Paul Kiparsky (Stanford)

Two approaches to locality in morphology

Carstairs (1987) discovered that contextual allomorphy is local and inward-sensitive: it can depend on the identity of the adjacent stemward affix and on the accumulated morphosyntactic features of the base, but not on affixes further away in the word. His empirical generalization has held up rather well, but its theoretical explanation has remained elusive. I present a version of lexicalist morphology that predicts it, and resolves some apparent counterexamples to it.

What Distributed Morphology predicts about morphological locality remains moot. It depends on whether spellout happens before linearization (Embick 2010, 2015) or after linearization (Arregi and Nevins 2012), and on whether its contextual window is limited by strict adjacency (Embick) or extends over spans of contiguous heads in an extended projection (Svenonius 2012, Merchant 2015). All versions, however, overgenerate by allowing at least some types of outward dependency. Some also undergenerate by disallowing root suppletion. Thus, the evidence from contextual allomorphy appears to favor lexicalist morphology over DM.