terms for which any two
members are antonyms: we can say that these sets exhibit incompat-
ibility. For instance, we can consider a set of colour adjectives such as
{black, blue, green, yellow, red, white, grey} to be incompatible, in that if
we say something is “blue”, it follows automatically that it is not black,
green, yellow, etc. – assuming that we are dealing with objects with a
single predominant colour. Similarly, a set of nouns denoting shapes,
such as {triangle, circle, square}, might also exhibit incompatibility: if
something is “a triangle”, it is not a circle or a square, and so on. A set
has incompatibility if every member of the set exhibits antonymy with
every other member of the set, so the diagnostics for incompatibility
will be essentially the same as for antonymy.
2.5 Hyponymy
The relation of hyponymy is about the different subcategories of a
word’s denotation. The pattern of entailment that defines hyponymy is
illustrated in (2.17).
(2.17) a. There’s a house on the riverbank.
b. There’s a building on the riverbank.
c. There’s a house on the riverbank ⇒ There’s a building on
the riverbank
d. There’s a building on the riverbank ⇏ There’s a house on
the riverbank
If it is true that there is a house on the riverbank, it follows that there is a
building on the riverbank, as indicated in (2.17c). This is because a house
is one kind of building. There are other kinds of building: school, church,
factory, and so on. Hence, the reverse entailment does not hold, as shown
in (2.17d): the building on the riverbank could be something other than
sense rel ations 27
a house. This pattern of entailment tells us that we are dealing with

hyponymy, and that house is a hyponym of building, or equivalently
building is a superordinate (sometimes called “hypernym”) of house.
Often, as in (2.17), a sentence containing a hyponym entails the cor-
responding sentence in which the hyponym has been replaced by its
superordinate. However, if the sentence is negative, this pattern may be
reversed, as shown in (2.18).
(2.18) a. There isn’t a house on the riverbank.
b. There isn’t a building on the riverbank.
c. There isn’t a house on the riverbank ⇏ There isn’t a build-
ing on the riverbank
d. There isn’t a building on the riverbank ⇒ There isn’t a
house on the riverbank
To put it another way, the fact of there being a building on the river-
bank is a necessary condition for there to be a house on the riverbank.
Hence, it would be reasonable to say that ‘building’ is a component of
the meaning of house: a house is a ‘building for living in’.
2.5.1 Hierarchies of hyponyms

House is a hyponym of the superordinate building, but building is in turn
a hyponym of structure, and structure is in turn a hyponym of thing. Like
synonymy, hyponymy is also transitive: if a is a hyponym of b and b is a
hyponym of c, then a is a hyponym of c. This means that house is also a
hyponym of structure and a hyponym of thing, which is fairly obvious if
we think of the definition above: if we replaced building by structure or
thing, the entailment patterns in (2.17) and (2.18) would stay the same.
Similarly, thing is a superordinate of house, and so on. These relations are
illustrated in Figure 2.1.
Incidentally, this means that we don’t have to worry about whether
we’ve captured all the layers in this hierarchy. If we discovered that
there was another category between structure and thing, say artefact,
which was a hyponym of thing and a superordinate of structure, that
would not affect the hyponymy relation between structure and thing.
As we move up a hierarchy of hyponymy, the senses of the words
become less specific and their denotations become larger and more
general. At lower levels, the senses are more detailed and the words
denote narrower ranges of things, as illustrated in Figure 2.2.
In Figure 2.2 we try to capture the meaning of each hyponym as the
meaning of its immediate superordinate with a modifier (e.g. ‘for living
in’). This captures the key idea that a hyponym is a special case of its
thing superordinate of structure, building, house, . . .
structure hyponym of thing, superordinate of building, house, . . .
building hyponym of structure, thing, superordinate of house, . . .
house hyponym of building, structure, thing, superordinate of . . .
Figure 2.1 An example of levels of hyponymy
thing ‘physical entity’
structure ‘thing with connections between its parts’ = ‘physical entity with
connections between its parts’
building ‘structure with walls and a roof’ = ‘physical entity with connections
between its parts with walls and a roof’
house ‘building for living in’ = ‘physical entity with connections between
its parts with walls and a roof, for living in’
Figure 2.2 Hyponym senses get successively more detailed as we go down

the tree
superordinate: it’s an instance of the superordinate that has certain prop-

erties that not all instances of that superordinate share. In practice, it’s not
always easy to provide useful hyponym definitions in this way, because
sometimes we run into circularity: a dog is a type of animal, but it’s difficult
to describe what type of animal it is without using the word ‘dog’ or some
related word like ‘canine’. It is, in effect, ‘an animal that is a dog’.
Even so, hyponym hierarchies are useful to us as language users
because they guarantee the validity of a large number of inferences. If
someone tells us facts about a particular house, we know that the things
they are telling us are also true of at least one structure, and at least one
building, and so on. To take a marginally more practical example, if we
know that platypus is a hyponym of mammal, we know that (2.19a) entails
(2.19b).
(2.19) a. Platypuses lay eggs.

b. There are mammals that lay eggs.
sense rel ations 29
entity
thing time person1 idea
place structure product plant animal1 student
dam building vehicle tool animal2 person2 freshman (post)

graduate
barn house garden tool kitchen utensil
saw
hacksaw
Figure 2.3 Part of the hyponym hierarchy of English nouns
Hyponym hierarchies exist for other parts of speech, such as verbs

and adjectives, as well as nouns. Amble is a hyponym of walk, which in
turn is a hyponym of move; (made of) oak is a hyponym of wooden, and
so on. Moreover, these hierarchies are potentially vast. WordNet is
a systematic database of English word meanings, which, at the time
of writing, contains entries for more than 155,000 words. In creating
WordNet, Miller and Fellbaum (1991: 204) discovered that a hyponym
hierarchy with twenty-six high-level superordinates (time, plant, animal,
and so on) ‘provides a place for every English noun’.
Figure 2.3 shows a tiny fraction of the WordNet noun hierarchy,
featuring just seven of the twenty-six superordinates (and, of course,
omitting the vast majority of their hyponyms). Note that some enti-
ties appear twice in this hierarchy: we distinguish two senses of person
(corresponding roughly to ‘human’ and ‘psychological individual’) and
two senses of animal (corresponding roughly to ‘living thing that is not a
plant’ and ‘living thing that is not a plant or a human’).
In addition to hyponymy itself, we might try to identify other sense
relations within a hyponym hierarchy. The hyponyms of a given
superordinate might be linked by incompatibility – house and factory
are different hyponyms of building, so we might argue that something
that is a house is not a factory – but actually this does not follow from
the definition of hyponymy we have been working with here. Indeed, if

house and dwelling are synonyms, then they are both hyponyms of build-
ing, but they are clearly not incompatible.
Summary
In this chapter, we discussed the notion of entailment and its relevance
to semantics. Based on this, we were able to define various distinct
sense relations that can obtain between expressions in a language: syn-
onymy, antonymy, complementarity, converseness, incompatibility and
hyponymy. Entailments between sentences are the evidence for sense
relations between words; or, going the other way, sense relations indi-
cate the entailment potentials of words. The idea of sense relations is
helpful to us, as analysts, when trying to give a semantic description of a
language, but also when we as speakers are trying to learn the semantics
of a language.
Exercises
1. The word dishonest means ‘not honest’. The following five words
also have ‘not’ as part of their meaning: distrust, disregard, disprove,
dislike, dissuade. Write a brief gloss for the meaning of each, similar
to the one given for dishonest. Thinking of sentences for the words
will probably help. Use the term scope (introduced in Section 2.2) to
describe the difference between the glosses.
2. Here is an unsatisfactory attempt to explain the meaning of not
good enough: not good means ‘bad or average’; enough means ‘suf-
ficiently’; so not good enough means ‘sufficiently bad or average’.
With the aid of brackets, explain why the phrase actually means
‘inadequate’.
3. Someone once said to me, “You and I are well suited. We don’t like
the same things.” The context indicated (and I checked by asking)
that the speaker meant to convey that we are well suited because
the things we don’t like are the same: but the sentence is in principle
ambiguous. Explain the ambiguity, and comment on unambiguous
alternatives.
4. Which of the following sentences entail which?
(a) The students liked the course.

(b) The students loved the course.
(c) The rain stopped.
(d) The rain ceased.
sense rel ations 31
5. Provide example sentences and write out a pattern of entailments

(comparable to (2.10a–d)) that establishes soundless, silent and noiseless
as synonymous.
6. Are awake and asleep complementaries? Give reasons for your answer.
Whether you have answered yes or no, how would you include half-
awake, half-asleep and dozy in an account of the meanings of awake and
asleep?
7. The hyponyms of footwear include shoes, sneakers, trainers, sandals, slip-
pers, boots and galoshes. Draw up a hyponym hierarchy for these words.
Is footwear the superordinate that you use for all of the hyponyms, or
are there other alternatives?
Recommendations for reading

Cruse (2011) provides a thorough discussion of oppositeness in meaning,
as well as hyponymy. Lappin (2001) provides a good article-length
introduction to formal semantics, and Saeed (2015) complements this
by dealing in greater detail with some relevant topics. WordNet is
available and browsable online at https://wordnet.princeton.edu/.
3 Nouns
Overview
Nouns form the majority of English words. They typically denote enti-
ties with rich and complex sets of properties. We can think of some of
these properties as being associated with hyponymy relations, as dis-
cussed in Section 2.5: because a dog is a type of mammal, we know that
dogs possess all the properties that mammals must have. But additional
sense relations apply to nouns, chief among them the has-relation,
which we discuss in this chapter. The has-relation captures the fact that
the things denoted by nouns can have parts, whether these are physical
or conceptual, and the question of which parts a noun has may be highly
relevant to its meaning. This chapter also discusses the distinction
between count and mass nouns: that is, between nouns that are treated
as denoting entities that can be separated and distinguished from one
other, and nouns that are not.
3.1 The has-relation

The everyday words square, circle and triangle are also technical terms in
geometry, where they have tight definitions. We might define a square
as a ‘closed figure which has four straight sides of equal length separated
by 90° angles’. The fact that a square has four sides is built into its defi-
nition. We can think of the entity a square as being associated with ‘four
sides’ by the has-relation. Like the relations discussed in Chapter 2, the
has-relation is important to semantics because it guarantees the truth of
certain entailments, as illustrated in (3.1).
That figure is a square ⇒ ‘That figure has four sides of equal

(3.1) a.
length’
That figure is a square ⇒ ‘That figure has four internal 90°
b.
angles’
32
nouns 33
Mathematical terms are somewhat atypical of natural language because

they have such unambiguous definitions. Trying to define nouns in
everyday use – part of the task of linguistic semantics – isn’t always
straightforward. In particular, the status of has-relations is sometimes
unclear. Consider (3.2).
Mary drew a face ⇒ ‘The picture that Mary drew includes
(3.2) a.
eyes’
Tom drew a house ⇒ ‘The picture that Tom drew includes a
b.
door’
If we think of the things that are denoted by the English word face,
the examples that spring to mind probably include eyes, a nose and a
mouth, among other features. That is to say, if we think of a prototype
of a face – a clear, central example of the denotation of face – it probably
has eyes, a nose and a mouth. Similarly, a prototype of a house probably
has a door, windows, a roof, and so on. (Conversely, there are numerous
features that many real faces and houses have but which are not likely to
be present in the corresponding prototype – say, a goatee, or a carport,
respectively.)
The inferences in (3.2) are based on the relations ‘a face has eyes’ and
‘a house has a door’. But these has-relations are really only valid if we
think in terms of prototypes. Something could be unambiguously a face
without having eyes, and unambiguously a house without having a door.
If Mary drew a picture of someone wearing shades, we would grant that
“Mary drew a face”, but the inference in (3.2) wouldn’t be valid.
In (3.1), then, we’re dealing with proper entailments of the kind
introduced in Section 2.1. But in (3.2) we have weakened the entail-
ments by making them conditional on prototypicality. It would be more
appropriate to write them down as something more like (3.3).
Mary drew a face ⇒ ‘If the picture that Mary drew is proto-
(3.3) a.
typical, it includes eyes’
Tom drew a house ⇒ ‘If the picture that Tom drew is proto-
b.
typical, it includes a door’
Essentially, it’s necessary to weaken things in this way because typical
English words are not as tightly defined as technical words like square
(in the geometric sense). But it is also useful for us to do this because
knowledge about prototypical properties is an important part of our
knowledge of what words mean.
In fact, for many nouns there are very few properties that are com-
pletely essential to the definition, even though a noun may have many
prototypical properties. This point has been argued by a number
of influential thinkers about language, so I want to dwell on it for a

moment. We can say that a square must have four sides, and anything
which does not have four sides cannot correctly be called a square.
But can we say anything similar about a face, or a house, or many other
things?
The classic example in this respect is the word game. Wittgenstein
(1953) argued that there were no features at all that were shared by all
the things we call a game. Not all games involve competition, or skill,
or physical ability; not all games have rules; not all games have playing
pieces or a scoring system. So, by this token, knowing that something
“is a game” doesn’t tell you anything about the has-relations that it pos-
sesses. Similarly, although we can think of a prototypical house as having
windows, a door and a roof, it would not automatically cease to be a
house if it had no windows, or lost its roof.
Despite this lack of obligatory properties, it still seems perfectly rea-
sonable to identify something as being a game or not being a game, or to
talk about a building being a house or not being a house. Consequently, a
useful idea is that word meanings are organised (at least in part) around
prototypes rather than obligatory properties – an idea called “prototype
theory”, which is particularly associated with Eleanor Rosch and col-
leagues. Calling something a game does not guarantee that it possesses
any one specific property associated with the prototypical game, but it
does suggest that that thing possesses enough of the prototypical game
properties, to a sufficient extent, to fall within the category. Similarly,
calling something a house doesn’t guarantee that it has windows, a door
or a roof, but it does strongly suggest (in the absence of indications to
the contrary) that it has most or all of these features, along with other
prototypical house features.
In the following sections we will look at some of the consequences of
the has-relation as applied to various aspects of noun meaning. In doing
this, we’ll generally be making some tacit assumptions about proto-
typicality when talking about word meanings, although we will also see
some of the limitations of this approach, for instance in the discussion of
the has-relation and hyponymy in Section 3.1.2.
3.1.1 Inferring existence from the has-relation

By appealing to the has-relation, we can infer the existence of entities
that haven’t been explicitly mentioned. Consider (3.4).
(3.4) Some kids walked up to a house, knocked on the front door and
ran away.
nouns 35
In (3.4), we have an indefinite article, a, and a definite article, the. A

noun phrase that first introduces its referent into conversation is usually
indefinite, whereas subsequent mentions of the same referent will
usually involve a definite noun phrase. This is why (3.5a) is reasonably
natural but (3.5b) is odd (assuming that a house denotes the same house
in both sentences) – (3.5b) attempts to refer to the already-established
referent with an indefinite noun phrase.
(3.5) a. Some kids walked up to a house. The house was old and
spooky.
b. Some kids walked up to a house. *A house was old and
spooky.
What we see in (3.4) is that, after mentioning a house, front door behaves
as though it has also already been mentioned: it’s acceptable to say the
front door, and it would be odd to say a front door. To put it another way,
if we say a house and then say the front door, the hearer is able to infer that
we probably mean ‘the front door of the just-mentioned house’. Clark
(1975) introduced the term bridging inference to describe this kind of
inference, as it involves connecting up the newly mentioned material to
that which has been mentioned before.
The pattern shown in (3.5) holds to some extent for non-prototypical
features. Earlier I mentioned that carport was a non-prototypical feature
that a house might have. In a context like (3.5), we can still use the with
carport, but it may also be fine to use a. That doesn’t work with front door,
as we see in (3.6). (The situation with door is a little more complicated
because a door might suggest that the house has multiple doors.)
(3.6) a. Some kids walked up to a house. The front door was to the
right of them.
b. Some kids walked up to a house. ?A front door was to the
right of them.
c. Some kids walked up to a house. The carport was to the right
of them.
d. Some kids walked up to a house. A carport was to the right of
them.
These examples show that we are conscious of the has-relations that
are associated with the nouns we mention, and these can influence
how we talk about things that we subsequently mention. Prototypical
parts may require the use of definite articles, whereas parts that are
not prototypical can be used with indefinite articles. To put it another
way, the hearer can reasonably infer from the speaker’s use of a house
that the door of that house exists, and they expect to encounter the
expression the door if the speaker intends to refer to that door. The
hearer may also be willing to accommodate the use of an expression
like the carport if they are willing to draw the inference that the house
has a carport, but as this is not part of the prototype, using a carport
is also fine. Chapter 10 will go into a little more detail about how the
use of definite and indefinite articles feeds into our understanding of
discourse.
3.1.2 Hyponymy, prototypes and the has-relation

The has-relation is obviously not quite the same thing as hyponymy,
discussed in Section 2.5, but these two relations interact in important
ways. Recall that hyponymy is about categories being grouped under
superordinate terms. To take another geometrical example, we could
say that square is a hyponym of quadrilateral, and so are rectangle, paral-
lelogram, kite, rhombus and trapezium. Hyponyms then “inherit” the parts
that their superordinates have (Miller and Fellbaum 1991: 206). By defi-
nition, a quadrilateral has exactly four sides – or, to put it another way,
it is connected to the attribute “exactly four sides” by the has-relation.
This same relation is inherited by square, rectangle, parallelogram, kite,
rhombus and trapezium. Quadrilateral is in turn a hyponym of polygon: by
definition, a polygon has straight sides. This relation, “has straight sides”,
is inherited by all the hyponyms of polygon, including quadrilateral and
all its hyponyms.
This kind of inheritance is important to our semantic knowledge. As
a result of it, when we learn about the hyponymy relations that a word
enters into, we automatically acquire knowledge about its has-relations.
However, although this idea is easy to apply to terms with clear defini-
tions, such as mathematical shapes, it becomes more complicated when
we are dealing with prototypes. The prototype of a hyponym does not
generally inherit all the has-relations from the prototype of its superor-
dinate. For instance, the Neolithic houses uncovered at Skara Brae in
Orkney had no windows: if we coin the neologism skara for a house that
resembles one of these houses, skara will be a hyponym of house, but the
prototypical skara will have no windows.
This observation applies not only to has-relations but also to other
properties. A classic example is that a property of a prototypical bird is
that it “can fly”. This property is inherited by most of the hyponyms of
bird, but of course not all: penguin is a hyponym of bird and the prototypi-
cal penguin cannot fly. We can’t even say with confidence that the prop-
erties of a prototypical superordinate will be inherited by most of its
hyponyms. Suppose we had 10,000 words for different kinds of penguin.
nouns 37
They would all be hyponyms of bird, but it wouldn’t alter the fact that a
prototypical bird can fly and a penguin cannot.
A hyponym will, of course, inherit all the obligatory properties,
including has-relations, from its superordinate(s). But, as we already
discussed in the case of game, some seemingly well-understood super-
ordinate categories don’t seem to have any obligatory properties – so
in these cases knowing about a hyponym–superordinate relationship
(for instance, knowing that oware is a game) doesn’t necessarily tell us
anything extra about the properties that the hyponym must have. Still,
it might be reasonable to think that prototypical properties of super-
ordinates are somewhat likely to be inherited by at least some of the
(prototypical instances of) their hyponyms.
3.1.3 Parts can have parts

Just as hyponyms can themselves have hyponyms, so the parts of an
object – the things that that object has – can themselves have parts. We
could say, for instance, that a suburb has houses, a house has windows, a
window has panes, and so on.
As far as the properties of these parts are concerned, remember that
again we’re dealing with the has-relation rather than hyponymy. Even
if properties are obligatory for a whole, they need not be obligatory of
its parts. A square obligatorily has four sides and four corners, but it is
not true that “each side has four corners” or “each corner has four sides”.
3.1.4 Spatial parts

For practical reasons, we’re often interested in talking about physical
objects which don’t necessarily have separate distinguishable parts.
Even so, we might want to talk about aspects of the physical nature of
these objects. We can do this by using words like top, bottom, sides, front,
back, and so on. Although not all physical objects have these particular
spatial parts, many different kinds of object do – heads, buses, screens,
wardrobes and pianos, to name but a few.
Some objects are associated with a particular physical orientation – a
bus falls into this category. There is no doubt about which end of the bus
is the front and which end is the back: these are intrinsic to the nature of
the thing. And arguably the same is true of top and bottom – even if a bus
is upside-down, there is a real sense in which the top of the bus still refers
to its roof. By contrast, other objects are not associated with a particular
orientation, or only partially associated with one. A tree has a top but
doesn’t have a front, for instance; and a rock doesn’t usually have any of
these things (assuming that the issue of which way up it happens to be is

a geological accident rather than a matter of design).
The lack of intrinsic orientation doesn’t mean that we can’t talk about
the front of a rock or the back of a tree. But it does mean that the use of these
terms depends on the context of utterance: that is to say, these expres-
sions are deictic. Like the examples of deixis we saw in Section 1.2.1, the
meaning of front or back depends on who is speaking and, in this case,
specifically how they themselves are located. A reasonable guess is that
what the speaker means by “the front of the rock” is the part nearest the
speaker, and what they mean by “the back of the tree” is the side of the
trunk that is furthest away from them.
In practice, things are often even more complicated than this. One
complication is that, although speakers usually use expressions like front
or back relative to where they are located, they may choose to use them
relative to where the hearer is located. A second complication is that
speakers sometimes use expressions like front or back relative to their
own position, or that of the hearer, even for objects which do have an
intrinsic orientation.
Suppose that there is a chair in the middle of the room that is pointing
in a northerly direction (that is, if you were sitting normally on it, you
would be facing north). Suppose that you are standing two metres to the
south of the chair, and I am standing two metres to the west of the chair,
and we’re both facing towards the chair. Now I say “Please place a coin
in front of the chair”. (This is very unlikely in real life, but psycholin-
guistics experiments are full of bizarre instructions like that.) There are
at least three ways in which you could interpret that request. Taking in
front of to refer to the intrinsic orientation of the chair, I’m asking you to
place a coin to the north of the chair. Taking it to be deictic based on the
speaker’s location, I’m asking you to place a coin to the west of the chair.
Taking it to be deictic based on the hearer’s location, I’m asking you to
place a coin to the south of the chair.
Essentially, in common with a lot of deictic expressions, many ref-
erences to the spatial parts and orientations of objects are technically
ambiguous. But, in common with a lot of technically ambiguous expres-
sions in language, much of the time we seem to use them perfectly well
without misunderstanding one another. The question of how we do this
is still an open one and continues to animate a lot of research.
3.1.5 Ends and beginnings

Long, thin objects, such as those listed in (3.7a), have ends. Sometimes
things are sufficiently orientable that it makes sense to distinguish two
nouns 39
different kinds of end, which we could broadly term beginnings and ends,
although for some objects we use more technical vocabulary to describe
these (a ship has a bow at one end and a stern at the other, a river has a
source at one end and a mouth at the other). Nouns denoting periods of
time, such as those listed in (3.7b), also have beginnings and ends, as well
as middles. By extension, we can think of events and processes that take
place over time as having beginnings, middles and ends, as in the exam-
ples in (3.7c). In Chapter 6, we will see some of the ways in which the
structure of events influences how we have to talk about them, from a
grammatical standpoint.
(3.7) a. rope, string, train, ship, road, river, canal
b. day, week, month, era, term, semester, century
c.
conversation, demonstration, ceremony, meal, reception,
match
3.1.6 Body part terms, metaphor and has-relations

Words for body parts are widely used in English (and many other lan-
guages) to describe many different kinds of object and their parts: the
foot of a mountain, the neck of a bottle, the mouth of a river, and so on. These
uses started out as metaphors, presumably because of some general
human tendency to interpret the world in terms of our own bodies,
although in many cases these have long since become established as
distinct senses of the relevant words (sometimes referred to as “dead
metaphors”).
There is more detailed discussion of figurative language, including
metaphor, in Chapter 9. Here I just want to make one point about how
has-relations work in some of these cases. Compare (3.8a), (3.8b) and
(3.8c).
(3.8) a. [I] hired a forest ranger on his weekend off, and started climb-
ing one of the faces of Mount Rushmore. (<https://www
.pbs.org/wgbh/americanexperience/features/rushmore
-hitchcock/>, retrieved 18 April 2022)
b. Between the mountains Machu Picchu and Huayna
Picchu form a face looking at the sky. (<https://www
.ticketmachupicchu.com/face-machu-picchu/>, retrieved
18 April 2022)
c. Hulking perilously above the village of Grindelwald in
Switzerland is the infamous north face of the Eiger. (<https://
www.ukclimbing.com/articles/features/north_face_of_the
_eiger_-_1938_route-3652>, retrieved 18 April 2022)
In (3.8a), face means ‘a face’ – well, more precisely, ‘a carving depicting

a face’. It is not metaphorical and does not refer to the geographical
feature of that name. We can expect that the face in (3.8a) inherits the
has-relations of a prototypical face: we expect the carving to depict eyes,
a mouth, and so on. By contrast, in (3.8b), face means something more
like ‘rock formation that happens to resemble a face’. This is more like
a metaphor: we understand the speaker not to be claiming that these
peaks literally constitute a face, but rather to be pointing out the resem-
blance. Here we might expect that some of the parts of a face will be
somehow discernible, but perhaps not all of them – just enough to make
the comparison plausible. And in (3.8c), face has a purely “dead meta-
phor” status. We don’t expect the Eiger to display anything recognisable
as the parts of a face: the word face is used simply because that is what
the side of a mountain is called.
In short, the has-relations of body part terms are applicable when
we are using the terms literally, inapplicable when we are dealing
with dead metaphors, and potentially partially applicable – but not to
be relied on – when we are dealing with new “living” metaphors. And
there is nothing special about body part terms in this respect; we can
expect to observe the same pattern for other expressions that can be
used metaphorically.
3.2 Count nouns and mass nouns

Another striking feature of noun meanings is that we can distinguish
two broad categories, which are represented differently in the grammar
of English: count nouns, such as loaf and coin, and mass nouns, such
as bread and money. As shown in Table 3.1, there are a lot of differ-
ences between count and mass nouns in terms of how they combine
with expressions of quantity, as well as the singular indefinite article a.
This reflects the obvious fact that count nouns denote distinguishable
whole entities whereas mass nouns denote undifferentiated, and hence
uncountable, substances.
As the examples in Table 3.1 illustrate, the division between count
and mass nouns cross-cuts other semantic differences – the words loaf
and bread enable us to talk about the same thing in a ‘count’ or ‘mass’
way. There are also words such as cake which can be freely used as
either count or mass nouns. As the table suggests, there are also cases
where what is usually a mass noun can be “coerced” into a count use – if
someone talks about many breads, you might infer that they mean ‘many
distinct varieties of bread’. There are also cases where what is usually a
count noun can be coerced into a mass interpretation – if someone says
nouns 41
Table 3.1 Distinguishing between count and mass nouns

Count nouns Mass nouns
This is a loaf. ?This is a bread.
This is a coin. *This is a money.
How many loaves are there? ?How many breads are there?
How many coins are there? *How many monies are there?
a large number of loaves ?a large number of breads
a large number of coins *a large number of monies
six loaves ?six breads
six coins *six monies
?some loaf some bread
*some coin some money
some loaves ?some breads
some coins ?some monies
*How much loaves are there? How much bread is there?
*How much coins are there? How much money is there?
there is egg all over the floor, we can easily interpret that as referring to the
contents of broken eggs rather than one or more unbroken individual
eggs.
In essence, where we have both count nouns and mass nouns for
the same thing, the speaker has a choice about how they are going to
portray reality: as containing individual objects like loaves or coins, or
as containing undifferentiated “stuff” like bread or money. Of course,
this does not mean that when a speaker uses a mass noun they become
incapable of distinguishing its parts: when we talk about furniture we can
still tell the difference between tables and chairs (and between individual
chairs), but we choose not to emphasise that point.
Hyponymy (and incompatibility) exists among mass nouns just as
among count nouns: velvet, corduroy, denim and so on are incompatible
hyponyms of the mass noun cloth, and are themselves mass nouns. In
principle, we might expect mass nouns not to enter into has-relations,
because homogeneous substance is not separable into distinct parts: but
this is perhaps not a clear-cut issue, simply because (as remarked above)
whether we treat something as mass or count doesn’t necessarily say a
great deal about its physical reality. If we agree that cloth has threads, then
this is a has-relation involving a mass noun, and it is inherited by the
hyponyms of cloth in the usual fashion.
Summary
In this chapter we have considered some of the characteristic proper-
ties of nouns, with particular emphasis on the has-relation and the
interplay between this and other sense relations. We have seen that
the has-relation is a potentially powerful tool for learning about the
entailments of hyponyms of nouns, but this is sometimes a complicated
matter because has-relations are often not obligatory, and prototypical
properties are not always inherited by hyponyms in the same way as
obligatory properties. We have also considered the spatial properties of
objects and some of the ways in which language allows us to talk about
these properties. And we considered the distinction between count
nouns and mass nouns as a way of portraying the world: labelling with
a mass noun treats the thing referred to as a homogeneous substance
without distinct parts, and this has consequences for how we can talk
about this referent. Rather like the choice of how we talk about spatial
configuration, the distinction between mass nouns and count nouns
does not correspond to a crisp distinction between two kinds of things
that exist in the world, but rather between two different ways in which
we can relate to the things in the world through language. We will
encounter similar ideas in the following chapters, which deal in turn
with adjectives and verbs.
Exercises
1. What parts does a prototype shoe have? Do those parts have parts?
2. If we wanted to describe the meanings of some spatial part words, we
might say something like this:
“The top of a thing is one of its sides, the side that is uppermost. The
bottom of a thing is one of its sides, the side that is down. The front is
one of the sides, the side that faces forwards. The back is one of its
sides, the side that faces away from the front.”
If this description is correct, what sense relations hold between the

words side, top, bottom, front and back? Give reasons to support your
answer.
3. Paper, glass and cheese are ambiguous between a count sense and a
mass sense.
(a) Devise a pair of example sentences for each of them that clearly
brings out the count–mass difference.
nouns 43
(b) Find some hyponyms for each of the words in each of its senses.
Use these to comment on the systematic difference in meaning
between the count and mass interpretations of these words.
4. In the question “Have you ever eaten rabbit?”, what difference does
the lack of an article (rabbit instead of a rabbit) make to the interpreta-
tion of the noun rabbit?
5. Why might we interpret left differently when we are describing
something as being to the left of the chair versus to the left of the stool?

For prototype theory, Mervis and Rosch (1981) give a detailed over-
view of some of the major ideas and their consequences. Johannsen
and De Ruiter (2013) and Coventry et al. (2018) report useful work
on how speakers choose spatially referring expressions; Weisberg and
Chatterjee (2020) specifically examine how ultimate Frisbee players do
this. Kearns (2011) is an excellent source for more about differences in
meaning between mass and count nouns.
4 Adjectives
Overview
In Chapter 2 we already talked about a number of the sense relations
that adjectives (along with other parts of speech) can enter into: syn-
onymy, complementarity, antonymy, incompatibility and converse-
ness. In this chapter, we focus on two more specific aspects of adjective
meaning – their tendency to express different degrees, or levels, of the
quantity that they denote, and the way they combine with noun mean-
ings. In both cases we will also see how contextual factors influence the
interpretation of adjectives.
4.1 Gradability
In Section 2.4, we talked about several sense relations that apply to
adjectives. One of them was converseness, as illustrated by (4.1).
(4.1) a. France is bigger than Germany.
b. Germany is smaller than France.
Converseness relies on the existence of comparative forms like bigger
and smaller, and these comparatives are possible because the adjectives
they are based on are gradable. That is to say, the language allows us to
express different degrees or levels to which nouns possess the qualities
that these adjectives denote. Some adverbs are also gradable, as we see
with quickly in (4.2a), but nouns and verbs generally are not. To make
any sense of (4.2b) we have to interpret fish as fishlike (an adjective), and
to make sense of (4.2c) we have to interpret more swims as swims more,
which isn’t quite the same thing.
(4.2) a. A salmon swims more quickly than a human.
b. *A salmon is more fish than a human.
c. *A salmon more swims than a human.
44
adjectives 45
The use of the comparative -er suffix in (4.1) is an indicator that we are
dealing with a gradable property. So is the use of more in (4.2a), as well
as the use of than in both (4.1) and (4.2a). Other examples of morphemes
that indicate the presence of gradability are shown, underlined, in (4.3).
(4.3) a. He is the rudest person I have ever met.
b. These questions are too difficult.
c. How long is this going to take?
d. The weather was very hot.
e. That is a good enough set of examples.
The adjectives in the examples in (4.3) are all members of antonym
pairs: rude–polite, difficult–easy, long–short, hot–cold, good–bad. Each adjec-
tive denotes what we can think of as a region towards one or the other
end of a scale. On the scale of “difficulty”, difficult denotes values towards
the high end and easy denotes values towards the low end. Difficult and
easy are not complementaries, in the terms introduced in Section 2.4,
because some things are neither difficult nor easy on the “difficulty” scale
but somewhere in between.
Curiously, these pairs of antonyms don’t behave quite the same way
within the language system. Consider (4.4).
(4.4) a. This will take a long time.
b. This will take a short time.
c. How long is this going to take?
d. *How short is this going to take?
In this particular context, if we want to ask a question about where we
are on the scale of “duration”, we can only use long – the option of short
seems ungrammatical. More often we have a situation in which one
member of the antonym pair is used to ask neutral questions and the
other is not, as in (4.5).
(4.5) a. How old is that player?
b. How young is that player?
(4.5b) would be an odd question to ask about a professional footballer
who is, say, thirty years old. It might be a reasonable question in a
context in which the player in question is or appears to be unusually
young, and it might be a reasonable question if the player ought to
be young (for instance, because it is a junior tournament) but doesn’t
appear to be. But (4.5a) is acceptable in any of these situations. In fact,
what seems to be happening here is that old is the preferred adjective
for talking about the “age” scale (as in I am thirty-eight years old) and if a
speaker chooses to use young instead, they invite the hearer to infer that
there must be some kind of special reason for their choice. In Section
8.2.4, we will discuss an account of pragmatic inference that proposes
a more general explanation of how meanings like this arise when a
speaker uses an unusual or disfavoured expression.
4.1.1 Non-gradable adjectives

Not all adjectives are gradable. In particular, the members of com-
plementary pairs are sometimes resistant to being interpreted this
way. In Section 2.4, I gave six examples of complementary pairs:
moving–stationary, same–different, right–wrong, true–false, intact–damaged
and c onnected–disconnected. All these pairs, by definition, partition all the
entities that they apply to into two, non-overlapping classes: everything
that is moving is not stationary, and vice versa. However, while things that
are moving move at varying speeds, we can’t really say that something
that is moving faster is “more moving” than something else, and we can’t
say that two stationary objects differ in “how stationary” they are. So
the moving–stationary pair separates objects into two categories but then
doesn’t really discriminate within those categories at all.
By contrast, complementary pairs like same–different do not follow
this pattern. In this case, if two things are the same, that’s absolute – we
can’t quantify “how same” they are. But different is gradable: chalk and
cheese are more different than apples and pears. In short, complementary
pairs vary quite considerably in how gradable they are, as shown by the
examples in (4.6).
(4.6) a. ?The jury heard evidence that was too false to be credible.
b. ?The jury heard evidence that was too true to disbelieve.
c. ?That is the rightest thing any politician has said on the topic.
d. ?That is the wrongest thing any politician has said on the
topic.
e. ?How intact is the vase?
f. How damaged is the vase?
g. I feel very connected to the community.
h. I feel very disconnected from the community.
Other kinds of adjectives can also be thought of as non-gradable. In
(4.3a), we had an example of a superlative, rudest. Superlatives denote
extreme ends of scales, and this makes them resistant to being further
graded: their position on the scale is not up for negotiation, so we have
no need to clarify it further. In the case of *bestest or *more fastest, we
could think of this as just reflecting a morphological or syntactic rule
that stops us combining a superlative or comparative form with another
adjectives 47
superlative or comparative form. But actually we see a similar pattern

with other adjectives that denote extreme values, as in (4.7): these are
also difficult to use gradably (and when we try, it might just be under-
stood as a form of emphasis that doesn’t really change the meaning, as
in (4.7c)).
(4.7) a. *The presentation was very excellent.
b. ?The weather was too freezing.
c. ?She is extremely unique.
Conversely, there are some things that can be done with these extreme
adjectives that can’t be done with ordinary gradable adjectives. For
instance, they can be modified with “maximiser” adverbs such as absolutely
and completely, whereas gradable adjectives cannot, as shown in (4.8).
(4.8) a. *The shrink wrapping was absolutely thin.
b. *Her performance was completely good.
c. You’ll look completely different with your hair restyled.
d. What you say is absolutely true.
e. Digital sound reproduction is completely perfect.
f. The weather has been absolutely freezing.
Even when extreme adjectives aren’t gradable, they can still influence
how we talk about gradable quantities. For instance, if a speaker has the
possibility of describing something as excellent, that may influence the
circumstances under which they will describe that thing as merely good
– and hearers are aware of this. The possibility of drawing additional
inferences based on the speaker’s failure to use a (relatively) extreme
description turns out to be very useful in communication. There will be
more on that in Section 8.2.2.
4.2 Combining adjective meanings with noun meanings

