LINGUIST List 14.3137

Sat Nov 15 2003

Review: Corpus Ling/Lexicography: Jones (2002)

Editor for this issue: Naomi Ogasawara <>

What follows is a review or discussion note contributed to our Book Discussion Forum. We expect discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in. If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for review." Then contact Simin Karimi at


  1. Diana Lewis, Antonymy: A Corpus-based Perspective

Message 1: Antonymy: A Corpus-based Perspective

Date: Sat, 15 Nov 2003 12:12:20 -0500 (EST)
From: Diana Lewis <>
Subject: Antonymy: A Corpus-based Perspective

Jones, Steven (2002) Antonymy: A Corpus-based Perspective, Routledge,
Routledge Advances in Corpus Linguistics 2.

Announced at

Diana Lewis, University of Lyon2.


_Antonymy: A Corpus-Based Perspective_ examines the use, in samples
taken from a corpus of contemporary English newspaper language, of
fifty-six antonymous pairs of lexemes. These include adjectives (e.g.
dry/wet), adverbs (e.g. directly/indirectly), verbs (e.g.
fail/succeed), and nouns (e.g. peace/war).

The aim of the study is threefold (pp. 25-26): (1) - to investigate
and quantify the intra-sentential functions served by antonymy in
written text; (2) - to generate co-occurrence statistics and textual
profiles for individual antonymous pairs; (3) - to examine variables
(word class, gradability, etc.) which might affect the function of

The book is likely to be of interest to corpus linguists and


The first chapter, describing the attractions of antonymy - for those
acquiring L1 or L2, for writers, for linguists, indeed for everybody -
is followed by a chapter providing a brief overview of some previous
antonymy studies. Antonymous pairs have previously been described and
categorized in terms of gradability, reciprocity, inverseness and so
on. The present study proposes a new classification of antonymy.

Chapter 3 describes the data and method of this study. A selection of
fifty-six antonymous pairs for analysis was made on the basis of the
author's intuition that these ''would be intuitively recognised as
'good opposites', to a greater or lesser degree'' (p. 35). Further
selection criteria were that adjectives, nouns, verbs and adverbs be
included, that there be both gradable and non-gradable antonyms, and
some 'morphological antonyms' (specifically, some antonyms based on
the prefix -un). Sentences containing one or more of these antonymous
pairs were extracted from a corpus of around 280 million words of text
from The Independent newspaper. 3,000 of these sentences were retained
for analysis, their selection determined by various criteria to ensure
they included large enough samples of each word class, of non-gradable
antonyms and of morphologically-related antonyms. The 3,000 examples
of antonym use were then analysed and grouped according to the
perceived function of the antonymous pair, eight categories emerging
from this analysis. Two of these categories, termed 'Ancillary
antonymy' and 'Coordinated antonymy', account for over three-quarters
of the occurrences, and are the subjects of chapters 4 and 5

The first major category, 'Ancillary antonymy', comprises occurrences
where the contrast between the members of the antonymous pair evokes a
contrast between the elements the antonyms modify, as in '''Stamps are
popular, but collecting is unpopular''' (p. 35). The function of the
antonymous pair is ''to instruct us to interpret a nearby pair of
words or phrases contrastively'' (p. 38). There follows a puzzling
sub- categorization of the 'nearby' pairs, called 'B-pairs', into some
heterogeneous classes. Thus in '''Kennedy dead is more interesting
than Clinton alive''' (p. 49), the B-pair Kennedy-Clinton is said to
illustrate the category of 'Human B-pairs', while in '''a happy end to
a sad chapter''' (p, 52), the B-pair end-chapter illustrates the
category of 'Meronymous B-pairs'. These sub-categories are clearly not
mutually exclusive and it is hard to see what purpose they
serve. Especially interesting in this chapter are the examples of the
way antonymy is often used to create syllepsis with humourous effect:
two concepts from different domains are incongruously linked by the
parallelism inherent in antonymy and by phonological parallelism such
as rhyme, alliteration and syllable count. Examples are ''regarded
male wrestlers as morons and female wrestlers as oxymorons'' (p.52)
and ''it denotes high butter mountains and a low boredom threshold''

