An Introduction to

Discourse Analysis

Compiled and Edited

By

Dr. Samira Mahmoud El Sherbeny

2025
 Preface

Discourse analysis is a way of studying language in use.

It focuses on how people communicate in real-life situations,

whether through spoken conversations, written texts, or even

visual media. Unlike traditional grammar, which looks at

sentences in isolation, discourse analysis examines how

language works in larger contexts. It asks questions like: How

do people use language to achieve their goals? How does

language reflect social relationships, power dynamics, or

cultural norms? And how do we interpret meaning beyond the

words themselves?

At its core, discourse analysis is about

understanding meaning in context. For example, if someone

says, “It’s cold in here,” they might simply be stating a fact

about the temperature. But in another context, the same

sentence could be a polite request to close a window or turn up

the heat. Discourse analysis helps us uncover these hidden

layers of meaning by looking at the situation, the relationship

between the speakers, and the cultural or social rules that shape

communication.

Why do we Study Discourse Analysis?

Language is one of the most powerful tools we have as humans.

It allows us to share ideas, build relationships, and influence

others. But language is not neutral—it is shaped by the world

around us, and it also shapes the world in return. For instance,

the way politicians speak can influence public opinion, and the

way we talk about social issues (like gender or race) can

reinforce or challenge stereotypes. Discourse analysis gives us

the tools to critically examine these processes and understand

how language shapes our lives.

Key Concepts in Discourse Analysis

1. Discourse: This refers to any form of communication,

whether spoken or written. It can also include non-

verbal elements like gestures, images, or even silence.

Discourse is not just about individual words or

sentences but about how they come together to create

meaning.

2. Context: Context is crucial in discourse analysis. It

includes the setting, the participants, their relationships,

and the cultural or social norms that influence

communication. For example, the way you speak to a

friend is likely very different from the way you speak

to a teacher or a boss.

3. Power and Ideology: Discourse analysis often looks at

how language can be used to maintain or challenge

power structures. For example, certain groups may

have more authority to speak in a given situation, and

 their language may reflect or reinforce their position of

power.

4. Cohesion and Coherence: These terms refer to how a

text or conversation holds together. Cohesion is about

the linguistic connections (like pronouns or

conjunctions), while coherence is about the overall

meaning and logic.

How to Do Discourse Analysis

Discourse analysis involves carefully examining texts or

conversations to identify patterns, themes, and underlying

meanings. Here are some common steps:

1. Choose a text or conversation to analyze.

2. Look for patterns in language use, such as repeated

words, phrases, or structures.

3. Consider the context: Who is speaking? What is their

relationship? What is the purpose of the

communication?

4. Interpret the meaning: What is being communicated

explicitly, and what is implied?

5. Reflect on the broader social or cultural implications of

the discourse.

Discourse analysis is a fascinating and practical way to

study language. It helps us understand how

communication works, how meaning is created, and

how language shapes our world. By learning to analyze

discourse, you’ll develop critical thinking skills and a

deeper appreciation for the power of language. As you

read this book, you’ll explore these ideas further and

learn how to apply them to real-life examples. This

book brings together essential readings to offer a well-

rounded introduction to discourse analysis. It draws on

two foundational chapters from Introducing Discourse

Analysis by David Nunan (1993), offering foundational

perspectives on the subject, complemented by twelve

comprehensive chapters from Discourse Analysis : A

Practical Introduction by Canning &Walker (2024),

which bring contemporary approaches and practical

applications. Together, these selections create a

cohesive resource that bridges classic and current

perspectives in the field.

Dr. Samira Mahmoud

 Contents

1 Discourse: language, context, and choice 1

Introduction 1
What is discourse? 1
The nuts and bolts of language and discourse 3
Text 13
Understanding context 16
Spoken and written discourse 20
Standard English (or the issue of convention) 22
Sociolinguistic variables 24
Conclusion 26
Further reading 27
Resources 28
2 Organising discourse: thematic and information structure 31
Introduction 31
Organising discourse 31
Structure of the English clause 32
Thematic structure 38
Information structure: Given and New 48
Given injustice: the case of Derek Bentley 53
Conclusion 56
Further reading 56
Answers to activities 56
3 Organising information in discourse: cohesion 62
Introduction 62
CONTENTS

Coherence and Cohesion in discourse 62

Reference 63

Ellipsis and substitution 68

Conjunction 70

Reiteration (lexical cohesion) 76

vii
 Conclusion 79

Further reading 80

Answers to activities 80

4 Analysing spoken discourse 84

Introduction 84

How spoken discourse is analysed 84

Spoken Interactions 85

Prosody in spoken discourse 87

Turns, turn taking, and turn transition 93

Functional analysis of turns 96

Sequencing 99

Backchannels 104

Overlapping talk 105

Summary of transcription conventions 105

Conclusion 107

Further reading 107

Resources 107

Answers to activities 108

5 Analysing meaning in discourse 113

Introduction 113

What do we mean by meaning? 113

Working out meaning in discourse: co-text and context 121

Entailment 122

Presupposition 125

Conclusion 132

Further reading 132

Answers to activities 133

6 Meaning and context 137

Introduction 137

Implicature 140

Doing implicatures in real discourse: memes 149

Conclusion 153

Further reading 154

7 Politeness 158

Introduction 158

Face 160

viii
CONTENTS
Using politeness strategies to mitigate face-threatening acts 164

Non-linguistic considerations 170

A short analysis of a football press conference 172

Conclusion 174

Further reading 175

Answers to activities 175

8 Metaphorical meanings in discourse: metaphor and metonymy 179

Introduction 179

What is a metaphor? 179

The conceptual basis of metaphors 181

Novel metaphors 186

Extended metaphors 191

Metaphors in political discourse 194

Metonymy 197

Metaphors and metonymies 200

Conclusion 201

Further reading 202

ix
 Answers to activities 202

9 Representing experience in discourse 209

Introduction 209

Different ways of telling 210

The transitivity model 214

The discourse situation 221

Case study: a case of domestic violence 223

Conclusion 231

Further reading 231

Answers to activities 232

10 Presenting other people’s speech, writing, and thought 235

Introduction 235

What is discourse presentation? 236

Different types of discourse presentation 240

Attribution of source of original discourse 252

Faithfulness 253

Conclusion 255

Further reading 256

Answers to activities 256

11 Corpus linguistics and discourse analysis 263

Introduction 263

Corpus linguistics 263

Corpus linguistics and discourse analysis 264

Using corpus methods to analyse corpora 267

CONTENTS

Conclusion 293

Further reading 294

 Corpus tools 294

Corpora 295

Answers to activities 296

12 Doing a project in discourse analysis 301

Introduction 301

Systematicity and the three Rs of research 301

Ethics 302

Copyright 308

Developing a research project 309

Data 314

Writing up your research – doing academic discourse 315

Conclusion 317

References

xi
 1 Discourse
Language, context, and choice

Introduction
In this frst chapter we explain what we mean by discourse and discourse analysis and
introduce some of the key concepts and linguistic terminology that we will use
throughout this book. We will discuss the notions of text, context, and co-text, before
going on to explore the differences between spoken and written discourse. We will
also examine the idea of a standard language and that some language varieties hold
more prestige than others. We will discover that when analysing discourse, analysts
consider the form of language (see levels of language in Figure 1.2), its function (e.g.
the purpose to which it is put; how it works to achieve certain goals), and the context
in which the language event occurs (e.g. a conversation between friends; a political
debate, an opinion piece in the press). Our starting point, perhaps unsurprisingly, is
‘discourse’.

What is discourse?
Discourse does not have one single defnition and has different meanings even within
linguistics. According to the Oxford English Dictionary online edition (‘Discourse’
1989), discourse can mean:

• a detailed and lengthy spoken or written discussion of a particular topic;

• spoken communication, interaction or conversation;
• a connected series of utterances by which meaning is communicated.

Although non-technical, these defnitions nonetheless provide important information

about what discourse is. Discourse is connected chunks of spoken or written language
(e.g. utterances; sentences) used in interactions for meaningful communication.
Discourse, then, is language being used in all its forms (including signed languages)
to communicate, interact, inform, and get things done. Simply put, and to quote two
pioneers of discourse analysis, discourse is “language in use” (Brown and Yule 1983:
1). Consequently, discourse analysis is not the study of linguistic forms in isolation;
it is, as Brown and Yule explain, the study of how linguistic forms function when
they are used in different contexts. Simpson and

DOI: 10.4324/9781003351207-1
Mayr (2009: 5) echo this distinction between linguistic forms and the function of
forms in use when they contrast language (as a system) with discourse (as language
in use):
1
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

Whereas language refers to the more abstract set of patterns and rules which
operate simultaneously at different levels in the systems [. . .] discourse refers to
the instantiation of these patterns in real contexts of use.
Importantly, they go on to say that:
discourse works above the level of grammar and semantics to capture what happens
when these language forms are played out in different social, political and cultural
arenas.
The defnition above chimes with that of another discourse pioneer, Mike Stubbs, who
described discourse as “language above the sentence or above the clause” (Stubbs
1983:1). Therefore, while discourse is, of course, made up of the building blocks of
language, it is greater than the sum of its parts. It is what results when these language
forms combine and connect in different ways in different contexts.
Figure 1.1 captures the preoccupations of language and discourse and their
relationship to each other. Discourse concerns all of that which pertains to language
(such as syntax, lexis, and morphology – see Figure 1.2) but, in addition, it involves
the context in which language is used (we will say more about context

Figure 1.1 Relationship between language and discourse

later in this chapter), the people using the language, and the purpose served by the
language in that context. Discourse, then, is structural because it involves the
linguistic building blocks set out in Figure 1.2, cognitive because it incorporates the
mental representations language users hold about the world (we explore this further
in Chapter 8), and social because “language users engaging in discourse accomplish
social acts and participate in social interactions [that are] embedded in social and
cultural contexts” (van Dijk 1997: 2, original emphasis) (we say more about acts in
Chapter 4).
2
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

What is discourse analysis?

As we established in the previous section, discourse analysis is not the study of
language rules and components in isolation; it is not just about considering clauses
and other structures and establishing what words or phrases go where. Instead,
discourse analysis studies how language is used in real-life, everyday settings. When
we ‘do’ discourse analysis we are looking at how meaning is conveyed between those
producing the language and those receiving the language. However, that is not to say
an understanding of the language system is not essential for analysing discourse,
because it is! Discourse analysis concerns analysing language forms and appreciating
their function in the context in which they occur. Moreover, it involves investigating
whether forms combine to create larger units of language, whether these have their
own structural patterns, and whether any such patterns relate to meaning. In short,
discourse analysis examines how meanings are made and interpreted through
linguistic and nonlinguistic behaviour in a given situational context. As you might
imagine, given the almost endless number of different situational contexts, this makes
discourse analysis a broad area of study. Indeed, in the preface to his 1997 edited
volume, Discourse as Social Interaction, Teun van Dijk acknowledges that given the
vast number of discourse genres (e.g. argumentation, storytelling), modes (e.g.
spoken, written, imagistic), and social domains (e.g. medical, legal, political), the
remit for discourse analysts is so wide that “even two volumes [of his edited
collection] are unable to cover everything” (xi). Twenty-fve years on, the information
and communication technology revolution has increased the scope of ‘everything’ still
further, with online interactions and social media now a commonplace way of ‘doing’
discourse.

The nuts and bolts of language and discourse

In this section, we introduce some of the ‘nuts and bolts’ of a language. This is because
for us to be able to analyse discourse in linguistic detail, we need to know how
language is constructed. This section will also provide a vocabulary that will enable
us to describe language and discourse. Language operates on several levels from the
smallest units through to discourse. Figure 1.2 below shows these structural levels,
which we describe below.

3
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

Figure 1.2 The levels of language

Morphology
Morphology is the study of the smallest units of meaning in language, known as
morphemes. Words comprise one or more morphemes. For example, ‘books’,
contains the morphemes ‘book’ and ‘s’ (see Figure 1.3). The morpheme ‘book’ is
what’s known as a free morpheme because it can stand on its own as a word. The
morpheme ‘s’ that is attached to ‘book’ is known as a bound morpheme because
even though it carries meaning (in this case, it means ‘plural’) it cannot stand alone
and have meaning; it must be bound to another morpheme. For example, we would
not say ‘s’ to answer the question
‘what are libraries full of?’

Figure 1.3 Free and bound morphemes in ‘books’

The addition of the bound morpheme ‘s’ to the free morpheme ‘book’ is an example
of affxation. In the example in Figure 1.3, the morpheme ‘s’ is a suffx because it
attaches to the end of ‘book’. This specifc type of suffx is known as an infectional
suffx because it carries grammatical information (in this case ‘number’). Infectional
suffxes can also signal tense, possession, or comparison (e.g. ‘-est’). Bound
morphemes can also function as derivational suffxes. For example, the verb
‘assassinate’ is created by adding the bound morpheme ‘-ate’ to the free morpheme
‘assassin’ (a noun). When the suffx is added, there is a change in grammatical class
from noun to a verb, so a new verb (‘assassinate’) is derived from a noun (‘assassin’).
Some bound morphemes attach to the front of a free morpheme and are known as
prefxes. These carry a variety of meanings but nonetheless cannot stand on their own

4
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

as words. For example, the bound morpheme, and prefx, ‘dis’ means ‘not’ or ‘the
opposite of’ so when attached to, say, ‘respect’, a new word is derived (‘disrespect’)
that means the opposite.
Morphological rules can be manipulated in discourse for a range of different
effects. For example, the poetry of E. E. Cummings often plays with morphology in
creative ways. For instance, in the poem ‘Love is more thicker than forget’,1
Cummings describes love as ‘moonly’ and ‘sunly’ which are unusual constructions
that, through the addition of the suffx ‘-ly’, change the word class of ‘sun’ and ‘moon’
from nouns to adjectives (you can fnd the full poem at the Poetry Foundation website).
New words can also be formed by the
merging of two free morphemes to form QR 1.1 Link to E. E. Cummings’s poem
compound words. For example, ‘bookworm’
is a combination of ‘book’ + ‘worm’. In 2020, the Oxford Dictionary’s ‘word of the
year’ was expanded to account for an “unprecedented” year and introduced new
compounds that included ‘bushfres’, ‘Covid-19’, ‘lockdown’, ‘circuit-breaker’,
‘support bubble’, and ‘keyworker’. As you can see, the words of the year can tell us
much about how we use existing language for new concepts, but it can also tell us
about the events of that year!

Phonology
Phonology is the study of the sound system of a language and is concerned with the
different sounds that carry meaning, known as phonemes. A phoneme is a distinctive
sound in any language that, when uttered, makes a difference to meaning and therefore
contrasts with other sounds. For example, the vowel sound in the word ‘pip’ is
different to the vowel sound in ‘pup’; indeed, it is the only difference between those
two words, and it is that difference that affects the meaning of the words. In
phonology, ‘pip’ and ‘pup’ are an example of a minimal pair, which is a pair of words
that differ in one sound only. Therefore, ‘pip’ and ‘pat’ are not a minimal pair because
they differ by two sounds. The idea of a minimal pair is to contrast particular sounds
to show that they make a difference to meaning. Phonologists aim to identify and
quantify the phonemes that comprise a language. For example, with Received
Pronunciation of English (see below), there are said to be 20 vowel sounds
(Cruttenden 2001: 91) and 24 consonants (Cruttenden 2001: 149). There are,
therefore, many more sounds in spoken English than there are letters in written
English, which has just fve vowels (aeiou) and 21 consonants
(bcdfghjklmnpqrstvwxyz). Consequently, phonologists use an expanded set of
symbols (including the letters we know and recognise) to represent the sounds of a
language which together make up the International Phonetic Alphabet (IPA). You can
fnd a copy of the IPA at the start of this book.
Table 1.1 presents a series of words and aims to demonstrate the 20 different vowel
sounds in English (assuming Received Pronunciation!) along with the IPA symbol
that represents that sound. The IPA symbol is placed between slashes, which is the
convention for phonemic transcription.
5
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

Table 1.1 The 20 vowel sounds in standard pronunciation (RP) of British English

pap /æ/ Parp /ɑː/ poser /ə/ poise /ɔɪ/

pep /e/ Perp /ɜː/ pun /ʌ/ pope /əʊ/
pip /ɪ/ Peep /iː/ pain /eɪ/ pow /aʊ/
pop /ɒ/ Paw /ɔː/ peer /ɪə/ pair /eə/
pup /ʊ/ Poop /uː/ pipe /aɪ/ poor /ʊə/

When we talk about speaking ‘standard’ English, we mean the variety of the
language that is conventionally accepted as the ‘norm’. In the IPA, the ‘norm’ is
Received Pronunciation of British English or ‘RP’ for short. The ‘received’ in RP
means ‘accepted’ or ‘approved’ and it is therefore the version of spoken English
approved by arbiters of the language (in this case, those policing it in the late 1800s).
RP then is the point of reference or model that the IPA is based on. This might seem
odd given that very few people actually use RP (around 3 per cent of the UK
population). In other words, RP’s approval or acceptance is more to do with perceived
social status than correctness (we say more about this later). Use QR code 1.2 to fnd
out more about RP at the British Library website and hear what it sounds like (think
1950s English TV/radio announcer at the BBC and
QR 1.2 Received Pronunciation you are there).
Of course, English is spoken in many different
regional accents across the world and different accents have a slightly different
inventory of phonemes (due to differences in pronunciation) and differ in which
phonemes contrast in meaning. For example, with some accents it is doubtful that the
difference between /ʌ/ and /ʊ/ is meaningful – e.g. in parts of the North of England,
/pʊb/ and /pʌb/ are both places to buy and consume beer (among other things).
According to Crystal (1995: 239), the most frequent vowel sound in English is /ə/,
which is known as schwa and represents a sort of short ‘uh’ sound. The frst 12 vowel
sounds in Table 1.1 are known as monophthongs (or pure vowels) because their sound
remains fairly constant when they are spoken. The remaining eight vowels are
diphthongs because there is perceivable movement (known as a glide) between two
different sounds. It is also possible to have triphthongs (e.g. words such as ‘power’,
‘prior’ and ‘player’). Such triphthongs involve a diphthong with the addition of a
schwa (/ə/) at the end. For example, ‘power’ is pronounced /paʊə/ in standard (RP)
English.
In Hiberno-English such as that spoken in the north of Ireland, some triphthongs
are spoken as monophthongs. For example, ‘power’ is often pronounced in Belfast as
‘par’ /pɑ:r/. Irish humour can often be self-refexive and has given rise to many books
on what is affectionately called the language of ‘Norn Iron’, itself a non-standard
phonological rendering of standard ‘Northern Ireland’. You cannot go far in Belfast
without seeing some form of this rich Hiberno-English variant marketed as T-shirts,
6
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

mugs, and more, as the following example in Figure 1.4 from T-shirt retailer Norn
Iron Tees shows.

Figure 1.4 An example of Hiberno-English: ‘power shower’

Lexis
This is the linguistic term given to the words (or vocabulary) of a language and their
different forms. Words can be divided into two general types: lexical words that refer
to things in the world (ideas, concepts, entities); and function words (also known as
grammatical words) that help to link the lexical words together to make clauses and
sentences (Freeborn 1995: 36). Lexical words are known as open class, because they
are being added to all the time as new words are coined to encode new experiences
and new ‘things’. Function words are known as closed class because they are static
(but not totally fxed – consider the more recent introduction of ‘Mx’ as a substitute
for the more conventional ‘Ms’ pronoun for women). Words are traditionally assigned
to what are known as word classes (also known as Parts of Speech) based on what
task they are performing in the text they occur. The different word classes with some
examples are shown in Tables 1.2 and 1.3. There are four open word classes: verbs,
nouns, adjectives, and adverbs, and six closed word classes: prepositions, pronouns,
determiners, demonstratives, conjunctions, and modal verbs. Note that the same word
can perform a different role in different texts and so can be assigned to different word
classes. For example, ‘fower’ can be a noun (as in ‘what a lovely fower’), but if you
ask,‘has your agapanthus started to fower yet?’, then ‘fower’ is doing the job of a
verb.

Table 1.2 Lexical word classes

7
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

Nouns thinker, book, worm, shelf, case . . .

Verbs think, book, saw . . .
Adjectives sunny, bookish, booky, quick . . .
Adverbs gingerly, bookishly, hurtfully, quickly . . .

Table 1.3 Function or grammatical word classes

Pronouns her, they, it, we, them, his, Mr, Mx . . .

Prepositions in, at, above, on, beside . . .
Determiners a, an, the, some, any, all . . .
Conjunctions for, and, but, so . . .
Demonstratives this, that, those, them . . .
Modal verbs should, shall, would, could, can, may, might, must, ought

Lexical creativity
New concepts require new words, known as neologisms. Neologisms are typically
achieved by compounding and blending existing words or by novel affxation. For
example, the new word ‘staycation’ (meaning to go on holiday without going abroad)
is a blend of ‘stay’ and ‘vacation’; ‘crowdfunding’ is a compound of ‘crowd’ and
‘funding’; and ‘metaverse’ (meaning a virtual meeting space), is formed by replacing
the prefx ‘uni’ (meaning ‘one’) in ‘universe’ with ‘meta’ (meaning ‘beyond’ or
‘higher order’2). An alternative to neologising is to give existing words new, additional
meanings. It was not so long ago that ‘mouse’ and ‘virus’ referred only to living
organisms, yet now they refer to inanimate objects or concepts as well. If we said to
you ‘We cannot get our mouse to work’, we doubt you would think we were exploiting
our pet rodent for material gain.
Poets and writers are often creative with language and will neologise willynilly to
suit their needs. An extreme example is the poem ‘Jabberwocky’ by Lewis Carroll, in
which there are so many neologisms that the poem seems totally nonsensical (at frst).
However, linguistic conventions are being adhered to, particularly word-class
conventions. In Activity 1.1 below, the frst activity in this book, see if you can tell
what word classes (noun, verb, adjective, adverb) the neologisms belong to. What do
you base your guesses on? How does what you know about language help you to
interpret the poem?

8
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

Activity 1.1 Making sense of neologisms

‘Jabberwocky’, by Lewis Carroll

’Twas brillig, and the slithy toves Did

gyre and gimble in the wabe: All
mimsy were the borogoves, And the
mome raths outgrabe.

“Beware the Jabberwock, my son!

The jaws that bite, the claws that catch!
Beware the Jubjub bird, and shun The
frumious Bandersnatch!”

He took his vorpal sword in hand;

Long time the manxome foe he sought— So
rested he by the Tumtum tree And stood
awhile in thought.

And, as in uffsh thought he stood,

The Jabberwock, with eyes of fame, Came
whiffing through the tulgey wood, And
burbled as it came!

One, two! One, two! And through and through The

vorpal blade went snicker-snack!
He left it dead, and with its head He
went galumphing back.

“And hast thou slain the Jabberwock?

Come to my arms, my beamish boy!

O frabjous day! Callooh! Callay!” He

chortled in his joy.

’Twas brillig, and the slithy toves Did

gyre and gimble in the wabe: All
mimsy were the borogoves, And the
mome raths outgrabe.

There are linguistic patterns in ‘Jabberwocky’ that indicate that the conventions of
English are being followed. Your knowledge of how sentences are constructed
(syntax) might help you to interpret ‘toves’ as a noun (and perhaps a material thing),
whereas your knowledge of morphology might lead you to conclude that ‘slithy’ is an
adjective that is constructed from ‘slith+y’ (much in the same way that ‘curl’ becomes
‘curly’). To take another example, you might have never heard of ‘The frumious
9
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

Bandersnatch’ but it will not take much effort to discern that ‘frumious’ is an adjective
that describes the noun following it, and that the noun is in fact a noun because it
follows a determiner (the defnite article ‘the’). You might also have deduced that the
‘Bandersnatch’ is a specifc name/has a specifc referent as it is capitalised (and so
graphologically marked – we introduce graphology below). These are just some of the
consistent patterns in this neologistically rich text.
One neologism that gained traction in 2017 was the word ‘covfefe’, used in a viral
tweet on 17 May 2017 by the then president of the United States, Donald Trump. The
full tweet read “Despite the constant negative press covfefe”. Widely acknowledged
as a typo for ‘coverage’, Trump refused to confrm his error and instead deleted the
tweet. When probed by reporters about the nonsensical reference, Trump’s
spokesperson, Sean Spicer, replied “I think the president and a small group of people
know exactly what he meant” (Estepa 2017). Sometimes discourse communities are
*really* small, it seems.

Graphology
Written language exists as marks on a page (or some other medium). This is known
as graphology and refers specifcally to such things as typography, punctuation, and
the arrangement of any marks that constitute the discourse. The word *really* in the
preceding paragraph is graphologically marked as it is enclosed within asterisks. Some
fonts, for example, carry meaning and/or are associated with particular discourse
types. For instance, this book is written using Times New Roman font because this
is seen as a ‘serious’ font that is suitable for this sort of text. The Comic Sans MS
font, however, with its comic book associations would probably be seen as not suitable
for a serious academic book. In some discourses, such as social media and other
computer-mediated communication, upper-case letters can be meaningful where they
communicate SHOUTING and/or anger. When we refer to the way the text is arranged
visually, we are also talking about graphology, and this includes all the meaningful
elements of that text (images, colour, space, etc.). The poster in Figure 1.5 is an artistic
representation of one that appeared on a few online vegan sites in 2015.

10
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

Figure 1.5 Vegan propaganda poster

Writing about the projected equivalence between racism and ‘speciesism’, Twitter
user and blogger, Claire Heuchan, shares the poster (citing @veganoso Twitter
account as the source) in her article condemning the use of “slavery as a tool to
promote vegan values” (2015). Heuchan argued that such comparisons were akin to
“vegan activists mak[ing] clear that vegan spaces are frequently racist spaces”. The
poster uses graphology to promote a relationship of semantic equivalence between
two disparate practices to persuade non-vegans to go vegan or in Heuchan’s terms, to
“trigger a dietary epiphany”. The division of the space into two equal parts sets up a
parallel between the two images, the two dates, and the two concepts ‘racism’ and
‘speciesism’. This nudges the reader to tease out a connection between them
(regardless of whether or not it exists in reality). By making the perceived connection
implicit through the graphological arrangement, the text producer relies on the reader
picking up the inferential connection. After all, we are more likely to be persuaded by
an argument we have helped construct than one that we have invested no time in
formulating.
11
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

Syntax
This is the linguistic term for clause structure. A clause is an organisational unit of
language that is made up of a subject and predicator. ‘We write’ is an example of a
clause. The linear line that comprises most European modern languages, as well
Southeast Asian, Indian, North American, and South American, is written from left to
right, syntactically speaking. Syntax refers to the different structural slots occupied by
language tokens such as nouns, adjectives, verbs, adverbs, and grammatical words
like conjunctions and so on. Conventionally, these syntactic slots follow a
conventional order, so a clause like ‘the dog ate the bone’ follows the conventional
Subject-Predicator-Object pattern in English. Sometimes, the syntactic order
switches, and we get constructions like that in Figure 1.6 ‘the bone, the dog ate’.

Figure 1.6 Conventional and unconventional syntax

Both constructions in Figure 1.6 mean the same thing. However, when we
encounter construction (b), it seems to communicate more than simply its
propositional meaning, which is the basic meaning of the clause relating to the
entities involved and their relationship to each other (see Chapter 5). Indeed, it implies
that the addressee already knows that the dog ate something, just not what it was. We
explore syntactic ordering and its effects in Chapter 2.

Semantics
Semantics is the study of meaning. An important concept in semantics is denotation,
or the thing, idea, action or concept that a word refers to. The denotive meaning of a
word is sometimes called its primary, core or literal meaning. A word may have a set
of possible denotive meanings one of which may be triggered by the surrounding co-
text and/or the context in which a word is produced. So, returning to our sentence, ‘the
dog ate the bone’, ‘dog’ refers to a domesticated canine animal, ‘bone’ refers to part
of an animal skeleton and ‘ate’ refers ingesting (chewing and swallowing) food via
the mouth. A word can also come with associated meanings, which are to do with the
feelings or psychological connections a word (or the concept it refers to) evokes. For
example, ‘dog’ is sometimes associated with concepts such as loyalty and devotion or
companionship. Another aspect of meaning is literal and non-literal (fgurative)
meanings a word can have. For example, a person can be ‘dogged’ by bad luck, and
you might need to ‘bone up’ before an exam. With these uses, ‘dog’ and ‘bone’ no
longer refer to their primary meanings but to metaphorical meanings. In Chapter 5,
we talk more about semantics including presuppositions and entailments, and in
Chapter 8 we deal with fgurative meaning.
12
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

Discourse
This brings us back to where we started this chapter – discourse. We defned discourse
as ‘language in use’. It concerns meaning above the level of the clause and sentence
(syntax). This sentence is a piece of discourse. This book is a bigger piece of discourse
characterised by academic writing, with an instructional purpose, with a high
percentage of domain-specifc lexis (the phrase ‘domain-specifc lexis’ being a case in
point). In fact, we are actively trying to avoid an overly academic register, and so we
hope that this chapter and the rest of this book is NOT characterised by long,
embedded sentences, passive structures, thirdperson pronouns, and lots of what are
known as ‘logical operators’ (Sinclair 2004: 7), which include, for example,
‘therefore’, ‘so’, and ‘consequently’. (Okay, apart from that overly long embedded
sentence.) Our purpose is to inform, not impress. We failed at the latter long ago.

Text
When we analyse discourse, we inevitably study texts because, as Halliday and
Matthiessen (2014: 3) point out, “[w]hen people speak or write, they produce text”.
Halliday and Matthiessen (2014: 3) go on to defne text as “any instance of language,
in any medium, that makes sense to someone who knows the language”. Text, then, is
a single language artefact or “unit of language” that can be spoken, written, signed or
otherwise (e.g. image) that is defned by “meaning” (i.e. it makes sense to someone)
rather than “form” (i.e. there are no formal restrictions on what counts as a text)
(Halliday and Hasan 1976: 1–2). Texts in written form are the typical object of
analysis for discourse analysts, which means to study and analyse spoken and signed
texts, we must record them in some way and create written transcriptions to allow
repeated scrutiny. Audio and video recordings can also be analysed using special
software, such as Praat (Boersma and Weenink 2023; ‘praat’ means ‘speak’ in
Dutch), but this is not something we will deal with in this book. Text, then, is the
object of study but, as Bloor and Bloor (2013: 8) point out, as discourse analysts we
must also appreciate that language is used within a particular context and texts emerge
from language events.
Language event is a general term often used (but less often defned) by linguists
(‘speech event’ and ‘communicative event’ are also sometimes used). A language
event is any event where language has a fundamental role or any event where language
happens. For example, chatting to a friend is a language event; it is diffcult to chat
without using language (in whatever form it might take). Giving a speech is a language
event because it would be very diffcult to give a speech without using any language.
Going for a run, however, is not a language event

13
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

because it does not require language, just working legs, ftness, and a certain amount
of willpower. Language events might well occur during a run though. If you run with
a friend, you might chat (in which case, kudos), or if you get lost you might stop to
ask someone for directions.
A discourse can involve one or more texts and, importantly, other factors including
the producer and intended receiver of the text, the location and date of production, and
purpose. Essentially, discourse is text and context. When we do discourse analysis, we
combine detailed linguistic analyses of formal features of text(s) with an examination
of context. By considering both language and its context of use we can gain a greater
understanding of how meaning is communicated in everyday settings. To illustrate
this point, try Activity 1.2.

14
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

‘HAVE YOU SHEETED?’ – discussion

This text forms part of a discourse. Indeed, to understand the text, we need to know
about the discourse it is part of. Moreover, we need to understand something about
the context in which the discourse occurs (we discuss context in more detail later in
this chapter). This text is printed on a sign at the exit of a quarry in the north of
England. The sign, which is made up of white words on a blue background with a
white border, contains a single sentence in the form of an interrogative. It is asking
whether the intended recipient of the text has done something. However, the text might
not make much sense, or communicate very much to readers who are not the intended
recipients. It is probably the word ‘sheeted’ that is hard to understand because it might
not be clear what action it refers to. The sign uses the word ‘sheet’ as a verb, but what
is ‘to sheet’ and what does it mean in this context? Quite simply, it means to cover the
back of tipper-trucks that leave the quarry fully loaded with gravel with a tarpaulin
sheet.3 The sheet is stretched taut over the back of the trucks to keep the gravel in the
hopper. Sheeting is therefore an important safety measure that all truck drivers should
do before they drive out of the quarry and on to the public highway.
Notice that the text is an interrogative and as such it asks a question probably on
the behalf of the quarry owners. However, the question is not intended to elicit a verbal
response from the truck driver. That is, the truck driver is not meant to go and fnd the
quarry owner and reply: ‘yes, I have sheeted’/‘no I have not sheeted’. Instead, the
interrogative functions as a reminder for the driver leaving the quarry with a fully
loaded truck to pause and check that their truck is sheeted. The presence of the sign at
the quarry exit might be because in the past some drivers left the quarry without
sheeting and this caused a problem, or that there is a legal obligation on the quarry to
remind drivers to sheet. Whatever the reasons that led to the placing of the sign, it is
a text that is part of a discourse between quarry owner and driver, and to understand
the text we need an appreciation of the context and the place the sign takes in the
discourse.
Discourse communities
The positioning of the text at a quarry exit and its lexically specifc terminology might
have led you to conclude, correctly, that it is directed towards a particular group of
people (quarry workers/truck drivers) who know what it means ‘to sheet’. Such groups
of people are known as a discourse communities (Swales 1988) because members
participate in and communicate through similar textual practices such as specialist
lexis (e.g. ‘sheeting’), and pragmatic meanings (e.g. ‘reviewer 2’ in academia).4
Although some scholars argue that discourse communities share common goals or are
“unifed by a common focus” (Porter 1992), this is slightly misleading because there
are many examples of discourse communities (e.g in academia or government) where
people frequently have different goals and are often not unifed! We can, of course, be
members of many different discourse communities, which means we adopt different
language practices depending on which community we are interacting with (and
therefore part of) at any particular time.

15
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

Meaning potential
The example in Figure 1.7 serves to demonstrate that as discourse analysts, we are
interested in how people use language in different settings to make meaning. Indeed,
meaning-making and meaning potential5 (Allwood 2003) is a topic we will deal with
in some detail over the course of this book. The example helps to show that context
is crucial to the notion of discourse and its analysis. The aim of our next section,
therefore, is to understand better what we mean by context and how we can
incorporate it into discourse analysis.

Understanding context
Language – and, by extension, meaning – does not exist in a vacuum. Language is
used by people in different locations, at different times, involves different participants
and is used for different purposes. In short, language occurs in different contexts.
Context refers to the various situational parameters that coincide with language and
which work together with language to create meaning. Language and context are
inseparable components in meaning making. J. R. Firth (a key fgure in linguistics)
made the following observation about the importance of context: ‘meaning’ is to do
with “the function of a complete locution in the context of situation, or typical context
of situation” (Firth 1935: 72). In other words, the intrinsic meaning of an utterance
(i.e. locution) only means what it means in relation to its context.
Brown and Yule (1983) discuss context at considerable length and in this section,
we summarise some of their main points. Citing Hymes (1962) and Lewis (1972), they
set out several situational parameters that help to delimit context. These parameters
can be thought of as answers to a set of questions that we can ask about the discourse
under analysis.
What is happening?
This question concerns the activity or event that the discourse is part of. The
answer might be, for example, a conversation with a friend, a university lecture, an
advertisement, a political speech, a celebrity interview.
Where is it happening?
The location of the discourse can be an important contextual factor and includes,
for example, the geographic location (e.g. country, town, street), the building or
physical space (e.g. university lecture theatre, conference hall, television studio),
or the virtual space (e.g. Snapchat, TikTok, Instagram).
When is it happening?
This question concerns the time of the discourse: hour, minute, second, year,
month, day.
What is being talked about?

16
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

This question refers to the topic of the discourse: medicine, shampoo, state health
provision, ‘my latest movie’.
What job is the language doing?
The purpose of the discourse: e.g. teaching, informing, selling, persuading,
entertaining.
Who is involved?
The people or participants involved: the addressor/speaker/producer, addressee/
hearer/receiver, and any observers or overhearers (audience). Also, the social
relationship between participants (e.g. friends, family, colleagues), and power
relations (is one participant more institutionally powerful?) e.g. lecturer and
students, advertisers and social media users, politician and delegates, interviewer,
interviewee and audience.
How is it happening?
This question considers the mode of the discourse – whether it is spoken (produced
orally by vocal cords and received aurally by ears), written (produced graphically
and received visually), or signed (produced gesturally and received visually) or a
combination (multimodal). The question also includes the means of conveyance, or
the medium of communication (e.g. phone message, email, webpage, print). There
might not be any media involved, of course, such as with face-to-face
communication.

A further possible consideration relating to how the discourse is happening is

whether spoken or signed discourse is happening spontaneously or involves some
level of planning. For example, political speeches are typically carefully prepared
and written out prior to their delivery whereas naturally occurring conversations
tend to be unplanned and spontaneous.

Examples of how discourse is happening include: face-to-face spoken conversation,

spoken telephone conversation, video lecture via (say) Zoom or Microsoft Teams,
printed handouts, slides projected on-screen, on-screen advert with words and
pictures, speaking from a script, participating in a face-to-face spoken Q&A,
participating in an online written Q&A.
Which language (variant) is being used?
For this aspect of context, Hymes (1962) uses the term code, which includes
language varieties and dialects such as, for example,American English, Haitian
Creole, British Sign Language, Hiberno-English, Yorkshire dialect.

These questions and their answers help to pin down contextual factors that are
potentially important for meaning making. For example, different contextual factors
infuence our language choices including the words we use (lexis), the way we say
them (phonology) and the way we arrange them (syntax). These choices, which are
made both consciously and subconsciously, help to tailor our language according to
what it is being used for, where it is being used, who is involved, and how the

17
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

communication is being mediated. Context is also important because it infuences our

understanding of language.

Expanding context (the role of background knowledge)

Consider this language event based on the experience of one of the authors, whom we
will refer to in the third person as person B.

On Wednesday 18 May 2022, at 06:45, B embarked on a train journey from

Inverness to Edinburgh (both in Scotland). B travelled alone and sat at one end of
a carriage. There were other people in the carriage. In particular, there were four
people travelling together who were sat at the opposite end of carriage being very
loud. Two were wearing bright orange sunhats (which is quite unusual for
Scotland). The conductor of the train entered the carriage to check people’s tickets.
He spoke to the four loud people and then came to the end of the carriage where B
was sat and addressed the people sat there. The conductor said that he had asked
the loud people not to be too loud; to let him (the conductor) know if they got too
rowdy; and that the train was quite empty so people could move to another quieter
carriage if they wished. The conductor fnished by saying: “Oh dear. I feel sorry for
Seville.”
B was confused. What did he mean by that? Clearly, B was lacking essential
information which the conductor seemed to be assuming his hearers shared. B
asked his close friend, Professor Google, for help and soon all became clear. He
simply typed in ‘seville’ on Google search on his telephone, and the frst result was
a news story headline: “Rangers in Seville: Police say 100,000 fans expected”.

All became clear: “Oh dear, I feel sorry for Seville” was said in the context of the
2022 UEFA Europa cup fnal that was to be played in Seville on the evening of
Wednesday 18 May 2022 between German club, Eintracht Frankfurt, and
(importantly) Scottish football club, Rangers. This was a Big Deal for Rangers fans
(apparently) and more generally for Scottish football, and Scotland. B is not a football
fan and is not Scottish – a perfect storm for non-understanding in this context.
However, once B had this important missing piece in the discourse jigsaw, he
understood that the loud people were Rangers fans and the signifcance of their orange
hats (Rangers are strongly associated with the colour orange). Notice though that the
search for meaning had to go beyond the local context (location, time, participants,
mode, purpose) of the train. The sentence was uttered in a national and international
context that relied on shared knowledge outside the carriage of the train. In this
instance, B did not share the knowledge of the football discourse community
ordinarily required to enable him to correctly interpret the reference to Seville.

18
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

Context and relevance

One issue with context that Brown and Yule (1983) highlight is how analysts (and
indeed language users) narrow down the particular contextual factors that are
important for understanding discourse. For relevance theorists (see, for example,
Sperber and Wilson 1986; Clark 2013) this issue is addressed by the notion of – yes,
you’ve guessed it – ‘relevance’ and the crucial role it plays in constraining contextual
factors and analysing discourse. Relevance theory posits that we make sense of
discourse largely through adhering to a principle of local interpretation, which
basically means that we only use what is relevant in the discourse situation to derive
meaning. Where local interpretation is insuffcient (as we saw in the train example
above), we might need to draw on a wider context to fll gaps in local knowledge. So,
we may use our own background knowledge in determining what is meant. We also
draw on our experience of discourse in different situations that develops over time and
use that to delineate basic patterns in discourse, which help us make sense of
interactions.

Co-text
Co-text is a specifc type of context that is not covered by any of the questions set out
above. It refers to the immediate linguistic environment of an utterance and concerns
how the meaning of a particular unit of discourse (word, phrase, sentence, utterance)
is infuenced by the rest of the discourse. Co-text constrains or infuences meaning of
lexical items and is particularly important for the interpretation of utterances in time
and space. The following sentences relate to the football event we mentioned above.
Consider the meanings of the words ‘teams’ and ‘goal’ in the following invented
sentences.

(i) Football fans arrive in Seville to support their teams.

(ii) Seville’s citizen security commissioner, Juan Carlos Castro Estevez, set his teams
the goal of calmly dealing with any problems that might emerge.
(iii) Joe Aribo scored a goal in the 57th minute putting Rangers ahead.
(iv) Rangers did not achieve their goal of winning the Europa League cup.

Hopefully, you will have noticed that ‘teams’ and ‘goal’ refer to different things in
the sentences above. In (i) the word ‘teams’ refers to ‘football teams’, and we reach
that conclusion because of the mention of ‘football fans’ at the start of the sentence.
In (ii) however, ‘teams’ refers to ‘security teams’ rather than ‘football teams’ and we
reach that conclusion because these are the teams of Seville’s security commissioner.
The pronoun ‘his’ in front of ‘teams’ references the security commissioner (see
Chapter 3 for more on reference) and signals belonging. It would make no sense for
multiple football teams to belong to the commissioner; the logical conclusion therefore
is that he has within his control some other sorts of teams relating to security.
Similarly, the word ‘goal’ in (ii) does not refer to getting a ball between two posts;
19
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

rather, it means ‘objectives’ or ‘desired outcomes’. However, in (iii), the co-

occurrence of ‘scored’, ‘57th minute’ and ‘Rangers’ leads us to read ‘goal’ as the
‘getting-the-ball-into-the-net’ type. Additionally, if you know the names of the people
playing for football teams, then ‘Joe Aribo’ might also be a clue. In (iv), however, we
understand ‘goal’ to mean ‘objective’ once again, and this is because of its co-
occurrence with the verb ‘achieve’ and its post-modifcation by the prepositional
phrase ‘of winning the Europa League cup’, which identifes the goal. We can see then
that the words ‘teams’ and ‘goal’ have different meanings in the sentences and co-text
helps us to discern those meanings.

Spoken and written discourse

Discourse is often divided into two modes – either spoken or written – or a
combination of both (i.e. multimodal). The spoken/written division characterises our
ability to physically produce language either orally using our vocal cords6 or, with the
right tools and materials, graphically. With the latter, combinations of symbols
(graphemes) convey meaning by representing the sounds of language visually (a
system referred to as orthography). Signs and symbols convey meaning in language
systems, including pictures. Additionally, we can communicate visually using
gestures and facial expressions, both of which are crucial for sign languages.
The spoken/written division also marks other important differences. For instance,
we learn to speak before we write and, certainly, human beings evolved to speak frst.
We are able to acquire spoken language through exposure to speech, but we must be
instructed on how to read and write. Naturally occurring speech is produced
dynamically and spontaneously and when in conversation, we must listen and
comprehend and plan our next turn. We cannot look back at what was said earlier, we
cannot fick forward to see what is coming up and we cannot backspace or delete
because we produce and transmit simultaneously – the words are out there as soon as
we have said them. Unless recorded in some way, they exist only in the moment they
are produced, disappearing as soon as the air molecules set in motion by our vocal
cords cease to vibrate, surviving only in the (imperfect and limited) memories of the
discourse participants.
Following Miller and Weinert (1998), we can summarise the features of unplanned,
spontaneous spoken language as follows:

• produced in real time, so without any planning and without editing;

• typically face-to face communication, so production and reception synchronous
with interlocutor;
• involves changes in voice pitch, amplitude (loudness), rhythm, speed of delivery,
shortening/lengthening of words and voice quality;
• also involves gestures, gaze, facial expressions, and body posture; •
infuenced by the limitations of our short-term memory.
20
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

(Adapted from Miller and

Weinert 1998: 22–23)

According to Miller and Weinert (1998), these features of speech production set out
above result in:

• information being produced in small chunks;

• low information density; for example, short phrases;
• a limited range of vocabulary and lexical repetition;
• simple embedding of clauses, one after another;
• interactive markers, back channels, fllers, discourse markers (see below);
• grammatically incomplete utterances;
• constructions and organisational features that are particular to unplanned speech.
(Adapted from Miller and
Weinert 1998: 22–23)

In contrast to (non-recorded) spoken discourse, written discourse is not transient; its

production immediately creates a record which can be referred to (albeit not
indefnitely). For this reason, written discourse is traditionally the object of study, and
the point of reference on which grammars of English are written and against which
spoken language is compared. If we wish to study spoken discourse, we must frst
record it and then transcribe it into written form (see Chapter 4).
There are marked differences between spoken and written discourse. Consider the
following made-up examples A and B below which contain phonological markers
(explained below).

A. So (.) I says to °her° eh, like, are YOU g-gonna, like, pick me up this avo or what
like?
B. So I said to her, “are you going to pick me up this afternoon or not?”
It’s not diffcult to tell which is which. The frst example (A) shows characteristic
markers of spoken discourse including false starts (g-gonna), pauses which are
signalled with a standard notation of parenthesis with a period in between (.), non-
standard contractions (‘avo’ for ‘afternoon’), fllers (a type of discourse marker) (‘eh’
and ‘like’), emphasis (YOU) and quiet pitch signalled by degree signs (°her°), all of
which can all be marked in written format using transcription notations (see Chapter
4). In contrast, the written version (B) contains standard English grammar and lexis.
None of the prosodic features apparent in A are discernible in B.

Discourse markers
One of the defning features of spoken communication, and which are present in the
above speech example, is the use of fllers or ‘discourse markers’ (Schiffrin 1987;
McCarthy 2004) such as ‘er’ and ‘um’ or ‘so’,‘right’,‘yeah’.Although such lexical
21
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

items may not carry much semantic meaning, they serve the pragmatic function of
showing our addressee or interlocutor that we are, in fact, listening to them (we deal
with pragmatics in Chapters 6 and 7). The linguistic term for this is ‘backchannelling’
and these little markers are more broadly understood as a type of language that is used
to maintain social connection (known as ‘phatic communion’, Malinowski 1923).
Next time you are chatting with a friend on the phone try not to ‘backchannel’ and see
how quickly your interlocutor asks
‘Are you still there?’

Standard English (or the issue of convention)

Earlier we introduced the notion of a ‘standard’ form when it comes to thinking about
a language. In written, spoken, and multimodal texts, linguistic behaviour is varied
and ranges from strict adherence to conventional grammatical and lexical forms, to
non-conventional grammar and lexis that incorporates vernacular forms, dialectal
markers and, in recent years, emoji, gifs, memes, and more. The choices we make are
context-dependent. This raises the issue of a reference point in terms of which form
of a language is considered the ‘norm’ or ‘standard’ against which different variants
of that language are assessed. That is to say, what do we mean by ‘standard form’?
When we speak, we often use language that suits the discourse situation we are in. For
example, suppose you are giving a talk in class for one of your courses. You are likely
to be more aware of the ‘correct’ forms of pronunciation and strive to adhere to the
‘rules’ of language in front of an audience, particularly one that may be judging your
intelligence!
Therefore, you might make more effort to enunciate your words.
The issue over what counts as ‘standard’ is typically one of prestige masquerading
as ‘correctness’. The advent of print in the 1400s meant that (English) language
producers needed linguistic uniformity or a standard form of the language in order to
decode written texts.The area of England that was chosen as the ‘standard’ form was
the southeast of England, named ‘Estuary English’ after the river Thames and its
estuary. ‘Estuary English’ was adopted as the new ‘normal’ as far as language was
concerned. The reason for this geographical variant was simply that London and the
surrounding area was the location of the courts, the government, and other socially
and politically important institutions, and garnered prestige based solely on this
connection to its learned users. The standardisation process took years of elaboration
and implementation but once codifed, the notion of a normative standard variant stuck
(Haugen 1972). Estuary English involved pronunciation that is now known as
‘Received Pronunciation’ (‘RP’), which we introduced earlier in the chapter. More
colloquially, RP referred to as ‘BBC’ or ‘Queen’s English’ (or now King’s), but in
reality, not many people talk like that.
In casual conversation, which is less rule-governed, adherence to Standard English
(SE) is less expected and non-standard forms abound. For instance, take the use of ‘-
22
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

ing’ in spoken words like ‘having’, we may well shift from a voiced velar nasal [ŋ] to
a voiced alveolar nasal [n] and say, for example,‘havin’ [hævin]. The social situation
you are in might encourage you to use more non-standard forms as a way of socially
bonding to the people with whom you are interacting (this is known in linguistics as
‘covert prestige’). In such interactions we may use more slang terms, or omit
grammatical elements, include expletives, and so on. Here is a complete spoken
exchange between two English speakers who come from the same family:

A: Lift’ll be in fve minutes, K? B:

Are you ACTUALLY?

The second speaker (B) appears to have asked an unfnished question, but the utterance
is complete as B’s intention is to convey surprise at the timing of the ‘lift’ in A’s
message. Note also that A uses non-standard grammar by omitting some elements of
the message. If it were SE, it would read ‘Our/Your/The lift will be here in fve
minutes, OK?’ (bold elements are omitted in the actual example).
Social attitudes underpin linguistic notions of prestige. Activity 1.3 asks you to
consider the ideological or attitudinal responses to how people use language in certain
situations and whether you consider their version of the language to be appropriate (or
not) in the context.

Activity 1.3 G-droppin’

To get a sense of how ingrained rules about standardisation are and the social snobbery that exists
among prescriptivists and language pedants in the UK, consider the criticisms levelled at Priti Patel
(the UK Home Secretary in the early part of 2022) on a popular

internet forum, mumsnet, because of her ‘G-dropping’ (conventionally known as ‘-ing dropping’)
during national briefngs and interviews (or should that be ‘briefns and interviews’?). You can
fnd the thread here,
QR 1.3 Link to G-droppin’ discussion on
mumsnet https://www.mumsnet.com/Talk/pedants_
corner/2077339-Priti-Patel-and-herdropped-
Gs, or use QR code 1.3).

To give you a favour of what is written on the forum, here are a few isolated comments from
the ‘reviews’ of Patel’s speech: “intensely irritatin’”, “absolutely awful”, “Gordon Bennett!”, 7
“causing me much stress”, “idiot”.

Do you think these criticisms of Patel’s linguistic behaviour are justifed? If so, why? On what
linguistic grounds? What might your response say about your own attitudes to language
variants? Our discussion below will help you to think about how you might answer some of these
questions.

23
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

Having considered Patel’s speech style, you might be interested to know that
popular British television presenter, Alex Scott, was criticised on Twitter in 2021 by
businessman and politician Digby Marritt Jones for habitually using the same non-

standard form in her broadcasts of the Tokyo 2020 Olympic games. Jones bemoaned
that Scott’s -ing-dropping (more specifcally, /ŋ/) “ruins the Olympics” and further
claimed – incorrectly – that -ing dropping was “wrong”.8 Simply put, ‘correct’
language is synonymous with SE or the standard variant of a given language. This
‘correct’ variant is the one that is used in offcial situations or situations where
conventions dictate that the language matches the agreed standard. It will be
characterised by strict adherence to standard grammatical structures and lexis. It is
important to note that what counts as standard can vary geographically. Indeed, if we
consider English, we would be more correct to think about different Englishes based
on their geographical location: British English, American English, Australian English,
Indian English, Hiberno-English, for example. Informal language usually includes
some non-standard forms and is what we use in everyday situations where social
conventions do not dictate that we adhere to the standard forms.

Sociolinguistic variables
While SE may garner more prestige on the whole than non-Standard variants, there
are times when the latter is preferred. Therefore, different sociolinguistic variables
(also known as situational factors) infuence linguistic choices. Interpersonal
distance, for example, can be a motivating factor in the decision to use non-standard
or standard forms. For instance, you might say ‘gimme me that book’ to your mate,
but ‘would you mind handing me that book’ to a stranger or a more socially powerful
interactant (see Chapter 7).
To offer another example, if you are chatting to friends over a messaging forum
such as WhatsApp, your linguistic choices are likely to refect, among other things, the
equal nature of your social relationship including your relative social proximity, your
generational status, and very likely your current mood. This might be manifested
linguistically via responses such as ‘K’ for ‘okay’ or acronyms with multiple
punctuation marks such as ‘WTF??!!!’, and orthographic duplication, e.g. ‘see
yaaaaaaa’ or ‘niiiiiiiiice’, or misspellings like ‘noice’ for ‘nice’. Non-Standard
examples like these would be well-received by your friends and indicate a high degree
of familiarity between you and them. However, email messages to, say, your teacher
are likely to adopt standard forms and be much less familiar; after all your teacher
would probably not take kindly to being addressed: ‘hiiiiiiii P-Dawg’. This is because
the power relations between friends and peers are usually equally distributed, whereas
between you and your teacher they are asymmetrical, and this has an impact on what
24
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

is conventionally expected in interactions (we return to this in Chapter 7). In messages

to your friends, politeness markers like ‘thanks’, non-standard lexis, like ‘gimme’, or
even the length or share of the conversational ‘turns’ (see Chapter 4) need not be so
important because, to use a term from sociology, you are on an equal footing.
However, you are not on an equal footing with your teachers and so politeness will be
more important (student readers please take note!) as will standard lexis. So, instead
of saying ‘gimme (more time to complete my essay)’ you might opt for the more
standard form ‘give me’ when communicating with your teacher.
Age is another sociolinguistic variable that impacts how we communicate. Your
lexis may also show generational differences, so ‘spill the tea’, ‘boomer’, or ‘woke’
may mean very different things to a much older teacher than to your fellow students
(unless, of course, your teacher is woke). And as one of our own students taught us,
even micro-features such as punctuation can carry social or contextual (pragmatic)
meaning as well as grammatical meaning. One of us gave lengthy positive feedback
to a student during a face-to-face tutor meeting where they thought they had not done
so well in their other classes. Their misunderstanding was fuelled by another teacher’s
answer to an online question on how they were progressing in that teacher’s course.
The teacher had answered (positively) with the following:

Fine.

The student was horrifed. When asked why, their response was ‘because it says “Fine
PERIOD”!’ Clearly, the presence of a period (full stop) at the end of a message shows
that the sentence has been completed (grammatically speaking), yet pragmatically, it
can signal displeasure or anger, depending on the context (we return to the punctuated
period in Chapter 6). For what it’s worth, this was not the intended meaning! The
student was indeed fne (no period needed).
Professional occupation is yet another sociolinguistic variable that can determine
what language choices we make. Where we work can determine what way we use
language. Profession-specifc lexis, such as ‘sheeted’ in Activity 1.2, is one such
example. Some socioeconomic domains are more likely to adhere to a standard form
of the language than others. For example, university and educational websites are
likely to use standard forms because the social and cultural context demands greater
adherence to standard linguistic conventions, not least so that they are uniformly
understood by most people in a wide geographical area. Because there is also a higher
level of social value or prestige attached to the standard form, its use in education is
advocated over non-standard or slang forms. So much so, that some schools and
educational organisations have taken to a form of language policing, proscribing what
students can and cannot say. One example is the policy (or should that be ‘policing’)
of preferred and dis-preferred language choices at Ark All Saints Academy, a high
school in southeast London which was the subject of a feature in the Guardian
newspaper in the UK (Booth 2021) for proscribing its students’ language choices. The
school banned such phrases as ‘he cut his eyes at me’, which means ‘throwing a bit of

25
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

shade’ and ‘giving side-eye’ (for the Boomers, this roughly translates as dismissing a
person by lowering the eyes and turning the head to the side).

Conclusion
In this chapter, we have touched on some of the important elements of discourse
analysis in everyday examples of communication. We have shown how the discourse
situation can impact on the language used, its meaning, and its meaning potential.
We also introduced the idea that what works in one discourse situation does not always
work in another and that part of our linguistic competence is knowing when to use
what and with whom. We also highlighted some of the differences and commonalities
between spoken and written discourse and acknowledged that the boundaries between
both modes are not so absolute, as our WhatsApp example showed.
Throughout this book we will highlight the key fact: language is choice. When we
choose to say something in a particular way, we are choosing from a varied, socially,
ideologically, and politically infected set of possible words and grammatical structures
to say it. Such choices are not always consciously made, but nonetheless we have
access to several ways of saying something, and how we say it is governed by what
we know about the world, our social expectations, and so on, all of which motivates
those selections. This knowledge of the world includes knowledge of interpersonal
relationships (e.g. friend vs teacher), age, and the discourse context. As we will show
in upcoming chapters, power relationships also have a lot to do with how and why we
make such choices. It is the remit of discourse analysis to tease out these relationships
between language and choice and to explore to what extent, and to what effect the
factors noted above shape discourse, and how discourse can shape our knowledge of
the world.
The remaining chapters present a toolkit for analysis that incorporates relevant
theories and methodological frameworks from linguistics that are instrumental for
analysing discourse. To help you practice discourse analysis, there will be activities
and tasks that explore the concepts, models, and frameworks we introduce. Answers
to activities will be either in the text immediately following the activity or at the end
of the chapter.

Notes
1 Cummings’s poems usually don’t have titles, so the frst line of the poem often becomes the proxy
title.
2 We’re not actually sure what ‘meta’ means in this context – we’d need to ask Mark Zuckerberg.
3 Thanks to Simon Garner of Hanson Aggregates for confrming the meaning of the sign and giving
us permission to use it as an example.

26
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

4 The phrase ‘reviewer 2’ sends fear through the bones of academics. When submitting an article
or other academic publication for review prior to it being accepted for publication, the author
usually receives two ‘blind’ and anonymous reviews from academics. Think of this like ‘good
cop, bad cop’. One will invariably be constructive and kind (known affectionately as ‘reviewer
1’) whereas the other will trash the fruits of your hard labour, destroy any hope or self-worth left
in your fragile ego-trodden soul, making you question your very existence in your academic
discourse community. Hello, reviewer 2.
5 While we prefer Allwood’s term ‘meaning potential’, Croft and Cruse (2004) use the term
‘purport’ but this sounds like somewhere you’d park your cat.
6 Also referred to as vocal folds.
7 Gordon Bennett is a phrased commonly used instead of a curse in some discourse communities.
8 In fact, he called it ‘g’ dropping. The correct term is ‘-ing dropping’. Jones is not a linguist. You
might be interested in Peter Trudgill’s (2021) piece on this specifc example. Peter Trudgill is a
linguist.

Further reading
Leech, Deuchar, and Hoogenraad (2006) provide an accessible introduction to the
model of grammar we use in this book.
For more on J. R. Firth, see Chapman and Routledge (2005: 80–86).
For more on language policing, see Cushing (2020) and Lampropoulou and Cooper
(2021).
Clark (2013) is a good place to fnd out more about relevance theory.
Nørgaard (2010) provides a useful introduction to graphological analysis of texts.
For an analysis of forensic discourse through relevance theory, see Lynn and Canning
(2021).

Resources
International Phonetic Alphabet with sounds: https://www.international
phoneticalphabet.org/ipa-sounds/ipa-chart-with-sounds

References
Allwood, J. (2003) ‘Meaning potential and context. Some consequences for the
analysis of variation in meaning’. In H. Cuyckens, R. Dirven, and J. Taylor (eds),
Cognitive approaches to lexical semantics, pp. 29–65. Berlin: Mouton de Gruyter.
Bloor, T. and Bloor, M. (2013) The functional analysis of English: A Hallidayan
approach. London: Routledge.
Boersma, P. and Weenink, D. (2023) Praat: Doing phonetics by computer [Computer
program]. Version 6.3.06. Accessed 31 January 2023. http://www. praat.org/

27
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

Booth, R. (2021) ‘Oh my days: linguists lament slang ban in London school’,
Guardian, 30 September. Accessed 30 September 2021. https://
www.theguardian.com/education/2021/sep/30/oh-my-days-linguists- lament-
slang-ban-in-london-school?CMP=Share_iOSApp_Other
Brown, G. and Yule, G. (1983) Discourse analysis. Cambridge: Cambridge University
Press.
Carroll, L. (2001) [1871]. ‘Jabberwocky’, Jabberwocky and other poems. Dover
Thrift Edition. New York: Dover.
Chapman, S. and Routledge, P. (eds) (2005) Key thinkers in linguistics and the
philosophy of language. Edinburgh: Edinburgh University Press.
Clark, B. (2013) Relevance theory. Cambridge: Cambridge University Press.
Croft, W. and Cruse, D. A. (2004) Cognitive linguistics. Cambridge: Cambridge
University Press.
Cruttenden, A. (2001) Gimson’s pronunciation of English (6th edition). London:
Edward Arnold.
Crystal, D. (1995) The Cambridge encyclopedia of the English language. Cambridge:
Cambridge University Press.
Cushing, I. (2020) ‘The policy and policing of language in schools’. Language in
Society, 49(3): 425–450.
‘Discourse’ (1989) The Oxford English dictionary (2nd edition) OED Online. Oxford
University Press. Accessed 30 April 2020. http://dictionary. oed.com
Estepa, J. (2017) ‘Sean Spicer says “covfefe” wasn’t a typo: Trump knew “exactly
what he meant”’, USA News Today. Accessed 3 June 2022. https://
eu.usatoday.com/story/news/politics/onpolitics/2017/05/31/sean-spicersays-
covfefe-wasnt-typo-trump-knew-exactly-what-he-meant/102355728
Firth, J. R. (1935) ‘The technique of semantics’. Transactions of the Philological
Society, 34(1): 36–73.
Freeborn, D. (1995) A course book on English grammar (2nd edition). Basingstoke
and New York: Palgrave.
Halliday, M.A. K. and Hasan, R. (1976) Cohesion in English. London: Longman.
Halliday, M. A. K. and Matthiessen, C. (2014) An introduction to functional grammar
(4th edition). London: Routledge.
Haugen, E. (1972) [1966].‘Dialect,language,nation’. In J. B. Pride and J. Holmes
(eds), Sociolinguistic, pp. 97–111. Harmondsworth: Penguin. (Originally
published in American Anthropologist, 68: 922–935.)
Heuchan, Claire (2015) ‘Veganism has a serious race problem’. Media Diversifed.
Accessed 22 October 2021. https://mediadiversifed.org/2015/12/16/veganism-
has-a-serious-race-problem
Hymes, D. (1962) ‘The ethnography of speaking’. In Thomas Gladwin and William
C. Sturtevant (eds), Anthropology and human behavior, pp. 13–53. Washington,
DC: Anthropological Society of Washington.
Lampropoulou, S. and Cooper, P. (2021) ‘The “grammar school pressure”: From
tolerance to distance, to rejection of “Scouse” in middle-class Merseyside
schools’. Linguistics and Education, 66: 1–13.
28
 DISCOURSE: LANGUAGE, CONTEXT, AND CHOICE

Leech, G., Deuchar, M., and Hoogenraad, R. (2006) English grammar for today (2nd
edition). Basingstoke: Palgrave.
Lewis, D. (1972) ‘General semantics’. In D. Davidson and G. H. Harman, Semantics
of natural language, pp. 169–218. Dordrecht: Reidel.
Lynn, N. and Canning, P. (2021) ‘Additions, omissions, and transformations in
institutional “retellings” of domestic violence’. Language and Law/ Linguagem e
Direito, 8(1): 76–96.
Malinowski, B. (1923) ‘The problem of meaning in primitive language’. In C.K.
Ogden and I. A. Richards (eds), The meaning of meaning, pp. 296–336. London:
K. Paul, Trend, Trubner.
McCarthy, M. (2004) ‘Spoken discourse markers in written text’. In G. Fox, M. Hoey,
and J. M. Sinclair, Techniques of description: Spoken and written discourse, pp.
186–198. Abingdon: Routledge.
Miller, J. and Weinert, R. (1998) Spontaneous spoken language: Syntax and
discourse. Oxford: Oxford University Press.
Nørgaard, N. (2010) ‘Multimodality: Extending the stylistic tool kit’. In D.McIntyre
and B. Busse (eds), Language and style, pp. 433–448. Basingstoke and New York:
Palgrave Macmillan.
Porter, J. (1992) Audience and rhetoric: An archaeological composition of the
discourse community. Englewood Cliffs, NJ: Prentice Hall.
Schiffrin, D. (1987) Discourse markers (No. 5).Cambridge: Cambridge University
Press.
Simpson, P. and Mayr, A. (2009) Language and power: A resource book for students.
London: Routledge.
Sinclair, J. M. (2004) ‘Written discourse structure’. In G. Fox, M. Hoey, and J. M.
Sinclair (eds), Techniques of description: Spoken and written discourse, pp. 22–
47. Abingdon: Routledge.
Sperber, D. and Wilson, D. (1986) Relevance: Communication and cognition. Oxford:
Blackwell Press.
Stubbs, M. (1983) Discourse analysis: The sociolinguistic analysis of natural
language. Chicago: University of Chicago Press.
Swales, J. (1988) ‘Discourse communities, genres and English as an international
language’. World Englishes, 7(2): 211–220.
Trudgill, P. (2021) ‘Digby Jones’ attack on Alex Scott’s accent wasn’t just snobbish,
it was wrong’. The New European. Accessed 29 September 2021. https://
www.theneweuropean.co.uk/digby-jones-attack-on-alex-scotts-accent- wasnt-just-
snobbish-it-was-wrong van Dijk, T. A. (1997) Discourse as social interaction (Vol.
2). London: Sage.

29
 Organising discourse
2 Thematic and information structure

Introduction
An important aspect of discourse analysis is considering how we structure the
information we are communicating when we speak and write. This might be at the
clausal level or at a broader, macro-level beyond the clause. In this chapter, we discuss
the organisation of information in the clause because the way the clause is structured
inevitably “fts in with and contributes to the fow of the discourse” (Halliday and
Matthiessen 2014: 88). One of the functions of the clause is to convey information, or
‘message’ (Halliday and Matthiessen 2014). The way information is organised in
clauses is constrained by the grammar of a language but there is nonetheless some
freedom within the grammar that allows the text producer room for choice. These
choices infuence how information is organised, how it is received, and the meaning
that is conveyed. For example, different syntactic structures can emphasise (and de-
emphasise) different clausal elements and focus our attention on specifc parts of the
message. It is these and other choices and their (potential) effects that are of interest
to the discourse analyst. In this chapter, we will discuss information structure using
ideas from Systemic Functional Linguistics: Theme and Rheme, and Given and New.

Organising discourse
Typically, when we communicate, whether in speech or in writing, we try to organise
what we say in a way that helps the other person (the reader/hearer) understand.
Indeed, we sometimes need to work extremely hard to structure the information in
discourse to make it understandable for the receiver. The interest of the discourse
analyst, then, is investigating “how speakers [and writers], having a given quantum of
information to impart, identify and package that information.” (Brown and Yule 1983:
176). What Brown and Yule are alluding to is that there are different ways,
linguistically, of delivering the same information and speakers and writers make
linguistic choices from the available possible alternatives within the rules of language.
These choices can affect meaning and reception.

DOI: 10.4324/9781003351207-2
The structuring and organisation of information in discourse was studied by
scholars at the Prague School of Linguistics1 (founded by Vilem Mathesius), who
showed that information structure shapes meaning and guides interpretation. Their
insights were adopted by Michael Halliday in, for example, his explorations of
transitivity (Halliday 1967) (see Chapter 9 of this book) and later his Systemic
Functional Grammar (e.g. Halliday and Matthiessen 2014). In Systemic Functional
30
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

Grammar there are two interrelated systems of analysis of the message: thematic
structure and information structure. The former involves what are known as Theme
and Rheme, and the latter Given and New information. These systems will be
described in this chapter. However, before we get to that, we frst need to introduce the
structure of the English clause.

Structure of the English clause

Simply put, the clause is an organisational unit of language that comprises (at least) a
grammatical Subject and a Predicator (e.g. ‘we write’). In Chapter 1, we briefy
introduced some of the ‘nuts and bolts’ of discourse including grammatical parts of
speech, which categorise words according to their grammatical class (e.g. noun, verb,
adjectives). We also briefy introduced syntax and noted that English is a Subject-
Predicator-Object language (SPO), where the Subject comes before the Predicator and
the Object comes after. Figure 2.1 shows these relationships using a made-up clause,
‘the donkey ate a carrot’.

Figure 2.1 Clausal and grammatical terminology

The labels Subject, Predicator and Object relate to function within a clause, and
Thompson (2013:19) suggests thinking of them as slots that are flled by particular
words and categories of words. Sometimes, instead of Predicator, the category label
Verb is used, refecting the fact that the Predicator slot is always flled by a verb or verb
phrase. However, this confates two different but nonetheless interrelated systems:
grammatical class and clausal function (Thompson 2013:20). In this chapter, we will
use Predicator to mean the clausal function and verb to mean grammatical class (or
part of speech).

Subject and Predicator

In English grammar, the Subject and the Predicator are the two main parts of a clause.
The Predicator is always realised by a verb or verb phrase (VP) containing a lexical
verb (e.g. ‘eat’) and, optionally, one or more auxiliary verbs (e.g. ‘is eating’, ‘has
eaten’, ‘has been eating’). The Predicator expresses some sort of action (physical,
verbal, cognitive) or state of being relating to the Subject and, where there is one, the
Object.
The Subject is typically a noun (‘donkey’), a pronoun (‘it’,‘he’,‘she’), or a noun
phrase (‘the hungry little donkey’) which can be structurally large containing, for
31
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

example, coordination (e.g. ‘the hungry little donkey and the child’) or a relative
clause (e.g. ‘the donkey who ate the carrot’). The canonical (prototypical) Subject is
who/what the clause is about, and/or who or what performs the action or takes on the
state realised by the Predicator. As we will see later in this section (and in Chapter 9),
this is not always the case. The position of the Subject in a clause differs depending
on clause type: declarative, interrogative, imperative. The different positions are
summarised in Table 2.1.

Table 2.1 Clause types and Subject position

Subject position Example

Before the Predicator The donkey eats
The donkey can eat
The donkey has eaten
Interrogative Between a form of the verb ‘do’ (or other Does the donkey eat?
auxiliary verb) and the main verb of the Did the donkey eat?
Predicator, or immediately after the Predicator Can the donkey eat? when
main verb is a form of ‘be’ Has the donkey eaten?
Is the donkey hungry?
Imperative Subject is elided but typically understood to Eat!
be ‘you’

Non-canonical Subjects
Subjects are prototypically noun phrases (NPs), but they can be realised by a variety
of other grammatical categories. For example:

1
(i) | To motivate a donkey | requires | a carrot
S P O
(ii) | Having a carrot to eat | was | the donkey’s dream |
S P C
(iii) | In the feld | is | where the donkey ate the carrot |
S P C
(iv) | Yesterday | is | when the donkey ate the carrot | S P C
In clause 1(i), the Subject is realised by a to-infnitive clause, in (ii) a gerund clause (a
gerund is an ‘-ing’ form of a verb which acts as a noun), in (iii) a prepositional phrase
(PP), and in (iv) the Subject is an adverbial. In 1(ii), (iii), and (iv), a form of the verb
‘be’ is the Predicator, and this is followed by a Complement which tells us more about
the Subject. We will say more about Complements after we have introduced Object.
32
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

Object
Objects, like Subjects, are typically realised by pronouns, nouns, or noun phrases.
While the Subject is usually (but not always) the thing that performs the action or
embodies the state expressed by the Predicator, the Object is the thing that the action
is performed on. For example:
2
(i) | The donkey | ate | the carrot |
S P O
(ii) | The hungry donkey | was eating | the juicy carrot | S
P O
In example 2(i), the noun phrase ‘the donkey’ (which comprises a determiner and a
noun) is the Subject of the clause and the eater of the carrot, the verb ‘ate’ is the
Predicator, and the noun phrase ‘the carrot’ is the Object and the thing that is eaten.
This type of Object is known as the direct Object (Od) because it is acted on directly
by the Subject. Some clauses also involve indirect Objects because some Predicators
(and therefore some actions) can involve more than one Object. For example, ‘gave’
typically involves something or other being given to someone/something. Consider
the following examples.
3
(i) | The child | gave | the donkey | a carrot |
S P Oi Od
(ii) | The child | gave | a carrot | to the donkey
| S P Od Oi

The examples demonstrate that the thing being given (a carrot) is being acted on
directly and is therefore the direct Object (Od), while the recipient (the donkey) is the
indirect Object (Oi). In 3(i), the indirect Object is realised by a noun phrase (‘the
donkey’) which is the canonical situation. Example 3(ii) shows a non-canonical but
nonetheless common situation where a prepositional phrase, which is made up of the
preposition ‘to’, the determiner ‘the’ and the noun ‘donkey’, acts in a similar way to
an indirect Object.2
Complement and Adjunct
As well as Subject, Predicate and Object, there are two other clause slots known as
Complement (C) and Adjunct (A). The Complement functions to tell us more about
the Subject and is typically an adjective, noun phrase, or prepositional phrase. The
Adjunct expresses extra information concerning, for example, where or when the
action happened. Adjuncts are typically adverbial or prepositional phrases. In the
following examples ‘A’ labels the Adjuncts and ‘C’ the complements.

33
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

4
(i) | The donkey | was | hungry | yesterday S
P C A

(ii) | Yesterday, | the donkey | was | hungry

A S P C

(iii) |The donkey | ate | the carrot | in the feld |

S P O A

(iv) | The donkey | was | in the feld | yesterday

S P C A

In 4(i) and (ii), the adjective ‘hungry’ provides more information about the Subject,
‘the donkey’. Examples 4 (i), (ii), and (iii), contain Adjuncts that provide information
about when (‘yesterday’) and where (‘in the feld’) the action takes place. Example 4
(iv) shows that the Complement can also be a prepositional phrase when the Predicator
is a form of the verb ‘be’. In this example, the Complement tells us where the carrot
eating took place, and the Adjunct tells us when. The position of Adjuncts is fexible.
The Adjuncts in 4(i) and (iii) are at the end of the clause, while in 4(ii) it is at the
beginning. Also, Adjuncts are not necessary for the clause to be grammatically
complete (we say more on this in Chapter 9). So, deleting the adverb, ‘yesterday’,
from 4 (i), (ii), and (iv) or the prepositional phrase, ‘in the feld’, in 4(ii) leaves a
grammatically complete clause. Removing ‘the donkey’ from any example, however,
leaves a grammatically incomplete and nonsensical clause. The
moveability/removability (or otherwise) of a clausal element is often a good test for
whether it is an Adjunct.
Adjuncts tend to be adverbs, adverbial phrases (AdvP), and prepositional phrases
(PP). They can also be single words (e.g. brightly), phrases (e.g. very brightly), or a
clause (e.g. even though the sun shone brightly). The latter type of adjunct, known as
the clausal adjunct, consists of a dependent clause. A dependent clause is one that
cannot stand on its own and needs the support of another, independent clause. For
example:
5
(i) To motivate a donkey, you have to give it a carrot
(ii) The donkey was content, because it had eaten a carrot
(ii) Having eaten a carrot, the donkey was content
In each example, the underlined section is a dependent clause, and the bold section is
independent. The independent clauses can stand alone and still make complete sense
while the dependent clauses cannot.

34
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

Passive clauses
Clauses in the ‘passive voice’ are created using a form of the verb ‘be’ and the past
participle of a lexical verb. A passive clause structures the message so the thing being
acted on, which would be the direct Object in a conventional SPO structure, becomes
the Subject. The thing doing the acting is not a necessary part of the clause but could
be added as an optional Adjunct. For example: 6
(i) | The donkey | ate | the carrot | S P
O
(ii) | The carrot | was eaten | by the donkey |
S P A
Examples 6(i) and (ii) communicate the same (or similar) information – one entity (the
donkey) acts on another entity (a carrot) in a particular way (eat). However, in 6(ii),
‘the carrot’ is Subject even though it is the thing being acted upon, while the doer of
the eating, ‘the donkey’, is now part of a prepositional phrase, ‘by the donkey’, and is
an Adjunct.
Either direct or indirect Objects can be Subject in a passive clause. Consider
examples 7(i), (ii), and (iii), all of which communicate a similar message. In 7(ii) ‘the
donkey’ (the receiver of the carrot) is Subject,‘the child’ (the giver) is part of prepos-
itional phrase which functions as an Adjunct, and ‘a carrot’ (the Object being given)
is the direct Object. In 7(iii) ‘a carrot’ is the Subject while both ‘the donkey’ and ‘the
child’ become embedded in prepositional phrases both of which act as Adjuncts. In
both examples, the Adjuncts can be removed, and the clauses will still be
grammatically complete – although in 7(iii), the clause would nonetheless ‘feel’
incomplete. 7
(i) | The child | gave | the donkey | a carrot |
S P Oi Od
(ii) | The donkey | was given | a carrot | by the child |
S P Od A
(iii) | A carrot | was given | to the donkey | by the girl |
S P A A
Note that when the Adjuncts are removed, the clause no longer communicates who or
what carried out the action – eating in 6(ii) and giving in 7(ii). Removing the doers of
actions is a topic we return to in Chapter 9, where we discuss the participants in clauses
and their roles.

Summary of clause structure

Subject, Object, Complement, Predicator, and Adjunct are labels given to elements of
the clause by virtue of their function within the clause. Particular grammatical entities
typically perform these functions: the Predicator of a clause is a verb or verb phrase;
35
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

nouns and noun phrases can be seen as the ‘norm’ for Subject and Object;
Complements are usually realised by adjectives, nouns/ noun phrases, and
prepositional phrases; and Adjuncts are typically prepositional phrases and adverbs.
However, this is only what is typical, and as we have seen throughout this section,
Subject, Object, Complement, and Adjunct can be realised by other grammatical
entities. A summary of clause structure is shown in Table 2.2. The bottom row of the
table indicates the grammatical categories that typically realise the clausal function.

