1 Introduction

1.1 Corpus Linguistics

Since the 1970s, the development of corpus linguistics has brought
about important changes in linguistics and applied linguistics. Corpus
linguistics is an approach to the study of language that involves
collecting large quantities of naturally occurring language and using
specialised software that manipulates that language to obtain infor-
mation about frequencies, co-occurrences and meanings. The lan-
guage may be spoken, written or signed, in one language variety or
more, and one register or more. It consists of language which has
occurred in natural contexts, not as the result of elicitation or intro-
spection. The components of the corpus are texts (whole or partial)
and thus consist of pieces of connected discourse. The quantity may
range from a few hundred thousand words to billions, though the
corpus usually contains more texts than could reasonably be read and
remembered by an individual.
What distinguishes a corpus from a collection of digitised texts is that
it is formatted such that the application of the software enables pattern-
ing to be observed that would be missed by conventional forms of
reading. The patterning might consist of collocations or phraseology,
or it might associate some language features disproportionately with
some parts of the corpus. The output from the software may be lists of
items, whether words, phrases or classifications, or it may be sets of
numbers visualised as tables, graphs or plots, or it may be simple sets
of concordance lines. Whatever the output, it is interpreted in terms of
the variety, register or community that the corpus represents.
Thus, doing corpus linguistics involves manipulation and observa-
tion of ‘what has been said (or written or signed)’. Although it may be
framed in terms of testing hypotheses about what will be found in
which contexts, the methodology is less about hypothesis formation
and more about moving from observation to generalisation and

https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

 2 Corpora in Applied Linguistics

categorisation of language features. This in itself has implications for

what the job of the researcher is taken to be.
Corpus linguistics has had a transformative effect on the study of
language. The impact may be said to be ‘inward’ and ‘outward’ facing,
that is, affecting the study of language(s) itself and affecting aspects of
life that are dependent on language. As this book will demonstrate,
corpus studies are often comparative and this has enabled varieties of
language distinguished by place, time or context to be studied in greater
detail than before. Models of the structure of language are also open to
test, but in addition new views of how language might be described
have emerged, for the most part giving lexis and phraseology a more
important role in that description than before.
Language, of course, is an integral part of many aspects of life and
the investigation of language in context has the potential to impact
those aspects. Because corpus linguistics studies naturally occurring
language, and corpora can be collected from specific contexts of use,
it has heavily influenced and extended the scope of applied linguistics.
Chapter 6 of this book discusses the impact of corpus linguistics on
materials for language learners and teachers. The corpora studied
consist of language produced by both experts and learners. The impact
is partly practical: learner dictionaries based on corpora can include
more detail about how words are used, for example, or the aspects of
discourse that are most difficult for a group of learners can be identified
from a corpus of their writing. It is also theoretical: large quantities of
speech or writing from learners can give insights into the processes
involved in acquiring a language and the extent to which individuals
and groups vary. Chapter 7 focuses on the role of corpora in investi-
gating how ideas are transmitted through discourse of all kinds, includ-
ing media discourse and academic discourse. Corpora allow large
amounts of text to be considered, allowing questions to be answered
such as ‘what is most often talked about in this context?’ or ‘what
attitudes are expressed or implied?’ or ‘how are texts structured?’. This
can show how coherence is achieved in texts, how knowledge is
constructed, and how views of society are normalised.
The availability of corpora has assisted other applications of language
study. Answering the question ‘who wrote this?’ has applications in
literary study and in forensic linguistics. Studying large amounts of
Twitter or blog data can answer questions about what topics are raised
in specific forums and how identities are construed on-line. This has
applications to discovering patient concerns about hospital services,
tracking the spread of political ideas, or identifying malicious communi-
cations. The linguistic strategies used to persuade can be identified
through corpora, and these are important to the development of

