we defined antonymy generally as the pair-wise relation of lexical items in context that are understood to be semantically opposite

A brief history of antonymy

Given that antonymy is 'the only sense relation to receive direct lexical recognition ineveryday language' (Cruse 2000:167) , one might expect to find that a robust system ofclassification has emerged. However, though antonymy has been the subj ect ofcontinual linguistic scrutiny, not all analysts have reached the same conclusions.Granted, most commentators categor ise antonymous pairs according to broadlysimilar criteria, but the terminology used to describe these categories is anything butstandard. Furthermore, an adequate definition of antonymy has yet to be agreed upon.Though all speakers can easily re call a lengthy list of 'opposites', describing this relationin a clear, concise fashion has proved problematic. Antonymy is a p henomenon bettersuited to exemplification than definition.

Defining antonymy
The word 'antonymy' was coined in 1867 by C.J. Smith as an opposite of (and byanalogy with) 'synonymy'. It

comes

from the Greek words anti for opposite and onym for name.
Whether antonyms are really the 'opposite' of synonyms isa moot point, especially as it is widely accepted that true synony my does not exist (e.g.Palmer 1976:94), but this does not challenge the validity of the concept. Since 1867,numerous atte mpts have been made to pin down the meaning of antonymy andformulate a workable definition of the term, but the proble m is that antonymy lendsitself more to illustration than description. Good 'opposites' are intuitively available tous all (old/y oung, down/up, lose/win, bad/good, etc.) but finding a definition whichadequately accounts for every example of antonym y is more problematic. In general,two ways of defining antonymy have emerged: the first involves semantic criteria; thesec ond involves lexical criteria. The relative merits of each approach will now beevaluated.

Semantic definitions
The Longman Dictionary of Applied Linguistics (Richards, Platt and Weber 1985)defines an antonym as 'a word which is opposite in meaning to another word'(1985:14). Similarly, in his Dictionary of Linguistics and Phonetics,

Antonymy: A Corpus-Based Perspective (Routledge Advances in Corpus Linguistics) (Steven Jones)

words or word senses are not related to each other directly, word to word, but only by way of their links to common background frames and indications of the manner in which their meanings highlight particular elements of such frames (Fillmore & Atkins 1992: 77)

The antonymy relation is one of the topics which have received extensive and intensive attention in the literature. It has also been proved that antonymy plays a pivotal role in organization of mental lexicon and discourse. Among the various types of antonymy relations, contraries or gradable antonyms account for the vast majority of the antonymous pairs. Also it is widely known that most gradable antonyms show the marked and unmarked distinction. This paper is an attempt to investigate the nature of antonyms from the perspective of semantics and pragmatics.

Antonyms in dictionaries The pervasiveness of antonymy in language and thought manifests itself in a number of ways. Cruse (1986) and Croft & Cruse (2004) point out that antonymy is the most robust among the lexico-semantic relations. Native speakers in all walks of life are intuitively aware of antonymy. It is important in language acquisition and learning (Jones & Murphy, 2005; Murphy & Jones, 2008; Tribushinina, submitted) as well as in lexicography and in lexicographical work targeting language learners (Fellbaum, 1998; Princeton WordNet4, Paradis & Willners, 2007; Storjohann, 2009). Some dictionaries systematically indicate antonyms of headwords that are considered to be frequently associated with a partner. One such dictionary is Collins Cobuild Advanced Learners English Dictionary (2003). The treatment of antonyms in Collins Cobuild Advanced Learners English Dictionary has been examined in a study by Paradis & Willners (2007). The rationale behind our investigation is that dictionaries in general and learners dictionaries in particular are important tools in the process of acquiring foreign languages and that the main goal of a learners dictionary informed by a large corpus, such as Collins Cobuild Advanced Learners English Dictionary, is to provide learners with idiomatic and useful information that will help them set up native-like links between words and meanings. Our investigation shows that, as expected, far from all headwords are provided with an antonym. The majority of headwords for which antonyms are given in the dictionary are adjectives (59%) followed by nouns (19%), verbs (13%) and adverbs and prepositions (9%). Most of the adjectival pairings are gradable, either UNBOUNDED meanings expressing a range on a SCALE such as goodbad, or BOUNDED meanings expressing totality, such as deadalive (as defined and developed in Paradis, 1997, 2001, 2008), but there are also non-gradable antonymous adjectives such as male female. An important observation in the study is that there are clear similarities across the meanings of the headwords for which antonyms are provided in the dictionary in that they are all what we might call inherently binary (we return to this issue in Section 6).

Murphy (2003, p. 45) argues for a general pragmatic principle, the Relation-by-Contrast, governing all semantic relations, i.e. antonymy, synonymy, hyponymy and meronymy. The principle defines relations on the basis of minimal differences. For antonyms the contrast relation holds among the members of a pair if and only if they have the same contextually relevant properties but one. This definition offers a semantic explanation for antonymy in language and thought and offers an account for why some lexical items form set couplling. Antonymy is characterzed as a binary contrast in which the two formmeaning pairings are used as opposites. This binary opposition is effected through a construal of comparison in which a contentful dimension (x) is divided by a BOUNDED configuration and the meanings on either side of the boundaries are used and understood as opposites.