When an adjective modifies a noun, how do the meanings of the adjec-
tive and the noun combine? This seems like a rather trivial matter, and
in some cases it is – but there are also cases where it is surprisingly
complicated. In this section I will just introduce a couple of cases that
illustrate that idea.
Let’s dispose of the relatively trivial case first. If we consider an
example like green bicycles, we could say that this adjective–noun combi-
nation denotes all the things that are both green and bicycles. If we use a
Venn diagram to represent the denotations of green (things) and bicycles,
as in Figure 4.1, then exactly those things that fall into the intersection
of the two ellipses are green bicycles.
green green
bicycles
things bicycles
Figure 4.1 The simplest case of an adjective modifying a noun is like the
intersection of sets
This seems to correspond reasonably well with our intuitions about

what green bicycles means. So, for cases like this, we can think of the
combination of an adjective and noun meaning as simply involving
the intersection of two sets – the set of things denoted by the noun and
the set of things to which the adjective applies. For that reason, we call
adjectives like this intersective adjectives.
Things get more interesting, and challenging, when we consider
adjectives that are not intersective. In the following subsections we’ll
briefly consider three kinds of non-intersective adjectives: those which
end up with potentially broader denotations than we would expect on
intersective grounds; those which end up with a denotation that doesn’t
overlap with the intersection at all; and those which are crucially
dependent on context for their interpretation.
4.2.1 Non-intersective adjectives with broader denotations

One slightly unexpected category of non-intersective adjectives
comprises words like alleged, assumed, supposed, and so on. When these
combine with a noun, the resulting phrase has a denotation that
potentially includes entities that fall within the usual denotation of the
noun as well as entities that do not. Consider the underlined phrases in
(4.9).
(4.9) a. Turkish authorities have cracked down on alleged sup-

porters of the U.S.-based Muslim cleric Fethullah Gülen
following a failed coup in 2016. (<https://greekcitytimes
.com/2021/12/03/17-turkish-asylum-seekers/>, retrieved
20 April 2022)
b. After months of boring assumed winners, the Oscars race
has been shaken up recently by shocker developments no
one saw coming. (<https://nypost.com/2022/03/26/oscar
-predictions-2022-who-will-win-vs-who-should-win/>,
retrieved 20 April 2022)
adjectives 49
c.
They lament the failure to prosecute any of the sup-
posed masterminds behind the insurrection. (<https://
www.washingtonpost.com/opinions/2021/11/10/merrick
-garland-capitol-riot-justice-department-pundits/>,
The speaker of (4.9a) is clearly signalling the possibility that not all
the people cracked down on are actual supporters of Gülen. In (4.9b),
the speaker is talking about people (and movies) that are or have been
assumed to achieve the status of winners in the future. And in (4.9c), the
speaker seems to be expressing scepticism towards the idea that any of
the people involved could really be called masterminds.
In general, then, this class of adjectives is clearly non-intersective.
What we get when we compose alleged and supporter is not ‘a special kind
of supporter, specifically one that also has the property of being alleged’.
Rather, we get a meaning along the lines of ‘someone who has been
alleged (by someone else) to be a supporter’. A similar pattern holds for
assumed and supposed. All these examples are compositional, in the sense
that the meanings are predictable from the meanings of the adjectives
and the meanings of the nouns: it’s just that the denotation of alleged sup-
porter is not a subset of the denotation of supporter.
The three examples above all involve adjectives that are derived
from verbs, and all of them result in a meaning that implicitly involves
the action of that verb (someone alleging, assuming or supposing
something). But there are also non-intersective adjectives that are not
directly based on verbs, as we see in (4.10).
(4.10) a. Miller continued to lead, looking a likely winner for 16 laps.
(<https://www.cycleworld.com/story/motorcycle-racing
/2022-grand-prix-of-the-americas-motogp-report/>,
b. If Tiger is at 2020 levels, then by that standard he would be
a possible winner but a long shot. (<https://fivethirtyeight
.com/features/if-tiger-woods-tees-off-at-the-masters-hell
-be-playing-to-win/>, retrieved 20 April 2022)
c. Giannis Antetokounmpo may be playing like an MVP, but
he’s an unlikely winner at this point. (<https://eu.jsonline
.com/story/sports/nba/bucks/2022/04/01/race-nba-mvp
-award-nuggets-nikola-jokic-leads-76-ers-joel-embiid
-and-bucks-giannis-antetokounmpo/7243205001/>,
In these examples, none of likely winner, possible winner or unlikely winner
is a specific kind of winner. In each case, what the adjective contributes
is some kind of evaluation of the probability of the individual being

talked about actually going on to be a winner of the relevant kind. The
class of possible winners for a given event is likely to be much bigger
than the class of winners. So, once again, these are non-intersective.
We use the word modality to describe the kind of meaning that these
particular adjectives express, and it will be discussed in more detail in
Section 7.1.
4.2.2 Privative adjectives

In the previous subsection we talked about the adjective unlikely, which
seems to suggest that the entity being talked about probably doesn’t
belong to the denotation of the following noun. However, we don’t
seem to be able to use impossible in quite the same way, as shown in (4.11)
– perhaps if we are using the word impossible, we are not licensed to call
the person a winner in the same breath.
(4.11) *Tiger is an impossible winner at this point.
Surprisingly, though, there are adjectives which do explicitly rule out
the possibility that the entity in question belongs to the usual denotation
of the following noun. These are sometimes called privative adjectives,
and the most widely discussed example is fake, as in (4.12).
(4.12) World’s costliest painting Salvator Mundi is a fake
Leonardo da Vinci, claims documentary. (<https://www
.aninews.in/news/world/europe/worlds-costliest-painting
-salvator-mundi-is-a-fake-leonardo-da-vinci-claims
-documentary20210420115925/>, retrieved 20 April 2022)
Here the use of fake Leonardo da Vinci explicitly excludes the possibil-
ity that the painting is due to Leonardo da Vinci. The denotation of
Leonardo da Vinci here includes all the works by Leonardo, rather than
the individual himself (that is to say, this is an example of the figure of
speech called metonymy, which is discussed a little more in Section
9.3). So, ostensibly, the denotation of fake Leonardo da Vinci doesn’t even
overlap with the denotation of Leonardo da Vinci.
Of course, the noun still contributes meaning to the expression, even
in this privative case. A fake Leonardo is not the same thing as a fake Picasso
or a fake Warhol. If you were to paint a version of Salvator Mundi, it would
be odd to describe the result as a fake Leonardo, unless you made some
special effort to pass it off as Leonardo’s work. So in this case the noun is
contributing the meaning of what the work is purporting to be, and the
adjective is contributing the meaning that it isn’t a genuine example.
adjectives 51
There is a possible analysis on which this kind of adjective isn’t really

privative at all, though: that would be if we take Leonardo not only to
denote all the works by Leonardo himself, but also those that closely
resemble those works or are claimed to be genuine examples. Then we
can separate the category of Leonardos into fake Leonardos and genuine
Leonardos. This analysis is not particularly intuitive as applied to fake,
but perhaps makes sense if we think about words like former, as in (4.13).
(4.13) Barack Obama is a former President of the United States.
Here, we could think of former as privative if we understand that Obama
is not the President of the United States, and cannot be President again
(because of term limits); or we could think of it as simply intersective if
we take President of the United States to denote anyone who has held that
office at any point in time.
4.2.3 Relative adjective meanings

In Section 4.2, we discussed green as an example of an intersective
adjective. Green, in its sense as a colour word (rather than its sense of
‘environmentally conscious’ or that of ‘naïve’, among others), is reason-
ably stable in meaning from context to context, although what counts as
green may vary from noun to noun (must the object be completely of that
colour, or only partially?). However, a lot of adjectives have meanings
which are much more context-dependent than this, as in (4.14).
(4.14) a. Sucrose is a big molecule.
b. The swimming pool was the size of a small lake.
If we wanted to describe big or small as intersective, we would have to
say that the denotation of big molecule is anything that is both a member
of the set of ‘big objects’ and a member of the set of ‘molecules’, and
that the denotation of small lake is anything that is both a member of
the set of ‘small objects’ and a member of the set of ‘lakes’. But this is
obviously wrong – it’s hard to imagine that the category of ‘big objects’
contains any individual molecules, or that the category of ‘small objects’
contains any lakes. Moreover, it seems intuitively obvious that what
the speaker of (4.14b) means is ‘the swimming pool was big’ – so clearly
the same thing can be small according to one standard and big according
to another standard.
What seems to be happening here is that we are interpreting the
adjectives big and small in a way that is relative to the noun they are
modifying. These adjectives (and many others, such as wide, narrow, long,
short, and so on) contribute a meaning that restricts the denotation of
the noun to examples that possess the relevant property to a sufficient

degree. These are sometimes described as relative intersective adjec-
tives. They are intersective only in the weaker sense that the denotation
of the adjective–noun combination is a subset of the denotation of the
noun alone: all big molecules are molecules, all small lakes are lakes, and so on.
But, in order to figure out precisely what subset we are talking about,
we have to consider the (perhaps rather idiosyncratic) way in which the
specific adjective combines with the specific noun. Clearly this is possi-
ble – we use relative intersective adjectives all the time without seeming
to be misunderstood very often – but the question of how exactly it
works is quite a difficult one.
There are also plenty of adjectives that have different senses which
attract different kinds of interpretation. Consider (4.15).
(4.15) She’s a good politician.
As applied to people, good has a possible sense in which it means some-
thing like ‘morally upstanding’; but in the context of (4.15), it also has a
possible sense in which it means something like ‘effective at doing the
job’. If what we mean by (4.15) is that the person being described is both
a politician and a morally upstanding person, then we are using good
as an intersective adjective. If we mean that the person is effective in
their job as a politician, then we are using good as a relative intersective
adjective.
In practice, we can sometimes exclude the possibility of fully inter-
sective interpretations by appeal to common sense: if I describe some-
thing as a good shovel, you can assume that I’m not trying to ascribe moral
qualities to an inanimate object, and for that reason I must mean that
it’s an effective implement for use in the act of shovelling. However, we
may sometimes need more sophisticated strategies than this for resolv-
ing potential ambiguities of this kind.
Summary
This chapter introduced some additional features of adjective meaning.
The characteristic of gradability, which applies to some adjectives,
relates to their ability to convey different levels or strengths of a
property, and has some interesting consequences for how we can and
cannot use and further modify those adjectives. We can also distinguish
adjectives by how their meanings combine with those of the nouns they
modify. In some cases, this takes place in the obvious way, but in other
cases the interaction of meanings is surprisingly complex. Indeed, a
single adjective may have multiple senses which modify the noun in dif-
adjectives 53
ferent ways, creating ambiguity. We are able to deal with some of these
ambiguities, however, by appealing to context and world knowledge, a
theme which will recur throughout this book.
Exercises
1. The adverb quite has two different meanings when it modifies adjec-
tives. In one sense it is a “downtoner”: quite friendly can be glossed
as ‘moderately friendly’. In another sense it is a “maximiser”: quite
exceptional is synonymous to ‘exceptional to the fullest extent’. More
specifically, quite is a downtoner with words such as clever, late, small
and unusual, but a maximiser with right, finished, impossible and alone.
What is the relevant difference in types of meaning between these
classes of words?
2. Which of the following adjectives would normally yield biased ques-
tions if you inserted them into the frame “How [adjective] is/was x?”,
and which would normally yield unbiased questions?
old young
rude polite
unpalatable tasty
weak strong
3. Which of the following phrases can be explained as examples of

intersective adjective use, and which cannot? Why?
royal visitor
royal correspondent
heavy eater
wise fool
4. Proxima Centauri is a small, cool, red star located near our Solar System. In
fact, it is the closest other star to the Sun. Comment on the meaning of the
adjectives in the context of these two sentences.

The relevant sections of general grammars of English such as Quirk
et al. (1985) and Huddleston and Pullum (2002) are rich sources for
detailed accounts of adjective meaning. Kennedy (2012) provides a
very readable discussion of adjective use and meaning, and some of
its broader philosophical implications. There are many more complex
papers on the semantics of gradable adjectives (for instance, McNally

2011), which get into some very interesting questions about the nature
of the underlying “scale of measurement” for these adjectives, but these
papers tend to assume a lot more prior knowledge.
5 Verbs
Overview
This chapter is about verb meanings. One way of thinking about what
makes up a sentence such as Robby brought me the news is that the verb is
at the heart of it and it “says something about” the referents of the noun
phrases that surround it (here Robby, me and the news). More specifi-
cally, we could think of the verb as explaining how these noun phrases
relate to one another. Although we can think of other parts of speech
in a similar way, it is customary to consider the verb as the semantic
centre of the sentence (or clause). In this chapter we’ll consider how
different verbs impose different requirements on the rest of the clause,
specifically in terms of how many other referents must be introduced,
and we’ll look at how the verbs cause different roles to be associated
with those referents.
5.1 Verb types and arguments

One way of categorising verbs is according to how many referring
expressions they interrelate: that is to say, how many noun phrases
(NPs) they require. Usually they require one, two or three, as illustrated
by the examples in (5.1), where the NPs are underlined.
(5.1) a. Billy sings. (one noun phrase)
b. Ella admires Beethoven. (two noun phrases)
c. I offered her a scone. (three noun phrases)
In practice, verbs do not necessarily require NPs in all these posi-
tions. Some verbs will accept preposition phrases (PPs) in certain posi-
tions, such as to her in (5.2a). And sometimes a position can be filled by a
more complex kind of structure, as in (5.2b–c).
(5.2) a. I offered a scone to her.
b. The report confirms that spring came early.
55
c.
That the daffodils are blooming confirms that spring came
early.
In (5.2b) and (5.2c), positions are filled by embedded clauses intro-
duced by that. A clause is a structure that usually has a verb of its own
and can carry a proposition on its own: for example, spring came early
contains the verb came and expresses a proposition about the start of
the season. However, in (5.2b–c), this clause is not free-standing but
has been embedded into another clause as the object of the verb confirm.
(5.2c) also contains another that-clause embedded in subject position.
The word that is one of the markers made available by English grammar
to mark a clause as embedded.
We use the term argument to cover all the kinds of obligatory con-
stituents that verbs require, whether they are NPs (the report), PPs (to
her) or embedded clauses (that the daffodils are blooming). (5.2a) has three
arguments; the main clauses in (5.2b–c) each have two arguments.
A verb that requires both a subject argument and a direct object
argument, such as admires in (5.1b), is called transitive. We could
add an additional argument to (5.1a) – for example, in the bath – but
the additional argument is not necessary for the sentence to be gram-
matical. (5.1a) is therefore intransitive. Other verbs, such as offer in
(5.1c), obligatorily require both a direct object and an indirect object:
we cannot normally say *I offered her or *I offered a scone (unless the
potential recipient is contextually obvious). (5.1c) is therefore ditran-
sitive. In (5.2a), we see a slightly different version of this in which the
indirect object is replaced by a PP, with a corresponding change in
word order. It is sometimes useful to distinguish one further class of
sentences, copular sentences such as John is my brother (or That’s what
I’m talking about, as in Chapter 1), where the verb is is used to “predicate
properties of a subject” (that is, label the subject as having particular
properties).
We can delve a bit further into the different categories of verbs.
Intransitives have been divided into two rather opaquely named kinds
(Trask 1993: 290–2) on the basis of the type of subject argument that the
verbs require, as follows:
• An unergative verb requires a subject that is consciously responsible
for what happens. Walk is such a verb, and tourists walk through the eco
park is an unergative clause. A good test for unergativity is accept-
ability with the adverb carefully, because taking care is only a pos-
sibility when an action is carried out deliberately.
• An unaccusative verb is an intransitive verb in which the subject is
affected by the action but does not count as being responsible for it.
verbs 57
Grow, drop and die are verbs of this kind. These verbs do not occur
readily with adverbs such as carefully: *Mort carefully died.
From the hearer’s point of view, knowing that a verb is unergative
potentially tells you something about the subject, namely that it is
something capable of conscious action. Of course, this is usually rather
trivial: in the example above, we already know that tourists are capable
of conscious action. However, we can also distinguish categories of
verbs which give rise to richer and more elaborate kinds of entailment,
as we shall see in Section 5.2.
5.1.1 Other kinds of arguments

In a way, the notion of arguments is more general than just ‘the obliga-
tory constituents that verbs require’. I mentioned in the introduction
to this section that we could think of other parts of speech as fulfilling
similar functions to verbs, and I meant that specifically in terms of how
they relate different arguments. Consider (5.3).
(5.3) a. Maya sent Bill the letter.
b. The letter moved from Maya to Bill.
We would typically analyse (5.3a) as a ditransitive sentence with the
verb sent, the subject Maya, the direct object the letter and the indirect
object Bill. But (5.3b) relates the same arguments in quite a similar way,
even though the verb moved is intransitive and both the arguments from
Maya and to Bill are optional. And abstractly there is nothing to stop us
from saying that (5.3b) is a sentence in which from Maya has two obliga-
tory arguments, a verb (expressing the nature of the motion from Maya)
and an NP (identifying the thing that moves), plus an optional argu-
ment (the destination). Sometimes it can be useful to think about other
parts of a sentence or clause in terms of the arguments that are needed
to combine with them in order to make a complete sentence. However,
usually we focus on the verb and its arguments, and that will be the case
in the rest of this chapter.
5.2 Causative verbs

In Section 2.2, we talked about the principle of compositionality: the
idea that the meaning of a complex expression is determined by the
meaning of its parts and the way those parts are put together. Focusing
on verbs, we can think of the meaning of a clause as being determined
not only by the meaning of the verb itself but also by the kinds of
a rguments that it has. The arguments can in fact reshape our impression
of the meaning that the verb contributes, as we see in (5.4).
(5.4) a. Robbers spray victims to sleep. (Fiji Post, 1 June 1995)
b. Yet Birthday’s reckless spontaneity has been focus-grouped
to death. (<https://www.theguardian.com/music/2020
/feb/21/tracks-reviewed-grimes-us-girls-jake-shears-anne
-marie-otoboke-beaver>, retrieved 22 April 2022)
In the case of (5.4a), we would normally understand spray as meaning
something like ‘distribute a substance in aerosol form’. We would expect
the clause to have an obligatory subject, as it does – robbers – and also
a direct object, identifying what the substance was sprayed at or onto,
which it does – victims. We might also see the substance itself being
identified, which it is not in (5.4a). Instead, we have this unexpected
argument to sleep. To make sense of this, we have to assume a different
interpretation for spray, one in which it is causative, because the reading
of to sleep as indicating the effect of the spraying is the only way to make
sense of its presence here as an argument.
Similarly, in (5.4b), we might expect the verb focus-group (meaning
something like ‘expose something to a focus group’) to have two argu-
ments, a subject and a direct object: here it occurs as a passive, so it has
no explicit subject (for more on passives, see Section 10.3.3). But it still
has two arguments, the direct object Birthday (denoting a song with
that title) and to death. Again, this suggests a causative interpretation of
focus-group, with the speaker of (5.4b) understood to be saying that, as a
consequence of being exposed to a focus group, the song Birthday has –
in some metaphorical sense – died.
The meaning expressed by a causative sentence is, in general, that
a situation is caused by whatever the subject noun phrase refers to, and
the situation is described by the additional argument. In (5.4a), that
situation is The victims sleep, and in (5.4b) it is Birthday has died. The
sentences in the left-hand column of Table 5.1 are further examples of
causatives, with each one entailing the sentence to its right.
The sentences on the right in Table 5.1 describe states or events,
without making reference to their causes. The causatives on the left
share the following properties:
• They include a causative verb: that is to say, one that expresses cau-
sation. In these examples, we have the causative verbs make, get, force,
cause, have, and prevent. Prevent is a negative causative.
• The subject of the causative sentence is used to refer to whatever
entity – concrete or abstract – brings about the situation described
by the sentence on the right.
verbs 59
Table 5.1 Examples of causative sentences with an entailment from each

Causatives Entailments
The thought made her gleeful. She was gleeful.
The children got the kite to fly. The kite flew.
Bad weather forces us to cancel the picnic. We are cancelling the picnic.
His inexperience is causing the decisions to go The decisions are going
unactioned. unactioned.
I had the students study this article. The students studied this article.
The lock prevented him from opening the door. He did not open the door.
• The causative has an argument which somehow carries the same

proposition as the sentence to its right. The sentence affirms that that
proposition is brought about by some cause.
The verb cause itself is arguably a superordinate for the other causative
verbs in Table 5.1: it expresses the vaguest and most general sense of
causation. For example, we can think of force as a hyponym of cause that
means more specifically something like ‘cause an unwanted or resisted
consequence’. In fact, as we will see in Section 8.2.4, using cause in
place of a more specific verb can sometimes turn out to have interest-
ing consequences, in terms of the meaning that the speaker is taken to
convey.
Semantically, causative verbs have a minimum of two arguments, one
denoting the causer and one denoting the caused state or event. We can
call this latter argument the embedded situation. The embedded situ-
ation itself contains arguments. For the causative sentence I made John
give Mary the book, the embedded situation ‘John gave Mary the book’ has
three arguments, John, Mary and the book.
5.2.1 Identifying embedded situations

The exact way in which the embedded situation is expressed syntacti-
cally in a causative sentence is beyond the scope of this book. We might
think of things like her gleeful or the kite to fly as being special kinds of
clauses that express the embedded situation, but it’s controversial as
to whether we can call them clauses, because they lack tensed verbs of
their own. But for our purposes it’s enough to say that these expressions
enable us to recover the embedded situations, such as ‘she was gleeful’
or ‘the kite flew’. In this latter case, the proposition receives its past tense
from the verb got in the causative sentence.
In other cases, embedded situations are even less clearly signposted.
We can think of certain sentences as involving “understood” embedded
situations, even when these are not made at all explicit. (5.5) is a poten-
tial example of this.
(5.5) The staff nurse gave Lucinda a key for the week.
As Tenny (2000) points out, for the week in (5.5) doesn’t modify the verb
give – the speaker of (5.5) is not saying that the action of ‘the staff nurse
giving Lucinda a key’ lasted the whole week. Instead, what for the week
modifies appears to be an understood embedded situation: that is, the
situation in which ‘Lucinda has a key’. That is to say, (5.5) is a concise
way of expressing the meaning that ‘the staff nurse gave Lucinda a key,
with the intention of causing Lucinda to have the key for the duration
of the week’.
There is some additional evidence that embedded situations which
are not syntactically visible might figure in language use and interpreta-
tion. Consider the situation described in (5.6).
(5.6) a. When the company started, their main selling point was
good value.
b. They then quietly doubled the price of their product.
c. After a public outcry, the company lowered the price again.
The use of again would usually be appropriate in a case where some-
thing had happened before – technically, it is often said to presuppose
that the thing in question happened before. (There is more on presup-
positions in Section 8.4.) But, in (5.6c), again appears to refer to the
event in which ‘the company lowered the price’, even though that didn’t
happen before. This use of again is called restitutive: it involves the
restoration of a previously existing state of affairs (Tenny 2000), in this
case the one in which the company sells its product at a lower price. A
good way of capturing this would be to say that again is actually refer-
ring to the embedded situation in which ‘the company sells its product
at a lower price’. The existence of this situation seems to be part of the
meaning of (5.6c), even though it is not explicitly expressed there.
Table 5.2 shows some more examples of causatives that involve just a
single clause, but which nevertheless entail a proposition about a caused
situation (as shown in the right-hand column). We could think of these
as involving various kinds of understood embedded situations, which in
these cases are bound up in the verb semantics: although feed and eat are
different verbs, an act of feeding naturally involves an embedded situa-
tion which contains eating, and so on.
The last two lines of Table 5.2 show causatives entailing unaccusa-
tives which use the same verb form: Gardeners grow vines ⇒ Vines grow;
He broke a bone ⇒ A bone broke. When we have two different types of
verbs 61
Table 5.2 Three kinds of one-clause causative with an entailment from each
Causatives Entailments
different verbs (e.g. feed–eat)
She fed the baby some mashed banana. The baby ate some mashed banana.
The bank has lowered its interest rate. The bank’s interest rate dropped.
Drought killed the lawn. The lawn died.
morphologically related verbs and adjectives (e.g. enrich–rich)
Nitrogen spills have enriched the soil The soil is rich here.
here.
The graphic artist enlarged the logo. The logo became larger.
His job deafened Dougie. Dougie became deaf (to an extent).
same verb form used causatively and non-causatively (e.g. walk–walk)
The guide walks tourists through the Tourists walk through the eco park.
eco park.
The gardener grew several vines. Several vines grew.
He broke one of his bones. One of his bones broke.
e xpression with the same form, we sometimes call this conversion or

zero derivation. Here, an unaccusative verb is formed by the conver-
sion of a causative verb (or vice versa) without any change in form.
(Conversion can also take place between other word classes: we saw an
example earlier, with focus group being a verb derived from the exist-
ing noun.) Fellbaum (2000: 54) notes that many English verbs exhibit
just this kind of causative-to-unaccusative entailment with conversion:
some common examples are bend, dry, hang, hurt, lean, pop, spill, split and
turn.
However, although this is clearly a productive and regular pattern,
there is nothing intrinsically special about this type of entailment:
similar entailments can arise between lexically unrelated word forms,
such as kill and die in Table 5.2.
5.3 Thematic relations

As we’ve already seen, verbs differ not only in how many arguments
they require, but also in how those arguments are related to the verb
semantically – whether they are the subject, the direct object, an
embedded situation, and so on. This is clearly related to the syntactic
structure of the sentence. However, we have also seen that sentences
with different syntactic structures can assign the same relations to their
arguments, as in (5.7).
(5.7) a. I offered her a scone.

b. I offered a scone to her.
The term thematic relation, also called thematic role or partici-
pant role, is used to describe the role that the referent of a noun phrase
plays in its sentence. In both (5.8a) and (5.8b), the relationship between
I and offered is the same: the individual denoted by I is performing the
action denoted by offered. Similarly, although the syntax is different, the
relationship between her and offered is the same in both cases: the indi-
vidual denoted by her is at the other end of the action denoted by offered,
in that she is the prospective recipient. And in both cases a scone denotes
the thing that is offered.
Of course, the relationships I just described are not a special property
of the two sentences in (5.7) but a generalisation about the verb offer in
general: it requires three arguments, one of which performs the action of
‘offering’, one of which ‘receives the offer’, and one of which is ‘offered’.
This is a more specific generalisation than just saying that offer is ditran-
sitive: that is, that it takes three arguments. Also, it makes clear that offer
taking three arguments is not an arbitrary fact about the verb, but rather
it’s intrinsic to the verb’s meaning: sentences involving offer must have
arguments that fulfil these vital thematic roles.
This also gives us another way to think about the classification of
verbs. Rather than just dividing them up according to how many argu-
ments they require, we can group them according to the specific set
of thematic relations that have to be fulfilled by the noun phrases that
complete the sentence. In the case of offer, we might call these thematic
relations Agent, Recipient and Theme – this last term describing an
entity that undergoes an action without changing its state (here, the
thing that is offered, in (5.7) a scone). We can then group offer with other
verbs that also require an Agent, Recipient and Theme, such as give,
donate, sell, pledge and (in at least one sense) promise.
Various other thematic relations can be distinguished, and doing so
might help us distinguish the meaning of verbs, even when they appear
to have very similar syntactic structures, such as in (5.8).
(5.8) a. Jane criticised Bill.
b. The kettle scalded Bill.
Here we could say that the subject of (5.8a), Jane, is deliberately per-
forming an action, whereas the subject of (5.8b), the kettle, is not. We
might want to distinguish these at the level of thematic relations by
allowing that Jane is an Agent but the kettle is something else which
we call a Stimulus. We could then distinguish those verbs like criti-
verbs 63
cise which require an Agent from those like scald which only require a
Stimulus (although in this case could also occur with an Agent).
We can also distinguish these from examples such as (5.9) in which
the subject of the sentence does not perform or accomplish anything but
is instead a recipient of some kind of input – in this case, we might call
the subject, Bill, an Experiencer.
(5.9) Bill heard the traffic.
In a similar vein, we can distinguish Recipients and Themes from other
classes of things that are acted upon in some way. In (5.10), the object of
the sentence undergoes an action and with it a change of state, unlike
a scone in (5.9), which remains unchanged: we could refer to the car in
(5.10) as a Patient.
(5.10) The tree damaged the car.
And we can also distinguish the Recipient role from the subtly different
role of Beneficiary: in (5.11), Valerie is presented as the person for whom
an action has been performed, but (5.11) does not commit us to the view
that Valerie actually gets to receive or experience the consequences of
this action.
(5.11) David wrote a poem for Valerie.
Numerous other thematic relations have been proposed – Instrument,
Location, Source, Goal, and so on. I won’t attempt to catalogue them all
here. Theorists are interested in knowing whether there are just a rela-
tively small number of possible distinct thematic relations, among which
the verbs of English (or any other language) select their requirements
– or whether there are lots of subtly different verb-specific thematic rela-
tions. For instance, was I justified in asserting that the person who is offered
something is a Recipient, even if they don’t actually receive anything?
Or should I categorise them as a Beneficiary, even if they don’t actually
benefit? Or am I ultimately forced to say that they are an “Offeree”, a
thematic relation that is unique to the verb offer? Similarly, we might
say that the verb hit requires two thematic relations to be fulfilled, Agent
and Patient, or we could argue that actually it specifies two much more
specific thematic relations, “Hitter” and “Hittee”, and simplifying this to
Agent and Patient would be invalid. If we can identify shared sets of the-
matic relations, this will be a useful way of identifying similarities in the
semantics of broad classes of verbs – but if we can’t simplify in this way,
thematic relations might not be a very useful idea for us.
Different approaches to grammar make different claims about the
connection between thematic relations and syntax. One idea of this
kind, due to Chomsky (1981), is the “theta-criterion”: this claims there

is a one-to-one correspondence between thematic relations (or their
near-equivalent in the theoretical framework used in that paper, “theta-
roles”) and syntactic arguments, and that any sentence in which there
is a mismatch between syntactic arguments and theta-roles will not be
well formed. We could think of examples such as (5.12) as illustrating
this idea. In (5.12a), there are not enough noun phrases to satisfy the
requirement that give imposes for there to be an Agent, a Recipient and
a Theme. In (5.12b), there are too many arguments for the verb cry,
which only requires an Experiencer. And in (5.12c), there is no possible
Agent, as required by the verb request, because both the entities denoted
by noun phrases in (5.12c) are inanimate.
(5.12) a. *John gave to Mary.
b. *Bill cried Jane.
c. *The rock requested water.
Summary
In this chapter, we have considered the role of verbs within a sentence.
It can make sense to think of the verb as requiring a certain number of
arguments, depending upon its semantics, in order for the sentence con-
taining it to be semantically complete. We can classify verbs according
to how many arguments they require, but we can also identify specific
categories of verbs that cause their containing sentences to give rise to
specific patterns of entailment. And we can drill further down into the
issue of how to complete a sentence containing a verb, by noting that
many verbs – on account of their meaning – require their arguments
to have specific semantic properties. We can think of this as the verb
specifying a set of thematic relations that must be fulfilled by the sur-
rounding noun phrases in order for the sentence to be complete. On
this view, each verb is associated with a complex set of requirements
that it imposes upon its containing sentence – and this property of verb
meaning is associated with the traditional idea of the verb as the seman-
tic core of the sentence.
Exercises
1. In February 2002 a UK government minister announced the resigna-
tion of a senior civil servant in his department. It was subsequently
reported that the civil servant only found out about his own alleged
resignation from listening to the radio. This led to a question being
verbs 65
asked by the media: Who is going to be resigned next? In fact, the civil
servant subsequently resigned three months later. Resigning is sup-
posed to be a conscious act performed by the person who quits the
post. If, in talking about the situation described above, someone had
used the expression The minister resigned the civil servant, would the
sentence have been causative? Would it have the same meaning as
The minister made the civil servant resign?
2. According to the nursery rhyme, after Humpty Dumpty’s accident,
All the king’s horses / And all the king’s men / Couldn’t put Humpty together
again. Given that no one had put Humpty together on any previous
occasion, what kind of verb is put in this sentence, and how does its
meaning relate to that of again?
3. What is each of the following sentences: unaccusative or unergative?
Give reasons for your answers:
(a) The kite flew.

(b) My heart sank.
(c) The students were reading.
4. We could think of pay as a hyponym of give. How does the argument

structure of pay relate to that of give? What is the difference in argu-
ment terms between John paid Mary and John paid the bill?