Table 2.2 English clause structure and typical grammatical realisations

Subject Predicator Object Object Complement Adjunct
(indirect) (direct)

The donkey is eating the carrot in the feld

The donkey is asleep in the feld

The donkey is a mammal

The child gave the donkey a carrot yesterday

The donkey was given a carrot by the child

Noun phrase Verb phrase Noun phrase Noun Adjective; noun Prepositional
phrase phrase phrase;
adverb;
adverbial
phrase

Activity 2.1 Label the clausal elements

In the following examples, taken from a variety of sources, label the clausal functions (SPOCA) and

say how they are realised by grammatical categories (NP, VP, PP, etc.) a) I volunteer at Andy’s care

centre

Dementia UK website, Sarah’s story, 28/07/2021 https://tinyurl.com/j3x2shzr

36
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

b) Oesophageal squamous cell carcinoma is the predominant subtype of oesophageal

cancer in Asian populations
BMJ 2022; 377: e068714, 19/04/2022
https://tinyurl.com/mtdttxfa

c) Kim Kardashian has broken the internet for the thousandth time in her life
Cosmopolitan.com UK, 29/06/2018
https://tinyurl.com/35ruwhwb

d) Improving how colleagues treat each other is a critical prerequisite to improving how the
department treats the public
Windrush Lessons Learned Review, March, 20203

e) In our analysis, we have attempted to mitigate this limitation [reporting autonomy] in a

number of ways
Cancer Immunology Research, 25/07/2019
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6066474 f)
Cedars-Sinai, like every other hospital in California, is overwhelmed with COVID-19 patients

The Jerusalem Post (online), 14/01/2021

https://tinyurl.com/4rtwu6dd g) Global macroeconomic developments and commodity supply
factors will likely cause boom-bust cycles to continue in commodity markets
World Bank Press Release, 11/01/2022
https://tinyurl.com/5ykkvt52

h) As a donkey owner, you need to be prepared for the cold winter months
The donkey sanctuary website, care of donkeys through winter, 18/12/2022
https://tinyurl.com/2v4ppb36 i) Coronavirus, also called COVID-19, is part of a family of viruses
that includes the common cold and more serious respiratory illnesses such as SARS.
Age UK website, Information about coronavirus, 07/04/2022
https://tinyurl.com/5mbjzvtc We give our answers at the end of the chapter.

Thematic structure
Now that we have established how the English clause is structured and familiarised
ourselves with the labels and functions of clausal elements, we can now move on to
another way of analysing clauses. This next framework focuses on thematic structure.
It separates the clause out into two parts that serve two functions called Theme and
Rheme.4 Together, Theme and Rheme make up the message of the clause.

37
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

Theme and Rheme

In both spoken and written English, Theme is the frst component or the starting point
of a clause. It is therefore defned by its initial position in a clause. Everything that
follows the Theme is known as the Rheme. The term ‘Rheme’, which was introduced
by the Prague School, is derived from the Greek word ‘rhema’, meaning ‘that which
is said’. Theme, in its primary position, “serves as the point of departure of the
message” (Halliday and Matthiessen 2014: 89) and consequently becomes the lens
through which the rest of the clause is read and understood. Consequently, Theme can
have a focusing function, highlighting certain information. The Rheme is what the text
producer has to say about the Theme and contains the main message of the clause.
Rheme, then, realises the communicative goal of the clause. Figure 2.2 shows a simple
example of how Theme and Rheme work. In the example,‘The donkey’ is the Theme
and the rest of the clause is the Rheme.

Figure 2.2 Simple example of Theme and Rheme

Theme in declarative clauses Subject as Theme

The example in Figure 2.2 is a declarative clause and it demonstrates the prototypical
situation where grammatical Subject and Theme are confated. We noted in the
previous section that English is an SPO (Subject, Predicator, Object) language and
when Subject and Theme coincide, this is said to be conventional and unmarked.
That is not to say that Subject and Theme are the same because they are not; Subject
is a clausal function whereas Theme relates to the message of the clause and is simply
the part of the message that comes frst.
The Subject of a clause can be structurally large and include, for example,
coordinated words and phrases, and pre-/post-modifed noun phrases. In such cases, all
parts of the Subject count as the Theme. Below are some (made-up) examples to
illustrate this point (in these and all subsequent examples, Theme is underlined).

8
(i) A carrot and an apple was eaten by the donkey
(ii) A carrot or an apple was eaten by the donkey
(iii) The donkey, equus asinus, is a carrot-loving hoofed mammal.
(iv) Bonnie, the donkey, loves a carrot
(v) The donkey who ate the carrot is in the feld.
38
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

(vi) The donkey, which had a reputation for being a greedy asinus, ate the carrot.

In examples 8(i) and (ii), the Subjects are realised by two noun phrases joined by
conjunctions (‘and’ and ‘or’), and in each case the whole Subject is the Theme. In
examples 8(iii) and (iv), both Subjects are realised by two noun phrases separated by
commas. This type of structure is known as apposition whereby two noun phrases
referring to the same entity are juxtaposed with the second noun phrase providing
further information. In 8(iii) the noun phrase, ‘the donkey’ refers to the species
generally, as does ‘equus asinus’ but provides an alternative label. In 8(iv) both noun
phrases have the same referent (Bonnie) with the second noun phrase providing further
information about the referent.
The Subjects in examples 8(v) and (vi) include relative clauses that provide more
information about the referent of the noun phrase. The noun phrase in 8(v) involves
what is known as a defning relative clause which helps to identify a specifc donkey;
8(vi) has a non-defning relative clause which simply provides more, non-essential
information about the referent of the noun phrase. In each case, these relative clauses
are included in the Theme.

Marked Themes
In principle, any part of the clause can be Theme, but when a clause constituent other
than Subject is Theme, then this is marked (Halliday and Matthiessen 2014: 98). It is
worth noting, though, that some non-Subject Themes are more marked than others. It
is not unusual, for example, to see an Adjunct as Theme since this clause constituent
has more moveability than others (Thompson 2013:
149). For example:
9
(i) In the middle of the feld, the donkey ate the carrot. (ii) Later
that day, the donkey ate the carrot.

You will probably agree that the sorts of Themes above are common in English
providing, as they do, a useful starting point and focus for the Rheme. A similar type
of marked Theme that is also common is when a dependent clause occurs frst in a
multi-clause sentence. For example:

10
(i) As the sun rose, the donkey ate the carrot.
(ii) Having eaten the carrot, the donkey had a snooze. (iii) Without
pausing, the donkey ate the carrot.

39
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

In each example, the underlined section is a dependent clause and the Theme; what
follows is the independent clause and the Rheme. A more marked (and much less
frequent) situation is when the Object or Complement is Theme, as follows.

11
(i) This carrot the donkey ate.
(ii) Always hungry was the donkey

A clause can start from any point and this starting point is chosen by the text
producer/speaker. The choices made by text producers can be infuenced by the context
of the discourse. For example, the context might require a marked Theme to direct
attention to a particular part of the clause. As Thompson (2013: 150) explains:

Subject is chosen as Theme when there is no good reason to choose anything else;
but when there are contextual pressures, such as the speaker’s wish to establish a
contrast or signal a particular form of organization in their discourse, another
element – Adjunct or Complement – may be chosen as Theme instead.

A marked Theme, therefore, is an emphasised Theme and can be meaningful. The

more marked a Theme is the more potentially meaningful (and noticeable) the choice
is. A summary of the marked/unmarked Themes in declarative clauses is shown in
Table 2.3.
Activity 2.2 uses examples from real texts (which you might recognise from
Activity 2.1) and aims to demonstrate a range of Themes.
Table 2.3 Summary of unmarked and marked Themes in declarative clauses
Status of Function Constituent Example
Theme

Unmarked Subject Pronoun She ate the carrot

Noun phrase The donkey ate the carrot

Marked Adjunct Adverb Hungrily, she ate the carrot

Preposition In the feld, the donkey ate the carrot

Clause Rather than feel hungry, she ate the carrot

Object Pronoun This the donkey ate

Noun phrase This carrot the donkey ate

Complement Adjective Always hungry was the donkey

40
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

Activity 2.2 Theme and Rheme in declarative clauses In each of

the following examples identify the Theme.

(i) I volunteer at Andy’s care centre

(ii) Kim Kardashian has broken the internet for the thousandth time in her life
(iii) Improving how colleagues treat each other is a critical prerequisite to improving how
the department treats the public
(iv) Global macroeconomic developments and commodity supply factors will likely cause
boom-bust cycles to continue in commodity markets.
(v) Cedars-Sinai, like every other hospital in California, is overwhelmed with COVID-19
patients
(vi) In our analysis, we have attempted to mitigate this limitation [reporting autonomy] in a
number of ways.
(vii) As a donkey owner, you need to be prepared for the cold winter months.
(viii) Coronavirus, also called COVID-19, is part of a family of viruses that includes the
common cold and more serious respiratory illnesses such as SARS.

We give our analysis at the end of the chapter.

Theme in interrogative and imperative clauses

Interrogative clauses prototypically ask a question to seek information but as we saw
in Chapter 1 (‘Have you sheeted?’) and as we will see in later chapters this is not
always the case (e.g. consider the question ‘Can you please be quiet?’). There are two
types of questions: those that seek a yes/no response; and those that seek information
using a wh- word (i.e. who, when, where, why, what, how). For yes/ no questions, the
unmarked Theme is an auxiliary verb + Subject, for example:

12
(i) Did the donkey eat the carrot?
(ii) Has the donkey eaten yet?
For Wh-questions, the unmarked situation is for Theme to be a wh-word, for example:
13
(i) What did the donkey eat?
(ii) Where did the donkey eat the carrot?

Imperatives typically instruct (or order!) someone to do something. The other person
(the Subject) is usually elided from an imperative clause. The exception is where the
imperative involves both the text producer and receiver and starts with ‘let’ (e.g.‘let
us go/let’s go’). For imperatives, then, the Theme is the Predicator or ‘let’s/let us’.
When the imperative is negative or emphatic then ‘do’ or ‘don’t’/‘do not’ are also
included, for example:
14
(i) Eat the carrot
41
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

(ii) Let’s eat the carrot

(iii) Don’t eat the carrot
(iv) Do eat the carrot

Marked Themes
It is possible for interrogatives and imperatives to have marked Themes, for example
where an Adjunct takes Theme position, as shown in 15.
15
(i) After the carrot, what did the donkey eat?
(ii) Yesterday, did the donkey eat a carrot?
(iii) Once in the feld, give the donkey a carrot.

Table 2.4 summarises unmarked and marked Themes in interrogative and imperative
clauses.

Table 2.4 Unmarked and marked Themes in interrogative and imperative clauses
Type of clause Theme Examples

Interrogative yes/no Unmarked

Aux verb + Subject Did she eat the carrot?
Is she eating the carrot?
Has she eaten the carrot?
Marked Adjunct
Afterwards, did she eat a carrot?
(Continued)
Table 2.4 (Continued)
Type of clause Theme Examples

wh- Unmarked wh-word

Who ate the carrot?
Marked Adjunct
Afterwards, what did she eat?

42
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

Imperative Unmarked
Predicator
‘let us’ Eat the carrot!
Aux verb + ‘let us’ Let’s eat the carrot!
Aux verb + Neg + ‘let us’ Do let’s eat the carrot!
Aux verb + Predicator Don’t let’s eat the carrot!
Aux verb + neg + Predicator Do eat the carrot!
Marked Don’t eat the carrot!
Adjunct
Afterwards, eat the carrot!
Tomorrow, let’s eat the carrot!
Test your knowledge of Theme and Rheme in imperative clauses by completing
Activity 2.3.

Activity 2.3 Theme and Rheme in imperative clauses

Identify the Themes in the following sentences typical of those you would fnd in a recipe.

Preheat the oven to 180⁰C/160⁰C Fan/Gas 4.

For the sauce, heat the oil in a small frying pan over a medium heat. Fry the
onion and garlic in the oil. After fve minutes, add the spices.

Answers are at the end of the chapter.

Theme in complex sentences

When a sentence consists of more than one independent clause, then each clause is
analysed separately. For example:
16
(i) The donkey liked carrots, but carrots were a rare treat.
(ii) The donkey ate the carrot and then it had a snooze.

Compound Themes
In the examples we have used so far, the Themes have simply contained the Subject,
Object, Complement, Predicator, or an Adjunct.Themes can also contain other
elements that can form part of the Theme yet cannot by themselves constitute the
Theme. An inventory of the possible additional elements is set out by Halliday and
Matthiessen (2014: Section 3.4), which we summarise below.

Continuatives
These are also known as discourse markers and can be used for emphasis or to mark a
change in the direction of the discourse. The examples provided by Halliday and
Matthiessen (2014: 107) are: ‘well’, ‘yes’, ‘oh’, ‘now’. For example:

43
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

17
Well now, the donkey ate the carrot.
The continuative ‘well now’ is bundled together with ‘the donkey’ to create a
compound Theme.

Conjunctions
These are linking words and include, for example: ‘and’, ‘but’, ‘then’, ‘when’, ‘while’.
These can be placed in Theme position, but do not form the Theme on their own. For
example:
18
But the donkey ate the carrot.

Conjunctive and modal adjuncts

While some Adjuncts can form the Theme on their own (e.g. ‘in the feld’), Halliday
and Matthiessen (2014) identify two types of adjunct that do not. These are
conjunctive and modal Adjuncts and are grouped in with the next element in the clause
to form a compound Theme. Conjunctive adjuncts include: ‘however’, ‘meanwhile’,
‘likewise’, ‘therefore’, ‘nevertheless’, ‘to sum up’. Modal adjuncts include:
‘regrettably’, ‘probably’, ‘surely’, ‘usually’, ‘apparently’, ‘to my mind’ (for a
comprehensive inventory, see Halliday and Matthiessen 2014: 109). For example:
19
(i) However, the donkey ate the carrot
(ii) Regrettably, the donkey ate the carrot
(iii) Surely, the donkey ate the carrot

Vocatives
These are personal names or other terms of address, such as: ‘madam’,‘sir’,‘your
honour’. For example:
20
Madam, your donkey ate my carrot.
Multiple elements
The Theme can contain any or all the elements set out above, as the following
examples demonstrate:
21
(i) But then suddenly to my surprise the donkey ate the carrot in the feld
(ii) And so, madam, as expected the carrot was eaten by the donkey

44
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

Special Themes
In this section we discuss four syntactic structures that create marked Themes that
have the effect of manipulating, emphasising, and prioritising (Jeffries 2010: 77) the
starting points of the message.

Clefting
Cleft clauses separate out a clause constituent from the rest giving it prominence.
Clefts consist of an impersonal pronoun (e.g. ‘it’) which is known as the cleft pronoun,
a copular verb (usually ‘be’), a noun phrase, known as the cleft phrase, and a relative
clause, known as the cleft clause. All these constituents – the cleft pronoun, copular
verb, and cleft phrase – combine to make the Theme in the cleft structure. Example
22 demonstrates cleft structures:
22
(i) It was the carrot that the donkey ate.
(ii) It was the donkey that ate the carrot.
With 22(i), the Theme includes the Object, while in (ii) the Subject. Cleft structures
foreground the Theme and signal the cleft phrase as being newsworthy (in our
examples ‘the carrot’ and ‘the donkey’). This emphasis might be to show that the cleft
phrase has been singled out from other possibilities and/or create a contrast between
the actual situation and what might have been asserted previously (e.g. it was the
donkey that ate the carrot – not the rabbit).

Pseudo-clefts
Pseudo-clefts (also known as wh-clefts) are cleft structures that consist of a relative
clause at the start of the sentence, followed by a copular verb, followed by a noun
phrase. For example:
23
(i) What the donkey ate was a carrot.
(ii) What ate the carrot was the donkey.
Like the cleft clauses in 22 above, these pseudo-cleft constructions can also be used
to indicate a contrast in assumed knowledge in a discourse or to contradict a previous
assertion (we will see how this relates to Given and New in the next section).
Additionally, they might be used to answer imagined questions for the hearer/reader
and help to prepare the hearer/reader for what comes next, therefore guiding the
hearer/reader (Thompson 2013: 155). Wh-clefts can also be inverted so that the
relative clause is at the end of the sentence, for example:
24
(i) The carrot is what the donkey ate.

45
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

(ii) The donkey is what ate the carrot.

These can serve a similar sort of purpose as pseudo-clefts, but also tend to refer to
previously mentioned elements in the discourse or commonly held assumptions/
knowledge (see Given and New in the next section).

Fronted or preposed Themes

A further way in which the starting point of a sentence can be changed is by a special
structure known as a fronted or preposed Theme. Fronting (or preposing) refers to
any construction where the Object, Complement or Adjunct of a clause is placed
before the verb. Fronting can emphasise different parts of the clause (it can also help
with cohesion – the topic of our next chapter). For example:
25
(i) This carrot, the donkey ate.
(ii) In the feld, the donkey ate the carrot.
(iii) Hungry was the donkey
A further type of fronting is where Theme is separated from the rest of the clause by
putting it frst in a dislocated phrase. This phrase is then followed by a clause that
contains a pronoun that refers back to the Theme. Here are some made up examples
(Thematic referent is underlined and the corresponding pronoun is in bold font):
26
(i) A carrot, that’s what the donkey ate.
(ii) The donkey, it ate the carrot.
(iii) In the middle of the feld, that’s where the donkey ate the carrot.
In each case above, the fronted Theme is non-canonical and so is marked and has
prominence.
Passive clauses
Passive clauses, as we noted earlier in this chapter, put the entity that would
canonically be the direct Object (the thing acted upon directly) into Subject position.
Thematically, therefore, passive clauses bring the affected entity into Theme position,
and therefore into prominence. In example 27, the carrot is the thing acted upon – it
has been eaten by the donkey – but it is the Theme of the clause.
27
The carrot was eaten by the donkey in the feld.
Now, try identifying Theme and Rheme for yourself by completing Activity 2.4.

46
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

Activity 2.4 Theme and Rheme in a news report

Below is a short extract from the beginning of an article from a British regional newspaper, the
Northampton Chronicle and Echo, reporting on the outcome of a court case involving domestic
violence (18 March 2021). The extract starts with the headline for the piece (in bold).

Drunk dad-of-two who set fre to his partner and threatened to kill her spared jail by
Northampton judge.

A phone engineer went on a 24-hour New Year’s vodka binge before threatening to kill his
partner, a court has been told.

Dad-of-two [name redacted], 39, had been drinking heavily after struggling to cope with the
demands of the long hours he was working to provide for his family, Northampton Crown
Court heard yesterday (Tuesday, March 16).

[Redacted name], a father of two children with additional needs, had been working
‘obsessively’ in the run up to the incident and had found it diffcult to relax.

He had been drinking bottles of vodka for 24 hours and taking his son’s ADHD medication in
when his partner went to visit her family, which he said invoked feelings of ‘paranoia and
jealousy’.

Identify the Theme and Rheme. What might the choice of Theme indicate about what the text
producer deems important or ‘newsworthy’? How might the choice of Theme infuence the way
you process the information in terms of perpetrator culpability?

Information structure: Given and New

In this section, we continue our examination of how the information in the messages
we communicate is structured by looking at another of the insights of the Prague
School: the information in a message can be divided into Given and New (we use
capitals to show the technical status of these words). In this section we will describe
what is meant by Given and New (hereafter Given-New) and show how information
status is signalled in discourse.
What is Given-New information?
Haviland and Clark (1974: 513) make the point that the primary purpose of language
is to impart information; speakers/writers use language to tell their hearers/ readers
something.That ‘something’ is typically information that the hearers/readers did not
already know. However, when speakers/writers do impart information, the message
will inevitably contain things that the hearers/readers do know. To communicate
successfully (i.e. to impart information) speakers/writers must make assumptions
about what their hearers/readers do and do not know, and they must choose their
language accordingly. Information that the speaker/writer assumes is already known
by their addressees is Given information, while information that is assumed not to be
known is New (Brown and Yule 1983: 154).

47
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

Signalling Given-New information

The status of information in discourse (i.e. the Givenness or Newness) is signalled by
language choices made by the speaker/writer based on their assumptions about the
hearer’s/reader’s knowledge. Information can be assumed to be known to their
addressee and therefore Given if it has already been mentioned in the discourse or is
physically present in the immediate context. Information can be assumed to be not
known to their addressee and therefore New if it is not obvious in the context or has
not already been introduced. Givenness can be (but not always) signalled by the
defnite article (‘the’) or pronouns, whereas Newness can be (but not always) signalled
by indefnite reference, which usually involves the grammatical articles ‘a’ and ‘an’.
For example:
28
Look at that donkey. It is eating a carrot. The carrot looks tasty.
In our invented (and rather thrilling) example, the demonstrative pronoun ‘that’
(highlighted in bold type) indicates that the donkey is physically present in the context
and so is taken as Given by the speaker. The pronoun ‘it’ refers to the previously
mentioned donkey so is also Given (we deal with reference in more detail in the next
chapter). Similarly, the defnite article ‘the’ signals that the speaker assumes that the
carrot in question is known to the addressee because it has already been introduced in
the discourse. The frst mention of the carrot, though, is preceded by the indefnite
article ‘a’ indicating that (even though the hearer might be able to see the carrot) the
speaker assumes (and signals) it is New information.
Givenness can be indicated in other ways; consider the two examples in 29.
29
(i) The carrot that the donkey ate was very tasty.
(ii) Q: Where’s Bonnie your donkey?
A: At the coast for the day.
In 29(i), the information in the subordinate clause that forms part of the Subject (‘that
the donkey ate’) is assumed to be known and therefore Given (see also presupposition
in Chapter 5). The New information in that clause concerns the tastiness of the carrot.
In (ii), Bonnie the donkey is elided from answer (A) completely, and this absence
signals Givenness.
Notice that in 29(ii) the defnite article in the noun phrase ‘the coast’ signals a
different kind of Givenness: the speaker assumes the referent is common or shared
knowledge. Other examples of this sort of Givenness include ‘the sun’, ‘the moon’,
and ‘the equator’, each of which might be assumed to be known to the hearer even
though they have not been mentioned previously in the discourse. In our invented
example, the referent of ‘the coast’ will differ from discourse community to discourse
community and location to location. In some situations, the assumption might be
misplaced; for example, if the conversation in 29(ii) happened in Nepal, then the next

48
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

question might well be ‘Which coast?’ The general assumption concerning the
existence of coasts still stands, though.
Another observation we can make about 29(ii) is that the Given-New distinction is
complicated. This is because that while ‘the coast’ is signalled as assumed shared
knowledge, it is nonetheless New information with regard to the whereabouts of
Bonnie. It is, after all, the information sought in the question (Q), which explicitly
communicates it as being not known to the question asker. The New information is
therefore the circumstantial relationship (see Chapter 9 for more on circumstances)
between Bonnie and the coast: Bonnie is at the coast. The example therefore combines
what Gundel and Fretheim (2004: 176–177) call referential Given-New and
relational Given-New. The former concerns what is referenced, and the latter is what
we fnd out about the relationship between the things referenced. In 29(ii) it is what is
asserted about Bonnie the donkey that is the New information.
To summarise, Given can be seen as information that is recoverable (Halliday and
Matthiessen 2014: 118) from the immediately preceding discourse, the immediate
context (discourse situation) or from shared knowledge and means that ‘this is not
news’. New information is not recoverable and means ‘this is news’ (Halliday and
Matthiessen 2014: 118). Over the rest of the chapter, we will look at Given-New in
relation to Theme and Rheme and examine how New information can also be signalled
by syntactic and/or phonological prominence (Halliday and Matthiessen 2014: 118).
A summary of Given-New information can be found in Figure 2.3.

Activity 2.5 Given-New in a spoken interaction

Consider the following short extract of a conversation taken from the spoken section of the British
National Corpus (BNC). The conversation is between two people discussing photographs (turns
numbered for ease of reference):

1 Stuart: That’s a good one.

2 Alison: Well where’s the one I took of you in the tree?
3 Stuart: It might be on the other ones. Oh the two of us in the tree?
4 Alison: Yeah. Look at him there. Oh my god. 5 Stuart: Look at me leg!

a Which information is signalled as Given and which as New? b What are the signals? c
What do you think are the reasons for the speakers assuming Givenness-Newness?

We give our answers at the end of the chapter.

49
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

Figure 2.3 Given and New information

Given-New and Theme and Rheme

The prototypical and unmarked situation is for Given information to coincide with
Theme, and New with Rheme (Halliday and Matthiessen 2014:120). However, this is
not always the case, not least because the start of any discourse will contain mostly
new information. Indeed, the clauses in a discourse may contain a complex sequence
of Given-New information rather than the two-way division that we saw with Theme
and Rheme. Given-New status depends on what is decided as being known to the text-
recipient and what is not. Consider the information on Covid19 taken from a UK
Government website. The information has been divided into Theme and Rheme; New
information has been underlined.

Table 2.5 Theme and Rheme, Given and New

Theme Rheme

Coronaviruses are a large family of related viruses that cause diseases in animals and
humans.
Some cause less severe disease, such as the common cold,

50
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

and others cause more severe disease, such as Middle East respiratory syndrome
(MERS) and severe acute respiratory syndrome (SARS).
They are a different family of viruses to the infuenza viruses that cause seasonal
infuenza.
The Government article that the information in Table 2.5 is drawn from provides
details about the history of Covid-19. By the time the reader gets to the frst clause
shown in the table, the concept of ‘coronaviruses’ contained in the Theme is assumed
(quite reasonably) to be known. The Rheme in the frst clause contains New
information which is indicated by the indefnite article ‘a’ at the start of the long noun
phrase.
The Theme of the second clause is the word ‘some’.As a determiner,‘some’ can
indicate indefniteness and sometimes signal Newness (e.g. ‘I want some carrots’ vs ‘I
want the carrots’). In our example, ‘some’ is acting pronominally, referring to the
Theme of the frst sentence (‘coronaviruses’). In this case, we suggest that it signals
Givenness. The Rheme of the second clause contains a combination of New
information and Given; the presence of the defnite article in ‘the common cold’
suggests that text producer is assuming this is shared knowledge.
The third clause is similar, with the Theme referring back to the Theme of the frst
clause, and the Rheme containing New information, but with MERS and SARS being
taken as Given.
The fourth clause is also similar – the Theme refers to ‘Coronaviruses’, so is Given,
the Rheme contain a combination of New and Given. Indeed, the noun phrase ‘the
infuenza viruses’ is post-modifed by a subordinate clause ‘that cause seasonal
infuenza’ assumes the existence of infuenza viruses and that they cause seasonal
infuenza. All that is taken as Given by the text producer and links to the notion of
presupposition which we discuss in Chapter 5.
The example in Table 2.5 does indeed have a pattern of Theme coinciding with
Given, with successive Themes point back to information established in the frst
Theme. It also shows that the Rhemes are a combination of New and Given. Readers
may well fnd different patterns with different texts. News stories, for example, are
likely to necessarily start with all New information before a pattern is established.

Given injustice: the case of Derek Bentley

Derek Bentley was the last man to be hanged in Britain. His crime – which he
didn’t commit and was (much) later pardoned for – was the murder of a police
offcer. The judge made a great deal about Bentley’s (later disputed) confession,
which contained the phrase “I didn’t know he was going to use the gun”
(Coulthard 2006). The phrase concerns Bentley’s ‘accomplice’, Chris Craig – the
person who actually murdered the offcer – and the gun that Craig was carrying
and used to shoot the offcer. The judge claimed that if Bentley was telling the
truth, that he did not know Craig had a gun, then why would he refer to it as ‘the
gun’ and not ‘a gun’? The answer, of course, is that once Craig used said gun to
shoot the police offcer it was completely obvious to all that Craig did indeed have
51
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

a gun! By using the defnite phrase ‘the gun’ (even though it was the frst mention
in the confession), Bentley appears to be assuming, quite reasonably, that the gun
is Given information because by that point, it is known to all concerned with the
case.

Given-New in spoken English

Given and New information is introduced (or brought to the attention of the addressee)
in discourse via defnite/indefnite expressions. In spoken English, Givenness-Newness
can also be signalled by intonation. Use of prosody can make information New even
if it is in Theme position and Given information can be signalled even when it is in
Rheme position. In the spoken data they analysed, Brown and Yule (1983: 166) found
that, by and large, New information was introduced with phonological prominence,
while Given information lacked phonological prominence. Brown and Yule (1983)
explain, though, that there is not a one-to-one correspondence between phonological
prominence and new information because there are other reasons why a speaker might
choose to make part of what they say phonologically prominent, such as contrast or
emphasis. Similarly, not all discourse that lacks phonological prominence is
Given.While the pattern described by Brown and Yule (1983) indicates that there is
a strong relationship between phonological prominence and New, it is down to the
speaker to decide what to emphasise, and to judge whether something is obvious by
the context or to be in the conscious of the hearer. Given-New, therefore, is decided
by the speaker and their “moment-to-moment assessment of the relationship between
what he [sic] wants to say and his [sic] hearer’s informational requirements” (Brown
and Yule 1983: 168).

Stress and prominence

Stress can also be used to give prominence to a part of an utterance compared with the
rest of the utterance. Such prominence is achieved by choices made by the speaker
about pitch, loudness, and length of syllables (O’Grady et al. 1996: 48). In the
following examples prominence is indicated by bold type and underlining.
30
i) I thought you were singing tomorrow (someone else did not think it) ii) I
thought you were singing tomorrow (and I was right) iii) I thought you were
singing tomorrow (not someone else) iv) I thought you were singing
tomorrow (but apparently, you’re not) v) I thought you were singing
tomorrow (and not dancing) vi) I thought you were singing tomorrow (not
another day)
The prominence afforded to the underlined words in the examples above suggest that
alternative scenarios are possible or have been suggested during the discourse. In all
examples, ‘I’ is in Theme position, but it is not always Given. Instead, the stressed
52
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

word has New status, while the rest is assumed to be known and therefore Given. (We
look at the meaning potential of intonation further in
Chapter 4.)

Given-New and special Themes

As we saw earlier in the chapter, cleft clauses are a special structure where the cleft
phrase forms part of the Theme and the cleft clause is the Rheme. However, in cleft
structures the Theme is what is New, and the Rheme is what is Given (and, as we will
see in Chapter 5, also presupposed). Clefting therefore brings a clause constituent to
Theme position, and also marks it as New information. For example:
31
Theme Rheme
|It was the Donkey | that ate the carrot |
New Given
Fronting also brings the New information into Theme position. This therefore
prioritises the newsworthy information and makes it prominent.
32
Theme Rheme
(i) | Eat the carrot | that’s what the donkey did in the feld |
(ii) | In the feld | that’s where the donkey ate the carrot. |
(iii) | The carrot | that’s what the donkey ate. |
(iv) | The carrot | the donkey ate it. |
(v) | Ate the carrot | The donkey did |
New Given
The sorts of fronted structures shown in 32 are likely in a question-and-answer
situation: What did the donkey do? Eat the carrot, that’s what the donkey did. In such
situations, it is possible to elide the Given information (and therefore the Rheme),
because the information is recoverable from the discourse situation. Note that in
spoken discourse, phonological emphasis leaves the information structure unchanged.
Instead, it provides contrast. For example (emphasis indicated by underlining and bold
type): 33
(i) Eat a carrot, that’s what the donkey did. (ii) Eat
a carrot, that’s what the donkey did.
In 33(i), even though ‘that’ is stressed, ‘eat a carrot’ is still New information. You
could imagine (i) answering the question: ‘I wonder what the donkey did in the feld?’
or contradicting the statement: ‘I heard that the donkey ate an apple in the feld’.
Similarly in 33(ii), emphasising ‘donkey’ phonologically does not change the
information structure; ‘eat a carrot’ is still New. The emphasis makes clear that it was

53
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

the donkey, but it is nonetheless Given. The example could answer the imaginary
question: ‘I wonder what the donkey did?’

Activity 2.6 Given and New in a news report

Consider again the news report from Activity 2.4 and identify Given and New information in it.
Focus on the following three sentences.

(i) A phone engineer went on a 24-hour New Year’s vodka binge before threatening to kill his
partner, a court has been told.
(ii) Dad-of-two [name redacted], 39, had been drinking heavily after struggling to cope with
the demands of the long hours he was working to provide for his family, Northampton
Crown Court heard yesterday (Tuesday, March 16).
(iii) [Redacted name], a father of two children with additional needs, had been working
‘obsessively’ in the run up to the incident and had found it diffcult to relax.

Our answers are at the end of the chapter.

Conclusion
In this chapter we have looked at how information is organised in the clause. The
thematic structure of a clause is divided into Theme and Rheme, and the information
structure is divided into Given and New. We examined how the choices available to
text producers, whether speaking or writing, can infuence what information is taken
as Theme, and what is Given (shared) and New (assumed not to be known by the other
person). We looked at the difference between marked and unmarked Themes, which
assumes a structural ‘norm’ in English whereby the Subject of the clause comes frst
and therefore realises the Theme. Marked Themes can emphasise different clausal
elements and bring into focus a particular part of the message that the clause is
communicating. This may raise the importance of what is contained in the Theme and
frame how the rest of the message (in the Rheme) is understood. The prototypical
arrangement is that Given information comes before New information, and that Given
coincides with Theme, and New with Rheme. However, we saw that marked Themes
can contain New information meaning that it is highlighted as newsworthy. The
choices made by the text producer can therefore infuence how information in the
clause is marked as news or taken for granted, and this may infuence how that
information is received.

Further reading
For clear and accessible introductions to Theme and Rheme, and Systemic Functional
Linguistics more generally, see Thompson (2013) and Bloor and Bloor (2013).

54
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

Answers to activities
Activity 2.1 Label the clausal elements
a) [S] I [P]volunteer [A] at Andy’s care centre
S = pronoun; P = verb; A = Prepositional phrase (PP)

b) [S] Oesophageal squamous cell carcinoma [P] is [C] the predominant subtype
of oesophageal cancer [A] in Asian populations S = Noun Phrase (NP); P = verb;
C = NP; A = PP

c) [S] Kim Kardashian [P] has broken [Od] the internet [A] for the thousandth time
in her life
S = NP; P=verb phrase (VP) made up of auxiliary verb, ‘has’ and past
participle ‘broken’; Od = NP; A = PP
d) [S] Improving how colleagues treat each other [P] is [C] a critical prerequisite
to improving how the department treats the public
S = NP (Present participle ‘improving’ functions as a noun and is known as a
participle noun or gerund; this type of NP also known as a Gerund Phrase);
P=verb; C = NP

e) [A] In our analysis, [S] we [P] have attempted [Od] to mitigate this limitation
[A] in a number of ways.
A = PP; S=pronoun; P = VP; Od = is not a prototypical NP, but an infnitive
clause; A=PP

f) [S] Cedars-Sinai, [A] like every other hospital in California, [P] is [C]
overwhelmed [A] with Covid-19 patients.
S = NP; A = PP; P = verb; C = participle adjective (i.e. past participle of
‘overwhelm’ acting as an adjective); A = PP
This example is not a passive construction; it is describing a state of being.

g) [S] Global macroeconomic developments and commodity supply factors [P] will
likely cause [Od] boom-bust cycles to continue in commodity markets.
S = two NPs coordinated by ‘and’; P = VP which includes adverb ‘likely’; Od
= is a non-fnite, to-infnitive clause which consists of NP+VP+PP.

h) [A] As a donkey owner, [S] you [P] need to be prepared [A] for the cold winter
months
A = adverbial phrase; S = pronoun; P = ‘need’ acts as an auxiliary and is followed
by a passive infnitive; A = PP.

Activity 2.2 Theme and Rheme in declarative clauses

Theme Rheme

55
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

I volunteer at Andy’s care centre

Kim Kardashian has broken the internet for the thousandth time in her life

Improving how colleagues is a critical prerequisite to improving how the

treat each other department treats the public
Global macroeconomic will likely cause boom-bust cycles to continue in
developments and commodity markets.
commodity supply factors In
our analysis, we have attempted to mitigate this limitation [reporting
autonomy] in a number of ways.
As a donkey owner, you need to be prepared for the cold winter months.

Activity 2.3 Theme and Rheme in Imperative clauses Themes are

underlined:
(a) Preheat the oven to 180⁰C/160⁰C Fan/Gas 4.
For the sauce, heat the oil in a small frying pan over a medium heat.
Fry the onion and garlic in the oil. After fve minutes, add the spices.

Activity 2.4 Theme and Rheme in a news report

Theme Rheme

1 Drunk dad-of-two who set fre spared jail by Northampton judge

to his partner and threatened to kill
her
2 A phone engineer
went on a 24-hour New Year’s vodka binge before
threatening to kill his partner, a court has been told.
3 Dad-of-two, [name redacted], had been drinking heavily after struggling to
39, cope with the demands of the long hours he was
working to provide for his family, Northampton
Crown Court heard yesterday (Tuesday, March 16).

4 [Redacted name], a father of had been working ‘obsessively’ in the run up to

two children with additional the incident and had found it diffcult to relax.
needs,
5 He had been drinking bottles of vodka for 24 hours and
taking his son’s ADHD medication in when his
partner went to visit her family, which he said
invoked feelings of ‘paranoia and jealousy’.
In each case, we have just analysed the main clause in each sentence. The sentences
are numbered in the table. In sentences 1, 3, and 4, apposition is used in the Theme to
provide more information. The three Themes the noun phrase ‘dad-of-two’ is used

56
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

twice and ‘a father of two children with additional needs’ once. Themes are the starting
point of the message and serve to frame the message, so the information in the Rheme
is framed by the ‘dad/father-of-two’ Theme on three occasions. In sentence 2, the
starting point is ‘a phone engineer’. The Thematic pattern focuses on the social roles
(father and engineer) of the perpetrator.
What might such an analysis indicate regarding the perpetrator’s culpability? Well,
the reader frst encounters the perpetrator not as a violent abuser, but as a ‘dad’, or
‘father of two children with additional needs’, ‘drunk’, ‘a phone engineer’. In every
case, his crime is backgrounded syntactically, while his familial and professional
status is brought to the fore. This might serve to mitigate the criminal activities by
suggesting he is a hardworking parent, rather than a violent abuser. Moreover, the
reference to him being ‘drunk’ could be read as excusing his violent behaviour.

Activity 2.5 Given-New in a spoken interaction

1 Stuart: That’s a good one.
2 Alison: Well where’s the one I took of you in the tree?
3 Stuart: It might be on the other ones. Oh the two of us in the tree?
4 Alison: Yeah. Look at him there. Oh my god. 5 Stuart: Look at me leg!
<laugh>
Given is double underlined; New is single underlined.
Defnite reference is being used to refer to Objects that are physically present in the
immediate context of the conversation or known to both participants. When Stuart
uses the demonstrative ‘that’ he is referring (and maybe even pointing) to a photo in
close proximity. When Alison refers to ‘the one I took of you in the tree’, she assumes
Stuart will know the photo and remember the tree. She is therefore assuming shared
knowledge and memories. The noun phrase ‘the other ones’ again presumably refers
to more photographs which both participants are aware of. Defnite reference is being
used to refer to things that are physically present in the immediate context and which
form part of the shared knowledge of the interlocutors. This information is taken as
Given in the discourse by the participants. In turn 1, ‘a good one’ uses indefnite
reference and is New information. Turn 3 answers a ‘where’ question. The answer
provides information that Alison seeks, but uses referentially Given information (‘the
other ones’). The relationship between ‘the one of you in the tree’ and ‘the other ones’
is New.

Activity 2.6 Given and New in a news report

(i) A phone engineer went on a 24-hour New Year’s vodka binge before threatening
to kill his partner, a court has been told.

57
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

(ii) Dad-of-two, [name redacted], 39, had been drinking heavily after struggling to
cope with the demands of the long hours he was working to provide for his
family, Northampton Crown Court heard yesterday (Tuesday, March 16).
(iii) [ Redacted name], a father of two children with additional needs, had been
working ‘obsessively’ in the run up to the incident and had found it diffcult to
relax.
The news article is written discourse where the text producer (a journalist) is sharing
information with a temporally and spatially displaced text receiver (a reader).The
journalist must make assumptions about what the reader does and does not know and
therefore about what is Given-New. The indefnite articles in (i) indicate that the
information is assumed not to be known by the reader and therefore New. This is
entirely reasonable since this is the opening of the article and the frst mention of the
people and events and is what we would expect at the start of a news article (and other
texts). The Subject and Theme of (ii) has the same referent as (i), but we are
nonetheless provided with more New information about the ‘phone engineer’. This is
a relational connection we must make, along the lines of ‘the phone engineer is a dad-
of-two / is called . . . / is 39’, where the ‘phone engineer’ is Given, and the extra details
about him are New. Similarly, in (iii) the repetition of the perpetrator’s name is Given,
but we are provided further information about the referent (underlined), which is New.
Notice in (ii) the NP that begins ‘the demands of . . .’ (double underlined). The use of
the defnite article signals Givenness, even though the reader cannot possibly know
this information, and imbues ‘demands’ with the status of common or shared
knowledge (rather like ‘the coast’ example we discussed). The NP also presupposes
these things to exist, which is a topic we take up in Chapter 5.

Notes
1 The Prague School refers to Vilém Mathesius, Nikolay Trubetskoy, Roman Jakobson, and other
scholars based in Prague in the 1930s.
2 This position is not agreed. Huddleston and Pullum (2002: 248), for example, say that the PP is
not an Object at all but a Complement.
3 This independent review authored by Wendy Williams is licensed under the terms of the Open
Government Licence v3.0 which can be viewed at: nationalarchives.gov.uk/doc/ open-
government-licence/version/3
4 Following Halliday and Matthiessen (2014), both Theme and Rheme have an initial capital to
show their status as a technical term.

References
Bloor, T. and Bloor, M. (2013) The functional analysis of English: A Hallidayan
approach. London: Routledge.

58
 ORGANISING DISCOURSE: THEMATIC AND INFORMATION STRUCTURE

Brown, G. and Yule, G. (1983) Discourse analysis. Cambridge: Cambridge University

Press.
Coulthard, R. M. (2006) ‘“And then . . .”: Language description and author
attribution’. Sinclair lecture. Birmingham: ELR (Birmingham University).
Gundel, J. K. and Fretheim, F. (2004) ‘Topic and focus’. In L. Horn and G. Ward
(eds), The handbook of pragmatics, pp. 175–196. Oxford: Blackwell.
Halliday, M. A. K (1967) ‘Notes on transitivity and theme in English: Part 2’. Journal
of Linguistics, 3(2): 199–244.
Halliday, M. A. K. and Matthiessen, C. (2014) An introduction to functional grammar
(4th edition). London: Routledge.
Haviland, S. E. and Clark, H. H.(1974) ‘What’s new? Acquiring new informa- tion as
a process in comprehension’. Journal of Verbal Learning & Verbal Behavior,
13(5): 512–521.
Huddleston, R. and Pullum, G. K. (2002) The Cambridge grammar of the English
language. Cambridge: Cambridge University Press.
Jeffries, L. (2010) Critical stylistics: The power of English. London: Palgrave.
O’Grady, G., Dobrovolsky, M., and Katamba, F. (1996) Contemporary linguistics: An
introduction. London: Longman.
Thompson, G. (2013) Introducing functional grammar. London: Routledge.

59
 Organising information in
3 discourse
Cohesion

Introduction
One of the important aspects of discourse analysis is understanding the ways in which
information is structured and links together, whether at a macro (whole text) or micro
(clause/sentence) level. In the previous chapter, we introduced key grammatical
concepts and terminology which are necessary to understand the relationship between
discourse and its structure. In this chapter, we continue by considering how structure
infuences the coherence and interpretation of discourse. We will look at Halliday and
Hasan’s (1976) infuential notions of coherence and cohesion and how, when we
produce language, we make it ft in with what has gone before and what is coming up,
as well as making it situationally relevant. We will see that cohesion and coherence
are created (in part) by organising and sequencing information as the communication
event (written or spoken) progresses (Thompson 2013: 145).

Coherence and Cohesion in discourse

When discourse is seen as logical, consistent, and understandable, has continuity, and
makes sense then it is said to be coherent. Coherence is achieved by organising and
structuring the information in discourse using the available language resources.
Cohesion refers to the lexical and grammatical choices (the language resources) that
combine to make a discourse a connected whole. Whereas coherence is a matter of
judgement by the reader/hearer, and so is a logical, common-sense phenomenon,
cohesion is a language and discourse phenomenon. Coherence and cohesion in
discourse were frst examined in detail by Halliday and Hasan (1976), who proposed
and described a system of cohesion in English. According to Halliday and Hasan
(1976), cohesion incorporates the following devices.

1 reference
2 ellipsis and substitution
3 conjunction
4 reiteration (lexical cohesion)

DOI: 10.4324/9781003351207-3
In the next four sections we will briefy describe these cohesive devices. In our
discussion, we will refer to both Halliday and Hasan’s (1976) original model, as well
as to the revised model described in Halliday and Matthiessen (2014).

60
 ORGANISING INFORMATION IN DISCOURSE: COHESION

Reference
Reference is the use of pronouns (e.g. ‘she’, ‘her’), demonstratives (e.g. ‘this’, ‘that’),
and comparatives (e.g. ‘more’, ‘less’) to point backwards or forwards to things that
have already been or that are about to be mentioned in the discourse. For example,
consider this short extract from a report commissioned by the UK Government on
racist immigration policies.
1
Veronica was invited to undertake work at The University of the West Indies as
a coordinator and lecturer of a number of social work programmes. She took up
the opportunity as she was delighted to use her expertise in a new context,
however, she eventually decided to return to the UK conscious of her
responsibilities, growing debt and gaps in NI contributions.
Windrush Lessons Learned Review,
19/03/20201

In example 1, the same entity, Veronica, is referred to again and again. Veronica is
frst identifed by name and then the pronouns, ‘she’, and ‘her’ (highlighted in bold),
refer back to Veronica. Halliday and Hasan (1976: 32) suggest that pronouns (and
other reference words) are signals that indicate the relevant information is to be
retrieved from elsewhere in the discourse. By relevant information, they mean “the
identity of the particular thing being referred to” which, in the case of example 1
above, is ‘Veronica’. The pronouns are semantically empty until flled by the thing
they each refer to, known as their “referential meaning” (Halliday and Hasan 1976:
32). In 1, the pronouns all refer to ‘Veronica’ and cohesion is created via this
relationship of co-referentiality. Thinking back to our previous chapter, we can also
notice that pronominal reference signals Given information, i.e. information that has
already been introduced in the discourse and is therefore known to the hearer/reader.

Endophoric versus exophoric reference

The link between a reference word (pronoun, demonstrative, and comparative) and the
thing being referenced is cohesive only if the ‘thing’ is introduced elsewhere in the
language of the discourse. This is the case in example 1 where ‘Veronica’ is introduced
at the start of the story. The pronouns therefore point to something mentioned within
the text, and this type of reference is known as endophoric. Reference to things
external to the language of the discourse is known as exophoric reference and is not
a cohesive device because it relies on context rather than co-text for meaning
(cohesion is a textual phenomenon). To help understand the difference between
endophoric and exophoric reference, let’s return to the extract of a conversation we
met in Chapter 2 that is taken from the spoken section of the BNC. Recall that the

61
ORGANISING INFORMATION IN DISCOURSE: COHESION

conversation is between two people discussing photographs (emphasis added; turns

numbered for ease of reference):
2
1. Stuart: That’s a good one.
2. Alison: Well where’s the one I took of you in the tree?
3. Stuart: It might be on the other ones. Oh the two of us in the tree?
4. Alison: Yeah. Look at him there. Oh my god.
5. Stuart: Look at me leg!
In example 2, Stuart and Alison refer to the photos in a few ways including the
demonstrative pronouns ‘that’ and ‘there’, and the noun phrases ‘the one’ and ‘the
other ones’. Additionally, when Alison says, ‘Look at him’, the pronoun ‘him’
references a person in a photograph. These noun phrases and pronouns are all
examples of exophoric reference because they point to and reference things in the
immediate context (the situation) in which the discourse, the conversation, occurs.
They do not point inwardly to the language of the discourse and to things already or
about to be mentioned. Consequently, the referents are only apparent and known to
the people sharing the same immediate physical context and visual perspective.
Readers of the transcription or hearers of the audio recording of the conversation can
only imagine the referents of the noun phrases and pronoun.
In turn 3, however, Stuart’s use of the pronoun ‘it’ is an example of endophoric
reference because it refers backwards within the conversation to something previously
mentioned by Alison in turn 2. The pronoun points inwardly to the language of the
discourse and establishes a co-referential link between ‘it’ and ‘the one I took of you
in the tree’ creating cohesion between the two lexical items. Notice, though, that when
Alison says, ‘the one I took of you in the tree’, this is exophoric reference because it
points outwards from the conversation and references a photograph that exists
somewhere in the situation of the conversation. It is, as we saw in Chapter 2, also
assumed to be shared knowledge and therefore Given information.
To summarise, endophoric reference points inwardly to the language of the
discourse and is a cohesive device because it creates links between the textual parts of
the discourse. Exophoric reference points outwardly to the discourse situation and is
not a (textual) cohesive device. Endophoric reference utilises a limited number of
lexical forms, mainly personal and demonstrative pronouns. Exophoric reference
utilises many other linguistics forms, such as nouns phrases, to reference things in the
context of the discourse.
Endophoric reference: anaphora and cataphora There are
two types endophoric reference:
• Anaphoric (pointing backwards)
• Cataphoric (pointing forwards).

62
 ORGANISING INFORMATION IN DISCOURSE: COHESION

Anaphoric reference (or anaphora) is the most common. For example, the use of ‘she’
and ‘her’ in example 1 and ‘it’ in example 2 are anaphora because the pronouns point
backwards to something already mentioned in the discourse (known as the
antecedent). Cataphoric reference (or cataphora) is less common than anaphora and is
where the referent of a pronoun comes after the pronoun in the discourse. For example,
in example 3 (which is taken from the spoken section of the BNC), the pronoun ‘she’
refers forward to ‘my mother’.
3
She was a good provider, my mother was.
A summary of the different types of reference we have discussed so far can be found
in Figure 3.1

Figure 3.1 Summary of reference

Personal, demonstrative, and comparative reference

We have seen so far that reference is a cohesive device when pronouns and other
reference items point forwards or backwards within the discourse to things already
introduced or that are about to be introduced. Halliday and Hasan (1976) describe
three types of cohesive reference: personal, demonstrative, and comparative. An
inventory of the lexical items used for each type of reference are as follows.

(i) Personal reference pronoun ‘he’, ‘she’, ‘it’,

‘they’
‘him’, ‘her’, ‘them’

63
ORGANISING INFORMATION IN DISCOURSE: COHESION

Possessive determiner ‘his’, hers’, ‘its’, ‘their’, ‘theirs’ (ii)

Demonstrative references
Pronoun/determiner ‘this’/‘that’, ‘these’/‘those’
Adverb ‘here’/‘there’, ‘now’/‘then’ Determiner ‘the’
(iii) Comparative reference determiner ‘more’, ‘less’,
‘another’, etc. adjective ‘similar’, ‘different’, etc. adverb
‘similarly’, ‘likewise’, etc

Personal reference
This type of reference employs personal pronouns and determiners to create cohesive
links between the reference item and the thing being referenced, which can be animate,
inanimate, or abstract. We encountered personal reference in example 1 where the
pronoun ‘she’ and the possessive determiner ‘her’ referred anaphorically to
‘Veronica’. The different pronouns refer to the same entity, Veronica, so they are co-
referential.

Demonstrative reference
With demonstrative reference, the reference items are demonstrative pronouns and
adverbs of space and time that typically mark proximity or distance. In other words,
they have a pointing function, known as deixis. For instance, in example 4,‘this’
makes anaphoric reference to ‘support’. Like personal reference, demonstrative
reference is co-referential because the reference item and the thing it references both
identify the same, real-world thing, which in the case of 4 is fnancial support offered
by the UK Government (emphasis added).
4
Find out what support is available in England to help increase your income. This
includes the Cost of Living Payment, Universal Credit and Carer’s Allowance.
GOV.UK, Cost of living support

Halliday and Hasan (1976) include markers of spatial and temporal deixis (‘here’,
‘there’, ‘now’, ‘then’) because in some circumstances, these can be used cohesively.
According to Halliday and Hasan (1976: 75) ‘then’ has the meaning “‘at the time just
referred to’” and ‘now’ “‘this state of affairs having come about’”.
The referential use of ‘then’ as a marker of temporal deixis is demonstrated in example
5 (emphasis added).
5
[CMC] became one of the frst IT companies to spot the prospects for software
services exports to the US and in 1991 made probably the frst crossborder
acquisition by an Indian IT company . . . By then it was already a Rs 100 crore
company.
CNBC18.com, 04/04/2022
64
 ORGANISING INFORMATION IN DISCOURSE: COHESION

In 5, ‘then’ (highlighted) refers back to ‘1991’ – the time just referred to.
Halliday and Hasan (1976) also include ‘the’ in the list of demonstratives. We saw
in the previous chapter how the defnite article ‘the’ can indicate that information in a
discourse is Given. This function is also cohesive because it indicates that “the item
in question is specifc and identifable; that somewhere the information necessary for
identifying it is recoverable.” (Halliday and Hasan 1976: 71). If the defnite article is
cohesive then the information is recoverable from the discourse, which means ‘the’ is
used anaphorically. Thompson (2013: 218) usefully sums up the meaning of
anaphoric ‘the’ as, “you know the one I mean [. . .] because I’ve already mentioned
it”.
For example, the noun phrase, ‘the opportunity’, in example 1 (Windrush Review),
anaphorically refers back to ‘work at The University of the West Indies’. Therefore,
the item ‘the opportunity’ is, to use Halliday and Hasan’s words, “specifc and
identifable” and “the information necessary for identifying it is recoverable” from the
discourse.

Comparative reference
Another form of reference is comparative reference, which involves lexical items that
set up a relation of contrast or similarity between the reference word(s) and something
already mentioned in the discourse. For instance, in example 6, which is taken from
the British Medical Journal, that ‘adenocarcinoma is more common’ references back
to another type of carcinoma mentioned at the start of the sentence.
6
Oesophageal squamous cell carcinoma is the predominant subtype of oesophageal
cancer in Asian populations, occurring in 90% of patients, whereas
adenocarcinoma is more common in North America and western Europe.
BMJ 2022; 377: e068714, 19/04/2022

Comparative reference therefore assumes that something has already been mentioned
in the discourse against which the thing currently being talked about is being
compared.
Activity 3.1 Personal, demonstrative, and comparative reference

Consider the following examples. Identify the cohesive reference, say whether it is personal,
demonstrative, or comparative.

(a) Kim Kardashian has broken the internet for the thousandth time in her life after a video of
her reacting to Kanye West staring at Rihanna has captured everyone’s attention.
Cosmopolitan.com UK, 29/06/2018
(b) [ ESTATE AGENTS] are excited to bring to the market a much sought after, spacious three-
bedroom detached house situated within the charming village of Logierait. This is an
excellent opportunity which will appeal to a broad range of potential buyers.
Rightmove, 08/02/2022

65
ORGANISING INFORMATION IN DISCOURSE: COHESION

(c) Adverse events related to treatment resulted in death in nine of 327 (3%) patients in the
sintilimab-chemotherapy group and six of 332 (2%) patients in the placebochemotherapy
group (tables S10 and S12). The rates of adverse events were similar when we grouped
patients by chemotherapy regimen.
BMJ 2022; 377: e068714, 19/04/2022

Our answers are at the end of the chapter.

Ellipsis and substitution

Substitution and ellipsis refer to different ways in which parts of the discourse can be
omitted from the discourse because they can be recovered or presumed from the co-
text. As we noted in the previous chapter, information that can be omitted because it
is recoverable is Given.

Ellipsis
With ellipsis, the text producer leaves gaps (i.e. omits words) which the receiver must
‘fll in’ using their knowledge of the preceding discourse or contextual information.
Ellipsis can be found in questions and answers, where answers are often elliptical.
Consider the following constructed examples.
7
(i) A Did the donkey eat the carrot?
B Yes [the donkey ate the carrot].
(ii) A Did the donkey eat the apple?
B No, [the donkey did not eat the apple, it ate] the carrot.
(iii) A What did the donkey eat?
B [the donkey ate] a carrot.
(iv) A I like carrots. Do you [like carrots]? B Yes. [I do like carrots]
In each example the words in square brackets are the possible missing words and these
are recoverable from the preceding discourse (i.e. the question).

Substitution
Substitution is where a part of the discourse is substituted with a ‘gap fller’ which can
be an adverb, pronoun, or determiner. Bloor and Bloor (2013) suggest a three-way
classifcation of substitution: nominal, verbal, and clausal. Nominal substitution is
66
 ORGANISING INFORMATION IN DISCOURSE: COHESION

where nouns or noun phrases are substituted by lexical items such as ‘one’, ‘ones’,
‘same’, ‘some’, and ‘any’. Various possibilities for this are demonstrated in the
following made-up examples (substitute word underlined; substituted words in square
brackets).
8
(i) I really like carrots. Can I have some? [carrots]
(ii) I really like carrots. Chantenay are my favourite ones. [carrots]
(iii) I really like carrots. Do you have any? [carrots]
(iv) You’re having carrots? I’ll have the same. [carrots]
Sometimes the Predicator is substituted for a form of the verb ‘do’ (typically, but not
always, accompanied by ‘it’ or ‘that’). When this happens, often the Object is also
substituted with a demonstrative pronoun (‘that’). This is shown in the following
examples (substitute word underlined; substituted words in square brackets).
9
(i) A: Your donkey ate all my carrots!
B: My donkey would never do that! [eat your carrots] (ii) A: So,
your donkey likes carrots.
B: Oh, it does. [likes carrots]
In 9(i) ‘do that’ acts as a substitute for Predicator and direct Object, ‘eat your carrots’,
while in 9(ii) ‘does’ substitutes for Predicator and direct Object ‘likes carrots’.
Some lexical items such as ‘so’ and ‘not’ can substitute for Subject, Predicator, and
Object (or longer stretches of discourse). The following made-up examples
demonstrate this (substitute word underlined; substituted words in square brackets).

67
ORGANISING INFORMATION IN DISCOURSE: COHESION

10
(i) A Did the donkey eat the carrot?
B Yes, I think so. [the donkey did eat the carrot] (ii) A Did
the donkey eat the carrot?
B I think not. [the donkey did not eat the carrot]
In 10(i), ‘so’ substitutes for ‘the donkey did eat the carrot’, while in 10(ii) ‘not’ is the
negative equivalent and substitutes for ‘the donkey did not eat the carrot’. In both
cases, Subject, Predicator, and Object are substituted by a single word.

Activity 3.2 Substitution

Look at the following extracts and identify the substitution and explain what has been
substituted with what.

(a) Is Katrina pregnant? Fans seem to think so.

Daily Times, Lahore, 18/08/2022

(b) JG: Now in ‘Men in Black II’ you show off the board in a different way, it’s where you have
the Michael Jackson cameo. How did you get him to do that?
(Question posed to director of Men in Black II, Barry Sonnenfeld by Insider reporter
Jason Guerrasio)
Insider, New York, 01/07/2022

(c) He told me he has changed his mind about a child and doesn’t want another but loves me
and if we end up having one, would just go with it rather than lose me.
www.mumsnet.com, 09/08/2018

Conjunction
Conjunction creates a relationship between clauses, sentences, and longer stretches of
text (e.g. paragraphs) by making links between what precedes and/ or what follows.
In Halliday and Hasan’s words, conjunction concerns how “linguistic elements that
occur in succession” relate to each other, and how that relationship is created in the
discourse (Halliday and Hasan 1976: 227). Conjunctions can be realised by single
words like the adverbs ‘however’, ‘actually’, and so on, as well as prepositional
phrases such as ‘as a result of’, ‘in spite of’. Halliday and Hasan (1976: 228) refer to
these types of connectors as conjunctive adjuncts because they are doing the job of
conjunctions and their clausal function is that of Adjunct. As you might recall from
our discussion of the English clause in the previous chapter, an Adjunct is a word or
phrase that adds extra information to a clause but is not necessary to make the clause
grammatically complete (we will see how that works in example 11 below). Halliday

68
 ORGANISING INFORMATION IN DISCOURSE: COHESION
and Hasan (1976) set out four types of conjunction: temporal, additive, adversative,
and causal.

Temporal
Temporal conjunctives signal time and sequencing relations. Consider the following
invented example.
11
I ate lunch. Afterwards, I had a drink
In 11, the adverb ‘afterwards’ is an example of a temporal conjunctive adjunct. It is
an Adjunct because the clause ‘I had a drink’ is grammatical on its own and does not
require the adverbial ‘afterwards’ to complete it, and it is conjunctive because it
creates a time (temporal) relation between ‘I ate lunch’ and ‘I had a drink’. Without
‘afterwards’, there is no explicit relation between the two events and no temporal
sequencing, so ‘afterwards’ is a “cohesive agent” (Halliday and Hasan 1976: 229).
Its presence in the sentence presupposes the presence of something happening earlier
in the discourse. If someone were to simply say, “Afterwards, I had a drink.”, then it
is likely that the hearer would ask, “After what?”
As well as signalling the sequencing of events described by the discourse, temporal
conjunction can also relate to progression in the “communication process”, which
means the sequence in which elements of a discussion, explanation, or description are
presented (Halliday and Hasan 1976: 263). This sort of cohesive sequencing is
achieved by adverbs such as ‘frst’/‘frstly’, ‘second’/‘secondly (etc.), ‘next’, ‘then’,
‘fnally’, and is prevalent, for example, in academic and management reports. For
instance, consider the following extract from a report prepared by the US government
National Centre for Biotechnology and Information (NCBI):
12
In our analysis, we have attempted to mitigate this limitation [autonomy] in a
number of ways. First, our central analysis was based on an agnostic approach
to trAE inclusion [. . .] Second, subanalyses were performed to address potential
sources of bias within each investigator’s AE reporting pattern [. . .] Finally, [.
. .] we performed a subanalysis of laboratory values for the most predictive trAE
in our cohort . . .
https://www.ncbi.nlm.nih.gov/pmc/articles/ PMC6066474,
25/07/2019
Also encompassed in temporal conjunction and related to the internal sequencing of
the discourse is what Halliday and Hasan label ‘here and now’ relations and
summary.‘Here and now’ uses lexis to refer inwardly to the discourse, while
summary expresses the culmination of, for instance, the discussion, argument,
description, or explanation in the discourse (Halliday and Hasan 1976: 264). For
example (emphasis added):

69
ORGANISING INFORMATION IN DISCOURSE: COHESION
13
The essential idea at this point in our discussion is that, from an ontological
viewpoint, there is no sharp line between a “signifcant” and a “nonsignifcant”
difference;
American Psychologist, 1989, 44(10): 1277

14
To sum up, the study presents important implications for predicting user
engagement in digital platforms and the social information signaling these
platforms can adopt.
Behaviour & Information Technology, 21/07/2022 doi:
10.1080/0144929X.2022.2101024

In example 13,‘at this point’ realises ‘here and now’ relations by referring intern- ally
to the discussion in hand and presupposes previous elements of the same discussion.
In 14,‘to sum up’ realises summary relations and presupposes the full extent of the
description of the research presented in the paper.

Additive
Additive conjunctions express that there is more in addition to what has gone before
and is realised by, for example,‘and’,‘also’,‘further’,‘what is more’,‘more- over’,
‘additionally’. Additive also includes alternative relations (e.g. ‘or’, ‘alternatively’),
comparative (e.g. ‘likewise’, ‘similarly’, ‘by contrast’, ‘conversely’) and appositive
(e.g. ‘that is’, ‘in other words’, ‘for instance’). The short extract from the trade
magazine Grocer in example 15, which reports on the experiences of a ‘secret
shopper’, shows the additive function of ‘moreover’ (emphasised) and forms a
connection between examples of messiness in the shop and low stock levels, all of
which connect back to a lack of staff.

15
The lack of staff resulted in a messy shop foor - there were several unmanned
shopping and restocking trolleys throughout the store. . . . Moreover, the food
counters were nearly depleted of options and fve of our required items were out
of stock.
Grocer, 16, 04/12/2021

Adversative
The adversative conjunctive signals a relationship of “‘contrary to expectation’”
(Halliday and Hasan 1976: 253) and contrast, and includes ‘yet’, ‘but’,
‘however’,‘nevertheless’, and ‘though’. For example,‘nevertheless’ in the extract
70
 ORGANISING INFORMATION IN DISCOURSE: COHESION
from a Times of India article about Bollywood actor Aahana Kumra suggests that
grabbing headlines is contrary to expectations and in contrast to not being a ‘star kid’
(emphasis added).

16
[. . .] soon she realised why things will not change for her so easily as she is not
a star kid. Nevertheless, she surely knows how to grab headlines [. . .]
Times of India, indiatimes.com, 23/01/2022

Included in the adversative category are corrective relations which are realised by,
for example, ‘instead’, ‘rather’, ‘I mean’, and dismissive relations which dismiss what
has gone before and can be realised by ‘anyhow’, ‘whatever’, ‘in any case’. The
following examples, taken from the record of UK Parliament (Hansard), helps to
show how corrective and dismissive adversative conjunctives operate (emphasis
added).

17
Tim Farron: [. . .] he [the Prime Minister] says nothing about outdoor education,
which is an industry of vast importance to us in the lakes and dales, and of great
value to young people right across the country. There are 15,000 people
employed in the sector; at least, there were, but some 6,000 have now lost their
jobs.
Hansard, Volume 689, 22/02/2021

18
Mr Dunne: [. . .] will my right hon. Friend consider offering occupiers of
listed premises in town centres [. . .] a VAT exemption on
repairs and maintenance of those premises [. . .]?
Mr Hammond: [. . .] under EU law, we cannot introduce a reduced rate of VAT
that is limited to repairs, maintenance and renovation of
listed buildings. In any case, VAT incurred on their
properties by VAT-registered businesses may be recoverable
from Her Majesty’s Revenue and Customs, [. . .].
Hansard, Volume 660, 21/05/2019

In 17, the conjunctive ‘at least’ makes a correction to what was said immediately prior
to it, and so according to Halliday and Hasan (1976: 25) expresses “‘as against what
has just been said’”. In 18, the conjunctive ‘in any case’ relates back to the request for
a VAT (value added tax) exemption made by Mr Dunne and, as a preface to stating
that VAT can be claimed back, serves to dismiss the request as pointless.

71
ORGANISING INFORMATION IN DISCOURSE: COHESION

Causal
Halliday and Hasan (1976) suggest that some causal conjunctives realise general
causal cohesive relations (e.g. ‘so’, ‘thus’, ‘therefore’, and ‘consequently’), while
others identify the specifc relations of reason (e.g. ‘for that reason’); result (e.g. ‘as
a result’) and purpose (e.g. ‘for that purpose’). For example, in the following extract
from an academic journal article, ‘for that purpose’ creates an explicit purpose relation
between using ‘a novel Bayesian approach’ and examining ‘the effects of monetary
policy’ (emphasis added).
19
We examine the effects of monetary policy on income inequality in Japan. For
that purpose, we use a novel Bayesian approach that jointly estimates the Gini
coeffcient from grouped income data and the dynamics of macroeconomic
quantities.
Empirical Economics, 62(5), May 2022

Coordinating conjunctions
Coordinating conjunctions are lexical items (such as ‘and’, ‘but’, ‘or’) that are
typically used to join together (co-ordinate) words, phrases and clauses (e.g. ‘the
donkey and the carrot’). When used in this way, these conjunctions are not cohesive
devices. However, if they are used at the start of a sentence this can be cohesive.
Consider the following abridged example from the US government web pages
(emphasis added):
20
In January 2021, Americans had very few tools to protect against COVID-19,
and the tools we did have were in limited supply. Over the last year, together,
with states, localities, and public and private partners, the Administration has
mobilized an unprecedented, whole-of-society effort to give Americans the tools
they need to protect themselves.
The Administration has put vaccines at the center of our COVID-19 response
because vaccines are the best tool we have to prevent hospitalization and death.
[. . .].
The Administration has also expedited the development, manufacturing, and
procurement of COVID-19 treatments, building a diverse medicine cabinet flled
with more treatments now than at any point in the pandemic. [. . .]
The nation’s testing supply has increased dramatically. We now have free
testing sites at 21,500 locations around the country. [. . .].
And the U.S. government has successfully put equity at the heart of a
nationwide public health response. Hispanic, Black, and Asian adults are now
vaccinated at the same rates as White adults. This is the result of an all-of-society
effort that got America to where it is today [. . .].
The White House, National COVID-19

72
 ORGANISING INFORMATION IN DISCOURSE: COHESION
Preparedness Plan, March 2022

The extract is part of an extended text that sets out the U.S. Government’s plan for
Covid-19 under Joe Biden. In this part of the text, the Biden Administration is stating
what they have done already to protect the country against Covid. This extends over
a series of paragraphs, the beginnings of which we have included above. The frst
sentence in the fnal paragraph of the extract starts with ‘And’. The use of ‘and’ is
conjunctive because it connects the point made in the fnal paragraph with those that
preceded it. It is also cohesive because, as Halliday and Hasan (1976: 227) explain,
lexical items that are “conjunctive” create cohesion “indirectly” by presupposing “the
presence of other components in the discourse”. The use of ‘and’ in example 20
therefore presupposes that at least one other related point has been made in the text
which this sentence/paragraph is adding to. Starting the sentence with ‘and’ also
provides a fnality to what is ostensibly a list of the “tools” that have been “given” to
Americans to protect themselves. It is the fnal element and the metaphorical ‘cherry
on the top’ stating as it does that everything has been “mobilized” with “equity”.

Conjunction summary
To summarise, the four main types of conjunction along with an abridged overview
of subtypes are presented in Table 3.12 with examples of the conjunctives (words and
phrases) that realise cohesion. It is worth noting that the categories of conjunction are
not agreed (on this point, see Thompson 2013: 226) and the list of conjunctives is not
exhaustive, potentially open-ended, and a word or phrase can belong to more than one
category.
Table 3.1 Categories and subtypes of conjunction based on the hierarchy presented in
Halliday and Hasan (1976: 244)
Types/subtypes Examples

Temporal just then, soon, meanwhile, afterwards

− sequential then, next, after that, frstly, previously, fnally, lastly
− here and now here, at this point, now
− summary to sum up, in summary, in short
Additive and, also, moreover, furthermore, in addition
− alternative or, or else, alternatively
− appositive that is, I mean, for instance, for example
− comparative likewise, similarly, in the same way, by contrast, conversely
Adversative but, yet, however, though, nevertheless
− contrastive in fact, actually, as a matter of fact, at the same time
− correcting instead, rather
− dismissive in any case, anyway, anyhow
Causal so, then, therefore, because, consequently
− reason for this reason, it follows
− result as a result, from this

73
ORGANISING INFORMATION IN DISCOURSE: COHESION
− purpose for this purpose, to this end
Activity 3.3 Conjunctive cohesion

Consider the following short extracts from news reports. Do they contain any cohesive
conjunctions? If so, where and what type(s)?

(a) Rapper and businessman Cassper Nyovest [. . .] no longer feels safe following the
brutal murder of his friend and amapiano pioneer DJ Sumbody. As a result, the
Destiny hitmaker has decided to beef up security, according to Zimoja.
https://www.myzimbabwe.co.zw, 13/02/2023

(b) In Los Angeles County alone [. . .] one in every fve people has tested positive for the
virus. On January 11, the county hit a grim milestone: 30,000 recorded COVID-19
deaths. Meanwhile, vaccines are being rolled out at a snail’s pace and frontline
workers are overwhelmed.
The Jerusalem Post (online), 14/01/2021

Our answers are at the end of the chapter.

Reiteration (lexical cohesion)

Reiteration refers to the use of semantically related lexis in the discourse to create
lexical cohesion. Reiteration is reference to the same entity or idea by the repetition
of a lexical item, or the use of other semantically connected lexis via synonymy,
antonymy, hyponymy, meronymy and collocation. These aspects of reiteration are
as follows.

• Repetition is the recurrence of a lexical item in a discourse including morphological

variants (e.g. eat, eaten, eat, ate).
• Synonymy describes a semantic relation of similar or equivalent meaning (e.g.
magic/sorcery; little/small).
• Antonymy describes a complementary (or opposite) relationship between lexical
items (e.g. little/large; clever/stupid).
• Hyponymy refers to ‘kind of’ relationships (e.g. a carrot is a kind of vegetable),
and ‘specifc-to-general’/‘general-to-specifc’ relationships (e.g. carrot is a specifc
kind of vegetable; vegetable is a more general term than carrot). Hyponymy
incorporates the related notions of superordination and subordination. For example,
‘carrot’ is a specifc member of the set ‘vegetable’, and so it has a hyponymic
relationship to ‘vegetable’, where ‘vegetable’ is the superordinate, and ‘carrot’ is
the subordinate. Also included in hyponymy is the idea of “general nouns”
(Halliday and Hasan 1976: 275), such as ‘stuff’ and ‘thing’, which can be seen as
general superordinate categories. For example, the general noun ‘thing’ can be seen
as a broad superordinate category that includes a vast range of other types of things

74
 ORGANISING INFORMATION IN DISCOURSE: COHESION
such as, for example, food. Food is a less general category that includes, for
example, vegetable, which is a superordinate category that includes carrot.
• Meronymy extends the notion of reiteration and lexical cohesion to include ‘part
of’/‘has a’ lexical relations. For example, an apple has a core and pips; core and
pips are parts of an apple.
• Collocation is the tendency for some words to co-occur with certain others. For
example, in the UK the word ‘fsh’ has a strong tendency to co-occur with ‘chips’
(partly due to the tendency in the non-textual world for actual chips to co-occur with
battered cod or haddock on a plate). We look at collocation again in Chapters 5 and
11.

Let’s consider reiteration in some of its forms and how it creates lexical cohesion by
looking at a UK Government press release on Covid-19 (all emphases added).

21
Young coronavirus (COVID-19) patients have told their stories of battling the
virus and suffering long-term debilitating effects as part of a new flm
encouraging people to get their vaccines.
The video features several patients who experienced serious symptoms of
COVID-19 or developed long COVID, as well as the doctors and frontline
staff who treated them, to warn of the dangers of the virus for those who are not
vaccinated. It is narrated by A&E doctor, Dr Emeka Okorocha.
It comes as people aged 16 to 17 in England are offered a COVID-19 vaccine
by today (Monday 23 August), meeting the government’s target. More than
360,000 have already been vaccinated and letters and texts were sent last week
to the remaining people inviting them to book an appointment with their GP or
visit their nearest walk-in centre.
All at-risk people aged 12 to 15 in England have also been invited for a
vaccination and young people are encouraged to take up the offer as soon as
possible to build vital protection before returning to school in September. The
latest fgures show that hospitals are seeing a rise in unvaccinated young adults
admitted with COVID-19. A ffth of COVID-19 hospital admissions in
England are aged 18 to 34–4 times higher than the peak in the winter of 2020.
Press release, UK Department of Health and
Social Care, 23/08/2021

The extract contains numerous examples of lexical cohesion. For example, there is
repetition of ‘COVID-19’, ‘young’, ‘people’, and ‘aged’ (all highlighted in bold), and
word forms that share the same stem (‘vaccine’/‘vaccines’/‘vac
cination’/‘vaccinated’/‘unvaccinated’; ‘offered’/‘offer’; ‘admitted’/‘admissions’) (all
underlined). The lexical items ‘flm’ and ‘video’ have a relationship of synonymy, as
do the ages ranges ‘16 to 17’/‘12 to 15’ and the phrase ‘young people’ and ‘18 to 34’

75
ORGANISING INFORMATION IN DISCOURSE: COHESION
and ‘young adult’ (outlined in boxes). There is also a rela tionship of synonymy
between ‘coronavirus’ and COVID-19, and ‘long-term debilitating effects’ and ‘long
COVID’. The press release that the extract is taken from concerns public health and
medicine and there are several lexical items relating to this semantic domain. For
example, there is mention of ‘doctors’, ‘frontline staff’ (presumably nurses, support
workers, and porters), ‘A&E doctor’ and ‘GP’ which are semantically related
hyponymically as these are all kinds of health worker. Similarly,‘hospitals’,‘walk in
centre’ are kinds of institutions providing healthcare (we might also include ‘GP’ if
we see that lexical item referencing a whole practice rather than an individual). There
is also a meronymic relationship between ‘hospitals’ and ‘patients’ and ‘A&E’
because hospitals have patients, and Accident and Emergency departments are part of
a hospital. We can also see a collocational relationship (i.e. a tendency to co-occur)
between ‘doctor’ and ‘patients’; ‘doctor’, ‘patients’, and ‘treated’; and ‘hospital’ and
‘admissions’. These and other lexical relations in the text are summarised below.
Repetition ‘COVID-19’, ‘young’, ‘people’, ‘aged’, ‘patients’, ‘offer’/‘offered’,
‘invited’/‘inviting’,
‘vaccine’/‘vaccines’/‘vaccinated’/‘unvaccina ted’
Synonyms ‘COVID-19’/‘coronavirus’/‘the virus’; ‘flm’/‘video’; ‘long-term
debilitating effects’/‘long-COVID’; ‘young people’/‘16 to 17’ and
‘12 to 15’; ‘young adults’/‘18 to 34’.
Hyponymy ‘Doctors’ and ‘frontline staff’ are kinds of health workers; ‘hospitals’
and ‘walk in centres’ are kinds of healthcare institutions.
Meronymy ‘Hospitals’ have ‘patients’ and ‘A&E’ depts. ‘Hospitals’ also
have ‘doctors’ and ‘frontline staff’.
Collocation ‘doctors’ and ‘patients’ and ‘treated’; ‘GP’ and ‘appointment’;
‘hospital’ and ‘appointment’; ‘hospital’ and ‘patients’; ‘hospital’ and
‘admissions’; ‘hospitals’ and ‘admitted’.
The short extract from the government press release is rich with lexical cohesion. We
can also see how the cohesive device of reference interacts with the reiteration. For
example, the frst mention of the new flm uses the indefnite article (‘a new flm’); the
next mention uses a different, semantically connected term, ‘video’, but we are guided
to understand that the same entity is being referenced by the defnite article (e.g. ‘the
video’) (our emphasis). Similarly, ‘the virus’ is used to reference ‘coronavirus’ and
‘COVID-19’ and the use of the defnite article (along with the semantically connected
word) guides us to understand that this noun phrase is referring to something that has
already been introduced in the text.

