Defining vs. diagnosing linguistic categories: A case study of clitic phenomena

Linguistic generalizations, e.g.,about phenomena labeled “clitics,” presuppose that we identify classes of phenomena inaconsistent way. But many grammatical terms (including the term“clitic”) are used for quite different phenomena in different languages. This is sometimes obvious, and sometimes less so. In this paper, I contrast two views about categories and their cross-linguistic applicability: The restrictivist approach assumes that there is a universal set of features and categories from which languages may choose, while the non-aprioristic approach makes no such assumption and proposes to compare languages on the basis of a special set of comparative concepts that are not closely related to language-specific descriptive categories. On the restrictivist view, categories exist independently of individual languages (as they are innate and thus given inadvance). Thus, they can be identified by diagnostics, much like diseases are identified through their symptoms, and different languages could exhibit different diagnostics. However, for clitics this approach fails demonstrably, as there is no agreement on how to distinguish clitics from affixes or free words. This approach allows no way to go beyond subjective judgements. The non-aprioristic view is that linguistic classes should be defined and thus identified rigorously and objectively. This applies both to language-specific descriptive categories and to comparative concepts.To illustrate this,I propose two new comparative concepts, plenimorph and minimorph, for making a few readily falsifiable claims about the “clitic” domain.