Swedish opposites - a multi-method approach to antonym canonicity

In spite of the fact that antonymy is considered to be an important organising lexico-semantic principle, very little research has been conducted to get a better grasp of what ‘goodness’ of binary opposition in human language and thought really is. The focus of this study is what distinguishes antonyms such as bra-dålig ‘good-bad’, lång-kort ‘long-short’, tunn-tjock ‘thin-thick’ from other types of contrast such as seg-mör ‘tough-tender’, dunkel-tydlig ‘obscure-clear’ and rask-långsam ‘speedy-slow’. There are indications in the literature that there are various converging reasons for perceptions of ‘goodness of antonymy’, e.g. frequency of co-occurrence, co-occurrence in certain constructions, stylistic co-occurrence preferences, pairwise acquisition and clarity of the antonymic dimension (e.g. Muehleisen 1997, Willners 2001, Jones 2002, Murphy 2003, Paradis et al.). The purpose of this paper is to examine some of those aspects through a multi-method approach to the study of ‘goodness of oppositeness’ of Swedish antonyms and to evaluate the converging and diverging results of the methods in the light of this particular issue. Three main methodologies are used. Corpus-driven methods ground the study of Swedish antonyms in text. These data are then used as the test set for an elicitation experiment and a judgement experiment to find patterns of ‘goodness of antonymy’ in Swedish