https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

 Introduction 3

knowledge in all fields of study, from physics to history, as well as in

applications such as political discourse or advertising. How languages
are like or unlike each other has been shown to be a question of relative
quantity rather than of sharp distinctions. For example, if each of two
languages uses a structure broadly equivalent to the passive, one lan-
guage may use the structure much more or less frequently than the other.
Corpora can be used to study the important field of translation, looking
both at ‘what translators do’ and ‘what translators need to know’.
Chapter 8 considers a range of applications of corpus linguistics.
Above I suggested that corpus studies can be considered ‘inward-
facing’, relating to the question of ‘what language is like’, or ‘outward-
facing’, considering ‘what language is used for’. In applied linguistics,
and in corpus linguistics, however, these two perspectives interact with
and inform one another. To take a case in point: Sinclair’s (1991; 2004)
influential study of individual words and their phraseology led to the
development of a number of concepts about the structure of language,
notably the unit of meaning and the open-choice principle / idiom
principle distinction. These constitute a way of describing language that
is based on (or driven by) the observation of words in a corpus. The
detailed, word-by-word description of English that acted as proof of the
unit of meaning concept was made possible by the work involved in
compiling a dictionary for learners of English (the Cobuild
dictionary, Sinclair 1987a). The act of deriving the theory of form
and meaning and the act of compiling the dictionary informed each
other, each of them impossible without the other.

1.2 About this Book

The first edition of Corpora in Applied Linguistics was written about
20 years before this one. At that time, corpus linguistics was a fairly
new way of approaching the study of language. It diverged from then
more mainstream approaches to language description in a number of
ways that have continued to be important in the field, as noted above:
a lot of language is collected; it has been produced in natural contexts;
software is used to manipulate the language and present it in innova-
tive ways; the language is observed and generalisations made.
Already 20 years ago it was clear that the observations from corpus
research could supplement other information. For example, studies of
language change over time were placed on a firmer footing because of
the large amount of evidence gathered (e.g. Mair, Hundt, Leech and
Smith 2003). The same is true of comparisons of registers, putting
register variation at the heart of language description (e.g. Biber,
Johansson, Leech, Conrad and Finegan 1999). Essentially, statements

https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

 4 Corpora in Applied Linguistics

about the differences between times, places and registers could be

made with much more confidence than before. Language descriptions
both relied on perceptions of difference but also contributed to how
those differences were conceptualised. It also, as noted above, led to
new concepts such as Units of Meaning (Sinclair 2004) or lexical
priming (Hoey 2005). Corpus linguistics could offer important sup-
port to other kinds of linguistic research but it was, arguably, at its
most significant when it disrupted perceptions of language.
Since the publication of the first edition, corpora and the techniques
used to study them have expanded in all directions. It has become
feasible to compile larger and also more specific corpora. The statistics
used have become more complex and are increasingly accompanied by
sophisticated visualisations of the data. The challenge for the corpus
linguist today might not be ‘how to see the wood for the trees’ but
‘how to see the trees as well as the forest’. Some concepts from corpus
linguistics have entered everyday life. Wordclouds based on word
frequency lists (or keyword lists) are a common way of representing
the ‘aboutness’ of a text. Ngrams tracked across decades of book
publication show trends in topic development. Commercial applica-
tions and an increase in digitally available texts have led to some of the
questions in corpus linguistics being tackled from more computational
perspectives, such as the development of algorithms to measure opin-
ions expressed in product reviews (see Chapter 8). Corpus linguistics,
however, continues to contribute to, and sometimes challenge, other
forms of language study. In particular, the corpus-inspired approach
to the unity of lexis and grammar both accommodates and questions
approaches to the same issue from cognitive linguistics and from
systemic-functional grammar (see Chapter 9).
Because of the vast growth in corpus linguistics in the two decades
between the first and second editions of this book, the two editions
cover very different ground. This second edition is broader, including
references to a greater diversity of approaches. It is also, necessarily,
highly selective. Research is exemplified rather than comprehensively
surveyed. For the most part, and to keep the project within manage-
able proportions, discussion is restricted to corpora of English.
Following this introductory chapter, the organisation of the book is
as follows. Chapter 2 describes types of corpora and discusses the
main issues raised in the compilation of corpora. It includes a list of
corpora of English mentioned in the book. Chapters 3–5 focus on
methodology. Chapter 3 exemplifies how patterns can be observed in
corpora and what conclusions can be drawn from these. This might be
called the qualitative approach to corpus investigation. Chapters 4
and 5 turn attention to quantitative research. Chapter 4 covers the

https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

 Introduction 5

older and more basic approaches to quantity while Chapter 5

addresses more recent developments. Chapters 6–8 focus on applica-
tions of corpus linguistics: to language learning and teaching in
Chapter 6, the study of discourse in Chapter 7, and other applications
in Chapter 8. Chapter 9 considers an open-ended question: the appli-
cation of corpus linguistics to language theory. It discusses this first in
relation to language as a mental phenomenon, specifically the poten-
tial alignment between corpus linguistics and construction grammar.
Then the issue is discussed in relation to language as a social phenom-
enon, specifically the relationship between corpus linguistics and
systemic-functional linguistics. Chapter 10 concludes the book with
an illustration of the application of corpus studies to an issue that is
key to life in the early 2020s: the COVID-19 pandemic.