Antonymy and negation Antonymy is a phenomenna that has been studied from many angles. In the literature, antonymy is recognized as the most robust of the lexico-semantic relations, important to both the mental organization of the vocabulary and the organization of coherent discourse (Cruse, 1986; Muehleisen, 1997; Fellbaum, 1998; Willners, 2001; Jones, 2002; Murphy, 2003; Croft and Cruse, 2004). Negation has been studied by philosophers, logicians, psychologists and linguists for a long time (for an overview, see Horn and Kato, 2000). Depending on whom you ask, negation may be a logical operator or a type of speech act, a basic element of semantic representations or a pragmatically loaded form of communicative interaction (Israel,2004:701). If we disregard the complexity of scopal properties, negation has a relatively straightforward role as a polarity item in the language of logic, in formal linguistic systems and literalist approaches to meaning. In natural language usage, however, the role of negation is anything but straightforward (Giora, this issue). Not only is not used as an expression of polarity and denial in common parlance as in I dont want to go to the cinema tonight or Sarah is not a music lover, but it is also frequently used as a hedging device in discourse as in I dont know, but I think its a good idea to go swimming, where no real sense of polarity and denial is conveyed (Tottie and Paradis, 1982; Tottie, 1991) or a modier of degree as in the water is not hot said about water that may be warm, lukewarm or cool (Bolinger, 1972; Horn, 1989; Israel, 2001; Giora et al., 2005a).

This article presents a new look at a very old topic. At the time of Aristotle, there was already a long-standing tradition of using opposites in both philosophy and rhetoric. Yet, Aristotle is considered to be the first thinker to classify and to provide a systematic account of opposites (Ackrill 1963: 31-38, Lloyd 1966: 1171). Oppositeness, or antonymy, which is the term used here, is a well-established notion in everyday contexts, as well as in many academic disciplines. It is a powerful relation in both thought and language, important to coherence in discourse as well as to the conceptualization of meaning through language and other modalities (Murphy, Paradis & Willners 2009). Characteristic of antonyms is that they share an important segment of meaning at the same time as they differ prominently along the same dimension (e.g. Cruse 1986). Antonymy comes in different guises in linguistic communication. At the one extreme, it shows up as conventionalized antonym pairs such as goodbad, heavylight, hotcold and slow

fast. At the other extreme, antonymy may be construed for purposes of originality or poetic effect, e.g. The most beautiful things are those that madness prompts and reason writes. (Andr Gide), Timid men prefer the calm of despotism to the tempestuous sea of liberty. (Thomas Jefferson) or A joke is a very serious thing. (Winston Churchill). In between those two extremes, there are numerous pairings which language users consider to be less good pairings, e.g. I prefer calm dogs to high-strung dogs, I prefer calm waters to flowing waters, I prefer a calm public to an agitated public and I prefer calm conversation to flame warring, where the various different antonyms of calm are more clearly bound up with highly specific domains and situations (Murphy & Andrew 1993; Paradis 2005). For decades, research on antonymy was tied up with the structuralist approach to meaning as a system of relations between words, leaving it separated from new insights into the dynamics of conceptual representations, and thus invisible to new observational techniques in linguistic research. With the growing sophistication of Cognitive Semantics and the development of computational facilities and experimental techniques, the foundation for research on antonymy has considerably improved. Based on a combination of a series of recent textual and experimental investigations (Jones et al. 2007; Murphy et al. 2009; Paradis et al. 2009; Willners & Paradis 2010), this paper offers a new take on antonymy as a linguistic category both by configuration (or schema) and by content. The primary goal is to propose a dynamic usage-based theoretical account for the category of antonymy that is capable of accommodating all kinds of antonym construals ranging from highly conventionalized lexico-semantic couplings to strongly contextually motivated pairings. The theoretical approach adopted is broadly that of Cognitive Semantics (Talmy 2000, Taylor 2003, Croft & Cruse 2004, Paradis 2005). Two questions are at the heart of the paper: What is antonymy? What are the differences between good antonyms, better antonyms and superb antonyms?

The structuralist position The basic assumption of the nature of meaning in Saussurean Structuralism, is that every language is a unique relational system in which words receive their meanings from their relationships with other words in the same language system, i.e. meanings are not substantive but relational. This means that a word does not have an independent existence but derives its meaning from its position in a linguistic network. This also means that the sense of a word is the set of sense relations the word has with other words in the same lexical field. Within this approach, language is regarded as an autonomous, self-contained system of paradigmatic and syntagmatic relations between words. Language is compared to a game of chess with rules and values (langue) and alternatives (parole). Langue is thus the language system that underlies the language put to use in a certain language community. Lyons (1977: 239240) makes use of the notions language system and language behaviour respectively. He points out that Saussures doctrine of the language system is not entirely clear. On the contrary, it has given rise to a lot of controversy in the literature. Saussure emphasized the supra-individual and social nature of the language system; and yet he held the view that it also had psychological validity in being stored