76
 ORGANISING INFORMATION IN DISCOURSE: COHESION

Activity 3.4 Collocation

Let’s play a little word association game. When you encounter each word below, write down the
frst word that comes into your mind:

Night
Knife
Hat
Tea
Smoke

Whatever you have written down is likely to be what often collocates (or co-occurs) with the
words in the list. For example, ‘Night’ often collocates with ‘Day’. Collocations can offer insight
into what a speaker deems culturally or socially prominent for them. For instance, we offered
this list to some family members and one of our relatives who is a new mum came up with ‘light’
(night light). Can you guess why?

Conclusion
In this chapter we have summarised Halliday and Hasan’s (1976) important
framework for cohesion which involves reference, ellipsis/substitution, conjunction,
and reiteration. As we have seen, cohesion in discourse is achieved via numerous
lexical means, and Halliday and Hasan’s framework systematically sets out the
various linguistic possibilities. There is an interplay between Theme and Rheme,
Given and New, cohesion and English grammar. The choices made at the syntactic
and lexical levels infuence thematic organisation, information structure and the way
the text coheres. Thematic patterns coincide with referential and lexical cohesion;
similarly, Given information can be elided as well as having referential connections.
Themes and information form sequences in texts and this sequencing is supported by
cohesion. In the next chapter we will develop the notion of sequencing by discovering
more about the sequences that occur in spoken interactions.

Further reading
For clear and accessible introductions to coherence and cohesion, and Systemic
Functional Linguistics more generally, see Thompson (2013) and Bloor and
Bloor (2013)

77
ORGANISING INFORMATION IN DISCOURSE: COHESION

Answers to activities
Activity 3.1 Personal, demonstrative, and comparative
reference
Identify the cohesive reference and say whether it is personal, demonstrative or
comparative.

( a) Kim Kardashian has broken the internet for the thousandth time in her life after a
video of her reacting to Kanye West staring at Rihanna has captured everyone’s
attention.
Cosmopolitan.com UK, 29/06/2018

Personal reference: the possessive pronoun ‘her’ (used twice) refers anaphorically to
the proper noun (the antecedent) ‘Kim Kardashian’. The noun and the possessive
pronoun are co-referential because they both refer to the same person in the real
world.

( b) [ESTATE AGENTS] are excited to bring to the market a much sought after,
spacious three-bedroom detached house situated within the charming village of
Logierait. This is an excellent opportunity which will appeal to a broad range of
potential buyers.
Rightmove, 08/02/2022
Demonstrative reference: at the start of the second sentence, ‘this’ refers
anaphorically to the house mentioned in the previous sentence (and perhaps to the
fact that it is for sale).

( c) Adverse events related to treatment resulted in death in nine of 327 (3%) patients
in the sintilimab-chemotherapy group and six of 332 (2%) patients in the placebo-
chemotherapy group (tables S10 and S12). The rates of adverse events were
similar when we grouped patients by chemotherapy regimen.
BMJ 2022; 377: e068714, 19/04/2022

Comparative reference: ‘similar’ forms a comparative relationship between rates of

adverse events when patients were grouped by chemotherapy regimen and adverse
events under different circumstances.

Activity 3.2 Substitution

Identify the substitution and explain what has been substituted with what.

( a) Is Katrina pregnant? Fans seem to think so.

‘So’ substitutes for ‘Katrina is pregnant’.

78
 ORGANISING INFORMATION IN DISCOURSE: COHESION
( b) Now in ‘Men in Black II’ you show of the board in a diferent way, it’s where you
have the Michael Jackson cameo. How did you get him to do that?
The pronoun ‘that’ substitutes for (something like) ‘cameo in the movie’.

( c) He told me he has changed his mind about a child and doesn’t want another but
loves me and if we end up having one, would just go with it rather than lose me.

The pronouns ‘another’ and ‘one’ are used to substitute for ‘a child’. The pronoun
‘it’ refers back to ‘having one’.

Activity 3.3 Conjunctive cohesion

( a) Rapper and businessman Cassper Nyovest [. . .] no longer feels safe following the
brutal murder of his friend and amapiano pioneer DJ Sumbody. As a result, the
Destiny hitmaker has decided to beef up security, according to Zimoja.
https://www.myzimbabwe.co.zw, 13/02/2023
There is a causal relation of result that is realised by ‘as a result’, and this creates a
cohesive causal link between Nyovest not feeling safe and him beefng up security.

(b) In Los Angeles County alone [. . .] one in every fve people has tested positive for
the virus. On January 11, the county hit a grim milestone: 30,000 recorded
COVID-19 deaths. Meanwhile, vaccines are being rolled out at a snail’s pace and
frontline workers are overwhelmed.
The Jerusalem Post (online), 14/01/2021

Temporal conjunction is realised by ‘meanwhile’ which presupposes that other events

have been described in the discourse that are happening at the same time as vaccines
being rolled out, etc.

Activity 3.4 Collocation

Here are the answers from our small sample of respondents:
Night – Day
Knife – Crime
Hat – Stand/Coat
Tea – Coffee
Smoke – Fire
‘Night’ is linked to ‘day’ because they are antonyms.‘Knife’ collocates with ‘crime’
which attests to the prevalence of knife crime in the recent press (January 2023) in
our small sample group’s region. ‘Hat’ collocates with ‘stand’ which could be said to
be a compound term ‘hat-stand’, but other respondents in our sample said ‘coat’. We
wondered what they might say if we asked them in summer! ‘Hat’ and ‘coat’ are both
subordinates of the superordinate term ‘clothes’, and so are cohesively linked along

79
ORGANISING INFORMATION IN DISCOURSE: COHESION
those lines. ‘Tea’ and ‘coffee’ are collocated along the same cohesive relationship as
‘hat’ and ‘coat’ in that the relationship is member of a set (‘hot drinks’). ‘Smoke’ and
‘fre’ collocate via a causal relationship, in that smoke indicates the presence of fre and
fre causes smoke to form. There is, after all, no smoke without fre.

Notes
1 This independent review authored by Wendy Williams, the title of which presupposes (some
might say erroneously) that lessons were learned, is licensed under the terms of the Open
Government Licence v3.0, which can be viewed at: nationalarchives.gov.uk/doc/ open-
government-licence/version/3
2 Halliday and Matthiessen (2014) describe an expanded hierarchical classifcation of cohesion,
which has three basic categories: elaborating, extending, and enhancing. These expand out to
nine further subtypes (of which additive, adversative, causal, and temporal are just four), and
these expand further to more than 20 sub-subtypes.

References
Bloor and Bloor (2013) The functional analysis of English: A Hallidayan approach.
London: Routledge.
Halliday, M. A. K and Hasan, R. (1976) Cohesion in English. London: Longman.
Halliday, M. A. K. and Matthiessen, C. (2014) An introduction to functional grammar
(4th edition). London: Routledge.
Thompson, G. (2013). Introducing functional grammar. London: Routledge.

80
 4 Analysing spoken discourse

Introduction
In this chapter we continue our exploration of discourse by looking at spoken
interactions. The analysis of spoken discourse is concerned with examining how
utterances between conversation participants are connected, structured, and organised.
Analysts attempt to fnd regularities in the function and sequencing of utterances and
explain how these regularities systematise conversation and contribute to meaning
making within discourse communities. Using approaches from Conversation Analysis
(CA), developed in the 1970s by Harvey Sacks, Emanuel Schegloff, and Gail
Jefferson (see, for example, Sacks et al. 1974) and the Birmingham School of spoken
Discourse Analysis (see, for example, Sinclair and Coulthard 1975), this chapter
explains how to analyse spoken discourse.
Analysis of spoken discourse involves looking at the meanings of the words used
(semantics), the order in which they occur (syntax), the way in which they are uttered
(intonation), their position within the discourse (co-textual sequencing), and the
situation in which they are uttered (context). We also need to understand what function
the words are performing (the pragmatic force) in the context they are uttered. This is
because when we speak, we are doing things with words; we are performing social
actions (Schegloff 2007: 2) or verbal acts (Sinclair and Coulthard 1975) such as
requesting information, issuing a directive, or making an offer. Therefore, as we will
see in this chapter, looking at the function of an utterance in a conversation is an
important part of analysing spoken discourse.

How spoken discourse is analysed

The analysis of spoken discourse centres on making observations about the way real
(rather than invented or scripted), naturally occurring spoken interactions unfold
dynamically between participants as they react and respond to each other. Such
analysis requires data collection, which involves creating:

(i) audio recordings of dialogue (once permission has been obtained; we deal with
the ethics of data collection in Chapter 12); and
(ii) written representations of the spoken interactions in the form of detailed
transcriptions.

DOI: 10.4324/9781003351207-4
Transcriptions of spoken discourse aim to record in as much detail as possible not
only what is said but also how it is said. This means that the phonological features
associated with spoken discourse (e.g. loudness, intonation, stress, pauses) are

81
ANALYSING SPOKEN DISCOURSE

represented graphologically in the written record. Transcriptions also identify

different speakers involved in the conversation to show who said what. All these
features are important for discovering patterns in spoken dialogue and exploring how
meaning is communicated.
Transcriptions can also use orthographic means to represent the pronunciation of
words (e.g. ‘whadja won’ instead of ‘what do you want’). Our observation is that this
convention is applied rather inconsistently and unsystematically across transcriptions
(on this point, see Walker 2013: 471), and apparently ignores the existence of the
International Phonetic Alphabet (IPA). The extent to which pronunciation is
represented in the transcription will depend on the analyst. Our approach in this
chapter is to restrict such orthographic representations to where they are absolutely
necessary; for example, elided syllables are not transcribed (e.g. ‘because’ is
transcribed as ‘cause’ if the frst syllable is not said). Whatever conventions are
adopted, they need to be applied consistently throughout.

Data used in this chapter

Notice that transcribing a spoken interaction is both analytical and interpretative.
When viewing a transcription prepared by another person, you are viewing their ‘take’
on the spoken dialogue using whatever transcribing conventions they use. Therefore,
where possible, it is important that analysing spoken dialogue should include listening
to the audio (or audio-visual) recordings as well as scrutinising the written transcript.
For that reason, throughout this chapter, we will be basing our discussion on data
obtained from the Santa Barbara Corpus of Spoken American English (SBCSAE) (Du
Bois et al. 2000–2005) and the spoken section of the British National Corpus (BNC).
We use these corpora because they are well documented, freely available online, and
include the audio recordings that the transcribed data is derived from (see the end of
this chapter for details). This means that readers can listen to the original conversations
themselves, which can be extremely helpful for understanding the transcriptions and
how the conversations proceed. We base our transcriptions on those from these
corpora, but we have used different transcribing conventions for some prosodic
features. We will introduce the conventions as we go along and present a summary at
the end of this chapter.

Spoken Interactions
When we talk, we do not produce words in a continual stream. Instead, as analysts
have long since recognised, our words form smaller, meaningful units (see, for
example, Crystal 1979).As Szczepek Reed (2010: 354) puts it,“speakers produce talk
on a chunk-by-chunk basis”. In written discourse we achieve ‘chunking’
graphologically using, for example, punctuation marks which help to divide the
82
 ANALYSING SPOKEN DISCOURSE

discourse and the information it contains into meaningful units such as clauses,
sentences, and paragraphs. In spoken discourse, however, full stops, commas, etc. are
not possible (although, see Activity 4.1). Instead, spoken discourse contains numerous
other important phonological features including pauses, ‘uhms’ and ‘ahs’, and
variation in voice pitch. It is these phonological means that help divide the speech into
units of meaning and contribute to meaning making and without them, some meaning
is lost. To illustrate this, read through example 1, which is a transcription of the words
used in a spoken conversation between two friends obtained from the SBCSAE. We
have omitted most punctuation (apostrophes are included) because when we are
speaking, we do not use full-stops (periods), commas and so on. Following Jefferson
(2004), we have used uppercase letters to mark proper nouns and the start of a turn
(i.e. the initial letter of frst word is capitalised). Look at the words from the
conversation and consider whether there are any places where you are not sure what
meaning is being communicated (turns are numbered).

1
1. ALINA: Remember Tyke lives next door to Mom
2. LENORE: Yeah
3. ALINA: Okay two weeks ago I’m watching tv and David Horowitz is going to
have this former car radio thief on
4. LENORE: It’s her boyfriend
5. ALINA: Yeah her ex-boyfriend Mike he’s the one that stole (name)’s radio
6. LENORE: How do you know
7. ALINA: Well cause well he he was a cocaine addict
(SBC006, Cuz, Santa Barbara Corpus)

Note: The name of person whose radio was stolen is unclear in the recording and has
been replaced with (name) in the transcription.

It is likely that when you read example 1, there are places where you intuitively
want to place punctuation to indicate, for instance, where you think a question is being
asked, where a sentence ends, or where there is a pause. There might be places,
however, where you cannot determine, for example, whether an utterance is a question
or a statement. This is because your intuitions are based solely on the words that the
speakers produced, their arrangement in the transcription, and your knowledge of
English grammar. However, when engaged in spoken conversation, participants also
create (and interpret) meaning using phonic (and other1) resources such as pitch,
loudness, word duration, pauses, and speed of delivery. Therefore, unless you listen
to the recording of this conversation, you are missing out on numerous phonic signals
that combine with the words and syntax to help create meaning (Selting 2010: 6; see
also Crystal 1979: 5). The phonological means we employ to help create meaning in
spoken discourse are usually referred to as prosody, or prosodic features. In the next
section we will examine how prosody helps in meaning making in spoken discourse.

83
ANALYSING SPOKEN DISCOURSE

Activity 4.1 Punctuation in spoken discourse

Back in the day, Victor Borge famously created a comedy sketch using the idea of sound effects
to indicate punctuation in texts he read from. His comedy act was not too far from reality,
because in spoken discourse we do use sound features to convey meaning in the same way
orthographic conventions do so in written discourse.

To fnd out what used to pass as comedy, search for Victor Borge QR 4.1 Victor Borge sketch
on YouTube, or follow this QR code.

As you read on through the sections on prosody in this chapter,

consider Borge’s sketch from a linguistic perspective and what (if any) problems there might be
with the central premise of his comedy.

Prosody in spoken discourse

Over these next two sections we will introduce two important prosodic features:
pauses and intonation. We will illustrate our discussion using the extract in example
1.

Pauses
Pauses are the short but noticeable silences that occur during spoken interactions as a
result of speakers hesitating, thinking about what to say next, and breathing. When
examining spoken data, analysts time pauses to the nearest tenth of a second and
transcribe them within the turn in which they occur using a system of symbols (see
below). Pauses that occur in between turns (i.e. after one person fnishes speaking and
before another person starts) are marked on a separate line of the transcription because
it is not clear who the pause “belongs to” (Du Bois et al. 1992). According to
Stenström (1994: 7), pauses can also be nonsilent or flled with voiced sounds such as
‘um’ or ‘erm’. Additionally, pauses can also include audible inhalation or exhalation
(which means they are flled with breathy sounds). The following pause categories and
transcription codes are based on those developed by Jefferson (2004), Du Bois et al.
(1992), and Stenström (1994).

Short pause – 0.2 seconds or less (.)

Medium pause – 0.3–0.6 seconds (. .)
Long pause – 0.7 seconds or over (e.g. 1.2 seconds) (1.2)
Short flled pause (um)
Longer flled pause (one or more colons lengthen pause) (u:m)
Audible inhale (H)
Audible exhale (Hx)

Pauses can be used by speakers to break their speech into units, and this can help to
make meaning. To illustrate, let us now return to example 1 (labelled 1a below) where
84
 ANALYSING SPOKEN DISCOURSE

we can see the same conversation between Alina and Lenore, but this time with pauses
marked using the coding conventions set out above.

1a
1.ALINA: Remember Tyke (. .) lives next door to Mom
(1.8)
2. LENORE: Yeah
(. .)
3. ALINA: Okay (. .)(H)(. .) two weeks ago I’m watching tv (. .) and David Horowitz
is going to have this former car (. . .) radio thief on
(. . .)
4. LENORE: it’s her boyfriend
5. ALINA: ( H) (.) Yeah her ex-boyfriend (. .) Mike (. . .) he’s the one that stole
(name)’s radio
(2.2)
6. LENORE: How do you know
7. ALINA: ( H) Well (. . .) cause well (. .) he (. .) he was a cocaine addict

The transcription now shows that the spoken dialogue involves pauses of different
lengths, some with audible inhalation. The addition of pauses to the transcription helps
to indicate how the spoken discourse is divided into units. This is particularly clear in
Alina’s frst turn, where the pause divides the turn into two meaningful units: (i)
‘Remember Tyke’ and (ii) ‘lives next door to Mom’. We can also notice at this point
that each of these units have a different discourse function; we will return to the
functional analysis of units later in the chapter.

Activity 4.2 Pause relocation

Try saying Alina’s frst turn, but this time with a pause after ‘remember’:

Remember (.) Tyke lives next door to Mom

What effect, if any, does this have on the information chunking and potential meaning of the
turn?

We give our answer at the end of the chapter.

Pauses can perform a similar role to the punctuation we fnd in written discourse by
dividing up the talk into meaningful units. While there is not a oneto-one relationship
between pauses and meaningful units of information, pauses are a good place to start
in the transcription process, once the words have been transcribed. Another crucial
prosodic feature in creating meaning in spoken discourse is intonation.

85
ANALYSING SPOKEN DISCOURSE

Intonation
According to Daniel Jones, an early pioneer in speech science, intonation is “the
variations which take place in the pitch of the voice in connected speech” (Jones
1960). Intonation, then, refers to the way in which speakers change their voice pitch
(or ‘tone’) during spoken interaction. Intonation can rise, fall, rise and fall
(rise-fall), and fall and rise (fall-rise). These changes can occur across syllables, single
words or groups of words and create what are known as pitch contours (Carr 2012:
107), and each contour helps to communicate meaning beyond the words uttered.
Some linguists suggest that intonation can indicate, for example, attitude to
information (doubt or certainty), emotions (happiness, sadness, boredom, annoyance),
and empathy (friendliness, hostility to hearer). While there is no universal agreement
on this, there is some consensus that changes in pitch can indicate whether the
utterance is complete or more information is to follow. For instance, it is generally
accepted that a fall in a pitch can indicate that a person is fnishing what they are saying
and is sometimes referred to as a terminal intonation contour. On the other hand, a
rise in pitch can signal incompleteness and is referred to a non-terminal intonation
contour. To illustrate these general tendencies, try recording yourself saying the
following out loud:

Events will take place in London Paris Washington and Sydney

Now, listen back to your recording. Think about where the intonation changes in your
voice as you say the words. Our hope is that you will fnd that your voice pitch rises
on all items in the short list of cities apart from the last one, where your voice pitch
falls. Rising intonation signals that there are more items to follow and falling
intonation signals that the list is complete.
One way in which intonation can be indicated in transcriptions, and the method we
use in this chapter, is by placing arrows either before the word or within the word
where the pitch movement occurs2 (see, for example, Carr 2012). For instance, if we
return to our list of cities, the intonation can be shown in the following way:

Events will take place in Lon↗don Pa↗ris Washing↗ton and Syd↘ney

The upward arrows indicates that ‘London’, ‘Paris’, and ‘Washington’ are said with
rising intonation, while ‘Sydney’ is said with a downward pitch movement which
starts on the fnal syllable. Other documented pitch movements include the fall-rise
(indicated by ↘↗), the rise-fall (indicated by ↗↘), and level tone (indicated by →).
These fve pitch movements are sometimes referred to as tones and their transcription
symbols are summarised in Table 4.1.

Table 4.1 Pitch movements and symbols

Pitch Movement Arrows

rise ↗

86
 ANALYSING SPOKEN DISCOURSE

Fall ↘
rise-fall ↗↘
fall-rise ↘↗
Level →
Intonation and syntax work together to create meaning, where falling tone is
typically used for statements, commands, and wh-questions (Carr 2012: 121–122), and
a rising tone is used for yes/no questions, to turn statements into questions, and non-
fnal clauses. For example:

(i) Statement: He asked you to ↘leave

(ii) Command: Get out of my ↘house
(iii) Wh-question: When are you ↘leaving
(iv) Yes-no question: Are you leaving ↗now
(v) statement as questions You are leaving ↗now
(vi) Non-fnal clause: He promised to ↘↗leave . . . . (but he ↘didn’t)

These intonation patterns are what tend to happen in spoken discourse and are not hard
and fast rules. Different intonation is possible, and any of the above may be said with
different intonation contours, which may communicate different meaning. For
example, commands can become polite offers if said with a rising tone and yes-no
questions said with a falling tone can give information rather than seek it, for example:

Have some ↗cake (i.e. would you like some cake?)

Would you like to ↘leave (i.e. do that again and you’re leaving!)

Even with the possible variations in tone contours, in spoken discourse the following
general pattern is recognised:

• rising tones indicate more is to come or information is being sought;

• falling tones indicate fnality or information is being given.

Intonation therefore plays an important role in how we create meaning in spoken

discourse not least because it can be instrumental in letting our interlocutor know
whether we are coming to the end of what we are saying or continuing. Together,
pauses and intonation help to divide up spoken discourse into units of information and
meaning. Let us now look at some examples of intonation at work in spoken discourse
by considering again the conversation between Alina and Lenore. In Example 1b,
below, we return to the transcription but this time with symbols added to indicate
intonation.

1b
1.ALINA: Remember ↗Tyke (. .) lives next door to ↗Mom
(1.8)
2. LENORE: ↗↘Yeah
87
ANALYSING SPOKEN DISCOURSE

(. . .)
3. ALINA: ↘Okay (. .)(H)(. .) two weeks ago I’m watching ↘↗TV (. .) and David
Horowitz is going to have this former →car (. . .) radio thief ↗on
(. . .)
4. LENORE: it’s her ↗boyfriend
5. ALINA: (H) (. .) ↗↘Yeah=her ex-boy↘friend (. .) ↘Mike (. . .) he’s the one
that stole (name)’s ↘radio
(2.2)
6. LENORE: How do you ↘know

7. ALINA: (H) ↗↘Well (. . .) cause ↘well (. .) →he (. .) he was a cocaine ↘addict

In turn 1, we can see that rising intonation makes it clear that the frst unit is a request
for information and is intended as a question. The second unit also ends with rising
intonation and, although it is giving information to Lenore, it is connected to the initial
request for information by the intonation.
In turn 4, Lenore makes a statement, but the rising intonation on ‘boyfriend’
suggests that she is seeking clarifcation and not stating a fact. This appears to be how
Alina interprets Lenore’s utterance because she responds with an affrmative ‘yeah’ as
though it were indeed a yes-no question.
The rise-fall intonation on ‘yeah’ at the start of turn 5 suggests that it is a separate
unit of information from that which follows – ‘her ex-boyfriend’ – which has its own
intonation contour. Note the lack of pause between ‘yeah’ and ‘her’. When two units
of information have no pause between them, this is known as latching, and the
convention is to use an equal sign (=) to mark where this happens. Latching (i.e. not
pausing) is one method of indicating that the turn is not over and that the current
speaker is going to continue.

Other prosodic features

There are, of course, other prosodic features associated with spoken discourse that we
do not have space to discuss in this chapter. These include the loudness or quietness
of the spoken words relative to the surrounding talk, the lengthening of words, and the
speed of delivery. The transcription conventions for these features are shown in Table
4.4 at the end of the chapter. Some of these features will be useful for tackling Activity
4.3.

88
 ANALYSING SPOKEN DISCOURSE

Activity 4.3 Okay

Key and Peele (2015), two male comedians, perform a sketch in which they acted as two female
QR 4.2 Key and Peele ‘Okay’ sketch characters, let’s call them Speaker 1 and Speaker
2 (fnd the clip on YouTube; top tip – search for
‘key and peele okay’ or follow the QR code
below).

In the sketch, speaker 1 begins by berating her unfaithful boyfriend, going through all the things
she will do to him for cheating on her and what she will do if he dares to cheat on her again. As
the conversation continues Speaker 1 moves the metaphoric goalposts a little each time so that
she relents a little more on how much she will tolerate from her cheating boyfriend. For
example, she says that her man gets just one chance . . . and then after that, he gets one more.
She goes on to say that she won’t let him cheat in the house . . . but then says he can, just not
when she’s at home. And so on. All the way through, Speaker 2 responds to each conversational
turn with only one word: ‘Okay’. Speaker 2’s meaning, though, varies widely and this is achieved
through different prosody.

Try to transcribe the different ‘okay’ responses from Speaker 2. How does prosody generate
different meanings of the same word?

Our answers are at the end of the chapter.

Syntax
Our knowledge of how words and phrases are conventionally arranged (syntax) also
plays a part in our understanding of what counts as a meaningful unit of information.
Often, when we speak, we do not use complete grammatical sentences. However, the
notion of syntactic completion is important for identifying what is a meaningful piece
of information.
For instance, in example 1, Alina’s frst turn contains two grammatically incomplete
units: ‘Remember Tyke’ and ‘lives next door to Mom’. Although ‘Remember Tyke’
is syntactically complete as an imperative, the rising intonation on ‘Tyke’ means that
it is intended as a question. As such, it is incomplete because it lacks an auxiliary verb
and Subject at the start of the clause (e.g. Do you remember Tyke?). Knowledge of
English grammar alongside the contextual situation (Alina is directing the question at
Lenore) means that the auxiliary verb and Subject are not essential for understanding
and can be elided. With ‘lives next door to Mom’, some sort of grammatical Subject
is missing, but we can still make sense of that utterance and interpret it as a meaningful
unit of language. It is possible, therefore, to make sense of these utterances and
understand them as complete and meaningful, even though they are grammatically
incomplete. However, consider how you would interpret Alina’s opening if she had
said the following:

Do you re↗member (. .) she lives next door ↗to

89
ANALYSING SPOKEN DISCOURSE

The question ‘do you remember?’ invites the hearer to remember something but is
incomplete because the ‘something’ cannot be retrieved from the context or from the
prior discourse (because Alina’s turn is an opening). The question therefore has an
empty slot that cannot be flled by the hearer, and would probably results in a response
such as ‘remember what?’ The second question, ‘she lives next door to?’ is
grammatically incomplete and only makes sense if it is seen as a request for
information (‘she lives next door to who?’). However, ‘she’ has no antecedent (see
Chapter 3), so the hearer is likely to be left wondering ‘who is she?’ In both cases, is
not possible for us to complete these two questions because their incompleteness is
not just a matter of grammar. It is probably possible to think of a situation where, if
the essential information was Given (see Chapter 3) then these questions could make
sense. Notice, then, that the position of an utterance within the conversation is also
key to making and understanding meaning.

Activity 4.4 Try saying this at home

Can you say the following sentence in two different ways to communicate two different
meanings?

Precisely at that point I called a taxi.

Think about what resources you use to change the meaning of the sentence.
Using the codes we introduced above to show pauses and intonation, transcribe the different
ways of saying the sentence.

We give our answer at the end of the chapter.

Turns, turn taking, and turn transition

An obvious observation about the transcription in example 1 that we have not
discussed so far is that it is divided into turns. This refects the fact that the actual real-
life conversation proceeded in that way – the two participants took turns to talk with
Alina speaking frst followed by Lenore, and so on. That the participants took turns to
talk might seem a rather banal remark, but research into spoken discourse has found
that turn taking and its organisation is of fundamental importance in spoken interaction
(Schegloff 2007: 1). In this section we will deal with the notion of turns and turn
taking.

Turns
Simply put, turns consist of words and meaningful sounds (such as ‘mmm’). A turn
can be defned as everything uttered by one speaker before another speaker starts to
speak (Stenström 1994: 4). As discourse analysts, it is an important task to separate a
transcription of a conversation into turns, which we can achieve by noticing when
different speakers are speaking. This can be tricky when there are more than two
90
 ANALYSING SPOKEN DISCOURSE

people involved in a conversation, and analysts must become very good listeners.
Additionally, the boundaries between turns can become blurred when, for example,
the talk of conversational participants overlaps. Turns can vary greatly in size from
one word (or meaningful sound such as ‘mmm’) upwards, with the context of the
conversation (e.g. job interview, casual conversation, classroom interaction)
infuencing the length of speaker’s turns (as well as content and function). For example,
it is likely in a job interview that many of the turns produced by the interviewer will
be short and aim to elicit information from the interviewee, while many of the turns
produced by interviewee will be comparatively long because they will aim to provide
as much relevant information as possible to demonstrate aptitude and suitability for
the vacancy.

Turn taking
Turn taking is, according to Clayman (2013: 150),“an elementary form of social
behaviour” because “interactants must act not as independent agents but in close
coordination with one another”. What Clayman is getting at is that turns are not pre-
determined by the participants before the conversation begins; turns are established
dynamically by the participants as the conversation progresses and so participants
must cooperate to create a dialogue. Participants take it in turns to contribute to the
conversation and tend to talk one at a time. If more than one person does speak at the
same time, then this usually happens very briefy with overlaps occurring in predictable
places (a point we will return to later in the chapter).

Turn transition
The ground-breaking work of Sacks et al. (1974) revealed that when people take turns
at talking, they signal that a potential end to the turn is approaching, and their
interlocutor uses those signals to anticipate the completion of a turn.
They also suggested that there is a tacit understanding that when someone is speaking,
they hold the foor and that transfer to another speaker only occurs at specifc places.
Turn taking is therefore managed by conversational participants, who intuitively
recognise when a turn is coming to an end and when there is the possibility of a
transition to another speaker (Schegloff 2007: 5). Sacks et al. (1974: 703) introduced
the term Transition-Relevance Place (TRP) for the point in an interaction where
another speaker may take a turn and thus, the foor. A TRP and therefore the possible
end of a turn can be signalled by syntactic completion and prosodically via intonation
and pauses. Typically, a rising or falling intonation contour on a potentially fnal word
or syllable in a syntactic unit can indicate a possible place for a transition between
speakers.
Let’s consider how TRPs work by returning to Alina’s frst turn in example 1
(labelled 1c below). We can see that there are two possible TRPs each at the end of an
information unit. Both end in rising intonation indicating that these are places where
91
ANALYSING SPOKEN DISCOURSE

Lenore could take the foor. The pauses strengthen the possibility of these being TRPs.
Transition occurs at the second TRP when Lenore takes the foor (in turn 2) and
answers the question. Lenore’s response has falling intonation contour suggesting it is
complete and completion is reinforced by the pause that follows. At that point, there
is a transition back to Alina. In turn 3, Alina’s frst information unit ends in a falling
tone but is followed by an audible intake of breath (after ‘Okay’). Such inhales can
indicate that the speaker has not fnished speaking and that there is more to come.
Indeed, Alina does have more to say, and Lenore, who stays quiet, correctly does not
see this as a relevant place to take the foor.

1c
1. ALINA: Remember ↗Tyke (.) lives next door to ↗Mom
(1.8)
2. LENORE: ↗↘Yeah
(. .)
3. ALINA: ↘ Okay (.)(H)(.) two weeks ago I’m watching ↘↗TV (.) and David
Horowitz is going to have this former →car (. .) radio thief ↗on
(. .)
4. LENORE: it’s her ↗boyfriend
5. ALINA: (H) (.) ↗↘Yeah=her ex-boy↘friend (.) ↘Mike (. .) he’s the one that
stole (name)’s ↘radio
(2.2)
6. LENORE: How do you ↘know
7. ALINA: ( H) ↗↘Well (. .) cause ↘well (.) →he (.) he was a cocaine ↘addict

Note that turn 3 ends in rising intonation and is followed by a pause. This suggests
that a question is being asked or a response sought and places an obligation on the
hearer to answer (perhaps through politeness – see Chapter 5), which creates a TRP.
Indeed, Alina seems to be handing the foor to Lenore here. The actions we perform
when we speak (questioning, answering, informing, directing, and so on) are therefore
also important for understanding turn transitions and sequences in conversations. We
return to the idea of performing speech acts in the next section.
To summarise, turns are all the words spoken by one person before another person
takes over (Stenström 1994: 4). A turn ends when the person stops (or is stopped from)
talking and someone else starts (although, we will see exceptions to this general idea
later in the chapter). So long as we can recognise that one person has stopped talking
and another person has started then we can work out where one turn ends and another
begins. Within turns, speakers create units of meaning using a combination of words,
syntax, and prosody. The hearer in a conversation knows when a point is approaching
where taking the foor might be appropriate or when it is their turn to talk; the speaker
signals (syntactically and prosodically) that a turn might be ending, or a point is
approaching where a transition is relevant. Conversational participants therefore use
what they hear to ascertain relevant points for taking a turn to talk.

92
 ANALYSING SPOKEN DISCOURSE

Functional analysis of turns

So far, our analysis of spoken dialogue has considered how the prosodic and syntactic
features of talk play a part in creating meaning. We now consider the function of what
people say in a turn because when people speak, they perform actions with their words.
For instance, we saw that in Alina’s frst turn (in 1c above) she performs the action of
asking a question. The importance of considering what actions talk performs is noted
by Schegloff (2007: 2), and the analytical framework for spoken discourse developed
by Sinclair and Coulthard (1975; see also Coulthard 1992) incorporates the idea of
talk as action whereby conversations proceed through predictable sequences of Acts.

Acts and actions

The notion of Acts is derived from the work of Austin (1979) and Searle (1976, 1979)
who developed research on ‘speech acts’. These refer to what we are doing with
language when we use it in certain ways and in certain contexts. Speech acts are
effectively actions that we perform, directly or indirectly, using words. They include,
for example, apologizing, promising, denying, offering, thanking, and threatening.
Sinclair and Coulthard (1975) use ‘acts’ in a similar way to Austin and Searle, but
their Acts3 form part of a grammar of spoken discourse. The number and type of Acts
are potentially open-ended. Both Francis and Hunston (1992) and Stenström (1994)
provide extensive lists that agree in general on the different classes of Acts but differ
in terms of labels and number. Table 4.2 provides a sample of Francis and Hunston
(1992) and Stenström’s (1994) classifcations. Following Stenström (1994), Acts are
enclosed in angle brackets.

Table 4.2 Selected list of Acts after Francis and Hunston (1992) and Stenström (1994)

Act Gloss
<greet> / <part> ‘hello’/‘goodbye’
<reply-greet> / <reply-part> Respond to <greet> / <part>; e.g. echoic ‘hello’ / ‘goodbye’
<offer> Offer somebody something
<request> Request permission or ask someone to do something
<accept> Accept <offer> or <suggest>; grant or agree to <request>
<reject> Turn down <offer>; deny or refuse <request>
<thank> Express thanks/gratitude
<question> Request information: wh-question, yes/no-question
<answer> Respond to <question> providing informative answer.
<inform> Provide information
<suggest> Suggest an idea
<acknowledge> Let the other person know ‘message received’; acknowledge
<thank> or <apology>
93
ANALYSING SPOKEN DISCOURSE

<apology> Say you are sorry

<summon> Attract someone’s attention (e.g. say ‘excuse me’ or use the
person’s name)
<reply-summon> Respond to <summon>
<evaluate> Provide an evaluation or judgement
Acts are verbal actions within turns and indicate what the person is doing by
uttering words with a particular prosody. Let’s return to the frst two turns of the
conversation between Alina and Lenore. These are repeated below but this time we
have labelled the Acts in both turns using categories from Table 4.2.

1d
1. ALINA: Remember ↗Tyke (.) <question>
lives next door to ↗Mom <inform>
(1.8)
2. LENORE: ↗↘Yeah <answer>

Simply put, this is a question-and-answer sequence: Alina asks a question and Lenore
provides an answer. However, in turn 1, Alina performs two Acts, which coincide with
the two speech units that are separated by a pause and end with rising intonation. The
frst functions as a yes/no-question which requests confrmation where either ‘yes’ or
‘no’ is an appropriate response. The second functions to inform Lenore (presumably
to prompt her memory) even though it is said with rising intonation like a question. In
turn two, Lenore gives an appropriate response to Alina’s information request by
providing confrmation.
In example 1d, it is not diffcult to decide what Act is being performed. But it is not
always so clear cut. Searle’s work (1976) on speech acts offers the following to help
identify what Act is taking place and what effects it generates:

1. Locutionary act: the words and prosodic features (i.e. the locution) used by the
speaker.
2. Illocutionary force: the intention of the speaker by uttering the locution.
3. Perlocutionary efect: the efects of the locution on the hearer intended by the
speaker.

Let’s consider the three terms above in relation to the conversation between Alina and
Lenore. In turn 1, Alina utters words in a particular way (locutionary act) which we
judge to have the intended illocutionary force of requesting confrmation from Lenore,
with the intended outcome of receiving the requested information. This appears to be
the way Lenore interprets Alina’s locution, since she provides the sought-for
confrmation. We therefore judge Lenore’s locution to have the illocutionary force of
an informative response. Thus, Alina’s request for confrmation has the desired
perlocutionary effect.

94
 ANALYSING SPOKEN DISCOURSE

Form and function

There is no one-to-one ft between the words uttered (the locution) and the Act they
perform since (i) different words can perform the same function, and (ii) the same
words can be used for different functions. A crucial factor in how an utterance
functions is how and where it is uttered within the discourse and in what context
(physical/social situation). For example, consider the following made-up example.

2
(a) (b)
A: Here’s your pie. A: Here’s your pie.
B: Thank you. (B eats pie)
A: You’re welcome. A: You’re welcome.

In scenario 2(a), the ‘thank you’ said by B offers thanks to A for the pie. The Act of
thanking could have been achieved by any number of different forms such as ‘cheers’,
‘nice one’, or ‘sound’. The utterance ‘you’re welcome’ in 2(a) functions to
<acknowledge> the thanks. In scenario 2(b), however, A does not offer any thanks
and, consequently, ‘you’re welcome’ has a different function. Instead of being a polite
acknowledgement, it is sarcastic and functions to <evaluate> or pass judgement on the
lack of thanks and implies something like ‘I want you to know that thanks were
expected but not received and that did not go unnoticed’ (readers may have their own,
equally valid, thoughts on the Act and the implied meaning). Part of the reason why
‘you’re welcome’ functions as a polite acknowledgement in 2(a) is because it follows
a <thank> Act. Similarly, part of the reason why ‘you’re welcome’ cannot be a polite
acknowledgement in 2(b) is because no thanks have been offered.
Therefore, what happens immediately prior to an utterance affects its meaning, and
one utterance can raise expectations about or predict what the next utterance will be
(in terms of function). A statement of fact functions as a response to a request for
information if it follows a request for information, ‘yes’ only grants permission if it
follows a permission request, and a question can be observed to be recognised and
understood as a question when it is followed by an answer of some nature (Wilkinson
and Kitzinger 2017: 75). The sequential position of an utterance in a conversation is
therefore crucial to assessing its function (see Schegloff 1968: 1083; Stivers 2013:
192). So too are social norms and expectations.

95
ANALYSING SPOKEN DISCOURSE

Activity 4.5 ‘I’d like a pie, please’

Consider the following (made-up) situations where ‘I’d like a pie’ is uttered. Does the statement
have a different function depending on the situation and discourse position it occurs in?

At home
A: What do you want to eat? B:
I’d like a pie please.

At the bakery
A: I’d like a pie, please.
B: Certainly. Here you are.

Now, consider the possible different discourse functions of the following: ‘hello’, ‘excuse me’.
Think of situations where the different functions might occur.

We give our answers at the end of the chapter.

Sequencing
Spoken discourse happens in sequences. Functional elements of dialogue are ordered
in specifc ways and infuence what happens around those elements. For example,
asking a question raises the expectation of an informative response. If the expectation
is not realised (i.e. the response is absent) then this is marked and goes against
expectations (Stivers 2013). In this section we will examine just some of the possible
sequences in spoken discourse identifed by scholars.
Adjacency pairs
The notion of adjacency pairs was introduced by Schegloff and Sacks (1973) and
describes the close sequencing of two turns where the second turn is related to the frst.
More precisely, though, it is the sequencing of the Acts performed in turns that better
describe adjacency pairs: one type of Act is closely followed by another particular and
predictable type. For example, a <question> is usually followed by an <answer> and
this forms a pair where the <question> is the First Pair Part (FPP) and the <answer>
is the Second Pair Part (SPP). The idea is that the FPP predicts the SPP, so when a
speaker asks a question, it usually means that the next utterance will contain an answer
(although, as we will see, there are exceptions to this). There are other Acts that can
be involved in adjacency pair sequences some of which are shown in Table 4.3.

Table 4.3 Examples of Acts that can form adjacency pairs

FPP SPP

96
 ANALYSING SPOKEN DISCOURSE

<greet> <reply-greet>
<offer> <accept> / <reject>
<request> <accept> / <reject>
<summon> <reply-summon>

Let’s look at an example, this time from the spoken section of the BNC. In extract
3, the context is that a mother is interacting with two children who are doing different
activities; her focus is on just one child when the other child speaks.

3
1. CHILD: ↘↗Mu:::m <summon>
2. MOTHER: Hell↗o: <reply-summon>
(BNC spoken section, fle KBW)

In turn 1, the child attracts the mother’s attention. This is an example of a

<summon>/<reply-summon> sequence where turn 1 is the FPP and turn 2 the SPP.
Turns 1 and 2 contain one Act each. The child’s locution has the force of summoning
or attracting her attention. The mother responds to the <summon> with ‘hello’. Notice
that the position in the sequence and the preceding Act (along with the rising
intonation) helps to determine that ‘hello’ is not functioning as a greeting in this case.

Support Acts
In example 3, the turns contain just one Act each and they form a neat pair, with one
Act following immediately after the other. However, you do not need to look at very
many spoken data to see that this is not always what happens. Our Alina and Lenore
example is a case in point. In example 1d above, we saw that the sequence involves a
recognisable <question> and <answer> pair. However, we also saw that, along with
the <question> Act, Alina’s frst turn contains an <inform> Act. This expands the
sequence. The <question> is of primary importance in the sequence – it takes the lead.
Without the <question> there would be no sequence (imagine if Alina had started with
‘lives next door to mom’). The <inform> has a supporting role; it assists in Alina’s
conversational goal of getting an informative response by offering relevant
information to help Lenore answer. The sequence in 1d therefore involves a pair of
Acts: a primary Act crucial for creating a sequence, and a secondary Act that helps
to achieve the conversational goal (for more on primary and secondary Acts, see
Stenström 1994: chapter 2).

Sequence expansion
Sequences in conversation can be more complex than so far described. For example,
one or more adjacency pairs can be embedded within the initial pair sequence in what
is known as insert expansion. For example, a request for information might be
97
ANALYSING SPOKEN DISCOURSE

followed by another request seeking clarifcation (Schegloff 1972: 76–79) which is

answered before the original request is answered. We can see how this works in
example 4, which shows more of the conversation between the mother and child we
met in example 3.
4
1. CHILD: ↘↗Mu:::m <summon>
2.MOTHER: Hell↗o: <reply-summon>
3. CHILD: Could you get me from the ↘shelf (.) the <request FPP1>
black felt ↘pen
4.MOTHER: Which ↘shelf <question FPP2>
5. CHILD: The big one with all the (unclear) on <answer SPP2>
↘top (.) there’s some ↘colours (.) there
isn’t black (. .) felt ↘pen

6.MOTHER: ↘Right (. .) I’ll see what I can →do. <accept SPP1>

(BNC spoken section, fle KBW)
As we established above, the frst two turns involve a <summon> and a
<replysummon>. Following that, in turn 3, the child makes a <request> – they want
their mother to do something. This forms the FPP of an adjacency pair. The SPP,
however, is in turn 6 when the mother agrees to the <request> with an <accept>. In
between this not-so-adjacent adjacency pair is an insertion sequence which is itself
an adjacency pair. Its purpose is to establish more information about the question and
its insertion suspends the SPP sought in turn 1. Once the mother gets the information
she needs, then she agrees to the <request>. It is, of course, possible for more
sequences to be inserted in between the initial question and the fnal answer. For
example, the mother could ask the child to say ‘please’ (which would be a <request>),
and only when the ‘please’ was forthcoming would she issue the SPP for the child’s
original <request>.

Sequence Coupling
As we saw with our Alina and Lenore example above, turns can contain more than
one Act; Alina elicited information from Lenore by issuing a <question> and an
<inform> Act. It is also possible for the Acts in one turn to conclude one sequence
and start another. In extract 5, obtained from the BNC, we can see in turn 1, Heidi’s
second Act is a <request> for permission (‘please may I . . .’) and constitutes a FPP in
a <request>/<accept> adjacency pair. In turn 2, Vicki performs an <accept> by giving
permission (with ‘Mhm’) and so provides a SPP to conclude that sequence. Vicki’s
turn contains a second Act which is a <request> (‘Would you like to . . .’) which
initiates another sequence. Notice that the <request> is in the form of a question which
seeks information about what Heidi would ‘like’. This is an indirect way of saying
‘get it yourself’ and is therefore requesting action (we discuss indirect ways of asking

98
 ANALYSING SPOKEN DISCOURSE

for things in Chapter 6). In turn 3 Heidi’s ‘Mhm’ functions as an <accept> and
completes the sequence with an acceptable SPP.

5
1. HEIDI: Excuse me ↘mum (. .) <summon>
↑please may I have some ↓more <request>
2. VICKI: ↘↗Mhm (. .) <accept>
Would you like to help yourself ↘love <request>
3. HEIDI: ↘↗Mhm <accept>
(BNC spoken section, fle KC3)

Pairs with three parts

Conversational sequences can have more than two parts. For example, consider
example 6 from the Santa Barbara Spoken Corpus.

6
[Door opens and closes; Karen enters room]
1. KAREN: Hi ↘sweetie (. <greet>
2. SCOTT: .) ↘Hey <greet-reply>
3. KAREN: Sweetie ↘frumptious <acknowledge>
(SBC034, What Time is it Now?)
In 6, Karen’s frst turn contains the Act of <greet>, where she says ‘Hi’ to her partner
on her return home from work. Scott’s turn contains a <greet-reply> where he
responds to the <greet> with ‘Hey’. In the third turn, Karen continues the greeting
with an <acknowledge> which continues the greeting with a further term of
endearment. In Conversation Analysis, this added extra is known as post expansion
of the adjacency pair. As <greet>/<reply-greet> conventionally form an adjacency
pair, the frst two Acts are predictable, but the third Act is not. The FPP is expected
given the situation (arriving home) and predicts the SPP; it would be unusual if the
<reply-greet> were not present. However, if Karen’s extra acknowledgement were
absent, this would not be marked.
There are situations, though, where post expansion of a sequence might be expected
and the lack of it marked. For example, if you are providing important information
(such as bank card details) following a <request>, then the lack confrmation might
lead you to wonder if the information was received. Similarly, the lack of thanks
following a <grant> of permission might not go unnoticed. The point is, though, that
while these third parts might be desired, their predictability (or otherwise) rests on the
circumstances of the exchange. For instance, Sinclair and Coulthard (1975) found that
in teacher–pupil interactions, where teachers ask questions to test pupil knowledge,
some sort of confrmation of the correctness (or otherwise) of the answer is required
by the teacher and is predictable in that context.

99
ANALYSING SPOKEN DISCOURSE

Activity 4.6 Adjacency pears

Consider the following two extracts taken from the spoken section of the BNC. Both examples
contain pears, but can you see any adjacency pairs or any other sequences? For each turn, say
what Acts are being performed and mark FPPs and SPPs.

(i)
1. CHRISTOPHER Why do you always have pears mummy?
2. DOROTHY Cos I just fancy them at the moment.
(BNC spoken section, fle BNC KBW)

(ii)
1. SIMON What can I take? (. .)
2. SANDRA Well you were going to take a satsuma and some Quavers weren’t you?
3. SIMON Yeah.
4. SANDRA We haven’t got a satsuma, what about a pear, packet of Quavers and a
pear?
(. .) Do you like pears or 5.
SIMON Yeah.
(BNC spoken section, fle BNC KDW)

We give our answers at the end of the chapter.

Preferred/dis-preferred responses
The initiating Act of an adjacency pair can often present different response options.
For example, an <offer> can be accepted or rejected. The different alternatives are
valued differently and are known as preferred and dis-preferred responses.Whether a
response is either preferred or dis-preferred is not motivated psychologically by the
interlocutors – i.e. not what the people involved in the conversation prefer, which will
differ from conversation to conversation. Rather a preferred response is one that will
enable the accomplishment of the action that the sequence of turns is aiming to
achieve, while the alternative is dis-preferred (Schegloff 2007: 59). For example, if
we assume that a successful offer is one that results in the giving of whatever is being
offered, then the acceptance of an offer is preferred over a rejection since this enables
the completion of the action. Dis-preferred responses tend to be more linguistically
complex because they often involve giving reasons for the response, which can
necessitate a longer turn than a preferred response. (In Chapter 7 we will see that dis-
preferred responses also link to politeness.)

Backchannels
When we are involved in spoken dialogue, participants expect some sort of audible
oral feedback from the other, no matter how minimal. Try keeping silent next time
100
 ANALYSING SPOKEN DISCOURSE

you are on the phone and see how long it takes for the other person to say something
along the lines of ‘Are you still there?’. Such feedback during spoken conversations
is known as backchannelling and is crucial for keeping the conversation going.
Backchannels inform the speaker that their interlocutors are still involved in the
conversation and include noises such as ‘ah’, ‘oh’, ‘mmm’, laughter, and words such
as ‘yes’, ‘sure’, ‘really’. Consider the following example from the SBCSAE.

7
1. FRED: last night I got into a hassle with James ↘Bo:yd (. .) (H) I’m in
the ↗cafeteria [and I] took a break (.) that was just a ↑LITTLE
BIT↑ too long ↘man (.) you ↘know
2. RICHARD: [yeah]
(SBC047, On the Lot)

Note: The up arrow (↑)indicates a step up in pitch. We have marked a step up across
words by enclosing the words with up arrows (↑this is all higher pitch↑).
Example 7 demonstrates that backchannels can overlap with an ongoing turn. In the
example, Fred is telling Richard about his shift in the factory, and his turn is
continuous from ‘last night’ to ‘you know’. Our transcription aims to show that
Richard saying ‘yeah’ overlaps with Fred saying, ‘and I’ by enclosing the overlapping
words in square brackets.

Overlapping talk
Sidnel (2010: 52) notes that there is a widespread view that people in conversation
interrupt and talk over one another. However, evidence from recorded and transcribed
data shows that this is not the case. Instead, overlapping talk tends only to occur at
Transition Relevance Places (see above) when one speaker’s turn could be coming to
an end. At that point, the other speaker might start talking a fraction of a second before
the previous turn ends, as they sense the other person is fnishing (Sidnel 2010: 52).
Such overlaps are thought to help the conversation advance because they indicate that
speakers are taking notice of and responding to what the other is saying (Sidnel 2010:
53). In example 8, Richard’s contribution in turn 3 overlaps with Fred’s talk, but is
supportive because it shows Richard pre-empts what Fred is about to say and joins in
with the ‘telling’, which demonstrates Richard’s engagement with Fred’s story.

8
1. FRED: (H) and then he comes into the ↘cafeteria (H) and I thought he was coming
in to chase everybody ↘away
2. RICHARD: [He was after you].
3. FRED: [(H) But he’s] coming after me.
(SBC047, On the Lot)

Note: Overlapping talk is indicated by square brackets.

101
ANALYSING SPOKEN DISCOURSE

The overlap occurs at a possible TRP because Fred’s frst turn is syntactically
complete and ends with a fall in pitch (a terminal intonation contour).
However, speakers can also talk over each other to retain ownership of the
conversational turn and control the conversation, but such incidents are marked and
noticeable. Interruptions are different from overlaps because (i) they tend not to occur
at Transition Relevance Places and (ii) rather than helping the conversation proceed
they can stop conversational actions progressing and signal that the speaker is
declining to participate.

Summary of transcription conventions

Analysing spoken discourse relies on accurately transcribing the spoken data in a way
that records the features that are important for meaning making. The following are just
some of the transcription codes used in CA. Many of the symbols in the table were
originally developed by Gail Jefferson (see Jefferson 2004) and have become the
default notation used by analysts. The idea is that the system provides details of vocal
production of spoken utterances. The overview in Table 4.4 is based on a combination
of Jefferson (2004) and the Discourse Transcription System in Du Bois et al. (1992,
1993).
Table 4.4 Summary of transcription symbols

Pauses between words and utterances

(.) dot in round brackets Small pause, 0.2 seconds or less
(. .) Two dots in round brackets Medium pause 0.3–0.6 seconds
(0.0) numbers in round brackets Timed pause 0.7 seconds and over
let’s equal signs Latching. No pause between turns
go==okay or between information units within a
turn.
Let’s [go] Single left and right square Start and end of overlapping
[okay] brackets speech
Characteristics of speech production
wor– dash/hyphen Cut-off
wo:rdwo:::rd Colon Prolonged vowel or consonant sound.
Multiple colons indicate longer
sound
word Underlining Emphasis or stress
WORD underlining + upper case Loud talk relative to the surrounding
talk
°word° degree sign (shortcut – Quiet talk relative to the
Alt 248) surrounding talk

<word> Left/right angle brackets Talk slowing down

>word< Right/left angle brackets Talk speeding up

102
 ANALYSING SPOKEN DISCOURSE

(H) Uppercase H in brackets Inbreath (more hs for longer

duration)
(Hx) Uppercase H lowercase x in Outbreath (more hs for longer
brackets duration)

w(h)ord bracketed h (or hs) within a word Breathiness while talking –

usually laughter within the word
↗ Upward diagonal arrow (short Rising tone
cut – 2197 followed by Alt x)
↘ Downward diagonal arrow (short Falling tone
cut – 2198 followed by Alt x) Upward
↗↘ and downward diagonal arrow Rise-fall

↘↗ Downward and upward diagonal arrow Fall-rise

→ Level arrow (shortcut – 2192 Level tone

followed by Alt x) Up
↑ arrow (2191 Alt x) Step up in pitch
↓ Down arrow (2194 Alt x) Step down in pitch
Transcriber notes
() Empty brackets inaudible/unintelligible word
(word) Bracketed word uncertain word
((sneeze)) Double brackets comments or descriptions of non-
((leaves verbal activity or sounds that are
room)) not accounted for by other symbols

Conclusion
In Chapters 2 and 3, we explored written discourse and examined how information
structure and meaning making is a matter of graphology and syntax. In this chapter,
we have examined how in spoken discourse it is a matter of phonology and syntax
(Brown and Yule 1983). What we have discussed covers only some of the many
features of spoken discourse. Its dynamic nature means that there are lots of methods
at our disposal for making meaning, many of which are constantly evolving and
adapting to social and interpersonal needs. These methods incorporate lexis,
semantics, syntax, pragmatics, and prosody, as well as turn taking and sequencing.
The dynamic nature of spoken discourse also means that interlocutors must pay
attention to linguistic and non-linguistic elements in interactions. These non-linguistic
factors can include hierarchical relationships concerning power and social distance, as
well as cultural differences and social expectations (see Chapter 7). In subsequent
chapters, we will look in more detail at this interplay between linguistic and extra-
linguistic phenomena.

103
ANALYSING SPOKEN DISCOURSE

Further reading
For an excellent practical guide to intonation with a companion website which
includes many useful audio fles, see Collins et al. (2019). There are numerous
introductions to CA; for example, Sidnel (2010) offers a comprehensive introduction
that covers the key elements of CA in an approachable and accessible style. Stenström
(1994) provides an accessible guide to analysing spoken discourse that combines the
ideas from CA and from the Birmingham School (Sinclair and Coulthard 1975).

Resources
Santa Barbara Corpus of Spoken
American English (SBCSAE) contains QR 4.3 SBCSAE link
approximately 250,000 words of spoken (https://www.linguistics.ucsb.edu/research/
santa-barbara-corpus)
American English from a variety of
different situations, including
face-to-face family chats, telephone conversations, and business meetings. Sound fles
can be found at www.talkbank.org in the ‘Conversation Banks’ under ‘CABank’.
BNC corpus of spoken and written English
contains 100 million words of spoken and QR 4.4 BNC link
written British English from the early (http://www.natcorp.ox.ac.uk/)
1990s: 10 million words spoken data; 90
million written.
Follow QR 4.5 for details of a webbased
QR 4.5 Web-based interface link interface to the BNC, including how to
(http://corpora.lancs.ac.uk/BNCweb/ sign up to use the corpus for free.
index.html)

Answers to activities
Activity 4.1 Punctuation in spoken discourse
We asked you think about whether there were any issues with the basic premise of
Borge’s comedy routine, in which he reads out loud from a book and indicates the
punction in the writing with sounds effects. However, as this chapter explains, we
already do that in a way using pauses and intonation, so no additional sound effects
are really needed.

104
 ANALYSING SPOKEN DISCOURSE

Activity 4.2 Pause relocation

The effect of a pause after ‘remember’ creates an imperative structure whereby Alina
instructs (or commands) Lenore to remember, and then provides the details of what
must be remembered. We could also see this as Alina telling Lenore something (that
Tyke lives next door to mum) rather forcibly.

Activity 4.3 Okay

Key and Peele’s ‘Okay’ sketch shows how one lexical item can, through prosodic
manipulation, communicate a variety of diverse attitudes towards Speaker A’s
propositions. For example, Speaker B begins by pitching her ‘Okay’ in a higher tone
(↑Okay↑) which could be said to signal support when she acknowledges her
interlocutor’s decision not to take any more crap from her cheating boyfriend. Later,
when A concedes that she will take some crap, but not all of it, B’s ‘Okay’ is
pronounced as a drawled ‘Oka:::::y’ with the vowel sound [eɪ] lengthened
considerably. There is some overlap as the sketch progresses, but as B starts to
question A’s approach to her boyfriend’s cheating, her ‘Okay’ changes from long to
short until she fnally interrupts exclaiming ‘↗O (.)↘KAY!’ with her two hands up as
if to stop her friend from speaking further. She has, at this point, clearly had – and
heard – enough. Then the cheating boyfriend walks in looking all buff and with a smile
that would melt an ice cream on the spot and she’s all like ‘Oka:: ↗y’. As speaker A
acknowledges, her girl ‘been speaking words of wisdom all night’.
Activity 4.4 Try saying this at home
The two ways we had in mind to break the utterance into two units in different places
are:

↘Precisely (.) at that point (.) I called a taxi Precisely at

that ↘point (.) I called a taxi

The difference is whether ‘precisely’ acts on its own as an agreement response or

modifes ‘at that point’.

Activity 4.5 ‘I’d like a pie, please’

At home
A: What do you want to eat? [request for information/offer: I will get you something
to eat]
B: I’d like a pie please. [informative response/acceptance of offer: get me a pie!]

The inference that an offer has been made might be cancelled if A says something like
‘then you’d better go to the bakery and buy one’.
105
ANALYSING SPOKEN DISCOURSE

At the bakery
A: I’d like a pie, please [request for action: sell me a pie!]
B: Certainly. Here you are. [acknowledgement of request + appropriate action]

‘hello’ can function as a greeting and as a way to attract someone’s attention. ‘excuse
me’ can be used to attract attention, request action (move!) and apologise.

Activity 4.6 Adjacency pears

(i)
1.CHRISTOPHER Why do you always have pears ↗mummy <question> FPP
2. DOROTHY Cos I just fancy them at the ↘moment. <answer> SPP

<question> / <answer> adjacency pair.

(ii)
1 SIMON What can I ↗take (. .) <question> FPP1
2 SANDRA Well you were going to take a satsuma and <question> FPP 2 some
Quavers weren’t ↘you
3 SIMON ↗↘Yeah <answer> SPP2
4 SANDRA We haven’t got a ↘satsuma (.) <inform>
what about a ↘pear = packet of Quavers <suggest> SPP1
and a ↘pear FPP3
5 SIMON mmm <accept> SPP3
6 SANDRA (.) Do you ↗like ↘pears ↘though <question> FPP4
7 SIMON ↘↗Yeah <answer> SPP4

<question> / <answer> adjacency pair with <question> / <answer> insertion sequence.

Note: <inform> in turn 4 provides a follow-up to the FPP2 and SPP2 and frames the
response to the initial <question> in turn 1.The response in SPP 1 is a <suggest> which
both closes the frst sequence and initiates another. Simon’s ‘mmm’ seems to <accept>
the <suggest> which closes that sequence.

Notes
1 There are also visual clues, such as gaze and gesture, that we don’t have space to explore in this
book.
106
 ANALYSING SPOKEN DISCOURSE

2 An alternative method – adopted by DuBois et al. (1993) – is to use backslash (\), forward slash
(/) and underscore (_).
3 Act has an initial capital letter to show its status as a technical term within Sinclair and
Coulthard’s (1975) discourse grammar.

References
Austin, J. L. (1979) ‘Performative utterances’. In J. O. Urmson and G. J. Warnock
(eds), Philosophical Papers (3rd edition), pp. 233–252. Oxford: Oxford
University Press.
Brown, G. and Yule, G. (1983) Discourse analysis. Cambridge: Cambridge University
Press.
Carr, P. (2012) English phonetics and phonology: An introduction. Oxford: John
Wiley & Sons.
Clayman, S. E. (2013) ‘Turn-constructional units and the transition-relevance place’.
In J. Sidnell and T. Stivers (eds), The handbook of conversation analysis, pp. 150–
166. Oxford: Wiley-Blackwell.
Collins, B., Mees, I. M., and Carley, P. (2019) Practical phonetics and phonology: A
resource book for students. London: Routledge.
Coulthard, M. (ed.) (1992) Advances in spoken discourse analysis. London:
Routledge.
Crystal, D. (1979) ‘Neglected grammatical factors in conversational English’. In
Sidney Greenbaum et al. (eds), Studies in English linguistics for Randolph Quirk,
pp. 153–166. London: Longman.
Du Bois, J. W., Cumming, S., Schuetze-Coburn, S., and Paolino, D. (eds) (1992)
Discourse transcription. Santa Barbara Papers in Linguistics 4. Santa Barbara:
University of California, Department of Linguistics.
Du Bois, J. W., Schuetze-Coburn, S., Cumming, S., and Paolino, D. (1993) Outline of
discourse transcription. In Jane A. Edwards and Martin D. Lampert (eds), Talking
data: Transcription and coding in discourse research, pp. 45–89. Hillsdale, NJ:
Lawrence Erlbaum.
Du Bois, J. W., Chafe, W. L., Meyer, C., Thompson, S. A., Englebretson, R., and
Martey, N. (2000–2005) Santa Barbara corpus of spoken American English,
Parts 1–4. Philadelphia, PA: Linguistic Data Consortium.
Francis, G. and Hunston, S. (1992) ‘Analysing everyday conversation’. In M.
Coulthard (ed.), Advances in spoken discourse analysis, pp. 123–161. London:
Routledge.
Jefferson, G. (2004) ‘Glossary of transcript symbols with an introduction’. In G. H.
Lerner (ed.), Conversation analysis: Studies from the frst generation, pp. 13–31.
Amsterdam: John Benjamins.
Jones, D. (1960) An outline of English phonetics (9th edition). Cambridge: Heffer.

107
ANALYSING SPOKEN DISCOURSE

Key, K. M. and Peele, J. (2015) ‘OK’. Uncensored. Series 5, Episode 6. Comedy

Central.
Sacks, H., Schegloff, E. A., and Jefferson, G. (1974) ‘A simplest systematics for the
organisation of turn-taking for conversation’. Language, 50, 696–735.
Schegloff, E. A. (1968) ‘Sequencing in conversational openings’. American
Anthropologist, 70, 1075–1095.
Schegloff, E. A. (1972) ‘Notes on a conversational practice: formulating place’. In D.
N. Sudnow (ed.), Studies in social interaction, pp. 75–119. New York: Free Press.
Schegloff, E. (2007) Sequence organisation in interaction: A primer in conversation
analysis. Cambridge: Cambridge University Press
Schegloff, E. and Sacks, H. (1973) ‘Opening and closings’. Semiotica, 8(4), 289–327.
Searle, J. (1976) ‘A classifcation of illocutionary acts’. Language in Society, 5(1), 1–
23.
Searle, J. (1979) Expression and meaning. Studies in the theory of speech acts.
Cambridge: Cambridge University Press.
Selting, M. (2000) ‘The construction of units in conversational talk’. Language in
Society, 29(1), 477–517.
Selting, M. (2010) ‘Prosody in interaction: State of the art’. In D. BarthWeingarten,
E. Reber, and M. Selting (eds), Prosody in interaction, pp. 3–40. Studies in
Discourse and Grammar, vol. 23. Amsterdam: John Benjamins,.
Sidnel, J. (2010) Conversation analysis: An introduction. London: John Wiley and
Sons.
Sinclair, J. and Coulthard, M. (1975) Towards an analysis of discourse. Oxford:
Oxford University Press.
Stenström, A-B.(1994) An introduction to spoken interaction. London: Longman.
Stivers, T. (2013) ‘Sequence organisation’. In J. Sidnell and T. Stivers (eds), The
handbook of conversation analysis, pp. 191–209. Oxford: Wiley-Blackwell.
Szczepek Reed, B. (2010) ‘Intonation phrases in natural conversation: A participants’
category?’ In D. Barth-Weingarten, E. Reber, and M. Selting (eds), Prosody in
interaction, pp. 191–212. Studies in Discourse and Grammar, vol. 23.
Amsterdam: John Benjamins.
Walker, G. (2013) ‘Phonetics and prosody in conversation’. In J. Sidnell and T. Stivers
(eds), The handbook of conversation analysis, pp. 455–473. Oxford: Wiley
Blackwell.
Wilkinson, S. and Kitzinger, C. (2017) ‘Conversation analysis’. In C. Willig and W.
Stainton Rogers (eds), The SAGE handbook of qualitative research in psychology,
pp. 74–92. London: Sage Publications.

108
 Introduction
Over this and the next chapter we are going to explore how we communicate meaning
through language. Our exploration will cover topics that are typically associated with
the linguistic subdisciplines of semantics and pragmatics, where the former is usually
glossed as the study of meaning of words and sentences and the latter the study of
meaning in context. The pragmatic meaning of an utterance might vary from context
to context, so is unstable and depends on its environment to make meaning. The
semantic meaning is more stable and tends to be tied to linguistic form and co-text
rather than contextual factors. Meaning is derived from several semantic and
pragmatic factors in combination, some of which we introduced in Chapter 1. This
chapter begins by outlining different types of meaning before going on to discuss
entailment and presupposition.

What do we mean by meaning?

The notion of meaning is far from simple, and the idea of semantic and pragmatic
meaning recognises that understanding what a piece of language means concerns not
just what is encoded by the language itself (the words and their order), but other factors
such as the social situation in which the language occurs. Dictionaries can give us a
range of meanings that a word is typically used to signify, but we only know a word’s
meaning once it is used in a specifc context. Back in 1923, Ogden and Richards
explored the idea that words had a variety of different meanings and suggested 22
possible types of meaning. Leech (1981) proposes a wholly more manageable seven,
namely conceptual (or denotative), connotative, affective, collocative, refective,
social, and thematic. The basic idea of Leech’s framework is that a word (or piece of
language) has a conceptual meaning (i.e. it denotes something) but also carries with it
various associations (connotative, affective, collocative, refective, and social). It may
also carry with it, as we saw in Chapter 2, thematic meaning depending on where the
word (or piece of language) occurs in the clause. The meanings in Leech’s framework
are

DOI: 10.4324/9781003351207-5

109
 ANALYSING MEANING IN DISCOURSE

summarised in Table 5.1 and over the rest of this section we will consider six of
Leech’s seven in more detail below (thematic meaning is not discussed here because
we dealt with that in Chapter 2).

Table 5.1 Leech’s seven meaning types (after Leech 1981: 23)
Meaning type Gloss

Conceptual Denotative or ‘dictionary’ meaning

Connotative Experiential meaning relating to the signifed concept
Affective Feelings and attitudes of speaker/writer
Associations Social Meaning to do with social circumstances
Refective Meaning transfer from other senses of same expression
Collocative Meaning transfer from co-occurring words
Thematic Organisation of the message (see Chapter 2)

Conceptual meaning
What do you do if you want to know what a word means? Probably, you look it up in
a dictionary. This gives us a word’s conceptual meaning. At the heart of conceptual
meaning is the study of ‘signs’ (known as semiotics). In short, language is comprised
of signs; which are words together with their meanings. All linguistic signs, therefore,
must have a label and a concept to have meaning, as we will now explain.

Signifier, signified, and referent

Think of a word and you will automatically think of what the word means. The word
together with its meaning constitutes a sign. If you have no meaning for a word, the
word is simply a ‘signifer’ with no ‘signifed’ concept (yet). Let’s consider the word
‘chair’ and what it means (see Figure 5.1).

Figure 5.1 This is a chair

When you were very young you probably learnt how to read using books with
sentences very similar to the one in Figure 5.1 with a matching picture. The word ‘c-
h-a-i-r’ is an arbitrary collection of letters in a particular order (if written), or an
arbitrary collection of sounds in a particular order (if spoken), that English writers and
110
 ANALYSING MEANING IN DISCOURSE

speakers use to signify the concept ‘chair’. The collection of letters is arbitrary because
there is nothing inherently ‘chair-y’ about the word ‘chair’. Furthermore, in languages
other than English, different combinations of letters signify the same concept of
‘chair’, for example, chaise, sedia, cathaoir, stoel, oche, kiti, and tuoli (see Table 5.2).
The combination of the word (or the sounds we make when we say the word) with the
concept it signifes constitute the ‘sign’ (Saussure 1983). As well as the sign, there are,
of course, the real-world entities or things (physical chairs) to which the sign refers,
and these are known as the referent. These realworld entities, or referents, will share
a set of qualities that make them chairs (e.g. a surface to place your backside,
something to support your back). To summarise:

• The word ‘c-h-a-i-r’, known in semiotics as the signifer.

• The concept ‘chair’, known as the signifed.
• The sign is the totality of the signifer and the signifed.
• The sign ‘chair’ can refer to real-world entities/things.
• The real-world ‘chair’ is the referent.

The signifer and signifed are mental constructs, that is, they are conceptual. So, the
sign itself is conceptual. The referent is a real-world example of the thing or concept
signifed by the signifer. The sign for ‘chair’ (as in Figure 5.1 above) will have at least
as many different signifers as there are languages, as Table 5.2 demonstrates.

Table 5.2 Some signifers for ‘chair’

Signifer Language

Stoel Dutch
Kurasee Hindi
Scaun Romanian
Cathaoir Irish
Kerusi Malay
Sandalye Turkish
Stól Icelandic
Cadira Catalan
Kursi Somali
Chaise French
Of course, what constitutes a ‘chair’ in the real world, in this case ‘a piece of
furniture used for sitting on’, may differ depending on where you are in the world.
Indeed, there may well be as many kinds of chairs as there are words for ‘chair’, but
there will be some essential features that make the different chairs, chairs. The point
is that the word–concept (signifer–signifed) relationship is arbitrary. This is the case
for most words in any language; the exception is onomatopoeic words which are
iconic in that they acoustically resemble the concepts they represent.

111
 ANALYSING MEANING IN DISCOURSE

So how do we know that ‘c-h-a-i-r’ means ‘chair’ in English? The answer is

convention. This is the word conventionally used to mean ‘chair’. Amazingly, we have
learnt these conventional referential relationships for thousands upon thousands of
words. This is what it means to have linguistic competence. Moreover, we can use
them and combine them with other words to make meaningful utterances. This is
evidence of our linguistic performance. We even know words for things for which
there are no real-world referents. If we were to ask you ‘Have you ever seen a
unicorn?’ you will reply negatively (we hope) but would have no trouble thinking of
the signifed concept (a horse-like creature with a single horn in the middle of its
forehead) triggered by our use of the signifer ‘u-n-i-c-o-r-n’. Similarly, if we ask you
about ‘love’, you will know what we mean even though you could not point to the
concept or real-world referent in the same way you can point to a real-life chair.
Meaning is not simply about signifers and what is signifed, however. Take a word
like ‘apartment’, that, according to the Oxford English Dictionary (OED), has the
following conceptual meaning (i.e. what it denotes):

a fat, typically one that is well appointed or used for holidays.

So, presumably ‘fat’ means the same thing? Not quite. The OED cites the meaning of
‘fat’ as follows.

A set of rooms forming an individual residence, typically on one foor and within a
larger building containing a number of such residences.

So, when do we use ‘fat’ and when do we use ‘apartment’? What is the difference
between a ‘fat’ and an ‘apartment’?1 The answer lies partly in social attitudes as well
as associations that the word carries with it, and this is what we discuss over the rest
of this section in relation to Leech’s other types of meaning.

Connotative meaning
In addition to conceptual meaning (those dictionary defnitions we just talked about)
meaning is also achieved through association. This is how connotative meaning
comes about. According to Leech (1981: 12), connotative meaning is “the
communicative value an expression has by virtue of what it refers to, over and above
its [denotative] content” and is derived from the “‘real world’ experience one
associates with an expression” (Leech 1981: 13). For example, the word ‘mother’
could be said to connote ‘nurturer’ or ‘carer’, but this is not part of the denotative
meaning of ‘mother’; rather they are further qualities that we expect ‘mother’ to have.
Therefore, conceptual (denotative) meaning of concepts can be elaborated upon and
imbued with connotative meanings that derive from our own knowledge of them in
the real world. The differences between ‘fat’ and ‘apartment’ might well depend on
our accumulated perceptions of the concepts gleaned from our social experiences as
well as our conceptual or denotative understanding of them. For example, if the
112
 ANALYSING MEANING IN DISCOURSE

environments in which we encounter ‘apartment’ consistently show more positive

realisations of the concept than ‘fat’, then we might tend to consider ‘apartment’ as
connoting something more socially attractive, possibly luxurious.
What the fat/apartment example demonstrates is that linguistic expressions (phrases
or lexical items) trigger access to what is known as encyclopaedic knowledge, which
includes knowledge of denotation (dictionary knowledge) as well as a large range of
other knowledge (see Evans 2019: 367, 387). When we think of a signifed entity, we
often think of the most common or prototypical example of that signifed entity. A
prototypical example of either a ‘fat’ or ‘apartment’ might be ‘a self-contained housing
unit that occupies part of a building’. However, the connotations of ‘apartment’ may
well differ from that of ‘fat’. Where we grew up (in the north of Ireland and England),
‘apartment’ had more positive connotations than ‘fat’ because the former was
associated with a more affuent, professional socioeconomic community, whereas the
latter was typically associated with a lower socioeconomic community.
The history of a word (its etymology) can also help explain why some words have
more positive connotations than others. Historically, Latinate lexis carries more social
prestige than Germanic terms2 and so it may not surprise you to note that ‘apartment’
is Latinate and ‘fat’ is Germanic. Note, though, because connotative meanings
(connotations) emerge over time, they are “relatively unstable”, “indeterminate”, and
“open-ended” (Leech 1981: 13).

Affective meaning
Affective meaning relates to the attitude of the text producer towards what is being
talked about and can be expressed denotatively or connotatively (Leech 1981: 15).
Think about the differences between referring to your neighbour’s canine friend as a
‘dog’ or a ‘mutt’, or referring to a person as being ‘discerning’ or ‘picky’. Whichever
we choose can say something about what we think of those characteristics or the
people who exhibit them. Similarly, terms like ‘expat’, ‘migrant’, ‘asylum seeker’,
and ‘refugee’ can say a lot about a speaker’s/ writer’s ideological positioning (see
Chapter 8 for more on this idea). Whatever the topic, the lexical choices we make
indicates something about our feelings or attitudes towards that topic. Of course, it is
not just the content of the discourse about which we have affective responses – our
choices can also refect our attitudes towards interlocutors, too. As we saw in Chapter
1, we might use more standard forms when addressing someone with whom we are
unfamiliar, e.g. ‘Please would you mind lowering the volume of your music?’,
whereas we might tell our fatmate to ‘Turn that racket down’ (we develop this example
in Chapter 7).

113
 ANALYSING MEANING IN DISCOURSE

Social meaning
This type of meaning is to do with the “social circumstances” (Leech 1981: 14)
associated with certain words or expressions. Social circumstances include dialectal
differences which relate to geographical location (regional or global). For instance, in
the north of England, in some regions a pathway between buildings is referred to as a
‘snicket’ while in others a ‘ginnel’. In other (global) locations, the word ‘alley’ or
‘alleyway’ is used. Other social circumstances include interpersonal relationships
(social relationships), discipline/domain (e.g. law, medicine, football), social activity
(e.g. lecture, consultation, sports commentary), and time period (e.g. 18th century).
Words and expressions may also carry with them associations to do with formality,
where some words can sound offcial, while others are colloquial or slang (e.g.
‘dwelling’, ‘house’, ‘pad’, ‘crib’). Additionally, some may carry associations of social
status (or class) (e.g. ‘grub’, ‘luncheon’). Any or all these different types of social
meaning may be carried by a word at the same time. Test your understanding of social
meaning by trying Activity 5.1.

Activity 5.1 Social meaning and synonyms

Consider what social (and other) associations are carried by these sets of (near) synonyms.

‘wee’, ‘small’, ‘diminutive’,

‘fat’, ‘apartment’, ‘condominium’
‘lunch’, ‘dinner’, ‘tea’, ‘luncheon’
‘tradie’, ‘tradesman’, ‘tradesperson’, ‘navvy’

We give our answers at the end of the chapter.

Reflected meaning
Many words and expressions have more than one conceptual meaning, and it is the
context of use and the surrounding co-text that enable us to tell which concept is being
evoked. Sometimes, though, some words and expressions, regardless of the context
and co-text, carry with them associations from their other conceptual meanings. Leech
(1981: 16) suggests that this is most striking with taboo words such as, for example,
‘erection’ and ‘ejaculation’ where their sexual associations “rub-off” (as Leech puts
it) onto their other senses. Similar refective associations might also be invoked by
‘rub-off’, especially when in close proximity to the words ‘erection’ and ‘ejaculation’
(or maybe it’s just us). The notion of associations transferring from nearby words leads
us on to the idea of collocation and Leech’s next type of meaning.

