Before embarking on a study of lexical semantics, even one which is avowedly descriptive rather than theoretical in orientation, it is necessary to make explicit certain basic assumptions concerning meaning, and to establish, as far as possible, a consistent method of studying it. The approach which is adopted in this book, and which is described in this introductory chapter, is a variety of "contextual" approach: it is assumed that the semantic properties of a lexical item are fully reflected in appropriate aspects of the relations it contracts with actual and potential contexts. The full implications of this will become clearer as the exposition proceeds. In theory, the relevant contexts could include extra-linguistic situational contexts. But there are good reasons for a principled limitation to linguistic contexts: first, the relation between a lexical item and extra-linguistic contexts is often crucially mediated by the purely linguistic contexts (consider the possible relations between 'horse' and the extra-linguistic situation in 'That's a horse' and 'There are no horses here'); second, any aspect of an extra-linguistic context can in principle be mirrored linguistically; and, third, linguistic context is more easily controlled and manipulated. We shall therefore seek to derive information about a word's meaning from its relations with actual and potential linguistic contexts. However, the combinatorial characteristics of words in utterances are constrained not only by their meanings, but also by their grammatical properties. Grammatical constraints may overlap and reinforce semantic ones, but they may also be semantically arbitrary. In order to be able to use contextual relations for semantic purposes, therefore, we need to be able to recognise and discount combinatorial peculiarities which are purely grammatical in nature. (Alan Cruse)