in peoples brains. Not all structuralists conceived of the nature of meaning as Saussure did, but rather as a reflection of the external world. Lyons goes on to say that linguists working within the structuralist framework argue about whether there is an underlying universal system or not, and many will deny that this system is internalized. What is clear is that the structuralist focus is on language as an externalized object and not on something in the mind of the language users. Lexico-semantic relations are of two types: paradigmatic and syntagmatic relations. Lyons (1977: 270317) is mainly concerned with lexico-semantic relations from a paradigmatic point of view, focusing on relations such as antonymy, synonymy and hyponymy. Paradigmatic relations hold between lexical items which are intersubstitutable in a given position in a syntagm. In contrast, the syntagmatic approach, or the contextual/usage approach, defines the meaning of a word as its uses across its grammatical occurrences (Firth 1957; Cruse 1986; Sinclair 1987). The Firthian dictum has it that You shall know the meaning of a word by the company it keeps and collocation is a key notion. Cruse (1986), which is an important piece of work in the structuralist approach to word meaning and sense relations, is a cross between the two approaches. The individual chapters of his book are devoted to the two different types of sense relations, but he explicitly says that the approach adopted is a variety of the contextual approach (1986: 1). Cruse assumes that the semantic properties of lexical items are fully reflected in the relations they contract within actual and potential contexts, semantic as well as grammatical. The main disadvantage of the structuralist approach to meaning is that it is a static system where words have (a) set meaning in the relational network. The theory does not provide any tools for explanations of lexical flexibility in how antonymy is used in language, as illustrated by the different examples in the introduction and, moreover, there is a want of empirical data in most of the literature on lexico-semantic relations in the structuralist tradition. It has become increasingly clear that it is of the utmost importance for a theory of semantics to be able to account for the flexibility of lexical meaning in language use and meaning making. Two questions are at the heart of this analysis. They concern (i) what antonymy is, and (ii) what the differences between good, better and superb antonyms are. In answer to the first question concerning the nature of antonymy, I argue that antonymy is a binary opposition of some content. Antonymy is a binary construal of comparison in which the contentful dimension is divided by a BOUNDED configuration. The configuration of BOUNDEDNESS constitutes an absolute and necessary requirement for meanings in a certain content segment to be used as antonyms (irrespective of whether the configuration of the opposing elements against which the contrast is profiled is BOUNDED or UNBOUNDED). In contrast to this definition of antonymy by configuration (or schema), the categorization of antonymy by contentful meaning structures forms a continuum in that some pairings are better pairings of binary opposition than others. The structure of the category of antonymy from the point of view of the content segment is one of prototypicality, with canonical pairings as core members and ad hoc couplings on the outskirts (Paradis et al. 2009). In answer to the second question, the article attempts to untangle the question of the nature of the meanings of the more canonical, conventionalized antonym pairings as opposed to all other pairings that have no such obvious partners, and it does so by appealing to their ontological set-up in terms of the salience of the contentful dimension as well as the configuration of the members that form part of the opposition, the symmetry of the members of the antonymy in relation to the BOUNDARY between the members of the pairs, contextual versatility and frequency. It is my contention that lexical items that are strongly coupled with an antonym partner in language as shown through different empirical indicators are all inherently

binary along a salient contentful segment.

The cognitive position A leading principle in Cognitive Semantics is that meanings are mental entities in conceptual space. Meanings are in peoples minds rather than relations within language, as in Structuralism. Words and expressions do not have meanings but are cues for making inferences that promote adequate reasoning and understanding (Verhagen 2005: 22). Words and expressions evoke specific profilings of conceptual structures when they are used in text and discourse. Lexical meaning is the relation between the relevant parts of the total meaning potential of words and expressions. Lexical meaning is constrained by encyclopaedic knowledge, conventionalized mappings between lexical items and concepts and conventional modes of thought in different contexts and situational frames (Cruse 2002; Paradis 2003, 2005). According to this view, meanings of words are always negotiated and get their definite interpretations in the specific context where they are used.

Antonyms are words belonging to the same part of speech, identical in style,expressing contrary or contradictory notions. V.N. Comissarov classified antonyms into two groups: absolute (root)antonyms (late - early) and derivational antonyms (to please to displease, honest dishonest). Absolute antonyms have different roots and derivational antonyms have the same roots but different affixes. In most cases negative prefixes form antonyms (un-, dis- non-). Sometimes they are formed by means of antonymous suffixes: -ful and less (painful - painless). The difference between derivational and root antonyms is also in their semantics. Derivational antonyms express contradictory notions, one of them excludes the other: active-inactive. Absolute antonyms express contrary notions. If some notions can be arranged in a group of more than two members, the most distant members of the group will be absolute antonyms: ugly, plain, good-looking, pretty, beautiful, the antonyms are ugly and beautiful. Leonard Lipka in the book Outline of English Lexicology describes three types of oppositeness: a) complementarity: male female. The denial of the one implies the assertion of the other, and vice versa; b) antonyms: good bad. It is based on different logical relationships; c) converseness: to buy to sell. It is mirror-image relations or functions:husband-wife, above-below, pupil-teacher.

L. Lipka also gives the type which he calls directional oppositions: up-down,consequence opposition: learn-know, antipodal opposition: North-South, East-West. L. Lipka also points out non-binary contrast or many-member lexical sets. In such sets of words we can have outer and inner pairs of antonyms: excellent, good,average, fair, poor. Not every word in a language can have antonyms. This type of opposition can be met in qualitative adjectives and their derivatives: beautiful-ugly, to beautify-to uglify. It can be also met in words denoting feelings and states: to respect-to scorn, respectful-scornful and in words denoting direction in space and time: here-there, up-down, before-after. If a word is polysemantic, it can have several antonyms, e.g. the word bright has the antonyms dim, dull, sad.

Antonym- a word that expresses a meaning opposed to the meaning of another word, in which case the two words are antonyms of each other. Antonyms - words of the same category of parts of speech which have contrasting meanings such as hat - cold, light dark, happiness - sorrow. Morphological classification: -Root words form absolute antonyms.(write - wrong). -The presence of negative affixes creates - derivational antonyms(happy - unhappy). Semantical classification: Contradictory notions are mutually opposed and denying one another, i.e. alive means not dead and impatient means not patient. Contrary notions are also mutually opposed but they are gradable; e.g. old and young are the most distant elements of a series like: old - middle - aged - young.

Incompatibles semantic relations of incompatibility exist among the antonyms with the common component of meaning and may be described as the relations of exclusion but not of contradiction: to say morning is to say not afternoon, not evening, not night.

