By now, the success with which corpora can be used to examine ‘opposites’ should be apparent. Primarily, various functions of antonymy within a 3,000sentence database have been identified, quantified and used as the basis for a new system of classification. However, this database has also enabled more subtle aspects of antonymy to be explored. For example, the fact that most antonymous pairs favour a given sequence in text has been established and discussed, as has the influence of word class on the function of antonymy, and the question of whether non-gradable antonyms operate in the same way as gradable antonyms. This research can be seen as a complement to pre-corpus investigations of antonymy and as a precursor to further data-based analyses. The purpose of this chapter is threefold: to look back at past investigations of antonymy so that the relationship between traditional classes and new classes can be assessed; to compare this research with other corpus-based studies of antonymy so that the present position about antonymy can be consolidated; and, finally, to make explicit the limitations of this study so that future research may focus on those aspects of antonymy not adequately explored here.

Antonymy has invariably been packaged together with a batch of other sense relations for research and teaching purposes. However, thinking about antonymy alongside synonymy, hyponymy, meronymy, etc. obscures the fact that these relations are all diverse in nature. Unlike other sense relations, ‘opposites’ are a key element of human communication, becoming entrenched in the psyche from a very early age and playing an important part in our daily lives. Antonymy is not just another member of the nymic family, as traditional semantic theory sometimes implies.