1.3 Terminology
In this section, some key terms that will be used throughout the book
are explained. They are: text, type, token, lemma, wordform, ngram,
concgram, tag, parse, annotate and metadata.

1.3.1 Text, type, token

Corpora are often described in terms of the number of texts they
contain, the number of tokens, and the number of types. Usually, a
text is one of the pieces of spoken or written language that have been
taken in their entirety from a natural context and compiled into a
corpus. For example, in a corpus of student essays, each essay is a text.
Sometimes, however, the word is used slightly differently, for example
to describe something that is longer or shorter than a naturally occur-
ring text. Some corpora, for example, consist of texts that are exactly
2,000 words long, which means that the corpus ‘text’ is not the same
as the ‘text’ from which it is taken, either being an extract from a
longer piece of writing or consisting of several texts combined.
The terms type and token both mean ‘word’, but in slightly different
senses. The following paragraph (from Simpson and Montgomery
1995: 140) illustrates this:
What elements make up a narrative? Providing an answer to this question has
become one of the central challenges for a stylistics of prose fiction. Much
work in modern narrative stylistics seems to isolate the various units which
combine to form a novel or short story and to explain how these narrative
units are interconnected. Having identified the basic units in this way, the next
task is to specify which type of stylistic model is best suited to the study of
which particular unit.

https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

 6 Corpora in Applied Linguistics

In one sense, there are 84 words in this paragraph. That is, there are
84 sequences of letters separated by spaces or punctuation. The word
token is used to mean ‘word’ in this sense: the paragraph consists of 84
tokens. It is also said that the paragraph consists of 84 ‘running
words’. However, many of these words occur more than once: a,
narrative, units and which occur three times each; stylistics occurs
two times; to occurs six times and so on. If each unique word is
counted only once, there are 60 words, or 60 types, in the paragraph.
As texts get longer, more words tend to be repeated, so the number of
types relative to the number of tokens goes down. Texts that are
carefully written and contain complex ideas, such as the one above,
tend to have more types relative to the number of tokens; texts that are
easier to read have fewer types relative to tokens. The type-token ratio
(TTR) is often used to compare texts. Corpora are often described in
terms of their total number of tokens and types, and the average
number of tokens and/or types per text.

1.3.2 Wordform, lemma, stem, ngram, concgram

One notable aspect of this account of types is that each wordform is
counted separately. It was said above that there are three instances of
the type units in the paragraph. There is also an instance of the word
unit, but this is treated as a different type. There is a sense, however, in
which the singular unit and the plural units are ‘the same word’. They
are said to comprise the same lemma. The same is true for the word-
forms eat, eats, ate, eating and eaten, which comprise the lemma EAT.
This book follows common practice in indicating wordforms cited
from corpora by lower case italics and lemmas by capital letters. It
also follows the practice of specifying lemmas by word class.
According to this definition, the wordform walk belongs to two
lemmas: the noun lemma WALK with the wordforms walk and walks
(as in ‘I went for a walk’); and the verb lemma WALK with the
wordforms walk, walked, walking and walks (as in ‘I walked two
miles’). Following the same principle, the wordforms evident (adjec-
tive), evidently (adverb), evidence (noun) and evidence (verb) belong
to four separate lemmas. Some software allows the user to specify
whether the search for a lemma will treat noun and verb walk as
belonging to the same lemma or to different ones. Some pre-processing
of a corpus has to be carried out if lemmas are to be identified. This
often relies on a dictionary containing information such as that ate and
eaten are instances of the same lemma. A more rough-and-ready way
of obtaining lemmas is to use a stem approach, which means that a
wild-card query is added to a word stem. For example, in building the