114
 ANALYSING MEANING IN DISCOURSE

Collocative meaning
Collocation is a term used to refect the relationship of co-occurrence between words.
Therefore, collocative meaning is infuenced by the company a word normally keeps.
The idea is that a word can become associated with other words it frequently co-occurs
with and acquire or take on some of their meaning. Leech uses the example of ‘pretty’
and ‘handsome’ (1981: 17) which, at the time of publication of Leech’s book, tended
to co-occur with females and males respectively. Both words might be said to mean
the same thing, and so are synonyms, but, like our ‘fat’/ ‘apartment’ pair, their use
demonstrates why no synonym is truly equivalent in actual use. The words each carry
(different) extra meaning that results from their collocational partners or habits. In the
case of ‘handsome’ and ‘pretty’, they are gendered. Handsome has a collocative
meaning of ‘maleness’, while pretty carries the collocative meaning of ‘femaleness’.
While these collocational tendencies probably seem intuitively sound, we can test our
intuitions empirically by looking at real discourse data. One way to do this would be
to read through lots of texts, looking for the words ‘handsome’ and ‘pretty’ and
assessing whether they describe male or female subjects. This method might take
some time.
An alternative means for those with less free time on their hands, and one that we
explore in Chapter 11, is to use a large collection of text data known as a corpus and
search through that using specially designed computer tools. For example, the British
National Corpus (BNC) is a 100-million-word corpus of spoken and written British
English from various genres compiled in the 1990s (we met the BNC in Chapter 4).
The corpus is freely available online via a web interface which allows for searches of
the corpus data. The web interface will also automatically calculate the collocates of
a word – i.e. show us the collocational tendencies of a word we are interested in. So,
we can use the corpus to assess the maleness of ‘handsome’ by looking at the frequent
collocates of the word in the corpus.
The word ‘handsome’ occurs in the BNC 1,577 times and among its most frequent
collocates are ‘man’ (which collocates with ‘handsome’ 132 times),
‘he’ (131 times), and ‘his’ (102). Other, less frequent, collocates include ‘her’ (45),
‘she’ (41), and ‘woman’ (23). These results begin to indicate that ‘handsome’ has a
tendency to co-occur with words denoting males rather than females which helps to
explain its collocative meaning of ‘maleness’. Searching for ‘pretty’ in the same
corpus offers contrasting results. The word ‘pretty’ when used as an adjective occurs
2,613 times in the corpus and its most frequent collocates include ‘she’ (209 times),
‘girl’ (122), ‘her’ (107), ‘woman’ (56), ‘girls’ (48). The only male collocate in the list
is ‘boy’. These collocation tendencies in the BNC data help to show why ‘pretty’ has
a collocation meaning of ‘femaleness’.
Have a look at Activity 5.2 below for more on collocation. We return to the concept
of collocation in Chapter 11 when we discuss corpus linguistics.

115
 ANALYSING MEANING IN DISCOURSE

Activity 5.2 Collocation

Consider the words below. Do you perceive them as negative? Positive? Neither negative nor
positive? Refect on your own attitudes to the concepts to which the words refer. Why might you
hold those attitudes? What do you think has informed them? Can you think of a context of use
in which these words might appear?

Commit
Juvenile
Happen
Cause

You can test the collocations of these words and phrases using the British National Corpus
‘search’ function.

(The additional online support materials on the book’s web page provide instructions.)

We now develop our discussion of meaning in discourse by taking a closer look at

how co-text and context build on or shape meaning. Following that, we discuss the
logical meanings that words carry when used in sentences that trigger presuppositions,
which are meanings that are not explicitly stated but rather implied or assumed.

Working out meaning in discourse: co-

text and context
Knowing what a word or phrase ‘means’ in discourse depends on two primary things:

(i) its co-text

( ii) its context

By ‘co-text’ we mean the linguistic company the word keeps in the text, so the other
words that occur around the word in the utterance. Look at the following sentence and
consider what concept is signifed by ‘chair’:

Ivan sat on the chair

There are other meanings of ‘chair’, including a person who calls or hosts a meeting.
It is reasonable to assume, though, that the ‘chair’ that Ivan ‘sat’ on is a piece of
furniture designed for doing just that (as in Figure 5.1), and this assumption is largely
down to co-text. Importantly, the verb ‘sit’ relates to the conventional function of
chairs as furniture and seeing that word in close proximity to ‘chair’ is likely to trigger
the ‘chair as furniture’ concept. Because of its co-text, then, you would know that the
meaning of ‘chair’ is not a person who hosts and manages a meeting (unless you attend
some very strange meetings). In short, even though we might know the different
conceptual meanings of the word ‘chair’, we can only tell which concept is being
signifed once we see ‘chair’ in its co-textual environment.
116
 ANALYSING MEANING IN DISCOURSE

We must also consider the context or situation of the utterance. If we are in a

meeting and the host has not yet arrived and someone asks ‘Where is the chair?’, we
would reasonably assume them to be referring to the person who hosts the meeting
and not a piece of furniture. This is because the situation determines that
‘chairpersons’ be present at meetings and the semantic understanding of ‘chair’ as
meaning ‘chairperson’ is triggered by the question (‘where is the chair’) in conjunction
with the lack of a host of the meeting. This is contextual-specifc knowledge which
means we need to know something about how meetings are organised, what roles
people play, and what is appropriate or necessary for a meeting to take place. We are
helped to the correct sense of ‘chair’ by the co-text as well. The defnite article ‘the’
signifes a very specifc ‘chair’, and therefore, unless there is a chair in the sense of ‘a
piece of furniture’ that everyone expects should be in the meeting but is not, the
meaning of ‘chair’ as chairperson will be easily – and correctly – reached.
The role of contextual knowledge in meaning making is the focus of ‘pragmatics’,
which we deal with in more detail in Chapter 6. Nonetheless, as the example above
shows, pragmatics and semantics are interlinked.

Activity 5.3 What’s in a word?

The words we choose to refer to concepts can have profound effects, some ideological (see
Jeffries 2010: Chapter 1). Back in 2019, the UK was involved in negotiations to leave the
European Union (EU). During the negotiations, Labour MP Hilary Benn (a member of the offcial
opposition party) introduced the European Union (Withdrawal) (No. 2) Act. The Act became law
in September 2019 and stipulated that the Prime Minister had until 19 October 2019 to either
make a deal with the EU and get it agreed in Parliament or get MPs to approve a no-deal exit
from the EU (known as ‘Brexit’).

In a heated debate in the House of Commons on 25 September 2019, the Prime Minister at the
time, Boris Johnson, repeatedly referred to the Act as ‘the Surrender Act’.

Johnson: The truth is that a majority of Opposition Members are opposed not to the so-called
no deal; this Parliament does not want Brexit to happen at all. Many of those who
voted for the surrender Act a few weeks ago said then that their intention was to stop
a no-deal Brexit.
(Hansard, Volume 664, 25/09/2019)

He also used the terms ‘capitulation Act’ and ‘humiliation Act’. The repeated use of the word
‘surrender’ caused uproar in the House and generated a lot of criticism of Johnson. Using your
knowledge of conceptual and associative meaning, suggest why the word ‘surrender’ was so
infammatory.

We will now move onto entailment and presupposition, which are areas of
meaning that for some are semantic, while for others pragmatic. The distinction, if

117
 ANALYSING MEANING IN DISCOURSE

there is one, lies partly in whether entailment and presupposition are discourse
phenomena or non-contextual. The standard position on entailment and presupposition
taken by semanticists (summarised by Saeed 1996) aims to explain the information
that we can derive from an utterance or sentence based on our language knowledge
(i.e. our linguistic competence).

Entailment
In this section, we introduce entailment and briefy show how it can be used in the
analysis of discourse. We owe much of our discussion here to Saeed (1996), which
would be useful further reading for anyone wishing to fnd out more about this topic.
Entailments are, according to some (see for example Kroeger 2018), a type of
meaning-based inference, whereby we make inferences based on meaning relations
presented in the text and not on context external to the text. Entailment can be
summarised as:

The truth of one utterance guarantees the truth of another utterance.

Consider the following statements.

1
(a) Rob is younger than his sister, Kim. (b)
Kim is older than her brother, Rob.

There is a relationship between the statements whereby the truth of statement 1(a)
guarantees the truth of statement 1(b). That means, if (a) is true then (b) must also be
true. We can say, then, that (a) entails (b). Entailment is symbolised by the logical
symbol (also known as the ‘double turnstile’): 1(a) 1(b).
Conversely, if statement (a) is false (i.e. Rob is not younger than Kim), then
entailment (b) must also false (i.e. Kim is not older than Rob). Furthermore, if (b) is
false, then it follows that statement (a) cannot be true. Saeed (1996: 95) states the
relationships involved with entailments as follows:

A sentence p entails a sentence q when the truth of the frst sentence (p) guarantees
the truth of the second sentence (q), and the falsity of the second sentence (q)
guarantees the falsity of the frst sentence (p)

Saeed’s defnition can be summarised as ‘if p is true then q is true/if q is false then p
is false’.
We can work out the entailment in 1(a) through the comparative relationship
between Rob and Kim that is signalled by the comparative adjective ‘younger’. The
entailment is therefore a logical conclusion that we can draw from the statements
derived from their linguistic composition and the meaning relations they encode. We

118
 ANALYSING MEANING IN DISCOURSE

use our knowledge of the meaning of lexical items and their relationship with other
lexical items within the discourse. We do not need to know anything about Rob and
Kim (such as who they are); the entailment depends on the comparative relationship
set up in the statement containing a comparative adjective. Other comparative
adjectives also trigger entailments (e.g. ‘smaller’, ‘heavier’, ‘richer’).
Let’s consider another example of entailment.

2
(a) Next door’s cat killed a mouse in my garden. (b) The
mouse is dead.

Statement 2(a) entails statement 2(b): 2(a) 2(b)

The entailment is derived from the relationship between ‘kill’, the action, and
‘dead’, the state resulting from the action. We know that if 2(a) is true then 2(b) must
also be true. Therefore, the logical conclusion we draw from 2(a) is 2(b). Put another
way, 2(b) is the logical consequence of 2(a). However, if 2(b) is not true (i.e.the mouse
is alive), then 2(a) cannot be true (i.e. the cat did not kill the mouse).
Entailments are also involved in hyponymic relationships. As we saw in Chapter 3,
hyponyms involve generic-specifc relations (see Leech 1981: 135–7). For example,
‘animal’ is a generic group within which there are numerous types including ‘cat’ and
‘mouse’. To see how that works with entailment, consider the following phrase
(which, at the time of writing, is a controversial hashtag on social media).

3
(a) All men are trash.
(b) Bob, who is a man, is trash.

Statement 3(a) entails 3(b) because 3(a) makes a generic, hyponymic statement that
applies to any group member. The logical consequence of the generic statement is that
since Bob is a man, he is part of the generic group and so is trash. We can present this
entailment in the following way.

All Xs are Ys
A is an example of X
Therefore, A is Y

In a similar way, where a specifc group member is specifed in a statement, then a

generic entailment can be derived. For example, in 4(a) a specifc type of animal is
identifed from which we can derive a generic entailment.

4
(a) Next door’s cat shat in my garden. (b)
An animal shat in my garden.

119
 ANALYSING MEANING IN DISCOURSE

If 4(a) is true, then it entails statement (b) because cats are a type of animal. The
reverse, however, does not hold:

5
(a) An animal shat in my garden
(b) *Next door’s cat shat in my garden

The truth of (a) does not guarantee the truth of (b). The hyponymic relationship
between animals and cats does not logically entail that if an animal shat in your garden,
then it was a cat, because many species belong to the generic group ‘animal’ any of
which might have paid your garden a visit (within reason).
Saeed (1996) also suggests that entailment can account for synonymy, as the
following examples illustrate:

6
(a1)Rob has an older sister.
(b2) Rob has an older female sibling and
(c1) Rob lied about his age.
(d2) Rob did not tell the truth about his age.

In the examples in 6, the statements (a1) and (c1) entail synonymous statements, where
‘sister’ and ‘female sibling’ are synonymous, as are ‘lied’ and ‘not tell the truth’.

Activity 5.4 Pinning the entailment on the Prime Minister

Consider this following short extract from the UK Westminster Parliament. The utterance is
made by Boris Johnson during Prime Minister’s Questions on 20/01/2021 (when Boris was
having a go at being Prime Minister of the UK).

I think it is very important that the Prime Minister of the UK has the best possible
relationship with the President of the United States – that is part of the job description, as I
think all sensible Opposition Members would acknowledge.
Hansard HC, Volume 687, column 958, 20/01/2021)

Work out the entailment in the statement and explain how the entailment works.

We provide an answer at the end of the chapter.

Presupposition
Presuppositions are to do with information that is contained in discourse but is not
stated explicitly; instead, it is assumed. There are two types of presupposition –
existential and logical – which are triggered in discourse by certain words and
structures. Presuppositions are different from entailments because whereas
entailments are logical consequences (i.e. if p is true then q must also be true)

120
 ANALYSING MEANING IN DISCOURSE

presuppositions are background assumptions or things that need to be taken for

granted for an utterance to be true.

Existential presuppositions
These are presuppositions derived from defnite noun phrases, which are noun phrases
with a determiner that is either the defnite article, ‘the’, or a demonstrative or
possessive pronoun (e.g. ‘the coast’, ‘that cat’, ‘her car’, ‘his coat’, ‘our nation’).
Recall that defniteness also indicates Given information (see Chapter 2), and there is
a connection between what is Given and what is presupposed. Existential
presupposition is fundamental to the way in which communication works because, if
we want to talk about the world, we often need to make some assumptions about what
exists in the world. Imagine trying to forecast the weather without making some
assumptions about the geography of a country – what is the north and the south and
so forth. Consider this example from the UK Met Offce website.

7
Outlook for Monday to Wednesday: Unsettled, rather mild and windy with rain at
times, heavy in the west. Colder for the north with some hill snow. Gales in the
south, and later in the north [our emphases].

The words in bold signal information that is necessarily presupposed to make the
forecast possible, and readers are likely to accept the information without objection.
Labels like ‘the north’ and ‘the south’ can be contentious, though, and can become
political. With political motivation in mind, here are two examples, both taken from
press releases about immigration policy produced by the UK Home Offce during the
Conservative and Liberal Democrat coalition government in 2012.

8
To play a full part in British life, family migrants must be able to integrate – that
means they must speak our language and pay their way . . .
UK Home Offce, 11/06/2012
https://tinyurl.com/46mh7wcr)

9
Settlement in the UK is a privilege. We are sweeping aside the idea that everyone
who comes here to work can settle
(UK Home Offce, 29/01/2012
https://tinyurl.com/ypftnbv9)

Extract 8 presupposes that ‘our language’ is a single language in the UK. Of course,
some language communities within the UK (where there are a few native languages
121
 ANALYSING MEANING IN DISCOURSE

including Welsh, Gaelic, and Irish) might disagree with that, or at least ask, who does
‘our’ refer to?
The comment in 9 presupposes that an idea exists (in bold). The whole noun phrase
and its contents are presupposed, which include the following.
• the idea (which implies a belief held by some people because
ideas do not just exist in the ether);
• people come to the UK to work; • people who come to
the UK can settle.

The noun phrase packages up these ideas and takes these things for granted (see
Jeffries 2010: Chapter 1).

Logical presupposition
Logical presuppositions are the assumptions that we can draw from the meaning of
some words and phrases. These sorts of presupposition can be triggered lexically and
syntactically.

Lexical triggers
Logical presuppositions can be triggered by some change of state verbs (e.g. ‘stop’,
‘start’, ‘return’, ‘continue’, ‘resign’, ‘fnish’). We met a change of state verb earlier in
example 2: ‘next door’s cat killed a mouse in my garden’. The verb ‘kill’ is a change
of state verb whereby something that was living becomes dead. We said that there was
an entailment – that if the statement is true that the mouse has been killed then it must
also be true that the mouse is dead. ‘Kill’ also carries with it a presupposition; it
presupposes an earlier state – if the mouse has been killed, then it must have been
alive before it was killed. Killing something presupposes that the something was alive
before it was killed and entails that the something becomes dead.
Some iterative adverbs, adjectives and verbs (e.g. ‘again’, ‘rewrite’, ‘fnal’) can
also presuppose earlier activities or states. Consider the following examples (the
convention is to use >> to mean ‘presupposes’):

10
(i) She rewrote history
>> history was already written
(ii) For the last time, shut up
>> the polite request has been made before
(ii) He lied again
>> he has lied before

122
 ANALYSING MEANING IN DISCOURSE

Consider now the following tweet below posted on Twitter on 28 December 2021 by
the Association of Flight Attendants (AFA), which contains logical presuppositions
(highlighted).
11
“There are two signifcant caveats in the guidance that recognize concerns raised by
our union. CDC recommends reducing quarantine to 5 days only if asymptomatic
and with continued mask wearing for an additional 5 days.” Read our full
statement: [statement included].
(AFA-CWA @afa_cwa (2021), our emphases)

The verb ‘reducing’ presupposes (in this context) the existence of a quarantine period
that is longer than the fve days; ‘continued’ presupposes that wearing face masks is
already happening; and ‘additional’ presupposes that there is already a specifed time
period in place.
Another group of lexical triggers for logical presuppositions are what are known as
factive verbs. These are a small group of verbs in English that presuppose the
Complement to be true.These include ‘understand’,‘knew’,‘realise’, and ‘regret’.
The above tweet in (11) includes the clause ‘recognize concerns raised’, where
‘recognize’ presupposes the existence of the concerns raised. Extract 12 below, which
is taken from the State of the Union Address given by George W. Bush on 28 January
2003, provides another example.

12
The British Government have learned that Saddam Hussein recently sought
signifcant quantities of uranium from Africa.

Bush’s use of the factive verb ‘learned’ presupposes the truth of the content of the
subordinate clause and leaves no room for Saddam Hussein’s actions to be mere
possibility.3 Lexical triggers therefore introduce relationships between what is
asserted linguistically and what underlying assumptions we need to accept as being
true for the asserted information to make sense.

Syntactic triggers
Presuppositions can be triggered syntactically by cleft and pseudo-cleft structures (we
introduced these in Chapter 2). In each of the examples below, >> outlines the
presupposition.

13
(a) It was his cat that annoyed me.
(b) What annoyed me was his cat. >> something annoyed
me.

14
123
 ANALYSING MEANING IN DISCOURSE

(a) It wasn’t me who walked mud into the carpet.

(b) What I did not do was walk mud on the carpet. >> someone
walked mud into the carpet.4 Presuppositions can be triggered by
comparative structures (in each example >> denotes the
presupposition).

15
(a) He was as ugly as next-door’s cat. >>
next-door’s cat is ugly.

16
(a) She was as corrupt as a politician. >>
politicians are corrupt.

In each case, the basis of the comparison is the aspect that is presupposed.

Testing presuppositions
Presuppositions are said to survive negation, meaning that they persist even when the
statement that creates them is negated; for example:

17
(a) It was his cat that annoyed me.
>> I was annoyed.
(b) It wasn’t his cat that annoyed me. >> I was annoyed.

Example 17 shows that we can negate the statement by adding a negative particle to
‘was’ and the presupposition that ‘I was annoyed’ (along with the existential
presupposition that the cat exists) remain intact, thanks to the cleft structure which
presupposes something ‘annoyed me’.

Presupposition and propositions

Presuppositions are not attached to propositions but the forms and syntactic structures
that realise a proposition. A proposition is the basic meaning that we can distil from
a sentence or utterance. Consider the following.

(i) The cat sat on the mat.

(ii) The mat is what the cat sat on.
(iii) It was the mat that the cat sat on. (iv) It was the cat that sat on the mat.
(v) The mat was sat on by the cat.

124
 ANALYSING MEANING IN DISCOURSE

Even though the sentences are grammatically different they involve the same two
entities, ‘cat’ and ‘mat’, and the same action, ‘sit’. The proposition of the sentence is
derived from the relationship between the two entities. So, no matter which way the
sentence is structured, the proposition remains the same. Now consider the following
two sentences.

18
(a) It was the cat that annoyed the dog.
(b) It was the dog that the cat annoyed.

In both sentences, there are the same two entities – the dog and the cat – and there is
the same relationship between them – the cat annoys the dog. Therefore, in both
sentences, the proposition is the same. However, as we established above, with cleft
structures, it is the content of the subordinate clause (the cleft clause) that is
presupposed. So, 18(a) and 18(b) above have different presuppositions, as we show in
19 below.

19
(a) It was the cat that annoyed the dog.
>> the dog was annoyed
(b) It was the dog that the cat annoyed.
>> the cat annoyed something or someone

Presuppositions in discourse
Presuppositions can be important in some discourses. For example, within the context
of courtroom cross examinations or police interviews, lawyers and police offcers can
ask questions that contain presuppositions that introduce assumptions into the record
that the witness or suspect must actively contest. Sidnel (2020) notes that
presuppositions in questions used during such questioning can prove diffcult for the
witness (or suspects) to disagree with or challenge. According to Ehrlich and Sidnell
(2006), the potential problem for witnesses on the stand in court is that the rules of
courtroom interactions may not permit them to confront the presupposition; instead,
they must simply answer the question that carries the presupposition. Presuppositions
in questions can also be seen in political discourse. This is true of the questionand-
answer exchanges routine within the UK parliament. For example, each of the
questions below asked in the House of Commons introduce a presupposition.

20
(a) When will the Prime Minister stop ducking responsibility, do the right thing and
reverse her kamikaze Budget, which is causing so much pain?
(Keir Starmer (Labour)
Engagements, Volume 720, 12/10/2022)
125
 ANALYSING MEANING IN DISCOURSE

(b) Can the Chancellor tell us at what point

in his predecessor’s so-called plan for
growth did he realise that it was a recipe for
economic disaster?
(Peter Grant (Scottish National Party)
Economic Stability, Volume 722, 15/11/2022)

(c) When will the Prime Minister start standing up for what is right . . . ?
(Anna McMorrin (Labour)
Leaving the European Union, Volume 653, 21/01/2019)

In 20(a) the lexical trigger ‘stop’ presupposes that the Prime Minister (on that day it
was Liz Truss) was, in fact, ‘ducking responsibility’. In (b), the factive verb ‘realise’
presupposes that the content of the clause that follows is a fact, that the ‘plan for
growth’ was a ‘recipe for economic disaster’. The primary thrust of the question is to
ask whether the Chancellor realised something while the ‘something’ (contained in
the subordinate clause) is taken for granted. In (c), the lexical trigger ‘start’
presupposes that the Prime Minister (Theresa May at that point) was not already
standing up for what is right. In each example, the presupposition introduces an
assumption into the debate that is potentially damaging to the addressee.

Activity 5.5 Presuppositions and the negation test

The following sentence is taken from a news story that appeared in an online police community
chat forum in October 2011. See if you can identify the presupposition(s) and what triggers
it/them. Try negating the trigger(s) – does the presupposition remain intact or is it cancelled?

An HIV positive couple are alleged to have attacked police offcers after refusing to stop having
sex in a public swimming pool.
Police Community, 24/10/2011
https://tinyurl.com/3ym2cyfd) We give our answer at the end of the chapter.

Presupposition and entailment

The line between presupposition and entailment is contested and not categorical. Some
people concede that there is no such thing as logical presuppositions but rather these
are types of entailment (Leech 1981: 286–287). Indeed, when tackling discourse
(rather than isolated examples) it is sometimes diffcult to separate out presupposition
from entailment. It is important to not become encumbered by categories of meaning
phenomena because categories are inevitably leaky. As discourse analysts, we should
use such categories and their descriptions as starting points for discourse analysis. Any
analysis of discourse should describe where meanings lie and how they can be
126
 ANALYSING MEANING IN DISCOURSE

accounted for rather than simply assigning labels to parts of a discourse. For us
entailment is what we can logically conclude from a statement (e.g. if your football
team did not win, then we can conclude that they lost or drew) while presupposition
is what must be assumed to be in place for the statement to hold true or make sense
(e.g. if your football team lost again, then we have to assume that they have lost before
for the statement to make sense).

Conclusion
This chapter has shown how meaning operates in a number of different ways. Meaning
is derived from forms (e.g. lexis) and from grammar (e.g. cleft structures). Meaning
can be polysemous and is often deducible from context, e.g. ‘chair’ as referring to
‘piece of furniture for sitting’ or ‘person who hosts a meeting’. The relationship
between words and things is arbitrary, but often follows linguistic conventions.
Meaning is explicit or denotative, but it is often connotative, which is to say, implicit
(we talk about implicatures in the next chapter). Interpreting meaning relies on our
knowledge of language but also on our knowledge of the world and how it works.
Encyclopaedic knowledge is built up over time and we draw on this to make meanings
from others’ utterances. Meaning is also social – we use what we know about social
and cultural conventions to determine what people mean when the form of their
utterances does not refect their intended meanings (such as when we ask a question
but mean it as a request). In the next chapter we explore pragmatic meaning and
consider the complex relationship between context and meaning potential. Chapter 7
develops pragmatic concepts further by considering why and how we ‘do’ certain
types of meaning.

Further reading
If you are interested in developing Activity 5.1 on collocational meaning, specifcally
its core concept, ‘semantic prosody’, see Louw (1993), Hunston (2007), and Morely
and Partington (2007). For a general introduction on types of semantics and other
types of meaning, see Saeed (1996) and Leech (1981). For more presupposition
triggers, see Levinson (1983: 181–184). If you want to learn more about
presupposition in critical stylistics, see Jeffries (2010).
Answers to activities
Activity 5.1 Social meaning and synonyms
‘wee’, ‘small’, ‘diminutive’ – ‘diminutive’ has formal or offcial associations, ‘small’
is conventional, ‘wee’ is colloquial and associated with Scotland and Ireland.
127
 ANALYSING MEANING IN DISCOURSE

‘lunch’, ‘dinner’, ‘tea’, ‘luncheon’ – whether a noon-time meal is referred to as ‘lunch’

or ‘dinner’ has regional and class associations; the same is true for whether ‘dinner’
or ‘tea’ is used to refer to an evening meal. So, it is what the word refers to that
communicates social meaning and not the word in isolation. However, ‘luncheon’ has
associations with upper social class regardless of what meal it refers to.
‘tradie’, ‘tradesman’, ‘tradesperson’, ‘navvy’ – ‘tradie’ is used in Australia to refer to
a tradesperson; ‘tradesman’ carries with it association of gender stereotypes;
‘tradesperson’ is more woke; ‘navvy’ is a labourer, but it is outdated and so has
association related to time period.

Activity 5.2 Collocation

The words in the list tend to be perceived negatively, but not always. This intuition
can be supported by corpus evidence, such as from the BNC corpus of spoken and
written English. Here are the words with some frequent collocates from the BNC (we
discuss collocates in corpora further in Chapter 11).
Commit – often collocates with negative things such as ‘crime’.
Juvenile – frequent collocates include ‘delinquency’, ‘crime’, ‘offenders’.
However, in some contexts, such as zoology and horticulture,‘juvenile’ is used
simply to refer to an age group (e.g. juvenile elephants are between fve and ten
years old) and does not hold negative connotations. Therefore, context of use can
affect the connotations a word carries.
Happen – ‘accidents’ and ‘disasters’ happen, but then so do ‘miracles’ (but not
overnight), so sometimes the evidence for collocational meaning is not clear cut.
Cause – frequent collocates include ‘problems’, ‘damage’, ‘death’, concern’,
‘trouble’.

Activity 5.3 What’s in a word?

The conceptual/denotative meaning of ‘surrender’ is quite negative since it
conventionally refers to ‘giving up’ and submitting to an opponent. The word brings
with it connotations of war, and therefore being on the losing side.
By using the word, Johnson signalled his attitude towards the Act, indicating that
it effectively handed power to the European Union. This attitude contrasts with MPs
who supported the Act who suggested that it protected the UK from leaving the
European Union without a deal.
In the debate that followed Johnson’s use of ‘surrender’, many MPs criticised its
use. Their discussion of it involved other words that appeared to be triggered by
‘surrender’, notably ‘betrayal’, and ‘traitors’.

Rachel Reeves: We have a Prime Minister who has broken the law and uses
dangerous language of betrayal and surrender, which sows division
and worse in the communities we all serve.
128
 ANALYSING MEANING IN DISCOURSE

(Hansard, Volume 664, 25/09/2019)

Tracy Brabin: We are hearing from the Prime Minister words such as the
“humiliation” Act, the “surrender” Act, and the “capitulation” Act.
All of these words suggest that we, because we disagree with him,
are traitors, that we are not patriots, but nothing could be further
from the truth.
(Hansard, Volume 664, 25/09/2019)

That ‘surrender’ suggests betrayal and traitorousness seems to indicate further

connotative meaning relating to experiential knowledge. So, as ‘mother’ connotes
‘nurture’, ‘surrender’ appears to connote ‘traitor’ and ‘betrayal’.
Another possible reason why ‘surrender’ was used by Johnson, and why it was so
effective in creating controversy is that (both intuitively and according to the BNC) a
frequent collocate is ‘unconditional’. This potentially carries over meaning
(collocational meaning) of a complete loss of power and sovereignty, which is of
course what Johnson was hoping to communicate, regardless of whether it was a true
refection of reality.
You can read the full debate on the Hansard website: https://hansard.parliament.uk

Activity 5.4 Pinning the entailment on the Prime Minister

I think it is very important that the Prime Minister of the UK has the best possible
relationship with the President of the United States – that is part of the job
description, as I think all sensible Opposition Members would acknowledge.

The underlined section includes an entailment. The entailment is similar to our ‘all
men are trash’ example. The statement entails that any member of the opposition party
who is sensible acknowledges Y. Boris Johnson is saying that:
All Xs [where X is ‘sensible Opposition member’] would acknowledge Y
Person Z is an X
Therefore, Z acknowledges Y

The logical consequence of the entailment is that if you are a member of the opposition
party but you don’t acknowledge Y, then you are not sensible.

Activity 5.5 Presuppositions and the negation test

The presuppositions are fronted by >> an HIV positive couple are alleged to have
attacked police offcers after refusing to stop having sex in a public swimming pool.
>> the couple were having sex in a swimming pool
>> the couple had been asked to stop
Negating the propositions leaves the presuppositions intact:

129
 ANALYSING MEANING IN DISCOURSE

An HIV positive couple are alleged to have attacked police offcers after NOT refusing
to stop having sex in a public swimming pool.
>> the couple were having sex in a swimming pool
>> the couple had been asked to stop

Notes
1 About £75,000, at time of writing.
2 The ‘Inkhorn’ controversy in the early 16th century started a bit of a row about borrowed words
for which native words already existed. The borrowings were considered to have more prestige
than their antecedents because most came from scientifc writings or classical literature (and so
connoted high learning and consequently, prestige). Think of it as a kind of sociolinguistics
snobbery.
3 It became apparent, however, over the years following the speech that the British intelligence –
which was part of the justifcation of invading Iraq in 2003 – was incorrect.
4 An implicature – that the speaker is accused of doing something – is likely to be generated here
through a fout of the maxim of Quantity via the negated particle ‘not’.

References
Bush, G. W. (2003) State of the Union Address. Offce of the Press Secretary.
The White House. Accessed 18 June 2021. https://georgewbush-whitehouse.
archives.gov/news/releases/2003/01/20030128-19.html 2003
Ehrlich, S. and Sidnell, J. (2006) ‘“I think that’s not an assumption you ought to
make”: Challenging presuppositions in inquiry testimony’. Language in
Society, 35(5): 655–676. doi: 10.1017/S0047404506060313
Electronic Immigration Network (2012) ‘UKBA: Automatic settlement for skilled
workers to end’. Press release. 20 February. Accessed 3 January 2023. https://
www.ein.org.uk/news/ukba-automatic-settlement-skilled-workers-end
Evans, V. (2019) Cognitive linguistics: A complete guide (2nd edition). Edinburgh:
Edinburgh University Press.
Home Offce (2012) ‘Radical immigration changes to reform family visas and prevent
abuse of human rights’. Press release, UK Government. Accessed 11 December
2022. https://www.gov.uk/government/news/radical-immigra tion-changes-to-
reform-family-visas-and-prevent-abuse-of-human-rights
Hunston, S. (2007) ‘Semantic prosody revisited’. International Journal of Corpus
Linguistics, 12(2): 249–268.
Jeffries, L. (2010) Critical stylistics: The power of language. London: Palgrave.
Kroeger, P. (2018) Analyzing meaning: An introduction to semantics and pragmatics.
Berlin: Language Science Press.

130
 ANALYSING MEANING IN DISCOURSE

Leech, G. (1981) Semantics: The study of meaning (2nd edition). Harmondsworth:

Penguin Books.
Levinson, S. C. (1983) Pragmatics. Cambridge: Cambridge University Press
Louw, B. (1993) ‘Irony in the text or insincerity in the writer? The diagnostic potential
of semantic prosodies’. In M. Baker, G. Francis, and E. Tognini Bonelli (eds),
Text and technology: In honour of John Sinclair, pp. 157–176. Amsterdam: John
Benjamins.
Morely, J. and Partington, A. (2007) ‘A few frequently asked questions about semantic
– or evaluative – prosody’. International Journal of Corpus Linguistics, 14(2):
139–158.
Ogden, C. K., and Richards, I. A. (1923) The meaning of meaning: A study of the
infuence of thought and of the science of symbolism. New York: Harcourt Brace
Jovanovich.
Saeed, J. I. (1996) Semantics. Oxford: Wiley-Blackwell.
Saussure, Ferdinand de (1983) [1916] Course in general linguistics (trans. Roy
Harris). London: Duckworth.
Sidnell, J. (2020) ‘Presupposition and entailment’. In J. Stanlaw (ed.), The
international encyclopedia of linguistic anthropology. doi: 10.1002/97811187
86093.iela0325

131
 6 Meaning and context

Introduction
This chapter deals with what is traditionally known as pragmatics which, like
semantics, is concerned with meaning. However, whereas semantics relates to the
relationship between linguistic form and meaning (i.e. dictionary or denotative
meaning), pragmatics deals with the function of utterances which involves taking into
account the context within which linguistic forms are produced (we introduced
‘context’ in Chapter 1). In this way, pragmatics is concerned not so much with
‘meaning’ as with ‘meaning potential’ (Allwood 2003). As we progress through this
chapter, we will explore context and meaning potential in discourse, and we will
consider linguistic choices made by language producers and the ways in which
receivers use them in discourse to make meanings.

What is pragmatics?
Imagine that you borrow your friend’s phone and then accidentally drop and break it.
Your generous but slightly upset friend might say, ‘I could KILL you’. The context of
situation, however, makes it clear that this utterance is not to be taken literally because
friends tend not to kill each other over minor breakages. The semantic meaning,
therefore, is not the intended one. The pragmatic meaning can be ascertained quite
quickly as an expression of annoyance at the broken phone even though the words ‘I
am very annoyed’ are nowhere to be found in the utterance. Contrast this with the
same utterance being said in a thriller movie by a kidnapper. The chances are that the
literal meaning prevails, which we can determine based on the context of the utterance:
the situation of a kidnapping, our background knowledge (see previous chapters) of
the things kidnappers usually do, the fact that we are watching a thriller in which we
expect bad things to happen, and so on.
Thus, context is relevant for the producer (speaker or writer) and the receiver (the
hearer or reader) of the language: the receiver must make meaning from the language
produced by the producer in any given situation, and the producer is

DOI: 10.4324/9781003351207-6
cognisant of the receiver’s potential interpretation.This involves what is referred to as
common ground between producer and receiver, which is shared knowledge and a

132
 MEANING AND CONTEXT

shared appreciation and understanding of the context in which the language is

produced.
Additionally, context accounts for interpersonal relationships, such as the
relationship between speakers. For instance, friends are socially close, share common
interests and, so, tend not to want to kill each other over accidental damage to their
possessions. On the other hand, social distance and different goals or purposes between
kidnapper and hostage put them at odds with each other, interpersonally speaking.
Finally, context includes the utterance in its wider linguistic environment, or co-text
(see Chapter 1). This allows meaning to be derived from what is said around it. To
offer an example of co-text, consider the following utterance taken from the BBC
World News headlines on 13 January 2021:
‘Trump impeachment: Several Republicans to join Democrats in House vote’
The word ‘house’ has a few meanings. Here are just some:
Noun
1. a family dwelling
2. a legislative assembly
Verb
3. to put something back in its place
4. to provide accommodation
Adjective
5. to assign a quality to a head noun, e.g. a type of music (‘house music’).
The grammatical structure of the sentence means the word ‘House’ is functioning as
an adjective because it pre-modifes the noun ‘vote’ as in sense 5 above. Because
‘House’ appears here in a sentence about political matters and is surrounded by
political lexis (‘impeachment’, ‘Republicans’, ‘Democrats’, ‘vote’) we interpret it as
meaning 2 above, ‘legislative assembly’. The graphological form of the word (its
initial letter is capitalised) may also lead us to this interpretation, but our common-
ground knowledge of headlines means that capitalisation is common and cannot be
relied upon as the only indicator of meaning 2.

Context summary

As we discussed in Chapter 1, contextual factors (and the contextual assumptions

associated with each of them) have a bearing on the discourse situation, and
include the following.
• Physical space: where the discourse is happening (the immediate location of
the utterance), e.g. a classroom, a bus queue, a restaurant, a street.
• Social situation: the people involved and the interpersonal relationship
between speaker and addressee, e.g. a teacher and student, two work
colleagues, a mother and child, two strangers, two friends, a kidnapper and a
hostage.

133
 MEANING AND CONTEXT

• Real-world knowledge: what the speaker and addressee know about the world
and the way the world works including shared cultural, ideological, and
habitual knowledge, e.g. people typically don’t kill each other when they break
stuff; in some cultures assertiveness is considered rude, whereas in others, it’s
empowering.
• Textual context: the way in which the discourse is happening. This includes
the co-text (i.e. the text around the utterance), which can be grammatical or
sentential, or structural, e.g. the sequencing of ‘turns’ in conversations.

Activity 6.1 depicts a language event that aims to help exemplify the concepts just
introduced.

But what do you think the recipient REALLY means? Or, what is the recipient aiming to (i)
communicate and (ii) not communicate to the gift giver? What is being pragmatically
communicated? What do you need to know of the context (of gift giving; of interpersonal
relationships, of cultural habits, etc.) in order to ‘get’ the speaker’s intended meaning? Consider
other possible ways in which the receiver could have responded. Think about these questions
before reading on.

134
 MEANING AND CONTEXT

Implicature
In the gift-giving scenario in Activity 6.1, you have probably worked out that the
intended meaning is that the recipient does not like the jumper and is asking for the
receipt so they can exchange it. The receiver could have said simply, ‘I don’t like it’.
This is truthful, to the point, and communicates exactly the amount of information
needed to convey the speaker’s intended meaning. Yet this (or something similar) has
not been explicitly uttered. Instead ‘did you keep the receipt’ is all we have to go on.
Therefore, arriving at the correct or intended meaning depends on the hearer doing a
bit of inferencing work. In this section we explore how and why inferences are reached
in discourse through the highly infuential work of Paul Grice (1975) and his ideas
about conversational cooperation and implicature.

Cooperative principle
Grice’s work appreciated that as communicators we want to communicate something
meaningful. Therefore, it is expected that we are cooperative conversationalists and
that our interlocutors (i.e. our conversational partners) are also being cooperative when
they interact with us. This is part of what makes us social beings. Grice (1975: 45)
suggested a Co-operative Principle (CP) which states:
Make your conversational contribution such as is required, at the stage at which it
occurs, by the accepted purpose or direction of the talk exchange in which you are
engaged.
Now, if we take Grice’s CP rather literally, our Christmas jumper example does not
really appear to be the most effcient way to communicate our dislike of the jumper.
Instead, it goes ‘around the houses’ a little bit, indirectly expressing the intended
meaning. But Grice is not commanding us to be completely transparent in our
utterances, nor is he saying ‘get to the point’ in every exchange.
We are not, after all, robots without feelings. The general idea with the CP is that
conversation can be approached directly and indirectly to different degrees, but
typically, conversational interactants intend their meanings to be picked up regardless
of the level of transparency they adopt. To that end, conversation tends to consist of
connected contributions that are:

characteristically, to some degree at least, cooperative efforts; and each participant

recognizes in them, to some extent, a common purpose or set of purposes, or at least
a mutually accepted direction.
(Grice 1975: 45, our emphasis)

So, even if the “common purpose” or “mutually accepted direction” of a conversation

is to get out of wearing that Christmas jumper, which, at the same time, may insult our

135
 MEANING AND CONTEXT

gift giver, we must cooperate (conversationally in the Gricean sense) to make and
understand meaning in our discourse exchange. Therefore, while we might not be
cooperating in the everyday sense of the word because we are arguing or disagreeing
or potentially insulting one another, we are still cooperating conversationally so that
our interlocutor knows what we are doing is arguing or disagreeing, insulting, and so
on. We are communicating these things in a way so that it is mutually understood.
Grice’s Co-operative Principle aims to explain how utterances can communicate
meaning that is additional to or different from the semantic meaning of the words
uttered. In Grice’s explanation, he separates out what we say from what we mean; he
reserves ‘say’ to indicate “something close to” the “conventional meaning” (1975: 45)
(i.e. semantic meaning), and ‘mean’ for what is implied, suggested, or hinted at by the
utterance. Grice suggests the verb implicate for the action of generating meaning
beyond what is said and the noun implicature for these implicated meanings. Speakers
implicate or generate implicatures by not adhering to basic expectations that underpin
the Co-operative Principle, as we will now explain.

Maxims
Within the CP Grice tentatively introduces some key principles of conversational
cooperation that he calls maxims of which there are four types. They are Quantity,
Quality, Relation, and Manner (Grice uses upper-case initial letters for these maxim
types). The general idea is that we observe these maxims for optimum transparency
and effciency. So, as cooperative participants in an interaction, there is an expectation
that we will not say too much or too little (maxim of Quantity), that we will be honest
(maxim of Quality), stick to the point
(maxim of Relation), and be coherent and unambiguous (maxim of Manner).
We set the maxims out below, quoting directly from Grice (1975: 45–46), and follow
them with examples.
The maxim of Quantity refers to “the quantity of information to be provided” so
that speakers should:
1. “Make your contribution as informative as is required (for the current purposes
of the exchange).”
2. “Do not make your contribution more informative than is required.”
The maxim of Quality refers to the truthfulness of the contributions so that
speakers should “try to make your contribution one that is true” and incorporates
two sub-maxims:
i “Do not say what you believe to be false.” ii “Do not say
that for which you lack adequate evidence.” The maxim of
Relation states simply:
1. “be relevant”.
136
 MEANING AND CONTEXT

The maxim of Manner relates to not what is said but how it is said so that speakers
should “be perspicuous” (i.e. be clear and easy to understand!) and:
i “Avoid obscurity of expression.” ii
“Avoid ambiguity.”
iii “Be brief (avoid unnecessary prolixity).” iv
“Be orderly.”
As with the Co-operative Principle, even though the maxims are phrased as
imperatives they are not directives or rules. Instead, they are conceived as a set of
conventional expectations we have when we engage in conversation (e.g. we expect
our interlocutors to avoid being vague or ambiguous). The idea is that if a
conversational turn does not fulfl these expectations (i.e. does not observe the maxims)
we might be prompted or “compelled” (Potts 2015: 185) to search for an alternative
implicated meaning because we assume the person is still co-operating
conversationally but for some reason is not observing the maxims at the level of what
is said.

Flouting maxims
This leads us to examine why interlocutors would not observe maxims. Grice’s
maxims are often exploited, perhaps to be polite or because it would not be appropriate
in some situations to be direct (we discuss politeness in Chapter 7). When a speaker
‘blatantly’ fails to adhere to a maxim and is being cooperative in doing so, then this is
known as fouting a maxim. When fouting, the speaker wants the hearer to notice that
a maxim has not been observed and intends for the hearer to derive an alternative
meaning (an implicature) from this fout.
This alternative meaning is not explicitly encoded in the language uttered but is
implicated by the fout. In other words, the implicated meaning cannot be traced to the
form of the utterance. Also, because there is an expectation that the speaker is
observing the cooperative principle, the hearer assumes that the speaker’s fout is
meaningful and not simply a ploy to derail the conversation just to be mischievous.
Given all these assumptions, the hearer therefore needs to work out how what was said
by the speaker is a meaningful turn within the ongoing interaction. To do this, the
hearer must infer what the speaker is implicating by the utterance.
Before going on to look at how Grice’s CP can help with the analysis of real
discourse, we will work through each of the maxims and demonstrate how fouting
might work using made-up examples. Each example centres on the hypothetical
scenario where an adult hearer is asked whether they have seen the animated Disney
flm, Frozen.

Flouting the maxims of Quantity maxim (1) –

be as informative as required:

137
 MEANING AND CONTEXT

A: Have you seen Frozen?

B: I might have. [meaning that they have but they are embarrassed to admit it]
maxim (2) – do not be more informative that required:
A: Have you seen Frozen?
B: Yes, I saw it on March 8th, March 19th, March 24th, March 29th, April 2nd,
April, and I think I also saw it last week. [Meaning that they have seen it a lot.1]

Flouting the maxims of Quality.

Recall that this category has one super maxim (‘Try to make your contribution one
that is true’), which is qualifed by two submaxims. Our examples deal with the
submaxims:
Submaxim (i) – do not say what you believe to be false:
A: Have you seen Frozen?
B: Only about a gazillion times. [Meaning that they have seen it a lot] Submaxim
(ii) – do not say what you lack evidence for:
A: I can’t believe you haven’t seen Frozen.
B: There are literally thousands of people who haven’t seen Frozen. [Meaning –
believe it! It’s not a big deal!]
Our made-up examples aim to show how the two submaxims of Quality differ. The
fout of submaxim (i) is defnitely not true because (a) gazillion is not an actual number
and (b) if gazillion were a number it would be potentially so huge that it would not be
possible to see the flm that number of times (based on average life expectancies and
typical length of flms, the most you could expect to watch a flm of 90 minutes,
assuming non-stop watching from birth to death, is around 430,000 times, give or take
a thousand2).
The second example is different because there is a good chance that there are many
people around the world that have not seen the flm; it sounds reasonable and might be
true. Person B, who utters the statement, might not believe it to be false and they might
suspect it is true. However, without some sort of survey, the statement lacks evidence.
It is fairly common to say things that lack evidence. We refer you to the previous
sentence as an example of that sort of thing. The key with Gricean implicature is that
the statements that lack evidence need to be made so as to exploit the Quality
submaxim, and therefore communicate an implicature. With our made-up example, B
implies that A needs to recognise that there are plenty of people in the world with
better things to do than watch feature-length cartoons.

Flouting the maxim of Relation

The category of relation has just one maxim: be relevant.
A: Do you want to go and see Frozen?
138
 MEANING AND CONTEXT

B: I’m 55 years old. [Meaning: No, I don’t want to see Frozen – why would I? I’m
a grown-up!]

Flouting the category of manner

Grice sets out one supermaxim (be perspicuous!) and four further submaxims for this
category (and says that there might be more). The submaxims describe various ways
in which one might observe the supermaxim.
A: [talking to a person of interest in the presence of young sibling/related child]
Please indicate the extent to which you would fnd it agreeable to accompany
me to the kinematograph this postmeridian to partake in ocular and aural
stimulation via F-R-O-Z-E-N. [Meaning: I don’t want the young child to
understand my utterance]
B: I am laundering my cranial follicular protein flaments. [Meaning: I don’t want
to go with you.]
In the example above, A is asking B on a date to the cinema to see the flm Frozen. In
our rather contrived scenario, A is blatantly not observing the category of Manner and
fouting the supermaxim by not being clear and easy to understand. The idea is that
while A wants B to understand that an invitation is being issued, A does not want the
small child, who is also present and within hearing distance, to understand the
utterance and to think that the invitation includes them. A therefore communicates the
invitation to B but in such a way that an implicature is also generated: that A wants B
to know that the child is not meant to understand and is not invited. A therefore
exploits the maxim of Manner by fouting it. Our view is that submaxim (i), avoid
obscurity of expression, is most pertinent here because:
• issuing an invitation in the form of an imperative when a more straightforward and
clear way would be to use an interrogative (e.g. ‘Shall we go to the cinema?’);
• the imperative is in fact a request for information relating to B’s emotional response
to an invitation, which is not the clearest way in which to offer an invitation;
• embedding the content of the request within a noun phrase (which starts with ‘the
extent’) again reduces the clarity of the utterance;
• using Latin phrases (postmeridian), rarely used variants of words (kinematograph),
and technical terminology (ocular and aural) are more obscure than the more
common alternatives;
• and spelling out Frozen is less clear than simply saying the word.
It is important to note that a single utterance can fout more than one maxim. In the
above example, the invitation is longwinded, so speaker A also fouts submaxim (iii)
of Manner by not avoiding prolixity. Note that this submaxim (be brief – avoid
unnecessary prolixity) seems closely connected to the category of Quantity. However,
while the Quantity category focuses on what information is communicated, the
Manner category focuses on how it is communicated. So, submaxim (iii) of Manner

139
 MEANING AND CONTEXT

reminds us that while the right amount of information is provided (not too much; not
too little) more words than necessary might be used to convey that information.
To summarise, to fout a maxim is cooperative because it requires the hearer to know
that the speaker’s non-fulflment of maxims is being done for a reason that they want
the hearer to ‘pick up’ (see below for further examples). The box below consolidates
some of the ideas above.

Doing implicature in a post-football match press conference

Football press conferences can be very entertaining discourse situations,

especially for those interested in pragmatic implicature. On 19 February
2020 Jose Mourinho, then manager of the QR 6.1 Jose Mourinho post-match
English football club Tottenham Hotspur interview
(Spurs), was interviewed at a press
conference following his team’s 0–1
defeat
to Leipzig at the former’s home ground. You can watch the interview by
following QR code 6.1 (and we re-visit this press conference in Chapter 7).

The extract opens with a forced choice question “Was it Spurs that were bad
or [. . .] Leipzig [. . .] good?” where neither option casts Spurs in positive light.
Mourinho answers the question with a yes-no question, “You think we were
bad?” and throws the frst choice in the initial question (‘bad’) back to the
reporter. The reporter’s response “Well, the result says you lost at home” fouts
the maxim of Relation (by not answering the question with either ‘yes’ or ‘no’)
and Quantity (by saying more than is required: either ‘yes’ or ‘no’). These fouts
give rise to the implicature that losing at home means that the team is ‘bad’. The
discourse context (a post-match conference) means that those in the interaction
will share some common ground knowledge: that away wins are more impressive
than home wins; that this is refected in the points awarded in deciding the winner
in tie-breaking situations3 and conversely, that, psychological factors such as an
advantage gained from increased home supporters and familiar surroundings
mean that winning should be easier for the home team.
With this shared background or common ground information, it’s pretty clear
that the implicature will be picked up. The assumption has to be that
conversational cooperation is the communicative goal here, unless there are
overriding reasons to suggest otherwise – in which case it wouldn’t be a fout and
there would be no implicature. Mourinho interprets the Relation fout (“Well, the
result says you lost at home”) as a hedged insult: that Spurs were bad and that
they should have won because they would have had the easier time of it given
that they played at their home ground. When Mourinho leaves the press
conference, as he walks away, he shouts “Great question, mate”. This is an ironic
sideswipe at the reporter and is a fout of the maxim of Quality as Mourinho
does not mean what he says. The reporter retorts “Great answer”, which may
well generate a similar implicature because it’s not clear whether he is being
ironic. If he is, then he, too, fouts the Quality maxim and in doing so distances
140
 MEANING AND CONTEXT

himself from the harm done by Mourinho’s side-swipe. Or maybe he simply

misses the implicature.

Violating maxims and opting out

Sometimes speakers do not observe the maxims, for example, for interpersonal reasons
such as politeness (see Chapter 7) or institutional constraints on what information can
be shared, and so on. Grice (1975: 49) sets out some ways in which the maxims may
not be fulflled. These include (a) violating a maxim and (b) opting out of the CP. Both
jeopardise the CP but in slightly different ways.

Violate
A conversational participant may “quietly and unostentatiously” not fulfl a maxim.
Grice notes that “in some cases” this is “liable to mislead” the other participant. So, a
lie would be a violation of the Quality Maxim. In violating a maxim, a speaker is being
intentionally uncooperative, in which case such unostentatious violations are not
relevant to what is communicated. However, as Jeffries and McIntyre note, violations
of the Quality maxim are ‘notoriously diffcult to spot in real life data, for the simple
reason that an effective violation of this maxim is one that is not noticeable’ (2010:
107).

An example of a violation

On 28 May 2022, the Champions League fnal, a high-profle football match, was
held at the Stade de France in Paris. The match was between Liverpool and Real
Madrid. The policing of the match was a shambles and the Liverpool fans were held
back by police where they patiently waited in very dangerous circumstances for over
two hours at the entrance to the ground. According to French news reports, French
Interior Minister Gerald Darmanin falsely claimed that “massive, industrial-scale
and organised fraud in fake tickets” was to blame for the scenes at and around Stade
de France, claiming further that 30,000 to 40,000 Liverpool fans arrived at the
stadium either “without tickets or with counterfeited tickets”. This was a lie intended
to mislead the public and defect attention away from the shambolic actions of UEFA
and the French police. The violation of the Quality maxim was later ‘found out’,
which led to a hedged retort from a parliamentary member. Responding to
Darmanin’s violation to a general audience, MP Manuel Bompard fouted the
Manner maxim: “When you make a mistake – and mistakes happen – the best thing
is to acknowledge your error, not to invent fake fgures to try to hide it” (France24
2022). Bompard here fouts the Manner maxim by being ambiguous in using what
appears to be generic ‘you/your’. By fouting the Manner maxim, and referring to

141
 MEANING AND CONTEXT

nobody specifcally, the hearer can easily pick up the implicature that the pronouns
refer specifcally to Darmanin. (See also Activity 5.4 in
Chapter 5.)

Explicitly opt out

Opting out is when a conversation participant makes it clear in some way or another
that they will not or cannot be cooperative. Grice’s example is “I cannot say more; my
lips are sealed.” Common phrases such as ‘no comment’ also fall into this category.
In opting out, a speaker is jeopardising the CP, but doing so ‘on record’, which is to
say they explicitly state that they will opt out. Opting out can be a conversational
strategy to avoid committing to a proposal or suggestion (think of political fgures who
refuse to be drawn on topics). On the other hand, opting out can be a noble or genuine
attempt to respect contextual constraints, such as when business decisions are
forbidden from being discussed, or there are legal implications to saying more.

Infringement of maxims
Thomas (1995) suggests that another way in which maxims are not fulflled that is not
mentioned by Grice4 is by infringement. Thomas uses infringement for when a
speaker breaks a maxim because her/his performance is impaired in some way. The
following is Thomas’s defnition:

A speaker who, with no intention of generating an implicature and with no intention

of deceiving, fails to observe a maxim is said to ‘infringe’ the maxim. In other
words, the non-observance stems from imperfect linguistic performance rather than
from any desire on the part of the speakers to generate a conversational implicature.
This type of non-observance could occur because the speaker has an imperfect
command of the language (a young child or a foreign learner), because the speaker’s
performance is impaired in some way (nervousness, drunkenness, excitement),
because of some cognitive impairment, or simply because the speaker is
constitutionally incapable of speaking clearly, to the point, etc.
(Thomas 1995: 74)

Neurodivergent people can often infringe maxims in interactions. In one study (Losh
and Capps 2006), children with Autism Spectrum Disorder were asked to respond to
the question: “Tell me about a time when you felt embarrassed”. In one instance, a
boy (ten years old) gave the rather under-specifc response: “Usually . . . when I say
things.” (Losh and Capps 2006: 814). Here, the child infringed the Quantity maxim
by saying too little and not making their contribution as informative as required. In
another study, Losh and Gordon (2014) found that autistic participants tended to
142
 MEANING AND CONTEXT

produced off-topic and irrelevant remarks, departed from main story themes, and
produced less coherent stories. These studies demonstrate that while participants’
contributions were intended to be cooperative, their responses appeared to be
inadvertently uncooperative at times.
One literary example of maxim infringement can be found in the interactions of
Mark Haddon’s character-narrator in The Curious Incident of the Dog in the Night-
Time. Christopher, the narrator, presents as an Autistic person which leads him to (at
times) unintentionally infringe conversational maxims. In Haddon’s novel,
Christopher’s social capabilities are markedly different from neurotypical people, so,
on the one hand he infringes maxims, while on the other hand he rigidly adheres to
them in ways that make him seem robotic or aloof. The following exchange between
Christopher and his neighbour, Mrs Alexander, who is trying to befriend him,
demonstrates this when she asks why he disappeared a few days earlier while she was
inside the house getting some biscuits for them to share out in the garden (we have
added speaker turns for ease of reference):

Mrs A.: What happened to you the other day?

CB: Which day?
Mrs A.: I came out and you’d gone. I had to eat all the biscuits myself
CB: I went away
Mrs A.: I gathered that
(Haddon 2004: 69)

When Christopher tells Mrs Alexander ‘I went away’ he is observing the Quality
maxim, but conversational conventions mean that Mrs Alexander wants to know more
than the obvious ‘truth’ of him leaving (as she has established that much already).
Therefore, Christopher’s response offers less information than is required to be
cooperative. Yet, as Christopher is not intending to communicate anything other than
the literal truth his response is not a Quantity fout – he has no desire to raise an
implicated meaning in his response (such as ‘I don’t want to talk right now’). Thus,
he has infringed rather than intentionally fouted the Quantity maxim. (for more on
Haddon’s use of infringement to characterise Christopher, see Semino 2014).

Doing implicatures in real discourse: memes

In this section, we will focus on digital communication. As noted above, an implicature
is generated when maxims are fouted. The hearer is encouraged to arrive at a meaning
through inference. This occurs in all types of interactions and so Grice’s ideas can be
applied to more contemporaneous discourse, such as electronic messaging. In social
messaging apps, like Instagram, WhatsApp, and Facebook, discourses are often
multimodal in the sense that words combine with images in meaning making. The
143
 MEANING AND CONTEXT

success of communication depends on shared or ‘common ground’ knowledge or

values. Memes and ‘stickers’ used in messaging apps are examples of multimodal
discourse and combine topical, culturally specifc images with words in novel ways to
implicate meanings. Because memes tend to employ indirectness, they require greater
cognitive processing similar to other indirect utterances (Shibata et al. 2011; van
Ackeren et al. 2012), and this extra effort may help develop or reinforce interpersonal
relationships between interlocutors. Memes are cooperative ‘utterances’ in the same
way that conversational utterances are cooperative because both rely on shared
common ground knowledge. Indeed, memes exploit common ground-ness and their
use can even strengthen the interpersonal relationship between sender and addressee.
This is partly because memes typically employ humour, which refects well on the
sender’s character because humour ‘tickles’ the addressee (Simpson 2001).
Additionally, the addressee invests greater processing effort to interpret the meme’s
indirect message and is rewarded for that, usually by ‘getting’ the humour. As such,
memes can reinforce in-group membership. In other words, if the implicature
generated through the meme is successfully decoded, the sender and receiver are
jointly successful in terms of meaning making on the one hand, and sharing (and
getting) the humour of it on the other. The interpersonal relationship can become
stronger as a result (O’Boyle 2021).

Context, culture, and implicature

Memes often use cultural fgures to reinforce in-group membership or “groupness”
(Brubaker and Cooper 2000). For example, O’Boyle’s (2021) study of a
“neighbourhood drinking group” called “the Man Shed” discusses how a small
community of residents communicated using Donald Trump memes during the early
part of the Covid pandemic in 2020. The study acknowledges the social function of
such discourse. O’Boyle concludes that the phatic function of communication “play[s]
an important though inconspicuous role in relationship maintenance” (2021: 460).
While Donald Trump is a globally recognised cultural fgure in the real world,
fctional characters can be equally thematised in social messaging exchanges. One such
fgure, a cartoon sponge, with which many Gen Z young adults (people born between
1997 and 2012; see Dimock 2019) may be familiar.5 The meme below in Figure 6.2
uses a character we have called Absorbent Andy to represent the meme. It shows him
wearily getting out of his chair with the text ‘Ight Imma head out’ added. The lexis is
African American vernacular: ‘ight’ is a clipped form of ‘Right’ or ‘Alright’ and
‘imma’ is a contraction of ‘I am going to’. Let’s assume the meaning in the original
context is literal (Absorbent Andy gets up and goes out). In the meme version, the
intended meaning is non-literal (or literal AND non-literal).
When used non-literally, users fout the maxim of Quality because the literal
interpretation (of physically exiting) is not the intended meaning. Rather, the ‘leaving’
is metaphorical – ‘I’m exiting this situation’ (in a polite, playful, awkward, or
embarrassing way). Second, the co-text, that is, the text produced by the speaker that
144
 MEANING AND CONTEXT

accompanies the meme, is needed to correctly decode the intended meaning. For
example, one social media user employs the meme in reference to

Figure 6.2 An artist’s representation of the ‘Imma head out’ meme featuring Absorbent Andy

an absent dad (Figure 6.3), where child abandonment is euphemistically encoded as a

temporary ‘heading out’.
This particular version of the meme (6.3) fouts the maxim of Quantity, because the
‘out’ is insuffcient information. In fact, it is an example of a scalar implicature, which
means that implicated meaning is informative to a degree on a conceptual scale
(‘scalar’) ranging from weakly informative to strongly

Figure 6.3 The absent dad variant

informative. The use of ‘out’ introduces a range of different kinds of ‘out’ that have
greater ‘communicative strength’ (Potts 2015: 18) (e.g. out to the shop, out for the rest
145
 MEANING AND CONTEXT

of the person’s life). The father in Figure 6.3 is both literally out (he has indeed left
the room) but also at the same time fguratively out for the remainder of the child’s
life. The British comedian Micky Flanagan immortalises this scalar meaning when he
talks about going ‘out’ (for a beer without the intention of staying out), rather than
‘out-out’ (a more intentional, planned night ‘out’). In the meme example above, the
‘out’ is semantically true (and so adheres to the maxim of Quality), but is not as true
as it could be because it does not specify where on the scale of ‘out’ the dad sits (which
appears to be ‘out for the whole person’s life’).
Another adaptation of the ‘Imma head out’ meme has relevance for a subset of Gen
Z interlocutors: fans of the American singer Halsey (Figure 6.4). The Halsey version
of the meme says ‘Ight, Imma head to the Badlands’. Badlands is a Halsey album that,
according to some listeners (personal communication), invokes melancholy, sadness,
or simply refection. Used this way, the meme means ‘I’m exiting this situation to be
sad/refect’. Decoding it relies on the receiver making the intertextual connection (to
Badlands), interpreting it as a fout of the Quality maxim (for there is no such physical
place as ‘Badlands’), and then picking up the implicature that the sender is in a
particularly pensive mood. Once the meme has gained cultural prominence, its
appropriation becomes meaningful, too. The ‘out’ becomes a kind of ‘time out’ for
quiet refection or introspection. Of course, the addition of the wig and outft in Figure
6.4 helps steer the reader to the meaning of ‘Badlands’, although we are not quite sure
what Halsey would make of her doppelganger’s aesthetic!
The meme can be used in lots of situational contexts, such as the university
classroom (Figure 6.5). This is one that we, as teachers, are rather familiar

Figure 6.4 The Halsey variant

146
 MEANING AND CONTEXT

Figure 6.5 The student variant

with. In this version, the accompanying text presents the teacher’s utterance
‘Attendance isn’t required’, but the fnal word is cut off (the rhetorical term for this is
aposiopesis) by the ‘half the class’ getting up to leave. The humour is partly derived
from the novel way that the meme has been appropriated, and partly from the
contextual familiarity that the target social group (students) has with the collective
common ground information invoked by the teacher’s announcement (that half the
students leave). The student meme generates an implicature that if a class is not
compulsory a lot of students will not attend; it is by picking up the implicature that the
humour is successfully carried. Here, the ‘out’ is literal.
As can be seen from these examples, memes and their various permutations require
cooperation from sender and addressee. The implicatures memes generate depend on
a shared understanding of context (e.g. university) and they incorporate sociocultural
themes or events (e.g. music genres, cartoons). The payoffs from working out their
intended meanings can build or reinforce social group membership and strengthen
interpersonal relationships.

Conclusion
In this chapter, we have discussed the cooperative principle and the maxims to which
speakers orientate their contributions. We showed how speakers often fout these
maxims to generate implicatures so that what they say is not often what they mean.
Shared knowledge and expectations of cooperation lead hearers to deduce implicatures
when it appears that maxims have been fouted. Flouts often occur for particular
reasons, namely social politeness (see Chapter 7), irony, or for humorous purposes.
Grice’s CP relates to real-life conversations, but the principle governs all kinds of
discourse, such as print ads, political poster campaigns, government warnings,

147
 MEANING AND CONTEXT

interpersonal app messaging, memes, and so on. Indeed, Grice alludes to the ubiquity
of the CP across a range of discourse situations in one of his examples which involves
a reference given by a teacher, person (A), about a former philosophy student, Mr X.
The reference goes like this: “Dear Sir. Mr. X’s command of English is excellent, and
his attendance at tutorials has been regular. Yours.etc.” (Grice 1975: 52). Grice states
that:
[‘A’] cannot be opting out, since if he wished to be uncooperative, why write at all?
He cannot be unable, through ignorance, to say more since the man is his pupil;
moreover, he knows that more information than this is wanted. He must, therefore,
be wishing to impart information that he is reluctant to write down. This supposition
is tenable only if he thinks Mr. X is no good at philosophy. This, then, is what he is
implicating.
(Grice 1975: 52)

Maybe the next time you ask for a reference, you might want to read not only what is
there, but what is not there!
The language used in any discourse situation, whether written, spoken, or
multimodal, conveys a particular semantic or conventional meaning that is tied to the
forms in the utterance. Yet as we have seen, pragmatic meaning is conveyed through
and around the language used. It is conveyed by the speaker but with the hearer in
mind, as implicatures must be capable of being picked up by the hearer to function
successfully. Pragmatic meaning, then, is a kind of joint effort (hence ‘cooperation’)
because when the speaker implicates meaning the implicature is not tied to the forms
of the sentence. As Grice (1981: 185) notes: “It is important that what is
conversationally implicated is not to be thought as part of the meaning of the
expressions that are used to get over [carry or communicate] the implication.”
This means that if a speaker wishes to generate a specifc implicature, whether it’s
to communicate their dislike of a gifted Christmas jumper or emphasise their
indignation on an issue, there are several strategies at their disposal for doing so.
Speakers can either fout a maxim or they can opt out. The situation and wider context
(which includes interpersonal relationships, mode of communication and so on) will
determine which route they take. It is also important to note that while different
utterances might implicate the same meaning they might also generate different
pragmatic, social, political, and cultural effects. Context, in all cases, is crucial.

Further reading
The sociolinguistic objectives in interpersonal interactions assume an element of
cooperation between interlocutors as Grice sets out in his seminal work.
However, Grice’s approach is not the endpoint of such discussions. Some scholars
have developed a robust and nuanced approach to implicature and pragmatics more
148
 MEANING AND CONTEXT

generally by focusing on one of Grice’s maxims, the maxim of Relation. This forms
the basis for the development of Relevance Theory, which has as its central tenet that
a speaker’s contribution to conversation carries ‘a presumption of optimal relevance’
(Wilson 2016). Robyn Carston notes the usefulness of Relevance Theory over the
Gricean model when she says that “Relevance-based pragmatics stands a better chance
of providing a psychologically sound explanation of utterance interpretation, since it
is grounded in a theory of how we process and represent information in general”
(1995: 213). Sperber and Wilson (1986), Clark (2013), and Scott et al. (2019) offer
accessible and insightful studies using Relevance Theory.
For more on scalar implicatures, see Zondervan (2010) and Potts (2015).
Additionally, Noveck (2001) explores the relationship between logical and pragmatic
reasoning when children and adults encounter scalar implicatures.

Notes
1 Give us a break. It’s been a long few years in lockdowns and isolation.
2 See note 1.
3 Until the 2021–2022 European Football season, if a two-legged tie was drawn between team A
and team B, and team A scored the most away goals, then team A would be deemed the winner.
So, if team A drew 1–1 at home and 2–2 away, they would go through to the next round on away
goals.
4 Grice does use the term ‘infringement’, albeit in a different way, to subsume fouting, clashes, and
explicit opting out.
5 We cannot include the original character here due to copyright, so Eve Canning produced a
generic sponge-shaped character that we have called ‘Absorbent Andy’.

References
Allwood, J. (2003) ‘Meaning Potential and Context. Some consequences for the
analysis of variation in meaning’. In H. Cuyckens, R. Dirven, and J. Taylor (eds),
Cognitive approaches to lexical semantics, pp. 29–65. Berlin: Mouton de Gruyter.
BBC World News (2021) ‘Trump impeachment: Several Republicans to join
Democrats in House vote’. 13 January. Accessed 2 June 2022. https://www.
bbc.com/news/world-us-canada-55642101
Brubaker, R., and Cooper, F. (2000) ‘Beyond “identity”’, Theory and Society, 29(1):
1–47.
Carston, R. (1995) Quantity maxims and generalised implicature. Lingua, 96(4): 213–
244.
Clark, B. (2013) Relevance theory. Cambridge: Cambridge University Press.
Dimock, M. (2019) ‘Defning generations: Where Millennials end and Generation Z
begins’. Pew Research Center, 17 January. Accessed 4 July 2023. https://
149
 MEANING AND CONTEXT

www.pewresearch.org/short-reads/2019/01/17/where-millennials-end-and-
generation-z-begins.
France24 (2022) ‘Champions League fnal “could have been better organised”, French
interior minister tells Senate’. 1 June. Accessed 13 February 2023.
https://www.france24.com/en/france/20220601-macron-backs-darmaninwants-
full-transparency-on-champions-league-chaos
Grice, H. P. (1975) ‘Logic and conversation’. In P. Cole and J. Morgan (eds), Syntax
and semantics vol. 3, pp. 41–58. New York: Academic Press.
Grice, H. P. (1981) ‘Presupposition and conversational implicature’. In P. Cole (ed.),
Radical pragmatics, pp. 183–98. New York: Academic Press.
Haddon, M. (2004) The Curious Incident of the Dog in the Night-Time.
Harmondsworth: Penguin.
Halsey (2015) Badlands [album]. Astralwerks.
Jeffries, L. and McIntyre, D. (2010) Stylistics. Cambridge: Cambridge University
Press.
Leech, G. N. (2014) The pragmatics of politeness. Oxford Studies in Sociolinguistics.
Oxford: Oxford University Press.
Losh, M. and Capps, L. (2006) ‘Understanding of emotional experience in autism:
Insights from the personal accounts of high-functioning children with autism’.
Developmental Psychology, 42(5): 809.
Losh, M. and Gordon, P. C. (2014) ‘Quantifying narrative ability in autism spectrum
disorder: A computational linguistic analysis of narrative coherence’, Journal of
Autism and Developmental Disorders, 44(12): 3016–3025.
Noveck, I. A. (2001) ‘When children are more logical than adults: Experimental
investigations of scalar implicature’. Cognition, 78(2), 165–188.
O’Boyle (2021) ‘WhatsAppening Donald: The social uses of Trump memes’.
European Journal of Cultural Studies 25(2): 458–462.
Potts, C. (2015) ‘Presupposition and implicature’. In S. Lappin and C. Fox (eds), The
handbook of contemporary semantic theory (2nd edition), pp. 168–202. Hoboken,
NJ: John Wiley and Sons Inc.
Scott, K., Clark, B., and Carston, R. (eds) (2019) Relevance, pragmatics, and
interpretation: Essays in honour of Deirdre Wilson. Cambridge: Cambridge
University Press.
Semino, E. (2014) ‘Language, mind and autism in Mark Haddon’s The Curious
Incident of the Dog in the Night-Time’. In M. Fludernik and D. Jacob (eds),
Linguistics and Literary Studies/Linguistik und Literaturwissenschaft, pp. 279–
304. Berlin: Mouton de Gruyter.
Shibata, M., Abe, J. I., Itoh, H., Shimada, K., and Umeda, S. (2011) ‘Neural processing
associated with comprehension of an indirect reply during a scenario reading
task’. Neuropsychologia, 49: 3542–3550. doi: 10.1016/j.
neuropsychologia.2011.09.006
Simpson, P. (2001) ‘“Reason” and “tickle” as pragmatic constructs in the discourse of
advertising’. Journal of Pragmatics, 33(4): 589–607.

150
 MEANING AND CONTEXT

Sperber, D. and Wilson, D. (1986) Relevance: Communication and cognition. Oxford:

Blackwell Press.
Thomas, J. (1995) Meaning in interaction: An introduction to pragmatics. London:
Longman.
van Ackeren, M. J., Casasanto, D., Bekkering, H., Hagoort, P., and Rueschemeyer, S.
A. (2012) ‘Pragmatics in action: Indirect requests engage theory of mind areas and
the cortical motor network’. Journal of Cognitive Neuroscience,
24: 2237–2247. doi: 10.1162/jocn_a_00274
Wilson, D. (2016) ‘Relevance theory’. In Y. Huang (ed.), Oxford handbook of
pragmatics. Oxford: Oxford University Press.
Zondervan, A. (2010) Scalar implications or focus: An experimental approach.
Amsterdam: Netherlands Graduate School of Linguistics.

151
 7 Politeness

Introduction
In the previous chapter we considered the importance of context and the ways in which
the relationship between producer and receiver infuence how the ‘message’ is
communicated. We showed how meanings are often not traceable to the semantic content
alone. We established that, as creative users of language, we adapt linguistic structures
so that our utterances can carry meanings that are either direct or indirect. We discussed
Grice’s Co-operative Principle, which accounts for utterances that are direct (and
generally adhere to conversational maxims) and those that are indirect (typically by
fouting maxims). It is by fouting maxims that we create implicatures which communicate
something different than what we actually say. We also looked at the ways in which
discourse makes use of shared sociocultural knowledge to generate and interpret
implicatures and showed, with reference to memes, that shared knowledge is often
specifc to a particular social group or discourse community.
In this chapter, we present a pragmatic framework that explains one of the reasons
why we use and invest so much cognitive effort in creating and interpreting implicatures
in everyday interactions. The framework we introduce offers a range of pragmatic
strategies that attend to linguistic meaning in its social, cultural, personal, and
professional contexts. It is the framework of politeness, frst introduced by Brown and
Levinson in the 1970s and developed in the 1980s and beyond.
Politeness in a non-linguistic sense is already familiar to you – it is those little markers
of good manners, like saying ‘thank you’, that help endear you to one another, not least
by showing that you appreciate what you do for each other. The pragmatic framework of
politeness focuses on a wider range of phenomena than good manners and attends to their
‘payoffs’ in social interactions. According to Leech: “politeness is concerned with
avoiding discord and fostering concord, only insofar as these are manifested through
communication, especially through what meanings are expressed or implicated” (Leech
2014: 88).
Leech sees politeness as having a social motivation and a psychological one. The
social motivation is to maintain “concord” and avoid “discord”, while

DOI: 10.4324/9781003351207-7

the psychological goal is to avoid damage to self-image (yours and your interlocutor’s).
Let’s start with an example. Imagine you are working from home and your roommate is
playing their music a bit too loud for you to concentrate on your work. You want to say
something that effectively communicates your wish for them to turn down their music,
but telling them to turn it down imposes upon them because you are making demands on
their freedom to do as they wish. At the same time, you risk offending them by suggesting
152
 POLITENESS

that you do not share their desire to listen to loud music. So, there is a tension between
achieving your goals but managing the imposition to maintain social concord with your
roommate. There are a number of ways you can achieve your goal linguistically, using
language. Arguably, the more direct you are, the clearer your message will be and the
less cognitive effort your roommate will need to process your utterance, so directness is
more economical. Theoretically, then, if you want someone to turn down their music you
can formulate a short, direct utterance that communicates your intended message in the
most economical way:

(1) turn down your music1

Being concise and direct in (1) means your message is economical and clear. After all,
you are adhering to Grice’s conversational maxims (see Chapter 6) – you are being
concise and to the point so observe the Quantity maxim; you are being truthful and so
uphold the Quality maxim; your utterance is relevant to the situation, so you adhere to
the maxim of Relation; and you are unambiguous and have avoided perspicuity, thus
adhering to the Manner maxim. In short, you are observing the cooperative principle in
the most direct way. So far, your utterance appears to satisfy the ‘effective’ part of your
wish to communicate. However, in uttering (1) you are perhaps less likely to achieve the
outcome you desire. This is because you risk potential discord, not just by the making
the imposition, but by the way you do it, in this case using an imperative structure.
Devoid of context, this utterance is essentially an order or command, which assumes a
relationship where you have more power than your roommate, (and this might be
something your roommate objects to or resents). So, while you have communicated your
wishes, you have done so in a way that has causes discord rather than maintains concord.
From this simple example, we can see that language is not simply a medium for imparting
information from one person to another in order to get things done. Language is also
instrumental in building and maintaining relationships. It is – and requires – negotiation.
This leads us to adjust our linguistic choices according to the kinds of interpersonal
relationships we have or wish to create with our interlocutors and the contexts in which
we use language. We make language choices that are mindful of the other person as well
as ourselves. This idea that our language choices are infuenced by social factors was
discussed at length by Erving
Goffman, who introduced the infuential notion of face (Goffman 1967).
In the following sections we discuss face and the contiguous concept, face-work, before
dealing with how face is addressed and redressed through the phenomenon of politeness.

Face
Face is a sociological concept that has great signifcance for discourse analysts as it
attends to the way language users acknowledge and respect cultural conventions and
handle empathy, power relationships, and self-presentation in communication. First
153
 POLITENESS

introduced by Goffman, whose work was based on earlier traditional notions of

politeness in East Asian cultures, the concept of face is: “the positive social value a
person effectively claims for himself [sic] by the line others assume he [sic] has taken
during a particular contact” (Goffman 1967: 5).
Following Goffman (1967), Brown and Levinson developed ‘politeness theory’ and
used face metonymically (see Chapter 8) to stand for the whole person, their reputation
and character. Much of our social communication is done cognisant of each other’s face.
Politeness is essentially face-work; when we ‘do’ politeness, we are doing face-work.
To unpack this a little, face is the “public self-image” (Brown and Levinson 1987: 61)
that a person wishes to project and protect in everyday interactions. This means we do
not want to ‘lose face’, nor do we wish to threaten our addressee’s face by what we say
to them. We like to be thought of as decent humans who make good choices in life and
we want our discourse to refect this. Therefore, if the ‘line’ (to use Goffman’s word –
think ‘stance’ or ‘approach’) someone takes in an interaction threatens our positive self-
image, that tends to bother us. And, equally, we generally take pains to ensure that the
‘line’ we take does not threaten the face of others so as not to bother them. In this way,
we demonstrate empathy, show that we are aware of socially appropriate linguistic
behaviour, and that what we say and how we say it respects and preserves our
interpersonal relationships.