Antonymy is unique among lexical semantic relations in that it requires one to-one relations, rather than one-to-many or many-to-many. We can observe this in the different ways we talk about antonymy and synonymy in everyday English. (1) Whats the opposite of interesting? (2) Whats a synonym for interesting? While question (1) presupposes a unique opposite, (2) allows for more than one answer. Within the Corpus of Contemporary American English (Davies 2008), the opposite of occurs 1,344 times but an opposite of only twice. On the other hand, the synonym for occurs only 4 times but a synonym for occurs 189 times. This peculiar binarity of antonymy means that some of the bestexamples of the relation are those that either belong to semantic sets that naturally have only two members or are the polar categories of something (a dimension, an object, an event) that can be described in terms of a scalar dimension. An example of the two-member-set type is femalemale the only sexes for which English has well-known names. In the polarity case,we have adjectives that describe scalar dimensions (shorttall, earlylate) and the poles of things or events in space or time (headfoot, startfinish). But while these kinds of naturally binary sets provide some of the clearest examples of antonymy, it is not sufficient to say that the existence of antonymy can be explained solely by the existence of binary sets and semantic dimensions with poles. This is because such an observation would not explain why two particular words form an (or the) antonym pair for a particular dimension/semantic field when other available synonyms are available (e.g. largesmall rather than largelittle), nor would it explain why some pairs are preferred over others in multidimensional semantic fields, such as taste (sweetsour or sweetbitter but not sourbitter) or emotion (happysad but not happy afraid). It has been established that, unlike for other relation types, people have strong intuitions that various types of opposite relation all count under an umbrella category of opposite words or antonyms (Chaffin and Herrmann 1984) and that antonym relations are mastered earlier in our metalinguistic development than synonym relations (Heidenheimer 1978). Murphy (2003:169) goes so far as to say that antonymy is arguably the archetypical lexical semantic relation. It is no surprise, then, that the advent of corpus linguistics has inspired a number of publications about antonyms and the antonym relation. Some of these (e.g. Mettinger 1994, Willners 2001, Jones 2002) have investigated the types of contexts in which antonyms typically co-occur in text, and some (e.g. Jeffries 2010, Storjohann 2010b) have considered the role of contextual properties in allowing for the construal of novel antonym relations. Other work (e.g. Paradis 2001, Murphy 2003, Croft and Cruse 2004) has remained on a more theoretical plane, emphasizing the context-dependence of antonym relations, in contrast with earlier Structuralist work. This book bridges the gaps between the theoretical and the empirical, the more entrenched and the more creative uses of antonym pairs, and the paradigmatic and syntagmatic aspects of antonymy. Using a variety of textual and psycholinguistic evidence, we build up a theoretical picture of the antonym relation: how antonym pairings are semantically and contextually licensed and how they are stored and/or derived in speakers minds. We discuss key contributions to the study of antonyms, moving from Aristotle to present-day perspectives, such

as Relation by Contrast (Murphy 2003) and the Cognitive Construal Approach (Croft and Cruse 2004). As we discuss each of these, we highlight unanswered questions and unsolved problems that deserve further investigation.

1.2 Defining antonymy and oppositeness We use antonymy to refer to the pair-wise relation of lexical items in context that are understood to be semantically opposite. Much of our work relies on the notion that antonym pairs can be judged to be better or worse exemplars of the category for semantic, pragmatic, or form-related reasons. While we hold that antonymy is context-driven and available to a broad range of lexical pairings, particular emphasis should be placed on conventionalized pairings, also known as canonical antonyms (following Murphy 2003) that is, pairs forming part of an antonym canon that is learnt through experience of the language. We use the term opposite to refer to the semantic relation between antonym pairs that is, antonyms are understood to have meanings that are opposed to one another in a given context. Factors that contribute to particularly good antonym pairings may relate to more than just the two items semantic oppositeness; for instance, the pairing of increase and decrease is supported by their rhyme and the perception of a parallel morphology, as well as their semantic opposition. Thus, when we speak about antonymy, we are speaking about the lexical and discourse instantiation of oppositeness, as well as antonym pairs stored in a language users memory. One could define oppositeness in terms of logical incompatibility that is, if a thing can be described by one of the members of an antonym pair, it is impossible for it to be described by the other. So, if a person is a man, he is not also a woman. If a piece of string is long with reference to some contextual standard, it cannot also be short with reference to the same standard. But logical incompatibility is an insufficient criterion for defining oppositeness, since many pairs of lexemes are semantically or logically incompatible, but this does not lead to their use as antonyms. So, while it is unlikely for something to be both a limerick and a pencil, this is not reason enough to think of limerick and pencil as opposites. The reason that limerick and pencil are unlikely to be construed as antonyms is that semantic opposition involves similarities as well as differences, and these two words are not similar enough. On the other hand, black and white are readily construed as antonyms because (a) they are incompatible, in that they cannot refer to the same colour, and (b) white shares with black more properties that are relevant to linguistic-semantic opposition than other possible antonyms for black, and vice versa, in that black and white are the only two basic colour terms that refer to unmixed, achromatic colours. This principle of minimal difference between members of antonym pairs has long been noted in the literature (e.g. Clark 1970, Hale 1971, Lyons 1977, Cruse 1986, Murphy 2003). Describing antonymy in terms of maximal similarity and minimal difference means that words may have different antonyms in different contexts depending on which of the words properties are most relevant to contrast within a particular context of use. Consider, for example, sentences (3) and (4), taken from British newspapers, which are discussed as contextual opposites in Jeffries (2010:7980): (3) The evil genius behind the strategy that has turned the party from unelectable to unstoppable in 10 years.

