# Words, paradigms and analogy

This chapter has summarized how a model whose limitations were well understood a half-century ago came for a while to occupy a dominant role within modern approaches to the analysis of word structure. To overcome these limitations, we now shift our focus outside the narrow frame of reference provided by the Post- Bloomfieldian tradition, to models that offer a more flexible analysis of the diversity of patterns exhibited by morphological systems.

The development of these models must be set against the backdrop of the ideas inherited from the classical model. This model is, as Hockett (1954:386) stresses, of more than purely historical interest. As “the traditional framework for the discussion of Latin, Greek, Sanskrit, and a good many more modern familiar languages”, the classical WP model survives largely intact in current pedagogical and reference grammars and dictionaries. Even for the time that it was eclipsed theoretically by morphemic models, the WP model remained the primary framework for broad- coverage grammars, and it is in the pedagogical and descriptive traditions that the assumptions of the classical WP model are perhaps most faithfully represented, albeit with idealizations.

The pivotal role that words and paradigms play in comprehensive descriptions naturally leads one to ask why pedagogical and reference materials should be organized in terms of words and exemplary sets of words, and whether this organization is of any general theoretical relevance. One possible answer is, of course, that the structure of descriptive grammars again reflects sheer intellectual inertia. Grammars of classical languages are organized into paradigms and principal parts, and the Western grammatical tradition merely perpetuates an established practice. While there may be some truth to this view, the continued relevance of the WP model points to the descriptive value of words and exemplary patterns. Pedagogical and reference grammars are driven more by considerations of utility than by theoretical commitments. As Matthews (1991: 187k) remarks, with regards to flectional languages at least, neither the Bloomfieldian model nor any theoretical approach developed since Bloomfield’s time organizes the facts of a language more transparently than the classical WP model. This organizational scheme contains three interacting sets of assumptions: assumptions about units, assumptions about grammatically significant patterns and assumptions about the mechanisms that relate the two. Chapters 3-5 now consider each set of assumptions in turn.