Face threats and face-work

In Brown and Levinson’s politeness theory, there are two types of face: positive face and
negative face. When we talk about positive face, we are referring to a person’s positive
public self-image, the right for them to have their opinions valued, their ideas
appreciated, their choices respected, and so on. For example, saying ‘I really like your
music’ considers the positive face of your interlocuter because it praises their choices.
On the other hand, when we talk about negative face, we are talking about the right to
be free from impositions. Asking your roommate to lower their music imposes on their
right to feel unimpeded, and hence, disregards their negative face-wants. Brown and
Levinson defne positive and negative face as follows.

positive face: “the want of every member that his [sic] wants be desirable to
at least some others’”
negative face: “the want of every ‘competent adult member’ that his [sic]
actions be unimpeded by others”
(Brown and Levinson 1987: 62)

Any language event that causes a risk to face is called a face-threatening act or FTA
for short. Impositions are considered face-threatening acts to negative face because they
place the hearer (H) under an obligation to do something, even if H does not feel obliged
(and even if they do not do it). In the example above, the roommate may feel obliged to
lower the volume of the music which impedes them from listening to their music as they
want, and it makes them do something they would not have otherwise done. On the other
154
 POLITENESS

hand, an FTA to positive face is any threat to a person’s public self-image – saying ‘I
really don’t like your music’ is a threat to positive face because it undermines the hearer
and their life-choices, desires, and so on.
The ‘positive’ and ‘negative’ terminology is unfortunate because the words already
have commonplace meanings that suggest they are diametrically opposed. However, in
the politeness framework we have to suspend our common ground understanding of the
terms because in the theory positive and negative are not oppositional, where one is
‘good’ and the other is ‘bad’, but rather each addresses different wants that a person may
be perceived to favour. Those wants (face-wants) are that our positive self-image is
preserved and respected (positive face) and that we don’t feel imposed upon in respect
of our time, energy, or resources (negative face). Of course, there are many times in our
lives when we simply have to impose on others. In doing so, we potentially threaten their
negative face. Or, we have to say something or do something that will likely cause them
to feel aggrieved or disappointed or hurt and so on, which potentially threatens their
positive face. Doing politeness, then, is about balancing the risks to face with the
demands our FTAs invariably carry, and then fnding an appropriate strategy or strategies
to minimise that risk. In politeness terminology, to minimise the threat is to mitigate it
or redress it with a politeness strategy or strategies.
There are many strategies for doing face-work and previous scholars (for example,
Brown and Levinson) have separated strategies into the categories ‘positive politeness
strategies’ and ‘negative politeness strategies’. However, we fnd this distinction might
be a little misleading because it could imply that certain politeness strategies are directed
exclusively to one face or the other (i.e. ‘positive politeness strategies’ can only be
employed to anoint positive face). In reality, politeness strategies can be employed to
redress FTAs to positive or negative face. So, rather than differentiate between them as
‘positive politeness strategies’ and ‘negative politeness strategies’ we refer to them
collectively as ‘politeness strategies’ – but it’s important to know to which face they are
oriented! Therefore, we will use terms like ‘politeness strategy oriented to positive face’
or ‘politeness strategy oriented to negative face’. There is, of course, plenty of overlap in
that some strategies can mitigate threats to either face, and more than one strategy can be
employed in an utterance. We present many of them in the table below (and you can
probably add your own strategies to these). It is important to point out at this stage that
different cultures will do politeness differently and we return to this point later. In this
section we will explore how politeness strategies work in discourse.

Table 7.1 Some politeness strategies

Politeness Strategy Examples 1

Acknowledge the imposition ‘I know this is a lot to ask but . . .’

(includes ‘indicate reluctance’)
‘I wouldn’t usually ask you this but . . .’
Hedges ‘Could you, er, close the window maybe?’
155
 POLITENESS

Be pessimistic ‘You wouldn’t just close the window?’

Be optimistic ‘You’ll close the window, right?’
Minimise the imposition ‘Does anyone have a tiny little bit of milk I could use for
my coffee?’
Apologise ‘I am so sorry to have to ask you, but . . .’
Ask forgiveness ‘Forgive me for being blunt but that outft needs its own
therapist’
Joke ‘You must’ve been born in a hospital with swinging
doors!’
(Sending humorous memes also falls into this category)

Use conventional indirectness ‘Could you pass me the salt?’

Use deference (i) Honorifcs ‘Dr Canning, I’d like to make an appointment’
Use deference (ii) humble self ‘I won’t be able to work this out by myself – could you
help?’
Impersonalise ‘It would be great if the table could be set for dinner’
(Includes ‘state the FTA as a ‘Students must return books’
general rule’) ‘Books must be returned’
‘Smoking is not allowed here’‘No smoking’
Provide overwhelming ‘Well, this morning, my alarm clock failed to go off
reasons three times, then I had a headache when I woke up, then
my dog ate my homework, then my granny got sick, so
could I have your homework to copy for class?

Politeness Strategy Examples 1

Seek agreement/avoid S1. ‘You been at the library?’

disagreement S2. ‘Well, I gotta go there tomorrow, but today I
was at the lab’
Use in-group markers ‘Hey mate, we are next in line’
(includes code-switching and ‘Spill the tea, sis’
dialectal forms) ‘OMG are you *sleggin?’ (*this means ‘joking’ or
‘winding me up’ in Belfast English)
Claim common ground ‘You like Grimes?2 Me too!’
Exaggerate ‘You’re the best friend EVER!’
Understate ‘I may have written a little tome’ (imagine it was said
by Stephen Hawking about his bestselling book)

Irony S1: ‘Dude, I broke your phone’

S2: ‘Well, aren’t you the bright spark?’
Use tautologies S1: ‘This coronavirus is a real bummer and I’m so sick
(note the overlap here with of staying in’
avoid disagreement) S2: ‘It is what it is’
Use rhetorical questions S1: ‘How many times do I have to invite you here
before you actually come?’
S2: ‘I know, sorry, I should make more effort’

156
 POLITENESS

Compliment ‘You have a lovely home/hairdo/family/work ethic . . .’

Activity 7.1 Politeness in real discourse

Examples of real discourse have been provided below with context. Example (A) is written email
communication from an academic in a British university to all staff inviting participants to respond to a
survey for their research project. The second (B) is a spoken exchange taken from the Santa Barbara
Corpus of Spoken American English (see Chapter 4) between a bank employee and the president of the
bank in Illinois, USA. The participants are Jim, who is the president, and Joe, who is the employee. Joe is
taking minutes for the meeting. Example (C) is from the UK version of The X-Factor (Cowell 2013), a
television talent show in which singers audition in front of four celebrity judges for a place in the
competition. In the exchange, the judges are Louis Walsh (music producer), Gary Barlow (from the
successful British group Take That), Nicole Scherzinger (from The Pussycat Dolls), and Simon Cowell (music
producer). The auditionees, Liddia and Ryan, are, in spite of their enthusiasm and energy, not terribly good
at singing.

Can you spot any potential FTAs? And to which face? Are any strategies used to mitigate the FTAs? What
are they? We offer some possible answers at the end of the chapter.

(A) The academic email

Happy St. Patricks Day! Sorry for cross posting (or if you have already completed) but I wanted
to send out a reminder for the [redacted] Questionnaire.
Responses have been amazing so far and I am having a fnal push for participant collection. I
would ask anyone who plays sports (either competitively or for fun) and is over the age of 18 to
take 5-minutes to complete my PhD questionnaire. Please fnd the link below:

[redacted]

This study has been approved by the University Ethical Approval system at [university name
redacted]. Many thanks

(B) The bank meeting

JIM: I got a couple things here, [. . .], just to kind of bring us up to speed, Matt has tentatively
accepted our offer for employment, but since he’s going to be a dual employee, th- you really
don’t need to put this in the minutes. JOE: Okay.

(C) The X-Factor audition

The context for this example is that X-Factor judges Gary Barlow and Louis Walsh rate an audition
by Liddia and Ryan poorly. Walsh calls them both ‘deluded’, and Gary Barlow tells
QR 7.1 Liddia and Ryan’s X-Factor audition them that they haven’t got a voice
between them. To watch the audition
and commentary, scan the QR code. The
relevant segment begins at 26:23
minutes.

157
 POLITENESS

Using politeness strategies to mitigate face-threatening

acts

So how do we ‘do’ politeness? If we return to our music-loving roommate, our desire to

have the music volume lowered means that we have to do an FTA. Activity 7.2 asks you
to have a go at identifying some of the strategies that might serve to reduce the threat to
face in the utterances (a)–(f).

Activity 7.2 Mitigation strategies

If we wanted our roommate to turn down their music, we could choose a number of ways of getting
our intended result. Let’s imagine a sort of Groundhog Day scenario where we do the FTA six times
in six different ways. In each scenario we utter one of the

options set out below ((a) to (f)), all of which are oriented to face to varying degrees. Refer back
to the politeness strategies in Table 7.1 and see if any are employed in (a) to (f). Which
utterances might work best? Which might not work well at all? Can you think of a reason why
(f) is a strategy? We give some possible answers throughout the remainder of the chapter.

(a) Turn your music down

(b) I can see how much you’re enjoying your music so I hate to have to ask if you wouldn’t
mind lowering the volume a little?
(c) It’s nice music, but a bit loud, so please could you lower it? (d) I can’t really concentrate on
my work
(e) Wow, you kinda like your music loud, huh?
(f) Ed fucking Sheeran3 at 400 decibels – who hurt you?

Even though all the utterances desire the same outcome, only (a), (b), and (c) make
the intention explicit (but with varying degrees of directness). We need to dig a bit deeper
to get to the intended meaning in (d), (e), and (f) as they fout maxims and so are risk
averse in their approach to the FTA (yes, even (f) could be considered risk averse to
negative face – not so much to positive face!). Yet because they generate implicatures
they are also risky in terms of whether or not the implicatures get ‘picked up’ by the
hearer, meaning that the speaker’s intended meaning must be inferred by the hearer
because it is not explicitly conveyed. For instance, (e) and (f) could be so indirect that
the implicit request is lost on the hearer.
According to Brown and Levinson, when approaching an FTA, there are a number of
options open to us. We can do the FTA:

1 on-record (directly, or ‘baldly’) 2

of-record (indirectly)
158
 POLITENESS

Or we can choose not to do the FTA and

3 opt out (say nothing in order to avoid the FTA)4

Once a speaker decides what approach to take (i.e. do the FTA on-record, off-record, or
opt out), they can then select from a number of redressive strategies that are oriented to
the hearer’s positive or negative face (or both). Figure 7.1 shows these approaches and
roughly plots their level of explicitness (on the left of the diagram) and how they are
executed strategically (on the right).
Whether speakers deliver the FTA on-record or off-record they can choose to ‘soften’
the FTA with redressive action by choosing a strategy or strategies that work to
minimise threats to face. We will consider each of our scenarios in (a) to (f) in turn, but
frst, we present them in their corresponding positions in the model in Figure 7.2 below.

Figure 7.1 Approaches to FTAs

159
 POLITENESS

Figure 7.2 Doing the FTA ‘turn your music down’

On-record FTAs
We start with the on-record utterances frst. Any politeness strategy mentioned will appear
in italics. Utterance (a) is maximally direct because it gets to the point. By being
maximally direct, however, it constitutes an FTA to the positive and negative faces of
the hearer. Consequently, the FTA is done ‘baldly’. It is also on-record because the words
used directly correspond to the intended meaning: grammatically it is an imperative, and
desired action is explicitly encoded (we will see later in the chapter that orders and
requests can be made using a range of grammatical forms).
Utterance (b), which is less direct than (a), mitigates the imposition by using a few
politeness strategies (italicised here and throughout the chapter). First, the speaker
acknowledges the value of the music to the roommate (‘I can see how much . . .’). In this
way the utterance “anoints” (Brown and Levinson 1987) the positive face of the
roommate by ‘buttering them up’ before hitting them with the imposition (Figure 7.2).
So, the speaker attends to both positive and negative face. To use Brown and Levinson’s
terms, the speaker is able to “pay back in face whatever [they] potentially [take] away by
the FTA” (1987: 71). Utterance (b) is doing other work – it acknowledges the imposition
the speaker is making to the hearer (‘I hate to have to ask . . .’), which foregrounds the
risk to their own positive face in asking. This is likely to predispose the hearer to respond
positively as a proportionate ‘return’ for the risk the speaker takes in asking.
Additionally, it presents the request in a pessimistic way, thus not demanding or even
expecting compliance (‘if you wouldn’t mind . . .’) which, even though this is only a
“token bow” to the negative face-wants of the hearer it still gives them an out (Brown
and Levinson 1987: 70–71). Finally, it lessens the imposition with ‘lowering the volume
a little’ foregrounding the fact that what the speaker is asking is small and so refusing
would seem unreasonable.
Similarly, utterance (c) encodes a small imposition through ‘a bit’ and butters up the
hearer by complimenting their music choice. Also, the choice of verb ‘could’, as opposed
to ‘would’, is indirect because even though it’s intended as a request to turn the music

160
 POLITENESS

down, grammatically it asks the hearer about their ability to do so. In this way we can
see that both positive and negative face are attended to and therefore both politeness
strategies are used together.
In these ways (from (a) to (c)), the speaker orients their utterances to the hearer’s face-
wants through a range of different politeness strategies while still remaining on-record
about their communicative intentions (to get the music turned down). In summary then,
speakers can show that they are attending to the negative face of hearers by using
politeness strategies such as acknowledge or lessen the imposition, and be pessimistic. In
addition, speakers can anoint the positive face of hearers by acknowledging the value (of
the loudness of the music) and by complimenting the hearer, which helps to soften the
blow of the imposition.
Off-record strategies
Utterances (d), (e), and (f) are off-record because the form of the utterance does not
semantically or syntactically correspond to a particular request/ demand to lower the
volume of the music. The intended meanings need to be picked up by H through
implicature. Going ‘off-record’ is a politeness strategy in itself. But even off-record
utterances can carry some risk to face. Here are (d) to (f) again:

(d) I can’t really concentrate on my work

(e) Wow, you kinda like your music loud, huh?
(f) Ed fucking Sheeran at 400 decibels – who hurt you?

Utterance (d) does not explicitly state but rather implicates the intended meaning by
offering a reason from which H must infer relevance in the context. As such, it fouts the
maxim of Relation. The reason given shifts the focus to the speaker’s incapacities rather
than the hearer’s actions. This is a form of deference as the speaker is humbling
themselves. Utterance (d) uses additional strategies, namely hedging with ‘really’. Think
of ‘hedging’ rather literally as a way of softening the hard edges of your driveway – it is
the same for language. Hedges are elements of language that are used pragmatically, even
though they may well have semantic meaning. That is, their presence is doing
interpersonal work in helping take the edge off the parts that we feel may cause a bit of
a threat/damage to the addressee. Some hedges include the adverbials ‘just’, ‘only’, and
‘really’. In utterance (d), ‘really’ weakens the proposition making it less face-threatening.
The hedge offers only partial commitment to the verb ‘concentrate’ so the roommate may
feel less responsible for the distraction their music caused than if no hedge was used.
Utterance (e) is a comment about the other person’s likes, so once again does not
address the goal of the speaker, making it an off-record utterance. It uses the hedge
‘kinda’ which softens the off-record FTA and offers only partial commitment to the verb
‘like’. It also employs a tag question (‘huh?’) which nudges the force of a grammatical
assertion towards a less face-threatening interrogative structure, yet with no onus on H
to reply.
Finally, you’ll notice that utterance (f) is not in Figure 7.2. We will address it
separately here so as not to complicate the diagram above. Utterance (f) is clearly off-
record because it does not contain any request or order to lower the music. As such, it
161
 POLITENESS

fouts a number of maxims. It fouts the Relation maxim in asking ‘who hurt you?’ which
does not appear to be linked to the preceding clause. Additionally, it does not explicitly
link to the speaker’s intention to get the roommate to lower the volume. The utterance
also fouts the Manner maxim (avoid prolixity) by not being brief (‘at 400 decibels’ versus
‘loudly’) and the Quality maxim by asserting ‘400 decibels’ (clearly an exaggeration).
Yet, the indirect request – essentially a threat to negative face5 – is delivered in such a
way that it also carries with it a threat to positive face (that the choice of Ed Sheeran is a
poor one6). It appears to go against the general idea of politeness which is to maintain
social harmony and avoid social discord. In fact, it’s possible to read it as confrontational.
So why include it here in a discussion of politeness? The answer is in the interpersonal
relationship between the two interlocutors.
Given that they are roommates, this confrontational utterance is an example of “banter”
(Leech 1983), a kind of ‘mock impoliteness’ which is done as a marker of solidarity and
intimacy (let’s face it, you’d not get away with that utterance to a stranger on the train).
We say a bit more about impoliteness in the section on further reading at the end of this
chapter.

The humble hedge

Brown and Levinson devote quite a bit of space to hedges in their discussion of the
politeness framework. We met some hedges in the examples above, and other hedging
strategies include the use of minimising or diminutive adjectives such as ‘little’ (or ‘wee’
if you come from Ireland or Scotland), ‘tiny’, and so on. Recall the jumper example we
discussed in Chapter 6. Let’s return temporarily to Christmas Jumper-gate in utterance
(1) and mitigate it in (1a) with adverbial hedges:

(1) ‘I don’t like this jumper’

(1 a) ‘I just don’t really think I like this jumper’

Here, hedging with ‘just’ puts some distance between the speaker (‘I’) and the ‘not
liking’, both syntactically by positioning the grammatical Subject further away from the
verb and fguratively by downplaying the imposition. Notice how this offers a little bit of
protection to the speaker’s own positive face as well as the addressee’s positive face
because they are minimising the not liking by using an adverb of degree (‘just’). In some
cases, adverbs like ‘really’ and ‘just’ tend to offer more interpersonal social value than
semantic value, as in our example. By downgrading the ‘don’t like’ to ‘don’t think . . .
like’ the speaker reduces the commitment to the not liking and consequently the force of
the FTA. The latter is an example of using a lexical verb (‘think’) modally to hedge a
speaker’s commitment to a face-threatening proposition, which in turn, reduces the face
threat of the proposition for both parties.
Impositions come in many forms, and even something low-stakes like asking to
borrow a book from a friend or asking a roommate to lower their music can constitute an
imposition. We impose on people all the time. Just take today, for instance; how many
times have you used communication in ways that amount to an imposition? (We did it in
162
 POLITENESS

the last sentence – we asked you to think about your communication and we mitigated
that imposition with ‘just’). Did you try to get a fatmate to give you ‘a wee taste’ of milk
for your coffee? Did you message a friend or family member to ask for a lift ‘just up to
the shop’? Have you emailed a teacher to ask for ‘a little more’ clarity on a scheduling
issue or ‘only’ one more day for that deadline? Or did you ask to borrow your friend’s
phone ‘for a quick sec’? In our classes, our students frequently email us saying ‘I have a
small question’, regardless of the size of their question. Whatever you have tried to
procure, chances are that you adapted your way of asking depending on who you were
asking and what you were asking for, and we would be willing to bet that you used a wee
hedge at some point.

Non-linguistic considerations
In the politeness framework language operates in conjunction with sociological
variables. This means that when we approach an FTA, we need to consider a few
sociological factors in weighing up what is, for the specifc encounter, the most effective
and socially appropriate option to take. Brown and Levinson note three factors that
should be considered when approaching an FTA. They are as follows.

the social distance between the interlocutors (abbreviated to D for

‘distance’. Genius, eh?)
the relative power of the interlocutors (P) the
ranking (or scale) of the imposition (R)

These can be explained using the six scenarios featuring our noisy roommate. In that
example, the relationship between the speaker and hearer will infuence the choices the
speaker makes when approaching the FTA. We consider the issue of power (P) frst.

Power (P)
If interlocutors are good friends, some of the on-record utterances may be less threatening
to face, or the FTAs are mitigated by social factors. This is because the friendship puts
them on equal footing and means that neither one has power (P) over the other. To use
the terms outlined in Brown and Levinson’s model (1987: 77), being on an equal footing
means that the speaker cannot impose their own plans at the expense of the hearer’s.
Compare this with a situation where there is a disproportionate power relationship
between the two interlocutors, such as when someone is giving you a lift in their car and
they are blasting the car radio. In that situation the driver has more power (they own the
car, they drive the car, they are doing you a favour) and the asymmetrical power balance
means it’s more face-threatening to tell the driver to lower the music than to tell your
roommate (all other things being equal).

163
 POLITENESS

Social distance (D)

Linked to power (P), distance (D) concerns the nature and degree of the social
relationship that exists between speaker and hearer. Brown and Levinson defne social
distance as being related to ‘frequency’ of interactions. This is premised on the argument
that the greater the frequency of interactions you have with someone, the less distant you
are from them interpersonally. It follows then, that the less social distance there is, the
more freedom speakers will have in tackling an FTA. For example, good friends, family
members, and couples are generally able to approach an FTA with less risk to face than
those with greater social distance because the familiarity places the relationship on a
more secure (and equal) footing. For instance, your mother might be able to tell you that
your outft would look better in the bin, but a sales assistant in a shop is not likely to be
so on-record because the interpersonal relationship is more distant (and arguably calls
for more politeness). Of course, you could argue that your mother’s FTA to your positive
face is more damaging because of the close relationship (and that the sales assistant’s
FTA is less damaging because the distance between you means that it won’t have
repercussions for your future relationship). You could also argue that the sales assistant
simply wants the sale, in which case this might be an instance of what we might call
covert motivation! Perhaps in such situations mothers rank higher on the power variable
than shop assistants. Maybe this wasn’t such a great example, but it does show the
complex interpersonal work our language choices have. In our roommate example, there
is likely to be less social distance (D) (because roommates frequently interact) and equal
power (P). The important point that Brown and Levinson make is that social distance is
only one key factor that must be considered in conjunction with the other two; so, if
power was equal and the imposition was small, the social distance would be the only
variable in play and speakers modify their approaches to FTAs accordingly. The example
they offer is as follows. i) Excuse me, would you by any chance have the time? ii) Got
the time, mate?

According to Brown and Levinson (1987: 80) utterance (i) contains more strategies and
would be used when the interlocutors are distant, and (ii) is more likely to feature where
they are socially ‘close’.
Social distance is dynamic and can change over time (even over a very short space of
time). For example, you may start the semester addressing your new teacher with ‘Dear
professor’, but a few weeks in as social distance shrinks you might change to show less
deference, e.g. ‘Dear Brian’, ‘Hi Brian’, but probably not ‘S’up B-Dawg’. On the other
hand, even though social distance is less in this situation, the power factor might mean
that you retain honorifcs (e.g. terms of deference such as ‘Dr/Prof.’) for the whole
semester.
It is important to note, however, that cultural differences give different weight to the
three variables. For example, Gu’s work (1992) on Chinese politeness demonstrates that
social distance is refected in address forms; only speakers considered socially close to
their hearers will use frst names as terms of address, whereas surnames are considered
the default “non-kin public address term” (Gu 1992: 250). It is not just in naming that

164
 POLITENESS

politeness strategies vary in different cultural contexts. Yangmeishan and Wencheng

(2019) show that in Chinese culture, where familial relationships are highly valued,
family members are not typically thanked for small favours. Indeed, they show that
making explicit expressions of gratitude to close family members is akin to ‘socio-
pragmatic failure’. Explicit expressions of gratitude are reserved for non-kin or socially
distant interlocutors.

Ranking of imposition (R)

Finally, the nature of an FTA itself can be gauged on a rank scale of imposition, which
answers the question how imposing is the imposition? This is culturally dependent. What
is meant here is that the FTA is gauged as weighty by the degree to which it “interferes
with an agent’s wants of self-determination or of approval” in that culture (Brown and
Levinson 1987: 77). Translating this to a request to a roommate to lower their music, the
risk to face is proportionately low in terms of its impediment on the hearer’s freedom to
act as they wish. (A high impediment would be if you were asking them to never play
their music again). Therefore, all other things being equal, going on-record is an
appropriate strategy because it carries less risk than if one or other factors were not equal.
Indeed, for sincerity to prevail, redressive action for any FTA would need to be
proportionate to the weightiness of the FTA; too much redress, as in (g), runs the risk of
sounding sarcastic, too little (h) and it sounds rude.

(g) I’m so sorry, I can’t believe I have to ask you this, especially since we’ve been friends
for years now, and it’s only because I cannot really concentrate on my work which
has to be in for tomorrow morning or I’ll fail my course and nobody wants that to
happen and certainly not you, as you are so supportive of me, but could you possibly
turn down your wee music honey?
(h) Lower that.

A short analysis of a football press conference

To close this chapter on politeness, we re-visit the extract from the previous chapter on
the exchange between Jose Mourinho and a journalist at a press conference following
Mourinho’s team’s defeat. In Chapter 6 we considered the conversational implicatures
in the exchange; in this section we consider politeness and show how the variables of
power (P), social distance (D), and ranking (R) intersect in a real discourse situation. Our
analysis explores the contextdependent nature of these variables and the importance of
social roles. We will examine how different values in D, P, and R can ameliorate the risk
to the speaker’s positive face when the speaker issues on-record FTAs.
Recall that the exchange between Jose Mourinho, the then manager of

165
 POLITENESS

Tottenham Hotspur (Spurs), and the reporter occurred during a press conference which
took place on 19 February 2020, following his team’s
home defeat to Leipzig. The QR code links to the
interview segment. QR 7.2 Jose Mourinho interview
Football press conferences are arenas where
powerful fgures (such as players and managers) are
interviewed by the press (who also have power) in front of a wide television audience.
This presents a contradiction as the manager is, to some degree, at the mercy of the
journalist asking the questions. There is also a cultural expectation in post-match press
conferences that the journalist will ask questions and that the manager will answer them.
Power (P) is at play, then, particularly when the journalist asks questions that offend the
manager or that imply negative capability (and so are highly ranked (R) FTAs). What
happens in our example could be construed as a tug-of-war over power (P) precipitated
by an FTA to Mourinho’s negative face that is ranked highly (R), compounded by the
social distance (D) between the two, which is likely to also be high. The reporter poses a
question (“Was it Spurs that were bad or Leipzig that were good?”) that presents an
either/or scenario, leaving Mourinho little option for a way out. Therefore, the reporter
imposes on Mourinho, threatening his negative face (his desire to be not imposed upon)
in the frst instance by trying to get Mourinho to commit to one of two (restricted) options,
that the opposing team was “good” or that Mourinho’s team was “bad”. Neither option
is complimentary towards Mourinho’s team because neither evaluate their performance
positively. The reporter’s question therefore also threatens Mourinho’s positive face
because it suggests that his team was either ‘bad” or not as good as the opposition, which
attacks the way his team played and thus his competency as a manager. Instead of
answering the question, Mourinho repeats it and sends it back to the reporter seeking
clarifcation: “You think we were bad?” However, Mourinho’s question focuses on the
frst of the two options originally offered by the reporter (which is more direct than the
reporter’s utterance) and obligates the reporter to provide a yes/no response. This could
be perceived as an FTA to the reporter’s negative face due to the obligation to commit to
a response. Yet Mourinho still offers the reporter a way out of committing to the
proposition that Spurs were “bad” as the latter can reply with ‘no’.
In his reply to Mourinho, the reporter avoids the preferred response (polar questions
require ‘yes’ or ‘no’) choosing instead to respond with “Well, the result says you lost at
home”. This is indirect because it answers the question via an implicature (that we
explained in the previous chapter) and commits an FTA to Mourinho’s positive face off-
record. As we discussed in Chapter 6, the response fouts the maxim of Relation and
Quantity giving rise to the implicature that losing at home means that the team is “bad”.
Additionally, the response is mitigated with a hedge, “well” which is a type of discourse
marker that signals justifcation is being offered for the implicature by prefacing what the
hearer, Mourinho, may not want to hear (Owen 1981; Jucker 1993). In Lakoff’s terms
(1973: 458), ‘well’ implies “a complex-answer” to a question and, according to Jucker,
is used “in cases in which respondents know that they are not giving directly the
information the questioner has requested” (Jucker 1993: 440) as is the case here. The
journalist attempts to preserve his own positive face by avoiding taking personal
166
 POLITENESS

responsibility for the ‘saying’ himself. Instead, he impersonalises the contentious claim
by deferring to an inanimate agent, “the result”. As impersonalisation is a politeness
strategy, the reporter mitigates the FTA. This allows him to put distance between him
(the reporter) and the perceived insult to Mourinho.
Mourinho uses his power as a prolifc manager to opt out when he says he is “not going
to answer”. It also allows him to assert baldly and on-record that he “does not like” the
question and thinks it is “out of order”, thus issuing an FTA to the reporter’s positive
face. Notice though, that he tells the reporter that it is his “question” that is out of order
and not the reporter himself who is out of order. It could be that Mourinho is still
attempting to maintain social concord with an interlocutor who has, by all accounts,
relatively high (P). As Mourinho gets up to leave the press conference he appears to
anoint the reporter’s positive face by issuing a compliment (“Great question”) and using
an in-group marker (“mate”), but this may have been done ironically (the non-verbal
communication will help in discerning this) and is actually an FTA to the reporter’s
positive face (it fouts the maxim of Quality as Mourinho means the opposite of what he
has said). The exchange culminates in an excellent side-swipe by the reporter who regains
some of his damaged face by replying with the echoic retort, “Great answer” which also
anoints Mourinho’s positive face if it is a genuine compliment, but it’s more likely that
it is meant ironically (via a Quality fout) and so threatens rather than anoints Mourinho’s
positive face.

Conclusion
Politeness is one of the primary reasons that we choose to communicate meanings
indirectly. When doing politeness, we employ a range of politeness strategies that
mitigate potentially face-threatening acts.These strategies can be used to mitigate threats
to negative face, for example when imposing on a hearer’s time, energy, or resources.
They are also used to “anoint” the positive face of our hearer, for example, when we
acknowledge their right to be liked, respected, and valued. In attending to the “face-
wants” of others, we are subtly protecting our own positive face. The context of the
interaction will often determine whether we go on- or off-record in our utterances.
Additionally, the context will determine what, if any strategies are used, and how many
are appropriate. Power, social distance, and the weightiness of the FTA (P, D, R) all
dynamically interact in helping us to make a plethora of communicative decisions in
every single interaction.

Further reading
In the discussion above about the music-loving roommate, we talked about how utterance
(f) exhibits impoliteness in that it appeared to invite discord as it resembled an insult and

167
 POLITENESS

as such an FTA to positive face. To read more on impoliteness, banter, and mock
impoliteness, see Leech (1983, 2014), Culpeper (1996), and Bousfeld (2008). For more
on how cultural differences infuence politeness, see Gu (1992).

Answers to activities
Activity 7.1 Politeness in real discourse
(A) The academic’s email
Happy St. Patricks Day! Sorry for cross posting (or if you have already completed) but I
wanted to send out a reminder for the Sporting Superstitions Questionnaire.
Responses have been amazing so far and I am having a fnal push for participant
collection. I would ask anyone who plays sports (either competitively or for fun) and is
over the age of 18 to take 5-minutes to complete my PhD questionnaire.
Please fnd the link below:

[redacted]

This study has been approved by the University Ethical Approval system at [University
name redacted].
Many thanks

The email’s objective is to invite staff to take part in a survey. This effectively imposes
on the recipients’ freedom to act, and as such, constitutes an FTA to negative face. As
responding to the survey is not a requirement, the sender must appeal to the addressee to
secure their participation or predispose them to participate. Thus, the sender begins with
a greeting – one that sends good wishes for “St. Patrick’s Day” (17 March). The sender
then acknowledges the imposition, i.e. the potential drain on the receiver’s workload for
having to read the email (perhaps more than once), “Sorry for cross posting (or if you
have already completed)”. This is mitigated frst by issuing an apology and then by
offering reasons for the message, “but I wanted . . .”. Before asking for participation, the
sender softens the imposition by issuing a compliment, describing the responses to date
as “amazing”. This anoints the positive face of the receiver. The FTA to negative face is
mitigated further through depersonalisation, in which the action required from the
receiver is frst characterised as noun rather than a verb: “a fnal push for participant
collection”. The nominal group “participant collection” is presented as a phenomenon
and does not need an Actor, grammatically speaking. This mitigates the role of the
sender. Moreover, the sender’s role is as a Carrier in a Relational process of having (see
Chapter 9), “I am having a fnal push . . .” rather than a Material process of doing (contrast
this with ‘I am pushing you for . . .’). The asking is hedged with the modal “would”,
which softens the force of the request and is depersonalised somewhat with “anyone”
rather than the more direct ‘you’. Finally, deference to a higher authority via the
university’s ethics committee offers a nod to the legitimacy of the survey, which would
have relevance for the receiver, and which shows goodwill.
168
 POLITENESS

(B) The bank meeting

JIM: I got a couple things here, [. . .], just to kind of bring us up to speed, Matt has
tentatively accepted our offer for employment, but since he’s going to be a dual
employee, th- you really don’t need to put this in the minutes. JOE: Okay.
Jim, the president, and therefore the hierarchically powerful (P) interactant, imposes upon
Joe, the employee, by asking him not to include something in the minutes. As minute-
taker, Joe would ordinarily decide what goes into the minutes and what doesn’t, and so
should be free to make those administrative decisions himself. Jim imposes his will on
Joe. This constitutes an FTA to negative face. However, the imposition is ranked as low
(R) because power is unequal (P) (because Jim is more senior, he has more freedom to
impose on employees). Even so, Jim hedges his FTA by using hedging terms like “really”
in “you really don’t need to put this in the minutes”. By framing the imposition as
something that Joe doesn’t “need” to do Jim demonstrates that he is cognisant of Joe’s
negative face-wants. Presented like this, Jim appears to be considerate, which mitigates
the imposition on Joe who is effectively being told what to do (or rather, what not to do).
Brown and Levinson offer this formula for ‘calculating’ the weightiness of an FTA,
where W=weightiness, x=FTA and (S, H)= speaker, hearer:

Wx = D (S, H) + P (H, S) + Rx

In our example of Jim and Joe, we can put it like this:

W(you really don’t need to put this in the minutes’) = D (Jim, Joe) +
P (Jim, Joe) + R ‘you really don’t need to put this in the minutes’

Wx = >D + >P + <Rx

(C) The X-Factor audition
The frst FTA is to the auditionees, Liddia and Ryan. It is a criticism of their performance
and so is oriented to positive face. The judge, Gary Barlow, delivers it with “you haven’t
got a voice between you. You couldn’t get a gig in an empty pub”. This is harsh, yet his
position as judge and the reputation of the show as the source of some acerbic insults to
wannabe performers means that the FTA is perhaps not uncharacteristic nor uncommon,
and so not as impactful within the culture of TV talent shows. This could be why it is not
heavily mitigated. He does say “There’s nothing to critique here”, which provides
reasons for his negative review. He fnishes with another mitigating approach, this time
a hedge in “There’s just no” before he is interrupted. While Barlow focuses his critique
on their singing ability, Louis Walsh goes further and insults their character, “You are
the two most deluded people I have heard all day”. This is a rather bald FTA to positive
face and constitutes ‘impoliteness’ (see further reading). More unusually, the contestant,
Ryan, hits back at Louis with an FTA to his positive face, “You wouldn’t know what
talent was . . .”. The power differential between them makes this more threatening than
Louis’s FTA. While judge Nicole exchanges some banter with them, Gary urges her to
decide “is it a ‘no’ . . .”, which constitutes an FTA to her negative face. He poses it as a
question rather than an imperative, and so mitigates it a little. He also offers reasons for
169
 POLITENESS

his FTA “We’re waiting . . . ”. At this point, Ryan interjects telling Gary “There is no
need to be rude”. This is an FTA to Gary’s positive face. Ryan mitigates it with a tag
question “is there?” which effectively seeks agreement. Ryan then utters a series of FTAs
to Gary’s positive face, ironically by saying what he will not say “’cos I’m not that rude”.
In uttering the latter, Ryan is effectively able to absolve himself from insulting Gary,
which he does indirectly, and at the same time preserves his own positive face “but I’m
not going to [say all that] ’cos I’m not that rude”. The cat is well out of the bag, however.
Which, ironically, is what their singing sounds like.

Notes
1 ‘Please’ is not a requirement. But it will probably increase the likelihood of getting the outcome you
want.
2 Grimes is a Canadian musician, singer, and producer.
3 Ed Sheeran is a British singer popular for his catchy ballads. If by some twist of fate Ed Sheeran
chances upon this chapter we would like it known that our example does not refect our views on his
musical career. We also think ‘Bad Habits’ is a banger.
4 This is not always a ‘way out’ of an FTA as there is some communicative force in saying nothing,
contrary to Keith Whitley (or Ronan Keating depending on your age), who claims ‘you say it best
when you say nothing at all’. https://www.youtube.com/ watch?v=xNU7iIdw7Ss
5 We say ‘essentially’ because this utterance does seek to have the music-loving roommate turn down
the volume (so constitutes an off-record imposition). Additionally, it does not avoid discord, but
actively courts it (‘who hurt you’?) and so, in certain contexts could be read as an FTA to positive
face.
6 Again, sorry Ed and Ed fans. Although, to be honest, one of us is still salty over ‘Galway Girl’.

References
Bousfeld, D. (2008) Impoliteness in interaction (Vol. 167). Amsterdam: John Benjamins.
Brown, P. and Levinson, S. C. (1987) Politeness: Some universals in language usage
(Vol. 4). Cambridge: Cambridge University Press.
Cowell, S. (2013) The X-Factor. Series 10. Syco Entertainment Thames.
FremantleMedia.
Culpeper, J. (1996) ‘Towards an anatomy of impoliteness’. Journal of pragmatics, 25(3):
349–367.
Goffman, E. (1967) Interaction ritual: Essays on face-to-face interaction. Chicago:
Aldine.
Gu, Y. (1992) ‘Politeness, pragmatics and culture’. Foreign Language Teaching and
Research, 4: 10–17.
Jucker,A. H. (1993) ‘The discourse marker well: A relevance-theoretical account’.
Journal of Pragmatics, 19(5): 435–452.
170
 POLITENESS

Lakoff, R. (1973) ‘Questionable answers and answerable questions’. In B. B. Kachru, R.

B. Lees, Y. Malkiel, A. Pietrangeli, and S. Saporta (eds), Issues in linguistics: Papers
in honor of Henry and Rente Kahane, pp. 453–467. Urbana, IL: University of Illinois
Press.
Leech, G. (1983) Principles of pragmatics. London: Longman.
Leech, G. N. (2014) The pragmatics of politeness. Oxford Studies in Sociolinguistics.
Oxford: Oxford University Press.
Owen, M. (1981) ‘Conversational units and the use of “well . . .”’. In P. Werth (ed.),
Conversation and discourse, pp. 99–116. London: Croom Helm.
Yangmeishan, Z. H. O. U., and Wencheng, G. A. O. (2019) ‘Socio-pragmatic failure of
Chinese non-English majors in intercultural communication’.
Sino–US English Teaching, 16(5): 209–215.

171
 Metaphorical meanings in
8 discourse
Metaphor and metonymy

Introduction
In Chapter 5 we explored semantics and different types of meanings. In this chapter,
we continue our exploration of meaning but this time we look at fgurative or non-
literal meaning. Focusing on metaphor and metonymy, we outline key theories about
fgurative meanings including Conceptual Metaphor Theory (CMT), and we align with
existing scholarship that posits that metaphors and metonymies are not simply textual
phenomena but conceptual phenomena too. Throughout the chapter we use examples
from a range of discourse contexts, including political commentary, TV sitcoms, song
lyrics, and everyday expressions. We explore how fgurative language can prompt us
to think about things (e.g. people, events, activities) in different ways and how this
might impact on our social and cultural understandings of such things.

What is a metaphor?
‘Metaphor’ derives from the Greek ‘meta+pherein’ meaning ‘over’ + ‘to carry’, which
becomes ‘to carry over’ or ‘transfer’. So, in a metaphor, something is ‘carried over’
or ‘transferred’. But what is transferred and why? And what is the outcome of the
transfer? These are questions we also need to address in order to answer our main
question: ‘what is a metaphor’. Most dictionary defnitions refer to ‘metaphor’ as an
‘expression’ (also Black 1955: 275) in which one thing is expressed to describe or
‘substitute’ for another unrelated thing. At its simplest, this is what a metaphor is and
does. It ‘transfers’ one meaning or sense of something to another. But metaphor is not
simply an ‘expression’ that operates on a linguistic level, although linguistic
expressions trigger metaphors. As we shall see, metaphors also operate on a
conceptual level because they invite us to think about one thing in terms of another.
Writing in 1936, I. A. Richards says this about metaphor:

In the simplest formulation, when we use a metaphor we have two thoughts of

different things active together and supported by a single word, or phrase, whose
meaning is a resultant of their interaction.
(Richards 1936: 93)

DOI: 10.4324/9781003351207-8
Richards’s view of metaphor is the basis for the cognitive view that we present in this
chapter – metaphor as an ‘interaction’ between two ‘ideas’ or, in current terms,
‘concepts’ and ‘conceptual domains’. However, already we are dealing with two terms
that need explanation: concept and conceptual domain.

172
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Starting with ‘concept’, recall that in Chapter 5, we talked about semantics and the
nature of signs. We used the example of a ‘chair’ as being a sign by virtue of it having
both a linguistic expression (the word ‘c-h-a-i-r’) and a mental construct (our own
mental image of a ‘chair’). The latter is a concept in that it is an image or idea
conceived in our minds regardless of whether it exists in the world. The sign, then, is
the totality of these two things; it comprises both a word and concept (and if we throw
a real chair into the mix we get a third part, the ‘referent’). When we use a metaphor,
we are using two mental constructs or concepts across which we transfer meaning(s).
Which brings us to ‘domain’. Studies of metaphor often use the term conceptual
domain to account for concepts at different sides of the transference. A conceptual
domain involves a broader set of meanings or ideas. So, for example, the concept
‘chair’ sits within a conceptual domain that includes other closely related concepts
such as legs, back, seat, and ‘to be sat on’. So, a domain is a complex of related
concepts and meanings; they are networks of conceptual knowledge and they are
activated by lexical items (Langacker 2008: 47). With a metaphor, a concept from one
domain (the target) is talked about in terms of concepts from another domain (the
source) The result is that attributes associated with one domain are ‘carried across’ or
mapped on to the other. So now we have a third term that needs some clarifcation:
mapping.
When a person encounters a lexical item being used metaphorically (like ‘virus’ in
‘computer virus’) they draw parallels between it (virus) and the concept to which it
likely targets (a piece of digital code that can be destructive to digital fles and
software).1 Thus, the source domain is used to understand or reconceive the target
domain. Properties are selected from the source domain (invited by the lexical term
that triggered the metaphor) and applied to the target, generating a more accessible,
nuanced, or novel way of understanding the target. By ‘parallels’ we mean properties
that could conceivably or plausibly apply to both domains, and thus interact. Figure
8.1 shows the relationship between source and target in a metaphor.
Often metaphors are used where a non-metaphorical expression could be used. For
example, a classmate with three essays to produce in one week may tell you they are
‘drowning in work’ when they could have said something like ‘I am really
overworked’. But metaphor is not simply a process of substitution. Therefore, the use
of the metaphor ‘drowning’ must generate some additional meaning to justify the
hearer putting in the processing effort to work it out. The effect, then, of using a
metaphor can be to amplify or enrich the interpretation of the target domain over and
above what can be achieved by a non-metaphoric expression. These additional or
enriched interpretations are not discernible from the more literal option of ‘I’ve got
too much work’. The metaphor then, gives a fuller picture of the state of the speaker.

173
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Figure 8.1 Metaphorical relationship between source and target domains

The conceptual basis of metaphors

Often, we have recourse to metaphors when we struggle to communicate an abstract
phenomenon. Mental health, for instance, is diffcult to discuss literally. We frequently
talk about our mood being ‘up’ or ‘down’ to signify when we are happy or sad. In
other words, we invoke one conceptual domain ‘orientation’ (up/down) to refer to
another (‘mood’). There are many expressions that draw on the concept of orientation
in relation to mental health (e.g. ‘I’m really low’), but, of course, we cannot literally
have a raised or lowered mood. The metaphors in the italicized comments below relate
to declining mental health; the frst from a celebrity ftness personality, Joe Wicks,
describing the effect of a Covid lockdown and the second from guidance issued by the
UK National Health Service (NHS) during the same period (emphases added).
(i) It’s okay to be upset and feel down right now [. . .] I don’t have anxiety or
depression, but right now I feel so low, and so down.
(Guardian, 5 January 2021)

(ii) At times like these, it can be easy to fall into unhealthy patterns of behaviour that
end up making you feel worse.
(NHS website, https://www.nhs.uk/every-mind-matters/
coronavirus/mental-wellbeing-while-staying-at-home)
Unsurprisingly, such metaphors are called orientational metaphors. They are
expressed more broadly as good is up and bad is down which points to their

174
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

conceptual origin from which various metaphorical expressions such as ‘I’m feeling
a bit down today’ are derived.2 The basis for orientation metaphors like good is up
and bad is down is, according to Lakoff and Johnson, physical: when we are sick, we
are usually physically lying down, when we are well, we are up and about.
Additionally, Lakoff and Johnson note that “drooping posture typically goes along
with sadness and depression, erect posture with a positive emotional state” (1980: 15).
In this way, we can see how the physical positioning of up/down can have useful
metaphorical applications.

Activity 8.1 Orientational metaphors – positive or

negative evaluation?

Read the following extract from an article by Stewart Lee in the Guardian newspaper on 10 July
2022 and try to identify the orientational metaphor – how is it lexically expressed? The piece
was written a few days after ex-British prime minister, Boris Johnson, had resigned following a
spate of controversies. Lee writes:

Johnson’s premiership, a mindless rotting meat zombie held together with the Sellotape™®
of lies and the sticky excretions of his own spaff-faucet, began a sudden and dramatic slide
into the ocean on Wednesday, like an arctic ice shelf made entirely of frozen shit.
(Lee 2022)

You might also want to consider the co-text – how might you interpret the orientational
metaphor in light of its appearance alongside so many other pejorative words?

The experiential basis of metaphors

As we can see from the orientational metaphors discussed above, metaphors can derive
from experiential phenomena. If we take an abstract concept such as ‘love’ as a target
domain, we can see that many ‘love’ metaphors are based on physical sensations and
sensory experiences of, for example, temperature (e.g. ‘the relationship is heating
up’), hunger (‘starved of love’), thirst (‘I can’t quench this desire’), pain (‘I’m aching
to see you’), light (‘you light up my life’), and sound (‘my heart sings when I am with
you’). Others are based on connections between experiences (Lakoff and Johnson
1980; Grady 1997).
For example, we can feel the heat emanating from another person when we are
physically close to them and, usually, we to choose to be physically close only to those
we care about. We can therefore connect physical proximity with warmth, both of
which we might also connect with affection. Consequently, we might understand the
concept of affection via the conceptual metaphors affection is closeness, and affection
is warmth (Grady 1997; see also Duffy and Feist forthcoming, 2023). We can also
connect other emotional states with heat via our experience. For example, we know
that we can get hot when we are stressed, embarrassed or angry and it is known that
physiological effects (raised body surface temperature and increased heart rate) are
triggered by a release of hormones in the body when we experience certain emotions.

175
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Such physiological experiences can explain conceptual metaphors such as love is heat
(see, for example, Goatly 2007) and anger is heat (see Lakoff 1987; Kövecses 1986,
1990). These conceptual metaphors (in small capitals) are also known as primary
metaphors (Grady 1997) and these can be realised linguistically via a vast array of
metaphorical expressions (words that trigger metaphors are sometimes called ‘lexical
metaphors’), as the above examples have shown. In fact, many metaphorical
expressions can draw on the same source domain. Conversely, the same source
domain can be used to metaphorically understand a range of different target concepts.
Consider the following metaphorical utterance:

(iii) ‘I see what you mean’

The concept ‘seeing’ in (iii) is a sensory one – perception – that is not literally true
when it applies to ‘mean’. We cannot, after all, ‘see’ what someone ‘means’ when
they say or do something because meaning does not have physical form; we can only
understand what someone ‘means’. Thus, ‘see’ is a metaphor in utterance (iii). So how
do we get from the realm of invisible cognition to being able to ‘see’ meaning? It is
useful to regard metaphors as operating on a generic level and a specifc one.
Conceptual metaphors operate at this generic level, and lexical metaphors are their
specifc instantiations (the relationship is structurally similar to that of genus to
species).3 In fact, several lexical metaphors might stem from one conceptual metaphor
and so share the same “underlying organisation” (Simpson 2014: 43). You might have
encountered expressions like ‘look at things from where I’m standing’ or ‘he’s got his
blinkers on when it comes to promotion’, or ‘I smell a rat’, and so on (‘rat’ is also
metaphorical!). All of these expressions systematically construe cognition as
something to do with ‘seeing’ or ‘smelling’ – in other words, ‘sensing’. They are all
conceptually linked to sensing. The conceptual metaphor, then, that captures these
corresponding metaphoric expressions is cognition is sensing. In this way such
“conventional ways of talking” about knowledge as something we can sense
“presuppose a metaphor we are hardly ever conscious of” (Lakoff and Johnson 1980:
5). The specifc lexical metaphors from this conceptual one derive from an overarching
source-to-target mapping. In Grady’s terminology, this hierarchical relationship of
governance from the primary (generic) level is the source of a “basic mapping” that is
“lacking in detail” (Grady 1997: 33). The lexical (specifc) metaphors provide the
detail by putting meat on the bones, as it were.
There are, though, different levels of conceptual generality when discussing
conceptual metaphors. Readers could, if they wanted, take the level of abstraction to
its most general (the ‘bones’ we refer to above) or to a lesser level (somewhere closer
to the ‘meat’) when analysing metaphoric mappings. In the diagrams and analyses that
follow we will use ‘CONCEPTUAL’ to refer to the most general level and
‘conceptual’ to signal less general mappings. This is because it can be useful to group
conceptual metaphors into categories on different levels of abstraction. We show this
in Figure 8.2.

176
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Figure 8.2 Mapping between conceptual and lexical metaphors

In Figure 8.2, we have tried to make knowledge visible! We show an overarching

CONCEPTUAL metaphor, which is rather abstract, leading down to a less general
mapping that is still conceptual, and then fnally down further to lexical metaphoric
expressions. The CONCEPTUAL level is ‘cognition is sensing’. The level below it
(labelled ‘conceptual’ in lower-case letters) refnes the CONCEPTUAL metaphor
somewhat and groups the lexical metaphors below it together. The bottom two levels
derive from the top and share its underlying structure.
In Figure 8.3 we show two different but related mapping trajectories at the
conceptual level and the lexical metaphors that can result. First, (1a), knowledge is
light, and second, (1b) knowledge is vision.4 You will see that (1a) and (1b) are
governed by the CONCEPTUAL metaphor above it.

177
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Figure 8.3 The different levels of metaphor production for ‘knowledge is light’ and
‘knowledge is vision’

In terms of doing discourse analysis, noting different levels of abstraction allows

us to draw together patterns in metaphor use. It also helps show how metaphors are
conceptual and not simply linguistic phenomena. Indeed, this is the basis of Lakoff
and Johnson’s Conceptual Metaphor Theory (CMT) (1980) which takes as its central
premise that we think metaphorically (e.g. we think about mental health as being ‘up’
or ‘down’).

Activity 8.2 Metaphor in literary fiction

Below is a text example from a short story by Haruki Murakami (1994). Which words in the short
extract are being used metaphorically? Can you identify the generic CONCEPTUAL metaphor
from which the lexical metaphors derive?

“potentiality knocks on the door of my heart”

Novel metaphors
Some metaphors are so entrenched in language that their meanings are accepted almost
instinctively. The lexical metaphor ‘enlightenment’ is one such example. Knowledge
as something ‘illuminating’ is so conceptually embedded and accepted that we no
longer consciously perceive of it as metaphoric. In fact, if you asked someone to spot
the metaphoric content in this noun phrase ‘the path to enlightenment’ they might offer
‘path’ frst (from the life is a journey conceptual metaphor). However, some
metaphoric mappings have less obvious connections between source and target.
Consider the metaphor in (iv) below.
You might fnd it unusual.
178
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

(iv) The shop was hiving with people

In this example, the word that acts as the lexical trigger for a metaphorical
interpretation and entry point to the metaphoric source domain is ‘hiving’. Before
reading further, refect on (iv) and note your responses to it. What meaning do you
think it is targeting?

Activity 8.3 Real examples of the ‘hiving’ metaphor

Let’s look more closely at the ‘hiving’ metaphor. Below are some real examples of speakers using
the metaphor (emphasis added) in online customer reviews (a, b, e) and news articles (c, d):

(a) “Discovered Muriel’s Restaurant after visiting the Natural History Museum and heading
for the tube back to our hotel. The place was hiving with people which is always a good
sign.”
(Tripadvisor reviewer from Ballygally, north of Ireland,
21 April 2015 https://tinyurl.com/3ya39ur5)

(b) “Good atmosphere – you can sit at tables, counter or bar. Lavery’s in general is buzzing
currently. I watched a game on TV there recently and the place was hiving. Two thumbs
up for The Woodworkers.”
(Tripadvisor reviewer from Belfast, north of Ireland,
16 August 2015 https://tinyurl.com/4c2aents)

(c) “The frst thing that was on our agenda was people’s safety,” said Mr Rainey [chairperson
of the centre hosting the event]. “We don’t want to be seen to be jumping to conclusions.
To me, I thought the place was hiving with security staff on Saturday night.”
(BBC NI News, 6 November 2013
https://tinyurl.com/2p885ua5)

(d) “‘I remember driving up – we’d obviously have been there a couple of hours before the
game – and even on the outskirts of the town, the place was hiving. The crowds were
unbelievable,’ remembers Strain.”
(Belfast Telegraph, 25 April 2020)

(e) “the number of people in the spa impacted the quality of our stay. After check in, we put
or robes on and went down to the spa – and it was hiving with people.”
(Tripadvisor reviewer from Belfast, Northern Ireland, 8 April 2019
https://tinyurl.com/bder3euu)

What, if any, common denominators characterise these examples? What might they tell you
about the relationship between discourse communities and certain metaphors? How might the
use of this particular source domain evaluate the activity it targets?

We provide answers at the end of the chapter.

179
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Analysing the ‘hiving’ metaphor

Let’s take a closer look at the ‘hiving metaphor. In it, the lexical trigger for the
metaphor is ‘hiving’ because shops (and shoppers) do not literally ‘hive’. So, a non-
literal meaning must be invoked in order to make the utterance make sense. The
mapping might go something like what we present in Figure 8.4.
In Figure 8.4, there are two domains: the source domain is ‘hiving’, which is to say,
the word ‘hiving’ triggers a metaphor that evokes the status of the shop as having
properties or characteristics of beehives (let’s call this ‘hive-ness’). They include
‘physical space’ (1), ‘contains bees’ (2), are ‘densely populated’ (3), are sites of
‘purposeful activity’ (4) making honey, and so on. The target domain is the shop’s
(very busy) status, and it has properties of its own – for the sake of consistency, we’ll
call these ‘shop-busyness’. To make sense of the utterance then, we select and map
properties from the source domain to the target domain. Which ones we map depend
on which appear to be the most contextually relevant and may depend on our cultural
background (Lakoff and Johnson 1980: 142–144). There could be variation, then, in
what we map from the source domain to the target and how we conceive of both
domains; as Steen and Gibbs point out:

such variation between people in a community may have important effects on their
experience of specifc linguistic expressions as conventional or new, easy or
diffcult, appropriate or inappropriate, and so on, and may infuence people’s
production and comprehension of specifc linguistic expressions in concomitantly
varying ways.
(1999: 3)

Returning to our source domain, we map some of the ‘hive-ness’ characteristics onto
the target domain of ‘shop-ness’. The presence of the source domain ‘hiving’ compels
us to project a “set of associated implications” (Black 1979: 28),

180
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Figure 8.4 Metaphoric mapping for ‘hiving’

that are capable of being asserted of the source domain. In other words, we map
characteristics of ‘hive-ness’ on to the target ‘shop-ness’ to generate an understanding
of the target domain that has a contextually meaningful parallel with the concept ‘hive’
(otherwise, why use the metaphor?).
Making sense of metaphors sometimes requires extra cognitive work on the
hearer’s behalf because the intended meaning of a metaphor is not that which is
literally stated. In other words, when we say the shop was ‘hiving’ our knowledge of
shops helps us understand that they do not tend to be densely populated by bees
working to produce honey. To make sense of the utterance we must ‘pick up’ the
metaphor’s implicated content which arise from what Black calls a “system of
relationships” between source and target domains (Figure 8.4) and construes as an
“implicative complex” (1979: 28, our emphasis) that demands uptake from a hearer.
This leads us to ask the following questions: If metaphors demand interpretation,
then what properties from the ‘system of relationships’ do we map? And why? And
what can these mappings do to our knowledge of each of the domains (source and
target)?
What do we map and why?
181
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

For the ‘shop’ to be ‘hiving’ we assume it to share a property exhibited (prototypically)

by hives, namely that the shop is uncharacteristically, even frenetically busy. In fact,
this is a more precise rendering of our target domain:

Source: ‘hiving’
Target: ‘the shop is uncharacteristically, even frenetically (threateningly?) busy’

It is important to note, as Lakoff and Johnson do, that “metaphorical structuring [. . .]

is partial, not total. If it were total, one concept would actually be the other, not merely
be understood in terms of it” (1980: 13). This means that there are things we do not
map, for example, the second property in Figure 8.4, ‘contains bees’, because it would
not make much sense to say that shops contain bees. We do map some of the properties
of bees, though (e.g. that they are densely populating the hive). This is only one of
several properties of hives (Figure 8.4). The metaphor prompts us to perceive shared
(or potentially shareable) characteristics between the two domains which we map from
source to target domain.
Earlier we suggested that commonplace knowledge of either or both domains in a
metaphor differs according to cultural and personal experience (real or imagined!) of
them. If you were badly stung by a bee or bees as a child and grew up being afraid of
bees and hives then you might map this fear from the source domain ‘hiving’ to the
target domain ‘shop’ and interpret the metaphor as conveying a fearful or potentially
dangerous or threatening shopping environment. It might lead you to view shopping
as a fearful activity even if you had not considered this before. In this way, the
metaphor can lead you to see the world in a certain way; it has “the power to create a
new reality” (Lakoff and Johnson 1980: 145). The converse of this is also true – if you
despise shopping even before encountering the metaphor then you may be primed to
pick out properties from the source domain that convey a negative view of shops and
shoppers. Alternatively, if you are a bee-lover this property may not transfer across in
the metaphorical mapping. The point is that your interpretation of a metaphor will be
shaped by and through your own personal experience, and this uptake can shape or re-
shape how you view the target/source domain.

Novel metaphors in song lyrics

Aristotle believed that the mastery of metaphor was “a sign of genius” because,
according to him, “a good metaphor implies an intuitive perception of the similarity
in dissimilars” (2008: 1459a 3–8). But what does it mean to have a ‘good’ metaphor?
(for more on this, see Black 1977; Giora et al. 2004; Littlemore et al. 2014). It depends.
It might mean that the metaphor does a good job of communicating the desired
mapping. However, there is a particular kind of metaphor that Aristotle may have in
mind here – the ‘creative’ or ‘novel’ type, often found in literary fction. Aristotle’s
view is partly responsible for the rather narrow view that metaphors are only literary
devices (as opposed to conceptual, everyday tools).

182
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Let’s start with Aristotle’s claim. If we assume by ‘good’ he meant ‘novel’, his
point appears to be that the ‘genius’ of creating a novel metaphor derives from your
ability to introduce a rather unusual way of seeing one thing – the target domain – in
terms of something else – the source domain – for which a connection is not so
obvious. Such metaphors self-consciously draw attention to themselves. Novel
mappings can be found in many discourse situations including fction (as noted above)
and song lyrics.
For example, Alanis Morissette sings about “transparent dangling carrots”
(Morissette and Ballard 1998) and Twenty One Pilots have a song called ‘Neon
Gravestones’ (Twenty One Pilots 2018) in which the gravestones “call” for the
“bones” of the speaker. Let’s assume that the target concepts in these examples are
‘motivation’ and ‘death’ respectively. These metaphors help us to look upon the target
in a fresh way. For example, we interpret ‘transparent dangling carrots’ as a reference
to the speaker’s motivation for which the primary conceptual metaphor is emotional
states are objects. This comes from the practice of dangling a carrot in front of a
donkey to motivate it to walk. As the donkey never really gets the carrot, there is often
a negative connotation to such ‘dangling carrots’, but whether this is carried over in
Morissette’s song is anyone’s guess. The metaphor is embellished with the addition of
‘transparent’, implying that the speaker is not able to visibly perceive the metaphoric
carrot, so the motivation is diffcult to pin down or the motivation was false or
meaningless. When Twenty One Pilots sing of ‘neon gravestones’ that ‘call’, the lyric
invokes an objects are animate entities metaphor by humanising a gravestone that is
capable of ‘calling’. That the gravestone is ‘neon’ leads the hearer to interpret it as
having the property ‘attractive’ or ‘welcoming’ (on the basis that neon signs tend to
attract visitors to events or locations). This is incongruous because gravestones are not
‘clubs’ that people are motivated to join or enter. Indeed, in our interpretation of the
song, we read it as an appeal to those considering suicide to resist the ‘call’ of ‘neon’
gravestones. In cases of novel metaphors5 like these, the ‘accepted commonplaces’ are
passed over in favour of what Black calls “specially constructed systems of
implications” resulting in “made to measure” mappings (1955: 290).
Before reading on, have a look at Activity 8.4.

Activity 8.4 Making meanings through metaphoric mappings

Consider the meaning of the following line from rapper Akala’s song, ‘Shakespeare’ (Akala
2007).

“keep spittin’ ya darts, mine is javelins”

Before reading further, can you identify the source and target domains in Akala’s song lyric
example? What properties do you think might be mappable? What meaning or meaning
potential might these mapped properties give rise to? Cognisant of Black’s interactionist view
of metaphor, do the mappings lead you to view the source or/and target differently in light of
the metaphor’s context? How so?

183
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Perhaps your answers to the frst question in Activity 8.4 yielded something similar
to the following.
Source domain: Sharp objects
Target domain: Rap
The conceptual metaphor that governs these lexical metaphoric expressions is
words are objects.
Regarding the second question, some of the following associated implications may get
mapped.
• sharp objects are gradable on several scales that include weight, size, reach
• “javelins” generates a potential for damage to a stimulus (it would be less damaging
to get hurt by a dart than a javelin!)
• superiority (darts are inferior instruments to javelins if what you want to do is throw
them a great distance)
The mapping encourages us to understand the quality of rap in a gradable manner in
the same way that the relationship between darts and javelins is scalable/gradable. By
this, we can conclude that Akala’s rap is more impactful, because it is sharper, and
has a greater reach or scope than his contemporaries’ rap. Notice that the comparison
generated by the metaphor introduces this scale as one of the implicatures that we
derive from the metaphor’s use. We may also import the associated implication that
‘sharpness’ carries in other metaphors such as ‘sharp wit’ (good),‘sharp tongue’ (bad,
but depending on whether you’re on the receiving end or not), ‘sharp dresser’, and
so on. We may be encouraged to view rap in new ways as a result of the mappings.
This new way of seeing can refresh or disrupt our perception of the target domain. It
can challenge us to see the target domain in a new way (sometimes forever!).

Extended metaphors
Sometimes, metaphors are creatively expanded or ‘extended’. This means that once a
source domain has been invoked, other properties of that source domain come into
play in lexical metaphoric expressions that help to elucidate the target concept. We
offer an example in the following activity. Have a look and try to work out what is
happening metaphorically before reading further.

184
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Activity 8.5 Extended metaphors

In the American medical comedy TV show Scrubs (Callahan and Kerns 2004), the character Dr
Perry Cox, who is known for his acerbic wit (his discourse offers many fne examples of ‘spittin’
javelins’), participates in an exchange with his psychology colleague, Dr Molly Clock, in which
both employ the same metaphor in different ways in a short exchange about types of people.
Whereas Clock claims that people are not “pure evil”, she concedes that some have a “hard
outer shell”. An exchange follows whereby Cox and Clock conceptualise human personality
types as being “nougaty” or having “creamy centres”, before Cox concludes the exchange with
QR 8.1 Comedic effect from extended the retort that people are mostly
metaphor “bastardcoated bastards with a bastard
flling”. You can scan QR code 8.1 to watch
the clip (from 5:55 mins).

There are different concepts at play in their exchange – can you tell whether there is a primary
conceptual metaphor at work, and if so, what is it?

So, what is going on in the example from Activity 8.5? First, Molly attempts to
elucidate the target domain “personality type” through a ‘container’ source domain
using a conceptual metaphor people are containers (see Lakoff and Johnson 1980 for
185
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

a full description of this type of primary/conceptual metaphor). Molly introduces a

metaphoric interpretation through the lexical terms “hard outer shell” and “creamy
centre”. These are examples of lexical metaphors. If you guessed a primary conceptual
metaphor, did you get ‘container’? The ‘container’ (of sorts) turns out to be chocolate.
The two properties “hard outer shell’ and “creamy centre” are likely to be from the
same source domain, so that “hard shell” does not, for example, refer to a turtle. This
is because the presence of two lexical metaphors from the same conceptual domain
(‘shell’ and ‘creamy centre’ are both properties of chocolates) are more likely than
two lexical metaphors from two different conceptual domains assuming that both are
being cooperative. Therefore, “shell” and “creamy centre” prime the hearer to interpret
a confectionery link rather than a reptilian one. Then, Molly extends the chocolate
container metaphor by introducing new conceptual elements of the chocolate source
domain, e.g. “nougaty”. So now we have “creamy centre” pitched against “nougaty
centre” in a kind cline of hardness that concretises ‘evil’ and ‘good’ personality types.
This is a common conceptual cline in which ‘soft’ is ‘good’ and ‘hard’ is
bad/diffcult/stubborn and so on. It is fairly conventionalised in lexical metaphors such
as ‘you’re such a softie, and ‘don’t be too hard on them’.6 Dr Cox initially negates
Molly’s chocolate metaphor but then merges it with his own creative metaphoric
source domain of “bastards” producing a mixed metaphor so that people can be said
to be “bastard-coated” and have “bastard flling”. Of course, he is not trying to say that
people are chocolates either, but he is capturing the same metaphoric mapping of
people as confectionery ‘containers’ where their “shell” and “flling” are more novel
because both are . . . well . . . ‘bastard’. In Dr Cox’s view of people, they are not simply
‘bastards’ in the nominal sense, but also in the biological or corporeal sense being
comprised of ‘bastard’ parts flled with ‘bastard’ centres. In this way, Cox’s mixed
metaphor “selects, emphasizes [. . .] and organizes features of the [target domain] by
implying statements about it that normally apply to the [source domain]” (Black 1955:
291–292, emphasis in original). Of course, it is also rather amusing which is really
more the point.

186
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Activity 8.6 Metaphorical extension and elaboration in BBC

TV’s Dragon’s Den

Dragon’s Den, a BBC reality series that broadcasts entrepreneurial pitches to an audience of
wealthy, shrewd investor ‘dragons’, incorporates a host of metaphoric puns in its narration of
each of the entrepreneurial pitches. The show’s presenter, Evan Davis, provides the voiceover
narration to the goings-on in the ‘den’. After reading about metaphorical extension, see if you
can match the investment-seeking pitches to the groups of metaphors Davis invokes in his
corresponding comments from episode 13 in series 18 (Dragon’s Den 2021). Then identify the
lexical metaphors and see if you can link any of them to an overarching conceptual metaphor.
Do you see any metaphorical patterns between theQR pitches and thethe
8.2 Match lexical metaphors?
metaphor group to the
entrepreneurial pitch
You can watch the episode by scanning QR code
8.2 (if you can access BBC iPlayer).

The pitches:
Pitch 1: A ‘Rev-off’ device to turn a regular push bike into an e-bike
Pitch 2: An optometrist duo pitch a screen that limits blue light exposure from digital devices.
Pitch 3: An entrepreneur is selling trees to offset the carbon footprint via an app Pitch 4: A
busy mum who is a teacher is pitching her ‘family-themed’ fruit preserves Pitch 5:
Comedian, Darren, is pitching his idea for a ‘dog poo’ lifter.

Metaphor group A
“Will he scoop up an investment?”
“. . . fnd out if there’s any brass where there’s muck”
“. . . [his pitch] is anything but a walk in the park” “[the
product] falls foul”

Metaphor group B
“[Will she] be the toast of the den?”

187
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

“[As a teacher and mum] she is spread pretty thin”

“stuck into questions”
“a ready-made partnership”
“sweeten a potential deal”
“sticky situation”
“sweet success”
“pump up the jam business”

Metaphor group C
“[The duo] have their sights set on two particular dragons”
“catch the eye of the dragons they’re about to go face-to-face with”
“[Will they] green light a deal for the blue light duo?” “see
eye-to-eye”

Metaphor group D
“stumps up any cash”
“dig deeper into the [business idea]”
“Will his digging unearth a valuation?”
“the sapling business”
“get to the root [of the fnances]”
“[He] has his investment dreams felled”

Metaphor group E
“back in the saddle [. . .] like riding a bike”
“gives entrepreneurs a rough ride”
“a smooth ride so far”
“hit a stump in the road”
“put the skids on their pitch”
“[The idea] was a non-starter”

Metaphors in political discourse

The use of some metaphors can promote a particular worldview in the mappings they
generate. This is because metaphors can have ideological outcomes. Their use in some
contexts can lead hearers to reconceive of the target domain in ways that reinforce
existing or establish new social and cultural stereotypes. For example, in the UK,
refugees are often discriminated against and negatively presented in public discourse
through metaphors that present their arrival as a negative phenomenon. Such
metaphors may convey the speaker’s attitude or stance towards the target domain. The
following comment by David Cameron, then Prime Minister of Britain, incorporates
a metaphoric mapping linking refugees to insects and reveals ideologically – and
culturally – damaging comparisons.
‘You have got a swarm of people coming across the Mediterranean, seeking a
better life . . .’
(Elgot and Taylor 2015)

188
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Cameron came in for quite a bit of criticism for this comment about refugees entering
Britain, and rightly so. The metaphorical use of ‘swarm’ was the issue for many, as
this “dehumanis[ed] desperate men, women and children” seeking refuge in the UK
(Lisa Doyle, quoted in Elgot and Taylor 2015). The metaphor is the linguistic
manifestation of a conceptual metaphor, people are insects. The source domain
‘swarm’ triggers associated implications, namely visions of insects invading a
physical space. These largely pejorative connotations (see Leech’s meaning types in
Chapter 5) are mapped across to the target domain so that refugees are reconfgured
linguistically and conceptually as parasitic insects. To be more specifc, insects in
‘swarms’ tend to do damage to crops, gardens, and so on (if you’ve ever had to deal
with vine weevil wreaking havoc on your agapanthus you’ll see the point here), so the
parasitic implications can extend to refugees. Indeed, Cameron more or less made this
point a year earlier when writing about his changes to the UK’s immigration policy
(Cameron 2014) in which he fuelled the ‘swarming’ metaphor by saying “most new
jobs used to go to foreign workers”.7 The implication that refugees take jobs from UK
residents is an erroneous, yet popular belief in the UK. The implicated comparison of
refugees to insects incorporates – and makes it easier to accept – this metaphorically
construed ‘invasion’ that extends beyond the land itself into the economy, parasitically
eating up jobs that sustain the inhabitants of that country. This in turn can change how
we look at ‘swarms’ of insects, even just by foregrounding some of their properties
over others.
The metaphor has become popularised to the point that the novelty has worn off.
This is insidious because it naturalises the pejorative discourse about refugees by
making the comparison part of the “background” information (Simpson 2014: 96) we
hold about refugees. Indeed, the metaphor that posits a mapping between insects and
refugees has often been extended as the following show (our emphasis in bold italic).

(a) President Trump in a tweet claimed that Democrats “want illegal immigrants, no
matter how bad they may be, to pour into and infest our country”
(Twitter, 19 June 2018)
(b) Rima and her children joined the stream of refugees on what has become known
as the “ant road”, from Turkey to Western Europe.
(Guardian, 19 December 2015; see Mujagić and
Berberović 2019 for other examples)
(c) Meanwhile, migrants continue to crawl into the country.
(Hart 2021; see also for other examples)

These metaphors incorporate corresponding properties of the source domain, such as

manner of movement (‘crawl’), density (‘swarm’), and so on, and lexicalises them to
develop the primary conceptual metaphor. But they also make it easier to map
properties that may not be intended by the metaphor producer such as ‘disease’ (see
Charteris-Black (2006) for more on the DISEASE metaphor), or parasitic behaviours
(such as feeding off British economy by taking jobs).
189
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Amnesty International (2020) reported on the growing concern over “dangerous”

language choices when referring to refugees and quotes the United Nations High
Commissioner for Refugees as saying that “this type of language stigmatizes refugees,
migrants and other people on the move, that gives legitimacy to a discourse of racism,
hatred, and xenophobia”. Such metaphors invariably stigmatise refugees, legitimises
racism, and perpetuates xenophobia. Of course, it is not the metaphor itself that does
this, but the metaphor producer(s), and to some degree, the metaphor-hearers can be
complicit in this process if, upon making the connection, they choose to commit to the
implicit mapping and accept it as a valid one.8

Activity 8.7 Making meanings through metaphors

What follows is a short exercise to test whether metaphors help us understand concepts in a
particular way when infuenced by a metaphoric source. Make a list of fve totally random nouns
or noun phrases that contain ‘concrete’ concepts. Ask a friend or classmate to make a list of fve
abstract concepts. Now swap your lists with each other. Compare the lists side by side. The list
of concrete nouns is the source domain list and the list of abstract nouns is your target domain
list. Now try to map the source on to the target to create a metaphor that reconstrues the target
according to the source. What properties do you map? Why? How do these mappings steer your
understanding of the target? Does your mapping generate a positive or pejorative position
towards the target?

Now try to fesh out your metaphoric mapping by offering a context or situation in which the
mapping ‘works’. One of us did this exercise with our students and you can see some of the
creative mappings they came up with at the end of this chapter. Can you see how some students
extended the metaphor?

These are the lists that we each came up with.

Abstract nouns is a Concrete nouns
hope sea urchin
power campsite

truth orange
patience tree
evil canal

Now to create some context! Let’s just take one as an example: HOPE IS A SEA URCHIN . This could
derive from a conceptual metaphor EMOTION IS A LIVING BEING . We aren’t too familiar with sea
urchins but working with what we do know, we might say that the property ‘rare’ characterises
sea urchins (relatively speaking) because they are only found in certain waters and at certain
depths, so hope can often feel like it’s ‘rare’ or out of reach; we have to dive deep to get it. Sea
urchins are prickly, so we may map the fact that we cannot hold hope easily in our hands or that
hope is elusive in a more general sense (note how it is almost impossible to make the mapping
without recourse to other metaphors). We might even conclude that the metaphor producer is
feeling hopeless. This affrms the point made earlier that metaphors can say something of the
producer’s view of the world.
190
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Metonymy
Metonymy is connected to metaphor in that metonymic structures are also mappings
and are used non-literally (-ish). In a metonymy, like metaphor, something stands for
something else. The difference is that in a metonymy, the mapping is not across two
separate conceptual domains (as it is in metaphor), but rather, resides within a single
(target) domain accessed through a trigger term that acts as a vehicle to the metonymic
interpretation.

Within-domain mapping
The relationship in a metonymic mapping is within the same domain and so we use
slightly different terms for the two parts of a metonymy. Instead of ‘source’ and
‘target’ the conventional terms are ‘vehicle’ and ‘target’. Because metonymic
mappings are based on a ‘stands for’ relationship, they are typically between a part of
a thing and its whole (part-to-whole). Some examples include producer for thing
produced, cause for effect, type for token, and so on. The relationship also works in
reverse so there are ‘effect for cause’ mappings and so on. Like metaphors, these
mappings operate at a generic primary level with lexically specifc instantiations and
applications. The following sentences contain specifc metonymies. The metonymic
vehicle is in bold type and it carries (hence ‘vehicle’) the metonymic relationship:
a) Have you seen the recent Tarantino? ([flm] producer for [flm] produced)
b) Why are Iceland going back to the polls? (country for the people of the country)
c) Don’t be such a Karen (a member of a set for characteristics of all members in
that set)
d) The bump won’t let me eat spicy food any more (sign of pregnancy for the state
of pregnancy)
The metonymic relationship is in parentheses after each example so that ‘Tarantino’,
a movie producer, is used to stand for the movie he has produced, and so on. In terms
of the primary mapping, example a) is creator for creation, (see Quintilian 1921); in
b) the primary mapping is place for people; in c) it is type for token; whereas in d) it
is effect for cause. Note also, that these metonymies only work if the cultural reference
is known to the interactants, so you’d need to know who Quentin Tarantino is in order
to understand that his flm is metonymically invoked in a), and so on. This emphasises
the point we made earlier that context is really important for making sense of
metonymic and metaphoric mappings.

Metonymy or metaphor?

191
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Grady (1997) offers a useful test for distinguishing between metaphor and metonymy
– the ‘is like’ test. In a metaphor the ‘is like’ test usually works in presenting the source
domain as being ‘like’ the target, for example:

‘The shop was hiving’ > The shop is like a hive.

However, in metonymy, the ‘is like’ test makes no sense, for example:

‘Why are Iceland going back to the polls?’ > *Iceland is like people
(‘Iceland’ stands for the people who live in Iceland)

Ubiquity of metonyms
Metonymies are as ubiquitous as metaphors. We use them every day. Let’s take a
particular type of communication to illustrate this. Emojis are essentially metonymic
signs that invoke a characteristic of a concept or domain to represent or substitute for
the whole concept or domain. In some cultures, champagne is considered a common
element in celebrations. Imagine, then, you sent the following message in Figure 8.5
to your friend or colleague after they fnished a particularly diff cult book chapter:

yay, you finished your

chapter!
Figure 8.5 WhatsApp message

The little champagne bottle conveys the meaning of ‘celebration’ by invoking a

key or salient element of celebration – champagne – through the metonymic mapping
symbol for thing symbolised. Therefore, following Simpson (2014), we can say that
metonymy takes a “salient characteristic from a single domain” and “upgrades” it to
“represent that domain as a whole” (45). Whatever text producers choose to ‘upgrade’
from the target domain is often contextually dependent. For example, if you were in a
hospital and overheard a doctor say to the ward staff offcer, ‘I need to go and prep bed
four’, you probably would not think they were off to tidy up the bedding and fuff the
pillows, but rather, that they were about to administer some pre-surgical drugs to the
person occupying bed four. While this might sound a bit impersonal to say the least,
the issue is that it is a contextually derived metonymy because it takes the salient
characteristic for the participants (i.e. the fourth bed in a ward) to stand for the entirety
of the domain (the patient in bed four).9 To use Forceville’s term, it has ‘relevance’ as
well as salience (2002). It also provides an economical (if impersonal) way of
signalling the target domain in a busy place with a high turnover of people like a
hospital ward. Therefore, the staff choose a characteristic that best captures the target

192
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

domain in the most effcient way for them. The same metonymy may not work so well
in a different context. Imagine you share a bedroom with your sister who has overslept
and is at risk of missing her commute to work. You probably wouldn’t offer to go and
‘wake up bed two’. There is no justifable contextual reason for referring to your sister
in this way, because you know her name, there is probably not a regular stream of
people arriving and leaving your house, and you probably don’t have to spend every
morning going round your family members and waking them up one by one (and if
you did, you’d probably use their names).

