“Philosophers and others from Heraclitus to Jung have noted the tendency of things to slip into their opposite states; and many have remarked on the thin dividing line between love and hate, genius and madness, overlook (a mistake) and (scenic) overlook . . . . The paradox of simultaneous difference and similarity is partly resolved by the fact that opposites typically differ along only one dimension of meaning: in respect of all other features they are identical, hence their semantic closeness; along the dimension of difference, they occupy opposing poles, hence the feeling of difference.” D.A. Cruse

Last week, I approached the idea of synonymy from the perspective of semantic feature theory. If two words share all the same semantic features, they qualify as synonyms. Hence, “girl” and “lassie” are synonyms on the basis of sharing the feature set ([+person], [-mature], and [-male]). It is necessary to concede that even though these two words ‘mean the same thing’ according to this theory, they are certainly not always naturally interchangeable. Thus, there are other things which determine the appropriateness of a word in a particular situation, having to do with the linguistic location of the speaker (is the person in the US or Scotland, for instance?), what is the intended referent (is the named object a young female human or a personified entity, like a pet dog?), the feelings that the speaker has for the referent (does the speaker wish to express emotional neutrality or avuncular familiarity?), the formality of the situation (is the word used in a professional conference or in a personal reflection at home in front of the fireplace?), the modality of the expression (written or spoken?) and so on. So the total information conveyed by any expression will be much more than the meaning associated with a word. One advantage of this theoretical approach to synonymy in comparison to Cruse’s careful descriptive approach is that it establishes that synonymy is something that is real, and not a hypothetical ideal at the unreachable end of a gradient scale.

In contrast with ‘synonymy,’ the definition of ‘antonymy’ in the framework of lexical semantic feature theory declares this condition to hold whenever two words have all the same semantic features, though for one and only one binary feature, the two words are oppositely specified. If we return to our thesaurus entry for ‘girl,’ the one antonym that is given is ‘boy.’ We can confirm this antonymy by noting that ‘boy’ will have all the same semantic features of ‘girl’ in terms of personhood and maturity level, but will be specified oppositely on the feature of sex.

Interestingly, in our explorations of the synonymy of ‘girl,’ we ended up ruling out many of the candidates provided by the thesaurus, narrowing down the list substantially. Thesaurus lists of antonyms tend to be much smaller than lists for synonyms (why might this be the case?). In this case, it is natural to see if we can extend the list somewhat. Now our theory provides us with exactly the direction to explore, as we look for other antonyms of ‘girl’: What other words have all the same semantic features as ‘girl,’ with the exception of one oppositely specified? If we hold personhood and sex the same, we can modulate the feature for maturity and get ‘woman’ (and all of its synonyms) as another antonym of ‘girl.’ Psycholinguistic tests could be constructed to investigate whether ‘woman’ is as intuitively antonymous to ‘girl’ as ‘boy’ is. Note that we do not have the option of finding non-person antonyms for ‘girl,’ that is, something that would be identical in sex and maturity level, but a non-person. So ‘heifer’ (young, female, bovine) would not have a feature specifying personhood, but would have one identifying its particular place in the animal kingdom, and, hence, would not qualify as an antonym. This outcome may alert us to potential words that do not yet exist! Something like a “non-girl” would not help to fill this gap, because the prefix ‘non-‘ would overshoot the class of potential antonyms by covering everything that did not have all the features of ‘girl.’

Note finally that a single referent could be named in a variety of ways leading to all kinds of different antonymous sequences. So Miska, the little girl in our photograph, is a natural referent for ‘young’ (antonymous with ‘old’), ‘child’ (antonymous with ‘parent’), ‘small’ (antonymous with ‘big’), etc.

Conceptually, the theory of semantic features is quite attractive, but, as we have already considered, just what constitutes a feature is a difficult problem to resolve. Again, the close descriptive work that Cruse carries out in his chapter on ‘opposites’ gives us some sense of the challenges facing any theory of semantic features.