https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

 Introduction 7

Coronavirus Corpus (see Chapter 2), Davies (2021) searches for

words such as contagious but also for words sharing a stem, such as
self-isolat*. This search will find self-isolate, self-isolates, self-isolated,
self-isolating, self-isolation and any other forms with that stem.
In the account of type and token above, the stylistics paragraph was
divided into individual words, which are sequences of letters separated
by spaces or punctuation, but it is possible also to divide it into strings
of words or -grams. These can be two words long (bigrams), or three
(3-grams), four (4-grams) or any number. The general term is ngram.
The second sentence from the stylistics paragraph can be divided into
ngrams from 2 to 5, as shown in Table 1.1. Ngrams of a given length
can be quantified and compared in the same way as individual words
or lemmas are. From Table 1.1, it might be expected that items such as
one of the (central), providing an answer to and an answer to this
question would appear many times in a corpus of English, while prose
fiction or even a stylistics of prose fiction might appear many times in a
specialised corpus of stylistics. (These items are shown in bold in the
table.) In many cases, however, the recognised phrase would include a
variable item. A researcher counting the frequency of providing an
answer to may wish to identify also providing an acceptable answer to
or providing a response to. The units which might be described as
‘ngrams with a variable slot’ are known as concgrams.

1.3.3 Tagging, parsing, annotation and metadata

These terms are applied to procedures that add information to the
material in a corpus. The process of adding the information may be
entirely automatic or entirely manual, but is often a combination of
the two. Metadata is information about a text, such as the date or
place of publication, the genre of the text, or the gender or language
background of the speaker(s). In some corpora, the metadata attached
to each text can be exploited to build a bespoke sub-corpus to meet the
researcher’s needs. For example, the MICUSP corpus (Römer and
O’Donnell 2011), consisting of papers written by students at US
universities, can be searched to obtain papers written by students at
a selected level, or by native or non-native speakers of English, or to
obtain papers in a given discipline or of a given genre. This is possible
because each text has metadata added to it indicating the level of the
student, the discipline and so on.
The term tagging is normally used to refer to the process of adding a
part of speech (PoS) label to each word in a corpus. This enables
searches and frequency counts that depend on part of speech to be
undertaken. It is possible, for example, to compare corpora in terms of

https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

Table 1.1 Ngrams in a sentence

Providing an answer to this question has become one of the central challenges for a stylistics of prose fiction.
2-grams 3-grams 4-grams 5-grams
providing an providing an answer providing an answer to providing an answer to this
an answer an answer to an answer to this an answer to this question
answer to answer to this answer to this question answer to this question has
to this to this question to this question has to this question has become
this question this question has this question has become this question has become one
question has question has become question has become one question has become one of
has become has become one has become one of has become one of the
become one become one of become one of the become one of the central
one of one of the one of the central one of the central challenges
of the of the central of the central challenges of the central challenges for
the central the central challenges the central challenges for the central challenges for a
central challenges central challenges for central challenges for a central challenges for a stylistics
challenges for challenges for a challenges for a stylistics challenges for a stylistics of
for a for a stylistics for a stylistics of for a stylistics of prose
a stylistics a stylistics of a stylistics of prose a stylistics of prose fiction
stylistics of of prose fiction stylistics of prose fiction
of prose
prose fiction
 Introduction 9

the number of nouns or verbs in them, or to find all instances of the

noun (but not the verb) walk, or to search for all the adverbs that
precede a specific adjective. Tagging is often carried out automatically,
but the accuracy of automatic tagging procedures varies, and manual
editing is often employed if the corpus is small enough. Tagging is also
used as the basis for parsing, where the text is analysed grammatically
and the constituents of clauses and groups are identified.
Tagging is one form of text annotation. Other common forms of
annotation are error annotation (or error tagging), which is applied to
texts written by learners of a language to identify and quantify error
types, and semantic annotation, which assigns each word in a corpus
to a predetermined semantic set. Annotation of errors is normally
carried out manually (and there may be considerable disagreement
about what constitutes an error), though there is work on automating
the process via machine learning (Buttery 2021). Semantic annotation
is carried out automatically, based on a pre-classification of words into
semantic sets. (See Chapter 4 for more information and discussion of
this type of annotation.)
All forms of annotation involve the development of a tag-set. This
may be a list of parts of speech, a list of error types, or a list of
semantic sets.