Kearns (2011) and Huddleston and Pullum (2002) both offer illuminat-
ing accounts of the topics dealt with in this chapter. Lists of candidate
semantic roles vary enormously between sources and between theories,
an issue discussed by Croft (1991) and Dowty (1991). Dowty proposes
to address this by positing more general Agent and Patient “proto-
roles”. Radford (2004) discusses Chomsky’s theta-criterion with refer-
ence to syntactic theory.
6 Tense and aspect
Overview
This chapter is about two important features of how English grammar
allows us to convey meanings. First, we consider how we express
when events occurred, especially relative to the time of speaking or
writing, which is the role of tense. Secondly, we consider how we can
convey by grammatical signals more about how an event occurred,
with respect to time: that it to say, is it an ongoing event, a repeated
event, or an event that took place at a single moment? This is the role
of aspect.
6.1 Talking about events in time

In his novel The Restaurant at the End of the Universe, Douglas Adams
claims that the major problem with time travel is that of grammar, and
refers to an essential grammar book which, among other things, tells you
‘how to describe something that was about to happen to you in the past
before you avoided it by time-jumping forward two days in order to
avoid it’. Given that we don’t have time travel, our needs are somewhat
simpler when it comes to describing events in time – but even then,
this process can become surprisingly complex. Consider (6.1), from an
article by Andrew O’Hagan (‘The end of British farming’, London Review
of Books, vol. 23, 2001).
(6.1) When I told people I was spending time with farmers, they’d
say: how can you stand it, they just complain all day and they’ve
always got their hand out.
(6.1) is especially rich in time reference, so let’s take a moment to
unpack what is going on here, and how time gets expressed and under-
stood. First, we have the verb told, which is a past tense form of the verb
tell. Specifically, it is the past simple form, so called because it is simple
66
tense and aspect 67
in aspect and does not involve any of the special aspectual meanings
that will be introduced later in this chapter. It indicates that the first
event described in (6.1), the speaker telling people that he was spend-
ing time with farmers, took place before the point in time at which
(6.1) was uttered. So the tense marking on the verb expresses this time
reference for the whole event, relative to another point in time. In this
case, the time of the event is expressed relative to the time of utterance
– and therefore the meaning that (6.1) expresses depends upon the time
at which it is uttered. Hence, we can think of tense as deictic, much
like the time adverbials (today, soon, and so on) we discussed briefly in
Section 1.2.1.
The next verb in (6.1), was spending, is in a form called past progres-
sive. The tense marking on was (by contrast with am) indicates that this
is a past form: English happens to mark tense on the first element in
a multi-word verb form like this. The combination of the auxiliary be
in front of the main verb (which could take the form be, am, is, are, was
or were, as appropriate) with the -ing suffix on the verb itself indicates
that we are dealing with progressive aspect, which portrays an event
as being in progress at the relevant time point. Thus we infer that the
event in which the speaker spent time with farmers was an ongoing one
at the time the speaker told people about it.
Next we have they’d say, which is a contracted version of they would
say. Would is the past simple form of the modal auxiliary will (Chapter 7
has more on modal auxiliaries). There is no suffix in English that we can
put on verbs as an indicator of futurity, so will is the main grammatical
device for signalling future time reference in English. What we have
here is the past form of a future marker: the speaker is taking us to some
point in time prior to the time of utterance and looking ahead to a time
when people say to him how can you stand it? In this case, the point in time
the speaker is taking us back to is that same point in time at which he
told people he was spending time with farmers: we can tell this because
of the use of when right at the beginning of the sentence.
The next verbs we come to are the verbs can . . . stand and complain.
These are present simple. But, of course, in this case, that doesn’t
mean that they refer to the time at which the speaker is uttering (6.1) –
we’re still back at the past time in which these people are talking to the
speaker about farmers. So, although we can use present simple forms
to refer to events taking place at the time of utterance, we can also use
them to refer to past or future events, if that is the current time refer-
ence point that we have set in the preceding context. In fact, they just
complain all day is a little more complex still: it seems to convey not that
‘farmers complain all day at the present moment’ but rather ‘farmers
habitually complain all day’. English does not have a grammatical

marker of habitual aspect, but it can sometimes be inferred from the
use of the present simple, and that seems to be an appropriate inter-
pretation in this case. It might also be an appropriate interpretation of
told – there are reasons to suppose that the speaker is not talking about
a single event, but rather a pattern of events. There will be more on
habituality in Section 6.3.1.
Finally, we have the verbs have . . . got, in they’ve always got their hand
out, which are in a form known as present perfect. This uses an idiom,
to have one’s hand out, meaning roughly ‘to be asking for money or gifts’.
Here it occurs in the present tense (we couldn’t say *They’ve got their
hand out an hour ago, for instance) – and once again, that doesn’t mean
that it necessarily refers to the time of utterance, but to the current
time reference point, which in this case is in the past relative to when
(6.1) is uttered. We can think of it as expressing the culmination of
a past event: the farmers were ‘getting their hand out’ and have now
done so. The use of always is best read as hyperbolic (see Section
9.6) and probably expresses something like habitual aspect – but for
convenience here, let’s think of it as just marking an ongoing event
more emphatically.
In sum, (6.1) relates a whole bunch of events to each other in time.
We can think of the events as follows:
(a) ‘speaker utters (6.1)’

(b) ‘speaker tells people he is spending time with farmers’
(c) ‘speaker spends time with farmers’
(d) ‘people say “how can you stand it . . .”’
(e) ‘speaker can (or cannot) stand spending time with farmers’
(f) ‘farmers (are alleged to) complain all day’
(g) ‘farmers (are alleged to) get their hand out’
(h) ‘farmers (are alleged to) have their hand out’.
We can then represent our best guess about the time relations pictori-
ally in Figure 6.1, where time is represented by the line running from
left to right. For simplicity I’m marking the habitual or possibly habitual
events as though they happen just once.
This representation obviously makes some assumptions and sim-
plifications: for instance, it’s not certain that event (c) actually finishes
before the time at which (6.1) is uttered, (a). This is arguably suggested
by the use of was spending rather than spend, but is not certain. Also,
as noted above, I’m treating (b) and (d) as single rather than habitual
events; I’m also treating (a), (b) and (d) as events that take place each
tense and aspect 69
(b) (d) (a)
(c)
(e)
(f)
(g) (h)
Figure 6.1 Time relations among the events described in (6.1). Time runs
from left to right: open-ended boxes indicate events with no endpoint
suggested
at a single point in time, ignoring the fact that these events each have a
duration of their own.
However, the crucial point for our current purposes is that this
complicated picture can be expressed using just a few common, and
sometimes barely noticeable, morphemes in what is still a reasonably
simple sentence. The rest of this chapter discusses more systematically
the options that English makes available to us for expressing these kinds
of time reference via our grammar.
6.2 Tense
In the names of the verb forms we’ve seen so far – past simple, present
perfect, and so on – the first part of the name indicates tense and the
second part indicates aspect. Nine different combinations of tense and
aspect are set out in Table 6.1. This section will discuss tense in more
detail, while Section 6.3 focuses on aspect.
Table 6.1 Two-part labels for tense–aspect combinations, with examples

Past tense Present tense Future tense
simple aspect past simple present simple future simple
saw see will see
progressive past progressive present progressive future progressive
aspect was/were seeing am/is/are seeing will be seeing
perfect aspect past perfect present perfect future perfect
had seen have/has seen will have seen
6.2.1 Preliminaries
As discussed in Section 6.1, a tense form can be used to convey various
different time points or intervals, depending on the time of utterance
and on other features of the sentence. In this respect, tense is deictic,
but the process of understanding the temporal meaning of tense is more
complicated than for other deictic expressions, as we have already seen.
A time adverbial such as yesterday, right now or in ten minutes’ time anchors
an event in the past, present or future, respectively, relative to the time
of utterance. Even though we refer to past, present and future tense,
the mapping between these expressions and the actual time periods
they denote is not always as straightforward as this: sometimes we need
more elaborate pragmatic reasoning in order to determine the intended
meaning of a tense form. As we will see in Section 6.3, the same is true
of aspectual marking.
The forms that explicitly encode tense and aspect in English are the
explicit markers listed below, although we should also be aware that the
“unmarked” forms of verbs (see, look, can, and so on) are also associated
with specific kinds of tense and aspect:
• Auxiliary verbs: will, have, be.

• Irregular forms of verbs: saw, seen, thought, blew, blown, is, am, etc.
• Inflectional suffixes:
●
past tense, usually written -ed
●
present tense, usually written -s when the subject is singular, not
the speaker and not the hearer
●
progressive -ing, for example am singing, was emerging
●
past participle -(e)n or -ed, for example has seen, have helped
We’ll now consider some of the time meanings that English speakers
express using these forms.
6.2.2 Present, past and future

(6.2) shows some examples of the present tense being used with refer-
ence to events and states that occur or exist in a period of time that
includes the time of utterance. Hence, we could tag now or at this precise
moment onto the end of any of these sentences and the result would be
coherent.
(6.2) a. He goes for goal. (Said by a sports commentator)

b. That dog is happy.
c. It’s wagging its tail.
tense and aspect 71
Present forms are also used for so-called “timeless truths”, as in

(6.3). Someone uttering one of these sentences is not just making a
claim about what happens to be the case at the time of utterance. The
adverb always could be added to those utterances without changing the
intended meaning, whereas adding at this precise moment would change
the meaning in a somewhat odd way.
(6.3) a. At sea level, water boils at 100ºC.
b. Dark clouds have a silver lining.
It makes some sense to use the present tense to talk about timeless
truths: after all, they are as true at the present moment as they are at any
other time. More surprisingly, we can use the present tense with adver-
bial markers of future time, such as next year and a week on Wednesday, as
shown in (6.4).
(6.4) a. You arrive in Australia next month in time for the Melbourne
Cup.
b. I am giving a talk in Newcastle a week on Wednesday.
Both of these sentences are unproblematic – it’s easy to imagine them
being uttered in connection with a travel itinerary or a diary entry –
and the use of the present tense doesn’t detract from the future time
reference of the sentence as a whole. We can’t generally do this for past
time reference, as shown in (6.5).
(6.5) a. *You arrive in Australia last month in time for the Melbourne
Cup.
b. *I am giving a talk in Newcastle two Wednesdays ago.
The only way to make sense of the sentences in (6.5) is to interpret them
as part of a narration in which the speaker has switched to using the
present tense for narrative effect (sometimes called the historic present
tense). One possible explanation for the asymmetry between (6.4) and
(6.5) is that Germanic languages tend to have two tenses, past and non-
past (Hewson and Bubenik 1997: 209), so we might think of the English
present tense as really being a non-past tense that is appropriate to use
in reference to present or future events.
For the past tense, the examples in (6.6) demonstrate its typical func-
tion of communicating about events and states located prior to the time
of utterance: in these examples, the times are indicated by the adverbials
at 7 o’clock this morning and yesterday.
(6.6) a. We ate at 7 o’clock this morning.
b. I heard it on the news yesterday.
However, as shown in (6.7), the past tense can also be used with refer-
ence to events that haven’t (yet) happened but could do so.
(6.7) If we introduced proportional representation, there would be
more coalition governments.
We have already seen that we can refer to future events using the
present tense. The nearest thing English has to a grammatical marker of
future time is will. This is a modal auxiliary verb, and historically had
the sense of ‘want’ or ‘wish’, which naturally refers to a future situation.
Subsequently it developed a further sense as a general marker of futu-
rity that doesn’t imply the existence of any wish or preference. (6.8a)
does not express something that Kiribati wants to happen, nor does
(6.8b) express something that the hearer wants.
(6.8) a. A small rise in sea-level and Kiribati will disappear under the
Pacific.
b. You’ll have to leave if you carry on like that.
An alternative way to mark futurity is by using going to. This has under-
gone a similar grammaticalisation process to will: although it still retains
a sense in which it can denote physical motion towards a place, it has a
distinct sense in which it conveys futurity without suggesting motion,
as in (6.9a). In some cases, such as (6.9b), this results in ambiguity as to
which sense the speaker intends to express – motion, or futurity without
motion.
(6.9) a. He’s going to stay at home and look after the kids.
b. I am going to work.
By contrast, will is less often ambiguous, although it can also be used in
at least one other way: for conveying timeless truths, just as the present
simple was in (6.3). In (6.10), for instance, it is clear that the speaker is
not making a prediction about the future action of a specific diamond
or a specific rising tide. If we replaced will cut with cuts and will lift with
lifts, this would make very little difference to the speaker’s meaning in
each case.
(6.10) a. A diamond will cut glass.
b. A rising tide will lift all boats.
6.2.3 Tense and adverbials

Past, present and future time (relative to the time of utterance) can
be expressed in sentences using various deictic adverbials, as we have
tense and aspect 73
Table 6.2 The compatibility of some deictic adverbials with past, present
and future time
Past time Present time Future time
then now then
last year at present next year
last Bastille Day nowadays tomorrow
yesterday in 45 minutes from now
today, this week, this year
already seen. Table 6.2 outlines how some of these adverbials can be
used for different times.
The adverbials listed in the ‘future time’ column can be used with
present simple and present progressive tense forms, as in (6.11a–c), but
the resulting sentences clearly describe future events. The adverbials in
the ‘past time’ column can also be used with present tense forms, as in
(6.11d), but these have to be interpreted as historic present forms which
are talking about past events.
(6.11) a. Mark Lawson is here in 45 minutes. (Radio announcer
describing the next programme)
b. She lectures in Milton Keynes tomorrow.
c. He’s visiting Scotland next year.
d. Last year, he loses his job.
Some deictic adverbials are compatible with past, present or future
time reference: these include today, this week and this year. These can
all make reference to the time of utterance, a time before the utterance
or a time after. Consequently, they can freely combine with all tense
forms.
6.3 Aspect
Tense provides pointers to the location of events in time, relative
to the time of utterance. Once you have “thought yourself into” the
appropriate point in time, aspect comes into play. Aspect is about how
we encode how an event occurs in time. Do we treat it as though it is
compressed into an instant, or mentally stretch it out? Are we interested
in the middle stages of ongoing events, or their culminations? The lan-
guages of the world provide various different ways of expressing aspect
through the grammar. In this section, we look at three forms of aspect:
first, habitual occurrences versus single events, and then two kinds of
aspect that are explicitly marked in the grammar of English, namely

progressive and perfect.
6.3.1 Habituality and simple aspect

We talked a little about habitual interpretation when discussing (6.1),
repeated below.
(6.1) When I told people I was spending time with farmers, they’d
say: how can you stand it, they just complain all day and they’ve
always got their hand out.
A possible interpretation of (6.1) is that the speaker is recounting what
happened on a single occasion during which he told people that he
was spending time with farmers. However, it is perhaps more natural
to think that the speaker is actually generalising over a set of experi-
ences, and that it was in some sense habitual for him to tell people that
he was spending time with farmers, and for those people to respond
in the way he describes. The actions that those speakers attribute to
farmers – complaining, and having a hand out – are also presented as
habitual. The difference is that these are qualified by the adverbials
all day and always. So, although there is no grammatical marking that
specifies habituality in English, it seems that we are able to recognise it,
either for contextual reasons or because of the presence of an appropri-
ate adverbial.
Table 6.3 presents a series of sentences, in past, present and future
tense, which admit a possible habitual interpretation even without any
specific adverbials that point in that direction. For instance, the past
tense sentence Tim drank decaf can clearly be a description of a single
past event: for example, it could be used to tell us what Tim drank after
dinner yesterday. However, it can also be interpreted as a statement
about Tim’s past coffee-drinking habits. The same applies to the other
examples in Table 6.3. We can force a habitual interpretation by adding
Table 6.3 A range of sentences which all have habitual as a possible

interpretation
Past simple Present simple Future simple
Maya loved music. Maya loves music. Maya will love music.
Tim drank decaf. Tim drinks decaf. Tim will drink decaf.
Little Maurice brushed Little Maurice brushes Little Maurice will brush
his teeth by himself. his teeth by himself. his teeth by himself.
tense and aspect 75
a suitable adverbial – in those days, nowadays, in the future – as in (6.12). But

in the absence of an adverbial, both possible interpretations are in play.
(6.12) a. Maya loved music in those days.
b. Tim drinks decaf these days.
c. Little Maurice will brush his teeth by himself in the future.
There is also a distinction between the first set of sentences and
the other two sets. In the drank decaf and brushed his teeth sentences,
we can force a single-event reading by using a regular time adverbial
like yesterday, today or tomorrow. This is a bit odd with the loved music
sentences. This is because those sentences denote states, rather than
activities, a distinction discussed by Vendler (1957): if we say that
Maya loves music, we are not pinpointing a single instance of her loving
music, but rather commenting that she is in the general condition
of loving music. Because of the nature of states, sentences like this
convey habitual meanings (or something very like habitual meanings)
automatically.
A further wrinkle is that the single-event interpretation is not so
readily available for the present tense form as the past or future form:
(6.13a) is a little odd whereas (6.13b–c) are fine.
(6.13) a. ?Tim drinks decaf today.
b. Tim drank decaf yesterday.
c. Tim will drink decaf tomorrow.
When we use a present simple form such as drinks, there is a strong
preference for the habitual reading (Miller 2002: 148), with perhaps a
few specific exceptions such as sports commentary (as in (6.2a)). But this
may be because a single event is usually more naturally described with
a different form, in this case is drinking, as in (6.14). The kind of aspect
exhibited by is drinking – progressive aspect – is the topic of Section
6.3.2.
(6.14) Tim is drinking decaf today.
To summarise: all of the simple aspect sentences in Table 6.3 allow
a habitual interpretation, and this is the only interpretation available
for the three state-denoting sentences in the top row. The other six are
open to both habitual and single-event interpretations, but the habitual
reading is usually preferred for present simple forms. Habitual inter-
pretations can be made obligatory by the use of certain adverbials, but
in the absence of these we have to rely on the context of the utterance
(and the other material in the sentence) to understand which of the two
possible interpretations is intended by the speaker.
6.3.2 Progressive aspect

Progressive aspect is marked in English by the use of the verb be (in
some appropriate form) and the suffix -ing. The use of progressive
aspect presents an event as an ongoing activity which occurs for a pro-
tracted period of time – it doesn’t focus on the beginning or the end
of the event. In narrative, background events are often presented in a
clause with progressive aspect, as in (6.15).
(6.15) The rain was falling as I left the house. (<https://blogs.fcdo
.gov.uk/thomasreilly/2019/04/04/thoughts-in-the-rain/>,
Here, the scene is set by a progressive clause describing the (ongoing)
weather, and a single event is presented as occurring against that back-
drop, namely the speaker leaving the house. If we wanted to focus on
the onset of the weather, we could switch the aspectual marking around,
as in (6.16).
(6.16) As I was leaving the house, the rain started to fall.
Progressive aspect can be used to present even a short event as an
ongoing activity, thus making it a possible backdrop for other events.
For instance, (6.17) presents the departure of a bus as an ongoing activ-
ity, which the hearer is presumably being invited to interrupt with the
single momentary action of boarding the bus.
(6.17) Hurry, the bus is leaving.
We can see how progressive aspect disregards the ends of events by
considering the entailments of sentences such as (6.18)–(6.20). Here, the
(c) sentences are in progressive form.
(6.18) a. The firm demolished the building.
b. The firm has demolished the building.
c. The firm was demolishing the building.
(6.19) a. The waiter folded a napkin.
b. The waiter has folded a napkin.
c. The waiter was folding a napkin.
(6.20) a. They drew up a contract.
b. They have drawn up a contract.
c. They were drawing up a contract.
In each of these cases, we see that the (a) and (b) sentences entail that
the relevant event was successfully completed, whereas the (c) sentence
tense and aspect 77
does not: it is compatible with a situation in which the action was

interrupted.
The above examples all involve “accomplishments”, in Vendler’s
(1957) taxonomy of events: they have a duration in time and they have
a well-defined point of completion. By contrast, verbs that encode states
tend to be less compatible with progressive aspect, as shown by (6.21a).
Where we can use progressive aspect with state verbs, the effect is
sometimes what Cruse (2011) calls “provisionality”: the sense that the
state is currently in effect but may not be in the future. For instance,
(6.21b) seems to suggest that the hearer’s keys can be retrieved, whereas
the non-progressive (6.21c) suggests that the speaker’s hopes cannot.
(6.21) a. *Who is knowing Danish?
b. Your keys are lying at the bottom of the swimming pool.
c. My hopes lie shattered.
6.3.3 Perfect aspect

In English, the combination of the auxiliary verb have and the past
participle form of a verb expresses what is called perfect aspect.
The perfect (from the Latin perfectum, meaning completed) is used to
describe occurrences in the aftermath of an event or state: that is to say,
in the period (however long) during which the event or state seems to
continue to have direct consequences. To put it another way, when we
use the perfect, we convey that the occurrences are in the aftermath of
some event or state.
This is a slightly elusive notion, but we can illustrate it with some
examples. Consider (6.22) and (6.23).
(6.22) a. I passed my driving test.
b. I have passed my driving test.
(6.23) a. I passed my school leaving exams.
b. I have passed my school leaving exams.
Immediately after passing their driving test or their school leaving
exams, a speaker might utter any of these sentences. But if the speaker
is looking back years after the event, (6.23b) seems much less appropri-
ate than (6.23a) – even though the event may have been very relevant
to the life they live now, its direct consequences no longer reverberate.
It is perhaps easier to imagine a context in which (6.22b) would still be
appropriate years later – perhaps the speaker is reassuring a nervous
passenger – because the fact of their having passed their driving test
may still have immediate relevance. As Quirk et al. (1985: 193) put it,
two common features of present perfect meaning are that ‘the relevant
time zone leads up to the present’ and ‘the result of the action still
obtains at the present time’.
Of course, if we are in the aftermath of an event, this entails that
the event has been completed. We see this in the (b) examples of
(6.18)–(6.20) as well as those of (6.22) and (6.23). In cases such as (6.24),
arguably being in the aftermath of an event tells us something more.
(6.24) a. The rain started.
b. The rain has started.
Both versions of (6.24) entail that the event of ‘the rain starting’ has
happened. But under normal circumstances (6.24b) also conveys that
it is still raining at present, whereas (6.24a) is compatible with a situ-
ation in which the rain started and stopped again prior to the time of
utterance.
Having said that, what counts as the aftermath – and consequently,
what a sentence like (6.24b) tells us – is potentially dependent on
context. If (6.24b) was spoken by a scientist who predicted that climate
change would bring rain to some previously arid desert, it could be used
to report signs that rain had fallen there, rather than to report that it is
presently raining there. In this case, we’re relying on a habitual inter-
pretation: the speaker is reporting that we are in the aftermath of the
change from it habitually not raining to it habitually raining.
Linguists have noted that present perfect forms tend not to accept
past time adverbial modifiers, as illustrated in (6.25a). However, Klein
(1992) pointed out that the present perfect unexpectedly accepts
members of a small class of past time adverbials, including recently. The
contrast between (6.25b) and (6.25c) illustrates that recently behaves as
a past time adverb, but (6.25d) shows that it fits with a present perfect.
(6.25e) shows that the same is true for a preposition phrase with since.
(6.25) a. *I have arrived yesterday.
b. *They go there recently.
c. They went there recently.
d. They have been there recently.
e. They have been there since last week.
This pattern apparently arises because of how the meanings of recently
and since last week make reference to the time of utterance: the periods
they describe run up to the present time, which makes them compat-
ible with the present perfect. Past time adverbials which denote times
that do not extend up to the present, such as yesterday, do not work so
well.
tense and aspect 79
6.3.4 Perfect aspect or tense?

We have discussed how tense locates events in time, whereas aspect
describes the time profile of those events. However, perfect aspect does
locate events in time, in that it is used only in the aftermath of those
events. Hence, from the rain has started, we infer that the event of ‘the
rain starting’ happened before the time of utterance, even if only a
moment before. This is one reason why the perfect form is sometimes
called the “perfect tense”, as it is in Huddleston and Pullum’s (2002)
authoritative grammar of English.
In the examples we have seen of the present perfect, we do also have a
potential place to mark tense, the verb have. If we replaced have with had,
the clauses would be past perfect, as in the underlined parts of (6.26).
(6.26) a. When I started work, I had passed my driving test.
b. As I left the house, the rain had started.
(6.26a) clearly places the event of the speaker starting work in the after-
math of them passing their driving test, and (6.26b) places the event
of them leaving the house in the aftermath of the rain starting. In both
cases, the first clause deictically points to a location in time before the
time of utterance, and the past perfect places its event at a time before
that.
Summary
Tense is deictic. It locates events in relation to the time of utterance:
past, present or future. In English, past forms usually appear with a
suffix, present forms unmarked or with an -s, and future forms with
various kinds of marking. Time adverbials also help to reveal the
mapping between tense forms and time, which can be surprisingly flex-
ible and context-dependent.
Aspect is about the time profile of events. The grammatically marked
forms in English are progressive (ongoing without attention to the
ending) and perfect (we are in, or talking about a time in, the aftermath
of the event). Habitual aspect is not grammatically marked in English,
but is readily available as an interpretation for numerous classes of
sentence.
Exercises
1. Table 6.2 presents various kinds of deictic adverbial showing the
different times – relative to utterance time – that they are compatible
with. Which group does recently belong in? And where does soon
belong? You will need to make up sentences and scenarios for past,
present and future tense and try them for compatibility with recently
and soon.
2. With reference to aspect, discuss the difference in meaning between
Arthur’s a tyrant and Arthur’s being a tyrant.
3. A tobacco company told the Czech government that they had saved many mil-
lions of dollars because people were dying early. Think of the sentence in
italics as part of a newspaper report (and note that the pronoun they
refers to the Czech government). Identify the combinations of tense
and aspect used in the sentence. If we wanted to draw a diagram like
Figure 6.1 to represent the relative timing of the events mentioned
in the sentence, what might be ambiguous about the sentence as it
stands?
4. Sentences (a) and (b) illustrate “be to verb” as a rather formal way of
marking the future. (a) could be used to notify students about a dead-
line; (b) could be used on the deadline day to remind them.
(a) On 11 May you are to submit a written solution to the exercise.

(b) You were to submit written solutions today.
Sentence (b) embeds a future tense within the past; were is a past
tense form and “be to verb” is, as illustrated in (a), a way of marking
future.
Now try to find some less formal ways of embedding a future in
the past. Suppose your friend promised you (yesterday) that they
would lend you a book that you asked for, but they have now forgot-
ten which book. To remind them, you could use a ‘future in the past’
form: past because the promise was made yesterday, future because
the book was to be lent at a then-future date. Suggest two reasonable
completions for (c) that involve a form of future marking with past
tense on a verb.
(c) You said you _______________ lend me One Day.
How do these compare with the words that you might have used in
asking to borrow the book in the first place?

Trask (1993) is a good first resort for looking up terms such as tense,
aspect, progressive and perfect that may be unfamiliar. Chapter 13 of Miller
tense and aspect 81
(2002) is a short, clear introduction to the meanings associated with

tense and aspect. Kearns’s (2011) account of English tense and aspect is
also highly accessible and systematic, and there are good discussions in
Cruse (2011) and Saeed (2015). Worthwhile generalisations, as well as
many interesting details, are available via the index entries for tense and
aspect in Huddleston and Pullum (2002) and Quirk et al. (1985).
7 Modality, scope and
quantification
Overview
So far we have mostly discussed sentences that categorically state facts
about the world: that is to say, they assert that something is true. In
practice, we often want to communicate less categorically than this: for
instance, to assert that something is possibly true, or that it ought to be
true, or that it is permissible for it to be true. We use the term modal-
ity to refer to the ways in which we can use language to qualify state-
ments of fact like this. English provides a number of ways of expressing
modality, including through the use of modal verbs and adverbial
expressions.
Modality is important to us in communication, but the expression
of modality often gives rise to ambiguity. One way in which this can
happen is when a modal expression interacts with another expression,
such as negation – in this case, we need to understand the scope rela-
tions between the two expressions, which we will discuss in Section 7.2.
Finally, we will look at how we can use language to quantify over things.
As we shall see, quantification has a lot in common with modality, not
least that quantifiers can also enter into complicated and sometimes
ambiguous scope relations.
7.1 Modality
To put it very generally, a clause characterises a situation, and modal-
ity expresses how the speaker relates to that situation. The underlined
expressions in (7.1) contribute modal meanings.
(7.1) a. Bill has to apologise.

b. You can come in now.
c. She’s not able to see you until Tuesday.
d. Working so many hours a week, he must be exhausted.
82
modalit y, scope and quantification 83
e. If the pub is open, there ought to be someone behind the bar.

f. Martians could be green.
In (7.1a), the verb has to qualifies the situation ‘Bill apologises’, and
expresses that there is an obligation to make that situation come about.
Can in (7.1b) similarly indicates that a situation (‘the hearer comes in’)
is permissible, and able to in (7.1c) indicates that it is feasible. In (7.1d–f),
the verbs signal how confident the speaker is in their knowledge about
the relevant proposition: must indicates that, in the light of available
information, the speaker considers the proposition (in (7.1d), ‘he is
exhausted’) to be certainly true; ought to indicates that they consider it
likely to be true; could indicates that they consider it possibly true.
The main carriers of modality are a set of auxiliary verbs called
modals: will, would, can, could, may, might, shall, should and must. However,
modality is also encoded by various other expressions, such as possibly,
probably, have (got) to, need to, ought to and be able to.
7.1.1 Modal verbs and tense

In Chapter 6, we saw that tense forms are not an entirely reliable indi-
cator of time: it is possible to use present forms when narrating past
events, and past forms for future events in conditional sentences, and so
on. For modal verbs, past forms are often used in connection with future
events. As shown by (7.2a–b), would behaves as the past form of will: but
as shown by (7.2c–d), either can be used to make a request that refers
to the future.
(7.2) a. *In the past, John will help me.
b. In the past, John would help me.
c. Will you help me tomorrow?
d. Would you help me tomorrow?
Can and could pattern the same way. When used to make requests
about the future, there is very little difference in meaning, although we
might intuit that the past tense forms are somewhat more tentative and
perhaps more polite than their present tense counterparts.
In Chapter 6 we discussed how will can signal future time reference
but can also be used to state timeless truths like (7.3a). We can compare
this with (7.3b), in which we replace will with a weaker modal, can.
(7.3) a. A diamond will cut glass.
b. A diamond can cut glass.
If we think of will in (7.3a) as a modal, we can understand the sentence
as expressing a high level of confidence on the part of the speaker that,
if you attempt to cut glass with a diamond, you will succeed. (7.3b)
expresses a slightly weaker meaning, along the lines that if you do this,
you may possibly succeed. But there doesn’t seem to be a difference
between (7.3a) and (7.3b) in terms of the time reference they express:
will does not contribute a future tense meaning to (7.3a). Thus, when
we interpret will, we have to consider whether it is contributing tense or
modality. In fact, marking of modality tends more generally to be com-
plementary to tense marking: English syntax often forces us to choose
whether a clause will have tense in it or modality instead. The default
pattern for English is for clauses to have tense marking but no marking
for modality.
Expressions of modality exhibit a rich array of useful meanings. As a
consequence, the modal auxiliaries are among the most frequently used
verbs in English: six of the top twenty verbs in English are will, would,
can, could, may and should, each averaging more than 1,000 occurrences
per million words of running text (Leech et al. 2001: 282). Biber et al.
(1999: 456) analysed a corpus of text samples totalling 40 million words,
from a range of genres, and found that modals were used in about 15 per
cent of the clauses that could have them. In short, modality is an impor-
tant topic in English semantics and pragmatics. Whole books have been
written about English modals and modality, for example Palmer (1990).
Against this backdrop, the aim of this section is naturally quite modest:
we will just look at a couple of the principal issues that make modality
an interesting topic to study, along with the terminology that is used to
discuss these issues.
7.1.2 Epistemic and deontic modality

Recall that examples (7.1a–c) involved modality modifying how situ-
ations were presented, whereas (7.1d–f) used modality to qualify the
strength of the speaker’s belief in the proposition being discussed.
This distinction broadly corresponds to the distinction between what
is called – borrowing terms from philosophy – deontic and epistemic
modality.
Epistemic meaning has to do with knowledge and understand-
ing. Markers of epistemic modality are understood to be expressing
the strength of someone’s certainty about the truth of a proposition.
Examples (7.4)–(7.7) present pairs of sentences (a) without and (b) with
modal markers, which are underlined.
(7.4) a. The whole hillside is slipping down into the valley.

b. The whole hillside could be slipping down into the valley.
(7.5) a. They meet in the final tomorrow.

b. They may meet in the final tomorrow.
(7.6) a. Jessica went by motorbike.
b. Jessica probably went by motorbike.
(7.7) a. The van was heavily laden, judging by the tyre marks.
b. The van must have been heavily laden, judging by the tyre
marks.
Speakers can produce modally unqualified sentences, like the (a) exam-
ples here, when they are confident in their facts; however, the possibil-
ity of marking epistemic modality makes it easy for a speaker to make
a less definitive claim if they are not certain, as in the (b) examples.
Epistemic modality comes in different strengths: (7.4b)–(7.7b) represent
a gradient from weak to strong modality. In (7.4), replacing is by could
be considerably weakens the claim that the speaker is making: in (7.4b)
they are merely saying that the whole hillside slipping into the valley is
not ruled out by the available evidence. At the other end of the scale, in
(7.7b), replacing was by must have been makes only quite a subtle differ-
ence to the speaker’s commitment to the claim that the van was heavily
laden.
Deontic meaning is about concepts such as duty, morality, laws and
rules. Deontic modality allows speakers to express whether a proposi-
tion relates to a situation that is obligatory, permissible, or somewhere
in between, as shown by the examples in (7.8), with modal verbs
underlined.
(7.8) a. You can borrow my bike any time.
b. The consul could have been more helpful.
c. You should email Tom.
d. Tax returns must be submitted by the end of September.
(7.8a) shows a common way of giving permission (using may instead of
can would also work, and is sometimes preferred prescriptively): the
speaker conveys that the situation in which the hearer borrows their
bike at any time is a permissible one, as far as they are concerned. In
(7.8b), the speaker uses could to convey the idea that it would have
been permissible for the consul to be more helpful (and thereby sug-
gests that they would have preferred the consul to be more helpful).
The use of should in (7.10c) conveys that the speaker takes the situa-
tion in which the hearer emails Tom to be a preferable one. And must
in (7.10d) conveys that the timely submission of a tax return is an
obligation.
As the same verbs can be used to express both epistemic and deontic
modality, it is quite possible for a sentence to be ambiguous between
these interpretations. In fact, some of the examples we have already
discussed are ambiguous in principle. An example like (7.9) makes this
more obvious.
(7.9) After the exam, Amy may take a cigarette break.
If (7.9) is uttered by someone who has authority over Amy’s smoking,
it is natural to interpret it as a grant of permission: that is, to take may
to express deontic modality. If it is uttered by someone who does not
have such authority, it is natural to interpret it as a prediction that it is
possible that Amy will take a cigarette break after the exam – and in this
case we are taking may to express epistemic modality.
In interpreting (7.9), we rely on context to establish whether the
modal expression is meant to be epistemic or deontic: specifically, we
rely on facts about the speaker and their likely authority. And often
it is reasonably straightforward to disambiguate modal expressions in
this way. Consider (7.10)–(7.12), in which the modal expressions are
underlined.
(7.10) a. It may be dark by the time we’ve finished.
b. If you wish, you may copy these two diagrams.
(7.11) a. Random numbers can appear to have patterns in them.
b. The pigeons can have this bread.
(7.12) a. At 95 metres, this has got to be one of the tallest trees in the
world.
b. He has got to be more careful or he’ll break something.
In the (a) examples, the speaker is discussing natural phenomena, and it
makes no sense for the speaker to attempt to grant permission for night
to fall, or for random numbers to exhibit apparent patterns. Nor does it
make any sense to place an obligation on a tree to be one of the tallest in
the world. These potential deontic meanings are sufficiently absurd that
we can dismiss them even without consciously entertaining them: we
naturally seem to understand the (a) sentences as expressing epistemic
modality. By contrast, the (b) sentences do seem to express authority
on the part of the speaker, and are naturally compatible with deontic
interpretations.
The ambiguity between epistemic and deontic modality also arises in
cases that do not involve core modal verbs. (7.13) provides an example
of this: the underlined content expresses epistemic modality in the
context of (a) and deontic modality in the context of (b). We are led to
this understanding by the surrounding context: in (a), Marie’s expecta-

tion is based on her prior experience, whereas in (b) it is an expression
of how she thinks things ought to be.
(7.13) a. Marie expected the coffee to be strong; she had ordered it
before.
b. Marie told the waiter that she expected the coffee to be
strong and that she would send it back if it wasn’t.
Under certain circumstances, the ambiguity between deontic and
epistemic modality may be resolved incorrectly, and this can have
important consequences. The physicist Richard Feynman identified
an example of this while working on the inquiry into the failure of the
Space Shuttle Challenger in 1986. He quoted NASA officials as making
the argument presented here as (7.14).
(7.14) Since the Shuttle is a manned vehicle, the probability of mission
success is necessarily very close to 1.0. (Feynman 1999: 154)
The underlined material in (7.14) admits two possible interpretations.
On an epistemic interpretation, it affirms that, given the available evi-
dence, the mission could be judged to be almost certain to succeed. On a
deontic interpretation, it affirms that the mission must be almost certain
to succeed: that is, if it isn’t almost certain to succeed, it should not take
place. Feynman noted that NASA’s statement was intended to be deontic,
but was widely misinterpreted as epistemic, giving rise to an irrational
belief that the Shuttle mission was a lot safer than it actually was.
7.2 Semantic scope

In the preceding section, we saw how ambiguity could arise between
two possible readings of modal expressions, one epistemic and one
deontic. But additional ambiguities arise with the use of modals, par-
ticularly when combined with certain other kinds of expression, such
as negation.
In Section 2.2, we discussed how words such as unlockable can sys-
tematically have two meanings, one in which the prefix un- combines
with the word lockable to give the meaning ‘not able to be locked’, and
one in which the suffix -able combines with the word unlock to give the
meaning ‘able to be unlocked’. We could describe these different deri-
vational processes by using brackets, and distinguishing (un(lockable))
from ((unlock)able).
Another way of looking at this is to appeal to the idea of scope: that is
to say, the material that a particular operator applies to. We could think
of the prefix un- and the suffix -able as operators on the meanings of
other lexical items, which achieve the effects respectively of ‘reversing
the direction of change in a verb meaning’ and ‘turning a verb into an
adjective of possibility’. Then we could say that the difference between
(un(lockable)) and ((unlock)able) is that, in the former, un- takes scope
over - able, while in the latter the reverse is true. So in the former case
the resulting word meaning denies the possibility of something being
locked, whereas in the latter case the resulting word meaning expresses
the possibility of something being unlocked.
The general point here is that, when we have two operators in the
same expression, we can get different meanings depending on which
operator includes the other within its scope. This is an important issue
for modality. Consider the deontic interpretations of the sentences in
(7.15).
(7.15) a. You must discuss the case.
b. You have to discuss the case.
c. You mustn’t discuss the case.
d. You don’t have to discuss the case.
(7.15a) and (7.15b) are virtually identical in meaning, but their negated
forms (7.15c) and (7.15d) are sharply different: (7.15c) conveys that dis-
cussing the case is forbidden, whereas (7.15d) merely conveys that it is
not obligatory.
We can think of these meanings as involving two operators: negation,
expressed here by n’t, and obligation, expressed here by must and have to.
In (7.15c), obligation takes scope over negation, and the sentence conse-
quently expresses an obligation towards a negated proposition (‘hearer
does not discuss the case’). In (7.15d), negation takes scope over obliga-
tion, and the sentence expresses a lack of obligation towards a positive
proposition (‘hearer discusses the case’). We can write these scope rela-
tions down using brackets, as shown in (7.24).
(7.16) a. obligatory (not (you discuss the case))
b. not (obligatory (you discuss the case))
In this example we have two very similar modal expressions convey-
ing different meanings because they interact differently with negation.
In (7.17), we can see two very different modal expressions conveying
very similar meanings – again, because they interact differently with
negation.
(7.17) a. They must not have received the invitation.
b. They can’t have received the invitation.
Either of these sentences could be used to convey an epistemic modal

meaning, which might be intended to explain why some expected
guests have failed to show up at an event. But the modals must and can
obviously differ in meaning. It is their interaction with the negation that
cancels out this difference. In (7.17a), just as in (7.15c), must takes scope
over the negation, with the resulting meaning that ‘it is certain that they
did not receive the invitation’. But in (7.17b), can falls within the scope
of the negation (as it typically does), with the resulting meaning ‘it is
not possible that they received the invitation’. And this is essentially the
same meaning as that yielded by (7.17a).
Negation in English is complicated to interpret because its semantic
scope is not entirely predictable from syntax alone. As a first approxima-
tion, we might guess that negation will take scope over a modal operator
if the negation comes first in the sentence: but (7.17b) shows us that this
is not always true. A more refined idea would be that the scope rela-
tions of operators have something to do with their relative positions
in the syntactic tree: however, as we shall see in the following section,
semantic ambiguities seem to arise from some sentences that appear to
have unambiguous syntactic analyses. In addition to this, English has a
widespread class of expressions, sometimes referred to as “neg-raising
predicates” (Horn 1989), which cause negation to be interpreted in a
way that doesn’t straightforwardly match the syntax. (7.18) presents
some examples of this.
(7.18) a. I don’t think Bill is going to win.
b. Elena doesn’t seem happy.
c. Tom doesn’t want to move to London.
d. Aliya doesn’t intend to take that risk.
The speaker of (7.18a) appears to be saying that ‘it is not the case that I
think Bill is going to win’: that is, they are expressing the negation of ‘I
think Bill is going to win’. However, in practice, we understand them to
be conveying ‘I think Bill is not going to win’: that is, we actually inter-
pret the negation as taking scope over going to win rather than think Bill is
going to win. Similarly, (7.18b–d) are all naturally interpreted as making
stronger statements than they appear to: we understand the speakers
to be saying that ‘Elena seems unhappy’, ‘Tom wants not to move to
London’ and ‘Aliya intends not to take that risk’.
In general, then, we can’t just read scope relations off the syntax: we
are going to have to combine syntactic information, individual word
meanings, and contextual information in order to understand what the
speaker meant to convey. This is true when we combine modality with
other operators, when we interpret negation, and when we combine
various other kinds of operator, including those we will discuss in the

next section.
7.3 Quantification
English provides a range of expressions that we can use to provide infor-
mation about the number, or proportion, of individuals that have certain
specific properties. These expressions include quantifiers such as some,
all, none, many and most, as well as numbers, modified numerals such as
more than 10, and so on.
We can think of quantified sentences such as those in (7.19) as making
statements about the number of distinct entities in sets of things. In
(7.19a), this is the set of cats that lack tails; in (7.19b), it is the set of men
who have walked on the Moon; in (7.19c), it is the set of countries in
Africa.
(7.19) a. Some cats lack tails.
b. Twelve men have walked on the Moon.
c. There are more than fifty countries in Africa.
For this reason, before we discuss the meanings of quantifying expres-
sions in general, it will be useful to introduce some ideas about sets as
used in semantics and pragmatics.
7.3.1 Some basics about sets