Metonymy and worldview

Metonymies, like metaphors, can shed some light on the text producer’s worldview.
To give an example, in her collaboration on the song ‘Girls like you’ with Maroon 5,
the artist Cardi B sings about wanting, among other things, a ‘white horse’ and ‘carats’
(Maroon 5 featuring Cardi B 2018).10 The ‘carats’ is a reference to diamonds through
a unit of measurement for thing measured metonymy. Her invocation of this specifc
characteristic can tell us something of how she (her singing persona) views diamonds.
In this metonymy, diamonds could be referenced by any of their characteristic or
salient ‘parts’: ‘sparkler’ (effect for cause), ‘knuckle-buster’ (effect for cause), and
many more besides. Cardi B’s reference to them as ‘carats’ encourages us to consider
their material worth over and above their aesthetic value as diamonds are valued in
marketable or fscal terms on the basis of their carat weight.

‘Karenymy’
Some metonymies are pejorative – not because they are inherently ‘negative’ but
because their use in particular contexts and by particular speakers makes them so. This
is perhaps less noticeable (and so more insidious) when their use has become systemic.
One example of this is the now common, if rather pejorative metonymy ‘Karen’,
which is used throughout social media to refer to middleclass white women who
complain a lot, usually to assert privilege – a token for type mapping in which the
name or ‘token’ represents or stands for all women of a particular type. Of course, this
information needs to be accommodated by a hearer in order for the metonymy to
work!11 Like metaphors, metonymies can be extended. For instance, there is a ‘Karen’
haircut, ‘a coronavirus Karen’ (she refuses to wear masks, is anti-vax), and so on.
Indeed, the Karen metonymy is so pervasive now that it has caused problems for actual
Karens, as the following tweets from a UK supermarket’s account show:

Sainsbury’s: ‘We won’t be challenging customers without a mask when they enter
or when they are in store since they may have a reason not to wear a mask. Karen’
C.G. (Twitter user): Highly offensive putting Karen at the end . . . stick it’

193
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Sainsbury’s: Hi [C], sorry for the confusion. Karen is the name of the colleague
who responded to the query. Andy’
(Sainsbury’s, Twitter, 23 July 2020)

Metaphors and metonymies

As we said at the beginning of this chapter, metaphors and metonymies are ubiquitous
phenomena. Often, they give rise to meanings by working together. The fnal examples
in this chapter explore this interrelationship through novel utterances that generate
metaphor/metonymy combinations.

A ‘moo’ point
In the American TV sitcom Friends, the character Joey Tribbiani is known for his
rather one-dimensional way of looking at the world. One notable instance occurs in
series 7, episode 8, which you can see by scanning QR code 8.3.
In a conversation with the character
QR 8.3 A ‘moo point’ metonymy Rachel, in which they are discussing
strategies on how to woo a man she likes,
Joey advises that if this man isn’t
interested in Rachel then all of “this” is a “moo point” (Lin and Bright 2000). Joey
‘explains’ himself by latching on to the metonymic element ‘moo’ (SOUND OF
ENTITY FOR WHOLE ENTITY) as representing a whole cow by further claiming
that the opinion of the man “is like a cow’s opinion” which, as he puts it, “doesn’t
matter”. Therefore, the “moo point” metaphor derives from a metonymy which you
can test using the ‘is like’ test: ‘A moo (Vehicle) is like a cow (Target)’.
But this does not make sense. Instead, we can say that the sound of a cow – the ‘moo’
– is substituting conceptually for the whole of the animal. So, when Joey invokes the
sound of the cow through the vehicle term, ‘moo’, he is intending his hearer, Rachel,
to understand the ‘moo’ as standing for the cow itself. He inadvertently invokes a
mapping chain that moves from metonymy to metaphor in the following ways: a
metonymy (sound of entity for whole entity) is the source domain in a metaphoric
mapping in which the target is the notion of irrelevance (‘it doesn’t matter’). So, the
metaphor (incorporating the metonymy) would be something like irrelevance is a cow.
The conceptual metaphor then, would be states are living entities (see Figure 8.6).
Presumably, Joey’s utterance makes sense if it’s premised on a belief that the opinion
of a cow is insignifcant and has no bearing on human relationships.

194
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Figure 8.6 The ‘moo point’ metaphor/metonymy

You’ll see that the meaning is not directly traceable to the grammatical or lexical
elements but resides in an implicature (so implicitly traceable, in a sense). This is the
case for all metaphors (remember that metaphors are not literally true and so fout
Grice’s Quality maxim).

Conclusion
This chapter has shown how metaphors and metonymies are not simply fgures of
speech that show a text producer’s linguistic skill or ‘genius’ but rather are
systematised ways of seeing the world, the entities that inhabit it, and the goings-on
that occur within it. It also shows how both tropes rely on implicated unstated
meanings in order to make sense so that in some ways they call upon the hearer to
actively participate in seeing the world the way the metaphor or metonymy frames it.
We have also shown how metaphors can develop through extension to form whole
discourses about communities, events, behaviours, and so on. And we have
demonstrated how much we rely on metaphors in everyday language and thinking,
too, and that contrary to Aristotle’s claim, they are not only the product of creative
genius (or we are ALL creative geniuses!) To employ another well-wrought metaphor,
the following chapter moves away from the content of discourse to examine discourse
structure. More specifcally, it considers the how experience is represented in
discourse.

195
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Further reading
Metaphors are often employed to make sense of abstract concepts. One such study is
Duffy and Feist’s work which examines the concept of ‘time’ in a booklength study,
Time, Metaphor, and Language: A Cognitive Science Perspective (2023). In terms of
the rhetorical value of conceptual metaphors, approaches to CMT have found some
conceptual metaphors to be more persuasive than nonmetaphorical utterances
(Brugman et al. 2019; Sopory and Dillard 2002; van Stee 2018). For instance,
metaphorically presenting depression in terms of ‘ups’ and ‘downs’ or ‘bright’ and
‘dark’ periods, in comparison to non-metaphorically as ‘positive’ and ‘negative’
periods, has been shown to increase the perceived effectiveness of depression
medications (Keefer et al. 2014).

Answers to activities
Activity 8.1 Orientational metaphors – positive or
negative evaluation?
Johnson’s premiership, a mindless rotting meat zombie held together with the
Sellotape™® of lies and the sticky excretions of his own spaff-faucet, began a
sudden and dramatic slide into the ocean on Wednesday, like an arctic ice shelf
made entirely of frozen shit.
(Lee 2022)

It’s fair to say that, while not the most interesting part of the utterance, this negative
comment has an orientational metaphor as part of the main verb phrase, signalled by
bold type, ‘a sudden and dramatic slide into the ocean’. Sliding into the ocean (as
opposed to ‘slide up’) suggests a bad is down metaphor. Therefore, we are encouraged
to read this as a negative comment on Johnson’s premiership.
Of course, the co-text makes this abundantly clear. We are told the premiership (or the
man?) is a ‘mindless rotting meat zombie’ (this confates several metaphors, one of
being without sense (‘mindless’) and another connoting decay (‘rotting’) and a
contiguous concept, death (‘zombie’). The ‘sticky excretions’ links cohesively (see
Chapter 3) to ‘spaff’ (a colloquial term for ‘semen’ or ‘ejaculation’) and ‘shit’.

Activity 8.2 Metaphor in literary fiction

‘Potentiality’ is the Subject in a clause in which it is said to ‘knock’ on something.
Therefore, the phenomenon (‘potentiality’) is animated by virtue of being able to
‘knock’. This triggers a conceptual metaphor, abstractions are animate objects. The
grammatical Object, ‘the door of my heart’ is also metaphoric as there is no literal
door on a heart. So, perhaps the speaker is conceiving of the heart as a space that
196
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

houses emotions (such as love) through the metaphor the heart is a container. In
metaphoric terms, the act of knocking acts as a source domain for the target concept
of desiring access (to the speaker’s emotions). Overall, the speaker’s meaning in
uttering ‘potentiality knocks on the door of my heart’ appears to be something like
‘this meeting is an opportunity for love’. But Murakami’s metaphoric version is
arguably much more poetic!

Activity 8.3 Real examples of the ‘hiving’ metaphor

Example (a) is from the Tripadvisor website and is part of a review of a restaurant in
London. The reviewer notes they are from Ballygally in the north of Ireland. Example
(b) is also from Tripadvisor and the Belfast author is reviewing a bar called the
Woodworker Tap House. Example (c) is from a BBC news report into overcrowding
at an event in Fintona, in the north of Ireland, in 2013, in which people attending a
music event were crushed and injured in the premises. Example (d) is from a regional
newspaper, the Belfast Telegraph, and reports on the excitement remembered at a
1990 football match in which local side Portadown won their frst league title. You
will see that the interpretation of ‘hiving’ is largely positive (a, b, d) and example (c)
may be perceived as negative or neutral. Example (e) is negative as it laments the fact
that too many people were at the spa when the author wished to use it.
All the authors are from the north of Ireland and therefore speakers of
HibernoEnglish. A simple online search using the phrase ‘place was hiving’, or ‘room
was hiving’ generates a very limited set of results, all of them Hiberno-English
speakers. This is interesting for sociolinguistic understandings of metaphors as
regionally specifc! You might want to think of any metaphors you use in your own
speech community that could be used to identify that particular speech community.
Activity 8.6 Metaphorical extension and elaboration in BBC
TV’s Dragon’s Den
Metaphor group E is from Pitch 1: A ‘Rev-off’ device to turn a regular push bike into
an e-bike
Metaphor group C is from Pitch 2: An optometrist duo pitch a screen that limits
blue light exposure from digital devices.
Metaphor group D is from Pitch 3: An entrepreneur is selling trees to offset the
carbon footprint via an app
Metaphor group B is from Pitch 4: A busy mum who is a teacher is pitching her
‘family-themed’ fruit preserves
Metaphor group A is from Pitch 5: Comedian, Darren, is pitching his idea for a
‘dog poo’ lifter.

197
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Activity 8.7 Making meanings through metaphors

Abstract noun Concrete noun Mapping (as presented by students from University
College Utrecht, Netherlands)

Terror Pancake terror is a pancake

It [pancake] is heavy and sticky. And just like the
pancakes that form a sort of tower, so do the terror and
the problems when they start accumulating. The syrup
that overfows on top is just like the thoughts that you
can’t get out of, it is sticky and stays on you; it feels
like it cannot be taken off even after you clean your
hands with kitchen paper.

Love Boat love is a boat

Love is a boat, rocks from side to side with strong
waves. It may sink, it may stay afoat, but it’s
constantly moving. It can carry you.
Love Soup love is soup
Love is soup, can warm you up from the inside, keep
you happy and warm when you’re cold, can also burn
you and leave you scarred.
Future Wheel future is a wheel
Always in motion and it goes forward. The wheel is
the frst invention, so it also refects the past.
Language Room language is a room
Language is like a room. You can decorate it as you
want, it takes various shapes and forms, it is a space of
possibility, but you have to work within the space.

Notes
1 We refer to linguistic expressions of metaphors as lexical metaphors.
2 This way of representing the conceptual nature of metaphors using small capitals and an ‘X is
Y’ formulation, as we have here, stems from Lakoff and Johnson (1980). It shows, as Kövecses
puts it, that “the particular wording does not occur in language as such, but it underlies
conceptually all the metaphorical expressions” that derive from the capitalized metaphoric ‘X is
Y’ structure (2010: 4, our emphasis).
3 Grady (1997) calls the conceptual or generic metaphor a ‘primary’ here but there are a range of
terms for such metaphors. Some widely used terms include ‘root metaphors’ (Pepper 1935;
Brown 1977; Goatly 1997), ‘primary systems’ (Black 1955), and ‘principal metaphors’ (Brown
1977). Having so many terms doesn’t really help, though. We have opted to go with ‘primary’
as stated in the chapter.
4 Reconceiving of ‘knowledge’ as something we can see is a culturally entrenched metaphor,
particularly in religious discourse (Canning 2012).
5 The reference to ‘gravestones’ invokes a metonymy in which the place-marker for the dead
(gravestone) stands for the state of being dead (LOCATION FOR STATE). We deal with
metaphor/metonymy combinations later.

198
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

6 Also note that ‘hard’ is used to denote ‘toughness’ as in ‘Oooo, you’re hard’, said to Akala after
he spits some javelins.
7 This was an unsubstantiated claim for which Cameron was rebuked. If you can bear to read the
British tabloid press you will fnd many examples of this metaphor.
8 Arguably hearers are oriented to make the connection, but not to accept it
9 Note how the metaphor test helps you see that it is a metonymy (*a bed is like a person). Asterisks
are used to show that the utterance does not work.
10 We don’t think Cardi B is referring literally to ‘a white horse and a carriage’ and she didn’t
answer our email to confrm.
11 Perhaps the case of Caren Z. Turner offers one example of a real-life ‘Karen’ (if social media is
anything to go by). Turner was a Port Authority Commissioner who arrived at the scene of a
traffc violation where two Port Authority offcers had pulled over her child and some friends.
Turner used her senior role to try to get the offcer who stopped her child and their friends to drop
the violation. He didn’t. Here is the link to the clip of the incident:
https://www.youtube.com/watch?v=Goj-1stJpoA

References
Akala (2007) ‘Shakespeare’ [song]. Comedy, tragedy, history [album]. Freedom
Lasso. Illa State Records.
Amnesty International (2020) ‘Why the language we use to talk about refugees
matters’. Accessed 10 July 2023. https://www.amnesty.org/en/latest/news/
2020/03/why-the-language-we-use-to-talk-about-refugees-matters
Aristotle (2008) Poetics (trans. S. H. Butcher). Accessed 6 January 2022. https://
www.gutenberg.org/fles/1974/1974-h/1974-h.htm#link2H_4_0003
Black, M. (1977) ‘More about metaphor’. Dialectica, 31(3–4): 431–457.
Black, Max (1955) ‘Metaphor’. Proceedings of the Aristotelian Society, 55(1):
273–294.
Black, Max (1979) ‘More about metaphor’. In A. Ortony (ed.), Metaphor and
Thought. Cambridge: Cambridge University Press.
Brown, R. H. (1977) A poetics of sociology: Towards a logic of discovery for the
human sciences. Cambridge: Cambridge University Press.
Brugman, B. C., Burgers, C., and Vis, B. (2019) ‘Metaphorical framing in political
discourse through words vs. concepts: A meta-analysis’. Language and
Cognition, 11(1): 41–65.
Callahan, B. (Writer) and Kerns, J. (Director) (2004) ‘My common enemy’. Scrubs,
series 4, episode 7. ABC Studios.
Cameron, D. (2014) ‘We’re building an immigration system that puts Britain frst’. The
Telegraph, 28 July. Accessed 6 July 2023. https://www.telegraph.co.uk/
news/uknews/immigration/10995875/David-Cameron-Were-building-
animmigration-system-that-puts-Britain-frst.html
Canning, P. (2012) Style in the Renaissance: Language and ideology in early modern
England. London and New York: Continuum.

199
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Charteris-Black, J. (2006) ‘Britain as a container: Immigration metaphors in the 2005

election campaign’. Discourse & Society, 17(5): 563–581.
Dragon’s Den (2021) Season 18, episode 13. BBC Studios Factual Entertainment
Productions.
Duffy, S. and Feist, M. I. (forthcoming, 2023) Time, metaphor, and language: A
cognitive science perspective. Cambridge: Cambridge University Press.
Elgot, J. and Taylor, M. (2015) ‘Calais crisis: Cameron condemned for
“dehumanising” description of migrants’. Guardian, 30 July. Accessed 6 July
2023. https://www.theguardian.com/uk-news/2015/jul/30/david- cameron-
migrant-swarm-language-condemned
Forceville, C. (2002) ‘Further thoughts on delimiting pictorial metaphor’. Theoria et
Historia Scientiarum, 6(1): 213–227.
Giora, R., Fein, O., Kronrod, A., Elnatan, I., Shuval, N., and Zur, A. (2004) ‘Weapons
of mass distraction: Optimal innovation and pleasure ratings’. Metaphor and
Symbol, 19(2): 115–141.
Goatly, A. (1997) The language of metaphors. New York: Routledge.
Goatly, A. (2007) Washing the brain: Metaphor and hidden ideology. Amsterdam:
John Benjamins.
Grady, J. E. (1997) Foundations of meaning: Primary metaphors and primary scenes.
Berkeley: University of California.
Hart, C. (2021) ‘Animals vs. armies: Resistance to extreme metaphors in
antiimmigration discourse’. Journal of Language and Politics, 20(2): 226–253.
Keefer, L. A., Landau, M. J., Sullivan, D., and Rothschild, Z. K. (2014) ‘Embodied
metaphor and abstract problem solving: Testing a metaphoric ft hypothesis in the
health domain’. Journal of Experimental Social Psychology, 55: 12–20.
Kövecses, Z. (1986) Metaphors anger, pride, and love: A lexical approach to the
structure of concepts. Philadelphia, PA: John Benjamins.
Kövecses, Z. (1990) Emotion concepts. New York: Springer-Verlag.
Kövecses, Z. (2010) Metaphor: A practical introduction (2nd edition). Oxford:
Oxford University Press.
Lakoff, G. (1987) Women, fre, and dangerous things: What categories reveal about
the mind. Chicago: University of Chicago Press.
Lakoff, G. and Johnson, M. (1980) Metaphors we live by. Chicago: University of
Chicago Press.
Langacker, R. W. (2008) Cognitive grammar: A basic introduction. Oxford: Oxford
University Press.
Lee, Stewart (2022) ‘I’m not answering my phone, in case it’s No 10 offering me a
job’. Guardian, 10 July. Accessed 28 July 2022. https://www.theguar
dian.com/commentisfree/2022/jul/10/boris-johnson-out-corruption-torymps-
resignation.
Lin, P. (Writer) and, Bright, K. (Director) (2000) ‘The One where Chandler doesn’t
like dogs’. Friends, Series 7, episode 8. Bright/Kauffman/Crane Productions, in
association with Warner Bros. Television. Accessed 30 January 2023:
http://www.friends-tv.org/epguide.html#seventh
200
 METAPHORICAL MEANINGS IN DISCOURSE: METAPHOR AND METONYMY

Littlemore, J., Krennmayr, T., Turner, J., and Turner, S. (2014) ‘An investigation into
metaphor use at different levels of second language writing’. Applied Linguistics,
35(2): 117–144.
Maroon 5 featuring Cardi B (2018) ‘Girls like you’ [song]. Red Pill Blues [album].
Cirkut and Jason Evigan.
Morissette, A. and Ballard, G. (1998) ‘Thank U’ [song]. Supposed Former Infatuation
Junkie [album]. Maverick Records.
Mujagic´, M. and Berberovic´, S. (2019) ‘The immigrants are animals metaphor as a
deliberate metaphor in British and Bosnian-Herzegovinian media’. Explorations
in English Language and Linguistics, 7(1): 22–51.
Murakami, H. (1994) ‘On seeing the 100% perfect girl one beautiful April morning’.
In Murakami, The Elephant Vanishes. London: Vintage.
NJ.com (2018) ‘Complete 1-hour video: Port Authority commissioner confronts
police during N.J. traffc stop’. 26 April. Accessed 10 July 2023. https://
www.youtube.com/watch?v=Goj-1stJpoA
Pepper, S. C. (1935) ‘The root of metaphor theory of metaphysics’. Journal of
Philosophy, 32(14): 365–374.
Quintilian (1921) Institutio Oratoria. Vol. III, Book 8, Ch. 6. Loeb Classical Library.
New York: G. P. Putnam’s Sons.
Richards, I.A. (1936). The Philosophy of Rhetoric. New York: Oxford University
Press.
Simpson, P. (2014) Stylistics. A resource book for students. New York: Routledge.
Sopory, P. and Dillard, J. P. (2002) Figurative language and persuasion. In J. Price
Dillard and M. Pfau (eds), The persuasion handbook: Developments in theory and
practice, pp. 407–426. London: Sage Publications.
Steen, G. J. and Gibbs Jr, R. W. (eds) (1999) Metaphor in cognitive linguistics:
Selected papers from the 5th international cognitive linguistics conference,
Amsterdam, 1997 (Vol. 175). Amsterdam: John Benjamins.
Twenty One Pilots (2018) ‘Neon Gravestones’ [song]. Trench [album]. Fueled by
Ramen.
van Stee, S. K. (2018) ‘Meta-analysis of the persuasive effects of metaphorical vs.
literal messages’. Communication Studies, 69(5), 545–566.

201
 Representing experience in
9 discourse

Introduction
In previous chapters, we looked at the pragmatics of discourse in different contexts.
In this chapter, we examine the discursive goings-on in one particular institutional
setting, that of forensic police reporting. Over the course of the chapter, we will
present a model of transitivity as a methodological tool for the analysis of clause
structures in police reports and show how the same actions and events can be rendered
or construed in different ways in discourse. This is because clauses can contain
constituents that refer to agents who ‘do’ actions, as well as affected entities who are
acted upon. As we saw in Chapter 2, clauses typically contain a grammatical Subject
and a Predicator (verb) (but see below), so ‘I ran’ is a clause, whereas ‘I ran because
I was afraid’ is a clause complex containing two clauses: a main clause (‘I ran’) and a
subordinate clause (‘because I was afraid’). So, a simple clause like ‘I threw a stone
at the window’ contains an agent, ‘I’, an entity acted upon ‘a stone’ an action, ‘threw’,
and a circumstance that elaborates upon the aforementioned constituents ‘at the
window’. To bring these grammatical functions and their clausal arrangements to life,
we offer a little crime pastiche in Activity 9.1 and ask you to imagine that you
witnessed it. Then we ask you to think about all the different ways you might produce
a narrative of the crime (as a witness) and to consider some of the semantic and
pragmatic meanings of those different ways of saying.

Activity 9.1 Rendering experience in language

The pastiche

You are out for a walk and you witness a person run up to an elderly lady, push her over, grab her
handbag and run off in the opposite direction. You call the police for help knowing that the bag-
grabbing lady-pusher has committed a crime and that the elderly lady may need your help as a
witness if she chooses to pursue justice. The police call-taker is the frst person to whom you tell
your story of the crime you just witnessed.

DOI: 10.4324/9781003351207-9
Telling and retelling
Give an account of this event to a friend and record it on your phone. Have them re-tell it back
to you and record that, too. How did you each tell it? What was left out in the telling? How did
you phrase the goings-on?
202
 REPRESENTING EXPERIENCE IN DISCOURSE

Now, imagine that you will tell your story of the crime again when you give a formal witness
statement to the police. And again, when you tell it to your friends and family. This amounts to
a lot of re-tellings. How might you phrase each one? What words might you choose? Why? Are
you more specifc with certain details to the police than with your friends? Do you focus on
different elements of the story depending on whom you tell it to? Why? What are they? What
elements are consistent across all the tellings?

Different ways of telling

To address some of the questions from Activity 9.1 we offer a sample ‘telling’.
Let’s say the frst telling of your story is to the police emergency number and it goes
like this (turns are numbered for reference).

1 Call-taker: Police, what’s the emergency?

2 Witness: Hi, I’ve just seen a man push over an old lady and steal her bag.
3 Call-taker: Okay, is the lady hurt?
4 Witness: I don’t think she’s badly hurt but I’m not sure.
5 Call-taker: Okay, what’s your location?
6 Witness: I’m at the junction of Anytown Road and Sometown Way, beside the
town hall at the trafc lights
7 Call-taker: Right, I’m sending someone now, they should be with you in around
6 minutes. Is the attacker still on the scene?
8 Witness: No he ran of down Anytown Road with the bag.
9 Call-taker: Okay, thank you, an ofcer should be with you very soon.

You have just given the frst (admittedly brief) ‘telling’ of the at-issue event (a more
neutral term for ‘crime’ at the investigative stage) to the police call-taker. You will
give another to police offcers in person when they arrive on the scene and, following
that, a third telling in your formal witness statement. This third retelling of your story
will become evidence for any future prosecution case. It is important to capture this
retelling thoroughly, carefully, and in a timely manner (to limit misremembering or
gaps in recall) so let’s pretend you do it the following day at the police station.
During the interaction between you and the police offcer who writes up your
statement, you are probed about the colour of the victim’s handbag.
After a bit of discussion about it, you conclude, tentatively, that you ‘think it might
have been red’, putting stress on the underlined word to emphasise the lack of
certainty. The offcer distils this information into the following on your statement:
‘I believe the handbag was red’.
This could be interpreted as having a greater commitment to the certainty of the
assertion (known as epistemic modality) than your spoken version (the latter contained
203
 REPRESENTING EXPERIENCE IN DISCOURSE

the modal verbs ‘think’ and ‘might’ which have been removed in the offcer’s version).
When the statement is fnished, you read over it and sign it to verify its content. Three
months later you are asked to be a witness in court for the case as the perpetrator has
been caught and charged with theft and assault. In an adversarial court system, your
statement is challenged by the perpetrator’s legal team who cross-examine you about
the bag’s colour. You repeat your uncertain stance in the courtroom saying ‘I’m not
sure what colour it was’. The defence points out this anomaly between ‘I am not sure’
vs ‘I believe’. Yet to you, these phrases amount to the same thing – you don’t really
remember the colour for certain. There is now doubt over the veracity of your version
of events which undermines your evidence. The prosecution case falls apart and our
baggrabbing lady-pushing perpetrator walks free.
What we can learn from this invented case is that you as a witness are, essentially,
reduced to your word on that statement. One of the key features of witness statements
(and indeed, police reports), is that they are narratives about who or what did what to
whom or what. Knowing this information is necessary because to produce a coherent
story of a crime (the at-issue event), who is doing what and to whom, needs to be clear
and unambiguous. While SPOCA identifes clausal components, transitivity tells us
more about what they are doing and their relationship to each other. Let’s take our
bag-grabbing perpetrator again and show the different constituents and their roles
within the story in the second turn of the made-up phone call with the police call-taker.
Hi, I’ve just seen a man push over an old lady and steal her bag
I [participant] ’ve just seen
[action] a man [doer] push
[action] over [circumstance] an
old lady [affected participant] and
steal [action] her bag [affected
entity]
Laying it out like this shows who or what did what to whom or what (the roles
performed by each grammatical unit are in square brackets). For instance, the frst
action reported is one of ‘seeing’ followed by ‘push’ and ‘steal’. It’s clear who the
agent of the action of ‘push’ is (‘a man’), and so on. We return to these constituents
later as they will defne the functional roles outlined in the clauses in our story. These
roles will be the focus of the transitivity analysis later.

Activity 9.2 SPOCA spotting

Recall that in Chapter 2 we introduced a SPOCA (Subject, Predicator, Object, Complement, and
Adjunct) analysis of the clause. This analysis identifes the clausal function of words and phrases
in clauses. Now we want to consider a complementary analysis that identifes what clausal
constituents are doing in the reality that the clause renders. This reality can be presented in any
number of different ways – for instance, the way the reality is identifed (e.g. using nouns and
adjectives), the actions performed in that reality (e.g. verbs) and positions in time and space
204
 REPRESENTING EXPERIENCE IN DISCOURSE

(e.g. prepositions, adverbs). Identify the SPOCA elements of the retelling above in the following
sentence.

Hi, I’ve just seen a man push over an old lady and steal her bag

We give our answer at the end of the chapter

Another way of telling

We now want to introduce another way of telling the same story that obscures some
of these story constituents/roles (we have numbered the sentences for reference). What
kinds of questions might this new version generate? What roles are missing? How
does this compare with your version in Activity 9.1?

(1) On my way home there was a robbery. (2) An elderly woman’s bag was stolen
and she was pushed over. (3) I called the cops.

Nominalisation
You may have noticed that our bag-grabbing lady-pusher is missing from this version.
His role has been backgrounded linguistically through a process called
nominalisation that is, turning an action into a thing, i.e. ‘robbed’ (a verb) becomes a
noun, ‘robbery’. Consequently, the text producer presents the doing as something that
‘is’ – it simply exists. A good place to fnd nominalisations is in journalistic discourse,
where the convention is to anchor news stories on headlines that encapsulate an action
or series of actions into a single noun phrase. For example, read the following BBC
news headline (28 December 2022).

Murder arrests over Birmingham nightclub stabbing

(BBC News 2022)
The act of murdering appears here as a modifer ‘murder’ for ‘arrests’ which is a noun
derived from a verb ‘to arrest’. The stabbing of a young man is presented here as a
noun (a gerund – see Chapter 2), ‘stabbing’. Such grammatical conversions may be
due to discourse constraints on headline text – often, the body copy that follows makes
the agency more explicit (e.g. “Two men have been arrested on suspicion of murdering
a man who was stabbed to death on a nightclub dancefoor in Birmingham”, BBC News
2022).
As well as the generic constraints and conventions of news stories, police offcers
often use nominalisations as a way of distancing themselves from reported or
unverifed actions. This could be a conscious attempt to present a neutral institutional
stance. For example, compare the use of the verb form ‘allege’ in ‘Victim alleges that
suspect did X’ against its nominalised counterpart ‘There was an allegation of X’. In
the former, the victim is the source of ‘allege’ which may imply that they are not
believed or there is a lack of evidence from the victim, whereas no agent can be derived
205
 REPRESENTING EXPERIENCE IN DISCOURSE

from the nominalised form ‘allegation’ and so no overt stance is taken towards the
victim. Police also use nominalisations in the categorization of crimes. For example,
an act of violence committed by someone towards another person is categorised as an
‘assault’ (Calligan 2000) and by doing so, the implicit grammatical transitive structure
of ‘X assaulted Y’ is packaged neatly into a noun ‘assault’ which obscures not only
the perpetrator of the ‘assault’ but also the victim as Figure 9.1 shows.
Depending on the context, nominalising actions can have the effect of ‘glossing
over’ or eliding who did what to whom for ideological reasons as we later show.

Figure 9.1 Nominalisation vs clause

Passives
Let’s return to the last retelling of our robbery story. Here it is again for reference:
(1) On my way home there was a robbery. (2) An elderly woman’s bag was stolen
and she was pushed over. (3) I called the cops.
In the frst sentence the ‘action’ (stealing the bag) has been rendered as a noun,
‘robbery’, doing away with both agent and affected entity. In the second sentence,
which contains two coordinated clauses, the victim’s bag, and then the victim (‘she’)
are made the Theme (see Chapter 2) so the reader’s focus is steered towards the
affected person upon whom the action was performed and not the doer who performed
the action. This can be seen in Figure 9.2.

Figure 9.2 Passive constructions

Although the actions themselves are documented (‘was stolen’, ‘was pushed over’)
we are left asking ‘by whom’? This version lacks agency and these omissions must be
inferred from the passive constructions – in other words, we know that someone did
the robbing and the pushing. Language offers many ways of rendering real world (and
fctional) experiences.

206
 REPRESENTING EXPERIENCE IN DISCOURSE

There are different types of actions (e.g. physical, mental, being) and the way we
present these actions can change how we conceptualise the roles therein, or even the
whole story. The transitivity framework offers a very serviceable model for
delineating these relationships between roles, specifcally agency, actions, and
circumstances, and makes patterns across these elements explicit. In high-stakes
discourse contexts like policing and prosecuting crimes, such elements comprise
evidence and so are relied upon in making sense of – and adjudicating on – who did
what to whom.

The transitivity model

Transitivity as a framework for text analysis emerged in the 1970s with Margaret
Berry’s model (1975) and later developed by Halliday (1994) and Halliday and
Matthiessen (2014), among others (see Simpson 1993, 2014, for example). The model
is based on the grammatical distinction between transitive and intransitive functions
of verbs. That means, rather basically, some verbs can take a grammatical Object
(transitive), and some verbs cannot (intransitive) or they can be used transitively and
intransitively. Examples of verbs that can be used transitively are ‘ate’, ‘took’, ‘hit’,
‘climbed’, and ‘forgotten’. All these are transitive when they accept a grammatical
object (O), as the examples in Figure 9.3 show.
Verbs that are unable to take a direct Object or that do not require one are
‘intransitive’ and so the actions they encode are not executable on another entity.
Examples of verbs used intransitively are ‘fell’ (as in Figure 9.3) ‘laughed’ (‘they
laughed a lot’), and ‘came’ (‘She came back early’).

Figure 9.3 Transitive and intransitive verbs

Representing ‘inner’ and ‘outer’ experience

The model of transitivity focuses on the clause as a primary unit of analysis (Berry
1975; Halliday and Matthiessen 2014) because it is at this level that goings-on are
most basically represented. In clauses, the predicator encodes whatever type of action
is taking place, has taken place, or will take place. Therefore, the verb is the central
207
 REPRESENTING EXPERIENCE IN DISCOURSE

concept in transitivity and is known in the framework as the process. There are many
kinds of processes that go on in the world. Some processes are considered outer
experiences because they take place in the world around us, and others are inner
experiences as they occur in the realm of consciousness or in the person. The other
clausal constituents include participants who either do or are affected by the action
of the verb, and circumstances that convey the scope of the goings-on (usually
through adverbial or adjective phrases). Processes and their relationships to the other
elements of the clause are outlined in the following sections.

Material processes
Material processes are defned by Halliday and Matthiessen as process by which
“a quantum of change in the fow of events” results from some “input of energy” (2014:
224). Not very helpful, but in effect, their point is that Material processes capture
physical actions that cause change. Mary Ellen Ryder developed this idea further by
introducing “parameters” that defne “prototypical events” so that the action of
Material processes is not limited to causing “change” but also covers “maintaining” a
particular state (Ryder 2008). In short, Material processes encode actions or maintain
actions that take place in the outer world.
With Material processes, there are two participant roles: the ‘doer’ of the process
and the affected entity. The participant who carries out the process (the ‘doer’) is
referred to as ‘Actor’, and the affected entity is termed the ‘Goal’. In the examples
shown in Table 9.1, Actor and Goal roles are flled by ‘the perpetrator’ and ‘the elderly
woman’ respectively.
Table 9.1 The different Material Processes
Material Process Type

i Material- The perpetrator robbed the elderly woman’s bag

Intention ii [Actor] [Goal] the elderly woman [of
Material-Event The crime robbed her bag]
[Actor] [Goal]
iii MaterialSupervention The elderly woman fell
[Actor]

Material processes can be enacted by human participants (‘the perpetrator’ in

example (i) in Table 9.1). This subtype is known as ‘Material-Intention’ processes.
But there are also options for having non-human entities do the ‘acting’. An example
is ‘the wind knocked over the bins’, where ‘the wind’ is responsible for the doing, yet
is inanimate. In (ii), the Actor is an inanimate participant (specifcally it’s a noun that
has been derived from a verb) which has the effect of distancing the person who
commits the crime from the act they commit. In such cases of inanimate Actors, the
Material process is known more specifcally as a ‘Material-Event’ process. In (iii), the
‘elderly woman’ is still the affected entity in the little pastiche, but now the process is
an involuntary one. This type of Material process is known as ‘Material-
208
 REPRESENTING EXPERIENCE IN DISCOURSE

Supervention’. In other words, the Actor (the elderly woman) is presented as having
acted unintentionally, ‘fell’, without the infuence of an external agent. Notice how she
is not grammatically positioned as the Goal element in this confguration.

Circumstances
Another element in the transitivity model relates to information around the process
and participants that elaborates on the how, where, and when. This information is
captured in the part of the clause known as the Adjunct and is thus a deletable
component. In transitivity, this constituent is known as the Circumstance. In Table
9.1 above the Circumstance in (ii) is in square brackets. In Table 9.2 below we have
adapted our examples from Table 9.1 to account for grammatical mood and include
optional elements such as the ‘Circumstance’ (in grey) that provide orientational
information or information about the ‘how’ of the process.
As you can see, there are different linguistic choices being made in all our
examples: the lexis and the grammatical structure of the clauses can infuence
interpretation, particularly in terms of what is considered salient information (see
Chapter 2 on Theme and Rheme). They also show the versatility of language and how
the relevant constituents can combine to make roles more or less explicit. In Table 9.2,
example (i), agency is explicit as is the case with
Table 9.2 Confgurations of Material Processes and their clause constituent elements
Material Process Circumstances

i - in active voice The perpetrator robbed the elderly after much

[Actor] woman’s bag planning
[Goal]
ii - in passive voice The elderly was robbed after much

woman’s bag [Goal] planning

iii - using an The elderly disappeared in broad

intransitive verb woman’s bag [Actor] daylight

all Actor+Process+Goal constructions, which are conveyed in the active voice. In (ii)
agency is obscured as the Actor has been excised through the use of the passive voice
but can be added if the text producer desires (note that all optional elements are in
grey type). In (iii) an intransitive verb form is used (note that, grammatically, you
cannot typically ‘disappear’ something, so the verb cannot conventionally take an
Object). The effect of this choice means no external agent is suspected.
Clause elements that are not compulsory, such as the Circumstances, are in grey
type. This basically means that text producers can choose to give this information
rather than being compelled to do so by grammatical necessity. We can then examine
the potential ideological effects of providing the Circumstances in such cases. In Table
9.2 (i) and (ii), the Circumstances offer additional information about intentionality (i.e.

209
 REPRESENTING EXPERIENCE IN DISCOURSE

that the process was done ‘after much planning’), and so we may look on the act with
more contempt than we might do if intentionality was not apparent. In other words,
including this information may well steer our appraisal of the action as more negative
or unacceptable. This may have a bearing on a jury’s decision!
A reasonable question to ask at this point is why we need these processes and
participants and new terms for concepts that appear to have grammatical labels already
– why not just use good old Subject, Predicator, Object, Complement, and Adjunct
(SPOCA)? In answer to that question, consider the differences between the
grammatical roles of SPO and the transitivity roles of ACTOR-PROCESSGOAL in
the following identical clauses in Figure 9.4.
In these different permutations, the words that fll the grammatical and transitivity
slots are defned by where they fall syntactically. However, the syntactic form of the
SPO labels tells us nothing about the functional relationship between what flls that
slot and the other constituents of the clause (i.e. the Predicator and Object). The
transitivity confgurations, on the other hand, do tell us about these relationships. You
can see this in Figure 9.4 where the ‘Goal’ in the transitivity roles remains consistent
across both active and

Figure 9.4 Grammatical roles vs Transitivity roles

passive clauses, but yet the role of Subject changes according to whether the clause is
active or passive. This means that when we come to analyse a text, we can see more
clearly the ideological force or impact of the ‘who or what does what to whom’ than
if we only relied upon an SPO analysis. We take this up later in the chapter.

Mental processes
Mental processes represent ‘inner’ actions. Halliday acknowledges that different
participant roles are required for this type of process. This is because it is not
appropriate to assign the (Material) roles of ‘Actor’ and ‘Goal’ to entities experiencing
a Mental process and the concepts or things mentally experienced as there is no other
entity being ‘acted’ upon. Therefore, the participant roles that better refect inner
experience are known as ‘Senser’ (the experiencing entity) and the ‘Phenomenon’ (the
thing/concept/feeling experienced). As there are different kinds of Mental
experiences, they are subcategorised into three process types, cognition, reaction, and
perception.

210
 REPRESENTING EXPERIENCE IN DISCOURSE

( a) Mental processes of cognition (e.g. ‘think’, ‘know’, ‘consider’, ‘believe’)

( b) Mental processes of reaction (e.g. ‘hate’, ‘like’, ‘fear’, ‘enjoy’)
( c) Mental processes of perception (e.g. ‘see’, ‘hear’, ‘taste’)
To return to our crime pastiche example, if we were to narrate the story of the robbery,
we could conceivably use any of the Mental process types outlined above. We have
presented three examples in Table 9.3 below.

Table 9.3 Different formulations of Mental Processes and their constituent elements
Mental Process Circumstances

i Mental-cognition The perpetrator devised a robbery after much

[Senser] [Phenomenon] consideration

ii Mental-reaction The elderly woman felt threatened during her

[Senser] [Phenomenon] evening walk

iii Mental-perception The elderly woman heard his approach from the left
[Senser] [Phenomenon]

Clauses (i), (ii), and (iii) are presented here in the active voice, so it is clear who is
the Senser and what is being ‘sensed’.1 Unlike Material processes, many Mental
processes are unverifable in that they are experienced in the consciousness of the
Senser and so cannot be evidenced easily. In fact, as they are subjective, their role in
evidence statements is likely to be infrequent and marked.

Representing states of ‘being’ and ‘having’: Existential and

Relational Processes
While it is likely that material actions of doing and mental actions of sensing are
prototypical ‘goings-on’, there is another dimension to the experiential realm that we
now need to consider. It is the dimension of ‘being’ and ‘having’. Some things are
said to simply ‘be’; ‘Patricia is Irish’ is a case in point. Presented as a state of being,
the fact of this proposition is taken for granted. We will consider Existential processes
frst.

Existential processes
Existential processes are those that present happenings as simply existing and often
“serve to introduce central participants in the Placement (Setting, Orientation) stage
at the beginning of a story” (Halliday and Matthiessen 2014: 308). This is often
achieved grammatically by using ‘there’ as a Subject as in Figure 9.5.
Existential ‘there’ is often referred to as a ‘dummy Subject’ which “cannot be
queried” (Halliday and Matthiessen 2014: 308). It has the effect of introducing

211
 REPRESENTING EXPERIENCE IN DISCOURSE

something into a discourse that is accepted unproblematically as existing. We can

relate this to our fake crime case. In the clause in Figure 9.5, a ‘robbery’ is asserted as
an existential fact. Nothing is being ‘done’ explicitly to anyone or anything else. To
use the terminology of transitivity, there is only one ‘participant’, the ‘Existent’, which
in this case is ‘a robbery’. Existential processes use a form of the verb ‘be’, as do many
Relational processes, which we now turn to.

Figure 9.5 Existential process

Relational processes
Relational processes posit a relation between two things. Something is said to ‘be’ or
‘have’ something else. The participant roles are ‘Carrier’ and ‘Attribute’; the ‘Carrier’
is the entity or thing that carries the ‘Attribute’. So, in ‘the robbery was awful’, the
Carrier is the ‘robbery’, and the Attribute is ‘awful’. The difference between
Relational processes of being/having and Existential processes of existing is that in an
Existential process no relationship is established between the Existent (the ‘robbery’)
and anything else, but in Relational processes there is a relationship established (e.g.
‘the robbery was awful’).
There are three types of Relational processes (summarised in Table 9.4):
a) Relational-Intensive processes account for ‘being’ and so are identifable through
the ‘be’ verb (as main verb) and its variants (e.g. ‘been’, ‘was’, ‘is’, and so on).2
An example is ‘Patricia is Irish’ which establishes the relation between Patricia
and her Irishness.
b) Relational-Possessive processes use a form of ‘have’ as main verb (e.g. has, had,
having, and so on) and express relationships of possession (e.g. ‘We had a short
holiday’).
c) Relational-Circumstantial processes upgrade Circumstances making them an
essential element so that they feature as a ‘participant’. You will be able to identify
them easily enough because they appear ‘Circumstance-y’ (in that they are usually
Prepositional/Adverbial constructions) but they cannot be deleted from the clause
without the clause losing its meaning. An example of a Relational-Circumstantial
process is ‘The rabbits are in the hutch’. Try losing the PP phrase ‘in the hutch’
and just telling your friend ‘The rabbits are’. Nope, we don’t make sense of that
either – unless of course you’re having a René Descartes moment. And we’ve all
had one of those.

Table 9.4 Relational Processes

212
 REPRESENTING EXPERIENCE IN DISCOURSE

Relational Process

i Relational-Intensive Patricia is Irish

[Carrier] [Attribute]

ii Relational-Possessive We had a short holiday

[Carrier] [Attribute]

iii Relational-Circumstantial The rabbits are in the hutch

[Carrier] [Attribute]

We have said above that the verbs ‘be’ and ‘have’ operate in Relational processes
as main verbs because we want to note the potential confusion between these and other
processes that use these verbs as auxiliaries. If we return briefy to Material processes,
you’ll see that they sometimes present the clausal constituents as passives or as past
tense forms, both of which make use of the ‘be’/ ‘have’ verbs as auxiliaries. Here are
three examples of Material processes that use the ‘be’ and ‘have’ verb as auxiliaries
(the corresponding transitivity components are in parentheses).
‘We | have walked | as far as we could’ (Actor | [auxiliary] Process | Circumstance
‘Selma | was walking’ (Actor | [auxiliary] Process)
‘Buster | was walked | by his owner’ (Goal | [auxiliary] Process | Actor)
Therefore, a good test of a Relational process is whether the ‘be’/‘have’ verbs are
the main verbs: in ‘Selma was walking’ the ‘was’ is marking tense rather than ‘being’.
If the ‘be’ verb is followed by a main verb, chances are it’s not Relational or
Existential.

Verbal Processes
These are processes of ‘saying’ or more accurately, of producing sound. Many entities,
both animate and inanimate, can produce sound. In these processes, there is a
participant, namely the Sayer (the entity or thing that produces the sound) and the
Verbiage (the sounds that are produced). Some verbs that fall into this category:
‘shout’, ‘said’, ‘moaned’, ‘groaned’, ‘whispered’, ‘creaked’, ‘yelped’, ‘claimed’, and
so on. The constituents of Verbal processes can be seen in Figure 9.6.

Figure 9.6 Verbal process constituents

Table 9.5 summarises the different process types and their constituents.3

213
 REPRESENTING EXPERIENCE IN DISCOURSE

The discourse situation

Why do we choose to use language like we do? For example, why would we say ‘there
was a robbery’ (an Existential process) as opposed to ‘The perpetrator robbed the
elderly woman? (a Material process)? Linguistic choices are frst and foremost
motivated by the discourse situation itself (see for example, Biber and Conrad 2019).
For example, you might choose informal, colloquial constructions to tell a friend (e.g.
‘He nearly knocked the wind from her!’) but more formal language to relate it to a
police offcer (e.g. ‘She was pushed roughly to the ground’). The discourse situation
might also determine what you focus on (or omit). For instance, in a judicial setting
in which you give your statement to the police, you might foreground different aspects
of the story, such as the clothing and build of the suspect, the specifc timings, and so
on. You may focus less on how you felt (Mental processes) and more objectively

214
 :
:
:

:
:
:
Relational CARRIER ATTRIBUTE Relational-Intensive: e.g. being Aaron is an assaulter
Relational-Possessive: e.g. having Aaron has an assault charge

Relational-Circumstantial: e.g. in trouble, over Aaron was in trouble

my head
Verbal SAYER VERBIAGE Aaron stated he had assaulted Brenda

Existential EXISTENT There was an assault

 REPRESENTING EXPERIENCE IN DISCOURSE

on what went on in the ‘outer world’. In telling and retelling the same story, a level of
variation is expected (Coulthard 2004; Canning 2022, 2023), but it would be
reasonable to expect consistency with the key elements – who the participants are and
what role(s) they play. But this is not always the case. There are times when obscuring
the agency and/or foregrounding (and backgrounding) different clausal components
can have particular effects. Applying the framework of transitivity can help us
understand how the ways in which text producers choose to present participants and
their actions (or not) can prime the reader’s judgement. This can have devastating
consequences in a high-stakes forensic environment. The next section outlines a case
study involving the crime of domestic violence and analyses the police report of the
case as it is prepared for prosecutors.

Case study: a case of domestic violence

The following sections relate to information reported in a real police case fle. The
crime under discussion is classifed as ‘domestic violence’ (sometimes used
interchangeably with ‘domestic abuse’). Domestic Violence (DV) is defned by the
United Nations as “a pattern of behaviour in any relationship that is used to gain or
maintain power and control over an intimate partner” (United Nations
n.d.). Any abuse inficted on a person through DV includes “physical, sexual,
emotional, economic or psychological actions or threats of actions that infuence
another person” (United Nations n.d.). In the UK, there is a disproportionately high
rate of non-conviction in DV cases with relatively few prosecutions; between March
2019 and March 2020 in England and Wales (excluding Greater Manchester), there
were 1,288,018 police-recorded incidents of domestic violence (DV), yet only 61,169
prosecutions (ONS 2020).
In the UK, when police respond to a callout regarding DV, or any report of a crime,
there are a number of options at their disposal in dealing with it. They can take
administrative action (essentially, do nothing further other than log it), or they can
take formal action. In the latter, they can formally interview the suspect if
circumstances necessitate it, they can take a witness statement if the witness feels able
to give one, and after these documents have been secured they will then fle what is
called in England and Wales a ‘Manual of Guidance’ report (MG3) that goes to the
Crown Prosecution Service (CPS). The CPS decides whether there is enough evidence
to secure a conviction and therefore send the case to court, or whether to ‘dispose’ of
the case in other ways, for example, with a ‘simple caution’. Figure 9.7 below (Lynn
and Canning 2021) shows the process step by step and accounts for the differing
discourse settings within this process.
The culmination of the investigative work of the offcer in the case (OIC) is the
Manual of Guidance report (MG3) in which the OIC reiterates the ‘gist’ of the crime
using whatever texts have been generated thus far from the witness

216
 REPRESENTING EXPERIENCE IN DISCOURSE

Figure 9.7 The trajectory of a DV case (England and Wales) (Lynn and Canning 2021)

statements, the offcer’s own notes, and/or the suspect’s interview. The MG3 is
therefore, by defnition, a reformulation of the various (and sometimes clashing)
stories that document the crime. In creating this ‘gist’ the offcer can employ many
linguistic strategies when making a case for prosecution. For example, they might
choose to foreground the victim’s injuries and/or the suspect’s previous criminal
history (if applicable), or to amplify the actions of the suspect by including attitudinal
elements in the telling. Conversely, the offcer can employ different strategies to not
make a case for prosecution, but rather, to dispose of the case by recommending,
implicitly or explicitly, a ‘simple caution’, which is effectively a formal warning.
Some ways of doing this could be to obfuscate the action of the crime, to background
the victim, background the suspect, or foreground the remorseful state of the suspect.
In other words, the offcer has the power to steer the case in a particular judicial
direction. It is likely, then, that the discursive construction of the police documents
can infuence the rate of escalation of these cases and ultimately the rate of prosecution.
Given the current statistics on the rate of DV convictions in the UK for 2020 was
around 6.4 per cent (ONS 2020), examination of this particular type of discourse is
crucial and may shed some light on why the conviction rate is so low.
For our analysis, we were fortunate to have rare access to this chain of forensic
reporting for a DV case including the OIC’s account of the victim’s testimony, the

217
 REPRESENTING EXPERIENCE IN DISCOURSE

full text of the OIC’s reformulation of the at-issue event on the MG3 form together
with the Gatekeeper’s commentary. The case was part of a monthly case load collected
from Devon and Cornwall police force in 2010 (Lea and Lynn 2012) as part of an
internal investigation into the disparity between regional outcomes of DV cases. We
present the case below and ask you to think whether the suspect was convicted.

Activity 9.3 A case of domestic violence (male perpetrator, female victim)

Questions to consider

Do you think the suspect got convicted?

If not, why not? Try to give reasons for your answer.

The case

The case concerns an incident of domestic violence. A male suspect threw a computer at a female,
his wife, hitting her on the leg with it. The female victim contacted the police to report the assault
and the police attended at their shared home following the callout. When interviewed, the
suspect did not deny throwing the computer at his wife.

To convict or not to convict?

Let’s think about what we would need to know to work out if the case in Activity 9.3
is likely to result in a conviction. First, we need to understand what the crime actually
is. As the charge is ‘common assault’, we can start with the defnition of that in the
Criminal Justice Act:
Any act, which intentionally or recklessly, causes another person to apprehend
immediate and unlawful personal violence.
(CJA 1988: Section 39)
In prosecuting a perpetrator for common assault there are a few criteria that must be
met in order to meet the evidence threshold. These are known in policing as “points
to prove” (Calligan 2000), and for common assault, these are that the perpetrator
“unlawfully” and “intentionally” “assaulted another person” (Police National Legal
Database 2019: 49 2.1.1). In other words, a clear understanding that the perpetrator
did act upon someone in a particular way is needed. Therefore, thinking about our
case, the fact that the male suspect threw a computer at the victim while behaving
aggressively would suggest that there was enough evidence to secure a conviction or
at least escalate the case to court. Let’s now see how the transitivity model introduced
above sheds light on who does what to whom in a police report, and whether (and
how) the report captures – and accurately refects – the points to prove.

218
 REPRESENTING EXPERIENCE IN DISCOURSE

The text below comes from the case that we paraphrased in Activity 9.3. It is an
anonymised account of a real DV crime and it was written by the ‘offcer in the case’
(OIC). The witness signed it to verify that she had given the account (the witness did
not make a formal statement).
she said that she had gone into the male’s [her husband’s] bedroom to see what he
was doing and he became aggressive and threw the computer equipment at her.
This included a keyboard which had struck the female on the top of her leg.
(OIC notebook)

Analysing the initial crime report

Let’s frst look at the Material processes presented in the text above, including Actors,
Processes, Participants, and Circumstances. These are summarised in Table 9.6.
As can be seen in Table 9.6, there are three Actors: ‘she’, ‘he’, and ‘a keyboard’.
The frst Actor is the victim, who is presented as an Actor in a Material process, ‘had
gone’. There is no Goal affected by her actions. In (ii), the perpetrator is the Actor,
but the Goal of the process is not the victim, but ‘computer equipment’. In other
words, the suspect acts upon an inanimate object. The victim is backgrounded by
being placed in a prepositional phrase (‘at her’) and is therefore presented as being a
(deletable) Circumstance. The victim is not presented as someone affected by the
actions of the perpetrator, but rather as an orientational marker (the trajectory of the
throwing). The third Actor is the computer ‘keyboard’, which is rendered as the entity
that performs the assault as it is presented as striking the victim’s leg. This presents
that action as a Material-Event process. Aside from the Material process ‘threw’, the
only other process in which the suspect engages is a Relational one of ‘being’ (‘he
became aggressive’). He is the Carrier (‘he’) of the Attribute ‘aggressive’. What we
can see from the analysis is that the perpetrator – the person accused of doing
something – actually did nothing to anyone. The entire report is mediated

Table 9.6 The OIC’s account of the witness’s story

Actor Process Goal Circumstance

i she ii had gone into the male’s

[Material-Intention] threw bedroom
he [Material-Intention] struck at her
[Material-Event]
the computer
iii a keyboard on the top of
equipment the
her leg
female

219
 REPRESENTING EXPERIENCE IN DISCOURSE

through a Verbal process of ‘said’ which presents the information as reported rather
than experienced directly by him. By reporting second-hand information the OIC can:
1) neutralise his stance thus presenting himself as impartial; and 2) distance himself
from the truth of the reported content without going on record to express doubt (the
academic term for this is ‘epistemic distance’). In this way, if he is challenged, or if
the story is told differently at a later date, the offcer can resolve any anomalies by
pointing out that he merely wrote what was reported to him.

Analysing the summary report (MG3) of the domestic violence

crime
After the offcer interviews the suspect (we do not have access to this interview) he
achieves a fuller picture of the case. He writes up his summary of the report, the MG3
form (see Figure 9.7 above). This is what the OIC writes (‘DP’ is ‘detained person’):
She stated her husband (the DP) had thrown a computer keyboard at her which had
struck her on the thigh causing injury. He states she came up and he was perhaps a
little agitated as he was tired and swung the computer around catching her on the
leg.
(MG3 form)
The entire case has now been reduced to two versions of one particular event. If we
consider only the Material processes, outlined below in Table 9.7, we can see how the
relationships between Actor, Process, and Goal become more abstract as we go down
the table.
In any report of a crime, one would reasonably expect Material processes to
predominate (it is, after all, a report of ‘doing’). Undoubtedly, this report

Table 9.7 Material Processes in the OIC’s reformulation of the at-issue event
Actor Process Goal Circumstance

1 her husband had thrown a computer at her

[Material-Intention] keyboard

2 [a computer had struck her on the thigh

keyboard] [Material-Event]

3 She came up
[Material-Intention]
4 [he] swung the computer around
[Material-Intention]
5 [the computer] catching her on the leg
[Material-Event]

220
 REPRESENTING EXPERIENCE IN DISCOURSE

contains ‘doing’. However, what is interesting about these Material processes is who
or what fulfls the ‘Actor’ and ‘Goal’ roles. The perpetrator is the explicit Actor in
only one Material process (of ‘throwing’) (1) and implicitly in another (‘swung’) (4).
In (1) which is mediated through the victim’s perspective, the ‘Goal’ is the keyboard,
not the female victim. She is relegated to the Circumstance (‘at her’) which maintains
the original text as entered into the OIC’s notebook. The second Material process
(‘swung’) (4) affects an inanimate ‘Goal’ (‘the computer’). Again, the perpetrator
does not act on the victim. The perpetrator can be perceived to have acted intentionally
in ‘swinging’ the computer. However, this choice of a Goal-less process removes
culpability; in other words, he cannot be considered to have ‘intentionally’ struck the
victim as he merely swung the keyboard ‘around’. The actual assault is presented in
(5) through the Material processes ‘catching [her on the leg]’ and (2) ‘struck’. If we
consider who carried out these actions, we can see that they are executed by non-
human Actors – thus, both are Material-Event processes.
Let’s shift our attention now to the Goal element as this captures the entity affected
by the actions of the Process. In Table 9.7, there are only two clauses where the female
is the affected entity or ‘Goal’. In both, the Actor acting upon her is inanimate, ‘a
computer keyboard’ (2) and ‘the computer’ (5). As stated, in (5) the use of ‘catch’
implies accidental and thus unintentional contact (contrast with ‘struck’ in the
original). In summary, the OIC presents the victim as the affected entity on two
occasions where she has been acted upon, both times by an inanimate object. The
perpetrator never does anything to her – when he acts, he does so only upon an
inanimate object. In this report, computer keyboards and computers act and go on to
cause harm to the victim.

Relational processes in the MG3 account of the crime

It is also worth looking at the Relational processes in the text. Table 9.8 below outlines
them.
Relational processes tend to convey states of ‘being’ or ‘having’ through the
participant roles of ‘Carrier’ and their relation to an ‘Attribute’. Often such attributes
are beyond the control of the Carrier. For example, in ‘Patricia is Irish’ then Irishness
is a quality or Attribute that Patricia inherently holds and there is not much agency
around the having or not having that Attribute. Of course, that

Table 9.8 Relational processes of ‘being’ from the OIC’s MG3 form
Carrier Process Circumstance Attribute

6 he was perhaps a little agitated

7 [he] was tired
is not true of all Attributes, but ‘being’ something encodes less agency and
intentionality than ‘doing’ something. The clauses ‘he was perhaps a little agitated as

221
 REPRESENTING EXPERIENCE IN DISCOURSE

he was tired’ contain Relational processes, namely ‘agitated’ and ‘tired’. The
Attributes are hedged with Circumstantial components, too, ‘perhaps’ and ‘a little’
which softens or mitigates the Attribute (see Chapter 7 for discussion of ‘hedging’).
These clauses precede the action of swinging the computer ‘around’. Using these
Relational processes to frame the violence can lead to the implicature (see Chapter 6)
that the perpetrator’s actions were compounded – and even excused – by his tired and
agitated state.

Analysing the Gatekeeper’s contribution to the case

Finally, the last link in the chain of policing the crime (pre-outcome) is the
contribution by the ‘Gatekeeper’ who reviews the OIC’s account before it goes to the
Crown Prosecution Service for a charging decision. The Gatekeeper in this case added
his contribution to the OIC’s account of the crime as follows.
In interview D/P [detained person] admits that he was tired and agitated and was
trying to sleep. An argument develops between himself and his wife resulting in
him picking up the computer and swinging it towards his wife. He is suitable for
diversion by way of a caution. He has no previous cautions or convictions. PND
[Penalty Notice] not possible for a potential domestic.
(Gatekeeper’s input to MG3 form)
An interesting feature of this account is its organisation or structure. The full account
leads with a Verbal process ‘admit’, which presents the suspect as playing a
facilitative role in the investigation. It’s important to point out that an admission of
the offence is a necessary precondition for a ‘simple caution’ (nonprosecutorial and
thus non-conviction) outcome. However, what the suspect is ‘admitting’ to is not the
assault, but the Relational processes of being ‘tired and agitated’. Notice how the
Attributes ‘tired’ and ‘agitated’ have greater epistemic certainty in the Gatekeeper’s
version than in the hedged OIC version. By leading with the Relational processes, the
Gatekeeper frames the internal state of the suspect as the lens through which to read
his criminal actions. In effect, by making this the Gatekeeper’s point of departure it
primes the reader to empathise with the suspect. Table 9.9 shows these clausal
confgurations.
The Relational-Attributive process in (10) ‘has no previous cautions or
convictions’ is used to support the recommendation that the perpetrator should receive
a Caution which would dispose of the case in a way that means the suspect does not
end up with a criminal record. This recommendation is presented through a Relational
process of ‘being’ (‘is suitable’) in (Table 9.9), implying that he meets the criteria
well.
Table 9.10 presents the Material processes in the Gatekeeper’s contribution.

Table 9.9 Relational processes from the gatekeeper’s input on the OIC’s MG3 report

222
 REPRESENTING EXPERIENCE IN DISCOURSE

Carrier Process Attribute Circumstance

8 he was tired and agitated and perhaps; a little for diversion by

trying to sleep suitable way of a caution
9 [he] is no previous cautions or
10 He has convictions

11 PND [is] not possible for a potential domestic

The Gatekeeper backgrounds the victim by relegating her to Circumstantial
constituents, ‘towards his wife’ (Table 9.10). The Gatekeeper then presents the
‘argument’ as an Actor in a Material-Event process, ‘An argument develops’. The DV
incident is presented as a direct result (‘resulting in’) of this ‘argument’.
None of the preceding texts in this case refer to an ‘argument’ (noun) or ‘argue’ (verb),
so this nominalisation is the frst appearance of the term. Taken together, the
nominalised ‘argument’ mitigates the suspect’s role in the crime because it doesn’t
name him as agent and it divides responsibility. This is different from the victim’s
account in the OIC’s notebook in which she is presented as going into his room ‘to
see what he was doing’, which is not the same phenomenon at all as ‘an argument’!
Framing the assault as preceded by an ‘argument’ could be read as an attempt to justify
the violence through provocation. And we have already encountered an ‘excuse’ that
appears to justify the DV incident in the OIC report (‘he was perhaps a little agitated
as he was tired’), so the ‘argument’ frame certainly gives more credence to a narrative
that positions the suspect as retaliating to provocation. In terms of what the suspect
actually does in the Gatekeeper’s text, there are only two Material processes that he
engages in, ‘picking up’ (13) and ‘swinging’ (14). These two processes do not affect
any other human entity. Indeed, there is no reported outcome of the ‘swinging’ of the
computer. The only information offered is its trajectory ‘towards his wife’,

Table 9.10 Material processes in the gatekeeper’s review of the OIC’s MG3 report
Actor Process Goal Circumstance

12 An argument develops between himself

[Material-Event] picking and his wife
13 him up
[Material-Intention]
14 [him] swinging towards his wife
the computer it
[Material-Intention]
[the computer]
another Circumstantial component (14). The object is not presented as having made
contact with her in any way. In summary, the Gatekeeper’s reformulation not only
strikes the act (of common assault) from the formal record but presents the suspect as

223
 REPRESENTING EXPERIENCE IN DISCOURSE

acting only on a computer. No human participants are affected in any way in this
retelling. Unsurprisingly, the case was disposed of out of court with a simple caution.

Conclusion
This chapter has introduced transitivity as a serviceable framework for analysing
agency in discourse. Using police crime reports, it shows how discursive
constructions of victims and perpetrators can predispose decision makers to outof-
court options by what they choose to report and what they choose not to report in
formal crime records. It also shows the utility of transitivity in judicial discourses
because it can help identify patterns of reporting that can have very real consequences
in the pursuit of justice. A more detailed analysis could determine whether or not such
linguistic choices are systemic (i.e. institutional-wide), and whether there is
consistency in the presentation of specifc roles (e.g. Actor/ Goal) or specifc
transitivity confgurations (e.g. Material-Intention vs MaterialSupervention; Material
vs Relational) that could help explain why some of the many dismissed or early
disposed of cases do not proceed through the criminal justice system. The chapter also
shows how transitivity can be usefully employed alongside analyses of implicated
meanings (see Chapter 6).The following chapter develops Verbal processes by
examining how self and others’ speech and thought is presented in discourse.

Further reading
A transitivity analysis can elucidate patterns of experience and ideas in a range of
discourse contexts. For an accessible introduction, see Bloor and Bloor (2013). One
clear example that pre-dates the analysis in this chapter, and which offers a good
introduction to the model is Kate Clark’s article (1992) on the transitivity patterns in
tabloid press reports of sexual violence. Clark exposed a male-centric ideology that
underpinned gender-based discrimination that consistently negatively appraised
women and, at the same time, defected blame from their male attackers. Canning’s
article (2022) looks at further examples of domestic violence crimes using transitivity
as an analytical approach. Beyond forensic discourse, transitivity has been applied in
literary fction to demonstrate limited mind style (Burton 1982; Ji and Shen 2004) but
see Halliday (1971). For more on nominalisation, see Fowler et al. (1979); Fowler
(1991); and Billig (2008).

224
 REPRESENTING EXPERIENCE IN DISCOURSE

Answers to activities
Activity 9.2 SPOCA spotting
Identify the SPOCA elements of the retelling in the following sentence.

Hi, I’ve just seen a man push over an old lady and steal her bag

I = Subject
‘ve [just] seen; push; steal = Predicator a
man; an old lady; her bag = Object over
= Adjunct. just = Adjunct

Notes
1 Of course, they could also appear in the passive mood (‘a robbery was devised’; ‘his approach
was heard’).
2 We don’t understand why they are labelled ‘intensive’. We like the term ‘inherent’ or something
even remotely connected to the sentiment of ‘being’, but we will stick with the terminology
commonly used in the transitivity model so as not to confuse the issue.
3 In earlier models of transitivity including Halliday’s, there is an additional process,
‘Behavioural’, which exists between Material and Mental processes. They account for typically
involuntary processes that the body produces (such as coughing, sleeping, and so forth), but most
can fall within Material-Supervention processes enough to problematise the distinction between
Behavioural and Material. In fact, our students have often questioned this category and we agree
that it makes the model unnecessarily complex. Therefore, we have omitted Behavioural
processes from our discussion in this book.

References
BBC News (2022) ‘Cody Fisher: Murder arrests over Birmingham nightclub
stabbing’. 28 December 2022. Accessed 13 February 2023. https://www.bbc.
com/news/uk-england-64108203
Berry, M. (1975) An introduction to systemic linguistics. London: Batsford.
Biber, D. and Conrad, S. (2019) Register, genre, and style. Cambridge: Cambridge
University Press.
Billig, M. (2008) ‘The language of critical discourse analysis: The case of
nominalization’. Discourse & Society, 19(6): 783–800.
Bloor, T. and Bloor, M. (2013). The functional analysis of English. London:
Routledge.

225
 REPRESENTING EXPERIENCE IN DISCOURSE

Burton, D. (1982) ‘Through glass darkly: Through dark glasses’. In R. Carter (ed.),
Language and literature: An introductory reader in stylistics, pp. 194–214.
London: George Allen and Unwin.
Calligan, S. (2000) Points to prove. East Yorkshire: The new police bookshop (5th
edition). Goole, Yorkshire: New Police Bookshop.
Canning, P. (2022) ‘Writing up or writing off crimes of domestic violence: A
transitivity analysis’. Language and Law /Linguagem e Direito, 8(2): 48–69.
Canning, P. (2023) Forensic stylistics’. In M. Burke (ed.), The Routledge handbook
of stylistics (2nd edition), pp. 521–541. London: Routledge.
CJA (1988) Criminal Justice Act Section 39. Common Assault. Act of the Parliament
of the United Kingdom.
Clark, K. (1992) ‘The linguistics of blame: representations of women in the Sun
reporting of crimes of sexual violence’. In Michael Toolan (ed.), Language, text
and context: Essays in stylistics, pp. 208–224. London: Routledge.
Coulthard, M. (2004). ‘Author identifcation, idiolect, and linguistic uniqueness’.
Applied Linguistics, 25(4): 431–447.
Fowler, R. (1991) Language in the news. London: Routledge.
Fowler, R., Hodge, B., Kress, G., and Trew, T. (1979) Language and social control.
London: Routledge.
Halliday, M. (1971) ‘Linguistic function and literary style: An enquiry into the
language of William Golding’s The Inheritors’. In S. Chatman (ed.), Literary
style: A symposium, pp. 330–368. Oxford: Oxford University Press. (Reprinted in
J. Weber (ed.) (1996) The stylistics reader, pp. 56–86. London: Edward Arnold.)
Halliday, M. A. K. (1994) An introduction to functional grammar (2nd edition).
London: Hodder Arnold.
Halliday, M. A. K. and Matthiessen, C. (2014) An introduction to functional grammar
(4th edition). London: Routledge.
Ji, Y. and Shen, D. (2004) ‘Transitivity and mental transformation: Sheila Watson’s
the double hook’. Language and Literature, 13(4): 335–348.
Lea, S. J. and Lynn, N. (2012) ‘Dialogic reverberations: Police, domestic abuse, and
the discontinuance of cases’. Journal of Interpersonal Violence, 27(5): 3091–
3114.
Lynn, N. and Canning, P. (2021) ‘Additions, omissions, and transformations in
institutional “retellings” of domestic violence’. Language and Law/ Linguagem e
Direito, 8(1): 76–96.
ONS (2020) ‘Domestic abuse prevalence and trends, England and Wales: year ending
March 2020’. Offce for National Statistics. Accessed 10 May 2021.
https://www.ons.gov.uk/peoplepopulationandcommunity/crimeandjustice/
articles/domesticabuseprevalenceandtrendsenglandandwales/yearending
march2020
Police National Legal Database (2019) Accessed 12 November 2021. https://
www.pnld.co.uk (subscription required).

226
 REPRESENTING EXPERIENCE IN DISCOURSE

Ryder, M. E. (2008) ‘Overhauling transitivity’. Conference paper given at annual

conference of the Poetics and Linguistics Association (PALA), University of
Sheffeld, Sheffeld.
Simpson, P. (1993) Language, ideology, and point of view. London: Routledge.
Simpson, P. (2014) Stylistics: A resource book for students (2nd edition). London:
Routledge.
United Nations (n.d.) ‘What is domestic abuse?’ Accessed 7 July 2023. https://
www.un.org/en/coronavirus/what-is-domestic-abuse

227
 Presenting other people’s speech,
10 writing, and thought

Introduction
Have you noticed that when you are in conversation with other people you sometimes
tell your conversational partner what someone else said to you in another
conversation? We do this a lot in everyday, naturally occurring conversations. This
might include, for example, what someone said to us directly (e.g. in another
conversation) or what we have overheard being said (e.g. what was said by someone
on the TV or in a podcast). We can also tell our interlocutors what other people have
written, for example, in a newspaper column, on Twitter, or in a blog. We even tell
people what we thought and what other people are thinking. And we can also talk
about what we are going to say to another person in the future (‘I’m going to tell her
I can’t work at the weekend’).
Similarly, you may have noticed that in news reports (such as those printed in
newspapers, published on websites, or broadcast on TV and radio) we are often told
what people have said or written elsewhere. Indeed, news reports are often about what
other people (such as politicians, scientists, and celebrities) have said or written. Other
people’s discourse (or bits of it) can create news and, consequently, news report can
be full of the presentation of other people’s words and sometimes even their thoughts.
The phenomenon we are referring to is known as Discourse Presentation (DP) and
is where we take words (whether spoken or written) and thoughts from (typically) an
earlier discourse (e.g. a spoken conversation, a Twitter feed, a web page) and present
them in the ongoing discourse. Discourse Presentation, which is also known as
Speech, Writing, or Thought Presentation (SW&TP), recognises that when we
communicate, we often draw on and include other discourses (or bits of them, at least)
in our own discourses.
Discourse Presentation occurs in many communicative contexts including, for
example, prose fction, academic writing, police statements and news report. As we
will discover in this chapter, when presenting the words (and sometimes thoughts) of
others, we have several options. Readers will, no doubt, be familiar with the idea of
quoting other people, but there are many other ways in which other people’s words
can be presented. Our discussion will be based on the framework of DP that was
originally formulated by Leech and Short (2007) and later developed in the work of
Semino and Short (2004) and Short (2012). We

DOI: 10.4324/9781003351207-10
will also draw on the ideas presented in Thompson (1996) which provides an
alternative but complementary discussion of discourse presentation. Our exploration
of discourse presentation in this chapter will focus mainly on speech and writing.

228
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

Many of the examples we use are drawn from a large online corpus of news report
(see Chapter 11).

What is discourse presentation?

Prototypically, discourse presentation is the presentation of speech, writing or thought
from an anterior (earlier) discourse in a posterior (later) discourse. This defnition is
illustrated by Figure 10.1, which depicts two conversations (labelled 1 and 2).
Conversation 1 happens earlier in time and is the anterior discourse situation. In this
conversation, Jules says to Ginny “What are you still doing here at this time?” Later,
Conversation 2 occurs, known as the posterior discourse situation. In this
conversation, Ginny tells Aled what Jules said. Discourse presentation therefore
involves the content (or some part of it) of an earlier discourse being embedded in a
later discourse.

Figure 10.1 The prototypical situation for speech presentation

The example in 10.1 shows just one of a few options open to Ginny for presenting
what Jules said. We will discuss the range of options in detail in the next section. The
example, which uses what is known as direct speech, exemplifes the different possible
elements of discourse presentation (see Thompson 1996: 507). First, there are words
that introduce or ‘signal’ (Thompson 1996) the discourse presentation, and then there
are words that reproduce or indicate the content of the original, anterior, discourse.
Within the signal, there is language that points to the source of the original discourse
(‘Jules’), and there is an indication of the speech act performed (see Chapter 4) by the
original discourse (‘asking’). Finally, there is language that takes an evaluative stance
or ‘attitude’ (Thompson 1996) towards the content of the original discourse, and/ or
the manner of production, and/or the source (‘fatly’). These elements are summarised
in Figure 10.2.

229
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

Figure 10.2 Elements of discourse presentation

Following Thompson (1996), when analysing discourse presentation in a text we

can use these elements to ask a series of questions about the presentation, as set out
below.

1) What language is used to indicate (‘signal’) the presence of another

discourse?
2) What content of the original discourse is presented and how is it presented?
3) What source is attributed to the discourse being presented?
4) Does the language used to present the anterior discourse suggest:
a. the speech/writing/thought act that was performed by the original
discourse; and/or
b. the manner of production of the original; and/or
c. the emotive state of the original source.
If so, how is it/are they indicated?
5) Does the language used to present the anterior discourse suggest that the
presenter is taking an evaluative stance (‘attitude’) toward the original
content and/or the source?

The frst two questions are about identifying the linguistic forms that indicate discourse
presentation is occurring in the text under analysis and, importantly, what content
(words and/or ideas) comes from the anterior discourse. Answering these two
questions can be tricky and is the crux of the analytical framework we will describe
below. The third question deals with whether the source of the original is identifed,
which can be important when thinking critically about the use of Discourse
Presentation to, for example, support claims or arguments (we return to attribution
later in the chapter). The fourth and ffth questions deal with the language choices made
by the person presenting the original discourse and how these can encode an evaluative
attitude toward the content of the discourse and the producer. The language choices

230
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

relate to how the original discourse is ‘signalled’, and typically involves verbs and
adverbs (but, as we will see in subsequent sections, other forms are possible).
The speech/writing/thought act is what the producer of the original discourse is
doing with their discourse (e.g. asking a question). The label applied to the act is a
matter of interpretation and the choice of the person presenting the discourse. As we
saw in Chapter 6, questions (interrogatives) do not always seek information; they can
function, for example, as requests for action. The labelling of the
speech/writing/thought act therefore depends partly on the structure of the original
(e.g. declarative, interrogative, imperative), and the illocutionary force (see Chapter
4). In our example, Ginny chose the verb ‘asks’ which neutrally labels the
interrogative form of original discourse. However, instead of using ‘asks’, Ginny
could have chosen ‘demands’, which evaluates Jules as having a particular attitude:
perhaps overbearing and/or assuming discursive power at that point in the
conversation. This might lead the hearer to view Jules negatively. As well as attitude
towards the person who produced the original discourse, verb choice can express an
attitude towards the content of the original discourse (see Thompson 1996: 521–522).
For example, the reporting verb ‘claim’ tends to suggest that the person presenting the
discourse is dubious about or is less committed to the propositions contained in the
content of the original.
The manner of production relates to prosodic features (e.g. loudness, voice
quality, speed of production) and is again down to the presenter to decide upon and
label using the reporting verb (e.g. ‘shouted’) and/or a supporting adverb
(e.g.‘loudly’). Depending on the choice of verb or adverb, the manner of produc
tion can suggest emotive states.
Emotive states can be described by the choice of reporting verb (‘sneered’) and/or
adverb (e.g. ‘angrily’). In our made-up example, ‘fatly’ evaluates the way the words
were said and suggests that Jules was either lacking in emotion or being frm. This
evaluation is supplied by Ginny, the person presenting the anterior discourse; Jules
did not say ‘I ask you fatly . . .’. The choice of adverb therefore suggests Ginny’s
impression of Jules’s attitude. It also invites the hearer to share in that evaluation.
As we have seen above, the choice of words used to label the act performed, the
manner of production and emotive states can suggest the attitude of the original
speaker/writer/thinker, and that the presenter is taking an attitude toward the original
content and/or the source.
In the following sections, we will consider the elements set out above as we
describe the different types of discourse presentation. Readers may wish to refer to
these questions as they read on.
Discourse presentation and different voices
For Thompson (1996: 506) discourse presentation is “signalled voices in the text”
which includes “any stretch of language where the speaker or writer signals in some
way that another voice is entering the text”. This defnition highlights the idea that the
presentation of speech, writing and thought brings other voices into a discourse, and
that the different voices are signalled so that the hearer/ reader recognises them as
231
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

belonging to someone (or something) else (although, as we will see, the free indirect
categories in the model make other voices diffcult to spot). Fairclough (2003: 41)
points out that different voices bring with them different points of view and
perspectives on the world including, for example, different ideas and opinions (as we
saw in Chapter 9).1 They can introduce information, knowledge, evidence, and
arguments to help develop a line of discussion or persuade an interlocutor of the
validity of whatever is being reported. Moreover, as we saw in the previous chapter,
the inclusion of other voices can also help to preserve impartiality or neutrality
towards what is reported. In the following sections, we will explore the notion of
different voices in discourses using mainly examples from news report. We will start
by describing in more detail the different options for discourse presentation to provide
the necessary tools for identifying the presentation of others’ discourse.