1.4 Commonly Used Resources

At the time the first edition of this book was written, there were few
publicly available corpora and associated software resources. There
are now many. A short list of the best known are listed here. A list of
frequently used corpora is given in the appendix to Chapter 2.
Antconc (laurenceanthony.net) is one of a suite of programs
developed by Laurence Anthony. Researchers can use it to perform
tasks such as concordancing, obtaining lists of collocates, finding lists
of keywords, etc. on their own corpora. Antconc is constantly being
revised to introduce new features.
English-Corpora.org (English-corpora.org) is a collection of ten US
corpora and seven other corpora, with associated software, compiled
by Mark Davies. The US corpora include the Corpus of Contemporary
American English, the Corpus of Historical American English, the
News on the Web corpus and the Corpus of American Soap Operas.
The corpora and software are accessed on-line.
#LancsBox (corpora.lancs.ac.uk/lancsbox) is a suite of corpora and
associated tools developed at Lancaster University by Vaclav Brezina
and others (Brezina, Weill-Tessier and McEnery 2020). It can be used
with the ready-made corpora included, or researchers can use it with

https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

 10 Corpora in Applied Linguistics

their own corpora. It includes annotation and visualisation tools. It is

downloaded for use on the user’s own computer.
SketchEngine (sketchengine.eu), developed by Adam Kilgarriff, is a
suite of tools including concordancing, collocations, Word Sketch, a
thesaurus and many others. It contains hundreds of ready-made cor-
pora, in over 90 languages. Users can upload their own corpora into
SketchEngine and use the tools on those corpora. The corpora and
software are accessed on-line.
Wmatrix (ucrel.lancs.ac.uk/wmatrix) provides tools such as con-
cordancing, collocations and keywords. It was developed by Paul
Rayson. Users can upload their own corpora, which can be annotated
with Part of Speech tags and with semantic tags using the USAS tag-set
(see Chapter 4, Section 4.8) and can obtain key PoS and semantic
categories as well as keywords.
Wordsmith Tools (lexically.net/wordsmith) is another suite of pro-
grams, this one developed by Mike Scott, that can be used on the
researcher’s own corpus. The set of tools includes condordancing,
keywords, a concgram finder and others. Like Antconc, it is frequently
revised. The 2020 version includes a facility to link concordance lines
to video files.

1.5 Corpus Linguistics: A Personal View

Corpora are often used to test hypotheses about language, but one
aspect of corpus research that is often stressed is the ‘serendipity’ of
corpus research, when looking at the output from a corpus investi-
gation tool leads to surprising and exciting insights. My own first
introduction to corpus linguistics (courtesy of a talk given at the
University of Surrey by Jyl Francis) included a sample of concordance
lines for the adjective possible. This is an adjective with a wide range
of uses, far more than similar adjectives such as probable or impos-
sible. Identifying those uses convinced me that this new way of looking
at language was both informative and exciting.
To conclude this chapter, then, I shall attempt to replicate that
experience. The corpus I use here is the Wordbanks Online corpus
from HarperCollins, which uses a version of the SketchEngine search
software (wordbanks.harpercollins.co.uk). There are 107,735 instances
of possible in this corpus. For the purposes of illustration I have selected
a sample of 250. That is too many lines to show here, so the lines are
first ‘shuffled’ (put in random order) and then 25 successive lines from
the middle of the set have been extracted and are shown in Figure 1.1.
In this figure the lines are numbered and letters used to show the source
of the line: B(ook); M(agazine); N(ewspaper); S(poken). The word