A set is a collection of distinct objects, which can be of any kind. The
objects in the set are referred to as the elements or members of the set.
We can define a set either by listing all of its members or specifying the
properties that its members must have in order to qualify as part of the
set. (This latter option will be more relevant to our needs here.) A given
element can be a member of multiple sets at the same time.
It’s customary to denote sets by capital letters. If two sets A and B have
exactly the same members, we can say that the sets are identical: we can
write this as A = B. If every member that is an element of one set, C, is
also an element of some other set, D, then we can say that C is a subset
of D, and write this as C ⊆ D. For instance, if C is the set of numbers
that are divisible by 9 and D is the set of numbers that are divisible by
3, every member of C is automatically a member of D, but the reverse is
not true. We write the empty set – the set with no members – with the
symbol ∅. (Technically we say that C is a “proper subset” of D if C ⊆ D,
C is not equal to D, and C is not equal to ∅.)
We use the word size or cardinality to describe the number of dis-
tinct, non-identical members in a set. The empty set ∅ has cardinality

zero. If A = B, both A and B have the same cardinality. We can write
these relations as |∅| = 0 and |A| = |B|.
We can combine sets in particular ways to form other sets. For
instance, we can define the union of two sets, E and F, as the set com-
prising all the entities that are members of E or members of F (or both).
We write this as E ∪ F. We can define the intersection of two sets, E
and F, as the set comprising all the entities that are both members of E
and members of F. We write this as E ∩ F. As these definitions suggest,
union and intersection are both symmetric operations: E ∪ F = F ∪ E,
and E ∩ F = F ∩ E.
Because of how these sets are defined, we can write down some state-
ments about the cardinalities of unions and intersections relative to the
cardinalities of the original sets:
• |E ∪ F| ≤ |E| + |F|. The biggest that the union of two sets can
be is the cardinalities of those two sets added together. The union of
the sets will only be this big if the sets don’t overlap at all – that is,
there is no single entity that is a member of both sets – in which case
we say that the sets are disjoint. In that case, the intersection will be
empty: E ∩ F = ∅.
• |E ∩ F| < |E|, |E ∩ F| < |F|. The intersection of two sets can’t
be bigger than either of the sets individually (because it can only
contain entities that are members of both sets).
We can also define the complement of a set as all the things that are not
in the set. For the set G, we’ll write the complement as G′. For any two
sets G and H, the following relations will be true:
• G = (G ∩ H) ∪ (G ∩ H′)
• H = (H ∩ G) ∪ (H ∩ G′)
That is to say, given any set H, we can divide any set G completely
into two parts: the members of G that are also members of H, and the
members of G that are not also members of H. However, one of these
parts may be empty: if G is a subset of H, then G ∩ H′ is empty. If G and
H are disjoint, then G ∩ H is empty.
In the following section, we’ll see how we can use some of this theo-
retical machinery to make sense of how we express quantity in English.
7.3.2 Simple quantifiers in terms of sets

Suppose we want to convey information about whether members of
one set have a particular property: for instance, whether pandas are
P V
Meaning of (7.20a)
V
P
Meaning of (7.20c)
Figure 7.1 Venn diagrams for the meanings of (7.20a) and (7.20c)
vegetarian. Depending on what we want to say, we might use one of the

sentences in (7.20).
(7.20) a. No pandas are vegetarian.
b. Some pandas are vegetarian.
c. All pandas are vegetarian.
Let P denote the set of pandas and V denote the set of individuals
that are vegetarian. (Here I’m using vegetarian in the sense of ‘being an
animal that does not eat meat’ rather than the more specific sense of
‘being a person who does not eat meat’.) We can think of the meanings
of the sentences in (7.20) in terms of what they say about the relation-
ship between the sets P and V. (7.20a) means that there are no members
of P that are also members of V: to put it more formally, that the inter-
section of P and V is empty, which we can write as P ∩ V = ∅. (7.20c)
means that every member of P is also a member of V: that is, that P is a
subset of V, which we can write as P ⊆ V. Figure 7.1 illustrates these two
meanings with Venn diagrams.
(7.20b) is a little bit trickier. Minimally, we could say that this sen-
tence means that there exists a member of P that is also a member of V:
that is, that the intersection of P and V is non-empty. However, if there
existed only one vegetarian panda in the world, you might judge (7.20b)
to be false on the grounds that ‘one’ is hardly ‘some’. So we might instead
say that (7.20b) means that there are multiple members of P that are also
members of V: the intersection of P and V has cardinality at least two
(or at least three). For now I’ll leave that controversy aside and assume
that some just requires the intersection of the sets to be non-empty (that
is, ‘there exists a vegetarian panda’). We’ll also ignore the fact that there
is a sense of some in which it is just a plural indefinite determiner, as in I
ate some cakes: that is not the sense that is relevant for (7.20b).
If we assume that the above analysis is correct, then we can write
down the meanings of (7.20) in terms of the set relations in (7.21).
(7.21) a. P ∩ V = ∅
b. |P ∩ V| ≥ 1 (or P ∩ V ≠ ∅)
c. P ⊆ V
This formalism makes it relatively easy to establish what the entail-
ments for these sentences are. For instance, we can see that – assuming
that pandas exist at all – (7.21c) entails (7.21b), because if P ⊆ V then
|P ∩ V| ≥ 1, unless P = ∅. So, according to this formalism, if all pandas
are vegetarian is true, then some pandas are vegetarian is also true.
In practice, people are split on how they judge the truth of sentences
with some in situations where the corresponding sentence with all is true.
There is an intuition that some actually means ‘some but not all’ and the
above semantic analysis doesn’t capture that. However, there are good
reasons to suppose that the semantic meaning of some is just that shown
above, and that the meaning ‘some but not all’ arises because of a prag-
matic enrichment. We’ll see a possible explanation of why this happens
in Section 8.2.
We can provide a similar analysis for numerical quantifiers, as in the
examples in (7.22). Rather than talking about pandas that are vegetarian
– it seems odd to try to count the number of pandas in the world that are
vegetarian – we’ll discuss some examples concerning how many pandas
are in the zoo.
(7.22) a. There are three pandas in the zoo.
b. There are more than three pandas in the zoo.
c. There are at most ten pandas in the zoo.
Let P denote the set of pandas and Z denote the set of animals in the
zoo. A straightforward semantic analysis of the sentences in (7.22) is
proposed in (7.23).
(7.23) a. |P ∩ Z| = 3
b. |P ∩ Z| > 3
c. |P ∩ Z| ≤ 10
On this analysis, (7.22a) means that there are exactly three dis-
tinct entities that are members of the set of pandas and that are also
members of the set of zoo animals. (7.22b) means that there are more
than three distinct entities that belong to both these sets, and (7.22c)
means that there are ten or fewer distinct entities that belong to both
these sets.
Unfortunately, in practice, the meanings of numerical quantifiers in
English are somewhat more complicated than this. Even plain, unmodi-
fied numerals, like three in (7.22a), are ambiguous: the sentence might
mean that the number of pandas in the zoo is exactly equal to three (the
analysis given in (7.23a)), or it might just mean that there exist three
pandas that are in the zoo, which might be the intended interpretation if
the speaker has something to say about these three pandas in particular.
In the latter case, the total number of pandas in the zoo might actually
be more than three, in which case the analysis given in (7.23a) doesn’t
apply. And it’s a similar story for expressions such as more than, at least,
up to, and so on. We can easily write down formalisms that capture their
mathematical meaning – we use > for more than in (7.23b), ≤ for at most in
(7.23c), and so on – but actually the way they are used in natural language
has some additional subtleties to it. However, for our present purposes,
we can assume that the analyses presented above are reasonably good
approximations to the meanings of these English-language expressions.
7.3.3 Proportional quantifiers

Most of the quantifiers discussed in the preceding section have a sym-
metry to their meaning. To say that Some pandas are vegetarian is equiva-
lent, in terms of set membership, to saying that Some vegetarians are
pandas. Similarly, saying that There are more than three pandas in the zoo is
equivalent to saying that More than three of the things in the zoo are pandas.
This is simply because, for any sets A and B, A ∩ B = B ∩ A, which
follows from the definition.
However, not all quantifiers possess this kind of symmetry. The
notable exception in Section 7.3.2 is all: all pandas are vegetarian does not
mean the same thing as all vegetarians are pandas. In terms of set member-
ship, this is because the relationships C ⊆ D and D ⊆ C are not equiva-
lent – in fact, the only time they can both be true at once is if C = D.
So the only sentences with all which can be inverted in this way are the
ones in which the two sets being referred to have precisely the same
members, as in (7.24).
(7.24) a. All even numbers are divisible by two.
b. All numbers divisible by two are even.
We could think of all as a special case of a larger class of items called
proportional quantifiers, which express the proportions of sets that
have particular properties. Some other examples are given in (7.25).
(7.25) a. Most pandas are vegetarian.

b. Less than 50 per cent of pandas are vegetarian.
c. Few pandas are vegetarian.
In terms of sets, we can think of these as making statements about the
proportion of members of the set P that are also members of the set V.
(7.25b) is perhaps the most clear-cut example: it appears to mean that
less than 50 per cent of the members of P are also members of V. We can
formalise that meaning in terms of sets in various ways. For instance,
recall that everything in P is either a member of V or a member of its
complement, V′, the set of animals that are not vegetarians. If less than
50 per cent of the members of P are members of V, more than 50 per
cent of the members of P must therefore be members of V′. This is
equivalent to saying that there are more things in the intersection of P
and V′ than there are in the intersection of P and V. So we can express
the meaning of (7.25b) as (7.26) – put into different words, ‘the number
of non-vegetarian pandas exceeds the number of vegetarian pandas’.
(7.26) |P ∩ V′| > |P ∩ V|
We could say something very similar for (7.25a): most seems to mean
‘more than half’, in which case we could write this meaning down as
(7.27): ‘the number of vegetarian pandas exceeds the number of non-veg-
etarian pandas’. In practice, there are reasons to think that the meaning of
most is a bit more complicated than that: for instance, it seems odd to say
that Most Americans are female, but that statement is factually correct if most
simply means ‘more than 50 per cent’ (see Solt 2011).
(7.27) |P ∩ V| > |P ∩ V′|
Example (7.25c) is also tricky to define precisely. Few in this sense
does not appear to refer to a stable number, or range of numbers: it
appears to be context-dependent. For instance, it appears to denote a
much smaller number in (7.28a) than in (7.28b).
(7.28) a. Few people have been into space.
b.
Few people voted in the elections to the European
Parliament.
One way to approach this might be to propose that few requires the set
of individuals with the property in question to be much smaller than
the set of individuals without that property: that is, in (7.25c), that veg-
etarian pandas are greatly outnumbered by non-vegetarian pandas. We
could write this down with the expression in (7.29), where << denotes
the relation ‘much less than’.
(7.29) |P ∩ V| << |P ∩ V′|

The problem is that, although this also works for (7.28a), it doesn’t work
for (7.28b) – in the 2019 elections the turnout was 50.7 per cent, so more
of the people eligible to vote voted than abstained. There is still a sense
in which we could call this few, notably by comparison with other elec-
tions (the UK General Election in 2019 had 67.3 per cent turnout, the
US Presidential Election in 2020 had 66.9 per cent, the second round
of the French Presidential Election in 2022 had 72.0 per cent, and so
on). But we can’t describe the voters in the elections to the European
Parliament as a minority, or as a small number of people by the stand-
ards of democratic participation (nearly 200 million votes were cast). So,
once again, trying to pin down the relevant sense of few that explains its
usage is a challenging exercise.
7.3.4 Distributivity and collectivity

Yet another distinction in the domain of quantification is illustrated by
(7.30).
(7.30) a. All pandas are vegetarian.
b. Every panda is vegetarian.
c. Each panda is vegetarian.
These three sentences encode the same meaning, in terms of set mem-
bership, namely P ⊆ V. However, there is a sense in which they differ
in meaning. Every and each are distributive quantifiers: we can think of
them as ranging over the members of a set and attributing a property (in
this case, vegetarianism) to each of those members. By contrast, all can
attribute that property to the totality of the set (of pandas) in one fell
swoop. As a consequence, all can give rise to collective readings that are
not possible with each and every. (7.31a) potentially describes a different
situation from (7.31b), and (7.31c) is acceptable where (7.31d) is not.
(7.31) a. All the students lifted a piano.
b. Each/every student lifted a piano.
c. All the students together lifted a piano.
d. *Each/?every student together lifted a piano.
Essentially, we have another case of ambiguity with sentences with all
like (7.31a). On one reading, this sentence attributes an achievement
(‘lifting a piano’) to all the members of a set (‘students’) considered
individually (the same reading that is obligatory for (7.31b)), while on
another reading it attributes the achievement to all the members of the
set considered collectively.
This is a complex issue, when considered in detail, but I sketch it

briefly here just to reinforce the point that many quantifiers have sub-
tleties of meaning that are not easily captured in a straightforward set-
theoretic analysis.
7.3.5 Quantifier scope

Like modals and negation, quantifiers can be considered as operators,
with the ability to take scope over things. In effect, they take scope over
clauses, which express propositions; the quantifiers themselves can be
thought of as propositional operators. And just as the scope relations
between modals and negation can vary, and give rise to different mean-
ings, so can the scope relations between quantifiers and other operators.
The examples in (7.32) involve a quantifier and negation.
(7.32) a. All the books were not available.
b. Not all of the books were available.
c. Some of the books were not available.
(7.32a) and (7.32b) both involve the quantifier all and negation, expressed
by not. (7.32b) is unambiguous and means ‘it is not the case that all the
books were available’. It also potentially conveys that ‘some of the books
were available’, but that can be explained as a pragmatic inference, as
we shall see in Section 8.2.2. By contrast, (7.32a) admits two different
interpretations: ‘none of the books were available’ and ‘it is not the
case that all the books were available’. The former is perhaps the more
obvious interpretation, when the sentence is presented out of context,
but the latter interpretation is available, especially if there is emphasis
on not, and if (7.32a) is a reply to Were all the books available?
Similarly, although (7.32c) has a preferred interpretation (‘for some of
the books, it is the case that they were not available’), it can also be read
in a different way (‘it is not the case that some of the books were avail-
able’: that is, ‘none of the books were available’). This reading is easier
to obtain if you consider (7.32c) as the answer to the question Were some
of the books available?, and again think of it with emphasis on not.
These pairs of readings are associated with different scope relations
between the quantifier and the negation. The more obvious readings of
(7.32a) and (7.32c) are those for which the quantifier takes scope over
negation – we could paraphrase these respectively as All of the books were
unavailable and Some of the books were unavailable. The less obvious read-
ings are those in which the negation takes scope over the quantifier: It is
not the case that all/some of the books were available.
Musolino et al. (2000) proposed a generalisation about sentences such
as these, which they call the “observation of isomorphism” – they argue

that the preferred readings are those in which the semantic scope rela-
tions match the syntactic scope relations. In the cases above, the syntax
suggests that the quantifiers should take scope over negation, and that
corresponds to the semantic meaning that we obtain. However, it is
curious that the “non-isomorphic” readings – those in which the seman-
tic scope relations don’t match the syntactic scope relations – are still
available, and may even be preferred in certain contexts (for instance,
in response to the questions suggested above) or with certain intona-
tion patterns. The details of how this works are still being teased out by
experimental work. It would probably be fair to say that hearers appear
to take into account not only the syntax, but also the discourse context
and any special features of how the utterance is made, when disambigu-
ating sentences of this kind.
Summary
Must, should, can and similar expressions encode modality. Markers of
modality are principally understood as either epistemic (related to the
perceived likelihood of something being true) or deontic (related to the
status of the situation as obligatory, forbidden, and so on). Many modal
expressions are, at least in principle, ambiguous between epistemic
and deontic interpretations. Quantifier meaning can be expressed in
terms of sets, and this way of looking at quantifiers elucidates the kinds
of ambiguities that can arise from the sentences that contain them. In
particular, sentences with multiple scope-bearing operators (such as
modals, quantifiers and negation) can give rise to multiple interpreta-
tions as a result of the interactions between these operators. For hearers
to understand the speaker’s intended meaning in the many potentially
ambiguous cases involving modals, quantifiers and negation, they may
have to integrate knowledge about the expressions’ meanings with
information about the context of utterance, an idea that will be further
explored in the coming chapters.
Exercises
1. There are differences in strength between modal verbs when they
are used to indicate how certain a speaker is about a conclusion.
What about using no modal verb at all – how strong is that? Consider
the following situation: Edward has seen crowds streaming into a
department store and says either There might be a sale on or There’s
a sale on or There must be a sale on. Rank these three in terms of how
confident Edward seems that there is indeed a sale on in the store.

Comment on what we can infer about speakers’ knowledge of a situ-
ation if they use a modal verb in talking about it.
2. Think about possible interpretations of the modality in the six sen-
tences below. Can they be understood as deontic, epistemic, both, or
neither? Give a reason for each answer.
(a) They must be made from buckwheat.

(b) We must get up early tomorrow.
(c) The email needn’t have been sent.
(d) I can hear you now.
(e) They might or might not make it.
(f) You better apologise.
3. Propose alternative scenarios for each of the following three sen-

tences that could lead to their being interpreted (1) epistemically
and (2) deontically.
(a) Guests may check in between 3pm and midnight.

(b) You must be a musician.
(c) He might say something.
4. In terms of relative scope, can’t P means ‘not (possibly P)’, deontically

as well as epistemically. The same holds for cannot P. What about may
not? They may not have an invitation can be understood either deonti-
cally (‘I forbid them having an invitation’) or epistemically (‘Perhaps
they do not have an invitation’). What is the scope of negation rela-
tive to the scope of modality for these two interpretations?
5. I once heard the question asked on the radio “Which category of
witness may | not be named in court?” The vertical line indicates
that there was a sharp intonational break after may. It was clear from
the context that the question was about unusual circumstances in
which a court would consent to certain witnesses having the protec-
tion of anonymity. For this meaning, what are the relative scopes of
modality and negation? To keep things simple, answer with respect
to the sentence The witness may | not be named.
6. In this part of the factory, one machine tests each product. The underlined
clause is ambiguous in terms of relative scope. State the two possible
meanings clearly.

Kearns (2011) provides good coverage of the topics covered in this
chapter. Huddleston and Pullum (2002) contains a substantial survey
of modality, with many persuasive examples, along with a useful treat-
ment of quantifiers. Van der Auwera and Plungian (1998) try to map out
the ways in which the encoding of modality can and can’t change in the
history of a language: this might be an interesting paper to look at if you
are studying Old English.
8 Pragmatic inference
Overview
Up to now, this book has mainly focused on semantics – the abstract
meanings associated with linguistic forms such as words and sentences.
We can think of semantic meanings as being present independently
of the context in which the linguistic forms are being used; or, to put
it another way, we can think of them as being contributed by these
forms whenever and wherever they appear. However, it’s clear that
context also plays an important part in meaning, and literally the same
word or sentence may convey a different meaning depending on the
circumstances in which it is uttered. To explain this, we need to appeal
to pragmatics, as defined in Section 1.3: the study of utterance meaning,
which is concerned with explaining how things can be communicated
without being literally said.
Pragmatics is relevant to us at a number of levels. At a high level, it
is about how we do things with words: that is, how language enables
us to perform social actions, which will be the topic of Chapter 11. At
a lower level, it is about how we interpret the meanings of individual
words, or referring expressions: should we take them ‘literally’ or in
some sense idiomatically, as in cases such as metaphor and hyperbole?
We will discuss this in Chapter 9. But in this chapter we start in the
middle, and focus on how additional sentence-level meanings (that is
to say, propositional meanings) can arise when sentences are uttered in
particular contexts. The questions of what meanings arise this way and,
perhaps more interestingly, how language users convey and recognise
these meanings have been major issues in linguistic pragmatics since the
work of Paul Grice in the mid-twentieth century (although the focus
here will be on the ideas themselves rather than their history).
An important insight into how we use language is that, as Levinson
(2000: 29) puts it, ‘inference is cheap, articulation expensive’. That is
to say, it often appears to take very little effort for a hearer to infer an
101
a dditional meaning that the speaker does not make explicit: this seems
to be something that humans are particularly good at, and we can do it
in hundredths of a second. But it is often much more effort to convey
that meaning explicitly: the speaker needs to expend time and energy
actually uttering the necessary words. And this applies not only to
speech but also to writing, typing, texting, using sign language, and
so on. In all these modes of communication, language users can save
themselves effort by producing utterances (we’ll continue to call these
“utterances” regardless of whether they are spoken, written or signed)
that are not completely explicit in their meaning, but can be enriched by
the hearer (the recipient) by taking the context into account.
We have already seen some examples of hearers using context to
enrich meanings: for instance, the bridging inferences discussed in
Section 3.1.1, and the inferences about spatial orientation considered in
Section 3.1.4. But in what follows we’re going to be interested in richer
kinds of inference than that – inferences which don’t just serve to dis-
ambiguate what the speaker is talking about, but allow the speaker to
convey all sorts of additional meanings without encoding them explic-
itly in words.
As an illustration of some of the ways in which we can use context
to enrich meanings – and how similar utterances can convey strikingly
different meanings as a result – consider (8.1).
(8.1) I was bitten by something in Berlin Zoo.
The hearer of (8.1) might guess that the something that bit the speaker
was probably an insect or similar creature. How? Well, the use of some-
thing rather than someone suggests that the assailant was an animal rather
than, say, an enraged zookeeper. But the use of something rather than a
specific noun makes this utterance much less informative than it could
be. In Section 8.1, we will discuss how the use of a relatively underin-
formative utterance like this typically invites the hearer to infer that the
speaker is not in a position to make a more informative statement: in this
case, the hearer infers that the speaker does not know precisely what it
was that bit them.
Note that things don’t always work out this way, as we see when we
contrast (8.2) with (8.1).
(8.2) I was struck by something in Berlin Zoo.
In (8.2), struck by is ambiguous between a sense in which something
physically hit the speaker and a sense in which the speaker thought of
something. If we adopt the first interpretation, we might again take some-
thing to convey that the speaker can’t be more specific about what the
pr agmatic inference 103
thing was; but if we adopt the second interpretation, something doesn’t

seem to work that way. Instead, we rather expect the speaker to go on
and tell us exactly what it was that struck them.
In short, pragmatic reasoning at the utterance level often seems
perfectly natural and unexceptional, but turns out on closer inspection
to be surprisingly subtle and complicated. Various theories and frame-
works have been proposed to help us better understand what is going
on when we reason at this level: Sections 8.2 and 8.3 briefly introduce a
couple of the most influential ideas in this respect.
Another important point about these examples is that the speaker can
make use of what they know about the hearer – and what kind of real-
world knowledge is likely to be shared – in order to say a lot with few
words. In (8.1), it’s common knowledge that the animals exhibited in a
zoo are usually identified by signs: hence, if the speaker doesn’t know
what bit them, it probably wasn’t one of the exhibits. The hearer can use
their knowledge to infer this – and the speaker knows that the hearer
can do that, and therefore doesn’t need to clarify this point explicitly.
In Section 8.4, we’ll look at how speakers can exploit hearers’ (assumed)
knowledge about the way things are in order to convey additional infor-
mation efficiently.
8.1 Some ways of conveying additional meanings

One of the crucial insights of pragmatics is that speakers and hearers
share a rich set of expectations about how communication usually
works. We can exploit these expectations to make our communication
more effective, by conveying additional information that goes beyond
what we “literally” say. Consider (8.3), in which A is enquiring about the
progress of B’s new relationship.
(8.3) A: Have you met Lucy’s parents?
B: I’ve met her mother.
From what B says, A is entitled to infer that B has not met Lucy’s father.
This is clearly not based just on the meaning of I’ve met her mother (‘B has
met Lucy’s mother’). It’s easy to imagine that sentence being uttered in
other contexts where it wouldn’t convey anything about whether or not
the speaker met anyone’s father. Even in (8.3), the inference is poten-
tially deniable: B might continue by uttering I already knew her father
without self-contradiction.
What is special about the context in (8.3) that suddenly makes this
extra meaning available? Essentially, it is that A has asked a question,
which we understand B to be answering. If B had met both of Lucy’s
parents, B could simply answer yes. Instead, B produces a lengthier

response which is, on the face of it, less informative than this yes
response would be. Why should B do this instead of just saying yes? The
obvious explanation is that B knows that it would be false to say yes, and
thinks that it would be inappropriate to make a false statement. If that is
the explanation we accept, we infer ‘B has not met Lucy’s father’.
We can tell a similar story about (8.4).
(8.4) A: Did you eat the cookies?
B: I ate some of the cookies.
Here, we intuit that B is saying that they ate some but not all of the
cookies: if B ate all the cookies, they could have said all instead of some.
That would be more informative, in that it would tell A about what
happened to all of the cookies rather than only what happened to some
of them.
Now, as I mentioned in Section 7.3.2, we could argue instead that some
just literally means ‘some but not all’, in which case B asserts in (8.4)
that they did not eat all of the cookies. But this doesn’t seem quite right.
For instance, if B had eaten all of the cookies but mistakenly thought
there were some left over, it’s not clear that they would have been
speaking falsely in (8.4). And it seems possible to utter (8.5) without
self-contradiction.
(8.5) I ate some of the cookies, perhaps all of them.
So the general consensus about this kind of example is that some liter-
ally contributes the meaning ‘a non-zero amount’ and the interpreta-
tion ‘not all’ arises because of an inference. Just like in (8.3), this is an
inference motivated by the expectation that the speaker would make
a more informative statement if it were true. And once again, there
are circumstances in which this inference might not be available: for
instance, if the speaker happened to be reluctant to give full informa-
tion (perhaps because they didn’t really want to admit to eating all the
cookies).
One way in which (8.4) does differ from (8.3) is that B’s utterance in
(8.4), considered on its own and out of context, still seems to convey
that ‘B ate some but not all of the cookies’. That is to say, this additional
meaning arises just as a consequence of the use of some. In (8.3), the
use of her mother conveys the meaning ‘not her father’, but only does so
because it occurs in a context in which both parents have been men-
tioned. If A had asked instead whether B had met Lucy’s mother and
aunt, B’s utterance would have conveyed ‘not her aunt’. By contrast, in
(8.4), all doesn’t need to be mentioned in order for some to convey ‘not
all’: it is almost as though all is always a relevant alternative, whenever

some is uttered.
The effects of context can be even stronger than in (8.3), and there
can be quite a tenuous relationship between what a sentence means
and what it can be used to convey. B’s utterance in (8.6) is an example
of this.
(8.6) A: Is Bill a good lecturer?
B: He has an interesting beard.
Here we come away with the impression that B means that Bill is not
a good lecturer, even though this is completely unconnected to what
B actually says. By changing the subject, and ostentatiously refusing
to answer A’s question, B conveys a reluctance to answer it. There are
several possible explanations for this – perhaps B does not want to pass
judgement on Bill’s lecturing, or perhaps B dislikes Bill and does not
want to praise him despite him being a good lecturer – but the one that
leaps to mind is that B thinks that Bill is a bad lecturer but does not want
to go on the record as saying so explicitly.
Whichever interpretation of B’s utterance we ultimately prefer, the
first step in understanding it is to recognise that its apparent irrelevance
is probably intentional. Cooperative speakers don’t tend to make irrel-
evant remarks in response to questions, so B’s utterance alerts us that
something is wrong here: for some reason, B can’t act in the way that we
expect them to. This realisation triggers us to look for a plausible expla-
nation for what B is actually doing, drawing upon our understanding
of other social conventions (for instance, politeness, which might make
someone reluctant to voice criticisms directly).
In the examples so far, we’ve seen how we can draw inferences based
on what a speaker does not say. Sometimes we can also draw inferences
based on the particular way a speaker says something. Compare (8.7a)
and (8.7b).
(8.7) a. This morning I got up, had breakfast and checked email.
b. This morning I checked email, had breakfast and got up.
Both versions of this state that the speaker performed the three specified
actions earlier that day, and neither version overtly mentions the order
in which these actions took place. Yet (8.7b) in particular seems very
strongly to suggest that the sequence of events was one in which the
speaker first checked email, then had breakfast, and only then got up.
That is to say, we seem to be disposed to think that the order in which
events are described ideally should match the order in which they hap-
pened. Given the choice, we assume that the speaker is likely to say
things in some systematic order – and in this example, chronological

order seems like a reasonable choice. The speaker can exploit this
expectation in order to convey that additional nuance of meaning
without using additional words.
8.2 The Gricean maxims

Language users may have been drawing inferences like those discussed
in Section 8.1 since time immemorial, but theorists have only rela-
tively recently attempted to get to grips with how this all works. The
most influential idea is due to Grice (1975), who suggested that a wide
range of these additional meanings – which he termed (conversational)
implicatures – could be explained in terms of how we reason about
communicative norms.
Specifically, Grice proposed that conversation participants could be
expected to adhere to certain norms of behaviour, which he termed the
Cooperative Principle. He tentatively proposed four features of coop-
erativity, which he articulated as a set of maxims, summarised below:
• Maxim of Quality: Be truthful. Do not say things that are false or for
which you lack adequate evidence.
• Maxim of Quantity: Give the appropriate amount of information,
not too little and not too much.
• Maxim of Manner: Be clear, brief and orderly.
• Maxim of Relation: Make your contribution relevant to the current
goals of the conversation participants.
The word maxim can be defined in various ways, one of which is ‘a
general principle that serves as a guide or rule’. Importantly, though,
Grice’s maxims are not intended as a guide to how we ought to conduct
conversations, or rules that we must follow in order to be properly
cooperative. Grice simply claimed that communication proceeds as
though speakers are generally guided by these maxims. In the following
subsections, we’ll revisit the examples discussed so far in this chapter,
and see how they are predicted to arise with reference to Grice’s
maxims. In Section 8.3, we’ll then briefly discuss an alternative approach
– Relevance Theory – that explains these pragmatic enrichments in a
somewhat different way.
8.2.1 Quantity implicatures

With Gricean cooperativity in mind, we can revisit (8.3), repeated
below, and describe it as an apparent violation of the Maxim of Quantity.
A asks whether B has met both of Lucy’s parents, but B overtly provides
less information than was asked for.
Assuming that B is a cooperative speaker and therefore tends to act in
accordance with Grice’s maxims, how can we explain this? A possible
explanation is that for B to utter the stronger statement (yes) would
somehow be incompatible with being cooperative. And the only way
that saying yes would be uncooperative is if saying yes would violate the
Maxim of Quality, which states that speakers do not make false state-
ments or statements for which they lack evidence. If we take this to be
the explanation, we take B’s utterance to convey that yes might be false
– in other words, that B cannot confidently assert that B has met both of
Lucy’s parents. And if we are willing to assume that B knows whether
or not this is true, we conclude that B hasn’t met both of Lucy’s parents.
In other cases, such as (8.8), we might not be able to take this last
inferential step, because the speaker might quite plausibly not know
whether the stronger statement is true.
(8.8) A: Is Rebecca in the office?
B: Her car’s in its usual space.
Here, B falls some way short of answering A’s question – Rebecca’s car
could be in its usual space whether or not Rebecca is presently in the
office. Assuming that B is being as cooperative as possible, we can again
conclude that B is blocked from simply answering yes by the worry that
doing so would violate the Maxim of Quality. But in this case it may be
more natural to assume that B simply doesn’t know whether or not it
would be true to say yes – so we cannot draw the inference that Rebecca
is not in the office. We have to stop at the weaker conclusion that B does
not know for certain whether or not Rebecca is in the office.
Inferences such as these are often called quantity implicatures,
because they can be explained in terms of the Maxim of Quantity:
specifically, the expectation that cooperative speakers provide as much
information as is necessary for the current discourse purpose. Quantity
implicatures are inferences about a stronger proposition than that which
was explicitly stated, and they are always to the effect that the speaker
cannot commit to that stronger proposition. If we can assume that the
speaker is knowledgeable about its truth or falsity, the quantity implica-
ture is that the stronger proposition is in fact false.
Quantity implicatures also rely crucially upon the Maxim of Quality.
A crucial step in their derivation is the inference that the stronger
a lternative would have violated Quality, because it would have involved

the speaker saying something for which they lacked adequate evidence,
or which they knew to be false. This only works because the Maxim
of Quality is part of the system, and in particular because speakers are
willing to sacrifice Quantity in order to maintain Quality: they would
sooner be underinformative than inaccurate. It was certainly Grice’s
own view that the Maxim of Quality was privileged over the others: he
wrote that ‘other maxims come into operation only on the assumption
that this maxim of Quality is satisfied’ (1975: 46). Adherence to Quality
is apparently the central plank of what it means to be a cooperative
conversation participant.
8.2.2 Scalar implicatures

Example (8.4), repeated below, represents a special case of quantity
implicature, called scalar implicature.
(8.4) A: Did you eat the cookies?
B: I ate some of the cookies.
In this context, we can treat B’s utterance in (8.4) just like B’s utterance
in (8.3). B could have made a stronger statement (I ate all of the cookies).
If B is cooperative, a plausible reason for their failure to do this is that
they do not know the stronger statement to be true. If we further assume
that B knows whether or not that stronger statement is true, we can infer
that it is in fact false, and B did not eat all of the cookies. Once again,
this inference is cancellable: B can deny it without self-contradiction, by
continuing and possibly all of them.
Recall that the striking difference between (8.3) and (8.4) is that the
implicature in (8.3) – like that in (8.8) – relies on the presence of a very
specific discourse context, whereas that in (8.4) does not. The inferences
in (8.3) that B has not met Lucy’s mother, and in (8.8) that B does not
know whether Rebecca is in the office, are not normally triggered by
the sentences that B utters. By contrast, the inference in (8.4) that B did
not eat all of the cookies appears to arise just on the basis of B’s utter-
ance, regardless of the context in which it occurs. To put that in Gricean
terms, I’ve met her mother and Her car is in its usual space don’t strike us as
likely violations of the Maxim of Quantity: however, I ate some of the
cookies does.
A popular explanation for this is that the word some is special, and
specifically that it stands in a special kind of relationship to the alterna-
tive all. The idea here is that, after hearing an utterance involving some,
our attention is drawn to the possibility that the speaker could have
said all instead, and thus the use of some triggers a quantity implicature.
We then interpret some as meaning ‘some but not all’ if the speaker is
assumed to know about the truth or falsity of the corresponding all
statement, and otherwise as conveying that the speaker is uncertain
whether all is the case.
On this view, <some, all> is one example of an informational scale
(also called a Horn scale, in reference to Horn (1972)), but there are
potentially many others. These don’t just involve quantifiers like some
and all: they can involve modal expressions, gradable adjectives or
adverbs, or verbs. Some examples are given in (8.9).
(8.9) a. It’s possible that the team will win.
b. This coffee is warm.
c. Yvonne likes Ali.
In (8.9a), we can think of possible as belonging to a scale with certain, and
this sentence consequently conveying that ‘it is possible but not certain
that the team will win’. In (8.9b), warm forms a scale with the stronger
adjective hot, thus causing the sentence to convey that ‘this coffee is
warm but not hot’. In (8.9c), likes forms a scale with loves, causing the
sentence to convey that ‘Yvonne likes but does not love Ali’. All these
inferences are cancellable, and all of them depend on the assumption
that the speaker knows whether the stronger alternative is true or false.
Scales like this reverse in direction when we’re dealing with negative
sentences, like (8.10).
(8.10) It’s not certain that the team will win.
In this context, replacing certain with possible would make the sentence
stronger. So we would correctly expect (8.10) to implicate ‘it’s possible
that the team will win’ – because if the speaker didn’t think that this
would be possible, they would have said so. In a similar way, (8.11a)
and (8.11b) potentially admit implicatures that ‘this coffee is warm’ and
‘Yvonne likes Ali’.
(8.11) a. This coffee isn’t hot.
b. Yvonne doesn’t love Ali.
8.2.3 Relevance implicatures