Report, representation, and presentation

In the version of the framework described in this chapter, the term ‘presentation’ is
used in the category names in preference to ‘report’ and ‘representation’, which are
both terms that have been used, sometimes interchangeably, in earlier versions of the
framework (e.g. Short 1996; Leech and Short 2007). However, Short et al. (2002) and
Short (2012: 19) make a distinction between these terms:

• discourse report assumes “a match between the lexis, deixis and grammar in the
anterior and posterior discourses”, when the direct forms are used; •
discourse representation “assumes a mismatch”;
• discourse presentation refers only to the forms used in the posterior discourse and
does not necessarily link back to an original, anterior discourse.

To refect this distinction, in later versions of the framework, the term ‘discourse
presentation’ is used in preference to discourse ‘report’ or ‘representation’ (see Short
et al. 2002: 336; Short 2012). The advantages of using the term ‘presentation’ include:

1) it does not make any assumptions about a match or mismatch between what is
presented and the original (anterior) discourse. Where there is an anterior
discourse that is accessible and available for scrutiny, the discourse analyst can
assess whether the discourse presentation in the posterior discourse is
representation or report;
2) it makes no prior assumptions about the existence of an anterior discourse, which
is the case with hypothetical discourse and that predicted to happen in the future.
Also, in fction there is no anterior discourse because it is all made up.

Therefore, in the model we describe in this chapter, rather than ‘R’ (for
‘representation’ or ‘report’), which is used with earlier publications, we use ‘P’ (for
‘presentation’).

232
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

Different types of discourse presentation

There are various ways of presenting other people’s words and thoughts from which
the person presenting the discourse (the presenter) makes a choice. The choices that
are made might be motivated by different reasons, some contextual, others to do with
stance and attitude toward the producer of the original (anterior) discourse and/or the
discourse itself. In this section, we will describe the different forms discourse
presentation can take. The categories we use, which are set out in Table 10.1, are based
on those in Short (2012) but with some minor modifcations that take into account the
work of Thompson (1996).
Table 10.1 shows three parallel clines for speech, writing, and thought that range
from direct and (prototypically) faithful rendering of the content of the original
(anterior) discourse (e.g. DS), to minimal forms that merely indicate that speech
writing or thought occurred (e.g. PS). The direct discourse presentation categories at
the top of the table (DS/DW/DT) involve only the supposed original speaker whereas
all other categories are a combination, in varying proportions, of the original speaker
and the person presenting the discourse.
Table 10.1 Three parallel clines of discourse presentation
Speech presentation Writing presentation Thought presentation

DS Direct Speech DW Direct Writing DT Direct Thought

FIS Free Indirect FIW Free Indirect FIT Free Indirect
Speech Writing Thought

IS Indirect Speech IW Indirect Writing IT Indirect Thought

PSA Presentation of PWA Presentation of PTA Presentation of
Speech Act Writing Act Thought Act

PS Presentation of PW Presentation of PT Presentation of

Speech Writing Thought

We will start our discussion of the framework by exploring in more detail direct
and indirect forms, referring mainly to speech and writing presentation. Each section
below will begin with some examples (drawn from various sources), which we will
then discuss in terms of their formal attributes and potential effects.
Direct Speech, Writing, and Thought (DS/DW/DT)
(1) “I deeply and bitterly regret that that happened,” he [Boris Johnson] said.
(The Times and The Sunday Times on
Twitter, @thetimes. 18/01/2022)

(2) Mr Blair said, “every avenue must be explored”.

(The Report of the Iraq Inquiry 2016: 372)

233
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

(3) “The subject ofence is extremely serious,” she wrote. “At the time of this ofence,
the accused was on bail in relation to a number of ofences.”
(ABC Premium News, Sydney, 19/08/2015)

(4) The minibar would not close properly after opening. Never mind, I thought.
(Corpus of London hotel online reviews3)

(5) “For God’s sake,” he said, “why did you say that about that girl in Strasbourg for?
Didn’t you see Frances?”
“No, why should I? If I know an American girl that lives in Strasbourg what the hell
is it to Frances?”
“It doesn’t make any diference. Any girl. I couldn’t go, that would be all.”
“Don’t be silly.”
(The Sun Also Rises, Ernest Hemingway, 1926)

Prototypically, direct speech, thought and writing comprises two clauses usually
known as a reporting clause and the reported clause.4 The reported clause contains
the quoted words while the reporting clause signals that someone else’s words are
being used. Examples (1), (2), and (3) demonstrate that the reporting clause can follow
(‘he said’) or precede the reported clause (‘Mr Blair said’) or be placed between two
reported clauses in medial position (‘she wrote’).
Within the reported clause, verb tense and pronouns are grammatically appropriate
to the original (anterior) discourse situation. So, in (1), the pronoun ‘I’ and the present
tense ‘regret’ are all appropriate for the discourse between Mr Johnson and his
addressees. The convention in written texts is for the directly quoted words to be
enclosed in quotation marks (inverted commas), although other graphological
indicators are also possible such as italics, indentation, and dashes. And, as example
(4) demonstrates, quotation marks can be missing altogether. In spoken discourse,
gestures (such as air-quotes) or a different voice quality might be used to signal direct
presentation (see Halliday and Matthiessen 2014: 447 and Clark and Gerrig 1990:
775–777).
Extract (5) is an example of what is known as Free Direct Speech and demonstrates
that direct presentation of discourse can be free (get it?) of reporting clauses. The
example is from Hemingway’s (1926) novel, The Sun Also Rises and is part of an
interaction between the main character and narrator (Jake Barnes) and another
character (Robert Cohn). The frst reported clause is (conventionally) followed by a
reporting clause (‘he said’), but subsequent reported clauses are not. Instead, the reader
must use the sequencing of the quoted speech to understand that these are alternate
turns in the conversation and to work out who says what.
One effect of direct discourse presentation can be, as Short (1996: 293) notes, to
bring the original speaker’s words (from the anterior discourse) to the foreground. This
effect is amplifed with Free Direct forms (as demonstrated in example 5), where the
presenter is backgrounded by the omission of presenting clauses.
The traditional view of directly quoted discourse is that it presents the exact words
(or thoughts) used in the original verbatim. With this view, direct discourse is
234
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

completely faithful to the original. However, the status of direct discourse

presentation largely depends on its context. For example, in spoken conversation, if
we quote what other people have said there is a tacit understanding between speaker
and hearer that any such quotes are likely to be inaccurate because in practice it is very
diffcult to remember exactly the words that someone used. In these cases, the primary
purpose of direct discourse presentation is to help make the story telling more vivid
and lively. In other contexts, however, such as news report and academic writing, our
expectation is that anything that is quoted directly will be largely faithful to the
original. We say ‘largely’ because even the most accurate quotation of spoken
discourse is likely to miss out repeated words, false starts, and other non-fuency
features such as ums and ahhs. Even with print news report, our expectation of
faithfulness can change depending on where the quotation occurs. For example, Short
(1988) discusses examples of Direct Speech in newspaper headlines that were not
actually said by the person they were attributed to. We return to the topic of
faithfulness later in this chapter.

Indirect Speech, Writing, and Thought (IS/IW/IT)

(6) Infectious diseases specialists told the Post on Tuesday that symptoms of the
Omicron variant emerged quicker.
(South China Morning Post, 12/01/2022)

(7) she [Charlotte Newton] thought that he was a soft target and that he would be
less likely to call the police because of their secret sexual arrangement.
(Hull Daily Mail, Hull [UK], 26/04/2022)

(8) By October 13, Taoiseach Micheál Martin said he did not guarantee that the
removal of restrictions on October 22 would proceed as planned
(Irish Independent, 01/01/2022)

(9) New Zealand school buildings were generally poorly-ventilated, Kvalsvig said.
(New Zealand Herald, 10/01/2022)
(10) Aaron Bell, member for Newcastle-under-Lyme, asked Boris Johnson if he took
him for a fool for following rules in the House of Commons on Monday.
(BBC News online, 01/02/2022)

(11) Win promised to campaign for a change in the law and has done so for the nearly
20 years since her husband died.
(Daily Mail, London, 03/02/2022)

Unlike direct discourse presentation, where the original discourse is (apparently)

quoted faithfully, indirect speech, writing, or thought paraphrases the original
discourse and presents the meaning carried by the original utterance (the propositional
content; see Chapter 5). The presenter is free to choose the words that they think
convey the meaning of the original. While some of the original words might be used
235
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

in IS, there is no guarantee of that so the hearer/reader does not know which words
belong to the original discourse and which to the presenter.
Indirect Speech, Writing, and Thought has a two-clause structure, with a reporting
clause and reported clause (which expresses the words from the anterior discourse).
However, with indirect discourse presentation, the reported clause is subordinate to
the reporting clause and does not incorporate quotation marks (because it is not
quoting but paraphrasing). The subordination might be marked by the subordinating
conjunction ‘that’ (as in 6 and 7), or it can be omitted (as in 8). Indirect presentation
involves shifts in deixis (e.g. ‘now’ becomes ‘then’, ‘here’ becomes ‘there’) and tense
and pronouns are changed to be appropriate to the posterior discourse. For example,
in (8):

• ‘did’ is used rather than the ‘do’ that would be present in the direct rendering (‘I do
not guarantee . . .’);
• ‘would’ rather than ‘will’; and
• ‘he’ is used instead of ‘I’.

There might also be a change in clause type (declarative, imperative, interrogative)

where, for example, an interrogative in the original discourse becomes a statement in
the posterior discourse. This is the case in example (7) where in the original, anterior
discourse, Aaron Bell says:
Does the prime Minister think I’m a fool?
Examples (9)–(11) demonstrate that indirect discourse has other structures. Example
(9) shows that, rather than the conventional fronted position, the reporting clause can
come after the reported clause (the use of the past tense verb-form ‘were’ suggests that
this is indirect presentation). Examples (10) and (11) demonstrate that other two-
clause structures feature in indirect discourse with an if-clause in (10) and a to-
infnitive clause in (11) carrying the content of the original (anterior) discourse.
Because indirect forms of discourse presentation involve paraphrasing, they are
seen as being less faithful to the (perceived) original than direct forms. They combine
the original (anterior) discourse with the (posterior) discourse within the reported
clause and allow the presenter to insert evaluations, judgements, or extra information
into what is being presented (in the reported clause). The voice of the text producer
therefore merges with the voice of the other person being presented which has the
effect of blurring the point of view distancing the reader from (i) the original discourse,
and (ii) the person that produced it.

Activity 10.1 Transpose Direct Speech into Indirect Speech

Test out your knowledge of indirect forms of discourse presentation by transposing the direct
speech used in Figure 10.1 into indirect speech: “What are you still doing here at this time?” We
give our answer at the end of the chapter.

236
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

Reporting clauses
The stretches of text that introduce discourse presentation are often referred to as
reporting clauses because they tend to be clausal structures and contain a reporting
verb such as ‘said’, ‘wrote’ or ‘thought’. Reporting clauses are an important aspect of
discourse presentation because of the extra information they can provide about the
original (anterior) discourse situation, including:

• mode of anterior discourse – indicated, for example, via reporting verbs (e.g. ‘said’,
‘thought’, ‘wrote’);
• attribution of source of original – typically who spoke, thought, wrote;
• the speech/writing/thought act performed – (e.g. ‘promised’, ‘threatened’) (see
Chapter 4);
• manner – indicated, for example, by verb choice (e.g. ‘shouted’) and adverbs
(e.g.‘loudly’);
• emotive states – indicated, for example, by adverbs (e.g. ‘angrily’).

It is important to remember that reporting clauses are part of the posterior discourse.
This means that they are formulated by the person doing the presenting and pertain to
their point of view. Therefore, the language choices for rendering the
speech/writing/thought act, manner, and emotive states are largely in the hands of the
presenter and how they evaluate the original discourse and its producer. Even the
choice of reporting verb can be subjective. For example, when introducing spoken
discourse presentation, instead of choosing the neutral reporting verb ‘said’, other
verbs such as ‘blustered’ could be chosen to suggest a more negative attitude toward
the original speaker. Such evaluative choices can be important since they can be
biased and infuence readers’ view of the producer of the original discourse. They can
also provide information about the attitude of the person doing the presenting toward
the original words or the original text producer (see Thompson 1996: 507–522).
Other linguistic structures are often used to introduce the presentation of other
people’s words and thoughts. Consider the following examples and, before reading
on, identify the words that act like a reporting clause by introducing or signalling
discourse presentation.

(12) One of the suggestions is that the amount ofered by Cold Weather Payment is
raised to address this crisis.
(Express online, London, 20/01/2022)

(13) The eventual declaration by the Minister of Finance that President Buhari does
not want Nigerians to sufer, hence the suspension of the implementation of
subsidy removal, came as the icing of the cake.
(This Day, Lagos, 30/01/2022)

(14) According to the documents, the waivers are related to work aimed at turning
Iran’s heavy-water Arak reactor into a less-dangerous light-water reactor
(Wall Street Journal online, 05/02/2022)
237
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

Extracts (12) and (13) are examples of Indirect Speech, where the noun phrases ‘One
of the suggestions’ and ‘The eventual declaration by the Minister of Finance’ signal
indirect discourse presentation. Extract (14) contains Indirect Writing presentation
where the adverbial ‘According to the documents’ introduces the propositional content
of the original discourse.
When analysing direct and indirect presentation it is important to consider both the
content of the reported clause and the choice of reporting verb and to remember that
the presentation of other people’s discourse is controlled by the person doing the
presenting.

Activity 10.2 Discourse presentation in naturally occurring conversation

Look at the extract below taken from the Santa Barbara Corpus of spoken conversations,
which demonstrates how words spoken by other people can be included in a conversational
turn. The extract is part of a longer conversation between two male friends (Fred and
Richard) where Fred is sharing a story about his shift at the factory he works at. Identify the
discourse presentation and any reporting clauses (or other signalling structures).

FRED: yeah, and so, he comes and says, well, he goes, I don’t know if you’ve, if you’ve, packed
this or not. You know. On your production card, all it says, you know, is that you did three
thousand sheets, but you did . . . but you didn’t pack it. So I go, yeah, I go look man, there
they are. You could see, there’s my name, stamped right on there. I just didn’t put it down.
And he goes, and what are you doing in the cafeteria so late. I’m just going, aw man, this is
the pits man.
(SBC047, On the Lot)

Free Indirect Speech, Writing, and Thought (FIS/FIW/FIT)

(15) One farmer quietly admits he worries about the growing numbers of racial
minorities; another enjoys hearing new accents at the grocery store.
(Asia News Monitor, 01/02/2022)

(16) ‘What does it mean, just a woman?’ he [Cipriano] said, quickly, sternly.
She hung her head. What did it mean? What indeed did it mean? Just a woman!
(The Plumed Serpent, D. H. Lawrence, 1926)

Free Indirect Speech, Writing, and Thought sit between direct and indirect forms of
discourse presentation on the cline shown in Table 10.1.They combine anterior and
posterior discourses but tend to be anchored to the presenter’s viewpoint. Unlike
Indirect forms of discourse presentation, Free Indirect Discourse presentation is ‘free’
from reporting clauses and has no stable set of formal features, which makes it
notoriously diffcult to identify. Typically, though, verb tense and pronouns are, as
with Indirect forms, aligned to the presenter, while other features such as lexis (e.g.

238
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

dialect, colloquialisms), terms of address, discourse markers (e.g. ‘Oh’), punctuation

(e.g. exclamation marks), syntax (e.g. incomplete sentences) and mood (e.g.
interrogatives) are, as with Direct forms, aligned to the original speaker/writer/thinker.
In our view, the underlined section of (15) is Free Indirect Speech. It is ‘free’
because there is no reporting clause to introduce what appears to be indirect speech
presentation of ‘another’ farmer. Arguably, ‘another’ can be viewed as an incomplete
reporting clause (something like: ‘another farmer says that they . . . ’). Often, it is the
immediate co-text that plays a crucial role in identifying Free Indirect forms such as
whether some other form of discourse presentation occurs immediately before or after.
In (15) the FIS is preceded by IS, and this steers the reader to understand that more
speech (the underlined part) is being presented rather than some observation made by
the writer. Sometimes, though, it is not clear whose point of view is being presented
and consequently it can be diffcult to ascertain whether evaluative language and other
markers of subjectivity relate to the posterior or anterior discourse.
Example (16) shows instances of Free Indirect Thought (underlined). We suggest
that the two questions and the exclamation are the protagonist’s (Kate Leslie) thoughts
to herself as she hangs her head and thinks about the question asked by Cipriano (a
Mexican general). Like (15), these examples of Discourse Presentation are Free
because there are no reporting clauses, and they combine features of Direct and
Indirect presentation. Both questions are structured as interrogatives, so retain the
same form as (we imagine) the storyteller would have used in the original (anterior)
discourse, but the verb tense is back-shifted (‘did’) and therefore aligns with the
posterior discourse. The internalised exclamation is grammatically incomplete and
includes an exclamation mark, perhaps indicating the strength of her thought. The
example shows how Free Indirect forms blend features of both Direct and Indirect
presentation, which is why these forms are positioned as they are on the DP cline.
The effect of Free Indirect Discourse depends on whether it is spoken, written, or
thought. According to Semino and Short (2004: 83–85), Free Indirect Speech (FIS)
distances us from the original discourse and can indicate the presenter’s evaluative
point of view towards the discourse and/or the person that produced it (see also Short
1996: 308; Leech and Short 2007: 268–270). Such distancing and point of view effects
may lead to irony, humour, or present an ideological stance. Free Indirect Thought
(FIT), however, brings us closer to the original thinker by providing access to their
thoughts, which may lead to a sense of intimacy and empathy (Semino and Short 2004:
124; see also Short 1996: 315).

Presentation of Speech, Writing, and Thought Acts

(PSA/PWA/PTA)
(17) She [Grace Tame] spoke of continuing her campaign to end sexual abuse against
children and others through education and legal reform.
(NCA NewsWire, 25/01/2022)

239
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

(18) The organisers of the protest have called for the forceful elimination of all
Covid-19 restrictions and vaccine mandates . . .
(Evening Standard, 31/01/2022)

(19) On Friday, she [Carrie Lam Cheng Yuet-ngor] ordered an investigation into the
conduct of the ofcials who attended the party.
(South China Morning Post, 11/01/2022)

(20) Australia has accused WeChat of taking down Prime Minister Scott Morrison’s
account and redirecting his followers to a site that claims to provide information
for Chinese Australians.
(CNN Business, 25 January 2022)

(21) It [pro-Beijing newspaper Ta Kung Pao] also blamed Cathay Pacifc air
crew for triggering the ffth wave of infections by violating home isolation
rules.
(South China Morning Post, 11/01/2022)

The next category of discourse presentation is Presentation of Speech/Writing/

Thought Act (PSA/PWA/PTA).5 These types of discourse presentation differ from the
indirect forms we discussed above because, rather than having a two-clause structure,
they comprise a reporting word (prototypically a verb) followed by a prepositional or
noun phrase which contains a summary of the discourse being presented. The
information relating to the content of the original discourse is therefore not a separate
clause but is embedded into the reporting clause. In example (17), the reporting word
is ‘spoke’ which is followed by a prepositional phrase beginning with ‘of’. Similarly,
in (18) ‘have called’ is the reporting verb phrase and this is followed by a prepositional
phrase beginning with ‘for’.6
Examples (19), (20), and (21) illustrate that some reporting words can indicate the
illocutionary force of the original discourse (e.g. ‘ordered’, ‘accused’, ‘blamed’) and
thus the speech/writing/thought act being performed. These sorts of reporting verbs
therefore help convey part of the content of the original discourse (Thompson 1996:
517). It is worth remembering, though, that the person presenting the discourse
chooses the reporting word so is also indicative of their evaluation of the original
discourse which may be biased and designed to steer the reader’s/hearer’s opinion.
Presentation of a speech/writing/thought act (PSA/PWA/PTA) does not present the
propositional content of original discourse but only a summary, so typically the
original (anterior) discourse cannot be reconstructed. This then is different from Direct
forms, which ostensibly renders the original discourse verbatim, and Indirect forms,
which paraphrase the original, typically using some of the same words. The amount
of information in the summary can differ considerably. Semino and Short (2004: 52–
53) found that in news texts the indications of content could be lengthy and detailed.
Example (20) contains the largest amount of summary information in a prepositional
phrase consisting of 23 words (including ‘of’). The summary is presented from the
point of view of the text producer (the presenter) who controls the lexical choices used
240
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

to summarise the original discourse and the summary may be interspersed with
additional information and opinions not in the original discourse. There is therefore
greater potential for the merging of anterior with posterior discourse with this type of
discourse presentation. For example, in (20) we might wonder whether the use of
‘claims’, which casts doubt on the information contained on the website, stems from
the original discourse, or was added by the text producer.
In their original conception of this type of discourse presentation, Leech and Short
(2007: 260) say that it is “on a par with other kinds of action”. This is partly because,
structurally, the summary of what is said, written or thought is embedded in the same
clause as the reporting word. Thompson (1996: 517) suggests that this is likely to place
the focus on the communicative act or event rather than the content, making the
content of the original discourse less salient.
Presentation of Speech, Writing, and Thought (PS/PW/PT)
(22) European ofcials have been scrambling to secure backup energy supplies in the
event that a confict disrupts fows from Russia, enlisting the help of the U.S. and
talking with gas producers such as Qatar and Azerbaijan.
(Wall Street Journal, New York, 01/01/2022)

(23) President Uhuru Kenyatta on Thursday held one-on-one talks with Rwanda’s
President Paul Kagame who was in the country on a working visit.
(AllAfrica.com, 03/02/2022)

(24) Johnson attended the party and gave a speech, it is understood, but he left
after a few minutes.
(Daily Mail, 02/02/2022)

(25) Warnings issued after group fned for luring 4m croc out of Proserpine River
(ABC Premium News, Sydney 02/02/2022)

(26) He [Pete Docherty] mumbled darkly and incoherently between songs.

(Sunday Herald, Glasgow, 25/09/2005)

(27) A Salvation Army leader threatened his wife after she caught him having sex
with a co-worker in the religious charity’s ofce. [. . .] The dad shouted and swore
before following Sara when she ran from the premises in tears.
(Daily Record, Glasgow, 03/02/2022)

(Note: Discourse presentation is underlined in examples 22–27)

Presentation of Speech/Writing/Thought (PS/PW/PT) are the most minimal types of

discourse presentation, and the categories recognise that sometimes we are told only
that speech, writing, or thought occurred. As the examples above (22–27)
demonstrate, with PS/PW/PT we are not given any details about the content of the
anterior discourse. So, in (22) we know that European offcials have been talking with
gas producers, but no details are given about the content of the talk (although we can
241
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

probably infer from the rest of the sentence that the general topic of conversation was
gas and its supply!). There is no summary either, as would be the case for
PSA/PWA/PTA.
PS, PW, and PT are manifest in a variety of forms. In example (22), the ‘ingform’
of the speech-verb ‘talk’ indicates that spoken discourse has occurred. In example (23)
it is the noun phrase ‘one-on-one talks’ (which is the direct object of the verb ‘held’)
that informs us of a spoken interaction. Similarly, in (24) and (25) it is the nouns
‘speech’ and ‘warning’ that reference anterior discourses and not the verb forms ‘gave’
and ‘issued’. With (24), the discourse is obviously spoken, however in (25) ‘warnings’
could be written and/or spoken.
The plural form nonetheless suggests reference to multiple discourses. Example (26)
demonstrates that words that reference the manner of production (in this case a verb
form – ‘mumbled’) can also signal the presence of an anterior discourse (in this case
spoken). Finally, example (27) highlights that some lexical choices can indicate the
illocutionary force of the discourse. The verb form ‘threatened’ summarises the
overall force of the salvation army leader’s verbal output (similar effects can also be
achieved with nouns, e.g. ‘threats’). The use of ‘swore’ strongly suggests that
expletives were used and therefore constrains and/ or suggests possible options for
what the man said.
Notice that in (22) that while the speech verb (‘talking’) is followed by a
prepositional phrase (‘with gas producers’) this is not an example of PSA because
rather than giving information about the content of the discourse, the phrase serves to
tells us who the other parties involved in the talking were. The lack of any sort of
summary of or content from the anterior discourse helps us see that this Discourse
Presentation is unlike any that we have seen so far in this chapter and is a case of PS.
With presentation of speech, writing and thought (PS/PW/PT), the text producer is
apparently in complete control of the presentation of discourse, and we are furthest
away from the original speaker/writer/thinker and their words. The distancing effect
is demonstrated particularly well in Example (27) where we are not privy to the
content of the abuse. All we are told is that a person shouted, threatened, and swore,
but we are not given any hint of anything that was said. With these forms, then, there
is only the reporting word, which is chosen by the text producer, and these alone
provide information about the original discourse. It is, however, possible for the choice
of verb to hint at the content or provide the essence of what was said (or written or
thought).

Activity 10.3 Josh gives his opinion

The following sentence is taken from a UK daily newspaper. Does the sentence contain any
Discourse Presentation? If so, can you identify which kind? What, if any, are the effects of the
language choices? We give our answers at the end of the chapter.

Hollywood actor Josh Hartnett has waded in on the Downing Street controversy and Boris
Johnson’s actions during lockdown.

242
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

Daily Mirror (UK), 01/02/2022

Summary
The categories of discourse presentation discussed in this section provide a framework
for identifying and analysing discourse presentation in texts. Each of the different
categories in the framework express varying degrees of ‘interference’ (Leech and
Short 2007: 260, 276) by the text producer, as well as claims to the faithfulness of the
reporting of the supposed original utterance. The categories, which we summarised in
Table 10.1, form a continuum ordered according to “[. . .] the amount of ‘involvement’
of (i) the original speaker in the anterior discourse and (ii) the person in the posterior
discourse presenting what was said in the anterior discourse [. . .]” (Semino and Short
2004: 10). The continuum ranges from categories at the bottom of the table to the
direct discourse presentation categories at the top of the table, which apparently
involve only the original speaker. Anything lying in between these two extremes is a
combination, in varying proportions, of the original speaker and the person presenting
the discourse. Moving up the table through the categories coincides with a gradual
transition in viewpoint, shifting more and more from the point of view of the person
presenting the discourse, to the viewpoint of the original speaker/writer/thinker.

243
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

Activity 10.4 Discourse presentation from news story on the BBC website

Below is an example of discourse presentation taken from a news story that concerns the
resignation of a UK TV historian from the Mary Rose Trust following comments he made during
an interview on a YouTube channel. First read an extract of the statement made by the Trust.

(A) we were appalled to hear Dr Starkey’s public comments on slavery today. Mary Rose Trust
is a charity that exists for the beneft of everyone and we have zero tolerance for such
comments. The board of the Mary Rose Trust has therefore accepted Dr Starkey’s
resignation
(The Mary Rose @MaryRoseMuseum, 6:48pm,
Thursday 02 July 2020)

Now read the reporting of the statement in the news story.

(B) The Mary Rose Trust said it was “appalled” by Starkey’s comments, adding on Thursday
evening they had accepted his resignation.
(BBC News, 03 July 2020: ‘David Starkey resigns from university
role over slavery comments’,
https://www.bbc.co.uk/news/entertainment-arts-53279273)

Answer these questions:

i. Establish what type of discourse presentation is being used in (B). Spell out the formal
features that lead you to your decision.
ii. Comment on the word in quotation marks in the news story. iii. Say whether you
think the discourse presentation is faithful to the original (A).

We provide our responses at the end of the chapter.

Attribution of source of original discourse

Earlier in this chapter we saw that one of the functions of reporting clauses (and other
reporting structures) that ‘signal’ discourse presentation is to attribute the source of
the original (anterior) discourse. As such, attribution relates to what is known as
evidentiality which is, according to Aikhenvald (2004: 3),“a linguistic category
whose primary meaning is source of information”. Aikhenvald goes on to say that for
many languages, evidentiality is a grammatical system realised, for example, through
morphology. English, however, has no such system; instead, it has numerous lexical
ways in which the source of information can be expressed, which include attribution
in reporting clauses. Aikhenvald (2004: 10) stresses, though, that these lexical means
are not the same as a grammatical category, not least because they are non-obligatory
lexical choices.7 That means that when another discourse (and therefore another voice)
is incorporated into a discourse the text producer can choose whether and how to
attribute the source. Exploring these choices, via the framework presented in this
244
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

chapter, can be important for discourse analysis because Discourse Presentation and,
more specifcally, attribution can be used, for example, to legitimate actions and
practices.

Legitimation
Legitimation refers to providing justifcation for doing something. In his work on
discourse and social practices, van Leeuwen (2008: 105) suggests that legitimation is
the use of language to answer the questions ‘“Why should we do this?” and/or “Why
should we do this in this way?”’ For example, consider the following invented
conversation between two people:

A: We must wear face coverings in public buildings.

B: Why must we?
A: Because medical experts say it helps to stop the spread of Covid.

In our example, person A legitimises her assertion by referring to medical authority

via Indirect Speech (or Indirect Writing). According to van Leeuwen (2008: 106–109),
“legitimation through authorization” can be achieved by attributing discourse to
those in a role of authority, (e.g. teacher, parent), those who have expert authority (e.g.
scientists, medics), role models (e.g. celebrities, ‘infuencers’), and impersonal
authorities (e.g. laws, rules). So, we often rely on others’ speech (or writing) to
legitimise what we say. In general, reference to another discourse can validate our own
experiences, provide support for our claims, and help to substantiate and increase the
credibility of our own stories and opinions.

Faithfulness
We have established that people can present the words (written or spoken) and
thoughts of other people. You may have been in the situation when someone has
quoted your own words back to you, telling you that ‘you said this’ or ‘you said that’,
but they get what you said wrong; they misquote you! You might reply with something
like ‘I never said that!’ Or maybe they don’t quote you but present what you said in
their own words (paraphrase) which does not quite capture the essence of what you
said. In that situation you might respond with ‘That’s not what I meant!’ You may
have also seen occasions when people say that they were quoted out of context. The
point is that when we quote other people (and when other people quote us) or present
the gist of what was said, it is usually quite important that we do it accurately and
honestly. This is known as being faithful to the original, and the notion of faithfulness
is a crucial element of Discourse Presentation. As readers or listeners, it is important
for us to have confdence that when other people’s words (and thoughts) are presented
they are faithful to the original.
For Short et al. (2002: 328) faithfulness refers to “specifying as accurately as is
feasible in context the precise communicative content of the discourse being reported”
245
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

and involves “the reproduction of the lexical items and grammatical structures used in
the anterior discourse”. Different kinds of discourse presentation bring with them
different assumptions about their faithfulness to the content of the original discourse.
With Direct presentation, the assumption is that the words in the quotation marks
are quoted faithfully. However, this assumption can vary depending on the type of
discourse being presented (speech, writing, thought) and the context of the presenting
situation. Quoting someone’s spoken words is quite hard to do from memory. Indeed,
to be accurate you would need to make a recording of the speech and then listen back
to it very carefully, and possibly transcribe it. As far as we know, this is not normal
everyday practice, so if you quote someone during a naturally occurring conversation,
it is likely that your interlocutor will understand that your direct presentation of
someone else’s words might not be fully accurate because you are quoting from
memory. However, when we see or hear direct presentation of speech in, for example,
a news report, we expect the news producer to have taken steps (such as making a
transcription from a recording) to ensure an accurate rendering of the other person’s
words. Our assumption, then, is that in ‘high stakes’ contexts (e.g. news, police
reports, academic writing) more care will have been taken to render the original
discourse faithfully. A key contextual difference between (for example) everyday
conversation and news report is that it is unlikely to be a matter of great concern if you
do not quote entirely accurately what your friend said to you, but there could be serious
consequences, even legal ones, if you misquote someone in a news report (and other
contexts such as academic writing).
The faithfulness associated with indirect presentation is less than that with direct
presentation simply because there is no claim being made that what is presented is
verbatim. However, there is still an expectation that the propositional content is
faithful to the original. This sort of faithfulness can be more diffcult to assess and,
because the person who is doing the presenting manipulates the original by, for
example, changing some words or missing some words out, we can never be sure how
close we are to the original unless we have access to it. Once again, we might make
assumptions about faithfulness claims based on who is doing the presenting. For
example, we might assume that news outlets are more diligent than conversational
partners, and we might trust some news outlets more than others. It is worth
remembering, though, that any presentation of any kind, including direct forms, is
mediated by the text producer because the text producer can choose (i) what to quote
and (ii) the surrounding co-text that frames the quote. Activity 10.5 invites you to look
at a real-life example of DP where faithfulness was an issue.

246
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

Activity 10.5 Fake news!

‘Fake news’ is defned by Young-Brown (2020: 5) simply as “any news story that contains false
information on purpose”. Similarly, Allcott and Gentzkow (2017: 213) describe it as “news
articles that are intentionally and verifably false, and could mislead readers”. These defnitions
highlight that fake news is false information that pretends to be true and that it is intentional.
The term ‘fake news’ was brought to the fore by Donald Trump during his presidential campaign
and eventual reign and emerged as a key sociopolitical phrase8 with a meaning broader than the
previous defnition suggests. Farkas and Schou (2018), for example, suggest that the term is used
by left-wing liberals to critique right-wing politics and media outlets and then by the right-wing
(especially Trump himself) to discredit mainstream media. They write that prior to becoming
president, “Trump had insinuated that the term ‘fake news’ was a political construct created in
order to attack and delegitimise his presidency”. They go on to say that once in offce:

he began what would become a continuous and highly systematic use of the “fake news” [. .
.] to attack and delegitimise what he saw as his direct opponents: mainstream media. Trump
started using the term to lash out at media companies [. . .]
(Farkas and Schou 2018: 306)

Bearing in mind the view of Farkas and Schou, look at the tweet below by Donald Trump and the
reporting of it on the Washington Post and CNN websites.

On 29 October 2018 Donald Trump, in the role of US President, tweeted the following (sentences
numbered for ease of reference).

247
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

(1) There is great anger in our Country caused in part by inaccurate, and even fraudulent,
reporting of the news. (2) The Fake News Media, the true Enemy of the People, must stop
the open & obvious hostility & report the news accurately & fairly. (3) That will do much to
put out the fame . . . of Anger and Outrage and we will then be able to bring all sides together
in Peace and Harmony. (4) Fake News Must End!

In the news, on the same day, the Washington Post printed this on its website:

President Trump lashed out anew Monday at the news media, calling it “the true Enemy of
the People,” and he again blamed what he called “fraudulent” reporting for anger [. . .]
(https://tinyurl.com/mr3dkrdd)

NBC, on its news web pages, posted a story about the tweet in which they said the following.

Trump used Twitter Monday morning to rip the media as “the true enemy of the people” and
blame the press for “great anger” in the U.S.
(https://tinyurl.com/z7h9keu3)

Now, consider whether the news reports were faithful to the original discourse. Give reasons
for your conclusions. Following that, consider what, if anything, analysing discourse
presentation and faithfulness can contribute to the discussion of ‘fake news’.

Conclusion
In this chapter we have discussed in detail the idea of discourse presentation and that
it introduces other voices into a discourse. Discourse presentation can provide other
opinions and views and the help to legitimate and validate your own ideas. Our main
aim in this chapter was to introduce a framework for the analysis of discourse
presentation that will enable you, as discourse analysts, to identify different types of
discourse presentation in texts. This is an important step in any analysis before further
evaluative and interpretative steps can be made. We have also discussed how the way
in which the discourse is framed by the person presenting it can affect how the
original, anterior discourse is viewed by hearers and readers. This is especially true of
minimal discourse presentation categories (e.g. Presentation of Speech), but also
where reporting clauses (and other signals) include subjective and evaluative
language.

Further reading
Semino and Short (2004) is a book length discussion using corpus data of the SW&TP
framework used in this chapter.

248
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

For discussion of the application of the same framework to spoken data see
McIntyre et al. (2004), and to historical data see McIntyre and Walker (2011, 2012)
and Walker and McIntyre (2015).
Fairclough (2003: Chapter 3), who discusses discourse presentation from a Critical
Discourse Analysis perspective, relates reported speech to the notion of intertextuality.
Walker and Karpenko-Seccombe (2017) provide an analysis of a BBC news report
using Short’s (2012) version of the Leech and Short framework.

Answers to activities
Activity 10.1 Transpose Direct Speech into Indirect Speech “What
are you still doing here at this time?” Jules asked.

Some options that we came up with are:

(i) Jules asked Ginny what she was still doing there at that time.
(ii) Jules wondered what Ginny was still doing there at that time.
(ii) Jules questioned Ginny about still being there at that time.
(iv) Jules gently quizzed Ginny about her continued presence there at that ungodly
hour.

Notice that in options (i) and (ii) ‘are’ shifts in tense to ‘was’. In option (i) the pronoun
‘you’ changes to ‘she’ while in (ii) and (iii) Ginny’s name is used. In all options the
deixis changes to be appropriate to the posterior discourse so ‘here’ becomes ‘there’
and ‘this’ becomes ‘that’. In (iv), we have taken more liberties with the paraphrasing
of the original and converted ‘still doing here’ to ‘her continued presence there’, and
‘at this time’ to ‘at this ungodly hour’. Readers may like to think about whether (iv)
remains faithful to the original.

Activity 10.2 Discourse presentation in naturally occurring

conversation
Hopefully, you will have noticed that the extract largely involves Fred presenting part
of a conversation he had with someone else at the factory (his line manager) and
contains some of the words apparently spoken by the manager as well as some of
Fred’s own words in response. We can spot the conversational turns that are from an
earlier discourse because they are introduced by verb phrases such as ‘he comes and
says’, ‘he goes’, and ‘I go’. If you listen to the spoken transcript (available at the Santa
Barbara website), you will hear that Fred also signals that he is presenting his
manager’s words by impersonating his manager’s voice. You may have also noticed
that in the last orthographic sentence of the extract, Fred appears to tell Richard what
he was thinking during his conversation with his manager. His thoughts, which are
apparently ‘aw man, this is the pits man’, are introduced with ‘I’m just going’. While

249
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

‘going’ is not specifcally a verb of thought, and ‘go’ is used earlier in the extract to
introduce speech, we reasoned that (i) it is unlikely that Fred said these words out loud
to his manager, and (ii) the progressive aspect which encodes ongoing action suggests
the thoughts are happening during the spoken conversation or at least while the
manager is speaking.

Activity 10.3 Josh gives his opinion

‘Waded in’ is used metaphorically (and idiomatically) to mean intervene. We can
wade into a debate, meaning that we intervene to give our two pen’orth9 (another
idiom meaning give one’s opinion). In this headline, then, it seems likely that ‘waded
in’ refers to some sort of discourse produced by Josh Hartnett and is therefore an
example of Presentation of Speech (PS).
To use ‘waded in’ to refer to someone giving their opinion has rather negative
connotations because it sounds like they are intervening forcefully and possibly
uninvited. It possibly makes Josh look like he is ‘sticking his nose in’ where it’s not
wanted. To assess whether the Daily Mirror is offering us a true and fair rendering of
Josh’s contribution, we need to see the original, anterior discourse. We also need to
appreciate the context. Josh was being interviewed about his new TV series, and the
interview followed on from an item concerning the controversial story concerning
parties being held at 10 Downing Street during a nationwide Covid-19 lockdown in
the UK. After exchanging greetings, the interview starts by Susanna asking Josh
whether he has been gripped by what is happening at in the UK government and
whether those involve in the parties should be afraid. Josh replies as follows.

Josh: [H] ((laughs)) how can you not ↘be [Hx] er it’s it’s all that’s on
the ↘news (.) er (.) yeah I mean it’s of ↑course↑ they should be afraid (.) this (.)
this doesn’t look ↘good people were people were having a very diffcult time at
that ↘time [H] (.) it’s (.) it’s hard to believe that they ↘were (. .) living it ↗↘up
[1.0]
The intonation contour and pause suggests that Josh has fnished what he had to say
and that this is a Transition Relevant Place (TRP) in the conversation. Susanna takes
the foor at this point, saying that the parties have “appalled” and “shocked” everyone.
She then moves on to ask Josh about his new TV series (which is the main purpose of
the interview). Did Josh wade in, or was his opinion solicited by Susanna? We think
that he was asked for his opinion, and he gave it, so ‘waded in’ seems to us like a
misrepresentation of the original discourse.

Activity 10.4 Discourse presentation from news story on

the BBC website
(A) 1 2

250
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

The Mary Rose Trust said it was “appalled” by Starkey’s comments,

adding on Thursday evening they had accepted his resignation.
3 4
The presentation in (A) has the following structure:

• reporting clause (1), which contains the past tense form of the verb ‘say’;
• subordinated clause (2), which contains the propositional content of what was said,
with the pronoun it and a past tense form of the verb ‘be’ aligned to the news report
(the posterior discourse);
• another reporting clause (3), which contains the -ing form of the verb ‘add’;
• another subordinated clause (4), which contains more propositional content with the
pronoun they and the past-tense form of the verb ‘have’ aligned to the posterior
discourse.

The structure of the sentence, along with the choice of pronouns and verb tense in the
subordinated clauses, makes it an example of indirect discourse presentation. The fact
that the written words from a tweet are being presented, makes it Indirect Writing
(IW). Notice, though, that the word ‘appalled’ is placed in quotation marks indicating
that the IW also contains the Direct presentation of one of the words used by the Trust.
This phenomenon of embedding direct quotation within indirect presentation is not
unusual in news reports and highlights (in this case at least) a word that defnitely was
used in the original discourse.
We can also assess whether the indirect presentation in the news report offers a
faithful rendering of the propositional content of the Twitter message. One difference
between the original tweet (anterior discourse) and the presentation of it in the news
report (posterior discourse) is that the Trust says that they ‘were appalled to hear’
while the news report says that the Trust ‘was appalled by’. Whether that difference is
important (i.e. whether the original statement and the reported statement amount to the
same thing) is a matter for conjecture.
Activity 10.5 Fake news!
To assess faithfulness, we need to assess lexical choices made by the news outlets
when they are reporting Trump’s verbal actions:
In the NBC extract: ‘rip’, ‘blame’.
In the Washington Post extract: ‘lashed out’; ‘calling’, ‘blamed’, ‘called’.
First, we can consider whether Trump’s tweet can be characterised as ‘lashing out’.
Whatever we decide, we need to appreciate that the characterisation is imposed on us
by those presenting the words of Trump; it is their choice of words for depicting the
verbal action that frame the presentation of what Trump said.
Both news outlets quote ‘the true enemy of the people’; additionally, the
Washington Post quotes ‘fraudulent’, while NBC quotes ‘great anger’. The quoted
words are defnitely used in the original, so in that sense they are faithful. However,

251
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

Trump equates ‘the enemy of the people’ with ‘the fake news media’. He does this via
what is known as apposition (we introduced this in Chapter 2). This is where two
words or phrases (in this case, two noun phrases) are placed side by side, separated by
a comma, and refer to the same entity (i.e. they have the same referent). By doing so,
the two noun phrases form a relationship of equivalence whereby ‘the fake news
media’ = ‘the true enemy of the people’.10 However, in the news reports, the label of
‘the true enemy of the people’ is assigned to ‘news media’ by the Washington Post,
and just the ‘media’ by NBC, so both miss out the word ‘fake’. This omission is
important because it suggests that Trump was attacking all the media and not just those
outlets that produce fake news (of course, one might wonder which news outlets
Trump had in mind when he tweeted his message). To present Trump’s use of
apposition, the Washington Post uses the verb form ‘calling’, while NBC uses ‘rip’
(the meaning of which is probably something like ‘verbally attack’ or ‘use verbal
aggression’). These verb forms are suggestive of a particular speech act (or in this
case, writing act) that might take the form of ‘X is (a/the) Y’ (e.g. you are the enemy).
This is effectively what Trump is doing, but via apposition, which is (perhaps) a less
obvious way of creating equivalence than an ‘X is Y’ structure.
Finally, both reports use a form of the verb ‘blame’. This presents Trump’s claim
of causation (the causative structure ‘caused in part’) and we would need to decide
whether forging a partial causative link between states and actions (anger caused by
fake news) counts as allocating blame in the original Trump tweet. Additionally, NBC
say that Trump blames ‘the press’, which he does not; rather he suggests a partial
(‘caused in part’) link between anger and ‘inaccurate’ and ‘fraudulent’ ‘reporting of
the news’. So, NBC’s report is not completely faithful. This example also shows the
importance of context – it is not enough to take the exact words ‘x, y, z’ and quote
them faithfully as ‘the true Enemy of the People’, ‘fraudulent’, and so on, as both
sources have done here, but to render them accurately in the context of utterance.
Faithfulness therefore also pertains to the immediate textual context of the presented
discourse.
The example helps to show how faithfulness to the original speech writing (or
thought) is a complex issue and can be quite diffcult to assess. We can also see that
faithfulness is important in news discourse, as it is in other discourses such as political
and scientifc discourses.

Notes
1 Also relevant is Bakhtin’s (1981) concept of heteroglossia (‘different tongues/languages’). 2 The
report of the Iraq Inquiry Executive summary is available at: https://assets.publishing.
service.gov.uk/government/uploads/system/uploads/attachment_data/fle/535407/The_
Report_of_the_Iraq_Inquiry_-_Executive_Summary.pdf and is used in accordance with the Open
Government licence, a copy of which can be found at: https://www.nationalarchives.
gov.uk/doc/open-government-licence/version/3/
3 We introduce and use this corpus in our next chapter.
252
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

4 Even though we prefer the term ‘presentation’, the terms ‘reporting clause’ and ‘reported clause’
are well established and widely used, so we will continue to use them here.
5 Our view on PTA/PWA/PTA diverges from that of Semino and Short (2004) and aligns with
Thompson’s (1996) category of ‘summary’.
6 If ‘called for’ is analysed as a phrasal verb, then the content of the original discourse is
summarised in a noun phrase beginning with ‘the’.
7 Aikhenvald (2004) points out that evidentiality simply refers to the source of information and
has nothing to do with probability, reliability, commitment to truth, or providing proof, which
are to do with modality. Aikhenvald explains that evidentially and modality are often
misunderstood and overlapped in ways that are confusing.
8 For discussion of sociopolitical keywords, see Jeffries and Walker (2017).
9 Short for two pence worth.
10 For more on textually constructed relationships of equivalence, see Jeffries (2009:
51–65).

References
Aikhenvald, Alexandra Y. (2004) Evidentiality. Oxford: Oxford University Press.
Allcott, H. and Gentzkow, M. (2017) ‘Social media and fake news in the 2016
election’. Journal of Economic Perspectives, 31(2): 211–236.
Bakhtin, M. M. (1981) The dialogic imagination: Four essays, ed. M. Holquist, trans.
C. Emerson and M. Holquist. Austin, TX: University of Texas
Press.
Clark, H. and Gerrig, R. (1990) ‘Quotations as demonstrations’. Language, 66:
764–805. doi: 10.2307/414729
Fairclough, N. (2003) Analysing discourse. London: Routledge
Farkas, J. and Schou, J. (2018) ‘Fake news as a foating signifer: Hegemony,
antagonism and the politics of falsehood’. Javnost – The Public, 25(3): 298–314.
doi: 10.1080/13183222.2018.1463047
Halliday, M. A. K. and Matthiessen, C. (2014) An introduction to functional grammar
(4th edition). London: Routledge.
Jeffries, L. (2009) Critical stylistics: The power of English. Basingstoke: Palgrave
Macmillan.
Jeffries, L. and Walker, B. (2017) Keywords in the press: The New Labour years.
London: Bloomsbury.
Leech, G. and Short, M. (2007) Style in fction: A linguistic introduction to English
fctional prose (2nd edition). London and New York: Longman.
McIntyre, D., Bellard-Thomson, C., Heywood, J., McEnery, A., Semino, E., and
Short, M. (2004) ‘Investigating the presentation of speech, writing and thought in
spoken British English: A corpus-based approach’. ICAME Journal, 28: 49–76.
McIntyre, D. and Walker, B. (2011) ‘A corpus based approach to discourse
presentation in Early Modern English writing’. The International Journal of
Corpus Linguistics, 16(1): 101–130.
253
 PRESENTING OTHER PEOPLE’S SPEECH, WRITING, AND THOUGHT

McIntyre, D. and Walker, B. (2012) ‘Annotating a corpus of Early Modern

English writing for categories of discourse presentation’. In F. Manzano (ed.),
Unité et Diversité de la linguistique, pp. 87–107. Les Cahiers du Centre d’Etudes
Linguistiques. Lyon: Atelier intégré de publication de l’Université Jean Moulin –
Lyon 3.
McIntyre, D., Bellard-Thomson, C., Heywood, J., McEnery, A., Semino, E., and
Short, M. (2004) ‘Investigating the presentation of speech, writing and thought in
spoken British English: A corpus-based approach’. ICAME Journal, 28: 49–76.
Semino, E. and Short, M. (2004) Corpus stylistics: Speech, writing and thought
presentation in a corpus of English writing. London: Routledge.
Short, M. (1988) ‘Speech presentation, the novel and the press’. In W. van Peer (ed.),
The taming of the text, pp. 61–81. London: Routledge
Short, M. (1996) Exploring the language of poems, plays and prose. Harlow:
Longman.
Short, M. (2012) ‘Discourse presentation of speech (and writing but not thought)
summary’. Language and Literature, 21(1): 18–32.
Short, M., Semino, E., and Wynne, M. (2002) ‘Revisiting the notion of faithfulness in
discourse presentation using a corpus approach’. Language and Literature, 11(4):
325–355.
Thompson, G. (1996) ‘Voices in the text: Discourse perspectives on language reports’.
Applied Linguistics, 17(4): 501–530. van Leeuwen, T. (2008) Discourse and practice:
New tools for critical discourse analysis. Oxford: Oxford University Press.
Walker, B. and Karpenko-Seccombe, T. (2017) ‘Speech presentation and summary in
the BBC News online coverage of a Russian TV interview with Vladimir Putin’.
CADAAD Journal, 9(2): 79–96.
Walker, B. and McIntyre, D. (2015) ‘Thinking about the news: Thought presentation
in Early Modern English’. In P. Baker and T. McEnery (eds), Corpora and
discourse studies. Basingstoke: Palgrave.
Young-Brown, F. (2020) Fake news and propaganda. New York: Cavendish Square.

254
 Corpus linguistics and discourse
11 analysis

Introduction
Corpus linguistics is the study of language using computers, specialised software, and
large bodies of electronic data known as corpora (which is the plural of corpus).
Corpus linguistic approaches and tools can help us to explore linguistic patterns in
large samples of language data in ways that would be impossible to do manually. Since
its beginnings in the 1960s, when computing technology started to become readily
available, corpus linguistics has been enhancing our understanding of language and
how it works. Discourse analysis using corpora is an exciting and developing area of
study that is becoming increasingly important for informing our understanding of
discourse. In this chapter, we will introduce some of the basic ideas, methods, and
approaches from corpus linguistics.

Corpus linguistics
We will start by explaining what a corpus is, and then go on to describe what is meant
by corpus linguistics.

What is a corpus?
A corpus (which is Latin for ‘body’) is a collection of naturally occurring language
data that has been selected to be representative of a particular type of discourse or
language variety. Corpora can be very large containing hundreds of thousands,
millions, or even billions of words. Many ready-made corpora are available to use for
research or, depending on what it is you want to investigate, you can create your own
corpora. Corpora are analysed using specially designed computer software, and so
corpus data must be stored electronically on a computer in a machine-readable format.
That means, for example, that while a collection of VHS video tapes1 containing
cookery programmes from the 1990s holds quite a large sample of language data, it is
not a corpus in the sense used in corpus linguistics because the spoken words on the
tapes are not in the form of computer readable fles. To turn such VHS relics into a
corpus, the spoken

DOI: 10.4324/9781003351207-11
discourses (typically between presenter and viewer) that form a major part of the
programmes would need to be transcribed onto a computer. In the same way, scanning
written texts (e.g. magazines) and saving them as, say, JPEG or PDF fles does not
make a corpus because these sorts of fle format do not render the language data (the
255
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

words) machine-readable. To create a corpus, then, the words from the original
discourse need to be transcribed and stored on a computer in machine-readable
(usually plain text) format.

What is corpus linguistics?

Corpus linguistics is the study of language using corpora and specially designed
computer tools. According to Biber et al. (1998), corpus linguistics is also empirical,
meaning that it is interested in actual language patterns observed in data (and not what
you think the language patterns might be or should be) and requires both quantitative
and qualitative analysis. So, corpus linguistics is not just about counting language
features, which computers are very good at, but also looking at the computer-
generated results, and using analytical frameworks or models from other areas of
linguistics to make sense of the results. Like all computergenerated results (whether
they are x-rays, graphs, or word frequencies), they require human interpretation,
which means that corpus linguistic approaches do not dispense with the linguist!

Corpus linguistics and discourse analysis

Combining discourse analysis with corpus linguistics involves:

a) getting a big sample of discourse data onto a computer and using computer tools
to analyse that data to fnd patterns that tell us something about the language use
within the sample;
b) using the analytical frameworks from discourse analysis to help analyse and make
sense of those results and patterns.

Corpus linguistic approaches (including building corpora and using corpus tools) can
help to analyse a lot more discourse data than would be possible manually and help to
explore patterns in the data that might provide an insight into the discourse(s) under
investigation. The effectiveness and success of corpus linguistic approaches depends
largely, in the frst instance, on the quality of your data and there are some important
principles that need to be considered (see Chapter 12 for more about data collection
generally). While these principles relate to any data collection in any research and are
not just to do with corpora, sampling, and representativeness have become
watchwords in corpus linguistics. In the following sections we will discuss these
important ideas in more detail.
Sampling, representativeness, and language variety
We have established that corpora tend to be defned as a representative sample of a
language variety. Within this defnition there are three important concepts that need
further consideration: language variety, sampling, and representativeness.
256
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

Language variety
In the early days of corpus linguistics, researchers aimed to represent national
languages (for example, British English), or periods of languages (for example, Early
Modern British English). However, over the years, the scope of corpus linguistics has
broadened, and corpora now also contain more constrained language varieties such as
the language of a particular genre (e.g. print news media) or even the output of a single
author (e.g. the novels of Charles Dickens).
As discourse analysts, we can think of a corpus as representing a discourse or
collection of discourses which can be spoken or written. For example, we might have
a corpus which contains transcribed spoken discourse consisting of samples of spoken
interactions from a variety of different contexts. A corpus might also aim to represent
a more narrowly defned discourse and contain, say, spoken discourse from just one
context. For example, the Hansard Corpus at EnglishCorpora.org contains transcribed
spoken interactions between politicians in the UK houses of parliament (the House of
Commons and the House of Lords) and is therefore constrained by several factors
including location and membership (only elected members may speak). Defning what
counts as the discourse you are investigating is an important step in any research
project, not least corpus research.

Sampling
Depending on the language variety or discourse you are investigating, sampling is
extremely important. Say, for example, you are investigating email interactions, it
would be practically impossible to collect every email ever written and sent in the
world. Consequently, no matter how big your corpus, it could never contain every
email. This is where sampling comes in. The idea with sampling, in theory at least, is
to collect enough examples of a discourse type to represent the entirety of that
discourse. So, if we are interested in studying emails, we would collect enough emails
to represent the entirety of email discourse. The entirety of whatever phenomenon is
being studied is usually referred to as the population, and as you might imagine,
sampling requires knowledge of the population you are sampling from (for example,
what sorts of emails are out there) as well as careful planning and a systematic
approach to data collection. Figure 11.1 aims to show this idea using shapes. The large
circle depicts the whole population, but within that population there is variety (shown
by different shapes) which might be to do with sociocontextual factors such as age,
gender, purpose, domain (e.g. business, education, social). Ideally, if we want to create
a sample that represents the whole population, we need to collect examples of all the
shapes that make up the population (Sample A). This would require knowledge of the
population and what it comprises. We would need to decide how many examples of
each different shape we needed in our sample and whether the proportions of shapes
in our sample refect the proportions in the population (e.g. if squares are the most
frequent shape in the population, do we have more squares than any other shape in our
sample?). Sample B in Figure 11.1 contains only one shape and, furthermore, the

257
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

shape is the least frequent in the entire population. This is an extreme example, but
one that might occur if, for example, the researcher’s knowledge of the population
extended only as far as their own discourse community and their data collection was
limited by their knowledge of the population. The sample therefore only represents
the discourse community from which it was drawn, and any fndings are not
representative of the wider population.

Figure 11.1 Sampling

Suppose we wanted to create a corpus of emails that represented the whole

population of emails, we would have to think about the range and sort of emails we
would need to collect. By that we mean things like the language the emails are written
in (e.g. English, French, Japanese), the social situation in which the emails are created
(e.g. at work between colleagues), and the purpose of the email (e.g. workplace
admin). We might choose to collect only emails that are written in English,2 that refect
three different social situations and purposes: personal ones between friends and
family, professional ones between colleagues (or between employer and employee),
and business emails between customers and clients. We might decide to exclude (for
reasons we would need to make clear) advertising and spam email. These choices,
which are based on our knowledge of emails, become the parameters which defne (and
inevitably restrict) our data and our corpus. We would also need to think about how
many emails to collect and from whom, whether to collect whole email messages or
extracts, and whether to collect equal amounts of email data from the different
contexts. We would, of course, also need to obtain the consent of the email producers
to use their emails for data (we discuss consent and ethics in Chapter 12).
Conversely, to take another example, if we were interested in examining witness
statements from just one particular incident using corpus techniques, then our corpus
258
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

would be much more constrained. That is because there would be a fnite number of
witness statements and once collected no more would be produced. So, depending on
the size of the case and the number of witness statements collected, the corpus could
easily hold every single example of that discourse.

Representativeness
The idea of sampling and using our knowledge of the language population to produce
an informed sample, is that we want a corpus that represents the whole of the
population. In Figure 11.1, Sample A represents the population more successfully than
Sample B, because the former contains a better spread of different shapes from the
entire population whereas the latter is far more restricted containing only one shape.
Creating a corpus that aims to be a representative sample is important because if we
want to make generalisations about the population from which our sample was drawn
then they will only be reliable if our corpus is a fair and representative sample of the
population. We will return to this point in the sections that follow below.

Corpus not always required!

It is worth acknowledging at this point that we do not need to build a corpus to use
corpus linguistic methods and tools. We can explore any group of texts or a single text
using corpus tools. Whatever way we choose to use corpus tools, we always need to
be clear about our research aims and aware that any results, while possibly indicative
of wider trends, might not be generalisable beyond the data.

Using corpus methods to analyse corpora

In this section, we introduce the main analytical methods used in corpus linguistics
that are typically offered by corpus software.
Word frequency analysis
Word frequency analysis is a useful quantitative observation we can make about
discourse data that shows the frequencies of all the different words in a corpus. Corpus
tools tend to present word frequencies in two ways:

• raw frequency – shows the number of observed occurrences of each word in the
corpus;
• relative frequency – shows the number of occurrences of each word divided by the
total number of words in the entire corpus. The result of this calculation can be a
very small number and typically, depending on the size of the corpus, it is multiplied
by 100 to give a percentage (%), by 1,000 to give a per thousand frequency (‰), or
even by a million to give a per million frequency (/mill).
259
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

Word frequencies and language change across time

Frequency analysis can be used to investigate lexical change across time (i.e.
diachronic change). We are going to demonstrate this idea using a ready-made corpus
called the Corpus of Historical American English (or COHA for short). COHA is
available via a web-based interface (details are at the end of this chapter) and contains
475 million words of written American English (AmE) produced from 1820 to 2010.
Importantly, the corpus is divided into 20 subcorpora, one for each decade (see Table
11.1). This subdivision makes the corpus useful for studying language change over
time.
Imagine, now, that you are interested in the modal auxiliary verb ‘shall’ and
whether its usage has declined, risen, or stayed about the same over time (you might
have your own intuitions about this). COHA, with its almost 200 years’ worth of
written data, can provide us with some insight on this. The raw and relative frequencies
of ‘shall’ in each of the sub-corpora are shown in Table 11.1 along with the size of
each sub-corpora. The graph in Figure 11.2 plots the relative frequency of ‘shall’ for
each decade represented by the sub-corpora that make up COHA and shows a decline
in the use of ‘shall’ in AmE (as represented by COHA) over time. COHA, then,
enables us to explore diachronic lexical change. As far as ‘shall’ goes, we could
explore the frequencies of this modal verb in other corpora (such as a corpus of British
English) to see if a similar trend occurs (for corpus explorations of modal auxiliary
verbs in different corpora see, for example, Leech 2003 and Biber 2004).
Notice that in Figure 11.2 we used the relative frequencies of ‘shall’ in each
subsection of the corpus. We calculated the relative frequencies by frst dividing the
raw frequency of ‘shall’ by the number of words in the sub-corpus and then
multiplying the result by 1 million to give the frequency per million words of data.
This process is sometimes referred to as normalising the results and is necessary when
comparing frequencies from datasets of different sizes. Normalising the frequencies
takes into account the different sizes of the subcorpora by presenting frequencies
relative to the size of the data in which they occur.

Table 11.1 Raw and relative frequencies of ‘shall’ by decade in COHA

Decade Sub-corpus Size Raw Freq. Relative Freq./mill.

1820 6,981,389 8727 1250.04

1830 13,711,287 14100 1028.35
1840 15,807,047 16916 1070.16
1850 16,536,003 15405 931.60
1860 16,936,560 15381 908.15
1870 18,788,467 14764 785.80
1880 20,067,205 15942 794.43
1890 20,426,783 14818 725.42
1900 21,977,250 13422 610.72
1910 23,103,098 10893 471.50
1920 25,700,422 10325 401.74

260
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

1930 27,707,879 8343 301.11

1940 27,399,750 6960 254.02
1950 28,661,274 5759 200.93
1960 29,122,676 5850 200.87
1970 28,829,225 4474 155.19
1980 29,851,580 3588 120.19
1990 33,149,318 2863 86.37
2000 34,821,812 2133 61.25
2010 35,452,806 1624 45.81

Figure 11.2 Frequency of ‘shall’ over time in COHA

Wordlists ranked by frequency
Corpus tools can provide a list of all the different words in a corpus (known as types)
ranked by their frequency. Such wordlists can offer a way of viewing a corpus that we
could not achieve by simply reading the data which can indicate: (i) the most common
topics contained in a corpus; and (ii) the frequent grammatical structures used in the
data. Words ranked by frequency can therefore suggest a focus for further analysis.
This idea is illustrated by Activity 11.1 which requires you to consider the top 50 most
frequent words (shown in Table 11.2) in a corpus that contains 3.5 million words in
total.

Activity 11.1 The top 50 most frequent words

Before reading on, take a moment to look at the information in Table 11.2 and consider what
(if anything) the top 50 most frequent words in our mystery corpus tells you about the data.

The table is set out in three columns with the frst column showing the rank of the word (1–50),
the second column contains the word, and the third column shows the raw frequency. We say
more about the corpus and what it contains below.

261
CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

Table 11.2 Top 50 most frequent words in the mystery corpus

Rank Word Raw Freq.

1 The 202018
2 And 120495
3 A 92170
4 To 80006
5 Was 65807
6 In 56144
7 We 50196
8 I 48650
9 Of 46057
10 For 39198
11 Is 38921
12 Hotel 36648
13 It 33774
14 Room 30155
15 With 26658
16 Very 25296
17 Were 24528
18 At 23984
19 You 21998
Rank Word Raw Freq.

20 This 21817
21 But 21710
22 On 20391
23 Our 20293
24 Had 20030
25 That 18971
26 Staff 17992
27 My 17885
28 As 17789
29 Not 16564
30 Are 16297
31 Have 15221
32 London 15151
33 From 14596
34 Stay 14134
35 All 13572
36 So 12742
37 Great 12492
38 Service 12481
39 Be 12332
40 There 12323
41 n’t 12285
42 breakfast 11697
262
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

43 Which 10954
44 Good 10899
45 Would 10788
46 They 10709
47 ’s 9907
48 One 9891
49 Us 9497
50 An 9325
We can reveal that our mystery corpus contains online reviews for London hotels
downloaded from a review website. We will refer to this corpus as the London Hotels
Reviews Corpus (LHRC for short). The reviews relate to nine of the most expensive
London hotels (e.g. The Dorchester) and 11 cheaper/lower budget hotels. A
breakdown of the review data, including the number of words devoted to reviewing
each hotel, is shown in Table 11.3.
Looking back at the top 50 most frequent words in Table 11.2, the nongrammatical
words ‘hotel’, ‘room’, ‘staff’, ‘London’, ‘stay’, ‘service’, ‘breakfast’ provide clues
about the corpus data, and these words give some indication of what the reviews are
about and are suggestive of the sorts of things that provoke comment. Grammatical
words are also important for understanding more about
Table 11.3 Constituent word totals for London Hotels Review Corpus (LHRC)
Top end Cheaper end

Name words words

Bulgari 44341 Apex A 247342

Corinthia 365227 to Z 10166
Dorchester 218839 City View 2107
Lanesborough 38767 Hartley 4122
Mandarin 135053 London Guest House 28229
Rembrandt 346226 Marble Arch 2538
Savoy 725460 Mondrian 471862
Wellesley 47886 Newham 30155
45 Park Lane 30527 Rhodes 135784
Ridgemount Xenia 150200
143290
Totals 1,952,326 1,225,795
Grand Total 3,178,121
a discourse. For example, in the LHRC the ffth most frequent word is ‘was’, which
could indicate a prevalence of X was Y syntactic frame. This makes sense if we
consider the sorts of comments that are likely in a review: ‘The
hotel/breakfast/room/service was good/great/awful’, and so on. Additionally, the
pronoun ‘we’ is slightly more frequent than ‘I’, possibly suggesting that fewer
reviewers were travelling alone, and/or that reviewers write the review from the
perspective of their couple/group status. All these things are worth investigating
further yet it would have taken a very long time to assess these frequencies manually.
263
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

Further investigation of individual words usually involves looking at concordance

lines, which is the topic of our next section.

Concordance analysis
Concordances are generated by corpus linguistic software and provide a way of
investigating how a word (or phrase) in a corpus is used. A concordance is a list of
all the occurrences of a lexical item under investigation (a single word or short phrase)
along with some of its surrounding co-text. This provides a view of the word in its
linguistic environment (its co-text) and can allow us to observe patterns in the data.
Figure 11.3 shows an extract of a concordance for the word ‘was’ in the LHRC.
The fgure shows just the frst 20 concordance lines out of a possible 65,807 and they
are listed in the order in which they appear in the corpus (notice that all the occurrences
are from reviews for the same hotel – The Apex). The concordance in Figure 11.3
shows the word under investigation in the centre with some co-text either side. This
format is sometimes known as a KWIC view

264
CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

which stands for Key Word In Context. That is essentially what a concordance is.
Here, key word is the lexical item under investigation and context means some of the
co-text (or words) either side of the key word (we will introduce another type of
keyword later in the chapter). The amount of co-text is known as the concordance
window and in Figure 11.3, the window is 50 characters either side of ‘was’. Some
corpus tools allow the window size to be adjusted to suit your needs.
The 20 concordance lines in Figure 11.3 help to give a sense of how ‘was’ is used
in the LHRC. As we mentioned earlier, ‘was’ is likely to feature in X was Y
constructions, where X is the grammatical Subject, Y its Complement, and was
operates as a copular verb. In Figure 11.3 we can see that this is the case in 19 out of
the 20 lines (in line 3 ‘was’ is an auxiliary verb). We can also see in the small sample
that these clausal structures are used for evaluating things like bed, breakfast, view
from the room, and location. However, Figure 11.3 shows just a small fraction of the
65,807 occurrences of ‘was’ in our data and as analysts our job is to assess the data
systematically and rigorously rather than choose a few examples that help to support
any hypotheses we might have about the data. Any such analysis of concordances is
largely manual and so can be quite daunting, especially when faced with thousands of
lines (in our case 65,807). In the following sections, we will consider some methods
for handling the analysis of concordance lines.

Sorted concordances
Many corpus tools can sort concordances alphabetically based on the words to the left
and/or right of the search term. Alphabetical sorting of concordances lines can help
considerably with data analysis. We illustrate this in Figure 11.4 which shows the frst
20 concordance lines of ‘was’ from the LHRC with alphabetical sorting switched on.3
We can see that we get a different set of concordance lines to those shown in Figure
11.3 because now they have been ordered differently (recall that in Figure 11.3 we
saw the frst 20 concordance lines for ‘was’ without any sorting). The alphabetical re-
ordering has pushed to the top of the list a different set of 20 occurrences of ‘was’ out
of the total 65,807 – notice that the flenames in the ‘File’ column show that the
occurrences are from different fles, and none are from reviews of the Apex Hotel.
Sorting, then, does not alter the data, it just changes the way we see it.
If you look carefully at Figure 11.4, you will see that the occurrences of ‘was’ are
ordered, in the frst instance, alphabetically by the frst word following ‘was’ (i.e. one
word to the right). So, we can see all the instances of ‘was’ when it is followed by
‘awful’, then ‘awkward’, then ‘awoken’ and fnally ‘back’. The concordances also
incorporate a second level of sorting which places these results alphabetically using
the second word following ‘was’ (i.e. two words to the right). So, for example, if you
look at concordance lines 3 to 11 you

266
CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

will see that ‘awkward’ is followed by ‘and’ on line 3 (the open bracket is ignored),
‘especially’ on line 4, ‘for’ on line 5, and so on. Alphabetical sorting can expose
patterns of usage of words in corpus data and in the next section we will explore a
possible analytical ‘next step’ where we assign to categories any patterns we fnd.

Categorising results
It is often useful to categorise the language feature under investigation in some way
based on its usage or function within the concordance lines and therefore the corpus
data. Any such categories are selected or created by the analyst for the purpose of the
analysis of the data in hand (i.e. ad hoc) and are often suggested by the data itself
(sometimes referred to as data-driven). Categories need to be well defned and based
on some sort of analytical framework, and the concordance data should be treated
systematically using these categories (we talk more about systematicity in Chapter
12).
For example, if we consider the use of ‘was’ in the LHRC, one possible
classifcation we could apply is whether ‘was’ is used as a copular or auxiliary verb.
In Figure 11.4 we can see these two uses in action. Lines 11 to 15, which group
together ‘was’ followed by ‘awoken’, are examples of ‘was’ being used as an auxiliary
verb in passive constructions, where ‘awoken’ is a past participle. The other 15 lines
are examples of ‘was’ as a copular verb. These two grammatical categories relating to
the use of ‘was’ are based on linguistic description established within grammars of
English and offer a potentially useful starting point for systematically analysing
concordances of ‘was’ in the LHRC. Activity 11.2 asks you to try out applying this
categorisation to some more data from the LHRC.

Activity 11.2 Copular and auxiliary ‘was’

Below are ten short extracts from the LHRC corpus each containing an instance of ‘was’
(underlined). For every example, decide whether the underlined ‘was’ is used as a copular or
auxiliary verb. Give reasons for your answer. Are some examples harder to categorise than
others? If so, why?

(i) Our room was lovely, very chic and absolutely spotless.
(ii) The room was badly dated, a bit tattered, smelled weird, and had unsightly stains (iii)
As promised, I was called once the room became available at 14:00 (iv) Our room was
called the Beauchamp Suite.
(v) The wif was absolutely brilliant and very fast.
(vi) I was disappointed by the food though.
(vii) every restaurant in the hotel was fully booked for lunch.
(viii) The room was fully equipped with lots of unique extra touches.
(ix) Continental breakfast was included in our rate.
(x) My bed was positioned at an angle which was a little funky.

268
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

Answers are at the end of the chapter.

Other ad hoc categories we could employ in the analysis of ‘was’ include, for
example, grouping occurrences of copular ‘was’ by the grammatical part of speech
(e.g. adjective, adverb, noun) of the Subject Complement (i.e. the word or phrase that
follows ‘was’). The occurrences of copular ‘was’ followed by an adjective could then
be further analysed to look at, for example, whether the adjective is positive (e.g.
‘good’), negative (e.g. ‘awful’), or neutral (e.g. ‘big’) and/or whether there are any
patterns relating to the subjects they describe. For example, do certain grammatical
subjects (e.g. ‘the bathroom’) attract particular types of adjective, such as adjectives
of size (e.g. ‘big’). In this way, a corpus analysis frst provides a focus from which we
can carry out more detailed analysis using linguistic models to categorise and further
understand the discourse the data represents.
A concordance is an essential tool that provides a window onto the data allowing
us to explore it in ways that would not be possible manually. Concordances help to
identify patterns for further analysis, which might involve grouping occurrences of
lexical items in a principled way (perhaps informed by existing linguistic models) that
are suggested by the data itself. In the next section we move on to another essential
corpus method for data analysis that is made possible by tools: comparison.

Making comparisons between corpora

So far, we have seen how word frequencies in corpora can offer some insight into the
discourses they aim to represent. However, to know whether such frequencies are
‘normal’ or peculiar to the data under investigation and therefore noteworthy, we need
a reference point against which we can compare them. For example, in our corpus of
hotel reviews (LHRC) we saw that the words ‘the’ and ‘and’ were the top two most
frequent words (see Table 11.2), but we do not know whether these frequency
rankings are unique to written hotel reviews or normal in written discourse generally.
To help fnd out, we need to compare the fndings from the LHRC (our research
corpus) with word frequencies from another corpus (a reference corpus).
There are numerous ready-made corpora that can act as a comparator against which
word frequencies can be evaluated. Some of these corpora can be downloaded
(sometimes for a fee) while others are only available via webinterfaces that allow you
to interrogate them but not download the data. COHA, the corpus we met earlier in
the chapter, is one such example (we list others at the end of this chapter). If a suitable
ready-made reference corpus is not available, then you can build your own, provided
you have access to the data you need. Deciding which reference corpus to use is a
matter of balancing what is practical and what available data fts best as a point of
reference.

269
CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

Choosing a suitable reference corpus

Comparing a research corpus against a reference corpus helps to highlight potentially
distinctive language features in the research data and the discourse it represents. The
features that are highlighted will depend to some extent on the choice of reference
corpus, which will be driven by a combination of research aims and what is practical
and possible. The reference corpus aims to act as a norm or yardstick against which
the research corpus is evaluated.
When choosing a reference corpus, it is often a good idea to match contextual
aspects of the corpus data and the populations they represent. To help with this it can
be useful to create a list of discourse parameters (see Chapter 1) that relate to your
data. For example:

Event Leaving a review for a hotel.

Time 21st century (2012).
Purpose Reviewing; informing other potential customers; complaining;
commending.
Participants Reviewers (everyday customers) and readers of reviews (potential
customers); hoteliers.
Code Different varieties of English depending on nationality of reviewer.
Medium Electronic web-based forum. Mode
Written.

These parameters are variables which potentially affect language choices. We can
consider which parameters, or variables, we want to investigate in relation to language
choices. So, say we are interested in the effect purpose and medium has on language
choices in the data, we would aim to build a reference corpus that matches the other
parameters. Matching discourse parameters in the data is one way of controlling
variables that might affect language use, and this can help to reduce ‘noise’ in the
results. By ‘noise’ we mean results that are not connected to the focus of the research
aims, which is to fnd out about language features associated with online hotel reviews.
So, for example, if we were to use a corpus of spoken discourse for our reference
point, then we are likely to get results that highlight difference that relate to written
versus spoken discourse. Similarly, if we were to compare our hotel reviews against
a corpus of 18th-century British English (BrE), we would fnd out something about the
differences between 21st-century and 18th-century BrE. The idea then is to try to
match discourse parameters that do not relate to your research interests.
We aimed to match the mode of the discourse under investigation (written) and the
language (code). Because the hotel reviews in the LHRC were posted by a variety of
English speakers including those whose frst language is not English, we built a
reference corpus that contains examples of many different English varieties from
around the world (e.g. UK English, US English, Indian English, etc.). To achieve this
breadth of variety (which we hoped would go some way to match the breadth of

270
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

variety in our corpus), we combined several small corpora to form our reference
corpus.
We used ten corpora from the International Corpus of English (ICE) family of
corpora, each of which have a similar structure and aim to represent different varieties
of English from around the world. The corpora are around 1 million words each and
contain both (transcribed) spoken and written data. Our reference corpus contains the
written sections from those ICE corpora that are freely available for download: ICE
Canada, East Africa, Hong Kong, India, Ireland, Jamaica, Nigeria, Philippines,
Singapore, and USA. As well as these corpora, we also used the 1-million-word
Freiburg, Lancaster, Oslo, Bergen (FLOB) corpus of written British English. The
resulting super-corpus of 11 different varieties of English (hereafter ICE-FLOB4)
contains approximately 5.8 million words. Table 11.4 provides summary details.