https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

1 M. ertwined with neutrinos. We’re very excited about the possible connection, says Nelson. This is the only theory of dark
2 N. season, Borys says, has been stretched. That’s partly possible, exhibitors and analysts say, because moviegoing has
3 N. fine? Email us at ADDRESS with your tale for possible inclusion in an upcoming USA TODAY story.
4 N.and services would be allocated to the greatest extent possible, but this was a way for the State to avoid giving special
5 N. discos, indeed anything to attract the largest crowd possible into their already overcrowded establishments. And all in
6 B.ng more than one sentence. Each item has more than one possible answer. Example: My dog barks, and I have to move out of my
7 N. These developments raise questions related to other possible routes of transmission of SARS. These may involve
8 N. gues in Europe and there is also the added benefit of possible post-career opportunities. Heavy has seen his crock
9 N. as three bedrooms, two upstairs. But somehow it seems possible to stuff quite a few people into it, more than three bedroom
10 N.ss san Jose Calif. It might have seemed to be the best possible week for boosters of the Linux open-source operating
11 B.cline of the Public Sphere. There is, however, a final possible explanation for the decline of social equality – the Civic
12 B.and looked at her carefully. Did I? You did everything possible, Morgan. I could see Andy just a few feet away.
13 B.of a sandy wash, and he had wanted to sleep as long as possible. Looking at the agitated old man, Wanda wondered if perhaps
14 B. it could again either halve or rise 1.5 times, giving possible interest rates of 2, 6, or 18 percent. Now look at the pric
15 B.everyone working for minimum rates, it would hardly be possible to make it for so little, and in fact the film actually came
16 N.d. Board members are expected to continue discussing a possible 2004 construction bond measure. Electric blanket may have
17 B.w that he could resolve everything in the simplest way possible. He mustn’t let himself sink into total oblivion so that he
18 S. his call for the union treaty to be signed as soon as possible, to be followed as planned by elections for a new federal
19 N. families (although their presence in the opener makes possible a finale, in which Aggie compels her to hug her equally
20 N. sidiary of Motibhai’s, with the best level of service possible to support the effectiveness and success of their duty free
21 B. made a pass at him a long time ago, but in the nicest possible way he said I wasn’t his type. You’re a lucky girl. The only
22 S.the president, Mr Gbagbo. So that means it will not be possible to engage any kind of disarmament process without voting
23 N. planning going on yesterday was focusing on the worst possible scenario. In the lead-up to the outbreak of a war, the
24 S. about nervous reaction from financial markets to a possible Lula victory, the former labor leader said markets are
25 B.Until then we travelled to the races together whenever possible. Not any more. It wasn’t possible once we were at each other

Figure 1.1 Concordance lines for possible from Wordbanks Online

 12 Corpora in Applied Linguistics

possible is an adjective and in some lines the expected behaviour of

adjectives is observed: it is attributive, as in the possible connection
(lines 1, 3, 6, 7, 8, 11, 14, 16, 24), and it is predicative, as in That’s
partly possible (line 2). In some lines it is followed by a to-infinitive, as
in it seems possible to stuff quite a few people into it (lines 9, 15, 22).
There is also the expected sequence ‘as + adjective/adverb + as +
possible’, as in as long as possible (lines 13, 18). The list of uses
continues with the sequence ‘the + superlative + noun + possible’, as
in the greatest extent possible (lines 4, 5, 17, 20) and a variant ‘a/the +
superlative + possible + noun’, as in the worst possible scenario (lines
10, 21, 23). There are three further lines illustrating other uses: did
everything possible (line 12), whenever possible (line 25), and ‘make +
possible + noun phrase’ – makes possible a finale (line 19). The random
25 lines illustrate 9 different patterns of use. Returning to the 250 line
sample, more instances of each of these can be obtained. Some
examples are shown in Table 1.2. Two additional uses with MAKE
have been found, exemplified by make the development possible and
make it possible to. . . The phrase if possible has also been found. These
are added to the end of the table, making 12 patterns in all.
The conclusion might be that possible has a large number of behav-
iours, or is used in a large number of patterns, or, to borrow the term
from cognitive linguistics, occurs in a large number of constructions
(see Chapter 9). Some of these are almost unique to possible (e.g. if
possible, as soon as possible), in that possible can be replaced by a very
limited range of adjectives, notably available, though it can be
replaced by a phrase with can (if we can, as soon as you can).
Others are shared with all other adjectives, being the regular attribu-
tive and predicative uses (e.g. the possible explanation / the red bicycle
or it’s possible / it’s desirable). Some patterns are shared with a sub-set
of adjectives including probable and likely (it is possible that. . . / it is
probable that. . .). Others are shared with a sub-set of adjectives
including easy and difficult (it is possible (for her) to. . . / it is easy
(for them) to. . .; make it possible / easy for him to. . .). These observa-
tions are summarised in Table 1.3.
All this suggests that possible has a wider range of meanings
than most adjectives. Some are associated with epistemic modality
(what might be the case: it is possible that. . .) and with
deontic modality (what should or can be done: it is possible to. . .).
Table 1.2 also suggests that another distinction can be made,
between possible associated with the superlative and extremes
(as soon as possible; everything possible; the best possible
situation; the greatest happiness possible) and possible associated with
cautious assessment (it is possible that. . .; it might be possible to. . .).