We can revisit (8.6), repeated below, and think about it in terms of the
Maxim of Relation.
(8.6) A: Is Bill a good lecturer?
B: He has an interesting beard.
On the face of it, B’s utterance here is a useless attempt to answer A’s
question: it seems very unlikely that the character of Bill’s beard is rele-
vant to his competence as a lecturer. Thus, B appears to be violating the
Maxim of Relation. However, once again, we can devise an explanation
of B’s behaviour that preserves the assumption that B is cooperative. In
this case, it’s that B has refrained from providing relevant information
because none of the possible answers are available to B. Specifically, we
infer that B cannot say yes (or even I don’t know) without violating the
Maxim of Quality, and hence that Bill is a bad lecturer. We can think
of this as a relation implicature, more usually called a relevance impli-
cature: it arises because the speaker has apparently violated the Maxim
of Relation in order to adhere to the Maxim of Quality. Of course, we
still require an explanation of why B doesn’t just say no – and as we dis-
cussed before, the natural explanation for that seems to be that it would
be impolite to do so. This doesn’t follow from Grice’s maxims, though:
to explain this, we would have to introduce some additional maxims
concerning politeness, as proposed by Leech (1983).
(8.6) is in some ways a slightly odd example, because B is deliberately
responding to A’s question with irrelevant information, in order to
convey the message that they are not willing to say anything relevant.
We can get more direct relevance implicatures, as in cases like (8.12).
(8.12) A: Are you going to the party?
B: It’s Wednesday.
Here it looks as though B is trying to say something relevant in answer
to A’s question, but is leaving us some work to do in order to figure out
how their utterance is relevant. In this case, we infer that the fact of it
being Wednesday is relevant to whether or not B is going to the party.
If we know about the context, we might be able to draw more specific
inferences: for instance, if we know that B works early on Thursday
mornings, we might infer both that B is not going to the party and that
this is because of that work commitment. As we will see in Chapter 11,
we have a strong expectation that questions are likely to be followed by
answers, so we will attempt to interpret the responses to questions as
answers, if at all possible. (8.12) is a case in which B may be exploiting
that tendency in order to convey a rich message to A in very few words,
by bringing about relevance implicatures.
8.2.4 Manner implicatures

Finally, let’s revisit (8.7), repeated below, and think about it in terms of
the Maxim of Manner.
(8.7) a. This morning I got up, had breakfast and checked email.
b. This morning I checked email, had breakfast and got up.
As discussed earlier, (8.7b) suggests that the three events mentioned
happened in the order in which they are presented in the sentence.
This could be because we assume that the speaker, being cooperative, is
adhering to the Maxim of Manner, which requires us to be orderly. The
speaker of (8.7b) has chosen to present the events in a specific chrono-
logical sequence, and it is reasonable to think that this might map to the
chronological sequence in which they happened.
Having said that, there are lots of other possible reasons why a coop-
erative speaker, adhering to the Maxim of Manner, might mention
events in a particular order. They might be placed in order of impor-
tance, for instance. Or they might be causally linked in a particular way.
Consider (8.13).
(8.13) a. We sold our car and bought a tandem.
b. We bought a tandem and sold our car.
In these cases, we might infer that the event of selling the car pre-
ceded that of buying the tandem in (8.13a), and followed it in (8.13b).
However, we can also read these sentences as expressing richer causal
relations. From (8.13a), we might infer that selling the car provided the
money that made it possible to buy the tandem; from (8.13b), we might
infer that the tandem turned out to render the car unnecessary. This
difference in interpretation may arise because causes logically precede
effects, so we expect the orderly speaker to mention the cause first (even
though the language provides us with ways of reversing that order of
mention, such as because).
Of course, sometimes the order doesn’t really matter. Speech is nec-
essarily linear: we can’t say everything at once, so we have to choose
an order in which to present multiple pieces of information, even if
they are not linked by any temporal or causal relation. (8.14) is of this
kind: the (a) and (b) versions appear to be essentially paraphrases of one
another, and both are equally compliant with the Maxim of Manner:
there is little scope to draw any pragmatic inference.
(8.14) a. Her name is Moira and his name is Jon.
b. His name is Jon and her name is Moira.
There is more to Manner than just being orderly. In his formulation
of the Maxim of Manner, Grice also points to the importance of being
brief. When an utterance is longer than seems necessary, it may invite a
manner implicature, as in (8.15).
(8.15) Helen caused the car to stop.
(8.15) appears to mean ‘Helen stopped the car’, but expresses that idea in
an unusually roundabout way. By violating our Manner-based expecta-
tions, the speaker seems to convey that something other than the usual
sense of ‘stopping the car’ is intended: that is to say, (8.15) suggests that
Helen stopped the car by some other means than applying the brakes.
As Levinson (2000: 136) summarises it, ‘What is said in an abnormal way
indicates an abnormal situation.’ However, like the other implicatures, the
manner implicature that arises is cancellable – it would still be possible
to say (8.15) if Helen had been driving the car and had braked it to a stop.
As with quantity implicatures, manner implicatures of this kind rely
on the idea that there was some alternative that the speaker could have
uttered in place of their actual utterance, but chose not to. However,
quantity implicatures are about the speaker not being able to utter
that alternative because it is, or might be, false. Manner implicatures
are about the speaker not being able to utter that alternative because,
although technically true, it might be misleading – it might give the
hearer a false impression of what is going on.
8.3 Relevance Theory

In the previous section, we used Grice’s maxims to give a more system-
atic account of the way in which hearers can calculate (and speakers
can convey) implicatures. However, although Grice’s maxims can be a
useful way of describing human communicative behaviour, we might
still be sceptical that they represent the best way of doing so. As we dis-
cussed earlier, the Maxim of Quality seems to have a privileged status,
and it’s not clear that it should be placed alongside the others (a point
that Grice himself made). It’s also difficult to provide precise and non-
overlapping definitions for each of the maxims, particularly the Maxim
of Relation and the Maxim of Manner.
Several more recent approaches to pragmatics, such as those of Horn
(1984) and Levinson (2000), have reformulated the system around
different sets of maxims or principles. A more radical departure from
Grice’s system is offered by Relevance Theory (RT), proposed by
Sperber and Wilson (1995).
As the name suggests, Relevance Theory proposes that the notion of
relevance is fundamental to communication. Sperber and Wilson argue
that we, as humans, have a natural tendency to look for relevance in the
stimuli that we experience, and speakers can exploit this tendency in
order to communicate more efficiently.
We can explore some of these ideas with reference to examples such

as (8.12), repeated below.
(8.12) A: Are you going to the party?
B: It’s Wednesday.
Here, B can communicate additional meanings – beyond the explicit
‘Today is Wednesday’ – by exploiting the fact that A will try to inter-
pret B’s utterance as an answer to the question. As discussed earlier, the
precise meanings that are conveyed will depend on shared knowledge.
The crucial point here is that A will proactively look for relevance in
B’s utterance, rather than first dismissing it as a violation of the Maxim
of Relation and then trying to figure out why B would do such a bizarre
thing. If A did not look for relevance, they might ultimately conclude
that B had misheard or ignored the question, and B would fail to achieve
their communicative aims.
Sperber and Wilson define relevance in terms of two components,
which they call “cognitive effects” and “cognitive effort”. A (positive)
cognitive effect is anything that makes a positive difference to the hear-
er’s representation of the world: for instance, knowing that a particular
proposition is true. The utterances we hear are a potential source of cog-
nitive effects, because we can use the information in them (potentially
combined with our background knowledge) to learn new things about
the world. Cognitive effort refers to the quantity of mental resources,
such as memory and inferential processing, that we use in order to obtain
these cognitive effects. All things being equal, an utterance (or any stim-
ulus) is more relevant if it yields greater cognitive effects, and it is more
relevant if it requires less cognitive effort to process.
The central claim of Relevance Theory is that human cognition
has evolved in such a way as to maximise relevance, in terms of how
we process linguistic inputs. As Wilson and Sperber (2002: 254) put it,
‘our perceptual mechanisms tend automatically to pick out potentially
relevant stimuli, our memory retrieval mechanisms tend automatically
to activate potentially relevant assumptions, and our inferential mecha-
nisms tend spontaneously to process them in the most productive way’.
Coupled to this principle, RT posits a second principle that applies to
linguistic communication (and certain other forms of communication),
namely that every utterance conveys the fact of its own relevance.
Thus, when we hear an utterance, we are entitled to assume that it is
worth our while to process it, and also that it is the most relevant thing
that the speaker could manage to produce.
These ideas are quite compatible with the inferences that we’ve
already discussed in this chapter. Let’s turn once again to (8.3).

In RT terms, it would have been more relevant for B to say yes, if that had
been true: that way, A would have access to more information (greater
cognitive effects) and would have had to process a simpler stimulus (less
cognitive effort). Assuming that B is attempting to maximise relevance,
we infer that there must be some reason why they couldn’t say yes, even
though it would be less effort for the hearer. We thus conclude that yes
wouldn’t have led to greater cognitive effects – and this must be because
A would have drawn incorrect conclusions, had B said yes. So it must be
the case that yes would not have been entirely true.
A similar argument applies to (8.15), repeated below.
(8.15) Helen caused the car to stop.
Again, we need to explain why it wouldn’t have been more relevant
to say Helen stopped the car, given that this would involve less cogni-
tive effort for the hearer to process. We’re forced to conclude that the
speaker intended additional cognitive effects that would not normally
be available to the hearer of Helen stopped the car. Given our background
knowledge about how one would normally stop a car, the inference
emerges from (8.15) that Helen caused the car to stop in some other way
than pressing the brake pedal.
Relevance Theory offers a very useful way to characterise certain
inferences, and offers an interesting perspective on how we perform
pragmatic inference in general. However, as an approach, it has its limi-
tations. There is no consensus about how to quantify cognitive effects
or cognitive effort, so in practice we can’t simply take utterances and
evaluate their relevance according to some scale. Rather, as observed
by Wilson and Sperber (2002: 253), we can compare pairs of alterna-
tives and argue that one is more relevant than the other. This works
reasonably well for simple examples with obvious alternatives like (8.3)
and (8.15), but in more complicated cases it is not always clear how to
compare alternatives. As a result, it is difficult to disprove any claim
couched in terms of relevance: if we draw an inference from a particular
utterance, we can say that the effects must have justified the effort, and
if we don’t draw an inference from that utterance, we can say that the
effects must not have justified the effort. In such cases, RT doesn’t very
clearly predict whether or not a particular pragmatic enrichment is
going to be available.
8.4 Presuppositions
When we communicate, we take some things for granted: for instance,

that the hearer knows the meanings of the words that we use, but also
that certain entities exist, certain events have taken place, and so on.
The assumptions that are necessary in order for an utterance to make
sense are called the presuppositions of the utterance. Presuppositions
are also communicatively useful, because speakers can exploit them in
order to convey additional information without having to state it explic-
itly, as we shall see.
Perhaps the most widespread example of presupposition in language
is the so-called “existential” presupposition that is carried by noun
phrases. When we use a noun phrase, we typically presuppose the exist-
ence of the thing it refers to. Consider (8.16).
(8.16) a. Carolyn Cole worked as a photojournalist in Iraq during

the war.
b. Carolyn Cole worked as a photojournalist in Liberia during
the war.
In (8.16a), the speaker uses the referring expressions Carolyn Cole, a

photojournalist, Iraq and the war. In order for (8.16a) to mean anything, all
the entities referred to by these expressions have to exist: (8.16a) would
be meaningless if there were no such person as Carolyn Cole, no such
profession as photojournalism, no such place as Iraq or no such thing as
“the war”. Here, the speaker might reasonably assume that it’s common
knowledge that there has been a war in Iraq, and can thus use the war to
refer to that war.
In (8.16b), the situation is a little different: it’s not so likely to be
common knowledge that there has been a war in Liberia. By referring
to the war, the speaker is acting as though it were common knowledge.
Assuming that the speaker is cooperative and isn’t making a meaning-
less statement, the hearer can learn from (8.16b) that there was a war in
Liberia, without the speaker explicitly telling them so. In this case, we
say that the existential presupposition of the war is accommodated: the
hearer has inferred something by identifying it as a presupposition of
what the speaker said.
The possibility of accommodation makes presuppositions useful in
communication. And not all presuppositions are merely existential:
English, like many other languages, provides us with many ways of
signalling the presence of presuppositions of various kinds. (8.17) shows
some examples.
(8.17) a. I had dinner with Jo again.

b. Simon managed to deal with the problem.
c. Amy’s brother is also a medical student.
The use of again in (8.17a) signals that the speaker dined with Jo on a
previous occasion; managed in (8.17b) conveys that the problem was dif-
ficult to deal with; also in (8.17c) probably conveys that someone other
than Amy’s brother – perhaps Amy herself – is a medical student. (In
this case, there is a potential ambiguity, because (8.17c) might mean
that Amy’s brother has some other interesting attribute as well as being
a medical student.) In all three cases, the hearer can accommodate the
information that is (likely) presupposed, on the basis that the speaker
has acted as though that information is already common knowledge.
Presuppositions are essentially background information. In the above
examples, the presupposed content behaves differently from the main
point of the utterance, the asserted content. We can see this by looking
at some more examples, (8.18)–(8.20).
(8.18) a. Hana forgot to post the letter.
b. Hana did not forget to post the letter.
c. Did Hana forget to post the letter?
d. Hana was supposed to post the letter.
(8.19) a. Dick has quit smoking.
b. Dick hasn’t quit smoking.
c. Has Dick quit smoking?
d. Dick previously smoked.
(8.20) a. The treatment has cured her uncle.
b. The treatment hasn’t cured her uncle.
c. Has the treatment cured her uncle?
d. Her uncle was ill.
In each of these sets, the (a), (b) and (c) sentences presuppose the corre-
sponding (d) sentence: the triggers for the presuppositions are the verbs
forget, quit and cure. We can see that (d) is a presupposition in each case,
because if it is not true, the statements (a) and (b) don’t make sense and
the question (c) isn’t well posed.
Because presuppositions are essential to something making sense at
all, they are not affected by negation, or questioning, in the way that the
main (asserted) content of the sentence is. Normally, in saying either
that Hana forgot or Hana did not forget, or in asking Did Hana forget?, we
take it for granted that Hana was supposed to remember. In this respect,
presuppositions are unlike entailments, which do not persist when we
negate the sentence (or question it). For instance, (8.20a) and (8.20b)
both presuppose (8.20d), but only (8.20a) entails Her uncle is now well;
(8.20b) does not.
Having said that, although presuppositions survive negation, it is pos-
sible to cancel presuppositions in some negative sentences, as illustrated
in (8.21). These examples don’t have a strong sense of self-contradiction
about them, although they do seem a little odd, and might need a par-
ticular intonation to be uttered felicitously (for instance, placing extra
emphasis on the presupposition trigger).
(8.21) a. Hana didn’t forget to post the letter; she didn’t know it
needed to go.
b. Dick hasn’t started smoking; he’s been smoking for years.
c. The treatment hasn’t cured her uncle; he wasn’t ill in the
first place.
The above examples are just a small sample of the full set of presupposi-
tion triggers in English. In addition to quit, we have items such as start,
begin, commence, stop, pause, and many others besides. Restitutive again,
discussed in Section 5.2.1, triggers a presupposition about a state or
activity having existed before, and so do resume, restart, return, and so on.
There is also a class of items, often called factives, that have been
extensively studied as presupposition triggers (see Huddleston and
Pullum 2002: 1004–11). These include verbs such as regret, matter, realise
and explain, and adjectives which combine with the verb be, such as odd,
sorry, aware, and so on. They can be used to introduce a clause that the
speaker, in normal circumstances, presumes to be true. A sample of
factive sentences is given in (8.22)–(8.24): again, in each case, (a), (b) and
(c) all presuppose (d).
(8.22) a. It matters that they lied to us.
b. It doesn’t matter that they lied to us.
c. Does it matter that they lied to us?
d. They lied to us.
(8.23) a. Jill explained that the train was late.
b. Jill didn’t explain that the train was late.
c. Did Jill explain that the train was late?
d. The train was late.
(8.24) a. Rob is sorry that the World Cup is over.
b. Rob is not sorry that the World Cup is over.
c. Is Rob sorry that the World Cup is over?
d. The World Cup is over.
For comparison, the sentences in (8.25) use a non-factive verb, prove.

Here, (8.25a) still requires (8.25d) to be true, but (8.25b) and (8.25c) are
acceptable whether (8.25d) is true or not. Hence, (8.25d) is not a presup-
position of these sentences.
(8.25) a. This proves that they lied to us.
b. This doesn’t prove that they lied to us.
c. Does this prove that they lied to us?
d. They lied to us.
Presuppositions can also arise for syntactic reasons. For instance,
time clauses with past reference trigger presuppositions: (8.26a) pre-
supposes (8.26b). So do it-clefts, which will be discussed in more detail
in Section 10.3.2: (8.27a) presupposes (8.27b). In both cases, negat-
ing the (a) sentences (You didn’t . . ., It wasn’t . . .) has no effect on the
presuppositions.
(8.26) a. You loved me when we were in Monterrey.
b. We were in Monterrey.
(8.27) a. It was a meteorite that hit the sofa.
b. Something hit the sofa.
All these kinds of presupposition are available for the speaker to
exploit in communication, by appeal to accommodation. Sometimes
this feels like a by-product of the utterance, as in (8.16b) – while that
may convey that there was a war in Liberia, it would be a rather odd
way for a speaker to try to get this specific point across. But in some
cases the thing that is formally presupposed may be the main thing that
the speaker wants to convey. This seems plausible in a case like (8.28a)
– although (8.28b) is formally just a presupposition of (8.28a), it does feel
like the speaker’s main message, rather than (8.28c).
(8.28) a. I just found out that Sergey is getting a promotion.
b. Sergey is getting a promotion.
c. I just found this out.
This is perhaps not a typical case. We could say of the speaker in
(8.28a) that they “told me Sergey is getting a promotion”. We wouldn’t
normally say of the speaker who said Would you like another cup of tea?
that they “told me I had already had a cup of tea”, even though that’s
presupposed by the use of another. What this kind of example does
indicate is that the single phenomenon of presupposition covers a lot
of ground, from merely establishing the prerequisites for a sentence to
have meaning, all the way to conveying brand new information. Across
this range, presuppositions are unified by their behaviour under nega-

tion and questioning, which distinguishes them systematically from
straightforward entailments.
Summary
Pragmatics is about the use of utterances in context, a particular point
of interest being how we manage to convey more than is literally said.
The additional meanings that we can infer by appeal to context enable
us to “do more with less” – speakers can successfully communicate
their messages in fewer words than they would need if they had to be
completely explicit about their meanings. Traditionally, since the work
of Grice, these implicatures have been treated as inferences that we
can draw by relying on generalisations about how cooperative speak-
ers behave. However, another way of looking at these enrichments is to
think of them as arising from the hearer’s natural tendency to try to find
relevance in the utterances they encounter. When dealing with presup-
positions, speakers are also able to make use of the hearer’s expectation
that their utterances will make sense, and thereby convey additional
information, again without expending extra words to make it explicit.
Later in this book we will look at how speakers and hearers are able
to exploit these enrichments further in order to achieve social goals
through the use of language.
Exercises
1. Consider the following exchange:
A: Who’s that?
B: It’s me.
B’s response appears unhelpful, because it’s something that any

speaker could truthfully utter, so B doesn’t seem to be telling A
anything that is not obvious. What is it that B probably manages
to communicate anyway? Which of Grice’s maxims is potentially
involved in interpreting the utterance? Why might this be preferable
to a more explicit alternative utterance?
2. Suppose that a student asks a university librarian “Where can I find
books on sociolinguistics?”, and the librarian replies “Linguistics is
on the fourth floor”. What might the librarian’s utterance convey,
and – particularly with respect to Grice’s maxims – how does it do
so?
3. Under what circumstances would it be possible to say the following

things without being committed to the presuppositions (given in
brackets)?
(a) Jessica didn’t regret arguing with her boss. (Jessica argued with
her boss)
(b) Tom didn’t go to Canada again. (Tom went to
Canada before)
(c) Vicky didn’t find her keys. (Vicky lost her
keys)
4. Consider a children’s story that begins Once upon a time there was a
beautiful princess whose name was Eleanor. Eleanor had a wicked stepmother,
and they lived together in an enchanted castle. How could that introduc-
tion be made more concise? What would be expressed by presup-
positions, in the concise version, that was stated explicitly in the
original version?
5. By contrast with example (8.16b), if someone told you that Carolyn
Cole worked as a photojournalist in Switzerland during the war, how would
you interpret that? Would you need additional context in order to
make sense of it?

Lycan (2019) provides a good introduction to philosophical accounts
of implicature and presupposition. Wilson and Sperber (2002) is a good
place to start with Relevance Theory. Davis (2019) gives a rapid but
detailed overview of some of the controversies arising in the study of
implicature.
9 Figurative language
Overview
In this book, we have seen a number of examples in which an expression
has multiple possible senses, and a hearer has to disambiguate between
these senses in order to understand what a speaker means. In this
chapter, we will look at various forms of figurative language, which we
can think of as cases in which expressions display different senses from
those we might expect. In particular, we’ll look at metaphor, metonymy,
simile, irony and hyperbole. In each case, an important question will be
how a hearer successfully latches on to the speaker’s intended sense,
even when this is not a sense that is customarily associated with the
expression being used.
9.1 Literal and figurative usage

In Chapter 1, we encountered the idea that people have semantic
knowledge simply by virtue of knowing a language: that is, they under-
stand what expressions mean. And as discussed in Section 2.2, meaning
is compositional: the meanings of sentences depend on the meanings
of the component expressions and the way they are put together. The
entailments that sentences have then depend on the meanings of the
expressions that go into them. Thus, for instance, we know that (9.1a) is
true (at a species level) because it follows from the hyponymy relation
in (9.1b) and the fact about the superordinate category given in (9.1c).
(9.1) a. Owls lay eggs.
b. An owl is a kind of bird.
c. All birds lay eggs.
In (9.1), we’re assuming that what is meant by owl is the usual sense of
owl: that is to say, the sense to which (9.1b) applies. But we can also use
owl in a different sense, as in (9.2).
121
(9.2) Edward is a wise old owl.

It clearly doesn’t follow from (9.2) that I mean that ‘Edward is a member
of an egg-laying species’.
One way of thinking about the difference between (9.1a) and (9.2) is
that the meaning contributed by owl differs between these sentences. In
(9.1a), it contributes what we might think of as its literal meaning, the
one which introduces all the sense relations that we expect to apply to
the concept of ‘owl’. In (9.2), it contributes a different meaning which
lacks at least some of these usual sense relations, while possibly adding
others (perhaps owl in this sense is a hyponym of person). In Section 1.3,
we introduced the term “explicature” to describe the basic interpreta-
tion of an utterance with details like this spelled out. We can think of a
sentence as having a literal interpretation if all the word meanings that
contribute to the explicature are literal meanings, and having a figura-
tive interpretation if any of the word meanings that contribute to the
explicature are not the literal meanings. These are also described by the
traditional term figures of speech.
Note that we can distinguish between figures of speech and idioms.
As discussed in Section 2.2, idioms like head over heels or kick the bucket
are non-compositional: you cannot work out their meanings from the
meanings of the words they contain. Instead, you have to learn those
meanings. Hence, idioms behave like individual words. Thus, a literal
interpretation of a sentence involving an idiom is one in which that
idiom contributes its regular, idiomatic meaning. Given that kick the
bucket means ‘die’, it is part of its idiomatic meaning that it can only
apply to living things, so there is a sense in which (9.3a) is literal and
(9.3b) is figurative.
(9.3) a. Bill kicked the bucket.
b. Bill’s startup kicked the bucket.
Another way of looking at this is that ‘figures of speech can be inter-
preted according to general cognitive principles, while idioms have to
be learnt’ (Grant and Bauer 2004: 49). A major question for semantic
and pragmatic work on figurative language is: what exactly are these
‘general cognitive principles’ that allow us to interpret figurative uses
correctly? As we shall see, beginning with metaphor in the following
section, different figures of speech illuminate different aspects of this
interpretative process.
figur ative l anguage 123
9.2 Metaphor
Metaphors are traditionally understood as figurative expressions of
the form “X is (a) Y” – figurative because they identify their subject
with a class of entities to which it does not literally belong. (9.2) was an
example of this kind, assuming that Edward is not literally an owl: (9.4)
presents some more celebrated examples.
(9.4) a. The Lord is my shepherd (Psalm 23)
b. All the world’s a stage (Shakespeare, As
You Like It, II, vii,
139)
c. That’s . . . one giant leap for mankind. (Neil Armstrong)
Metaphor has been studied since classical times, and the traditional
Aristotelian view regarded metaphors such as these as implicit com-
parisons (‘all the world is like a stage’, and so on). In recent linguis-
tic research, this idea has competed with the idea put forward by
Glucksberg and Keysar (1990: 3) that metaphors are in fact ‘exactly
what they appear to be: class-inclusion assertions’.
What would it mean for a metaphor like those in (9.4) to be a true
assertion of class inclusion? In essence, the figurative expression within
the metaphor – traditionally called the vehicle – would have to be
interpreted more broadly than is usually the case. The word shepherd
normally means ‘individual who looks after sheep’, but in (9.4a) it would
have to be interpreted as meaning something like ‘individual who looks
after anyone or anything’. Similarly, stage in (9.4b) means ‘a location in
which dramatic events take place’ rather than specifically ‘a location in a
theatre where a play is performed’, and leap in (9.4c) means ‘sudden pro-
gress’ rather than specifically ‘an instance of propelling oneself forward
through the air’.
To put it another way, on this analysis we would be saying that each
metaphor vehicle denotes a (potentially new) superordinate category to
its usual literal meaning. The members of this superordinate category
– for instance, the things that are metaphorical leaps – share some of
the properties with the members of the literal hyponym. The metaphor
itself then identifies the thing it describes – sometimes called the tenor
of the metaphor – as a member of this new superordinate category.
If we think about how concise the metaphors in (9.4) are, compared
with the attempts I made above to sketch out their meanings in literal
terms, it is easy to see why metaphor is so effective in communication.
Metaphors allow us to compress a cluster of attributes into a single
expression, perhaps a single word, and then ascribe those properties to
a referent using a simple copular form (a form of be). This economy of

expression makes metaphor powerful in poetry, and relevant to scien-
tific theorising as well as to scientific communication.
However, compared with literal language, metaphor places a much
greater burden on the hearer, because it falls to the hearer to work out
which of the attributes usually associated with the metaphor vehicle
are actually relevant on this occasion. This applies whether we think
that the metaphor is actually expressing the meaning ‘X is like Y’ or the
meaning ‘X is a member of a broader class of things which I shall call Y’.
Consider (9.5).
(9.5) My manager is a bulldozer.
To understand this metaphor, somehow we have to work out that
the relevant properties of bulldozer are probably ones like ‘forceful’
and ‘heedless of obstacles’, rather than things like ‘pollutes the envi-
ronment’, ‘heavy’, ‘expensive’, or any of the other attributes that we
associate with bulldozers and which might conceivably be associated
with people. Particularly in the case of poetic metaphors, the intended
meaning may be pointed to only very approximately, and the hearer
may struggle to identify the features of the metaphor vehicle that are
likely to be relevant.
If a metaphoric interpretation is called for, we also have to be able
to suppress the literal meaning of the metaphor vehicle. This presents
a particular problem in cases where multiple metaphorical expressions
are brought together, but their literal meanings are incompatible – a
situation traditionally called a “mixed metaphor”. Shakespeare’s to take
arms against a sea of troubles is an example of this kind: (9.6) presents a
couple of even more striking instances.
(9.6) a. The sacred cows have come home to roost.
b. Mr Dewey would have been wielding a double-edged sword
in the shape of a boomerang that would have come home
to plague him. (Denys Parsons, letter in the New Scientist,
25 March 1971, p. 704)
These examples naturally strike us as absurd, but that absurdity arises
because the literal meanings of the metaphor vehicles clash: cows do
not roost, we can’t be plagued by a sword, and a sword can’t be the
shape of a boomerang. But at a figurative level these sentences are per-
fectly coherent: for instance, double-edged sword and boomerang are both
metaphorically used for something that could damage the person who
tries to damage someone else with it. Hence, the awkwardness of mixed
metaphor suggests that we aren’t always successful, at a cognitive level,
in suppressing the unwanted literal meanings that are associated with

metaphor vehicles.
It should be noted that the potential interpretation challenges dis-
cussed above apply to novel metaphors, or metaphors that are not fully
established in the language. Metaphors that are more conventionally
used for particular meanings can be resolved fairly straightforwardly:
(9.5) is arguably of this type. If a metaphor is so conventional that there
is no doubt about its intended meaning, then no additional processing
is required. For instance, when we encounter (9.7), we don’t have to ask
ourselves ‘which of the properties of foot does the speaker mean to asso-
ciate with this part of the hill?’ Foot, in this case, just means ‘lowest part’.
(9.7) The ball rolled down to the foot of the hill.
Expressions of this kind are sometimes referred to as dead metaphors.
We can probably think of them as just adding a new sense to a word.
Hearers will still need to disambiguate this based on the context of
utterance, but there is no particular challenge involved in doing this,
just because the word in question was at some point in the past a meta-
phor vehicle.
9.3 Metonymy
Back in the discussion of privative adjectives such as fake in Section
4.2.2, we encountered an example that is clearly figurative, by the defi-
nition given in this chapter: it is repeated here as (9.8).
(9.8) World’s costliest painting Salvator Mundi is a fake
Leonardo da Vinci, claims documentary. (<https://www
.aninews.in/news/world/europe/worlds-costliest-painting
-salvator-mundi-is-a-fake-leonardo-da-vinci-claims
-documentary20210420115925/>, retrieved 20 April 2022)
Clearly (9.8) doesn’t mean that a documentary is claiming that a paint-
ing is falsely pretending to be a Renaissance polymath. In (9.8), Leonardo
da Vinci means ‘a work by Leonardo da Vinci’. This kind of usage, in
which works are referred to by reference to their creators, is so familiar
as to pass without comment, as in cases such as (9.9).
(9.9) a. I enjoy reading Dickens.
b. There are too few women on philosophy reading lists.
The figure of speech in which we refer to a person or object using
a related referring expression is called metonymy. Apart from the
producer-for-product metonymy discussed above, other familiar kinds
of metonymy include the use of an associated building or location to

refer to a national government, as in (9.10a–b), and the use of kinds of
garment to refer to the people wearing them, as in (9.10c).
(9.10) a. The White House confirmed the new policy.
b. Brussels ratchets up pressure on Budapest over alleged
corruption. (<https://www.ft.com/content/52215b5f-6e44
-4363-b630-30f7516dcdc6>, retrieved 29 April 2022)
c. Will France’s Yellow Vests come back to haunt Macron
on election day? (<https://www.france24.com/en/europe
/20220401-will-france-s-yellow-vests-come-back-to
-haunt-macron-on-election-day>, retrieved 29 April 2022)
In cases like these, we might argue that the metonyms are actually just
novel senses of the words, like the dead metaphors discussed in the
previous section. But these metonymic processes seem to be productive,
in that they can apply to new referents. For instance, if we learn that a
country has a new capital city, we immediately understand that refer-
ence to that capital city might be intended as reference to the country’s
government, even if we have never seen that particular city’s name used
in that way before. And if someone says I enjoy reading Pierre, we can
immediately infer that Pierre is the name of a writer and the speaker is
referring to that writer’s work.
It also seems possible to apply metonyms in context-specific ways:
for instance, in a restaurant, it might make sense for a member of staff to
refer to a customer by their order, as in (9.11).
(9.11) The black coffee wants the bill.
However, compared with metaphors, even these novel metonyms are
relatively straightforward to interpret. The context specifies what kind
of referent we are looking for – whether that’s a book, a government,
an individual, or something else – and we also know that the referent
must be closely associated with the expression that the speaker has
used. Hence, it seems reasonable to suppose that resolving metonymy
is straightforward at a processing level. This makes metonymy a con-
venient and reliable way to refer to things which might otherwise need
lengthy referring expressions (the works of Charles Dickens, the United States
Government, the customer at the corner table, and so on).
9.4 Simile
As an alternative to saying that something is something else, as in the
case of metaphor, we have the option of saying that something is like
something else, as in (9.12).
(9.12) a. The Lord is like my shepherd.
b. All the world’s like a stage.
c. That’s like one giant leap for mankind.
I mentioned in Section 9.2 that the traditional analysis of metaphor is
that it involves implicit comparison: that is to say, the examples given
in (9.4) actually convey the meanings that are made more explicit
in the corresponding examples in (9.12). The term simile is used for
metaphor-like comparisons which are made explicit, typically using
words such as like or as.
Similes are usually regarded as a form of figurative language, and
some definitions of simile require that it involves a comparison between
two essentially dissimilar things. Hence, we could say that (9.13a) is not
a simile, but (9.13b) is.
(9.13) a. A pomelo is like a large grapefruit.
b. Running a university department is like herding cats.
A problem with drawing this distinction is that like is a rather vague
expression: there is no clearly defined threshold of similarity that has
to be exceeded before we can use like truthfully. As, by contrast, seems
to require a high degree of similarity, but (except in archaic uses, such
as (9.14a)) also requires the speaker to specify which attribute is being
talked about (in (9.14b), beauty).
(9.14) a. His eyes are as the eyes of doves by the rivers of waters,
washed with milk, and fitly set. (Song of Solomon 5:12, King
James version)
b. She was beautiful as southern skies the night he met her.
(‘Train in the Distance’, Paul Simon)
In effect, then, we have a couple of different kinds of simile: those
in which we say that the person or thing being described possesses a
specific attribute to the same extent that some comparator does (‘she
was beautiful to the same extent that southern skies are beautiful’,
in (9.14b)), and one in which we say that the person or thing being
described has a similarity to the comparator, but we don’t specify where
that similarity resides.
In either case, it’s not clear that similes really fall within the realm of
figurative language, as defined in this chapter: we do not need to inter-
pret the comparator non-literally, like we do for a metaphor vehicle,
in order for the simile to make sense or to be true. However, what we
do have to do – at least in the like cases – is to infer, based on the utter-
ance, precisely what points of comparison the speaker wants to draw to
our attention. So we still have to solve the same problem that we do for
metaphor – to identify which aspects of the comparator, or vehicle, are
actually relevant for the speaker’s purposes, and to remove from our
interpretation those aspects which are irrelevant. Hence, the proposal
by Stern (2000: 340) that ‘similes should be analyzed on the same model
as metaphors’ might be appropriate even if we think that metaphors are
figurative but similes are not.
9.5 Irony
Consider an utterance such as (9.15), made on a rainy day in Edinburgh
in June.
(9.15) Summer is here.
If a young child hearing this remark was puzzled by it, we might try to
explain it by saying that “they really mean the opposite, that summer
isn’t here yet”. This would be of some help: we would be advising the
child to treat (9.15) as conveying not the literal meaning of the sentence
but rather the negation of it.
Traditionally, irony can be thought of as the figure of speech in which
the speaker intends to convey the opposite of what is said. In practice,
there is clearly a bit more to it than that: the speaker of (9.15) seems to
express rather more, particularly in terms of their attitude towards the
weather, than they would by simply remarking Summer is not here. Two
related challenges for a linguistic account of irony are to explain what
this additional meaning is (and how it arises), and how we are able to
recognise utterances as ironic or potentially ironic in practice.
A crucial observation here is that we do not go around simply making
arbitrary false statements under the guise of irony. An influential idea
about what we are doing instead was articulated by Sperber and Wilson
(1981), who see verbal irony as ‘a type of echoic allusion to an attributed
utterance or thought’ (Wilson 2006: 1724). That is to say, when we utter
something like (9.15), we are acting as though the literal meaning of the
sentence had been expressed or endorsed by someone else. The effect of
this echoic allusion is somehow to call attention to the erroneous nature
of that literal meaning.
We can see this playing out more directly in B’s utterance in