(4) Let the professionals remember that the politicians that the public likes best are not the aloof ones but the human ones. In (3), unelectable and unstoppable, two words that derive their minimal difference in part from being morphologically alike, are placed into a frame (from X to Y) in which antonyms are regularly found. In (4), the opposition between a pair of non-canonical antonyms (aloofhuman) is accentuated by a contrastive conjunction and the parallelism of the [X/Y] ones. These examples demonstrate that, though some pairs can be described as canonical antonyms, any opposition can be licensed within an appropriate context. Our approach to antonymy makes no particular claims about the types of words or semantic structures that are contrasted in an antonym pair. As such, we use the term antonym to apply to any relation of lexical oppositeness, in contrast to some theorists (e.g. Cruse, Lyons, Lehrer), who have restricted the use of this term to certain types of opposition (particularly, contrariety). So, downup, hatelove, manwoman, and northsouth are antonyms, as are alivedead, longshort, happysad, and so forth. Common adjectives have antonyms more often than common nouns or verbs do. For example, adjectives constitute the majority of headwords for which Collins COBUILD Advanced Learners English Dictionary (CCALED) lists antonyms (Paradis and Willners 2007), as shown in Table 1.1. Adjectives affinity for antonymy can be attributed to their relative semantic simplicity. They often describe a single property that can be had to greater or lesser degrees as opposed to the complex conglomerations of properties that many nouns typically represent and the temporal and argument-structure complexity of verbs. To illustrate this point, Table 1.2 shows the five most common verbs, nouns, and adjectives in English according to the Oxford English Corpus (OEC). Most frequent English words and their canonical antonyms Verbs Nouns Adjectives OEC frequency ranking Antonym [substitutability]? 1. be 1. time 1. good bad [95%] 2. have 2. person 2. new old [99%] 3. do undo [2%] 3. year 3. first last [90%] 4. say 4. way 4. last first [50%] 5. get 5. day night [64%] 5. long short [74%]

As Table 1.2 shows, all of the top five adjectives have unambiguous, conventionalized antonyms, whereas we can identify conventional antonyms for only one noun and arguably one verb in this list. Where antonyms are available for nouns or verbs, they are not as available across contexts as the adjectival antonyms. This is demonstrated by the percentage figures in Table 1.2, which show that the contexts for the top adjectives can usually support substitution of a single antonym, whereas this is not the case for most of the top nouns and verbs. The numbers in Table 1.2 were arrived at by searching for the term in the British National Corpus (BYUBNC, Davies 2004), then testing a random sample of 100 sentences to see if the proposed antonym would be grammatically and idiomatically substitutable. Though substitutability alone is not an indicator of antonymy, it is a good measure of semantic similarity. On a minimal difference definition of oppositeness, the substitutability of the adjectival antonyms indicates that they have a better fit as potential antonyms (i.e. have fewer differences) than the noun and verb pairs in the table. So, for instance, sentences containing good, such as those in (5) (from the BNC data), would be equally grammatical and interpretable if bad had been substituted for good, and that was the case for 95 of the 100 sentences sampled. On the other hand, around one third of the cases of day could not be substituted by night, as illustrated in (6) and almost none of the cases of do could sensibly be replaced by undo, as shown in (7). (5) a. Still, me ears aint as good (bad) as they was. b. Many had had a good (bad) war c. children are not a good (bad) investment (6) a. she could have eaten it all day (night) long b. I bought her those the day (night) before she died c. He was justly celebrated in his day (*night) as a populariser of science (7) a. How do (*undo) you know about the state hes in? b. I did (*undid) very well c. it was only natural to do (*undo) so. The closeness of the antonyms meanings is a contributing factor to canonicity. As well as being a diagnostic for minimal semantic difference, substitutability may also contribute to strengthening canonical relations because antonym pairs become conventionalized through their co occurrence in text. That co-occurrence is facilitated by contextual parallelism (recall examples (3) and (4)) which is only possible when the two members of the pair are substitutable in terms of their grammatical and collocational properties. All of this is not to say that there are more adjectival antonyms in absolute terms than nominal antonyms, nor that adjectival antonyms are more commonly used in text. Lobanova et al. (2010), for example, found more examples of cooccurring nominal antonym pairs than adjective ones in a Dutch corpus. It is to say, however, that antonym relations are more central to the adjective classes than to other classes. Other

symptoms of the antonymadjective correlation are the prevalence of antonym responses for adjectives in word association tests (Deese 1965) and in dictionary and thesaurus entries for adjectives. This has led some (e.g. WordNet: Miller 1990, Fellbaum 1998) to posit that the adjectival lexicon is unlike the nominal and verbal lexicons in being organized primarily by semantic opposition. Since we do not start from an assumption of an organized lexicon (see below), we do not assume that the antonym relation is different for adjectives than for other word classes. Instead, it is our intention to address the semantic roots of the adjectiveantonym correlation and to explore the textual symptoms of such central members of the antonym category. 1.3 Key perspectives on antonymy and opposition Having determined what antonymy is, the next step is to explore some of the ways in which it has been handled in the literature. Each of the following subsections, therefore, introduces a different theoretical perspective on antonymy, notes its influence, and points towards ways in which this book incorporates the approach or seeks to develop it further. 1.3.1 Classical and Structuralist perspectives Much modern thinking about the antonym relation harks back to categories of propositional opposition devised by Aristotle (Ackrill 1963, Lloyd 1966), who created a diagrammatic representation (the square of opposition) for universal and particularized affirmations and negations (Every S is P; No S is P, etc.). This diagram introduced a range of terminology, including contradictory and contrary, that have been adopted in more modern philosophical and linguistic approaches, including Structuralist and Generativist (e.g. Bierwisch 1989) approaches. Work on relations at the lexico-semantic level are of key importance to the ideas within Structuralism, more precisely within European, or Saussurean, Structuralism. At the most general level, the central thesis of the approach is that every language is a unique relational system (Lyons 1977), the units of which derive their meanings from their relationships with other words (Saussure 1959), who is among the most prominent figures in Structuralist semantics, points out that Saussures doctrine of the language system has given rise to controversy because his published writings are not entirely clear about questions related to the basis of meaning in language. On the one hand, Saussure emphasized the supra-individual and social nature of the language system; however, on the other hand, he also held the view that the system has psychological reality, as it is represented in the minds of individual language users. For Lyons (1977), meaning is a system of relations between words. He contrasts his own position to that of Trier (1931), who claimed that every word calls forth its opposite in peoples consciousness. Lyons does not express an opinion about whether oppositeness in the linguistic system is caused by a universal dichotomizing tendency or whether it is due to the pre-existence of a large number of opposed lexemes in language, but says that it is a fact of which the linguist must take cognizance, that binary opposition is one of the most important principles governing the structure of languages; and the most evident manifestation of this principle is antonymy (Lyons 1977). The relation is taken as a primitive and a universal principle by other writers in the Structuralist tradition too for example, Lehrer and Lehrer (1982) and Cruse (1986) divide opposites into four main types: Complementaries comprise pairs that in their default interpretations

