Corpus Approaches to Antonymy

The word “antonymy” was coined in 1867 by CJ Smith to describe word-pairs commonly known as “opposites” such as hot/cold, girl/boy and buy/sell. Some linguists (e.g. Lyons (1977) and Cruse (1986)) apply the term “antonymy” restrictively and would only identify the first of these three pairs as being truly antonymous. However, this seems counter-intuitive in many respects, and this paper will use “antonymy” in its broader sense to refer to all word-pairs which could reasonably be identified as “opposites” by speakers of English. Antonymy is often defined simply as “oppositeness of meaning” (Palmer 1976: 94). However, the problem with an exclusively semantic definition is that it fails to explain, or even acknowledge, the tendency for certain words to become enshrined as “opposites” in language while others do not. For instance, rich and poor would be regarded as antonyms because they occupy opposite ends of the same scale, namely the scale of wealth. Affluent and broke also occupy opposite ends of this scale, but, intuitively, one would be reluctant to describe them as antonyms. Therefore, antonymy should be defined according to lexical as well as semantic criteria. It is a phenomenon “specific to words rather than concepts” (Justeson & Katz, 1991: 138). This paper will examine some of the ways in which advances in corpus technology enable antonymy to investigated afresh. Firstly, the categories to which antonymous pairs have been logically assigned will be summarised. Secondly, a set of new, corpus-based categories will be presented. Thirdly, a statistical analysis of co-occurrence rates among antonyms will be offered. And finally, ways of identifying new antonyms, again using corpus data, will be explored.