Table 11.4 ICE-FLOB reference corpus

Constituent corpus No. of words

ICE Canada written 435170

ICE East Africa written 802341
ICE Hong Kong written 498924
ICE India written 438694
ICE Ireland written 424186
ICE Jamaica written 415198
ICE Nigeria written 386405
ICE Philippines written 452198
ICE Singapore written 436339
ICE USA written 435155
FLOB 1128063
Total 5, 852, 673

Comparing LHRC against ICE-FLOB

Table 11.5 sets out the top 20 most frequent words in LHRC and those in ICEFLOB.
Take a moment to look at the information in the table and consider whether the
comparison helps to offer further insights into the reviews data. The frequencies in
Table 11.4 are per 1,000 words (denoted by the symbol ‰), which was calculated by
dividing the raw frequency by the total number of words in the corpus and multiplying
the result by 1,000. This makes the frequencies from the two corpora (which differ in
size) comparable.
Note: Choosing the frst 20 words as a cut-off is purely a matter of convenience; we
could have decided just as easily to show the frst 23 or 127 words. The choice then is
completely arbitrary because it is not made on any principled basis other than it is a
round number that fts easily onto the page and gives enough information for us to talk
about.
If we assume that ICE-FLOB provides a reference point for international varieties
of written English including British English, then comparing the ranks and word

271
CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

frequencies of the words can help to show what language features are particular to the
hotel reviews in our data. Looking at the information in Table 11.5

Table 11.5 Top 20 words by frequency in LHRC and ICE-FLOB

LHRC ICE-FLOB

Word Freq.
Word
Rank Freq ‰ ‰
1 the 63.45 the 62.16
2 and 37.85 of 32.56
3 a 28.95 to 26.53
4 to 25.13 and 26.46
5 was 20.67 a 22.17
6 in 17.63 in 20.18
7 we 15.77 is 11.34
8 I 15.28 that 9.92
9 of 14.47 for 9.65
10 for 12.31 it 7.79
11 is 12.23 I 7.34
12 hotel 11.51 on 6.83
13 it 10.61 as 6.78
14 room 9.47 s 6.65
15 with 8.37 was 6.58
16 very 7.95 be 6.39
17 were 7.70 with 6.33
18 at 7.53 are 5.52
19 you 6.91 by 5.16
20 this 6.85 he 4.80
tells us that ‘the’ is the most frequent word in both corpora (and the relative
frequencies are very similar), so the LHRC data are not special in having this as the
most frequent word. However, ‘and’ has a noticeably higher ranking and relative
frequency in the LHRC list, and this could be indicative of a syntactic style prevalent
in our online reviews. We can also confrm that the relative frequency of ‘was’, which
we mentioned above, is considerably higher in LHRC, which might also be indicative
of a recurring syntactic structure. The comparison also highlights that (unsurprisingly)
‘hotel’ and ‘room’ have a high ranking in the LHRC data. Additionally, we can see
that in the review data ‘very’ is the 16th most frequent word, whereas in the ICE-
FLOB it is ranked 104 (so not shown in the table) with a relative frequency of around
one occurrence per 1,000 words.
The comparison between two datasets where one is used as a point of reference can
be useful in highlighting lexical and possibly syntactic features that are peculiar to a
discourse and can offer a good starting point for further investigation. The next step

272
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

in an analysis is to investigate these lexical items within the data to establish how they
are used and whether their usage is distinctive of that discourse. This next stage is
achieved (certainly in the frst instance) using concordance lines.

Keyness and keywords

In the previous section, we made manual comparisons between word frequencies in
the LHRC (our research corpus) and ICE-FLOB (our reference corpus). By using
ICE-FLOB as a comparator, we were able to observe differences in word frequencies
between our review data and a corpus of general international English and use our
own judgement to decide whether the differences were salient and worth further
investigation. However, a substantial problem with this method of judging saliency is
that it is subjective. A further and equally sizeable problem is that we only evaluated
differences for the 20 most frequent word types and not all the word types in the
corpus.
A more principled and systematic way of comparing frequencies and deciding
which differences are noteworthy is to use statistical testing to calculate what are
known as keywords. Keywords are those words that are over-represented5 in the
research corpus by comparison with a reference corpus and are determined by
comparing word frequencies using statistical tests. Many corpus tools will compute
keywords automatically using either statistical signifcance or effect size calculations,
or a combination of both.

Statistical significance
Simply put, statistical signifcance refers to results that are assessed as important or
noteworthy using statistical tests. Results that are statistically signifcant are said NOT
to have happened by chance but have been caused by some factor that is particular to
the population from which the sample was drawn. Just what that factor might be and
whether it is peculiar to the population is for the analyst to investigate. Statistically
signifcant results, then, are those highlighted as being potentially important and
worthy of further investigation. Keywords are those words in a corpus whose
frequencies are found to be signifcantly different from the corresponding word
frequencies in the reference corpus.
Statistical signifcance depends on what is known as hypothesis testing. Recall that
at the start of this chapter we discussed the idea of corpora being a representative
sample from an entire population. When we calculate keywords, we are making a
comparison between two corpora and therefore (if we assume the corpora are
representative samples) two populations. Hypothesis testing relates back to the
populations from which the corpora were sampled and a default assumption that there
is no difference between them. Consequently, any differences in word frequencies
observed in the corpora are assumed to be due to sampling error or chance. This

273
CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

default assumption is called the null hypothesis and is denoted by H0. The alternative
hypothesis, which is denoted by H1, is that any observed differences in the corpus
data do refect a true or real difference in the populations. To summarise:

H0: There is no difference between two populations the data came from, and any
observed differences in the data are due to sampling error or chance rather
than a true difference between the populations.
H1: There is a true difference between the two populations, so observed differences
in the data relate to differences in the population.

The idea of hypothesis testing is that the null hypothesis holds true unless there is
suffciently strong evidence against it to reject it. This is where signifcance testing
comes in. A signifcance test, of which there are many to choose from, is a
mathematical calculation that provides us with evidence by computing the probability
of getting a result if the null hypothesis were true. If the probability is high, then that
is evidence to accept the null hypothesis and reject the alternative. However, if the
calculation returns a low probability, then that is evidence to reject the null hypothesis
and accept the alternative hypothesis. When we do that, we are saying that the result
is statistically signifcant. To summarise:
High probability that the result would have been obtained if H0 were true = Evidence
to accept H0 and reject H1; result not signifcant.
Low probability that the result would have been obtained if H0 were true = Evidence
to reject H0 and accept H1; result is signifcant.
If we relate this idea back to our LHRC and ICE-FLOB corpora, the default
assumption is that there is no difference between the two populations from which
these two corpora were drawn. Population here refers to the discourses that the two
corpora represent, and difference refers to persistent patterns of language use within
the discourses. So, H0 assumes there is no pervading difference in language use
between the online review discourse and the various discourses in ICE-FLOB. If we
reject H0 then we make an inference about the populations – that they are different in
relation to persistent language choices. We might also make an inference about the
contextual variables that infuence such language choices (such as medium, purpose,
producer/receiver). These inferences might become additional hypotheses that are
explored further. For instance, we might hypothesise that the language is infuenced
by some nonrandom factor in the discourse such as being online, or by the purpose of
reviewing.
Statistical tests provide a numerical indication of the probability of the results being
obtained if the null hypothesis were true. The probability is suggestive of the
confdence with which we can reject the default assumption of there being no
difference. Later in this section, we will discuss what level of probability counts as
suffcient evidence to confdently decide whether to reject the null hypothesis or not.

274
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

A popular signifcance test used in linguistics for calculating keywords is

loglikelihood (Dunning 1993). This test was designed especially for text analysis and
corpus research and is the test we will refer to over the rest of our discussion of
keywords. With log-likelihood, the higher the number returned from the calculation,
the lower the probability that the result would have been obtained if H0 were true.
Results with a high log-likelihood (LL) are therefore potentially statistically
signifcant, due to differences in the populations, and worthy of further investigation.
Statistical signifcance tests help to establish that a difference exists between a
research corpus and a reference corpus. They do not, however, give an indication of
the scale of the difference between observed results. To do that we need to use effect
size calculations, which we will look at next.

Effect size: measuring the scale of the difference

While statistical signifcance tests indicate the confdence with which we can say that
our results are not a chance happening, they do not tell us whether a difference is big
or small. Consequently, keywords investigations increasingly incorporate a second
statistic known as effect size because this measures the scale of the difference between
results (known as the effect). Whereas signifcance testing is a type of inferential
statistic because it allows us to make inferences about the population from which a
representative sample was drawn, effect size is a descriptive statistic which means it
simply describes the data. The basic idea with effect size is that small differences are
less important and therefore less worthy of further investigation than big differences.
As with signifcance testing, there are many different measures of effect size to
choose from, and each one returns a value that indicates the scale of difference
between results. Log ratio, for example, was developed by Andrew Hardie especially
for keyword research (see Hardie 2014). With log ratio, a value of 1 means that the
frequency in the research corpus is twice as big as in the reference corpus; a log ratio
of 2 means that the difference is four times as much, and so on (see Table 11.6).
Therefore, the higher the number returned by the log ratio calculation, the bigger the
scale of difference.

Table 11.6 Scale of frequency difference between research and reference corpora
based on log ratio of relative frequencies (after Hardie 2014)
Log ratio of relative Frequency in research corpus compared with
frequencies Reference corpus
0 Frequency in research and reference corpora is the
same
1 2 times more frequent in research corpus
2 4 times more frequent " " "
3 8 times more frequent " " "

275
CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

4 12 times more frequent " " "

5 16 times more frequent " " "
... ...

Keyness
Associated with keywords is the notion of keyness which, according to Scott and
Tribble (2006), is “a quality words may have in a given text or set of texts, suggesting
that they are important” (Scott and Tribble 2006: 55–56).6 Important words for
discourse analysts are those which are indicative of language use that is peculiar to
the discourse under investigation and therefore symptomatic of a contextual variable
associated with that discourse. However, as Scott and Tribble state, keyness only
suggests importance and so offers a starting point for further investigation of
particular words in a corpus. So, keywords are potentially important, and their
importance is indicated in the frst instance by statistical tests but then evaluated by
further investigation. We will return to this point later in this chapter.
What counts as a keyword? Using statistical cut-offs to
decide
In combination, signifcance testing and effect size offer a principled way in which to
establish which words in your data are key. Fortunately, corpus tools will carry out
the calculations for us, so we do not need to do them manually (we do not have space
here to show the formulae, but see ‘Further reading’ below). Statistical tests and effect
size calculations both return numerical results, but we still need to know which values
count as enough evidence against the null hypothesis and therefore offer enough
confdence for us to reject it. For this we require what are known as critical values.
The result from a statistical test can be seen from two points of view; they indicate:

( i) the probability that the result would have occurred if the null hypothesis were true;
( ii) the level of confdence you have that a result is signifcant.

The lower the probability of getting the result if H0 is true, the more confdently we
can infer the result is due to a true difference in the population from which the data
was drawn. To summarise: low probability that the result would high confdence that
we can assume
= have been obtained if H0
were true the result is signifcant.

The analyst decides on the level of confdence at which they are happy to reject the
null hypothesis. The critical value is the line you draw in your results below which
you say ‘I do not have suffcient evidence to reject the null hypothesis, so these results

276
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

are not signifcant’, and above which you say ‘I do have suffcient evidence to reject
the null hypothesis so these results are signifcant’. The idea of a critical value, then,
is that anything that falls below a chosen value is ignored; anything on or above the
value is accepted as a keyword. Different statistical calculations have different critical
values that relate to particular levels of confdence. The critical values for log-
likelihood (LL) are shown in Table 11.7. The table shows that an LL score of at least
15.13 means that the probability of obtaining that result if H0 were true (i.e. no
difference in the populations from which the data were drawn) is very low at 0.01%,
which means we can be 99.99% confdent that the result is signifcant. A result with a
LL score of at least 10.82 brings with it slightly less confdence of signifcance at
99.9%.
Table 11.7 Log-likelihood critical values (based on Rayson et al. 2004)
Log-likelihood Probability of getting result Level of confdence that
critical value if H0 true (%) result is signifcant (%)

3.84 5 95
6.63 1 99
10.82 0.1 99.9
15.13 0.01 99.99

When calculating keywords, the usual practice is to set a level of signifcance below
which results are ignored. This means setting a minimum statistical score (in our case
log-likelihood) for the results. With log-likelihood, if you want to be very confdent
about your results, then you set the minimum value high at, say, 15.13. Any results
returned with a LL score lower than that are ignored.
Keywords can also be calculated using effect size calculation such as log ratio.
However, there are no agreed critical values for this statistic and here researchers must
rely on their own judgements. We choose to ignore any keywords with a log ratio of
less than 1 which means we only look at keywords that are at least as twice as frequent
in the research corpus as they are in the reference corpus. Applying this cut-off
therefore eliminates keywords where the scale of difference is very small.
NOTE: Keywords can also be restricted using frequency cut-offs, which means
deciding on a minimum raw frequency in the research corpus below which keywords
are eliminated. This is rather different from using signifcance results because it has
nothing to do with hypothesis testing, probability, or confdence. Instead, researchers
must decide whether limiting results using frequency is useful for answering their
research questions. For example, keywords that have a low raw frequency in the
research corpus might be eliminated because they are unlikely to form patterns across
the whole corpus and therefore unlikely to be indicative of language patterns in the
discourse the corpus is representing.

Keywords in LHRC
Let’s now return to the comparison between frequencies from LHRC and ICEFLOB.
We carried out a keyword analysis of the LHRC and, using AntConc (Anthony 2022),

277
CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

automatically compared all the words and their frequencies in the LHRC corpus
against the corresponding words and frequencies in the ICE-FLOB corpus. We
delimited our results using the following cut-offs: loglikelihood of 10.83 (so, our level
of confdence in the differences is at least 99.9%), log ratio of 1 (so, keywords that are
at least twice as frequent as those in the reference corpus), and raw frequency of 100,
which we judged to be suffcient to observe meaningful patterns in the data. The
comparison produced 1,028 keywords, which is a lot of keywords to assess. By assess
we mean establish whether a keyword is symptomatic of discourse under investigation
and tells us something about the linguistic nature of the data. This is a crucial step in
any keyword analysis because while a difference might be statistically signifcant, it
does not necessarily mean that it is telling us something important about the discourse
we are investigating. We therefore aim to work out whether a keyword is not just
statistically signifcant but also interpretatively signifcant.
Ideally, then, we need to work through each keyword in the list systematically
looking at their co-text in detail using concordance lines. However, this can be
timeconsuming, and when there are thousands of keywords, impractical. Therefore,
we will discuss methods for dealing with large numbers of keywords, which we will
do using just the frst 20 keywords in the LHRC for illustrative purposes. Table 11.8
shows the frst 20 keywords listed in order of their log likelihood (LL) score.

Table 11.8 Keywords in LHRC when compared against ICE-FLOB

‰ LL

LHRC
Freq. ‰
36647 11.51 533 0.09 71694.5 7.0
2 room 30155 9.47 1637 0.28 51680.2 5.1 3 we 50196 15.77 17130 2.93 43680.0 2.4 4 was
65807 20.67 39764 6.79 32488.4 1.6
5 staff 17992 5.65 1343 0.23 29051.8 4.6
6 very 25296 7.95 6054 1.03 27419.7 2.9
7 london 15150 4.76 832 0.14 25872.0 5.1
8 stay 14134 4.44 751 0.13 24262.2 5.1
9 breakfast 11697 3.67 161 0.03 22894.5 7.1
10 stayed 8913 2.80 205 0.04 16848.9 6.3
11 rooms 8915 2.80 266 0.05 16459.0 5.9
12 great 12492 3.92 2132 0.36 15820.1 3.4
13 service 12481 3.92 2129 0.36 15809.5 3.4
14 our 20293 6.37 7925 1.35 15804.1 2.2 15 location 8635
2.71 348 0.06 15407.4 5.5
16 bar 7883 2.48 285 0.05 14252.1 5.7 17 i 48650
15.28 41442 7.08 13389.7 1.1

278
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

18 savoy 5790 1.82 5 0.00 12024.2 11.1 19 friendly 6444

2.02 208 0.04 11797.3 5.8
20 lovely 5984 1.88 172 0.03 11085.2 6.0

In Table 11.6, ‰ = per thousand frequency, LL = log-likelihood, and LR = log ratio

Like concordances, placing keywords into groups is often a useful starting point
for an analysis. Groupings can be driven (in the frst instance) by the grammatical class
of the keywords. For example, looking at our top 20 keywords, three are pronouns
(‘we’, ‘I’, and ‘our’) and these grammatical words are suggestive of the personal
nature of reviews with the reviewer being present within the review. Nine of the
keywords are common nouns (‘hotel’, ‘room’, ‘staff’, ‘stay’, ‘breakfast’, ‘rooms’,
‘service’, ‘location’, ‘bar’) and these indicate the discourse topic of the data and
potentially the focus of many of the comments in the reviews data. For this reason,
such keywords are sometimes referred to as ‘aboutness’ (Phillips 1989) keywords (see
Scott 2000; Scott and Tribble 2006). There are also two proper nouns (‘London’ and
‘Savoy’), both of which are indicative of the content of the corpus. It is usual for
proper nouns to be keywords in any corpus comparison and often they are of low
interpretative value. Three of the keywords are adjectives (‘great’, ‘friendly’, ‘lovely’)
as well as one adverb (‘very’) and these are indicative of the evaluative nature (or
purpose) of the discourse that the corpus represents. That three positive adjectives are
in the top 20 keywords could suggest that the reviews in the corpus contain more
positive evaluations than negative or that there is greater lexical variety used for
negative evaluation therefore reducing the frequencies of individual negatively
evaluative words. Finally, there are two verbs: ‘was’, which is indicative of repeated
syntactic structures within the data; and the lexical verb ‘stayed’, which is another
example of an aboutness keyword.
Keywords might also suggest categories that are semantic rather than grammatical.
For instance, we could form groups such as ‘food and drink’, ‘room, fxtures, and
fttings’, ‘facilities’, and so on depending on what other keywords are in the list. We
could also have categories such as ‘positive evaluation’ and ‘negative evaluation’.
Whatever we decide, placing a keyword into a category requires investigation of the
keyword in its surrounding co-text using concordance lines. Even when assigning
keywords to grammatical categories this is required. For example, before categorising
‘service’ we ascertained that in the data it is always used as a noun rather than a verb,
often as part of a two-word phrase ‘room service’.

279
CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

Intra-corpus comparison
As well as utilising a reference corpus for comparison, it can sometime be useful to
carry out intra-corpus comparisons, whereby different sections of the same corpus are
compared against each other. Whether this is possible or useful will depend in the
structure of the corpus under investigation. With our data, we could split the LHRC
on the basis of the price/star-rating of the hotels with nine being ‘top-end’ and the
remaining 11 being ‘cheaper-end’. Intra-corpus keyword comparisons are carried out
in much the same way as described above.
We compared the nine top-end hotel reviews against those of the cheaperend hotels,
and vice versa to obtain two sets of keywords. These are shown in Activity 11.3, which
asks you to guess which keywords belong to which group.
Activity 11.3 Comparing subsets of the London Hotels Reviews Corpus (LHRC)

Which set of keywords do you think are from the reviews of top-end hotels and which from the
cheaper-end hotel reviews? Give reasons for your answer.

List A: clean, small, modern, friendly, views, very, room, great, walk, tube, helpful, rooftop,
breakfast, shower, balcony, owners, owner, bus, good, nice, on, prosecco, free, foor, beans,
close, defnitely, stairs, location

List B: afternoon, tea, service, pool, butler, spa, experience, deco, beautiful, concierge,
sandwiches, luxury, foyer, suite, scones, world, the, best, cakes, dinner, grand, truly, wonderful,
history, treat, elegant, pianist, doormen, class, fowers

Once you know which keywords belong to which subcorpus, consider whether there is anything
that is consistent in top end/cheaper-end hotel reviews?

Our answers are at the end of the chapter.

Collocation
We met collocates earlier in the book in Chapter 3. Collocates are those words that
tend to occur near to the word under investigation (sometimes referred to as the node
word; see Sinclair et al. 2004: 10). The term ‘collocation’ is usually attributed to J.
R. Firth (1957) and relates to the co-occurrence relationships that words are involved
in. Different discourses and different discourse topics will produce different patterns
of collocation for a word, and corpora and corpus tools can help to explore these
patterns.

Calculating collocates
Collocates are calculated within a collocation window, which is a span of words
either side of the node word. The window can usually be set in the computer tool
being used and differently sized windows can be chosen to suit your analytical focus.
Sinclair et al. (2004: 35) suggest the optimum window is four words before and after

280
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

the node word which, when the node word is included, gives a nine-word collocation
window.
Collocates of words can be assessed using raw frequencies. So, for example, Table
11.9 shows the top ten collocates (out of over 2,500) of ‘was’ in the LHRC within a
nine-word window (i.e. four words either side of the node word).
We can see in Table 11.9 that the most frequent collocate of ‘was’ is ‘the’, co-
occurring a total of 42,871 times: 27,033 to the left of ‘was’; 15,838 to the right. To
illustrate left and right co-occurrence of ‘the’ with ‘was’, Figure 11.5 shows two
examples from the corpus. In the fgure, the collocation window is indicated by a
rectangle and any word that occurs within that window counts as a collocate.
Frequency can help us to assess the salience of collocates for a word. However, as
Table 11.9 shows, many of the high-frequency collocates are grammatical
Table 11.9 The most frequent collocates of ‘was’, 4L-4R
Rank Collocate Total Freq. Freq. Left Freq.
Right
1 the 42871 27033 15838
2 and 24746 11120 13626
3 a 14405 3159 11246
4 room 12156 10341 1815
5 it 11943 9985 1958
6 i 9189 5685 3504
7 to 8175 2670 5505
8 very 7797 1180 6617
9 in 7533 4094 3439
10 was 6380 3190 3190

Figure 11.5 Examples of ‘the’ as right- and left-hand collocates of ‘was’

words which are also likely to collocate with lots of other words. That means we
cannot be sure whether these co-occurrences are peculiar to ‘was’ or happen with lots
of other words in the data. For that reason, other statistical measures can help us to
assess whether a collocate of the node word is noteworthy or not.
Some statistics, such as log-likelihood, compute the confdence with which we can
assume that there is an association between two words. Other statistics evaluate the
strength of association between two words (i.e. the node and its collocates). For
example, mutual information (MI) is a statistic common to many corpus tools and
assesses strength of attraction between words by looking at the other company the
node word and the collocate in question keep within the corpus (see Clear 1993: 280;

281
CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

Manning and Schütze 1999: 182). So, while ‘the’ and ‘was’ may occur together a lot
in the LHRC, we know that both ‘the’ and ‘was’ frequently hang around with lots of
other words. So, in order to assess whether ‘the’ and ‘was’ have a particularly close
association, the co-occurrence of these two words is compared against their co-
occurrence with the other words in the corpus.
Church and Hanks (1990) recommend using a combination of statistics to calculate
collocates. First, a measure of strength of association (e.g. MI) and then a measure of
the confdence with which we can assume that there is an association (e.g. log-
likelihood). This method helps to eliminate low frequency collocates which are
potentially chance co-occurrences.
Using statistical cut-offs to decide what counts as a collocate
As with keywords, the usual practice with collocates is to set statistical levels below
which results are ignored. For example, if MI is used, the lowest score that is
indicative of a strong association is 3 (see Stubbs 1996; Hunston 2002). For log-
likelihood any of the critical values shown in Table 11.7 can be used. Minimum raw
frequencies can also be used on the basis that low frequency collocates might not be
indicative of patterns that run through the whole corpus and are therefore less
important as a feature of the discourse under investigation.

Statistically salient collocates of ‘was’ in LHRC

Using AntConc, we calculated collocates for ‘was’ again, but this time using statistical
measures. AntConc uses combination of MI and log-likelihood (the LL critical value
is pre-set to 3.84). So, MI is calculated frst and then LL, and any collocates that do
not achieve an LL score of at least 3.84 are eliminated. We also applied an MI cut-off
of three and a minimum raw frequency cut-off of 20 to eliminate low frequency
collocates. These settings produced a list of 917 collocates. Table 11.10 shows the frst
ten collocates of ‘was’ (out of 917) in order of frequency.

Table 11.10 The most frequent 4L-4R collocates of ‘was’ with an MI above 3 and an
LL above 3.84
Rank Collocate Total Freq. Left Freq. Right Freq. MI

1 the 42871 27033 15838 3.36

2 and 24746 11120 13626 3.31
3 room 12156 10341 1815 4.28
4 it 11943 9985 1958 4.09
5 i 9189 5685 3504 3.19
6 very 7797 1180 6617 3.90
7 our 5726 4009 1717 3.77
8 but 5400 2619 2781 3.59
9 breakfast 4930 4107 823 4.35

282
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

10 that 4369 2778 1591 3.48

We have listed the collocates in Table 11.10 in order of frequency for ease of
comparison with Table 11.9. Using the information shown in Table 11.10 we can see
that ‘the’ and ‘and’ are not only the most frequent collocates but are also attracted to
‘was’ because they have an MI score that is above the minimum required. Also,
because AntConc uses LL to assess confdence of association, we can be reasonably
confdent that there is an association between ‘was’ and these words. However, the
highly frequent collocates ‘a’, ‘to’, ‘in’, and ‘was’ that were in Table 11.7 do not
appear in Table 11.8 because these are not statistically salient when assessed using a
combination of MI and LL. Instead, ‘but’, ‘breakfast’, and ‘that’ have now moved into
the top ten. The collocates shown in Table 11.10 are therefore both high in frequency
and have statistical salience so we can be reasonably confdent that these associations
are not chance happenings. As we mentioned earlier, high-frequency collocates are
likely to be indicative of patterns of language use that extend across much of the
corpus.
Once collocates are calculated (of which there might be many), manual effort is
required to make sense of the results, usually achieved by returning to concordance
lines. As we discussed in the section above, any such manual analysis might involve
creating data-driven categories, which can help to ascertain collocational tendencies
and patterns.

N-grams
N-grams (sometimes referred to as lexical bundles or multi-word expressions) are
contiguous sequences of words where a single word is a 1-gram, a two-word sequence
is a 2-gram, a three-word sequence is a 3-gram; and so on. N-grams are calculated by
corpus tools, which can rapidly search for repeated sequences of n words (imagine
trying to do that manually!). Typically, the number of words in a sequence (the ‘n’) is
specifed by the analyst, and results can be fltered based on frequency and range, which
is the spread of n-grams across the fles that make up a corpus. So, a minimum
frequency and range can be specifed which will instruct the corpus tool to disregard:
(i) low frequency n-grams which are likely to be of limited use for discourse analysis;
and (ii) n-grams that appear in only a small number of the fles that make up the corpus.
N-grams have been studied extensively in both spoken and written discourses by
corpus linguists (see, for example, Biber and Conrad 1999; Cortes 2004), with Biber
and Barbieri (2007: 270) describing them as “important building blocks in discourse”.

N-grams in the LHRC

Using AntConc, we calculated 3-grams (i.e. three-word sequences) in the LHRC with
a minimum frequency of 20 and a range of at least 15 out of 20 fles, meaning that the
3-gram had to appear in at least 75% of the fles that make up the corpus for it to be
included in the results. This produced a list of 1,143 3-grams, the top ten of which are
shown in Table 11.11. Four of these 3-grams (the frst, ffth, seventh, and tenth) have a

283
CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

range of 20 so appear in all of the corpus fles; the rest have a range of either 17, 18,
or 19, so are well spread across the data.
The 3-grams shown in Table 11.11 reiterate the importance of clauses that contain
a form of the verb ‘be’ in the data, as well as the (unsurprising) foci of the reviews
(‘room’, ‘hotel’, ‘staff’). Concordances can be used to explore n-grams further. For
example, looking at the frst 3-gram in Table 11.11 we found the most frequent word
immediately to the left of ‘the room was’ was ‘very’, and the most frequent words
following that were ‘clean’, ‘comfortable’, ‘small’, or ‘spacious’.

Table 11.11 The top 20 3-grams in the LHRC

Rank 3-gram Freq. Range

1 the room was 3422 20

2 the hotel is 2977 19
3 one of the 2122 17
4 we had a 2090 18
5 the staff were 2036 20
6 of the hotel 1924 19
7 in the room 1738 20
8 the rooms are 1646 19
9 stayed at the 1548 18
10 our room was 1477 20
Note that the higher the value of ‘n’ in an n-gram, the fewer n-grams that are
returned in a search. So, for example, using the same parameters we searched for 4-
grams in the LHRC and this returned just 156 results, none of which occurred across
the full range of the corpus.
We do not have space here to discuss our n-grams further, but we can observe that
sequences of words can provide yet another window on our data that would have been
impossible to obtain manually. Each n-gram also provides a ‘way in’ to a large dataset
providing a focus for further, more detailed analysis.

Conclusion
Rather than a sideline followed by only those analysts interested in computers, corpus
linguistic approaches are coming more and more to the forefront of Discourse
Analysis. This is because there are several advantages that come with using corpus
data, including the following.

• Corpus data gives access to quantitative information about language, which can
often be used to support qualitative analysis.
• Insights into language gained from corpus analysis are often generalisable in a way
that insights gained from the qualitative analysis of small samples of data are not.
284
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

• Using corpora forces us to acknowledge how language is really used (which is often
different from how we think it is used).
• Exploring large datasets using corpus tools can be enlightening and open new
avenues of research.
Computer tools provide a quantitative way into large amounts of data as well as the
means to explore the data qualitatively. However, the analyst needs to fnd ways to
assess the results in a systematic and rigorous way to establish which results are
important for the analysis of discourse. Often there are too many results to be assessed
so the analyst needs to establish principled ways in which to reduce the data or make
selections that do not amount to ‘cherry-picking’. That means integrating linguistic
models and frameworks in corpus analysis to make sense of the computer-generated
results.
This chapter has provided only a brief snapshot of some of the analytical
approaches possible using corpora and corpus tools. We limited our discussion to one
online corpus (COHA) and one corpus tool (AntConc) as these are freely available
and reasonably straightforward to use.There are many other tools avail- able, of course
and we encourage readers to explore the possibilities of discourse analysis using
corpus linguistic approaches. We provide some examples below.

Further reading
Read more about choosing a suitable reference corpus in Scott (2009).

Readers interested in knowing how log-likelihood is calculated can fnd details on the
Lancaster University website at this URL: http://ucrel.lancs.ac.uk/llwizard.
html
Tip: type ‘log likelihood wizard’ into your search engine.

An explanation of the log ratio calculation can be found on Andrew Hardie’s

blog: http://cass.lancs.ac.uk/log-ratio-an-informal-introduction/ Tip:
type ‘Andrew Hardie log ratio’ into your search engine

For an overview of the statistical measures we have introduced, see also McIntyre and
Walker (2019), chapter 5.

285
CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

Corpus tools
AntConc (Anthony 2022) – Free-to-download software for corpus analysis:
http://www.laurenceanthony.net/software.html
Wmatrix (Rayson 2009) – Online corpus annotation and analysis tool (requires
registration): https://ucrel.lancs.ac.uk/wmatrix
Wordsmith Tools (Scott 2022) – The corpus analysis tool that made the notion of
keywords popular (available to download for a small fee): https://
lexically.net/wordsmith
Sketch Engine (Kilgarriff et al. 2014) – A powerful online corpus analysis tool which
allows you to upload your corpus to a server and carryout various quantitative
analyses, including those we have described in this chapter.
https://www.sketchengine.eu/

Corpora
British National Corpus (BNC)
100 million words of written and spoken British English from a variety of discourse
situations from the early 1990s. Freely available with restrictions via BNCWeb
(hosted by Lancaster University, UK at http://bncweb.lancs.ac.uk/
bncwebSignup/user/login.php). Also, at English-Corpora.org (see below).

Brown Family of corpora

• Brown – 1 million words of 1960s written American English
• LOB (Lancaster-Oslo-Bergen) – 1 million words of 1960s written British English
• FROWN (Freiburg-Brown) – 1 million words of 1990s written American English
• FLOB (Freiburg-LOB) – 1 million words of 1990s written British English

All available to purchase as part of the ICAME Corpus Collection (http://clu.

uni.no/icame/newcd.htm) and freely available with restrictions via CQPweb (see
below).

CQPweb (Corpus Query Processor)

Online concordancer hosted by Lancaster University, UK that gives access to
numerous corpora including English corpora (e.g. BNC, Brown Family) and South
and East Asian language corpora (e.g. Thai National Corpus): https://
cqpweb.lancs.ac.uk

286
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

English-Corpora.org
Online concordancer (https://www.english-corpora.org) that gives access to numerous
very large corpora including:

• Corpus of Contemporary American English (COCA) – 560 million words of written

American English from 1990 to 2017
• Global Web-Based English (GloWbE) – 1.9 billion words of textual varieties of
English from 20 different countries
• News on the Web (NOW) – Monitor corpus of over 6 billion words of webbased
newspapers and magazines from 2010 to the present time; updated daily

International Corpus of English (ICE)

A collection of parallel corpora of different Englishes worldwide, for comparative
studies. Some of the corpora are available to download for free (under licence).
Further details are at the project website: http://ice-corpora.net/ice/.

Log-Likelihood and effect size calculator

Log-Likelihood wizard (http://ucrel.lancs.ac.uk/llwizard.html)

The LHRC
We prepared the LHRC using a dataset built by PromptCloud. The data was
downloaded from Kaggle and used under a creative commons license CC BY-SA 4.0
(https://creativecommons.org/licenses/by-sa/4.0/). For the purposes of writing this
chapter, we edited the data to remove non-English contributions.
https://www.kaggle.com/datasets/PromptCloudHQ/reviews-of-londonbased- hotels

Answers to activities
Activity 11.2 Copular and auxiliary ‘was’
In many of the examples it is not straightforward to decide whether ‘was’ is acting as
a copular or as an auxiliary. These are the examples where ‘was’ is followed by a
participle form of a verb. This is because we tend to say ‘was’ + past participle of a
verb = passive construction (in which case, ‘was’ is an auxiliary verb). However,
participle forms of verbs can also function as adjectives (known as participle
adjectives). For example, compare ‘the motivated donkey’ and ‘the donkey was
motivated (by a carrot)’; ‘motivated’ is adjectival in the former but a verb in the latter.

287
CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

So, not everything that looks like a passive construction is actually . . . a passive
construction. So, when analysing ‘was’ + past participle, the issue is differentiating
between participle forms when used as a verb (e.g. ‘I was called’) and when used as
an adjective (e.g. ‘the hotel was fully booked’). In other words, we need to decide
whether the participle is acting as a verb, or acting as an adjective. If it is acting as a
verb, then the ‘was’ that accompanies it is auxiliary, and if it is acting as an adjective,
we can say that the ‘was’ is copular.
Examples (i), (iii), and (v) are reasonably straightforward to analyse:
(i) Our room was lovely, very chic and absolutely spotless. Copular
(iii) As promised, I was called once the room became available at 14:00
Auxilliary
(v) The wif was absolutely brilliant and very fast. Copular

In both (i) and (v) ‘was’ is copular with a ‘standard’ adjective in Complement position;
in (iii) ‘was’ is auxiliary forming a passive construction with the past participle
‘called’ (‘I was called’ presumably by someone, telephonically).
The other examples pose more of a problem and, when analysing, it can help to
keep in mind the following:

• participle form is acting as an adjective when describing a state of being (e.g. the
room was dated);
• participle acting as a verb when describing actions (e.g. I was called).

So, if ‘was’ + participle describes an action, we can say that ‘was’ is auxiliary.
However, to make things complicated, ‘was’ + participle acting as a verb can also
describe states of being! (e.g. ‘I was disappointed’). To help out, Huddleston and
Pullum (2002) provide some further tests that can be used to identify how participle
forms of verbs are being used. For verbal usage:

V1. Can a by-phrase that identifes the agent of the action be inserted into the
example without signifcantly altering the meaning? V2. Can an active version of
the example be constructed?

For adjectival usage:

A1. Can the words ‘too’ or ‘very’ be inserted without signifcantly changing the
meaning,
A2. Can ‘was’ be replaced with verbs like ‘seemed’, ‘appeared’, ‘looked’,
‘remained’.

Using these tests, we analysed the rest of the examples as follows (we have put the
test that helped in brackets):

288
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

(ii) The room was badly dated, a bit tattered, smelled weird, and had unsightly stains
Copular (A1 and A2)
(iv) Our room was called the Beauchamp Suite. Auxiliary (V1 ‘by the hotel’; V2 ‘the
hotel called the room the Beauchamp Suite’)
(vi) I was disappointed by the food though. Auxiliary (V1 – the by-phrase is already
there for us; V2 ‘the food disappointed me’; however, A1 ‘I was very
disappointed by the food’; but A2 ‘I seemed/appeared/looked/ remained
disappointed by the food’ changes the meaning). Our view is that the by-phrase
pushes for an auxiliary interpretation.
(vii) every restaurant in the hotel was fully booked for lunch. Copular (A2 e.g.
‘remained fully booked’); cannot construct an active alternative ‘*X fully
booked every restaurant for lunch’)
(viii) The room was fully equipped with lots of unique extra touches. Copular (A2
e.g. ‘seemed fully equipped’)
(ix) Continental breakfast was included in our rate. Auxiliary (V2 ‘our rate included
breakfast’/ ‘the hotel included breakfast in our rate’)
(x) My bed was positioned at an angle which was a little funky. Auxiliary (V1 ‘the
bed was positioned at an angle by the hotel’/ and V2 ‘the hotel positioned the
bed at an angle’; BUT A2? Does ‘the bed seemed/appeared/ looked positioned
at an angle’ carry the same meaning? We felt it did not, so opted for auxiliary.

This activity highlights that notionally straightforward categories are sometimes

anything but and need to be informed by existing theoretical approaches and research.
While some of the interpretations seemed counterintuitive, we were guided by the
analytical framework set out by Huddleston and Pullum (2002). In this way, a
framework can help produce a consistent analysis.

Activity 11.3 Comparing subsets of the hotel review corpus

List B is from the top-end hotels.The keywords that perhaps indicate top-endness are
words that may be perceived to have higher social class connotations (see Chapter 5):
‘pool, ‘butler’, ‘spa’, ‘concierge’, ‘suite’, and ‘pianist’. ‘History’ might also be an
indicator since the top-end hotels tend to have a long history.
The consistencies between the two lists relate to the class of keywords rather than
the words themselves. Both lists contain adjectives – indicating that both sets of
reviews evaluate – and reference to food and meals. The difference is, of course, in
the adjectives actually used (e.g. ‘good’/‘great’ vs ‘best’; ‘nice’ vs ‘beautiful’) and the
types of food (‘tea’/‘scones’ vs ‘beans’). The inconsistencies relate to, for example,
the cheaper-end keywords indicating that proximity to/types of transport is/are
mentioned a lot (e.g.‘walk’,‘tube’,‘bus’). Each of these keywords would need to be
investigated using concordance lines to check out these intuitions.

289
CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

Notes
1 Video tapes are ancient artefacts for recording television programmes pre-internet and Smart TV
era.
2 There are many varieties of English (British English, Australian English, American English, and
so forth) so this restraint on the data would need to be refned further in a study.
3 We used AntConc to do this with our data.
4 Yes, we also thought it sounded like an ineffectual anti-hero from an X-Men comic book.
5 Negative keywords can also be calculated, and these are words with a signifcantly low frequency
by comparison and therefore under-represented in the research corpus. We do not have space to
deal with negative keywords in this chapter, but the principle behind them is the same as
positive/over-represented keywords.
6 The notion of keyness extends beyond keywords. For example, key-clusters are statistically
signifcant strings of words, and Wmatrix (Rayson 2009) can calculate key grammatical and
semantic categories.

References
Anthony, L. (2022) AntConc (Version 4.2.0) [computer software]. Tokyo: Waseda
University. Accessed 10 July 2023. https://laurenceanthony.net/software.html
Biber, D. (2004) ‘Modal use across registers and time’. In A. Curzan and
K. Emmons (eds), Studies in the history of the English language II: Unfolding
conversations, pp. 189–216. Berlin: Mouton de Gruyter.
Biber, D. and Barbieri, F. (2007) ‘Lexical bundles in university spoken and written
registers’. English for Specifc Purposes, 26, 263–286.
Biber, D. and Conrad, S. (1999) ‘Lexical bundles in conversation and academic
prose’. In H. Hasselgard and S. Oksefjell (eds), Out of corpora, pp. 181–190.
Amsterdam: Rodopi.
Biber, D., Conrad, S. and Reppen, R. (1998) Corpus linguistics: Investigating
language structure and use. Cambridge: Cambridge University Press.
Church,K.W. and Hanks,P.(1990) ‘Word association norms,Mutual Information
and lexicography’. Computational Linguistics, 16(1): 22–29.
Clear, J. (1993) ‘From Firth principles: computational tools for the study of
collocation’. In M. Baker, G. Francis, and E. Tognini-Bonelli (eds), Text and
technology: In honour of John Sinclair, pp. 271–292. Amsterdam: John
Benjamins.
Cortes, V. (2004) ‘Lexical bundles in published and student disciplinary writing:
Examples from history and biology’. English for Specifc Purposes, 23, 397–423.
Dunning, T. (1993) ‘Accurate methods for the statistics of surprise and coincidence’,
Computational Linguistics, 19(1): 61–74.
Firth, J. R. (1957) Papers in linguistics 1934–1951. London: Oxford University Press.

290
 CORPUS LINGUISTICS AND DISCOURSE ANALYSIS

Hardie, A. (2014) ‘Log ratio: An informal introduction’. Blog post: ESRC Centre for
Corpus Approaches to Social Science (CASS). Accessed 23 March 2022.
http://cass.lancs.ac.uk/log-ratio-an-informal-introduction
Huddleston, R. D. and Pullum, G. K. (2002) The Cambridge grammar of the English
language. Cambridge: Cambridge University Press.
Hunston, S. (2002) Corpora in applied linguistics. Cambridge: Cambridge University
Press.
Kilgarriff, A., Baisa, V., Bušta, J., Jakubíček, M., Kovář, V., Michelfeit, J., Rychlý,
P., and Suchomel, V. (2014) ‘The Sketch Engine: Ten years on’. Lexicography,
1: 7–36.
Leech G. (2003) ‘Modality on the move: the English modal auxiliaries 1961–1992’.
In R. Facchinetti, F. Palmer, and M. Krug (eds), Modality in contemporary
English. Berlin: Mouton de Gruyter
Manning, C. D. and Schütze, H. (1999) Foundations of statistical natural language
processing. Cambridge, MA: MIT Press.
McIntyre, D. and Walker, B. (2019) Corpus stylistics: Theory and practice.
Edinburgh: Edinburgh University Press
Phillips, M. (1989) Lexical structure of text. Discourse Analysis Monograph 12.
Birmingham: English Language Research.
Rayson, P. (2009) Wmatrix: A web-based corpus processing environment. Computing
Department, Lancaster University. Accessed 21 April 2022.
http://ucrel.lancs.ac.uk/wmatrix
Rayson, P., Berridge, D., and Francis, B. (2004) ‘Extending the Cochran rule for the
comparison of word frequencies between corpora’. In G. Purnelle, C. Fairon, and
A. Dister (eds), Le poids des mots: Proceedings of the 7th International
Conference on Statistical analysis of textual data (JADT 2004), Vol. II, pp. 926–
936. Louvain-la-Neuve: Presses universitaires de Louvain. Conference held in
Louvain-la-Neuve, Belgium, 10–12 March 2004.
Scott, M. (2000) ‘Focussing on the text and its key words’. In L. Burnard and T.
McEnery (eds), Rethinking language pedagogy from a corpus perspective, Vol.
2, pp. 103–122. Frankfurt: Peter Lang.
Scott, M. (2009) ‘In search of a bad reference corpus’. In D. Archer (ed.), What’s in
a word-list? Investigating word frequency and keyword extraction, pp. 79–91,
Farnham, Surrey: Ashgate Publishing.
Scott, M. (2022) WordSmith Tools version 8 (64 bit version) Stroud: Lexical Analysis
Software. Accessed 10 July 2023.
Scott, M. and Tribble, C. (2006) Textual patterns: Key words and corpus analysis in
language education. Amsterdam: John Benjamins.
Sinclair, J., Jones, S., Daley, R., and Krishnamurthy, R. (2004) English collocational
studies: The OSTI report. London: Continuum.
Stubbs, M. (1996) Text and corpus analysis. Oxford: Blackwell.

291
 Doing a project in discourse
12 analysis

Introduction
In this fnal chapter, the focus is on the practicalities of doing discourse analysis. The
‘analysis’ part of discourse analysis is only part of the process because there are other
considerations when conducting an analysis of any discourse. These include how to
select and handle data, how to approach the analysis of data methodologically, what
ethical concerns need to be addressed and how to address them, and how to defne and
then refne your research idea into a question that is manageable and valid. In keeping
with the ‘practical’ focus in this book, the chapter includes some ideas for doing
research in discourse analysis.

Thinking about discourse as the focus of a project

Throughout this book, we have considered discourse as language in use that needs to
be analysed within its social, political, ideological, interpersonal, and cultural
contexts. There are an infnite amount of situational contexts in which discourse
occurs. Therefore, whatever discourse you choose to analyse, your methodological
approach must be one that is consistent and appropriate within the context you are
working in. In this chapter, we will discuss some of the issues you need to consider
when designing and executing a research project. We will consider data collection and
some of the ethical concerns involved with using other people’s discourses and then
go on to discussing how to develop a research idea and formulate research objectives
and questions. We will fnish by outlining the process of writing up your research,
which is a separate process from doing your research. We will start though with an
overarching concern that embraces a whole project from design to dissemination:
systematicity.

Systematicity and the three Rs of research

A systematic analysis is one that is organised and well-structured and deals with all
the data involved in the research in the same way. To help achieve systematicity,
Simpson (2014: 4) suggests the three Rs of doing research: rigour, retrievability, and
replicability.

DOI: 10.4324/9781003351207-12
• Rigour means being thorough and careful in your analysis and accounting for all
your data rather than simply ‘cherry-picking’ examples that help to support your
 hypothesis (if you have one). Rigour can be helped by using an analytical framework
to help structure your analysis.
• Retrievability means making explicit and explaining the framework, principles and
terminology that are used to organise your analysis. It also means making explicit
how you reached your conclusions, describing the steps you took to get from the
data to an interpretation, and providing evidence to support your claims.
Retrievability is therefore concerns being transparent about your design, data, logic
and reasoning, and the route to your conclusions. In this way, other scholars looking
at your research will not have to take you at your word because they can see for
themselves what you did.
• Replicability is achieved through retrievability – when your methodology,
framework, and terminology are retrievable, other researchers can see what you did
and how you did it, and can replicate your study. This means that, given the write-
up of your project, other scholars can replicate your research and potentially verify
your results or your method.

Say, for example, that you wish to analyse the discourse of hotel reviews using a
corpus (as we did in Chapter 11) and that you are interested in the verb-form ‘was’
when it is used as a copular verb in Relational-Intensive clauses (see the Transitivity
framework discussed in Chapter 9). As we saw in Chapter 11, a corpus approach can
quickly provide us with a list of all the occurrences of ‘was’ in a corpus. However, to
differentiate between copular ‘was’ and ‘was’ when used as an auxiliary verb, we need
some criteria (drawn from existing grammars of English) to inform our categorisation.
These criteria become our framework for analysis of ‘was’ in the data which need to
be applied systematically and rigorously to every occurrence. Describing the
categories used for the analysis, referencing the grammar that inform the criteria, and
explaining the process of working through the data makes the research retrievable
and replicable. Therefore, anyone wanting to do a similar analysis would be able to
examine the source used to establish the criteria, examine (and maybe critique) the
categories, and perform a similar analysis.

Ethics
The second important factor in discourse analysis is an ethical one. Each discourse
context may carry different constraints ranging from how freely you can access the
data to how appropriate it is to share the data publicly. Additionally, you should
consider what ramifcations exist for the producers of that data if it is studied or shared,
and what gains you receive from analysing it.
Guidelines for ethics can be found on numerous websites, such as: the British
Association of Applied Linguistics (BAAL); UK Research and Innovation (UKRI);
Professional and ethical codes for socioeconomic research in the information society
(EU): National Council on Ethics in Human Research (Canada); and the Association
of Internet Researchers (AoIR). The ESRC, which is part of UKRI, summarise what
is meant by ethical research using these six core principles:

1 research should aim to maximise beneft for individuals and society and
minimise risk and harm;
2 the rights and dignity of individuals and groups should be respected;
3 wherever possible, participation should be voluntary and appropriately
informed;
4 research should be conducted with integrity and transparency;
5 lines of responsibility and accountability should be clearly defned;
6 independence of research should be maintained and where conficts of interest
cannot be avoided they should be made explicit.
(UKRI 2021)

We can see from these principles that ethics permeates all aspects of research from
project design and data collection to the analysis and reporting of results. It also covers
how we as researchers should conduct ourselves within the research and wider
communities.
Let’s consider data in terms of ethics. Data for discourse analysis might include,
for example, naturally occurring conversations, interviews, focus group discussions,
or social media interactions. When data is collected from people, the notion of ethics,
particularly relating to ESRC/UKRI principles 2, 3, and 4 above, becomes important
because data collection of this nature is bound by ethical guidelines summarised by
BAAL:

Applied linguists should respect the rights, interests, sensitivities, privacy and
autonomy of their informants in all research contexts, including those in which
users’ rights are not so clear-cut, such as easily accessible internet sites.
(BAAL 2016)

The latter part of the quotation hints at the increasing amount of data that is gathered
from the internet, and the ethical questions that collecting and using such data can
raise (for a discussion about using the internet for research purposes, see Larner 2014).

Informed consent
BAAL states that “the cornerstone of ethical research” is informed consent. This is
the practice of obtaining consent from informants (participants) to use their discourse
as data prior to the data being collected and after they have been informed about the
research as fully as is practically possible. Informants can therefore make an informed
decision about whether they wish to participate in the research or not. Gathering
examples of discourse data from informants without their knowing – known as covert
research – is now acknowledged as ethically unacceptable. This is different from
conducting research with your participants’ consent but where you do not share full
294
details of the study’s aims because it might compromise the data collected (see the
section on ‘Observer paradox’ below).
Obtaining informed consent is reasonably straightforward with some forms of data.
For example, collecting spoken discourse from informants in focus groups requires
the analyst to select or solicit the informants in the frst place, at which point they can
be presented with all relevant information concerning the research study. This
information should include what happens after the study (e.g. how you will use the
data). However, informed consent becomes less straightforward in some situations
when collecting, for example, spoken discourse data in a public place because it is
likely to be impossible to locate all the people who might appear on the recording
whose language might potentially end up in the data. Additionally, depending on the
context, there may be extra challenges for securing informed consent because there
might be more than one party involved. For example, in forensic situations there may
be additional data protection issues that constrain data collection from host
organisations or institutions such as the police or court. If you wanted to analyse the
discourse of police interactions with witnesses, you need the informed consent of
witnesses to both observe their interaction and collect the data from that interaction,
but you will also need the consent of the police force who will want to know for what
purposes the data will be used, how it will be stored, for how long, and who has access
to it. In Appendix 2, we have included a sample document that seeks informed consent
from participants willing to take part in a study involving data collection using the
discussion group example that we outlined in Case study 12.1. Appendix 1 shows a
sample statement declaring consent for a study.
Informed consent can also be diffcult to obtain when data is downloaded from
online forums, chatrooms, and other forms of electronic social media (ESM), such as
Twitter and Facebook. For example, in this book we referenced Twitter users and we
wrote to the authors to ask for their consent to use their posts, but they did not respond.
Even though their utterances are publicly available (the account holders have public
profles and one of the tweets was subsequently picked up by the press) we felt,
nonetheless, that the right thing to do (ethically) was to inform them of our (pedagogic)
use of their tweets. This brings us to the notion of ‘implied consent’ (Martin and Knox
2000), which is where participants are offered the opportunity to say that they do not
want their contributions to be used as data, and if no reply is received then consent is
implied.
A more general point about using publicly available material is whether consent is
needed, since much of what can be found on the internet is arguably in the public
domain. The general guide here is that if the discourse data is password-protected,
then it is not for wider public consumption and cannot be used without consent. Doing
things like signing up to a chatroom and ‘lurking’ to collect data could be viewed as
covert research and a form of deception, because the researcher is posing as a ‘normal’
chatroom user. This is viewed as an ethically questionable research practice.1 In certain
online situations it might not be clear who the participants are or how to contact them,
so asking for consent is likely to be diffcult. Additionally, if there are thousands of
participants (e.g. certain Twitter feeds), then the task of asking for consent might
become extremely diffcult and practically impossible. While there appear to be no
specifc guidelines on how to deal with these sorts of problems with obtaining consent
for these kinds of internet data, generally, researchers should do what they reasonably
can to inform participants about the research and how their data will be used and give
them the opportunity to withdraw.

Observer paradox
Aside from the practicalities of gaining consent, a signifcant problem with informed
consent is what is known as the observer’s paradox (Labov 1972). This is where
discourse participants are aware that they are being studied or ‘observed’ and change
their linguistic performance because they are conscious that their language production
is ‘on show’ and under scrutiny. If such changes in performance occur in informants’
production of otherwise naturally occurring discourse, this can skew results and, at
worse, make the data unusable.
One option to help alleviate the observer’s paradox is a form of covert research
known as justifed deception. This is where some details about the specifc purpose of
the research are withheld from informants until after the data is collected. In such
situations, participants would know that their language is being studied (or some
aspect of their contributions is informing the research), but they would not know what
aspect until the data has been collected. For example, in Case study 12.1, which we
discuss below, the speech habits of students under formal and informal conditions
were studied, but the linguistic phenomenon that was being investigated was not
shared with informants prior to the data collection in case they were primed in some
way to use (or avoid) the phenomenon during the data collection period.
With justifed deception, the research is not wholly covert in the sense that we
describe above, but only partly so. The UKRI guidance addresses the issue of covert
research as follows.
covert research should only be used when no other approach is possible, where it is
crucial to the research objectives and design, or where overt observation may alter
the phenomenon being studied. The broad principle should be that covert [. . .]
research should not be undertaken lightly or routinely. It is only justifed if important
issues are being addressed and if matters of social and/or scientifc signifcance are
likely to be discovered which cannot be uncovered in other ways.
(UKRI 2021)

So, in certain special, socially signifcant (but unspecifed) circumstances, research that
does not fully disclose the aims of the study is allowable but should be handled
carefully to address the key ethical consideration set out by BAAL, quoted above.

296
 Anonymity
Also important in data collection from informants is the right to anonymity, so even
where informed consent is obtained (or implied), the data must be anonymised. This
can also raise several practicalities, not least if there is a considerable amount of data.
The extent of the anonymisation must also be considered (for example, any place or
institutional names, or other personal names mentioned) so that there is no way in
which participants can be identifed by their contributions.
In summary, there are numerous guidelines for ethical research that can be
consulted if there is any doubt about whether research practices are ethical or not,
particularly around data collection from informants.

Case study 12.1

Research question: Does ‘prestige’ manifest in our speech?
This is a rather general research question, so we need to narrow it down to answer
it. One of the ways we can do this is by testing a hypothesis in a sample of
language users. To do this, we will need:
a) a discourse community in which to test whether prestige exists;
b) some linguistic variables to test for prestige (and linguists need to agree what
counts as a prestige marker to make the research retrievable and replicable);
c) and some speech environment(s) that represent difering speech styles in
which prestige is likely to be more or less manifest.
One possible answer . . .
In a Belfast High school, one of the authors of this book set up a small focus group
discussion with teenage male students on the topic of ‘mobile phone use’. This
was in 2003 when mobile phones were pre-smartphone and certainly not as
widespread as they are today. The participants were told that the point of the
discussion group was to share thoughts on owning and using mobile phones and
their consent was sought and secured for the audio recording of the discussion
group and its use in a research study. While the experiment was indeed seeking to
collect their thoughts, this was only as a vehicle for eliciting and assessing the use
of two linguistic variables conventionally considered to mark prestige in English:
the presence of ‘-ing’ [-ɪŋ], as in ‘going’ vs its absence, as in ‘goin’ (see Chapter
1), and the presence of the standard ‘th’ [ð] as in ‘brother’ versus its absence as
in ‘broer’. The school was located in what is perceived to be a working-class area,
so the discourse community was more likely to exhibit non-standard features such
as ‘-ing dropping’. It was hypothesised that prestige markers would be largely
absent in informal speech.
The point of the study was to test a version of Labov’s ‘fourth foor’ experiment
(1972) in which he elicited casual and emphatic utterances of the rhotic [r]2 across
three New York department stores distinguished by social class (lower, middle, high).
Labov wanted to test whether the presence of the prestige marker [r] aligned with the
perceived prestige attached to the store – so higher-class stores would result in more
[r] whereas lower-class stores return less [r]. In the school experiment in Belfast, the
goal was to elicit as many utterances as possible of words containing the variables [-
ɪŋ] and [ð] in two speech situations, formal and informal, and then to analyse the
presence or absence across the two situations for patterns that discriminated the two
styles. Formal styles were assessed by getting the students to read aloud. Essentially,
this was done by the researcher asking the students to select a piece of paper from a
pile containing short statements that included words containing the linguistic
variables which they were to read out loud to serve as a discussion starter. Two such
examples read ‘My brother keeps stealing my phone’, and ‘My mother hates me
texting at the dinner table’, and so on. This was the ‘formal’ condition, so-called
because students are likely to be more conscious of the way they speak when reading
aloud to a small audience than when they are spontaneously speaking to each other
in a group discussion. The presence or absence of the variables in formal speech could
then be tested. The discussions that followed, some very animated, would then
provide the data for informal speech. The discussion talk satisfed the ‘informal’ or
‘casual’ condition.

Results from the high school experiment

‘-ing’ [-ɪŋ]: Results showed a marked disparity between formal and informal speech
in every instance of ‘-ing’. This variable was dropped 100 per cent of the time during
the informal discussion yet was present in most instances of the (formal) reading
aloud condition.
‘th’ [θ], [ð]: Results showed a marked disparity between formal and informal
instances of ‘th’: it was dropped in most instances during the informal discussion
(e.g. [ˈbrʌðər]>[ˈbrʌʔər]), yet it was present in all but one instance of the reading
aloud condition.
A really interesting fnding from the research was that in the formal condition
the same student resorted to non-standard forms, dropping ‘ing’ and ‘th’ (so from
[ð] to [ʔ]) which may indicate covert prestige of the nonstandard forms in all
conditions. Covert prestige is when non-standard forms of a language are given
more value than the standard forms amongst specifc social groups. This means
that the speaker may have believed that using non-standard variants in the formal
condition may have gained him some kudos with his peers. It would be interesting
to test whether this student used these non-standard forms in the formal condition
in one-to-one readings without his peers listening in!
We concluded that in defaulting to standard forms (higher prestige) in the
formal condition, students were aware of the social differences of both forms and
that the presence of these variables signalled overt prestige. Their absence,
therefore, may serve to connect them to their peer speech community reinforcing
their social bonds, and so constitute covert prestige.
A fnal word on the ethical considerations of this study: by not telling the
students what speech phenomenon was being studied we preserved the integrity
of the data. As all but one of the participants used the standard forms in the formal
298
 condition it suggests that they were conscious of the formality constraints on
speech styles and adjusted their speech accordingly. If we had told them what
linguistic phenomena was being elicited and why, they may well have changed
their informal speech behaviour, too.

Copyright
If you are collecting and storing discourse data, sharing your data, building a corpus,
and/or publishing your work along with your data, then you might need to think about
copyright. Copyright protects the work of others and gives them the exclusive right to
use it (e.g. print or publish it). If you want to use work that is in copyright, then you
must ask the rights holder’s permission. For example, the ‘Norn Iron’ T-shirt image
in Chapter 1 is included with permission from the source ‘Norn Iron Tees’. Images
are not the only medium that requires copyright permission. Whether you use news
discourse, advertising discourse, or political discourse, for example, you need to check
whether you need permission from the rights holder(s). Copyright laws are different
in different parts of the world, so you need to check carefully whether you need to
consider copyright or not. In the US, for example, copyright extends across the
owner’s lifetime plus another 50 years, although with works published before 1978
where the copyright was renewed after 28 years, the period of copyright after the death
of the owner is 75 years. In the EU, copyright extends across the owner’s lifetime,
plus another 70 years.
Provided that the research is for non-commercial purposes, it is possible to use short
extracts of discourse for research under what is known as fair dealing (or fair use).
Fair dealing permits the use of extracts without needing permission, and currently
allows up to 400 words from a journal article or book chapter, but no more than a total
of 800 words from an edition of a journal, or a whole book. For other genres, including
newspapers and magazines, no such fair dealing possibilities exist, and permission
must always be sought. If you are not intending to make a proft from your research,
then there is a good chance that permissions will be granted, but without permission
you cannot legally use and disseminate data.

Developing a research project

A research project that examines and analyses discourse is no different from any other
research project in the social and pure sciences: they all require planning and design.
The planning can begin with a series of questions, as follows.

• What do you want to fnd out and why?

• What are your research questions?
• What data do you need to answer those questions?
• Can you access the data?
• Are there any other problems with collecting the data?
• How will you analyse the data?
• How will you present your fndings?
• Does your data reveal anything worth reporting?
• If you do answer your questions, are your fndings generalisable?

Any project in discourse analysis should have clear research aims and research
questions that frame the research and help to direct it. The analysis of your discourse
data will aim to answer your research questions and achieve your research aims.
Research projects should begin with a hypothesis from which follows a question or
questions. Our example case study begins with a broad question, which is developed
until it becomes a measurable, testable hypothesis (see Figure 12.1).

Deductive and inductive research

So, how should you approach your research? Two possibilities are inductive research
(sometimes called ‘bottom-up’) and deductive research (conversely,‘top- down’).
Inductive research is where you look for patterns in your data with no prior hypothesis
about the possible patterns that might occur. Using our earlier example of the hotel
reviews data this would mean reading through the data to see what, if any, patterns
occur and how such patterns were categorisable. For example, you might notice that
‘super’ occurs very often as an adverb (e.g.‘super fun’) in reviews of higher-end
hotels but rarely in lower-end hotels and this might lead you to hypothesise that ‘super’
marks prestige. You can then select

300
Figure 12.1 Refning a research question

this variable and study its occurrence in the data. These patterns might lead to the
formulation of a hypothesis. Deductive research is where you start with a hypothesis
– that the adverbial function of ‘super’ is a marker of prestige’ – and study the data to
assess whether your fndings support this hypothesis or not. It’s important to note that
disproving a hypothesis is as informative and helpful as proving a hypothesis.
Research can be a combination of both inductive and deductive approaches.

Hypotheses
Hypotheses might be formulated based on our intuitions, or they might stem from the
literature on whatever topic you happen to be interested in, or something else such as
some previous research or observations as in our case study. Intuitions about the
phenomenon you are investigating can be important for formulating hypotheses.
Intuitions might be based on personal observations, knowledge, and might develop
previous research. For example, the hypothesis that people from a lower social class
speak differently in formal and informal situations could come from your intuitions
(or observations) about discourse:

I have a feeling (or I have witnessed) that people use standard English when they
are more self-conscious of their speech than when they are not.
Conversely, it could come from existing research, for example:
 William Labov conducted an experiment in the 1970s that found differing speech
styles across different social classes and I want to test it in a new speech community
with different variables to see if the phenomenon of social stratifcation exists in
that speech community.

The idea of a hypothesis is that it aims to explain a particular phenomenon. For

example, imagine that you are a European naturalist working in the 1700s and
someone sends you the stuffed remains of a duck-billed platypus. You have never seen
anything like it; you observe that the animal has a bill and webbed feet like a duck,
but a fur coat and tail more like a beaver. You might hypothesise based on your
knowledge of the natural world that the person who sent you the animal is playing a
practical joke on you, that the animal has been faked by a talented taxidermist and
does not exist.3 That the animal is fake is the hypothesis that explains what you are
observing.
A hypothesis must also be testable. That is, a hypothesis must be formulated in such
a way as to make it possible, after appropriate analysis, to reject it as an explanation
of the phenomenon under investigation. This idea of rejecting a hypothesis is known
as falsifcation, whereby we do not try to prove a hypothesis to be correct; instead, we
try to fnd data that forces us to reject the hypothesis. So, to test our duck-billed
platypus hypothesis, we would need to sail to Eastern Australia (it is the 1700s,
remember, so we’re going by boat) and look around for a real, naturally occurring
duck-billed platypus. As soon as we observe one, then we can reject our hypothesis
that the stuffed specimen scaring the neighbours from our home offce window is a
fake (no matter how painful and damaging to our reputation that might be).

Activity 12.1 Testing hypotheses

Suppose you hypothesise that vocabulary size makes the discourse in a particular tabloid
newspaper (Tabloid A) distinctive from other newspapers. How would you test that hypothesis?

Before reading on, consider the following questions (we have offered answers to the frst few).

• Q.1 What do you want to fnd out and why?

• Answer: Whether vocabulary size makes the discourse in a particular tabloid newspaper
(Tabloid A) distinctive from other newspapers.

• Q.2 What are your research questions?

• Answer: Does vocabulary size serve as a marker of discursive distinctiveness in a tabloid
newspaper?
• What data do you need to answer those questions?
• Can you access the data?
• Are there any other problems with collecting the data?
• How will you analyse the data?
• How will you present your fndings?
• Does the data reveal anything worth reporting?

302
 • Are the fndings generalisable?

Objectives and research questions

Using the activity question above, if we hypothesised that the discourse in a particular
tabloid newspaper (Tabloid A) is distinctive from other newspapers in terms of
vocabulary size, an over-arching or global research question might be:
Is the size of the vocabulary in Tabloid A distinctive from other newspapers?

The question raises other questions which we would frst need to answer before we
can:
(i) either reject or confrm our hypothesis; and
(ii) posit an answer to our over-arching research question.

To determine what these other questions are, we need to think about what our research
objectives are. These might be as follows (O = Objective):
O1: Determine the size of the vocabulary of Tabloid A and of other newspapers.
O2: Collect a sample of the discourse in Tabloid A. O3:
Collect a sample of discourse from other newspapers O4:
Count the number of words in both samples.
O5: Determine the number of diferent words in the samples.

To achieve these objectives, we need to answer the following research questions

(RQs):
RQ1: What is the vocabulary size of (i) Tabloid A, and (ii) other newspapers?
[Addresses O1]
RQ2: What counts as a representative sample of Tabloid A? [Addresses O2] (see
Chapter 11 for how to ensure representativeness in sampling)
RQ3: What counts as a representative sample of other newspapers? [Addresses O3]
RQ4: How many words are in each sample? [Addresses O4]
RQ5: How many diferent words are in each sample? [Addresses O5]

RQ1 is the global research question. The answer to this question can only be posited
after the other research questions have been answered. The other questions can be seen
as local research questions (see McIntyre and Walker 2019: 123–124). Answering
local research questions enables us to achieve objectives and therefore decide whether
to reject or confrm the initial hypothesis. Whether we reject or confrm our hypothesis
is determined by the answer to our global research question (RQ1).
Research questions help to direct your analysis as well as indicate to your reader
what it is you are looking for (and what you are not looking for). You do not have to
wait until you have decided on your data before start to develop research questions;
you can be thinking of questions before that, and those questions might dictate the
sorts of data you need to look at and eventually collect.
To summarise, one way to start thinking about research questions is to decide what
it is you want to know. If you have a hypothesis about a discourse, then this could be
stated as a question. It is perfectly normal for research questions to start off quite
general but then become more honed to address, say, specifc discourse practices (see
Figure 12.1). Do activity 12.2 which invites you to devise research questions from a
hypothesis about advertising discourse for gambling websites.

Activity 12.2 Hypothesis and research questions

Suppose you are interested in the advertising discourse of a gambling company and how they
present . . . well . . . gambling and how it positions you as a potential gambler. These general
ideas about what it is that you want to know can be seen as the aims of your project. From your
aims you can develop a research question or questions that help you to achieve your aims. So, if
your hypothesis is that ‘gambling advertising highlights the autonomy of the gambler’ (i.e. appears
to put the gambler in control), then what research questions could you ask?

Defining terms
An important step in any research project is to defne as clearly as possible the
phenomenon that you are investigating. This links back to the three Rs we mentioned
earlier, which require that you defne the terminology you are using so that everyone
is clear what you are researching and what your terms refer to. For example, if you are
interested in researching pauses in stand-up comedy, then you frst need to make clear
what is meant by the term ‘pause’. This might seem intuitively simple, but as we saw
in Chapter 4, a pause is only a pause if it lasts for more than a certain length of time.
Furthermore, a pause is not a fxed length, so your research might need to monitor and
account for different lengths of pause. Similarly, say we are investigating whether a
particular discourse is formal or informal (as in our Case study 12.1), we need to be
very clear what these terms mean, and which linguistic features combine to create
formality/ informality. The same goes for the term ‘prestige’ and what counts as
markers of prestige in a particular language – we might use conventional examples of
prestige markers suggested in previous scholarly research, for example. Research
projects that do not defne terms clearly are not replicable because it is not clear what
is being researched.

Controlling variables
It is important in any research project that variables are controlled. Variables are things
that might affect your results. If you want to know what has caused your result (which
variable) you need to be clear which variables are in play when you do your research.
304
So, for example, if you hypothesise that females swear less than males in spoken
discourse you need to be aware of and control the other variables that might be in play
when it comes to collect your data. Variables are either dependent or independent
and research aims to investigate the effects of independent variables on dependent
variables. In our swearing example, the amount of swearing is the dependent variable
(i.e. the thing being tested) and the research aims to establish what effect, if any, the
independent variable speaker gender has on that. Speaker gender is not the only
independent variable that might affect the amount of swearing in spoken discourse.
Other independent variables, such as age and socio-economic background of the
speaker, might infuence the amount of swearing. If these other variables are not
controlled (e.g. by only using speakers from the same age group in the research) then
it will be diffcult to draw conclusions from the results because it will not be clear
which independent variable is affecting the dependant variable and thus causing the
results. This means making sure that the males and females you collect data from are
of a similar age, socioeconomic background, and discoursing in the same or similar
context (e.g. at home with friends). Additionally, you will also need to defne what you
mean by swearing (see discussion above), and for that you might need to use an
existing taxonomy of swearwords (see, for example, McEnery 2005).

Data
As we have observed throughout this book, discourse data can be written, spoken, or
images (or multimodal). Remember, though, that spoken data must be transcribed frst.
The data you collect will aim to answer your research questions and fulfl your research
aims. One question that is often asked by students is: How much data do I need for a
project? The amount of data you need depends on what you are aiming to do. One
fairly short piece of discourse data can be enough, but it rather depends on what your
aims are. Some projects will make more than one example of discourse a necessity.
For example, you might want to carry out a comparison between discourses from
different newspapers or different politicians on the same topic. So clearly you need
more than one set of data to do that. Or you might want to investigate the way in which
someone presents someone else’s discourse by comparing the original (anterior)
discourse against the posterior discourse (see Chapter 10). Again, you would need
more than one text for that. In the school Case study (12.1) the whole discussion from
which the data was derived took about 30 minutes. It took about four hours to
transcribe. Be warned – transcribing data can take a very long time even for a few
minutes’ worth of talk! The data you need and what you do with it will be directed by
what it is you are trying to fnd out.
 Analysing your data
Discourse data can be analysed using any number of linguistic tools and approaches,
some of which we have outlined in this book. You need to approach your analysis in
a systematic and rigorous way. You need to go through all the data bit by bit looking
for all the occurrences of whatever discourse feature you have chosen to focus on.
Even though you might have a hypothesis about what you expect to fnd in the data,
you need to be quite open-minded about whether you will fnd it, or not. This is very
important.
You are not doing an analysis that sets out to fnd features that support your
hypothesis about the data, you are doing an analysis that goes through all your
data looking for all the features that you are looking for.
You cannot ignore or ‘cherry-pick’ examples to suit your hypothesis. You need to
take all occurrences of a feature into account. Once you have been through all the data
then you can see whether there are any patterns and begin to think about how the
discourse makes meaning. If your analysis supports your initial ideas, then that’s
something you can report. But if it does not, then that is still something to report –
and, as we said earlier, actually might be of greater interest. Doing systematic and
rigorous analyses means that you might fnd things that you did not anticipate. And
you gain a more detailed and more in-depth understanding of the discourse.

Writing up your research – doing academic discourse

If you write up your research (why wouldn’t you?), your report should bring together
data collection, your analysis, how you went about it and what you found. The report
should have a clear structure that is detached from the messy stuff that you might have
experienced while you were doing the research. Let’s just take a look now at what
your report might include.

306
Introduction

You should have an introduction where you set out what you are doing and your
research questions. You may want to contextualise your study.

Methodology

• You need to talk about your method of analysis and maybe your method for
collecting data in the frst place.
• You could state how and why you collected your data, how much you
collected, if you experienced any problems. As a minimum you should say
what your data is and where it came from.
• You should also say how you went about your analysis. It is important to say,
for example, that you went through looking for the feature that you were
interested in and that you looked for every single example of that feature. Or
if you took another approach and focused on some part of the data or one
participant in the data, then you need to say that and explain why.
• Try to show that you have been as objective as possible and tried as much as
possible to avoid your own biases about the data. Your readers need to know
that you have been systematic and rigorous and that you have not just looked
through for the interesting examples that help you to make the point you want
to make or that satisfy your hypothesis about the data. One of the best ways of
assessing objectivity is to ask questions of your own study as if you were its
greatest critic.

Analysis, results, and discussion

Here you present the results of your analysis, talk about what you found, and
suggest the impact of what you found. For example, you can discuss any patterns
you found. At the discussion stage, you write about your results in a narrative
format, as a kind of story about your data that makes sense. So, you have two
different processes and two different skills: doing the analysis on one hand, and
then writing up what you did and what you found in your analysis on the other.
Try not to do them both together at the same time. You could (should!) include
any tables or fgures of data in your narrative of the analysis so that you can fesh
out the results of your fndings for your reader. Remember to keep the reader with
you – do not assume they will know what you did or how you did it. You need
to tell them!

Conclusion

Conclusions can be hard to write. They might also be boring to write because you
are repeating yourself, but try not to make them boring to
 read. It is important that you draw out the main fndings and reiterate what is
important about your research. Don’t just copy and paste bits from other parts of
your study into the conclusions. Try to start with a blank piece of paper and try
to summarise your study just from your memory and say what it is you found
out. You can then fll in the detail if you need to once you have got the basic
structure down. You should also return to your research questions and refect on
whether you answered your questions or not. Make it easy for your reader to see
that you have kept the focus on the research questions that you started with.

Appendices

If you want to include additional materials relevant to the analysis, you could
add them as appendices. These might include ethical permissions and participant
consent forms, or longer stretches of data.

Conclusion
In this chapter we have considered some of the issues that need to be addressed before,
during, and after conducting analyses of discourse. These include ethical concerns
such as getting consent from participants, clearly informing participants in your study
what their contribution will be, what you will do with the data you collect, and who
will see it. We also offered suggestions regarding discourse contexts in which you
might want to conduct analyses. We encouraged you to think about your own
motivation for conducting discourse analysis – what kinds of questions you are
seeking to answer, who or what will beneft from your analysis, and what you (and
others) will learn from doing (and reading) your study. One of the key takeaways from
this chapter is that your research must observe the three Rs: rigour, replicability, and
retrievability. Finally, we offered a structural template for presenting your research.
In the appendices, we offer a sample document for acquiring informed consent from
participants in your study, along with a sample ‘information for participants’
document. When your study involves other humans or their language use (as it
invariably will!), you need to ensure that it is crystal clear about what you are aiming
to do and why. Following these guidelines means that you, too, will be very clear
about your role. Good luck!

Further reading
For more on doing project and for project ideas try Wray and Bloomer (2012); the
latest edition includes guidance on doing research on computer-mediated
communication.

308
For more on ethics and social media, see Townsend and Wallace (2017). For more on
testing hypotheses and designing research questions, see McIntyre and Walker (2019),
chapter 4.

Answers to activities
Activity 12.2 Hypothesis and research questions
Hypothesis: gambling advertising highlights the autonomy of the gambler.
RQ1: Does gambling advertising present gamblers as having autonomy?
For this RQ you would need to ask: How does gambling advertising discourse
suggest that gamblers have control? One way you could investigate this is through
transitivity choices. So, you could ask more specifcally: Do the transitivity choices of
the text producer suggest that the gambler is in control? Of course, if this were your
project, it would be helpful to make sure from the outset that it’s clear what you mean
by ‘in control’! This might be done by stating whether gamblers are Actors in
Material-Intention processes in the advertising discourse. This might indicate that they
have agency over their actions rather than being acted upon. Further questions that
you would need to answer your research question would relate to which data to collect
and how much. You might focus on one company or collect the adverts produced by
several.

Notes
1 Note that ‘research practice’ is different from forensic investigations. An example of the latter is
a police operation in which decoy participants are deployed to act as real participants in online
chatrooms in investigations into cases involving, for example, child exploitation.
2 Rhotic /r/ or ‘rhoticity’ is the term given to the pronunciation of the post-vocalic /r/ sound in
words like ‘car’ and ‘worked’. In some speech communities the /r/ is absent (i.e. not voiced) –
these are ‘non-rhotic’ pronunciations. In other speech communities the /r/ is voiced – these are
‘rhotic’ pronunciations. Culturally, and depending on where you are geographically, rhoticity can
carry prestige. In Labov’s New York study rhoticity was a marker of prestige. Rhoticity is not to
be confused with the rotating cooking method of roasting meat.
3 True story (sort of) – At the end of the 1700s, the biologist George Shaw struggled to accept that
the specimen of a duck-billed platypus he had been sent was legitimate.
 References
BAAL (2016) Recommendations on good practice in Applied Linguistics. Accessed
12 April 2023. https://baal.org.uk/wp-content/uploads/2016/10/
goodpractice_full_2016.pdf
Labov, W. (1972) ‘Some principles of linguistic methodology’. Language in Society,
1(1): 97–120.
Larner, S. (2014) Forensic authorship analysis and the world wide web. Basingstoke:
Palgrave Macmillan.
Martin, J. I. and Knox, J. (2000) ‘Methodological and ethical issues in research on
lesbians and gay men’. Social Work Research, 24(1): 51–59.
McEnery, T. (2005) Swearing in English. Bad language, purity and power from 1586
to the present. London: Routledge.
McIntyre, D. and Walker, B. (2019) Corpus stylistics: Theory and practice.
Edinburgh: Edinburgh University Press.
Simpson, P. (2014) Stylistics: A resource book for students (2nd edition). London:
Routledge.
Townsend, L. and Wallace, C. (2017) ‘The ethics of using social media data in
research: A new framework’. In K. Woodfeld (ed.), The ethics of online research,
pp. 189–207. Advances in Research Ethics and Integrity, Vol. 2. Bingley,
Yorkshire: Emerald Publishing Limited.
UKRI (2021) Framework for research ethics.Accessed 12 April 2023.
https://www. ukri.org/councils/esrc/guidance-for-applicants/research-ethics-
guidance/ framework-for-research-ethics/our-core-principles/#contents-list
Wray, A. and Bloomer, A. (2012) Projects in linguistics and language studies.
London: Routledge.

310
 References

Canning, P. & Walker, B. (2024) Discourse Analysis: A Practical

Introduction.(1sted.). Routledge Taylor & Fracis Group.

Nunan, D. (1993) Introudcing Discourse Analysis. Penguin.