exhaustively bisect a domain into two sub-domains, as for alivedead, closedopen, falsetrue. Contraries denote degrees of some property, e.g. fastslow, longshort, thickthin. Structuralists typically reserve the term antonym for members of this subcategory. Reversives denote change in opposite directions between two states, as in dressundress, fallrise. Converses denote two opposed perspectives on a relationship or transfer for example, buysell, childparent.

Lexico-semantic relations in the Structuralist framework are of two fundamental types: they are either paradigmatic or syntagmatic relations. A paradigmatic relation is one in which the related words constitute a set of potentially substitutable expressions, including antonymy, synonymy, and hyponymy. A paradigmatic approach to lexical relations (e.g. Lyons 1977) is therefore one that focuses on the semantic properties that define such sets. Syntagmatic relations are relations of collocation and co-occurrence. A syntagmatic (or contextual/use) approach describes the meaning of a word as its uses across contexts (Firth 1957, Sinclair 1987), hence the Firthian dictum You shall know the meaning of a word by the company it keeps. Cruses 1986 approach to lexico-semantic relations is a cross between the syntagmatic and the paradigmatic. While much of his research is devoted to paradigmatic relations, he also points out that it is impossible to ignore the syntagm: The two types of relation have their own distinctive significance. Paradigmatic relations, for the most part, reflect the way infinitely and continuously varied experienced reality is apprehended and controlled through being categorised, subcategorised and graded along specific dimensions of variation. They represent systems of choices a speaker faces when encoding his message. Syntagmatic aspects of lexical meaning, on the other hand, serve discourse cohesion, adding necessary informational redundancy to the message, at the same time controlling the semantic contribution of individual utterance elements. (Cruse 1986:86) In this book we expand on some aspects of the agenda set in Cruses contextualist account. In particular, we explore antonym co-occurrence in texts and what it reveals about the nature of the relation, both in language and in language users minds. Where we need vocabulary for types of semantic opposition, we make use of the terminology from Cruse (1986). Structuralist approaches treat relations as stable properties between words, but, since the latter part of the past century, linguistics has witnessed a widespread reaction against seeing language as a stable system of contrasts within which we make choices. With the development of new theoretical insights in language and cognition and new computational methods, research on lexical relations has experienced a revival and has seen further development. However, the notion that lexical meaning is organized according to stable relations among words is maintained in the now-global WordNet project (Miller 1990, Fellbaum 1998), which represents lexicalsemantic knowledge as networks of links between word senses. These

links are labelled by relational type (e.g. synonym, antonym, hyponym), and the antonym relation is presented as the primary means of organizing the adjective lexicon and as a minor relation for other word classes. Antonym relations in WordNet can be either direct, in which case there is an antonym link between two word senses, or indirect, in which case one or more of their synonyms are direct antonyms. The pairs that WordNet scholars label direct bear a superficial resemblance to those that we characterize as canonical. However, the term canonical, unlike direct, does not imply a particular model of lexical organization. 1.3.2 Relation by Contrast Murphy (2003) develops a theoretical model (begun in Murphy 1995, 2000) in which antonym relations and indeed all paradigmatic lexical relations obtain between words in use. As such, it presents an explicit argument against the position that lexico-semantic relations are central to the organization of the lexicon. Taking the example of black and white, then, Murphy claims that there is no need to represent the knowledge that they are antonyms in the lexicon, since their opposition is predictable from a pragmatic principle of minimal difference, which she terms Relation by Contrast. Different types of semantic relation arise through different applications of relation by contrast that specify the nature of the contrasting property. Antonymy is categorized as a binary realization of a more general relation of lexical contrast. (9) Relation by Contrast Lexical Contrast (RCLC) A lexical contrast set includes only word-concepts that have all the same contextually relevant properties but one (Murphy 2003). The key difference between this approach and semantic approaches that make reference to minimal difference is that RC does not refer to particularly semantic properties of the contrasted words. Instead, it holds that lexical relations are metalexical, rather than represented in the lexicon. Relations obtain between word-concepts that is, conceptual knowledge about words, rather than lexical or semantic representation of the words. This means that the goodness of fit between two members of an antonym pair can rely on properties of the words (or their use) other than (as well as including) their semantic properties. Thus, collocational preferences, morphological properties, rhyme, alliteration, connotation, social register, and so forth can come into play in judging word pairs as related (e.g. antonymous) or not, as well as the particular communicative demands of the context. The requirement that the related word-concepts be as similar as possible in contextually-relevant properties determines the semantic nature of the relation, since it would be extremely rare that meaning was less relevant to a communicative context than other aspects of words, such as morphological complexity or rhyme. Because the Relation-by-Contrast approach does not rely on any particular theory of semantic representation, it is consistent with any theory for which semantic relatedness itself is not a determinant of meaning. As a pragmatic approach, it focuses on the ways in which antonym relations are derived in contexts of use, offering ample evidence that antonym choice is context-sensitive. However, Murphy acknowledges the psycholinguistic evidence for an antonym canon which entails knowledge of antonym pairings, not just knowledge that allows the derivation of antonym pairings. She argues that the psycholinguistic methodologies for determining relational entrenchment are evidence for metalexical, rather than intralexical, knowledge and processes, and therefore the knowledge that, say, black is the