(9.16).
(9.16) A: Why don’t you just ask him not to do it?
B: Yeah, that’s a great idea – that will totally work.
Here, B can be seen as attributing to A the belief that A’s suggestion is a
great idea and will totally work. Somehow, by doing this, B conveys a nega-
tive and mocking attitude towards that belief. Of course, in practice, A
needn’t believe that their suggestion is great – perhaps they just thought
it was worth trying – so B’s attribution of this belief to them is unfair, but
it still serves a useful rhetorical function.
This kind of approach also helps to explain why (9.15) is an appropri-
ate ironic utterance on a miserable summer’s day, but not appropriate
on a miserable winter’s day, even though the ironic meaning would be
equally true in such a case. The crucial point is that the literal meaning
of (9.15) would be a perfectly imaginable thing for someone to say on a
normal summer’s day, and therefore the attribution of it to some imag-
ined non-ironic speaker makes a degree of sense. On a winter’s day, it
would be an odd thing for anyone to say, so we can’t imagine anyone
saying it: the attribution to an imagined speaker doesn’t really work and
the ironic effect is lost. To put it differently, in order to recognise an
utterance as ironic on the echoic allusion account, we have to under-
stand that someone could have produced the same utterance sincerely
under slightly different circumstances, and we can thereby infer that the
speaker is calling attention to the disparity between those circumstances
and the ones that are actually in effect.
There are other approaches to the analysis of irony. Grice (1975)
originally thought of irony in terms of a deliberate violation of the
Quality maxim that triggered reparatory inferences. Clark and Gerrig
(1984) take verbal irony to be a kind of pretence: the speaker is pre-
tending to convey the literal meaning of the sentence, but knows
perfectly well that the hearers will see through this pretence. The
challenge for any such account is to explain exactly why the speaker
ends up conveying the polar opposite meaning to the literal meaning –
rather than just any old alternative meaning – and why this is so often
associated with an attitude of criticism or mockery towards the literal
meaning and anyone who would agree with it.
9.6 Hyperbole
Hyperbole, or dramatic overstatement, represents a further class of
non-literal language use, as illustrated in (9.17).
(9.17) a.
Everybody’s talking about value-based health care.
(<https://www.forbes.com/sites/sachinjain/2022/04/12
/what-is-value-based-healthcare-really/>, retrieved 29
April 2022)
b. Nobody cares about value-based health care.
c. The bathwater was freezing.
In each of these examples, an expression that literally denotes the
extreme point on a scale (everybody, nobody, freezing) is being used figu-
ratively, in the sense that it appears to denote less extreme points that
we would literally express differently (many people, few people, very cold).
Rather like the case of metonymy (Section 9.3), correctly understand-
ing a hyperbolic expression involves identifying the related (and in this
case weaker) meaning that the speaker actually intends to convey. In
principle, this is once again an easier task than understanding metaphor,
because the space of possible meanings that we have to search through
is much more constrained in hyperbole than in metaphor. Compare
(9.17b) with (9.18).
(9.18) OK, Bob cares about value-based health care, but he’s nobody.
Here, nobody functions as a metaphor vehicle rather than a hyperbole:
and even though it is not a very promising vehicle, still a whole range of
possible meanings open up for us to choose among, perhaps including
‘Bob doesn’t really exist’, ‘Bob is unimportant’, ‘Bob is atypical’ and ‘Bob
is ineffectual’. By contrast, when we interpret nobody as hyperbolic, it is
immediately clear that we need to look at the lower reaches of the scale
running from nobody to everybody. Hence, in principle, hyperbole should
be easier to interpret than metaphor, and there is some experimental
evidence in support of that idea (Deamer et al. 2010).
A similar point can be made for dramatic understatement, or meiosis,
as in (9.19a), and the specific form of understatement called litotes,
which involves stating a negative in order to affirm a positive, as in
(9.19b).
(9.19) a. A lottery jackpot win would probably cheer me up a little.
b. Kicking back on the Seine watching the sun set isn’t the
worst way to see out the day. (<https://hauteliving.com
/2021/12/meet-emily-in-paris-new-leading-man-lucien
-laviscount/705406/>, retrieved 30 April 2022)
In these cases, we once again don’t have to search far to find the stronger
meaning that the speaker actually means to convey. As so often in this
chapter, the challenge is to recognise that we need to search for an alter-
native meaning at all.
Summary
This chapter has sketched out some of the major cases of figurative
interpretation, and looked briefly at the processes that may underlie
those interpretations. Semantically, words and sentences have literal
meanings. A literal interpretation of a sentence is one that involves only
literal meanings, whereas a figurative interpretation involves one or
more non-literal meanings. However, in some cases, it can be difficult
to distinguish between non-literal meanings as opposed to literal mean-
ings that have been extended beyond their original domain, as we saw
in the discussion of dead metaphors. Figurative interpretation is some-
times challenging because, as hearers, we must both identify the need to
break away from the literal interpretation, and identify the meaning that
the speaker actually intended to convey. In cases such as metonymy and
hyperbole, this involves searching in a constrained space of possibili-
ties; in cases such as metaphor, the interpretative process is much more
open-ended and we may struggle to recover the speaker’s intended
meaning precisely. Nevertheless, figurative language offers some pow-
erful and economical means of expression, and is correspondingly wide-
spread in everyday communication.
Exercises
1. Talking of a pair of garden birds in early summer: They’ve got two
hungry beaks to feed. What figure of speech is seen in the sentence
in italics? What prior knowledge is needed to understand this
example?
2. Muhammad Ali famously described his own fighting technique with
the words Float like a butterfly, sting like a bee. Using the technical terms
introduced in this chapter, try to identify the figure of speech he was
using.
3. A headline in GQ’s football section in 2018 ran José Mourinho
is a very stable genius. What figurative interpretations might this
evoke?

Cruse (2011) is a thoughtful discussion of the topics in the present
chapter. Lycan (2019) gives an accessible account of metaphor from the
perspective of a philosopher of language. In Huddleston and Pullum
(2002: 651ff., 682) there is an informative discussion of dead metaphors
among English prepositional meanings. Traugott (2000) provides a
short but somewhat technical survey of the roles that metaphor and
metonymy play in meaning changes in the history of languages.
Schumacher (2019) presents a review of recent research on metonymy,
including some experimental work.
10 Utterances in context
Overview
Connected utterances make up a discourse. By this definition, a dis-
course might be a conversation, a TV interview, an email, a letter, or
even a whole book, to the extent that writer and reader keep track of
the connections between the utterances in it. Discourse pragmatics is a
wide field, so this chapter concentrates on just one important aspect of
it, namely how we adapt our utterances to connect them to the current
interests and the existing knowledge of the other discourse participants.
As we will see, English presents us with various ways of packaging
information to achieve this purpose – for instance, placing stress on
certain material within the utterance, or using one or another distinct
syntactic pattern.
10.1 Tailoring utterances to the audience

In Section 8.2, we discussed Grice’s Cooperative Principle, and the
Maxim of Relation, summarised as ‘Make your contribution relevant to
the current goals of the conversation participants’. In Section 8.2.3, we
focused on the implicatures that might be available as a consequence
of a speaker seeming to ignore the Maxim of Relation. But here we’re
going to focus more on speakers adhering to the Maxim of Relation – as
they/we usually do.
It is unusual behaviour for a speaker to enter a room in which people
are having a conversation and, ignoring what those people are currently
saying, just to blurt out something that the speaker wants to tell them.
Admittedly, sometimes this might be a perfectly reasonable thing to
do – for instance, if the speaker wants to tell them that the building is
on fire, or something comparably urgent and important. But more typi-
cally, as suggested by the Maxim of Relation, a speaker is interested in
figuring out the current state of the discourse and then tailoring their
133
contribution to be coherent with this state. This improves the chances of

the speaker providing information that is useful to the hearer.
When we communicate with people, we generally have access to a
large repository of information about those people’s likely knowledge
and interests. If we already know the people we’re talking to, we proba-
bly have a clear impression of the kinds of things that they are or are not
likely to be interested in. But even if we have never met them before,
as the discourse progresses, we increase our stock of information about
their knowledge and preferences, and we expand our shared experience
with them, which now includes experience of the current discourse
itself. And there are rich assumptions that we can make without prior
knowledge of the people involved. All humans share an environment
(comprising the Earth, sun and moon, and so on), and certain traits and
dispositions (such as the capacity for pain, for love, and so on). Within
a given culture, we share additional norms about how to behave, in a
broad sense. We can also generally assume that our interlocutors have
opted into the discourse, which tells us something about their interests,
just as it is reasonable for me to assume that the readers of this book are
interested in meaning (and therefore for me to go on about the topic
at great length, in a way that clearly wouldn’t be appropriate at, say, a
party). If we are to be cooperative, the state of our shared knowledge,
as well as our understanding of our hearers’ current discourse goals,
should influence the kind of information that we attempt to contribute
to the discourse.
As speakers, it is also potentially helpful for us to signal which aspects
of the content that we’re conveying are intended to be new to the hearer,
and which aspects we assume are already established. We have already
seen one example of this, in Section 8.4, where we discussed presuppo-
sitions. In particular, we considered how the use of a presuppositional
expression might be intended to signal that the corresponding meaning
is a background assumption, but might also be used to convey new
information. In this chapter, we look at some of the ways in which we
can package information as established or novel, and correspondingly
how our choice of information packaging may have consequences for
the pragmatic meaning that we are taken to convey.
10.2 Definiteness
In Section 3.1.1, we saw that it is possible to refer to an entity as a house
or the house when it is first mentioned in this discourse: however, on a
second mention, the use of a house to refer to the same house is odd. Our
example of this is repeated below as (10.1).
ut ter ances in conte x t 135
(10.1) a. Some kids walked up to a house. The house was old and
spooky.
b. Some kids walked up to a house. *A house was old and
spooky.
The determiners a and the differ in respect of what we call definiteness,
which is a significant feature of the grammar of English. The use of the,
often called the definite article, signals that the reference is constrained
by prior knowledge: that is to say, the speaker is referring to something
that the hearer knows about. The use of a, the indefinite article, corre-
spondingly suggests that this is not the case. In (10.1a), house is introduced
to the discourse using the indefinite article, and can subsequently be
referred to using definite reference, because the speaker knows that the
hearer is aware of the house as a consequence of what has just been said.
It would not normally be appropriate to refer to a house as the house at
first mention, unless it was entirely clear from the context which house
was being referred to. By contrast, we can use definite reference for the
first mention in a discourse of a referent that we are all familiar with, as
in (10.2).
(10.2) a. Look outside, there’s a weird green glow in the sky.
b. Look at the weird green glow in the sky.
The speaker of (10.2) can use the sky to refer to the single, shared sky
that we are all under. However, in (10.2a), they cannot use the weird green
glow to refer to the phenomenon they wish the hearer to notice: in fact,
the whole point of the utterance is to direct the hearer’s attention to
something they do not know about and that cannot be inferred from the
existence of the sky. The sky is a topic: that is to say, ‘what the utterance
is primarily about’ (Huddleston and Pullum 2002: 236). The topic is not
the new information presented in the sentence, but is an entity that is
easily accessible given our background knowledge at the point of utter-
ance. To say that the sky is a topic is essentially saying that the speaker is
confident that the hearer knows what the sky is going to refer to, without
further pointers. In (10.2a), weird green glow is not a topic, and cannot be
referred to using the. By contrast, in (10.2b), weird green glow is presented
as something that is immediately identifiable to the hearer – ‘you know
which weird green glow I’m talking about’ – which would make sense if
the speaker and hearer are already looking at the same sky to start with.
We also saw examples in Section 3.1.1 that involved bridging infer-
ences, in which the existence of certain referents could be inferred
based on likely has-relations. (10.3) presents an extended version of an
example we discussed there.
(10.3) Some kids walked up to a house. The front door was on the
right. It had a brass bell-push and a big old iron door-knocker.
One of them rang the doorbell.
In (10.3), the is used in conjunction with front door because the exist-
ence of a front door can be inferred from the existence of a house. The
speaker then lists specific features that the door has, using the indefinite
article, putting them in the mind of the hearer and making them poten-
tial topics. Definite reference with the is then appropriate when the
speaker wants to refer to one of these features again, even though they
use a different expression (doorbell as opposed to bell-push). On hearing
the doorbell, the definite article cues us to search through the background
knowledge we have just established, looking for something that could
match with this referring expression.
Taken as a whole, (10.3) shows how we can bring new referents into
a discourse and how we can re-mention established referents. Where
we have indefinite marking, the speaker signals to the hearer something
like ‘this entity is new within the mental file you have opened for this
discourse’ and invites the hearer to consider it as a potential topic in
what follows. Here, this applies to the bell-push and door-knocker, both of
which are introduced with the indefinite article a: it also applies to kids,
who are introduced with the plural indefinite marker some. (Note that
this is not the same thing as the quantifier some, discussed in Section 7.3;
for instance, the indefinite marker some can be pronounced with a much
shorter vowel sound.) Where we have definite marking, the speaker
signals that the referent is familiar, in this example either because its
existence can be inferred (front door) or because it has been already
referred to, albeit by a different expression here (doorbell).
There are numerous other ways of encoding definiteness and indefi-
niteness in English, but the observations made above about their func-
tions are pretty general. The list in Table 10.1 gives some indication of
what is available.
10.3 Given and new material

We can think of the marking of definiteness or indefiniteness as one
device for managing shared background knowledge: specifically, it helps
the hearer to understand whether we are talking about established ref-
erents or introducing new referents. More generally, English provides
us with a variety of strategies that we can use, as speakers, to explain
how the meaning that we are communicating relates to already-shared
background knowledge. We can think of this as the distinction between
Table 10.1 A selection of indefinite and definite forms in English

Indefinite Definite
proper names
Aberdeen, Zoroaster
determiners determiners
a, an, some, another, several, most, no, the, this, that, these, those, its, most,
enough, any their, her, his, your, my, our
absence when head noun is plural
_cities worth visiting,
_famous people
indefinite pronouns personal pronouns
something, someone, somebody, anything, it, they, them, she, her, he, his,
anyone, anybody you, I, me, we, us
given and new information: that is, between what is already assumed to
be known, and what the speaker wishes to present for the first time. The
following subsections discuss some of these strategies.
10.3.1 Pseudo-clefts
In June 2004, a meteorite fell through the roof of a house in New
Zealand and bounced off the sofa. We could describe this truthfully
using any of the sentences in (10.4), which are all paraphrases of each
other.
(10.4) a. What hit the sofa was the meteorite.
b. What the meteorite hit was the sofa.
c. The meteorite hit the sofa.
(10.4c) represents the standard word order for an English declarative
sentence. By contrast, the structures in (a) and (b) are called pseudo-
clefts. Pseudo-clefts have three distinguishing characteristics:
• a clause headed by a “wh-word” (we’ll call it a wh-clause) with an
unspecified argument (what hit the sofa does not specify the subject,
and what the meteorite hit does not specify the object)
• a noun phrase that fills in the missing details for the wh-clause
• the main verb be (appearing as was in these examples).
Although the wh-clauses do not specify their arguments fully, they do
presuppose that these arguments exist. For example, both (10.5a) and
(10.5b) presuppose (10.5d). By contrast, (10.5c) does not.
(10.5) a. What hit the sofa was the meteorite.

b. What hit the sofa wasn’t the meteorite.
c. The meteorite didn’t hit the sofa.
d. Something hit the sofa.
In Section 8.4, we saw how expressions that trigger presuppositions
are usually appropriate to situations in which it is already agreed that
the presuppositions are true. Consequently, (10.5a) is appropriate in
a conversational context in which it is already agreed that something
hit the sofa. We also saw that presupposition triggers can sometimes
be exploited to convey new information – hearers will accommodate
the necessary presupposition – but that doesn’t seem to work well for
pseudo-clefts, possibly because we would usually just use the (10.4c)
form in such a case. Hence, pseudo-clefts are not appropriate in con-
texts in which the idea of anything having hit the sofa is new. For this
reason, the exchanges in (10.6a) and (10.6b) seem acceptable, but (10.6c)
is odd, unless it has already been established earlier that something hit
the sofa.
(10.6) a. A: I heard that something hit the sofa.
B: What hit the sofa was the meteorite.
b. A: Did something hit the sofa?
B: What hit the sofa was the meteorite.
c. A: Did the meteorite hit anything?
B: *What hit the sofa was the meteorite.
In short, then, an appropriate pseudo-cleft is one that matches
the conversational background. It does two things: by means of its
wh-clause it indicates the presupposition, and it presents a noun
phrase as the value of the missing argument (that is, it provides specific
detail in place of the indefinite something in (10.6a) and (10.6b)). The
noun phrase carries the new information provided by the utterer of a
pseudo-cleft.
10.3.2 It-clefts
(10.7) a. It was the childminder who took Judy to the cinema.
b. It was Judy who the childminder took to the cinema.
c. It was the cinema that the childminder took Judy to.
d. The childminder took Judy to the cinema.
It-clefts highlight a noun phrase, often in order to contrast it with
another. Although all the examples in (10.7) are paraphrases of one
another, they differ in where they place emphasis. For instance, (10.7a)
might be used to convey that ‘It was the childminder, rather than
anyone else, who took Judy to the cinema’.
It-clefts have similar distinguishing traits to pseudo-clefts:
• a clause with an unspecified argument (who took Judy to the cinema does
not specify the subject, for instance)
• a noun phrase that specifies the missing argument
• the main verb be (appearing as was in these examples)
• it as the grammatical subject.
As with pseudo-clefts, the clause with the unspecified variable is pre-
supposed. (10.8a) and (10.8b) both presuppose (10.8d), whereas (10.8c)
does not.
b. It wasn’t the childminder who took Judy to the cinema.
c. The childminder didn’t take Judy to the cinema.
d. Someone took Judy to the cinema.
In Section 8.4, we also discussed how some presuppositions are
potentially cancellable in negative sentences. This may be possible
for the presuppositions of it-clefts, as we see in (10.9) – the positive (a)
version of the sentence requires the presupposition, but the negative
(b) version of the sentence can do without it. The presuppositions of
pseudo-clefts are perhaps harder to dispense with in this way, as we see
in (10.9c).
(10.9) a. *The childminder took Judy to the cinema; actually, no one
took Judy there, she went by herself.
b. It wasn’t the childminder who took Judy to the cinema;
actually, no one took Judy there, she went by herself.
c. ?What hit the sofa wasn’t the meteorite; actually, nothing
hit the sofa, it just has a dent in it.
Even so, the circumstances under which we could utter something like
(10.9b) are quite limited, as was also the case for the presupposition can-
cellation examples in Section 8.4. Usually speakers would use it-clefts
only when the presuppositions hold true. The most obvious context for
the utterance of (10.7a) is perhaps in response to a question like Who
took Judy to the cinema?, and the asker of that question is clearly already
assuming that someone took Judy to the cinema, so that point is not
controversial. We will return to this issue in Section 10.4.
10.3.3 Passives
(10.10) a. The conspirators rented the flat in South London.

b. The flat in South London was rented by the conspirators.
Sentence (10.10b) is of a type called passive. Grammarians call the
standard form of the sentence, such as (10.10a), active, when contrasting
them with passives (see Miller 2002: 26). A passive is generally longer
than the corresponding active, because it contains extra grammatical
morphemes – be (occurring as was in (10.10b)), the preposition by, and
for some verbs a past participle form. The other obvious difference
between actives and passives are that the grammatical subject and object
are interchanged between an active sentence and its corresponding
passive. However, such pairs of sentences are nevertheless paraphrases
of one another, which is true of (10.10a) and (10.10b).
Because the passive construction allows this exchange between
subject and object positions, it plays a role in the meshing of new and
given information. There is a tendency – although not an invariable
rule – in English, and perhaps all languages, for utterances to present
given information ahead of new information (Huddleston and Pullum
2002: 1372). Intuitively, this seems like a reasonable way to proceed:
start with the knowledge that you already expect the hearer to have, use
a topic expression to indicate which bit of that knowledge you want to
build on, then present the new information. In (10.10), if the hearer is
expecting to learn more about what the conspirators did, the (a) version is
preferable; if they are expecting to learn more about how the flat in South
London entered into the picture, the (b) version is preferable.
Another tendency in English usage is one that favours using the
subject slot for reference to animate entities, as in (10.10a) (see Biber
et al. 1999: 378). Altogether, then, (10.10b) is quite a striking departure
from the norm: we are using a passive, which is more verbose, and we
are placing an inanimate entity in subject position. It would be quite
reasonable for the hearer of (10.10b) to notice this (if only uncon-
sciously) and consider whether the speaker has chosen this utterance for
some particular communicative reason, as in the discussion of manner
implicatures in Section 8.2.4. A possible explanation might be that the
speaker has made the flat in South London a very obvious topic, because it
is a crucial link to the prior discourse but not easy to recover: perhaps
the previous mention of this referent is some way back, or expressed in
other words (the Streatham hideout, say).
Huddleston and Pullum (2002: 1366) discuss and label a variety of
additional syntactic structures that can be used to change the way in
which English sentences present their information, without actually

changing the semantic content (for instance, The childminder, she took
Judy to the cinema). However, for reasons of space, we will not discuss
these various options and their discourse effects any further here.
10.3.4 Lexical and syntactic converses

The previous subsection focused on one of the functions of passives,
namely putting arguments into subject position, a basic slot for topics.
However, passivisation is not the only way in which this can be accom-
plished. In Section 2.4, we discussed the sense relation of converseness
that holds between some pairs of expressions, such as bigger than and
smaller than. One such pair is like and please, and therefore (10.11a) and
(10.11b) can be regarded as paraphrases of one another.
(10.11) a. The conspirators liked the scheme.
b. The scheme pleased the conspirators.
This also achieves the function of exchanging the arguments in subject
and object position, while preserving semantic equivalence. Just as like
and please exhibit lexical converseness, we can think of active sentences
and their corresponding passives as syntactic converses. The similar-
ity between syntactic and lexical converseness is further exemplified by
(10.12) – the (a) and (b) forms here are paraphrases of one another and
of the sentences in (10.11).
(10.12) a. The scheme was liked by the conspirators.
b. The conspirators were pleased by the scheme.
Not very many pairs of verbs exhibit converseness, though, so in prac-
tice the passive is a more useful and widespread way of switching the
order of arguments.
10.3.5 Focal stress

As well as lexical and syntactic factors, intonation can also play a role in
the way information is packaged. The intonation of spoken English gen-
erally gives extra weight to one syllable within a clause (or, more pre-
cisely, within an organisational unit that often coincides with a clause). A
syllable is a unit of pronunciation, but the kind of stress we are discussing
is associated with syntactic units: by default it tends to occur on the right-
most word of a phrase (Giegerich 1992: 252–4), usually a content word.
Focal stress, also known as focus, is syntactically located intona-
tional prominence that does semantic or pragmatic signalling work.
Earlier in this book, in the discussion of quantifier scope and of presup-

position, there were glimpses of the kind of role that focal stress can
play in semantic and pragmatic disambiguation. The present section
gives a sketch of its use as a signal of new or contrastive information in
pragmatics.
Example (10.13) presents an unremarkable exchange, which shows
how focal stress (indicated here by capital letters) marks new information.
(10.13) A: Did you go to Newcastle by BUS?
B: I went by TRAIN.
A’s use of focal stress indicates that their question is about the means
of transport B used, and that they are presupposing that B went to
Newcastle by some means. In effect, by bus is the new information that
is being introduced for B to confirm or deny. Similarly, B’s response
introduces the new information by train while leaving the presupposi-
tion untouched.
(10.13) is a relatively straightforward example because the focal stress
of A’s utterance unambiguously indicates what is new. Things can be
more complicated than this from a syntactic perspective, as illustrated
in (10.14), based on De Swart and De Hoop (2000: 123).
(10.14) a. Could you [email [her [new BOSS? ]]]
b. No, but I could email her new SECretary.
c. No, but I could email her uniVERsity.
d. No, but I could email MEEna.
e. No, but I could GO there.
The question in (10.14a) has focal stress on the word boss, but it is not
immediately clear from this what precisely the speaker is taking to be
‘new’: is it boss, new boss, her new boss, or email her new boss? Correspondingly,
all the responses indicated in (10.14b–e) are potentially appropriate,
with the indicated stress. To take one example, (10.14d) treats (10.14a)
as presenting her new boss as new information and presupposing ‘the
hearer can email someone’. (10.14d) is appropriate under these circum-
stances because it takes what is presupposed and supplies an argument
that fits as the recipient of the email.
Compare (10.13) with (10.15). In (10.15a), the question’s focal stress
is on you. Three different appropriate replies are given as (10.15b–d).
(10.15) a. A: Did YOU go to Newcastle by bus?
b. B: I went by TRAIN. (Focal stress on both I and train)
c. B: I went by TRAIN. (Focal stress on train alone)
d. B: LORna went by bus.
The speaker and hearer are often automatically treated as part of the
background (because it is difficult to have a conversation without them),
in which case I and you do not carry focal stress. However, in (10.15a)
and (10.15b), these pronouns can be stressed. The effect of placing
stress on you in (10.15a) is to suggest a presupposition ‘someone went to
Newcastle by bus’. The stress on I in (10.15b) does not mark this out as
new content – after all, it refers to the same person as you did in (10.15a)
– but it does seem to contrast the speaker with unnamed other people
who went by bus, thus reinforcing the presupposition of (10.15a). It is
the stress on train that indicates the new content. In (10.15c), the stress
on I is omitted and the response appears to lack the contrastive effect
of (10.15b). (10.15d) also endorses the presupposition ‘someone went to
Newcastle by bus’, and specifies a person who did so: strikingly, it does
not directly answer the question posed in (10.15a), although it does
implicate that the answer is no, as otherwise the speaker should say so,
for reasons discussed in Chapter 8.
10.4 The Question Under Discussion

When we discussed example (10.14), we saw how the same utterance
could be interpreted as asking various different questions, depending on
what was assumed already to be given and what was taken to be new.
One potentially helpful approach to trying to capture precisely what is
going on at each move in a conversation, and thus what is being com-
municated by the discourse participants, is the idea of Question Under
Discussion (Roberts 2012). The Question Under Discussion (QUD)
is defined as the immediate topic of discussion. This can literally be
the question that was asked, but not all information is offered up in
conversation just as the response to a question: so we can construe the
QUD more broadly as the question that an utterance can be taken to be
answering.
On Roberts’s account, a QUD “proffers” a set of relevant alternatives,
and an appropriate response is one that selects from among this set of
alternatives. For instance, a question like (10.16a) proffers alternatives
such as yes and no, whereas a question like (10.16b) proffers lists of names
of individuals as its alternatives.
(10.16) a. Did Mary come to the party?
b. Who came to the party?
Within the QUD approach, what we understand to be communicated
by a given utterance depends on what the QUD is. This is not always
obvious, because the question is not always made explicit: sometimes
we have to try to infer what the QUD is, based on the properties of the
utterance itself (and on information we might have about the discourse
context). The reason that the idea of QUD connects to the other con-
cepts discussed in this chapter is that these various techniques for pack-
aging information can be interpreted as signals that the speaker gives to
the hearer about what the QUD is.
Consider, for instance, the case of focal stress. As discussed in the
previous section, we could consider the focal stress in (10.17a) to be
emphasising that train is new information, and that in (10.17b) to be
emphasising that John is new information.
(10.17) a. John went to Newcastle by TRAIN.
b. JOHN went to Newcastle by train.
However, it’s perhaps more straightforward to think of the difference
simply in terms of QUD: the stress pattern of (10.17a) suggests that the
QUD is “How did John go to Newcastle?”, while the stress pattern of
(10.17b) suggests that the QUD is “Who went to Newcastle by train?”
By appealing to the QUD, we’re able to give an account of how some
of the pragmatic enrichments arise that are associated with uttering
these sentences. For instance, (10.17b) might implicate that the speaker
did not go to Newcastle by train, because if the QUD is “Who went to
Newcastle by train?” and the speaker did so, the speaker should say so.
But (10.17a) does not convey this implicature, because if the QUD is
“How did John go to Newcastle?”, the issue of whether the speaker also
went there is not immediately relevant.
The idea of QUD is clearly closely related to the notions of given and
new information, in that only new information can be the answer to a
QUD – in normal conversation we don’t tend to ask questions to which
we already know the answers. So we can think of “newness”, signalled
by focal stress – or by any other means – as a possible hint that we can
use to figure out what the QUD must have been, in cases where we don’t
already know this. This in turn could lead us to calculate specific impli-
catures that depend heavily upon the precise question that was asked.
This is striking in the case of it-clefts and pseudo-clefts, for instance
(10.7), repeated below.
b. It was Judy who the childminder took to the cinema.
c. It was the cinema that the childminder took Judy to.
d. The childminder took Judy to the cinema.
As remarked in passing earlier, (10.7a) is naturally construed as an
answer to the question “Who took Judy to the cinema?” If that is the
QUD, we might infer that (10.7a) is an exhaustive answer: that is to say,

no one else took Judy to the cinema. This is merely an implicature, and
cancellable, but it is nevertheless a distinct meaning that doesn’t arise
from the other sentences. (10.7b) is naturally construed as an answer to
“Who did the childminder take to the cinema?”, and as such may impli-
cate that the childminder didn’t take anyone else to the cinema, but it
doesn’t implicate that no one else took Judy there. A similar argument
goes for (10.7c). As noted earlier, this helps us to explain some of our
intuitions about how it-clefts and pseudo-clefts are appropriately used
in conversation.
In a similar spirit, the QUD-based approach might give us a slightly
more rigorous account of examples like (10.10), repeated below.
(10.10) a. The conspirators rented the flat in South London.
b. The flat in South London was rented by the conspirators.
Here, both the conspirators and the flat in South London are marked as defi-
nite. That being the case, neither the scheme nor the conspirators can
be considered “new” here, so we can’t readily explain the point of pas-
sivising (10.10a) by appeal to the given/new distinction. I noted earlier
that the (a) version is preferable if the hearer is expecting to learn more
about what the conspirators did, and the (b) version is preferable if they
are expecting to learn more about how the flat in South London entered
into the picture. From a QUD standpoint, we could say that (10.10a)
suggests a QUD to the effect of “What did the conspirators do?”, while
(10.10b) suggests a QUD to the effect of “What happened to the flat
in South London?” That is to say, we can potentially understand the
motivation for producing passive sentences in terms of how they answer
particular QUDs.
In summary, then, the idea of QUD appears to be a potentially useful
way of bridging between the grammatical marking of discourse features
and the pragmatic consequences of doing so. It is also an intuitively
appealing idea. What is not clear at this stage, however, is whether
we actually have a mental representation of the notion of QUD – and
if so, how we go about constructing it, based on the bits and pieces of
information that the discourse provides us with. That is to say, it is not
yet clear whether QUD has psychological reality in addition to being a
helpful expository device.
Summary
This chapter has presented an introductory survey of some of the struc-
tures and devices in English that indicate what the speaker takes to be
already known to the hearer when making an utterance, and what the
speaker takes to be new. We have discussed definiteness, pseudo-clefts
and it-clefts, passives and focus marking, but this does not by any means
constitute an exhaustive list. We have also considered the notion of
Question Under Discussion (QUD), which is one way in which we can
try to capture the diverse effects of some of these grammatical manipu-
lations within a unified analysis, and derive predictions about the inter-
pretation of these sentences. However, much more could be said on all
of these topics, in addition to the various alternative approaches that are
available, beyond the given/new distinction, for talking about informa-
tion structure in utterances.
Exercises
1. Why is there no need for a preparatory introduction of topic before
giving the warnings Keep your head down and Mind the step, where the
underlined phrases are definite?
2. Pseudo-clefts can be inverted: for example, The meteorite was what hit
the sofa as opposed to What hit the sofa was the meteorite. Is the presup-
position the same or different between these two examples?
3. Tom says that “It was the ATlas that Lucy borrowed”. You know
that Tom is wrong: (a) Mary borrowed the atlas, and (b) Lucy bor-
rowed the dictionary. Indicate how to correct Tom by filling in the
following to make a complete sentence:
“No, you’re wrong: _______________________________”.
Which of the scenarios, (a) or (b), does your completion relate to?
How does this fit with the presupposition pattern of it-clefts dis-
cussed in the chapter?
4. Earlier in the book we discussed how presuppositions can be can-
celled: under certain circumstances, a speaker can say Jessica didn’t
regret arguing with her boss without being committed to the presup-
position that ‘Jessica argued with her boss’. What kind of Question
Under Discussion might license the use of a presuppositional expres-
sion like this ( . . . didn’t regret . . .), even though the presupposition will
later be cancelled?

