|Title:||Suffix Ordering in Bantu: A Morphocentric Approach|
|Authors:||Larry M. Hyman|
|Abstract:||In this paper I address the issue of templatic vs. compositional determination of affix ordering. Rather than constituting two distinct, i.e. incompatible types, I show that both determinants can in fact co-occur in the same language. Proponents of the “mirror principle” (Baker 1985), which claims an identity between the order of syntactic and morphological operations, as reflected in the order of affixation, find support in cases of apparent compositionality in Bantu verb suffixation, and proclaim the Mirror Principle to be “an exceptionless generalization” (Alsina 1999:6). However, in this paper I present morphological, syntactic, and phonological arguments that there is a default order of Bantu verb suffixes, schematized in the partial template:
e.g. Chichewa -its-ir-an-idw-
I show that there is a 'tension' between this template and compositionality (whereby an outer suffix should have scope over an inner suffix). I propose that language-specific overrides are responsible for licensing atemplatic suffix+suffix combinations, when they occur. This conflict is aptly captured via via a general TEMPLATE constraint vs. specific MIRROR constraints. I show that these constraints can have different rankings with respect to different suffixes and in different languages. Importantly, when they override TEMPLATE constraints, MIRROR constraints typically have a much more restricted and often idiosyncratic character. Since affix ordering is not directly predictable by appealing to the semantics (e.g. scope; Bybee's 1985 'relevance') or the syntax (e.g. Baker's Mirror Principle), this argues not only for the reality of templates in morphology--but for a morphology that is autonomous. The low ranking of a (syntactic OR semantic) Mirror Principle as a predictor of suffix ordering in Bantu shows that it is not universal in the “no exceptions” sense, but rather in the (violable) OT sense.