antonym of white is recorded in the conceptual representation of knowledge about the words, rather than lexical knowledge that contributes directly to the formation of grammatical and sensible utterances. At the same time, however, there is (a) textual evidence for canonical antonymy (that is, co-occurrence of pairs across a range of frames known to associate closely with antonyms see Chapter 2) and (b) the propensity for canonical antonym substitution in speech errors (see Hotopf 1980). Both of these contribute evidence that preference for canonical pairings is not just present in the artificial experiments, but also in on-line utterance generation. In order to reconcile the position that the mental lexicon is not semantically organized with the evidence that canonical relations influence utterance construction in terms of co-occurrence and collocation, Murphy 2003 has to propose a roundabout way of involving word-concepts as well as words themselves in utterance generation. The present book develops and improves upon Murphys 2003 account by (a) providing further evidence for the antonym canon and the means for measuring antonym canonicity, (b) adopting an adaptation to Construction Grammar (following Murphy 2006) that allows for linguistic representation of paradigmatic relations without proposing fixed semantic relations in the lexicon, and (c) paying attention to the semantic detail that allows generation of antonym pairings with minimal relevant difference as per Relation by Contrast.

The notion of oppositeness (antonymy) The observation that words frequently evoke words that denote their opposite meaning can be accounted for by the notion of conceptual frame or domain. In order to understand the meaning of a word it is helpful or even necessary (though certainly not sufficient) to know what its antonym is. Lexicographers and language teachers alike have had this insight for a long time and have applied it in the compilation of dictionaries and integrated it into language teaching. One may thus conclude that an important feature of conceptual frames is that they contain, apart from an account of the meaning proper of the lexical item in question, information about opposite concepts. The term antonymy is used in a broad and a narrow sense. The narrow sense restricts antonymy to binary opposites (contradictories) such as same different, single married, dead alive, pass fail (a test), and polar opposites (contraries), typically exemplified by gradable adjective pairs like young old, good bad, wide narrow. Binary antonyms have specific logical properties: they are neither both true nor both false of a thing (of the right category). For example, animate beings cannot be both dead and alive, nor can they be neither dead nor alive. Polar antonyms cannot both be true of the same thing, but they may both be false. Thus, although a person cannot be both young and old, s/he can be neither young nor old. In other words, polar antonyms involve scales with intermediate values. In this chapter, we understand antonymy in the broad sense. Apart from binary and polar opposites, we include multiple incompatibilities (e.g. spring summer fall winter), converse opposites (e.g. buy sell, parent child), and reverse opposites (e.g. push pull) in the category of antonyms. Given the notion of frame, this is a natural consequence. For example, as is well known, in order to understand what buy means, the opposite (converse) notion of selling is crucial and must be incorporated into the COMMERCIAL EXCHANGE frame. We can now characterize ideal antonymy as follows:

(1) Two lexical items are antonyms if a. they correspond to one of the types of antonymy mentioned above, and b. they are formally substitutable for each other in a construction without resulting in ungrammaticality. Criterion (1a)the existence of word pairs that stand in a relation of oppositionis (obviously) often fulfilled. It is more difficult to find word pairs that satisfy criterion (1b) since the substitution of a lexical item by its antonym may entail changes in argument structure, or, at least, in the way that arguments are coded (by NPs, varying PPs, etc.). Ideal antonymy is therefore relatively rare. In what follows we assume a somewhat looser conception of antonymy that fulfills criterion (1a) given above, neglecting the formal criterion (1b).

Antonymy for rhetorical purposes in language use it happens not infrequently that words are used by speakers in a sense opposed to their conventional meaning, in order to achieve certain rhetorical effects. Typical examples are utterances like the following: (8) a. Boy, this food is terrific! (Akmajian et al. 2001: 378; italics ours) b. That argument is a real winner. (Akmajian et al. 2001: 378; italics ours) In (8a,b) the meanings of terrific and winner are, via conversational implicature, turned into their antonyms, viz. terrible and loser, respectively. Sometimes, such ironic speech acts becom e completely conventionalized: (9) a. You are a fine friend. You are a bad friend4 b. We are in a nice mess. We are in a bad situation In other cases, the auto-antonymous senses belong to different registers. In particular, in certain subcultural contexts, lexemes might be used with a meaning diametrically opposed to their sense in the standard language. Vohagen (1999) investigates such uses and proposes treating them as metonymies. Often-cited examples are bad good, wicked excellent, pretty in e.g. pretty ear deformed ear, cauliflower ear.

The next major kind of antonymy is that of Converses, which denote relational antonymsregarding relationships between things, states, and times; for example, tenant andlandlord, above and below, before and after. Kempson tells us that A pair of lexicalitems form a converse pair if for two items x and y, the following sentence relations hold: AxB implies ByA, and AyB implies BxA . (Kempson, 1977, p. 85). For example, X is thelandlord of Y entails and is entailed by Y is the tenant of X (Jones, 2002, p. 17).

The more restrictive the categories become, the more tenuous grows the link between the technical term antonymy and the universally recognised concept of opposites. (Jones,2002, p. 20).