Chapter 5 explores 'Coordinated antonymy', that is, antonymous pairs
that signal ''inclusiveness or exhaustiveness'' (p. 37). Coordinated
antonyms designate a whole scale, as in '''.. we cannot prove or
disprove the idea that God is angry''' (p. 62). Instead of sub-
categorizing this type of antonymy, as was attempted for Ancillary
antonymy in the previous chapter, the author here identifies its
typical structural frameworks. These are 'X and Y'; 'X or Y'; 'neither
X nor Y'; 'X as well as Y', where X and Y represent the antonymous

Ancillary antonymy and Coordinated antonymy together account for over
77% of the 3,000 tokens of antonymy in the database. The remaining
tokens were attributed to six 'minor classes', to which chapter 6 is
devoted. Where antonymy occurs in a comparison, such as '''..light
crude is more easily broken down than heavy crude''' (p. 76), it is
said to be 'Comparative'. Where the focus is on the ''inherent
semantic dissimilarity'' of the antonyms, there is 'Distinguished
antonymy'. This typically occurs in frameworks such as '[the
difference/gap] between X and Y', as in '''one must distinguish
between hard and soft drugs''' (p. 83). 'Transitional antonymy' is
the label given to the expression of a change of state, as in
'''..from success to failure''' or '''..moves the gas from the hot
part of the cylinder to the cold''' (p. 85). Antonymy in an
antithetical relation is labelled 'Negated antonymy'. Typical
frameworks are 'X not Y', 'X instead of Y', 'X as opposed to Y': An
example is '''..government must play an active, not passive, role..'''
(p. 88). 'Extreme antonymy' describes antonyms in ''a framework that
unites the outer-most areas of their given semantic scale'' (p. 91),
as in '''.. conditions are either too dry or too wet for racing on
turf..''' (p. 92). Again, typical frameworks are identified. Finally,
a class of 'Idiomatic antonymy' is posited, comprising idioms built on
antonymy, such as 'the long and the short of it', 'easy come, easy
go', agree to disagree' (pp. 93-4). The chapter ends with a
description of some 'residual' occurrences of antonymy that resisted
placement in one of these eight categories.

Chapter 7 assesses the prevalence of antonymy in the 280 million-word
corpus by comparing the actual co-occurrences of the selected
antonymous pairs with the expected co-occurrence rate ''if those words
occurred at random'' (p. 106). It is not clear how these results are
to be interpreted, as there seems to be no way of ensuring that the
two lexemes concerned are actually in an antonymous relationship each
time they occur in the same sentence (or indeed are not in such a
relationship when they occur in adjacent sentences). Statistical
criteria for gauging what constitutes a pair of 'good opposites' are
suggested, but the discussion in this chapter ends rather

Chapter 8 discusses markedness and typical orderings of antonymous
pairs: 'good and bad' rather than 'bad and good'; 'success or failure'
rather than 'failure or success', and so on. It is observed that
positive tends to precede negative, and base forms precede derived
forms. Magnitude, chronology, gender, phonology and idiomaticity are
also posited to play some role, while some orderings are simply
conventional. It would have been useful here to see if there was a
correlation between ordering and 'framework'.

Chapter 9 considers the relationships between the functions of
antonymy and (a) the word class of the antonyms and (b) the
gradability of the antonyms. It concludes that neither word class nor
gradability significantly correlate with the textual functions of

Chapter 10 takes a closer look at some of the structural frameworks
identified in chapters 5 and 6 to see whether, by looking at the
framework and seeing what lexemes appear in the slots, it may be
possible to identify emergent antonyms. Specifically, emergent
antonyms for the target words 'good', 'natural' and 'style' are sought
by retrieving from the corpus the frameworks 'both X and Y', 'between
X and Y' and 'whether X or Y', where X is the target word. For 'good'
and 'natural', opposites predominate in the Y position; e.g. 'flawed',
'nasty', 'wicked' with the former, and 'artificial', 'man-made',
'social' with the latter. The numbers here are too small for
generalizations to be made, but it would be fascinating to see this
method applied to larger and more diversified corpora. The exploration
of all the recurrent frames identified in this study as associated
with antonymy would make an excellent starting point for frame-based
research into lexical relations.