https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

Table 1.2 Patterns of possible in 250 random lines from Wordbanks Online
Pattern Examples
possible + noun . . .a possible merger. . .
. . .a possible leak or fire. . .
. . .a possible evacuation of 1,000 Canadians in the
. . .face possible deportation
. . .looking for possible connections
. . .the extent of possible brain damage. . .
One possible avenue. . .
. . .the possible resumption of diplomatic ties. . .
two of the possible intermediaries. . .
link verb + possible Anything’s possible.
. . .to come back if it’s possible
. . .that love is possible for all of us
I still think pressure is possible
A deal for devolution is still possible. . .
Was such a thing possible?
it + be + possible + to It’s possible to provide information to the public
it will not be possible to engage any kind of disarmament pro
It is possible to support the battle against terror
. . .it is possible to rid yourself of any fear. . .
It must be possible for people to have access to the air
. . .why it wasn’t possible to think of nothing at all
It was possible to see clearly what war had wrought
(continued)
13
https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

Table 1.2 (cont.)

14

it + be + possible + that . . .so it’s possible I could play in the future

. . .it is possible that families with multiple occurrences. . .
It is possible that those persons of rank. . .
It is possible the bouncing movements of running could
Was it possible that all the vermin in the world..
It is quite possible that Jem became a member of this brother
as + adv/adj + as possible . . .he’s as far away as possible from the trophy. . .
. . .had wanted to sleep as long as possible. . .
. . .try and help them as much as possible. . .
. . .dealt with as quickly as possible. . .
. . .are as similar as possible. . .
. . .call her as soon as possible. . .
appoint an administrator as soon as possible. . .
. . .murdering people as violently as possible. . .
the + superlative + noun + possible . . .attract the largest crowd possible
. . .the greatest happiness possible. . .
. . .in the most efficient way possible. . .
. . .in the most horrible way possible. . .
a/the + superlative + possible + noun . . .the best possible situation you could be in
. . .make the best possible recovery from her ordeal. . .
He has kept in the closest possible touch
. . .the highest possible price
. . .in the nicest possible way. . .
. . .at the simplest possible level of argument. . .
. . .the worst possible scenario. . .
https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

proform + possible . . .did everything possible. . .

. . .tried everything possible. . .
. . .whenever possible. . .
. . .wherever possible. . .
make + possible + noun phrase . . .made possible a much closer relation. . .
. . .a man who made possible the initial contact. . .
. . .and thus make possible a form of competitive coexistence.
make + noun phrase + possible . . .that made such a development possible
The trip was made possible by. . .
make + it + possible + (for noun) + to the appointment made it possible for him to pursue resear
Mechanization would make it possible to produce in 25 working
if + possible . . . are told to hold fire if possible. . .
I like to use Scottish wood if possible
. . .by Christmas if possible. . .
15
https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press

16

Table 1.3 The adjective possible compared with other adjectives

Pattern Example Pattern is used with. . .
possible + noun . . .a possible merger. . . any adjective
link verb + possible Anything’s possible any adjective
it + be + possible + to It’s possible to provide information. . . adjectives such as easy, difficult, impossible
it + be + possible + that It’s possible I could play in the future adjectives such as probable, likely, impossible
as + adv/adj + as possible . . .as soon as possible no adjectives, but replaceable by you can
the + superlative + noun + possible . . .the largest crowd possible. . . possible, available
a/the + superlative + possible + noun . . .the best possible situation. . . possible, available
proform + possible . . .tried everything possible possible, available
make + possible + noun phrase . . .a man who made possible the initial contact adjectives such as feasible, achievable
make + noun phrase + possible The trip was made possible by. . . adjectives such as feasible, achievable
make + it + possible + (for noun) + to The appointment made it possible for him to pursue research. . . adjectives such as feasible, achievable
if + possible . . .by Christmas if possible no adjectives, but replaceable by you can
 Introduction 17

This illustrates the interconnection between form and meaning, and

more specifically the importance of patterning, or phraseology, in
recognising meaning. It also illustrates the value of seeing many
examples of a specific form, possible in this case, in a format that
highlights similarity in the immediate co-texts (in Figure 1.1 and
Table 1.2).
Much of this information about possible could be found from other
ways of encountering language, including introspection. To me,
though, seeing the patterning in concordance lines offers a unique
perspective that provides affective as well as intellectual satisfaction.
This book will discuss methods that go far beyond concordancing, but
the joy of identifying what was previously unknown will be a
recurring theme.

https://doi.org/10.1017/9781108616218.001 Published online by Cambridge University Press