A comprehensive and readable general account of information packag-
ing is given in chapter 16 of Huddleston and Pullum (2002). De Swart
and de Hoop (2000) provide an excellent survey of relevant theories,

although parts of this are quite hard, and the same goes for Rooth’s
(1996) work on focal stress. The idea of Question Under Discussion is
set out in full in Roberts (2012).
11 Doing things with words
Overview
What is the point of using language? That may seem like a strange ques-
tion to ask, particularly in the last chapter of a book about language. But
most of the preceding chapters have focused specifically on utterances
that make statements of fact. Passing on information is certainly an
important function of language, but it is not the only one. Much of what
we do with language is not just about making statements of fact. We
exchange greetings, we ask questions, we issue requests, we complain,
we apologise, we warn, we thank, and much else besides.
J. L. Austin was one of the first people to focus attention on this ques-
tion. His 1955 lecture series, published posthumously in 1962 under
the title How to Do Things with Words, made the point that many of the
things we utter cannot really be said to be ‘true’ or ‘false’, but they are
nevertheless meaningful. To capture that meaning, we need an analysis
that goes beyond notions of truth and falsity, and engages with the way
in which utterances are used to perform social actions.
In this chapter, we will begin to consider what kinds of social action
can be performed through the medium of language, and how speakers
and hearers manage to perform and recognise these actions.
11.1 Speech acts

Here we will follow Searle (1975, 1979) in using the term speech acts to
talk about the social actions that we can accomplish through the use of
language. (11.1) lists a small sample of speech acts, along with examples
of utterances that could be used to bring them about.
(11.1) a. statement: “I lived in Edinburgh for five years.”

b. order: “Pay this bill immediately.”
c. question: “Where are you from?”
148
doing things with words 149
d. greeting: “Hello.”
e. invitation: “Help yourself.”
f. felicitation: “Happy New Year!”
g. apology: “I’m terribly sorry.”
It is difficult to produce an exhaustive list of possible speech acts: Austin

(1962) suggested that there could be hundreds of distinct items on such
a list. Austin zoomed in particularly on a large category of items that he
called performative, which are units of language that appear to perform
social acts just by being uttered (under the appropriate circumstances).
(11.2) gives some examples of performatives.
(11.2) a. I name this ship the Queen Elizabeth.

b. I now pronounce you husband and wife.
c. I apologise for my lateness.
By uttering these words, the speaker of (11.2a) causes the ship to be

called the Queen Elizabeth, when this was not previously the case; the
speaker of (11.2b) causes the hearers to become married to one another;
and the speaker of (11.2c) apologises for their lateness. Clearly this
doesn’t happen because the words are imbued with some magical
powers; rather, it happens because of agreed social conventions. In the
case of (11.2a) and (11.2b), it’s also part of this social convention that
the words only take effect if the appropriate person is uttering them.
Thus, if an imposter takes the place of the priest or registrar and utters
(11.2b), the utterance fails to bring about the act of marrying the couple.
By contrast, anyone could utter (11.2c) and could normally be said to
have apologised by virtue of doing so. The conditions that have to be
satisfied for a speech act to succeed in its effect are often referred to as
the felicity conditions for that act, and we describe the act as felicitous
if the conditions are met and infelicitous if they are not.
One possible way of recognising performative utterances is that the
word hereby can be inserted into them, reflecting the fact that the social
action is brought about by uttering the words. For instance, (11.3a)
and (11.3b) are acceptable, whereas (11.3c) and (11.3d) are not: (11.3c)
doesn’t perform the action it names, and (11.3d) reports on the speaker’s
state of mind, which does not depend on these words having been
uttered.
(11.3) a. I hereby name this ship the Queen Elizabeth.

b. I hereby apologise for my lateness.
c. *I hereby sing.
d. *I hereby doubt that Tom will go to the party.
By this definition, there are many performative or potentially per-

formative verbs in English, including apologise, complain, protest, resign,
object, declare, open, close, vote, propose, and so on. We could think of these as
each corresponding to distinct social actions; alternatively, we could try
to group them into more general classes. Various classification schemes
along these lines have been suggested, Searle’s (1979) proposal being
particularly influential, but for reasons of space I won’t delve into the
details here.
As hearers, we might also be interested in trying to distinguish
between different varieties of a single kind of speech act, particularly if
these different varieties require us to respond to them in different ways.
For instance, we can distinguish several different kinds of question, as
illustrated in (11.4).
(11.4) a. What’s your name?
b. Is today Tuesday?
c. Today’s Tuesday, isn’t it?
All of these are potentially genuine, information-seeking questions: the
speaker is performing the social action of trying to obtain information
from the hearer. (We might want to think of so-called “rhetorical ques-
tions” as not really being “questions” in speech act terms, because they
fulfil a different social function than seeking information.) But there are
important differences between these three examples in terms of how
they constrain the hearer. (11.4a) is a wh-question, using the question
word what, and admits arbitrarily many possible answers. (11.4b) is
essentially a yes/no question, also called a polar question, and admits
only the answers yes or no (although the hearer could opt out of answer-
ing by saying I don’t know). (11.4c) is also a polar question but it has a pre-
ferred response, namely yes. This kind of question is sometimes called
a tag question, because it tags a question onto the end of a statement:
it is used to seek confirmation of something that the speaker already
believes to be the case. So, although the three utterances in (11.4) are
all questions, it’s not obvious that they really are performing the same
social action, at least not if we consider the effect on the hearer: they
place quite different requirements on the hearer, in terms of what con-
stitutes an acceptable or appropriate response.
In short, we can think about speech acts at a fine-grained level, or we
can group them into a smaller number of broad categories. For some
purposes, the choices we make about how to think about speech acts
may have practical consequences, for instance if we’re building and
training artificial dialogue systems. Clearly the question of how many
different speech acts there are will be relevant to the question of how
our system recognises and distinguishes them, a point we’ll touch on

later in this chapter.
It is also sometimes useful to be able to distinguish between different
levels at which we can describe an utterance, in terms of its social effects.
Austin (1962) noted that we could distinguish between what he called
locutionary acts, illocutionary acts and perlocutionary acts. Broadly
speaking, the locutionary act is the act of speaking, the illocutionary act
is what was meant by it, and the perlocutionary act is the effect that it
had. Suppose that you are at a dinner and another diner, Lila, says Could
you pass the salt? You respond by passing Lila the salt. We could describe
Lila’s action in different ways, as illustrated in (11.5).
(11.5) a. Lila said Could you pass the salt?
b. Lila asked you to pass the salt.
c. Lila got you to pass the salt.
Here, (11.5a) is the description of what Lila did as a locutionary act:
it identifies the linguistic action that she performed. (11.5b) is the
description of what Lila did as an illocutionary act: it identifies her as
performing the social action of requesting something of you. (11.5c) is
the description of what Lila did as a perlocutionary act: it describes the
consequences of the illocutionary act that Lila performed.
When we talk about speech acts in linguistics, we’re usually talking
specifically about illocutionary acts. The locutionary level doesn’t
really involve social actions, and so it is not particularly interesting to
us here – Austin (1962: 94–5) talks about identifying the locutionary act
just to distinguish it from the other types of act. The perlocutionary act
is much more interesting at a social level, but most of what happens at
that level falls outside the domain of linguistics. For instance, performa-
tive utterances can bring about illocutionary effects, but they cannot be
guaranteed to bring about perlocutionary effects, such as persuading
someone of something. Thus, (11.6a) is felicitous, but (11.6b) is not.
And the question of how we achieve effects such as persuasion, based
on social actions such as warning, is not really a linguistic issue: it goes
much deeper into the psychology of decision-making.
(11.6) a. I (hereby) warn you that it is snowing heavily.
b. *I (hereby) persuade you not to go out.
11.2 Indicators of speech acts

In order for a speech act to work, in bringing about its intended social
action, the hearer has to be able to recognise what the speaker is
attempting to do with it. In some cases this is fairly straightforward – for

instance, for the performative examples discussed earlier – but in other
cases, the challenge of speech act recognition is in principle a difficult
one. The following subsections discuss some of the major kinds of cue
to speech act recognition that a hearer might exploit in order to under-
stand what a speaker is trying to achieve.
11.2.1 Syntactic cues and indirect speech acts

One major category of cues to speech act type is syntactic. The speech
acts in (11.1a–c) were placed at the head of the list because they corre-
spond with the three major sentence types of English (and many other
languages). These sentence types are declarative, commonly associ-
ated with statements; interrogative, commonly associated with ques-
tions; and imperative, commonly associated with orders or requests. In
English these sentence types are clearly distinguished syntactically, as
the example sentences for (11.1a–c) illustrate.
Even so, we can’t always identify speech acts just on the basis of the
sentence type. There isn’t a one-to-one matching between sentence
type and speech act type: as we’ve already seen, there are many kinds
of speech act, but most of these do not have their own special sentence
type. Consider (11.7).
(11.7) a. I promise you that I will come back later.
b. I warn you that I will come back later.
c. I advise you that I will come back later.
d. I will come back later.
We can interpret (11.7a) as a promise, (11.7b) as a warning and (11.7c) as
(at least ostensibly) advice – three different speech acts – even though
they use the same syntactic frame, differing only in one choice of verb.
And (11.7d), although it uses a simple declarative sentence type, might
be used to achieve any of these effects, depending on the context. So we
see that declarative sentences are associated with a wide range of pos-
sible speech acts.
In fact, it’s not just declaratives that have this flexibility: interrogative
sentences aren’t always uttered to ask questions, and imperative sen-
tences aren’t always uttered to give orders or make requests. In (11.8),
we see an example of how the same effect (in this case a request) can be
brought about using any of these sentence types.
(11.8) a. I’d be grateful if you would close the window.
b. Could you close the window?
c. Please close the window.
Formally, (11.8a) is a declarative sentence: it expresses an aspect of the

speaker’s state of mind. (11.8b) is an interrogative, and superficially
appears to be asking about an aspect of the hearer’s capability. But in
practice all three utterances can readily be used to perform the same
speech act, in this case a request, even though this is only directly
expressed by (11.8c), which uses the imperative sentence type.
In a similar vein, (11.9) shows how each of these three sentence types
can be used to perform a speech act that isn’t directly associated with
any of them, specifically an offer.
(11.9) a. You’re welcome to have some cake.
b. Can I get you some cake?
c. Help yourself to some cake.
When a sentence type is used to perform a speech act that is not cus-
tomarily associated with that sentence type, we have what is sometimes
called an indirect speech act (see Searle 1975). Indirect speech acts
are by no means unusual in language use: Levinson (1983) argues that
the majority of speech acts are in fact indirect. In (11.8) and (11.9), there
is only one direct speech act, namely (11.8c) – and, compared with the
other examples, (11.8c) might strike us as potentially impolite and only
suitable for use in a limited set of circumstances.
Several accounts have been proposed as to how we might get from
the “literal” understanding of a speech act to the indirect understand-
ing, when it is appropriate to do so. Consider (11.8b), which is literally a
question concerning the hearer’s ability to close the window, but which
we actually understand as a request to close the window. Gordon and
Lakoff (1971) propose that, in such a case, the speaker already knows
that the answer to their question would be yes, and the hearer knows this
too: consequently, the hearer can infer that the purpose of the utterance
is not simply to ask a question, and has to arrive at a different interpre-
tation. Searle (1975) offers a slightly more detailed account: he argues
that asking the question (11.8b) signals that the answer (known to be yes)
is relevant to the speaker. The hearer can infer that the reason it is rel-
evant whether the hearer can close the window is that the speaker wants
the hearer to do so. For that reason, the question can be reinterpreted as
an indirect request.
We can construct similar arguments for various other cases of indirect
speech act interpretation, which involve the hearer realising that the
literal interpretation would be absurd and then performing some kind of
reparatory inference to figure out what the speaker is actually trying to
achieve. However, since the 1970s, the consensus view in the literature
has moved away from the idea that this is what actually happens, and
towards an alternative view in which we are not obliged to consider the

literal interpretation first. On this view, while syntactic cues are relevant
to how we recognise speech acts, they are not so privileged that we nec-
essarily take them as the starting point for our reasoning.
11.2.2 Lexical cues

For performative utterances, we have already seen how the presence
of a particular lexical item indicates that the utterance is a particular
kind of speech act. This applies if we are talking about utterances in
the first person and with present time reference: if we are talking about
other people or other times, the presence of a performative verb does
not make the utterance performative. Hence, we can insert hereby in
(11.10a), but not in (11.10b–d); (11.10a) is itself an act of naming but
(11.10b–d) merely report on acts of naming.
(11.10) a. I name this ship the Queen Elizabeth.
b. King Charles names this ship the Queen Elizabeth.
c. I will name this ship the Queen Elizabeth.
d. I named this ship the Queen Elizabeth.
Many other lexical items are potentially useful cues to the hearer
recognising the kind of speech act that is being performed. Consider
(11.11).
(11.11) a. Could you remind the class that the office hour is
cancelled?
b. Why not remind the class that the office hour is cancelled?
As we discussed with respect to (11.5), beginning an utterance could you
. . . superficially suggests that the utterance will ask about the hearer’s
capability to do something. However, our experience suggests that could
you . . . is actually an idiomatic way of making a polite request. In a similar
way, why not . . . appears to ask for an explanation of why a particular
course of action is not being undertaken, but is actually an idiomatic way
of making a suggestion. Intuitively, we know this before we’ve encoun-
tered the rest of the utterance, so it appears that we do not need to go
through the kind of reasoning process discussed in Section 11.2.1. Rather,
we can just rely on our past experience of utterance use to infer that one
beginning could you is somewhat likely to be a request, and one beginning
why not is somewhat likely to be a suggestion.
Some lexical markers of speech acts serve as even more emphatic
cues than this. We might think of sorry as a marker of apology: of course,
it’s possible to use sorry in utterances that are not apologies, but hearing
sorry might well incline us towards the belief that something is an

apology. Hello is a very clear-cut marker of greeting, and its other uses
in conversation are rather limited.
Even smaller pieces of lexical information might offer us some help
in identifying the speech act being performed. For instance, it might
make a difference whether the utterance includes first-person pronouns
(I, we) or second-person pronouns (you): that might indicate whether the
speech act is likely to be one that obligatorily involves the hearer (such
as requesting or advising) or one that does not (such as stating or prom-
ising). This could be another possible hint that helps the hearer towards
the correct interpretation.
11.2.3 Cues based on conversation structure

Another kind of background knowledge that the hearer can use to
recognise a speech act is knowledge about the way conversations take
place, and the kinds of social actions that are likely to be performed
at various stages of those conversations. Perhaps the most obvious
example is that conversations often start with greetings, so we might
have a reasonable expectation that the first utterance in a conversation
is likely to be a greeting – even if it looks like something else, such as
a question: How do you do? Conversations often end with leave-takings,
so we might well conjecture that the final utterance of a conversation is
of that kind. We can even have utterances that serve as both greetings
and leave-takings, such as Ciao! In theory, this might be confusing, but
in practice it is usually easy to tell, just based on the situation, whether
a speaker is likely to be performing one act or the other.
As language users, we also know that there are rich interdependen-
cies between social actions. For instance, questions and answers are
closely related. There’s a logical dependency – something can’t be an
answer unless it corresponds to a question – but also a strong expecta-
tion that questions will be almost immediately followed by answers, if
the speakers are cooperative. As discussed in Chapter 8, we naturally try
to interpret something as an answer if it is said in response to a question
– (11.12) presents another example of this.
(11.12) A: Did Bob turn up on time for the meeting?
B: Has hell frozen over?
Here, B responds to A’s question by asking another question, which
could be answered no. We infer from this that the answer to A’s question
is no (and we also get some hint of the strength of B’s feeling about this).
This interpretation doesn’t make any obvious logical sense – the answer
to B’s question is in principle completely independent of the answer

to A’s question – but seems readily to be available because of our clear
expectation that B’s utterance is going to constitute some kind of answer
to A’s question.
There are also strong dependencies between other kinds of speech
acts: for instance, we expect invitations to be followed by utterances
that either accept or decline those invitations. In (11.13), we can readily
understand B to be declining A’s invitation.
(11.13) A: Would you like to go for a coffee with me?
B: I have to finish this assignment.
Strikingly, in this case, B could achieve a similar effect just by a short
silence: even a brief hesitation could be understood as communicatively
meaningful. In all these cases, we’re using our expectations about how
the conversation should proceed, given the turn that has just taken
place, in order to fill in the gaps left by what the speakers have actually
said.
This idea was developed by Schegloff and Sacks (1973). They used
the term adjacency pair to describe the relationship between two
conversational turns that are highly likely to happen sequentially,
such as question–answer. When we encounter the first part of an
adjacency pair being performed as a speech act, we can quite strongly
expect that the next speech act will be the second part of the adjacency
pair. Interestingly, Clark (2004) argues that this kind of relation can
also exist between wordless communicative acts such as gestures,
raising the issue of whether we can have things like question–answer
sequences playing out in purely gestural communication. However, we
won’t delve further into that possibility here (see Clark 2012 for more
detailed discussion).
As well as knowing about adjacency pairs, we also all know about the
kinds of actions that can reasonably be performed by a given speaker at
a given point in an interaction. It seems likely that we use this informa-
tion to constrain our guesses as to what the speaker is trying to do. For
instance, if someone utters their own name, they could be introducing
themselves, but this is only a possible interpretation if there are people
present who might not know them. If a customer goes into a restaurant
and utters the words A table for two, that’s recognisable as a request, but
if the head waiter utters those words to the customer, it is probably
something more like an offer. These inferences don’t just rely on our
linguistic knowledge – they also depend on our understanding of how
particular kinds of interaction work, and the social acts that different
discourse participants are likely to be trying to achieve.
11.2.4 Integrating the information

In the preceding three subsections, we briefly looked at three kinds
of cue that might provide a hearer with information about the kind of
speech act that a speaker is performing: the syntactic structure of the
utterance, the meanings of the words it contains, and shared knowledge
about the way speech acts are put together in conversations. This is not
an exhaustive list of possible kinds of cue: prosody is also relevant, and
so is gesture. But a question that naturally arises is how a hearer puts
together all these potential sources of information and uses them to
arrive at an understanding of what the speaker is trying to achieve.
As discussed in Section 11.2.1, the traditional viewpoint on this was
that we initially read the speech act directly off the syntax, and if the
result is anomalous – for instance, because it conflicts with other things
we know about what is going on in the conversation – we perform addi-
tional pragmatic inferences in order to correct this understanding. But
this turns out not to be an entirely convincing idea, for several reasons.
For one thing, it suggests that we have to perform a great deal of prag-
matic reasoning to get at the correct meaning of utterances which seem
to have very obvious interpretations, such as Could you pass the salt? It
would also mean that we initially ignore very strong non-syntactic cues,
such as those based on the current status of the conversation or the
typical use of particular words: that is to say, we would have to inter-
pret a conversation-starting How do you do? or How’s it going? literally as
a question before subsequently reinterpreting it as a greeting. The fact
that we can respond to indirect speech acts very rapidly and without
obvious effort suggests that we may be using shortcuts to the correct
interpretation when we can.
What kind of shortcuts might we be taking? One possibility is that we
adopt a top-down strategy: when we participate in a conversation, we
try to recognise the immediate goals of the speaker based on an under-
standing of what their more general goals are likely to be. Going back to
the restaurant example, if we go in and the head waiter says to us A table
for two, we can recognise this as a way of realising part of the plan that
the head waiter has for the overall interaction: that is to say, that they
will seat the customers, serve them food and drink, receive payment and
say goodbye. Specifically, we recognise it as part of the plan to seat the
customers, and we’re happy to collaborate in achieving that goal.
This idea underlies the plan-based approach to speech act recogni-
tion, set out by Perrault and Allen (1980). From this perspective, the
speaker is seen as a rational agent trying to bring about certain changes
in the world – which sounds rather grandiose, but in practice could be
something as simple as causing a window to be closed which is cur-

rently open. The speaker constructs a plan to achieve these changes,
which may involve performing speech acts. The hearer can draw infer-
ences about what the speaker is doing at a given moment, and use these
inferences to better understand what the speaker is ultimately trying
to achieve – but the hearer can also use their understanding of the
speaker’s likely goals in order to draw conclusions about what they are
trying to do in the moment.
An alternative possibility is that we take more of a bottom-up
approach, simultaneously considering all the evidence that could bear
upon the question of what speech act is being performed, and weigh-
ing it probabilistically. This point of view, referred to as the cue-based
approach, has come to dominate computational approaches to speech
act recognition – notably, for artificial dialogue systems – over the
past forty years or so. From this perspective, while syntax is helpful in
speech act recognition, it doesn’t have primacy over other sources of
information. Hence, thinking in terms of direct versus indirect speech
acts is something of an irrelevance: there is nothing special about direct
speech acts; they just happen to be the ones for which the syntactic cue
ends up agreeing with the consensus.
In the end, which kind of explanation is better may depend on exactly
what we want to achieve – are we interested in building an effective
artificial dialogue system, or do we want a rational reconstruction of
how we can recognise speech acts in principle? Or are we ultimately
interested in knowing about the mental processes that underpin speech
act recognition in humans? Trying to address any of these questions in
detail would take us some way beyond the scope of this book. Even so,
the debates around this topic can be illuminating when we consider how
to describe the speech acts of English. We have evidence that speech
acts are signalled by a range of linguistic devices, including syntax, but
also that syntax does not play quite such a central role as originally
assumed by Searle and colleagues in the 1970s. We also have evidence
that the recognition of speech acts depends on context, but this includes
not only the linguistic context but the cultural context, and brings in
encyclopedic knowledge about how the world works. In fact, if we take
the treatment of speech acts to be a part of linguistic semantics and
pragmatics, it becomes increasingly difficult to draw a clear boundary
between linguistic knowledge and world knowledge.
Summary
We can achieve a wide range of social actions specifically through the
use of words, and this is one reason why linguistic communication is
so crucial to our social interaction. In most cases, the intended social
purpose of the utterance – its illocutionary force – is not explicitly
marked and must be inferred. Syntactic features of the sentence, such as
the sentence type, are one helpful indicator of speech act type, but many
speech acts are “indirect” in the sense that their purpose mismatches
with the sentence type. Accounts differ as to precisely how we do it,
but it seems clear that we typically draw upon many distinct linguistic
and non-linguistic factors when we perform the task of recognising the
purpose of an utterance – a task we perform almost every time we use
language.
Exercises
1. Using the notions of speech acts and presupposition, give a brief
description of the wording of this notice seen in a bus: “Thank you
for not smoking. MAXIMUM FINE £100.” (In the same frame there
was a picture of a cigarette with a slash through it, inside a red circle.)
2. For each of the following sentences, name the kind of direct speech
act that would “normally” be associated with it, given the sentence
type, and say what indirect speech act a speaker would probably be
performing by uttering that sentence in practice.
(a) Can’t you stop talking?

(b) Help yourself to milk and sugar.
(c) Have you heard, our team’s leading 18 to 15?
(d) You have my sympathy.
(e) Don’t imagine that entailment and implicature are the same
thing.
(f) Accept my profound condolences.
(g) Have I ever let you down?
(h) I recommend that you keep a copy of the letter.
3. What kind of function does the utterance I should let you go tend to
serve in a conversation? How can we explain this effect?
4. It is claimed that, when Queen Elizabeth II opened the Inner Ring
Road in Birmingham in 1971, she was expected to name a tunnel “the
Queensway”. However, she misspoke and said, “I name this road the
Queensway.” How can we analyse the effect of this utterance?

Lycan (2019) and Saeed (2015) offer good introductory treatments of
speech acts. Searle (1975) is worth reading as a more detailed treatment
of how indirect speech acts can be calculated from a literal starting
point. Chapter 5 of Levinson (1983) is also a good guide to the problem
of speech act identification, and the limitations of Searle’s approach.
Jurafsky (2004) gives a lucid account of the competing approaches to the
recognition of speech acts from a computational perspective.
Suggested answers to the
exercises
Chapter 1
1. Arriving denotes a change from not being in/at a place to being in/at
it. Leaving denotes the reverse transition: from being in/at a place to
not being in/at it. In parallel to this, learning can denote a transition
from a state of not knowing to knowing something, and forgetting is the
reverse transition from knowing to not knowing. Establishing this par-
allel involves considering word meanings without regard to context,
and is therefore part of semantics.
2. Although we cannot be certain, it’s reasonable to infer that the grade
was probably low. By not making a more strongly positive state-
ment such as “You did very well”, the tutor can naturally be seen to
suggest that the student’s grade was near the threshold for failure:
that is to say, the most positive thing that could be said about it was
that it wasn’t a fail. However, the tutor’s utterance is still compatible
with the student having done well: we could, for instance, imagine a
situation in which the tutor is only allowed to tell students whether
or not they failed, until the official results are released. Or it could
be that the tutor thinks that the student is especially worried about
the possibility of failure, and wants above all to dispel that worry.
Because we are considering alternative utterances that could have
been used, this is pragmatic reasoning (which usually gives rise to
uncertain conclusions). We’ll discuss examples like this in more
detail in Chapter 8.
3. Pick the right lock can mean ‘choose the correct lock’ or ‘without a
key, open the correct lock’. At least two different propositions are
involved, one for each possible meaning that the sentence can have.
We could arrive at still more distinct meanings – and more distinct
propositions – by also considering the ambiguity of the words right
(‘correct’ vs ‘located to the right-hand side’) and lock (‘security
device’, ‘installation on a canal’, ‘piece of hair’, etc.). For instance,
161
Pick the right lock might also mean ‘choose the canal lock on the right-
hand side’.
4. The kangaroo example, like the window example discussed in Section
1.2.1, is a classic example of the limits of ostension. Here, one person
points to an object, and the other person produces an utterance: but
without a common language to mediate this interaction, we can’t be
at all certain that the utterance has anything to do with the name of
the thing being pointed at. The speaker could actually be saying ‘I
don’t understand you’, or naming the colour of the animal, or a part
of it – or indeed giving the specific name by which that particular
individual is known to them, rather than the general name for the
species of animals to which it belongs. Ultimately, to resolve this
with any confidence, we would need a much fuller understanding of
the language that was being spoken, which would enable us to ask
about the meaning of the expression directly.
5. The sign clearly isn’t supposed to mean that carrying a dog is a
precondition for using the escalator. Nor does it impose an obliga-
tion on people to take care of dogs for which they wouldn’t usually
be responsible, while on the escalator. Essentially the sign means to
convey that people who have dogs with them when using the escala-
tor should carry those dogs. The deictic your could be added to this,
with an optional by you, to indicate precisely which dogs (relative to
the hearer) need to be carried, and optionally who is responsible for
doing that: Your dogs must be carried (by you) on this escalator. In prac-
tice this seems like overkill because no competent user of language
would really misunderstand the intention behind the original sign.
6. In principle, just as (1.6a) suggests that only men can apply for the
job, (1.6b) suggests that only women can apply for the job. In one
way, this inference should be stronger than the inference from
(1.6a), in that female pronouns in English are not customarily used
for referents of unspecified gender, so the ‘person’ sense of (1.6b) is
not as readily available as it is for (1.6a). However, against this, in
recent times people have sometimes deliberately opted to use female
pronouns in precisely this way, to counter the historical imbalance
of male pronouns being used in a thoughtless and exclusionary way
in cases like (1.6a). So a possible interpretation of (1.6b) is that there
is no gender restriction on application, but the speaker is making a
point about gendered language use. In principle, (1.6c) has a sense
in which it suggests that only people who use they pronouns can
apply for the job, but a more likely interpretation is that the speaker
is using they to refer to any individual without particularising their
gender (as is done in this book).
suggested answers to the e xercises 163
Chapter 2
1. Distrust, disregard and dislike can be glossed respectively as ‘not trust’,
‘not regard’ and ‘not like’. In these cases, the verb itself falls within
the scope of not. Disprove and dissuade can be glossed as ‘prove not
(to be the case)’ and ‘persuade not (to do)’. In these cases, the verb
does not fall within the scope of not, although other material in the
sentence meaning does.
2. The incorrect analysis of not good enough corresponds to the bracket-
ing ‘(not good) enough’. In fact, the usual bracketing for this expres-
sion would be ‘not (good enough)’. Good enough means ‘adequate’ and
here it falls into the scope of not and is negated, so not good enough
means ‘inadequate’. (The negation encoded by not can alternatively
be encoded by the negative prefix in-, which fits with adequate.)
3. The intended meaning of the sentence could be represented as ‘we
(don’t like (the same things))’, as opposed to ‘we (don’t (like the same
things))’. On the latter interpretation, the meaning of the sentence
would be that it is not the case that the speaker and hearer like the
same things. This interpretation would be odd given the preceding
context, the speaker having just said You and I are well suited, unless
one or other of the sentences was intended ironically. However,
there are alternative utterances that would have removed the ambi-
guity entirely: for instance, the speaker could have said We dislike
the same things. This only admits the first of the two interpretations
discussed above: when the concept of ‘not liking’ is expressed by the
single word dislike, its status as a constituent is not in doubt.
4. (b) entails (a): we could argue that there is a sense of love that doesn’t
entail like, but that is not the sense of love that is usually applicable to
the relation between students and a course, even at the best of times.
(c) and (d) entail each other: they are paraphrases.
5. (a) They were soundless. (b) They were silent. (c) They were
noiseless.
(a) ⇒ (b), (a) ⇒ (c), (b) ⇒ (a), (b) ⇒ (c), (c) ⇒ (a), (c) ⇒ (b)
6. Awake and asleep can be considered complementaries because She is
awake entails that She is not asleep and She is not asleep entails that She is
awake. (Furthermore, She is not awake entails that She is asleep and She
is asleep entails that She is not awake: these are logical consequences
of the first two entailments.) I take half-awake, half-asleep and dozy to
denote different ways of being awake, rather than denoting an inter-
mediate region between ‘awake’ and ‘asleep’. Note, for instance, that
He’s dozy but still awake is not semantically problematic, but *He’s dozy
but still asleep is awkward; and that, in response to someone saying
“You’re asleep”, it’s acceptable to respond “No, but I admit I’m

dozy/half-asleep/only half-awake”.
7. Of this set of words, the immediate hyponyms of footwear seem to be
shoes, slippers and boots. Sneakers and trainers are synonyms, and these
along with sandals are hyponyms of shoes; galoshes are a hyponym of
boots. Although shoes enters into this hierarchy at a level at which it
contrasts with boots and slippers, there is potentially another sense of
shoes in which it could be considered roughly synonymous with foot-
wear, and used in its place as a superordinate: for instance, we would
normally go to a shoe shop rather than a ?footwear shop to buy boots and
slippers as well as shoes in the narrower sense of the word.
Chapter 3
1. We could think of a prototypical shoe as having an upper, a sole and
a heel, although we could also think of the heel as being a part of the
prototypical sole. The prototypical upper has a tongue, and possibly
some kind of fastener, although there is not an obvious superordinate
term for the various incompatible possibilities (laces, buckle, Velcro,
etc.).
2. Under this set of definitions, side is a superordinate for top, bottom,
front and back. The description names these as four different kinds
of side, and the relation of incompatibility holds between them,
because they are introduced with different and incompatible
descriptions. However, if we wanted also to use the word side in
the familiar way, to describe the spatial part that is not the top, the
bottom, the front or the back, we would have to consider this to be a
distinct sense of side that is a hyponym of the more general sense of
side used here.
3. Count senses of these words are illustrated by There is a paper on my
desk, How many glasses shall I wash? and Whole cheeses are on sale at that
stall. Mass senses are illustrated by We use too much paper, It can be
expensive to recycle glass and Cheese is made from milk.
In the count sense, taking paper to mean ‘newspaper’ (rather than,
say, conference paper), its hyponyms include broadsheet and tabloid
(and more recently Berliner, as well as freesheet and so on). Hyponyms
of glass in the count sense include goblet, wineglass, tumbler, and so on.
Hyponyms of cheese include any kind of cheese that is formed in dis-
tinct units: a Brie, a Stilton, an Edam, and so on.
In the mass sense, hyponyms of paper include newsprint, printer
paper, and so on. Hyponyms of glass include window glass, bulletproof
glass and safety glass. Hyponyms of cheese include Brie, Stilton and
Edam, but also types of cheese that do not form distinct units, such as
cottage cheese, soft cheese, and so on.
As a generalisation, the mass nouns (including the hyponyms)
denote materials or substances. The count nouns (including the
hyponyms) denote kinds of thing: newspapers, drinking vessels and
formats in which cheese is produced.
4. In Have you ever eaten rabbit?, the noun rabbit, which would normally be
a count noun, is coerced into a mass noun interpretation in which it
denotes a substance. If the article a were present, this reading would
be difficult to obtain. A similar effect holds even if we replace rabbit
with something less familiar as a potential foodstuff such as bear or
crocodile. Removing the count sense perhaps makes the individual
identity of the animal less salient: Have you ever eaten a rabbit? perhaps
has a slightly more accusatory flavour, and I couldn’t eat a rabbit
implicitly explains the speaker’s choice in terms of how they feel
about rabbits in a way that I couldn’t eat rabbit does not.
5. As discussed in the chapter, to the left of is generally ambiguous, in
that it’s not clear whether the speaker means ‘to the left from the
speaker’s perspective’, ‘to the left from the hearer’s perspective’ or ‘to
the left from the perspective of the object being used as a reference
point’. However, unlike chairs, stools do not tend to have an intrinsic
orientation – they are more symmetrical than chairs, and lack a back
– so what would be meant by ‘to the left from the perspective of the
stool’ is impossible to determine. (I assume that ‘to the left from the
perspective of the chair’ effectively means ‘to your left, if you were
sitting on the chair the normal way round’.) Consequently, to the left
of the stool can’t really be used that way, and realistically has to mean
‘to the left from the speaker’s perspective’ or ‘to the left from the
hearer’s perspective’. In that sense, it’s actually slightly less ambigu-
ous than to the left of the chair.
Chapter 4
1. The adjectives for which quite is a downtoner are gradable: it makes
sense to talk in terms of the extent to which someone or something is
clever, late, small or unusual. These could, for instance, also be modified
by very, which would convey a stronger meaning than quite in each
case. The adjectives for which quite is a maximiser are not gradable:
you could think of them as corresponding to the endpoints of scales.
They can’t be modified by very but they could be modified by com-
pletely. It’s just about possible to imagine quite acting as a maximiser
on a word like clever, but there seems to be a strong preference for
interpreting it as a downtoner (at least in British English) when the

opportunity presents itself (that is, when the adjective is gradable).
2. The adjectives young, rude, unpalatable, tasty and weak seem generally
to yield biased how-questions. Old, polite and strong do not, in general.
Tasty seems to give rise to a biased question where palatable would
not, perhaps because it is a stronger positive adjective than the
“neutral” alternative.
3. Taking royal visitor to mean ‘a royal person who is visiting’, this admits
a straightforward intersective analysis. Royal correspondent would nor-
mally mean ‘journalist who reports on royalty’, and on that reading is
not intersective, because the meaning contributed by royal can only
be understood relative to the domain of correspondent. If, on the other
hand, we meant the phrase to mean ‘a royal who writes letters’ (Queen
Christina was Descartes’ royal correspondent), an intersective analysis
would work. Heavy eater is not intersective, as it means ‘someone who
eats heavily’ rather than ‘someone who eats and is heavy’. We can
be reasonably confident about this because eater is hardly ever used
as a free-standing expression to mean ‘someone who eats’. Wise fool,
in order to be understood as a non-contradictory expression, must
mean ‘a professional fool (i.e. clown or jester) who is wise’, and can
therefore be interpreted intersectively.
4. When we describe Proxima Centauri as small, cool, red and near, we
intend these adjectives to be interpreted relatively, and specifically
with reference to the standards by which we judge stars. However,
when we describe it as the closest other star to the Sun, this is an
absolute description, and holds irrespective of whether or not we
consider the actual distance involved (4.24 light-years) to be large
or small.
Chapter 5
1. Compared with The minister made the civil servant resign, the sentence
The minister resigned the civil servant is a single clause, and we would
expect it to encode direct causation (whereas the two clause causa-
tive made . . . resign can encode either direct or indirect causation; see
Section 8.2.4 for a similar example). Given the lengthy gap between
the minister’s announcement and the civil servant’s actual resigna-
tion, it seems more likely that the causation was indirect: that is,
that the minister’s announcement did not by itself have the effect of
causing the civil servant to resign. That being the case, resigned isn’t
intended as a causative verb in the question Who is going to be resigned
next? The question cannot be analysed as meaning ‘Who will be made
to resign next?’, but something more like ‘Whose resignation will be

announced, without it actually having taken place, next?’ The appar-
ent oddness of how this question is formed turns out to be quite an
effective way of getting people to think about the meaning of the
word resign, and potentially to reflect upon how the minister is trying
to exercise their power.
2. Humpty was ‘together’ (intact) before his ‘great fall’. The soldiers
failed in the task of getting Humpty back into this previous state of
togetherness. Put is a causative verb, and the restitutive adverb again
modifies an embedded proposition ‘Humpty is together’, rather than
the main clause action verb put.
3. Sentences (a) and (b) are unaccusative, because the referent of the
subject does not consciously carry out the action, as confirmed by
the peculiarity of these sentences with carefully: *The kite carefully flew,
*My heart carefully sank. Sentence (c) is unergative: reading is some-
thing that students do consciously and they can do it carefully.
4. As discussed in the chapter, give requires three arguments, corre-
sponding to a giver (Agent), a Recipient and something to be given
(Theme). We can think of pay as representing a specific case of giving
something, usually money, in return for goods or services rendered.
Perhaps because the verb pay encodes information about the Theme,
this does not need to be made explicit in the sentence, although it can
optionally be specified. Thus we can say John paid Mary or John paid
Mary £1,000, whereas we cannot say *John gave Mary although we can
say John gave Mary £1,000. In fact, as the contrast between John paid
Mary and John paid the bill indicates, we can even omit the Recipient
with the verb pay: Mary fulfils the role of Recipient, but the bill is more
like a Theme (we can’t say *John paid the bill £1,000). If the Theme
and Recipient are both inferrable, we can use pay in what looks like
an intransitive construction (We went to the restaurant; John paid),
although there is some debate about whether these various senses of
pay are really analysable as the same verb.
Chapter 6
1. Recently goes with the past time group that includes yesterday. Note
the unacceptability of *He is happy recently, *He shops at the corner store
recently and *I will do it recently. (We noted that recently can be used
with present perfect forms such as I have been there recently, where it
indicates that the period between the occurrence reported and the
time of utterance is relatively short, but the period assessed as short
is all before now; that is to say, in past time.) Soon is like then in Table
6.2, in that it is acceptable with past and future times, but not present:
*I’m eating cake soon is no good if reference is to the present, but is fine
with future reference in I’m catching a train soon, even though both of
these utterances are present progressive in form. In such cases, soon
is directly anchored to the time of utterance, with the meaning ‘a
short time after now’. But when soon is used with past reference, as in
It began to rain, but soon stopped, the deixis is indirect, and soon means ‘a
short time after that time’. In the example, ‘that time’ is the time that
the past tense form began points to deictically.
2. Arthur’s a tyrant seems to express a long-lasting situation (a state), and
probably represents a judgement about Arthur’s personality. Arthur’s
being a tyrant is a longer form, which encourages a pragmatic line of
reasoning: why did the speaker not use the shorter alternative? In this
case, the use of being seems to convey some kind of “provisionality”
and invites the inference that, although Arthur is presently being a
tyrant, the speaker does not consider ‘being a tyrant’ to be a general,
non-time-specific facet of Arthur’s personality.
3. The verb told is past simple; had saved is past perfect; were dying is
past progressive. The report does not allow certainty over how far
the duration of “people dying early” should extend to the right: as
far as the claim is concerned, it could be that people were still dying
early at the time of the report, or the deaths had stopped before the
company told the government about it, or that the deaths stopped
sometime between the telling and the report. That is, we cannot
tell whether the company told the government that people are dying
early or that people were dying early or that people have been dying early. It
would also be relevant to the diagram to know whether the company
told the government you have saved a lot of money or you saved a lot of
money. These two possibilities can both be reported by means of a
past perfect.
4. You said you would . . . or You said you were going to . . . are possibilities
for (c). Would is a past tense form of the future marker will; were going
to combines the past tense were with the future marker going to.
The request to borrow the book might have been made with will
you . . . or would you . . ., although in this case the use of would conven-
tionally marks politeness rather than past tense. Can you . . . or could
you . . . would also be options; again, the use of could is about politeness
rather than tense. Are you going to . . . would also be a possibility. In fact,
these options are all forms of indirect request, in that they all appear
to ask whether the speaker intends to lend the book, or is capable of
doing so, rather than actually stating a preference for them to do so.
There is more on indirect requests in Section 11.2.1.
Chapter 7
1. As this example hopefully makes clear, the default description
without a modal verb is generally stronger. It is appropriate to use
a modal when the speaker lacks direct information about a state of
affairs but is presenting a conclusion based on reasoning about the
available evidence. Therefore the presence of modal marking gener-
ally invites the inference that the speaker is not really sufficiently
sure of the facts to make a non-modal utterance. Of the modal
options, must indicates more confidence than might, but still not as
much confidence as would be indicated by the complete omission of
the modal marking.
2. They must be made from buckwheat can be either deontic (a demand or
strong recommendation that buckwheat be used) or epistemic (an
inference, perhaps from the colour or taste of the item, that buck-
wheat is an ingredient).
We must get up early tomorrow is deontic, and specifies something
that, in the speaker’s view, needs to be the case. We could imagine it
being epistemic in a fictional setting in which the speaker has access
to information about what will happen the next day – or indeed a
situation in which the speaker is acting as though they have such
information, for instance in response to a fortune teller predicting
that Tomorrow you will go on a long journey. Otherwise, what will happen
tomorrow is inherently too uncertain to justify epistemic must.
The email needn’t have been sent can bear either interpretation. On
the deontic interpretation it conveys that there was no necessity for
the sending of the email; on the epistemic interpretation it conveys
that it is possible that the email has not yet been sent. For the author
the former interpretation is preferable (and the latter would be better
achieved by . . . might/may not . . .).
I can hear you now is a slightly unusual case, in that it indicates
capability: sound level, transmission and reception conditions mean
that what is coming from you is now being heard. Some semanticists
take this sort of modality to be similar to deontic modality: physics
and physiology are permitting something to happen. Others would
classify this differently, potentially as a kind of epistemic modality.
A pointer to this example being an unusual case is that the modal can
be removed without having any substantial effect on the meaning: I
hear you now is almost a paraphrase of I can hear you now.
In They might or might not make it, it is somewhat implausible to
think of might as encoding deontic modality, perhaps because it is
hard to imagine permission being given for people to succeed or not
succeed. A more natural interpretation of the sentence involves might