Through studying the lexical semantics of antonymy and the other sense relations, I came across a reoccurring problem in the literature; while the linguists seek to describe and definethe antonymy sense relation, they are restricted when it comes to inquiring where antonymyis derived from in the first place and why it is such an important element of our language. For example, in his 1977 work Semantics, John Lyons says that:We can leave to others to enquire whether the tendency to think in opposites, tocategorize experience in terms of binary contrasts, is a universal human tendencywhich is but secondarily reflected in language, as cause producing effect, or whetherit is the pre-existence of a large number of opposed pairs of lexemes in our nativelanguage which causes us to dichotomize, or polarize, our judgements and experiences. It is, however, a fact, of which the linguist must take cognizance, thatbinary opposition is one of the most important principles governing the structure of languages; and the most evident manifestation of this principle, as far as thevocabulary is concerned, is antonymy. (Lyons, 1977, p. 271).This problem is echoed in a similar fashion by Steven Jones in his 2002 work Antonymy. He says that:the integral position held by antonymy in the mental lexicon is something of achicken-and-egg situation. It is almost impossible to know whether language simplyreflects existing oppositions in the outside world or whether we, as humans, aresubconsciously predisposed to impose such dichotomies. Whatever the cause, theconsequence is that opposites hold a key place in language and this is reflected bythe pull of antonymy to learners of a first or other language. (Jones, 2002, p. 3). Is it a universal human tendency OR the pre-existence of opposed pairs in ournative language; which causes us to to think in opposites, to categorize experience interms of binary contraststo dichotomize, or polarize our judgements andexperiences (Lyons). Jones does say later that: Language reflects the urgewe have to dissect, an urge which may be related to the fact that we are all either female or male ., which seems to point to language reflecting, or in some way being caused by, naturaloppositions in the world; and also that: Humans seem prone to organise the world in termsof oppositions; antonymy is simply a linguistic reflection of this., which seems to point tothe subconscious conditioning of language and the world (Jones, 2002, p. 169). The second theory is called Linguistic Relativity but is also known as Whorfianism(named after its most famous proponent Benjamin Lee Whorf). The basic position of Linguistic Relativity is that the structural properties of our language and/or its categoriesdetermine, condition, or influence, how we conceptualise or even perceive the world. This isusually employed to discuss the affect that the different languages spoken by people indifferent cultures, do or do not have on their conceptualisation or perception of reality.(Lyons, 1977, p. 245f). A particular type of linguistic relativity may be devised that assertsthat antonymy imposes dichotomisation on thought and the world. This is somewhat in tunewith the human tendency

The argument for antonym pairs as constructions involves canonical antonym pairs (Murphy 2003) that is, pairs of words in binary semantic opposition that are associated by convention as well as by semantic relatedness. English examples of canonical antonym pairs include black/white, fall/rise, alive/dead, off/on, private/public and so forth. In this category, we also include words that may not be antonymous according to some logical definitions of antonym, but which are nevertheless incompatible terms that are conventionally paired, like cat/dog and niece/nephew. The notion of canonical antonymy contrasts with simple semantic opposition in which the meanings are incompatible, but the words are not necessarily conventionally paired for example, cold/scorching, blue/orange, descend/rise, etc. There is plenty of evidence that canonical antonym pairs are closely linked in our minds. The members of these pairs tend to elicit each other (and not other semantic opposites) in psychological tests such as free word association (Deese 1965; Charles & Miller 1989), and people are faster at recognizing them as antonyms than non-canonical opposites (Herrmann et al. 1979; Charles et al. 1994). Charles, Reed and Derryberry (1994) also found that canonical (or in their terms, direct) antonym recognition is not affected by the semantic distance between members of the pair, whereas semantic distance in non-canonical opposites delays reaction times for those pairs. In semantic priming tests, canonical antonyms prime each other more strongly than non-canonical opposites (Becker 1980). These types of evidence have been used to argue that the relation between the words is accessible without recourse to semantic processing (e.g., Gross et al. 1989; Charles et al. 1994). If, as I am about to argue, such pairs constitute complex lexical items in their own right, such effects can be accounted for. Comparing canonical antonym pairings to morphologically derived antonyms provides further evidence that the lexical (i.e., not morphologically derived) antonym pairs are directly represented in our language production faculties. While morphological opposites are readily available in English, we tend to prefer lexical opposites in contrastive constructions. For example, Jones (2002) searched for the word natural in various contrastive constructions. Artificial and man-made were the most frequent opposites found, while the morphologically derived antonym unnatural did not occur at all in the majority of the contrastive constructions investigated by Jones, and occurred at much lower rates than artificial and man-made in the constructions in which it did occur. So, while we have the ability to make predictable morphological opposites for many words, we often prefer to use morphologically unrelated words as opposites in contrastive constructions. The regularity with which we use the same words as antonyms in the face of the range of other available semantic opposites indicates a learned preference for pairing particular words. Antonymy is usually classed as a sense relation (e.g., Lyons 1977), meaning that it relates word senses rather than the words themselves, which have a range of senses. When a member of a canonical antonym pair acquires a new sense, the opposition can be carried into the new semantic field, thus indicating that we perceive the words as related even apart from their usually opposed senses. So for example, black/white are opposed as achromatic colour extremes, but when one is used to describe some other state, the other is still available as an opposite, as in (19).

(19a) black coffee / white coffee (unadulterated/with milk) (19b) black market / white market (illegal/legal)

White market plants include those that have been legally collected by botanical gardens (www.sarracenia.com/pubs/ncsp.doc) (19c) black people / white people (non-European/European genetic heritage) (19d) black box testing / white box testing (also called glass box testing) (involving notvisible/visible internal processes) (19e) white light / black light (visible/invisible types of radiation)