The use of models in linguistics
Before moving into the subject proper, I should perhaps say some thing about the term ‘model’ and the way in which it is, or has been, employed by linguists. As Chao (1962) pointed out in his article on this very topic, it is a term that is used in somewhat different senses in different disciplines. For the mathematician and mathematical logi cian, a model is a formal system considered from the point of view of its interpretation, or application to some practical problem, rather than abstractly for its own sake. However, when the social scientist or physical scientist employs the term ‘model’, he usually means some deliberately restricted and abstract representation of the phenomena whose structure or behaviour is being studied. Typical examples of
models, in this sense of the term, are a physicist’s representation of atomic structure or an economist’s analysis of monopolistic competi tion. Since any model of this kind is necessarily based upon an idealization of the data that it is designed to describe or explain, how one decides which variations in the data are of significance and which variations can be discounted becomes a question of crucial importance; and the answer to this question will depend upon the nature of the correspondence that is assumed to hold between the data and the model, and upon a fairly precise prior specification of what it is that the model is intended to explain or describe.