Some lexical structures and their empirical validity¹

Discussions of a semantic theory of language can flounder on the determination of the data which that theory is intended to describe. In this paper we shall avoid questions of the legitimacy of particular goals for semantic theory. Nor are we concerned primarily with the nature of the computational device used in analyzing the meanings of whole sentences. Rather we report some recent investigations of several kinds of lexical devices which are necessary formal prerequisites for the description of semantic phenomena.

The empirical reality of any descriptive device is initially supported by the facts it describes. It is further supported by the validity of predictions which it justifies. In this paper we show that certain hierarchical structures proposed for an adequate description of semantic lexical structure also correctly predict the potential occurrence of certain types of words and interpretations and correctly reject the potential occurrence of other types of words and interpretations.

The empirical phenomena and the formal devices which describe them are reflections of problems and solutions which arise in other behavioral phenomena (e.g. in anthropology or psychology). The basic phenomenon we explore is the representation of classes of words by features, like ‘animateness’, ‘plantness’, and ‘ livingness ’. These features are themselves organized hierarchically (e.g. anything that is ‘living ’ must be a ‘ plant ’ or ‘ animate ’). Finally, certain hierarchies express relations between particular words rather than between classes or words. These relations are maintained in the metaphorical extensions of the literal interpretations of these words. At each point we show how the linguistic facts justify certain formal constraints, and we demonstrate the empirical extensions of each of these descriptive devices.