On the 'Art' of Explanation
This work largely grew out of a dissatisfaction with the way the term 'explanation' has been used in linguistics. The standard practice of scientific inquiry is the fourstep procedure of identification, description, explanation and prediction. A particular phenomenon is first identified, then described, and then explained within a theoretical model which serves to derive predictions about other, hitherto unexplained phenomena. Scientific progress involves the formulation, testing, and reformulation of theoretical accounts. In generative linguistics, which may be regarded as one of the most, if not the most influential school of thought,1 the distinction between description and explanation is not generally respected. While generative linguists do not shy away from the term 'explanation', they often use it either as a mere paraphrase of description or in a rather superficial sense. Let us take an example from a recent textbook on phonology. In a discussion of onset and coda structures in various languages, Goldsmith ( 1990) states that a greater number of phonological contrasts are usually accommodated in onset than in coda positions. For instance, many languages allow a contrast between voiced and voiceless consonants in prevocalic though not in postvocalic loci. Goldsmith accounts for this asymmetry by introducing the concept of 'licensing'. He asserts that onset and coda segments are licensed by different 'licensing agencies', the former by a primary licenser (the syllable) and the latter by a secondary licenser (the coda itself). The greater restrictions on codas are 'explained' by their being less freely licensed. This is either a (tautological) restatement of the basic observation or the invocation of hard-to-justify assumptions whose ad hoc nature severely curtails their explanatory value.
Two aspects of Goldsmith's procedure are characteristic of linguistic research in general--the appeal to generalizations and to formalisms. Much linguistic work, past and present, is inspired by the desire to uncover 'linguistically significant generalizations'. This is undoubtedly a first step towards an explanation, although such statements tend to be rather more descriptive than explanatory in kind. However,____________________