Graded salience: probabilistic meanings in the lexicon

This paper draws together ideas published in Allan (1976; 1980; 1981; 2000; 2001) which relate to a potential variety of senses for a given lexeme when it is listed in the lexicon (as a listeme), each sense being appropriate to a particular set of contexts of use. The problem for the lexicographer (if this is the correct term for someone modelling the mental lexicon) is how to tag the various senses such that they can be readily accessed in appropriate contexts. This tagging within the lexicon raises questions of salient and default meaning. Allan (2001 Chapter 3) and Allan (2006) give my model for the structure of a lexicon and its relation to a (mental) encyclopedia. In my view the encyclopedia is a general knowledge base of which the lexicon is a proper part which stores information about the formal, morphosyntactic, and semantic specifications of listemes. The network of relationships among the components of a lexicon and the encyclopedia are shown in Figure 1, where formal data, F, is represented as a triangle, morphosyntactic data, M, by a circle, semantic data, S, is a rectangle, and encyclopedic data, E, is an ellipse. It illustrates my assumptions and the reader does not need to approve the supposed configuration because in this chapter I am only concerned with component S.