encoding epistemic modality. Although it is generally possible to
use might to report permission having been given, Biber et al. (1999:
491) found that almost all the usages of might in their large sample of
conversational and academic English were epistemic.
You better apologise is deontic. This is a reduced form of You had
better apologise (or You’d better . . .). The idiom had better is not used to
express epistemic modality (see Huddleston and Pullum 2002: 196).
We label this an idiom in part because, despite containing the form
had, it is not used to talk about the past.
3. Many right answers are possible here. For instance:
Guests may check in between 3pm and midnight is epistemic if
someone at a hotel is explaining to a new member of staff on the
front desk when they should expect people to come to check in. It is
deontic if it is the text on a notice showing the permitted checking-
in hours.
You must be a musician would be epistemic if said by a taxi driver to
a passenger carrying a cello case. If it were spoken by a music teacher
to a promising pupil, it could be taken as deontic.
He might say something would be epistemic if it concerns speculation
about whether a shy individual at a party will eventually overcome
that shyness and join a conversation. It would be deontic if it consti-
tuted a complaint that that individual was bringing the mood down
by his silence and ought to speak, out of politeness.
4. Deontic may not is similar to can’t: negation has wider scope, giving
rise to the meaning ‘not (possibly (they have an invitation))’.
However, epistemic may not behaves like mustn’t: modality has
wider scope, giving rise to the meaning ‘possibly (not (they have an
invitation))’.
5. In the situation described, The witness may | not be named is deontic,
with relative scope ‘possibly (not (the witness be named))’. This
is different from the general pattern for may (as in the preceding
question), where there is wider scope for negation when the inter-
pretation is deontic, and wider scope for modality with epistemic
interpretations.
6. Based on the “isomorphism” idea, we might expect one to have
wider scope because it is in the subject noun phrase. On this reading,
the sentence describes a single versatile machine that no product
escapes being tested by: ‘there is one machine (as for every product
(the machine tests it))’. However, in this case, there is an opposing
tendency for each to have wider scope, which makes an alternative
reading possible: ‘as for every product (there is one machine (the
machine tests that product))’. On this reading, we are still assured

that every product is tested by a machine, but it could be a different
machine for each different type of product.
Chapter 8
1. In this example, B probably succeeds in communicating their iden-
tity to A, because A can infer this from hearing their voice. Just in
propositional terms, it looks as though the Maxim of Quantity is
being disobeyed, because no new information is supplied. However,
this can be seen as reasonable if it is taken as a signal that no addi-
tional information is needed because the voice should be immedi-
ately recognisable. If this is the case, then this minimal utterance
is adequate for the communicative purpose, and a more explicit
utterance – such as B giving their name – would arguably violate the
Maxim of Quantity (by providing more information than needed) or
the Maxim of Manner (by being more verbose than necessary).
2. Assuming that B is cooperative, the failure to give a more specific
answer to A's question could be taken to implicate that B doesn't
know where specifically on the fourth floor the sociolinguistics books
are to be found. Alternatively, B may think that the information they
have given is sufficient, for instance because the linguistics section
can easily be located once the hearer is on the correct floor, and
sociolinguistics can easily be found within it.
3. There are various possible answers. One class of situation in which it
seems to be acceptable to say something like these sentences is when
the presupposition has already been introduced by someone else.
For instance, if someone affirms that Jessica regretted arguing with her
boss, but you know that they are actually thinking of someone else
(say, Jane), it seems to be OK to respond Jessica didn’t regret arguing
with her boss before continuing You’re thinking of Jane. It would be clear
that you weren’t committed to the truth of Jessica having argued
with her boss, even though you had uttered a sentence that formally
presupposed it. Similar contexts of utterance would rescue the other
two example sentences as well.
4. The most concise alternative that captured all the information
would be something like The beautiful princess Eleanor lived with her
wicked stepmother in an enchanted castle. In the short version, the exist-
ence of both the princess and her stepmother are introduced by
expressions that carry existential presuppositions, which the hearer
can accommodate. In the original longer version, their existence is
explicitly stated (there was, Eleanor had).
5. Given that there was no war in which Switzerland was involved,

either as a location or a belligerent, interpreting the war in this
example is not straightforward. It requires us to identify a contextu-
ally salient war which has the properties of occurring during Carolyn
Cole’s career and of having finished (which is suggested by the past
tense and the use of during). Given that numerous wars potentially
fit the bill, we would likely need more context to disambiguate the
speaker’s intended meaning.
Chapter 9
1. This is an example of metonymy: beaks is used as a metonym for
birds. This is supported by the knowledge both that birds have beaks,
and that the corresponding metonymy mouths is used for human
dependents.
2. Based on the use of like, it initially looks as though Ali’s maxim
involves two similes: he is saying that his boxing style is like a but-
terfly with regard to his floating, and like a bee with regard to his
stinging. But actually, of course, Ali’s style did not literally involve
floating or stinging, so in fact both of these clauses are metaphorical.
3. The expression very stable genius was famously used by Donald
Trump in reference to himself just before that headline was written,
and was subsequently widely quoted, largely by people who thought
that that self-assessment was considerably off-target and used it
ironically. Thus, the article title suggests an ironic interpretation of
the expression. There is perhaps a suggestion of a metaphorical iden-
tification of José Mourinho with Donald Trump, if we take very stable
genius conventionally to refer to Trump (which itself might be a case
of metonymy). In fact, the article supports this latter interpretation,
in that it characterises Mourinho as an expert in drawing attention
to himself to deflect it from other issues, an attribute also associated
with the then President.
Chapter 10
1. It is a reasonable assumption about prototypical hearers that each
has a head, which justifies immediate use of definite reference. And
the warning about the step would typically be given in a situation in
which it is possible for the addressee to experience the step directly,
for example by looking or by tapping it with a stick, again making it
part of the background without further ado. The answer is not simply
that warnings of this kind have to be issued in a hurry, as there is no
barrier to using indefinites in such cases, for instance where the

danger is not so clearly evident: Careful, there’s a snake in there.
2. Between these two versions of the pseudo-cleft, the presuppositions
are the same, namely (in this case) that ‘something hit the sofa’. This
is still what is conveyed by the negated versions of both forms of
the sentence: It wasn’t the meteorite that hit the sofa and What hit the sofa
wasn’t the meteorite.
3. It would be reasonable to correct Tom by saying either “No, you’re
wrong: Lucy borrowed the DICtionary”, or “No, you’re wrong: it
was the DICtionary that Lucy borrowed”. Both of these relate to sce-
nario (b) in which Lucy borrowed something, although it wasn’t the
atlas. This is because Tom’s it-cleft presupposes that ‘Lucy borrowed
something’, and that presupposition is satisfied by the information in
(b) rather than that in (a).
4. It seems that these kinds of usage are acceptable if the Question
Under Discussion in some sense involves the presuppositional
expression. For instance, if the QUD is “whether Jessica regret-
ted arguing with her boss”, it is reasonable to respond Jessica didn’t
regret arguing with her boss before explaining the presupposition
failure (You’re thinking of Jane). The specific case discussed in an
earlier exercise, in which the preceding turn involves someone
erroneously asserting that Jessica regretting arguing with her boss, is a
special case of this, because that sentence could also be interpreted
as answering the QUD of “whether Jessica regretted arguing with
her boss”.
Chapter 11
1. The speech act of warning bus riders not to smoke (or perhaps
threatening them with adverse consequences) is mitigated by pre-
senting part of it as a speech act of thanking. Thanking presupposes
that the addressee has done something that is appreciated by the
bus company. The mitigation is perhaps addressed to habitually
non-smoking passengers, who might otherwise have been affronted
by the apparent presumption that they needed to be warned not to
smoke.
2. (a) Literally a question, this has the force of an order.
(b) Literally an order, this has the force of an offer.
(c) Literally a question, this has the force of a statement.
(d) Literally a statement, this has the force of an expression of sym-
pathy or condolence.
(e) Literally an order, this has the force of a statement.
(f) Literally an order, this has the force of an expression of

condolence.
(g) Literally a question, this has the force of a statement (to the
effect of ‘I have never let you down’) – it is what is called a “rhe-
torical question”.
(h) Literally a statement, this has the force of a recommendation.
3. “I should let you go” appears to serve primarily as a device for bring-
ing about the end of a conversation. Formally it makes a deontic
statement about what the speaker should permit to happen. By
articulating this, the speaker could be said to implicitly give this
permission. However, giving this permission further suggests that
the recipient might wish to avail themselves of this permission. By
calling attention to the fact that the hearer may leave the conversa-
tion, the speaker invites the inference that the speaker thinks the
hearer should leave the conversation, and hence that the conversa-
tion should come to an end. Compared with “I should go”, the form
“I should let you go” is arguably more polite, in that it suggests that
the pressing calls on the hearer’s time (rather than on the speaker’s
time) are the reason that the conversation should end. However,
to the extent that it has become euphemistic for “I wish to end this
conversation”, that effect of additional politeness is arguably being
eroded by custom.
4. We might expect the actual utterance to be infelicitous, on the basis
that the issue of what the Queen would be naming had been settled
in advance, and this did not extend to the entire road. In practice, the
question of whether the Queen actually had the authority to name
the entire road was unclear, and raising that issue might have been
embarrassing for reasons of protocol. Thus, in practice, the Queen’s
utterance did have the effect of renaming the entire road – whether
it was erroneous, as the legend has it, or not.
References
Austin, J. L. (1962), How to Do Things with Words, Oxford: Oxford University

Press.
Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad and Edward
Finegan (1999), Longman Grammar of Spoken and Written English, Harlow:
Pearson.
Chomsky, Noam (1981), Lectures on Government and Binding, Dordrecht: Foris
Publications.
Clark, Herbert H. (1975), ‘Bridging’, in R. C. Schank and B. L. Nash-Webber
(eds), Theoretical Issues in Natural Language Processing, New York: Association
for Computing Machinery, pp. 169–74.
Clark, Herbert H. (2004), ‘Pragmatics of language performance’, in L. R. Horn
and G. Ward (eds), Handbook of Pragmatics, Oxford: Blackwell, pp. 365–82.
Clark, Herbert H. (2012), ‘Wordless questions, wordless answers’, in J. P. de Ruiter
(ed.), Questions: Formal, Functional and Interactional Perspectives, Cambridge:
Cambridge University Press, pp. 81–100.
Clark, Herbert H. and Richard J. Gerrig (1984), ‘On the pretense theory of
irony’, Journal of Experimental Psychology: General, 113: 121–6.
Coventry, Kenny R., Elena Andonova, Thora Tenbrink, Harmen B. Gudde and
Paul E. Engelhardt (2018), ‘Cued by what we see and hear: spatial reference
frame use in language’, Frontiers in Psychology, 9: 1287.
Croft, William (1991), Syntactic Categories and Grammatical Relations: The Cognitive
Organization of Information, Chicago: University of Chicago Press.
Cruse, Alan (2011), Meaning in Language, 3rd edn, Oxford: Oxford University
Press.
Davis, Wayne (2019), ‘Implicature’, in E. N. Zalta (ed), The Stanford Encyclopedia
of Philosophy, Fall 2019 edn, <https://plato.stanford.edu/archives/fall2019
/entries/implicature/>, last accessed 16 June 2022.
Deamer, Felicity, Richard Breheny and Nausicaa Pouscoulous (2010), ‘A
contrastive look at the processing of metaphor and hyperbole’, UCL Working
Papers in Linguistics, 22: 1–16.
De Swart, Henriëtte and Helen de Hoop (2000), ‘Topic and focus’, in L. Cheng
and R. Sybesma (eds), The First Glot International State-of-the-Article Book,
Berlin: Mouton de Gruyter, pp. 105–30.
175
Dowty, David (1991), ‘Thematic proto-roles and argument selection’, Language,

67: 547–619.
Fellbaum, Christiane (2000), ‘Autotroponymy’, in Y. Ravin and C. Leacock (eds),
Polysemy: Theoretical and Computational Approaches, Oxford: Oxford University
Press, pp. 52–67.
Feynman, Richard P. (1999), The Pleasure of Finding Things Out, New York: Perseus.
Giegerich, Heinz J. (1992), English Phonology: An Introduction, Cambridge:
Cambridge University Press.
Glucksberg, Sam and Boaz Keysar (1990), ‘Understanding metaphorical com-
parisons: beyond similarity’, Psychological Review, 97: 3–18.
Gordon, David and George Lakoff (1971), ‘Conversational postulates’, in
Papers from the Seventh Regional Meeting of the Chicago Linguistic Society, Chicago:
Chicago Linguistic Society, pp. 63–84.
Grant, Lynn and Laurie Bauer (2004), ‘Criteria for re-defining idioms: are we
barking up the wrong tree?’, Applied Linguistics, 25: 38–61.
Grice, H. Paul (1975), ‘Logic and conversation’, in P. Cole and J. L. Morgan
(eds), Syntax and Semantics, vol. 3, New York: Academic Press, pp. 41–58.
Grice, H. Paul (1989), Studies in the Way of Words, Cambridge, MA: Harvard
University Press.
Hewson, John and Vit Bubenik (1997), Tense and Aspect in Indo-European
Languages, Amsterdam: John Benjamins.
Horn, Laurence R. (1972), ‘On the semantic properties of logical operators in
English’, PhD thesis, University of California, Los Angeles. Distributed by
Indiana University Linguistics Club.
Horn, Laurence R. (1984), ‘Towards a new taxonomy for pragmatic inference:
Q-based and R-based implicature’, in D. Schiffrin (ed.), Meaning, Form, and
Use in Context: Linguistic Applications, Washington: Georgetown University
Press, pp. 11–42.
Horn, Laurence R. (1989), A Natural History of Negation, Chicago: University of
Chicago Press.
Huddleston, Rodney and Geoffrey K. Pullum (2002), The Cambridge Grammar of
the English Language, Cambridge: Cambridge University Press.
Johannsen, Katrin and J. P. de Ruiter (2013), ‘The role of scene type and priming
in the processing and selection of a spatial frame of reference’, Frontiers in
Psychology, 4: 182.
Jurafsky, Dan (2004), ‘Pragmatics and computational linguistics’, in L. R. Horn
and G. Ward (eds), Handbook of Pragmatics, Oxford: Blackwell, pp. 578–604.
Kearns, Kate (2011), Semantics, 2nd edn, Basingstoke: Macmillan.
Kennedy, Chris (2012), ‘Adjectives’, in G. Russell and D. Graff Fara (eds),
The Routledge Companion to Philosophy of Language, New York: Routledge,
pp. 328–41.
Klein, Wolfgang (1992), ‘The present perfect puzzle’, Language, 68: 525–52.
Lappin, Shalom (2001), ‘An introduction to formal semantics’, in M. Aronoff
and J. Rees-Miller (eds), The Handbook of Linguistics, Oxford: Blackwell,
pp. 369–93.
references 177
Leech, Geoffrey (1983), Principles of Pragmatics, London: Longman.

Leech, Geoffrey, Paul Rayson and Andrew Wilson (2001), Word Frequencies in
Spoken and Written English, Harlow: Pearson Education.
Levinson, Stephen C. (1983), Pragmatics, Cambridge: Cambridge University
Press.
Levinson, Stephen C. (2000), Presumptive Meanings: The Theory of Generalized
Conversational Implicature, Cambridge, MA: MIT Press.
Lycan, William G. (2019), Philosophy of Language, 3rd edn, London: Routledge.
McNally, Louise (2011), ‘The relative role of property type and scale structure
in explaining the behavior of gradable adjectives’, in R. Nouwen, R. van
Rooij, U. Sauerland and H.-C. Schmitz (eds), VIC 2009: Papers from the ESSLLI
2009 Workshop on Vagueness in Communication (Lecture Notes in Artificial
Intelligence 6517), Berlin: Springer, pp. 151–68.
Mervis, Carolyn B. and Eleanor Rosch (1981), ‘Categorization of natural
objects’, Annual Review of Psychology, 32: 89–115.
Miller, George A. and Christiane Fellbaum (1991), ‘Semantic networks of
English’, Cognition, 41: 197–229.
Miller, Jim (2002), An Introduction to English Syntax, Edinburgh: Edinburgh
University Press.
Musolino, Julien, Stephen Crain and Rosalind Thornton (2000), ‘Navigating
negative quantification space’, Linguistics, 38: 1–32.
Palmer, Frank (1990), Modality and the English Modals, 3rd edn, London:
Longman.
Perrault, C. Raymond and James F. Allen (1980), ‘A plan-based analysis of indi-
rect speech acts’, Computational Linguistics, 6: 167–82.
Quirk, Randolph, Sidney Greenbaum, Jan Svartvik and Geoffrey Leech (1985),
A Comprehensive Grammar of the English Language, London: Longman.
Radford, Andrew (2004), Minimalist Syntax, Cambridge: Cambridge University
Press.
Roberts, Craige (2012), ‘Information structure: towards an integrated formal
theory of pragmatics’, Semantics and Pragmatics, 5 (6): 1–69.
Rooth, Mats (1996), ‘Focus’, in S. Lappin (ed.), The Handbook of Contemporary
Semantic Theory, Oxford: Blackwell, pp. 271–97.
Saeed, John I. (2015), Semantics, 4th edn, Oxford: Blackwell.
Schegloff, Emanuel A. and Harvey Sacks (1973), ‘Opening up closings’,
Semiotica, 8 (4): 289–327.
Schumacher, Petra (2019), ‘Metonymy’, in C. Cummins and N. Katsos (eds),
Oxford Handbook of Experimental Semantics and Pragmatics, Oxford: Oxford
University Press, pp. 316–30.
Searle, John R. (1975), ‘Indirect speech acts’, in P. Cole and J. L. Morgan (eds),
Syntax and Semantics, vol. 3, New York: Academic Press, pp. 59–82.
Searle, John R. (1979), Expression and Meaning: Studies in the Theory of Speech Acts,
Cambridge: Cambridge University Press.
Solt, Stephanie (2011), ‘How many mosts?’, in I. Reich, E. Horch and D. Pauly
(eds), Proceedings of Sinn und Bedeutung 15, Saarbrücken: Universaar, pp. 565–79.
Sperber, Dan and Deirdre Wilson (1981), ‘On verbal irony’, in P. Cole (ed.),
Radical Pragmatics, New York: Academic Press, pp. 295–318.
Sperber, Dan and Deirdre Wilson (1995), Relevance: Communication and Cognition,
2nd edn, Oxford: Blackwell.
Stern, Josef (2000), Metaphor in Context, Cambridge, MA: MIT Press.
Tenny, Carol (2000), ‘Core events and adverbial modification’, in C. Tenny and
J. Pustejovsky (eds), Events as Grammatical Objects: The Converging Perspectives of
Lexical Semantics and Syntax, Stanford, CA: CSLI, pp. 285–334.
Trask, R. L. (1993), A Dictionary of Grammatical Terms in Linguistics, London:
Routledge.
Traugott, Elizabeth Closs (2000), ‘Semantic change: an overview’, in L. Cheng
and R. Sybesma (eds), The First Glot International State-of-the-Article Book,
Berlin: Mouton de Gruyter, pp. 385–406.
Van der Auwera, Johan and Vladimir A. Plungian (1998), ‘Modality’s semantic
map’, Linguistic Typology, 2: 79–124.
Vendler, Zeno (1957), ‘Verbs and times’, The Philosophical Review, 66: 143–60.
Weisberg, Steven M. and Anjan Chatterjee (2020), ‘Reference frames in spatial
communication for navigation and sports: an empirical study in ultimate
frisbee players’, Cognitive Research: Principles and Implications, 5: 53.
Wilson, Deirdre (2006), ‘The pragmatics of verbal irony: echo or pretence?’,
Lingua, 116: 1722–43.
Wilson, Deirdre and Dan Sperber (2002), ‘Relevance theory’, UCL Working
Papers in Linguistics, 14: 249–90.
Wittgenstein, Ludwig (1953), Philosophical Investigations, Oxford: Blackwell.
Index
Note: Bold print page numbers indicate where some aspect of a technical term is
explained. Closely related forms are listed together.
accommodation, 115, 118, 138 cancellability, 13, 103–4, 108

accomplishment, 77 cardinality, 90–1
active, 140, 141 causal relation, 111
addressee, 2; see also hearer causatives, 57–61
adjacency pair, 156 causative sentence, 58
adjectives, 23–6, 44–54, 165–6 causative verb, 58–9, 61
privative adjectives, 50–1 clauses, 19, 55, 56, 59, 82, 84, 97
relative adjectives, 51–2 embedded clauses, 56
adverbs, 44, 47, 53, 56–7, 67, 70–5, 78, wh-clauses, 137–8
165–6 clefts
downtoner adverbs, 53, 165–6 it-clefts, 138–9, 144–5
maximiser adverbs, 47, 53, 165–6 pseudo-clefts, 137–9, 144–5
time adverbials, 67, 70–5, 78 coercion, 40–1, 165
aftermath, 77–9 cognitive effect, 113–14
ambiguity, 5, 6, 8, 20, 38, 52, 72, 86–9, 96, cognitive effort, 113–14
98 collectivity, 96–7
and, 111 comparative forms, 25–6, 44–5
animacy, 64, 140 complement (set), 91
antonymy, 24–6, 45 complementarity, 23–4, 25, 46
arguments, 55, 56–7, 58–9, 61–4, 141 compositionality, 19, 21, 49, 122
articles, 35–6, 40, 135–6 context, 2, 3, 4–5, 9–11, 12, 38, 51–2, 75,
aspect, 66–9, 73–9 78, 86–7, 89, 95, 101–2, 104–5, 125,
habitual aspect, 67–8, 74–5 126, 138, 144, 152, 158
perfect aspect, 68–9, 77–9 contradiction, 6, 18
progressive aspect, 67, 69, 76–7 contrast, 138, 142–3
simple aspect, 66–7, 69, 74 conversational implicature see
asterisk (notation), 6 implicature
Austin, J. L., 148, 149, 151 converseness, 25–6, 44, 141
auxiliaries, 70, 83–4 syntactic converses, 141
modal auxiliaries, 83–4 conversion, 61
cooperation, 11, 105–11, 134
background, 76, 116, 136–7, 155 Cooperative Principle, 106
179
copular sentences, 56, 123–4 hearer, 2–3, 5, 12, 19, 35–6, 38, 47, 98,
count noun, 7–8, 40–2, 164–5 101–3, 112–14, 115–16, 124, 134,
135–6, 140, 143, 150, 151–5
declarative, 17, 137, 152–3 hereby, 149, 151, 154
definiteness, 35, 135–7 historic present tense, 71, 73
deixis, 9–10, 38, 67, 70 historical change, 12
denotation, 1–2, 7–8, 27, 33, 47–52 hyperbole, 129–30
deontic modality, 84–7, 169 hyper(o)nym see superordinate
discourse, 10, 133–6, 143–5 hyponymy, 26–30, 36–7, 123
disjoint, 91
distributivity, 96 idiom, 2, 17, 21, 101, 122
ditransitive verb, 56, 62 illocutionary act, 151
‘downtoner’, 53, 165–6 imperative, 152–3
implicature, 11, 13, 106–12, 144–5
element (of set), 90 manner implicature, 110–12, 140
embedding, 56, 59 quantity implicature, 106, 107–9, 112
embedded clauses, 56 relevance implicature, 109–10
embedded situation, 59–60 scalar implicature, 13, 108–9
encyclopedic knowledge, 12, 103, 158 incompatibility, 26, 29–30, 41
entailment, 17–19, 21–7, 32–3, 60–1, 93, indefiniteness, 34–6, 40, 135–7
116–17 indirect causation, 112
epistemic modality, 84–7, 169–70 indirect object, 56
existential presupposition, 115 indirect speech act, 153–4, 157, 158
explicature, 11, 122 inflection, 70
expression, 1, 7–9 inheritance, 36–7, 40, 41
intention, 2–3
factives, 117–18 interrogative, 152–3
felicity conditions, 149 intersection, 91
figurative language, 121–32 intersective adjectives, 48–52
figurative interpretation, 122 intonation, 98, 117, 141–3
figure of speech, 122 intransitive verb, 56–7
focus, 141–3, 144 irony, 128–9
focal stress, 141–3, 144 isomorphism, 97–8, 170–1
formal semantics, 7–8 lexical meaning, 6, 7, 10, 29, 34, 122
future, 67, 69–75, 80, 83–4 literal meaning, 11, 21, 103, 122, 124–5,
128–9
general knowledge see encyclopedic literal interpretation, 122, 153–4
knowledge litotes, 130
given information, 136–7, 140, locutionary act, 151
144–5
goals (of speaker), 69–70 manner implicature, 110–12, 140
gradability, 42–7, 109, 165–6 mass noun, 7–8, 40–1, 164–5
Grice, H. Paul, 2, 106, 112, 129 maximiser, 47, 53, 165–6
Grice’s maxims see maxims maxims (Grice), 106–12, 129, 133–4
Maxim of Quality, 106, 107–8, 110,
habitual aspect, 67–8, 74–5, 78 112, 129
has-relation, 32–7, 39–40, 41, 135–6 Maxim of Quantity, 106–8
inde x 181
Maxim of Manner, 106, 110–12 past progressive, 67, 69

Maxim of Relation, 106, 109–10, 112–13 past simple, 66–7, 69, 74
meaning perfect aspect, 18, 68, 69, 77–9
lexical meaning, 6, 7, 10, 29, 34, 122 perfect tense, 79
literal meaning, 11, 21, 103, 122, 124–5, performative, 149–50, 151, 154
128–9 perlocutionary act, 151
notation for, 4 polar question, 150
sentence meaning, 5–6, 10–11, 12, politeness, 105, 110, 153, 154, 168, 174
16–17 pragmatics, 1–3, 10–11, 12–13, et passim
speaker meaning, 5 predication, 2, 56
utterance meaning, 5–6, 10, 16–17 premises, 17–18
word meaning, 7, 10, 29, 34, 122 prepositions, 20, 23, 26, 55
meiosis, 130 preposition phrases, 20, 55, 78
member (of set), 90 prescriptivism, 12
metaphor, 39–40, 123–5, 126, 127–8, 130 present, 18, 67–8, 69, 70–3, 74–5, 78, 83,
dead metaphor, 125 154
mixed metaphor, 124–5 historic present, 71, 73
tenor, 123 present perfect, 18, 68, 78, 79
vehicle, 123–5 present simple, 67–8
metonymy, 50, 125–6 presupposition, 60, 115–19, 134, 137–8,
modality, 50, 82–9, 109 139, 142–3, 171
modal auxiliaries, 67, 72, 84 accommodation of, 36, 115–16, 118, 138
deontic modality, 84–7, 169–70 existential presupposition, 115
epistemic modality, 84–7, 169–70 triggers, 116–17
morphemes, 19–20, 21 privative, 50–1
progressive aspect, 67, 69, 70, 73, 76–7
negation, 19–20, 25, 87–90, 97–8, 116–17, proportional quantifiers, 94–6
170 proposition, 17–18, 56, 83–5, 97
neg-raising, 89 prototypes, 33–4, 35–7
new information, 134, 135, 137–40, 142–5 provisionality, 77, 168
nouns, 7–8, 29, 32–41, 44, 47–52, 55, 61,
62, 115 quantification, 40–1, 82, 90, 91–8
count nouns, 7–8, 40–2, 164–5 quantifiers, 90–8, 109, 136
mass nouns, 7–8, 40–1, 164–5 proportional quantifiers, 94–6
noun phrases, 35, 55, 62, 115 quantity implicature, 106, 107–9, 112
number, 90, 93–4 question mark (notation), 6
numerical quantifiers, 93–4 questions, 45–6, 103–4, 105, 107, 110, 113,
116–17, 143–5, 150, 152–3, 155–6
operators, 87–90, 97 polar question, 150
ostension, 7, 162 tag question, 150
wh-question, 150
paraphrase, 21–2 Question Under Discussion (QUD),
participant role, 62 143–5
passive, 140–1, 145
past, 66–7, 69–73, 74–5, 79, 83, 118 Reference, 9–10, 35, 62, 115, 126, 135–6
past participle, 70, 77, 140 Referent, 9
past perfect, 69, 79 referring expressions, 9
relative adjectives, 51–2 thematic relation, 62–4

relative scope see scope thematic role, 62
relevance implicature, 109–10 theta-role, 64
Relevance Theory, 112–14 time, 9–10, 39, 66–79, 83–4, 118, 154
restitutive, 60, 117 adverbials, 70, 72–5, 78
Rosch, Eleanor, 34 future, 67, 69–75, 80, 83–4
past, 66–7, 69–73, 74–5, 79, 83, 118
Scales, 13, 45, 46, 108–9, 130 present, 18, 67–8, 69, 70–3, 74–5, 78,
scalar implicature, 13, 108–9 83, 154
scope, 20, 87–9, 97–8, 170–1 topic, 135–6, 140
Searle, John R., 148, 150 transitive relation, 23
Semantics, 1–2, 7–8, 10–12, et passim transitive verb, 56
formal semantics, 7–8 truth, 17, 18 106, 148
sense, 8–9, 19, 21, 27 timeless truth, 71, 72
sense relations, 8–9, 16–30, 32
sentence, 1, 4–6, –5, 6, 8, 10, 11, 16–17, unaccusative verb, 56–7, 60–1
18, 19 unergative verb, 56–7
sentence meaning, 5–6, 10–11, union, 91
16–17 unmarked forms, 70
sentence type, 17, 152–3 utterance, 4, 5, 9, 10–11, 16, 17, 67, 102,
set theory, 47–8, 90–1 103, 104–5, 108, 111–12, 113, 115,
simile, 127–8 133, 143, 148, 151
simple aspect, 66–8, 69, 73, 74–5 utterance meaning, 5–6, 16, 101
size (cardinality), 90–1
spatial terms, 8, 37–8 vehicle, 123–5
speaker, 2, 3, 4–5, 9, 16, 38, 101–3, 104, verbs, 26, 29, 44, 49, 55–64, 67, 70, 77, 83,
105–6, 108, 111–12, 115, 128–9, 109, 116, 140, 141, 150
133–4, 143, 149, 151–2, 157–8 ditransitive verb, 56, 62
speaker meaning, 5 intransitive verb, 56–7
speech acts, 17, 148–58 transitive verb, 56
statement, 17, 148, 152 unaccusative verb, 56–7, 60–1
superlative, 46–7 unergative verb, 56–7
superordinate, 27–8, 36–7, 121, 123–4
synonymy, 21–3, 24, 25, 26 wh-clause, 137–8
syntax, 20, 61–2, 63–4, 84, 89, 97–8, wh-question, 150
118, 140–1, 142, 152–4 will, 67, 69, 70, 72, 74–5, 83–4
syntactic converses, 141 Wittgenstein, Ludwig, 34
word meaning, 7, 10, 29, 34, 122
tag-question, 150 WordNet, 29, 31
tenor, 123 world knowledge see encyclopedic
tense, 10, 59, 66–73, 74, 75, 79, 83–4 knowledge
historic present tense, 71, 73
perfect tense, 79 zero derivation, 61

Patrick Griffiths (Editor) - Chris Cummins (Editor) - An Introduction To English Semantics and Pragmatics-Edinburgh University Press (2016)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Patrick Griffiths (Editor) - Chris Cummins (Editor) - An Introduction To English Semantics and Pragmatics-Edinburgh University Press (2016)

Uploaded by

Copyright:

Available Formats

An Introduction to English Semantics and

titles in the series include:

Visit the Edinburgh Textbooks on the English Language website at www​

Cover design & illustration: riverdesignbooks.com

Edinburgh University Press Ltd

Typeset in 10.5/12 Janson by

ISBN 978 1 3995 0460 7 (hardback)

List of figures and tablesviii

9 Figurative language 121

Suggested answers to the exercises161

When Laura Williamson from Edinburgh University Press asked me

To illustrate some of the differences between semantics and prag-

In (1.1), that denotes something that is assumed to be obvious to the

This way of looking at communication has several important

1.1 Sentences and utterances

The abstract linguistic object on which an utterance is based is a

1.2 Types of meaning

From the above examples, we might conclude that finishing is a

1.2.1 Denotation, sense, reference and deixis

nouns, like honey, denote substances; singular names denote individuals;

• time:  now, soon, recently, ago, tomorrow, next week . . .

1.3 Semantics vs pragmatics

1.3.1 A first outline of semantics

1.3.2 A first outline of pragmatics

6. Example (1.6a) is potentially problematic as an utterance because,

2.1 Propositions and entailment

The number of sentences in a human language is potentially infinite, so

and is one of the most fundamental concepts in logic: arguments are

If (2.3a) is understood and accepted as true, then none of its entail-

­ ultiplication, but negation (or reversal) performed by the prefix un-

At the same time, our language also contains expressions which

2.4 Complementarity, antonymy, converseness and incompatibility

These pairs are complementary terms – so called because the comple-

(2.16) a. France is bigger than Germany.

a house. This pattern of entailment tells us that we are dealing with

2.5.1 Hierarchies of hyponyms

thing superordinate of structure, building, house, . . .

structure hyponym of thing, superordinate of building, house, . . .

building hyponym of structure, thing, superordinate of house, . . .

house hyponym of building, structure, thing, superordinate of . . .

Figure 2.1 An example of levels of hyponymy

thing ‘physical entity’

Figure 2.2 Hyponym senses get successively more detailed as we go down

superordinate: it’s an instance of the superordinate that has certain prop-

(2.19) a. Platypuses lay eggs.

thing time person1 idea

place structure product plant animal1 student

dam building vehicle tool animal2 person2 freshman (post)

barn house garden tool kitchen utensil

Hyponym hierarchies exist for other parts of speech, such as verbs

the ­definition of hyponymy we have been working with here. Indeed, if

(a) The students liked the course.

5. Provide example sentences and write out a pattern of entailments

Recommendations for reading

3.1 The has-relation

That figure is a square ⇒ ‘That figure has four sides of equal

Mathematical terms are somewhat atypical of natural language because

of influential thinkers about language, so I want to dwell on it for a

3.1.1 Inferring existence from the has-relation

In (3.4), we have an indefinite article, a, and a definite article, the. A

3.1.2 Hyponymy, prototypes and the has-relation

3.1.3 Parts can have parts

3.1.4 Spatial parts

Visit the Edinburgh Textbooks on the English Language website at www

List of figures and tablesviii

9 Figurative language 121

Suggested answers to the exercises161

• time: now, soon, recently, ago, tomorrow, next week . . .

ultiplication, but negation (or reversal) performed by the prefix un-

the definition of hyponymy we have been working with here. Indeed, if

(4.9) a. Turkish authorities have cracked down on alleged sup-

e xpression with the same form, we sometimes call this conversion or