A final chapter summarizes the findings.


Intuition, or convenience, drives many a research design, and it is
always disappointing to encounter hollow post hoc justifications of
data choices or methods. No such disappointment here. The design, data
and method are clearly presented and for the most part admirably well
explained. What is not explained, however, is why the sentence was
chosen as the unit of analysis, nor what its relevance to antonymy is.
Intuitively, it is a strange choice: at best the sentence is a
syntactic unit, while antonymy as defined in this book as a lexical
semantic relation which cuts across syntactic categories and
units. The author's characterizations of the categories he identifies
purport to be semantic, and bear little relation to sentential
structure. Yet the conclusions he arrives at often hang on the choice
of the sentence as the unit of analysis.

For instance, both the 'Ancillary antonymy' pattern and the 'Negated
antonymy' pattern, as described in the text, occur frequently across
sentential boundaries, while the 'Coordinated antonymy' pattern does
not. The quantitative analyses are therefore likely to be severely
distorted. 'Sentence' is not defined in the text, but from the
examples cited, it seems to correspond to a sequence between full
stops. Examples such as ''The bad news is now largely behind; the
good news is to come'' (p.50) are included, while there are no
examples across full stops, such as ''The bad news is that he failed
to reach the second round of the event. The good news is that he is
still alive.'' (The Independent, 09.08.2002). Inter-sentential
antonymy is only mentioned in passing (p. 104). Intuitively, a span
(number of words) designed to maximize recall while retaining enough
precision to make manual sorting feasible would seem to make more
sense. And since antonym use involves a type of reference, it would be
of interest anyway to see at what point diminishing returns set
in. Another qualm is the absence of significance indicators. It would
be nice for readers to be given some indication of the statistical
significance of the results obtained, rather than be left to try to
work it out for themselves from the figures (tricky, given the
sampling method).

Given that, as the author points out, writers often choose an
entrenched antonymous pair for argumentational reasons, it is
disappointing that there is no attempt to examine the information
structure of the sequences containing the antonyms. For instance, the
discussion in chapter 4 of the role of conjunctions in 'Ancillary
antonymy' (pp.57-8) fails to appreciate that the key difference
between the use of 'and' and the use of 'but' to conjoin two
contrasted clauses of the type 'a(x) and/but b(y)' is one of scope:
'and' takes scope over the whole propositions, while 'but' takes scope
over the predicates only. Other things being equal, 'and' thus
produces a symmetrical information structure and 'but' an asymmetrical
one. The examples given throughout the text suggest that many uses of
antonymous pairs are closely tied to asymmetrical information
structure and that information structure may therefore be relevant to
the discussion of antonym sequence in chapter 8.

The final chapter on profiling is by far the most interesting and the
reader is left wishing this framework retrieval had been extended to
include several of the other interesting frameworks identified in
chapters 5 and 6 and had used many more target words from among those
fifty-six selected as having 'good opposites'. Not only does this
procedure test intuitions about 'good antonyms', it can also tell us a
lot about polysemy networks. Indeed, it is a pity that issues of
polysemy and contextual modulation (Cruse 1986: 52), which are key to
understanding the antonymous behaviour of many of the lexeme pairs
selected for this study, are not addressed.

Overall, the text with its wealth of examples amply illustrates that
antonymy does have a special appeal; that it has indeed, as Jones puts
it, ''transcended its role as a mere sense relation'' (p. 1).


Cruse, D. A. (1986) Lexical semantics. Cambridge University Press.


Diana Lewis is an associate professor of English linguistics at the
University of Lyon2. She has research interests in lexical semantics
and pragmatics, language change and variation, corpus linguistics and
contrastive linguistics.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue