Speaker: Donca Steriade (MIT)
Title: Constraint Interactions derive Morphomic Identity
Time: Thursday 10/21, 12:30—1:45pm
Location: 32-D461
Morphological paradigms sometimes display systematic identity (or syncretism) between cells that
share no syntactic properties. Such phenomena are widespread enough to have generated new
descriptive technologies: e.g. the rule of referral (Zwicky 1985, Stump 2001), the morphome
(Aronoff 1992) and the thematic space (Bonami and Boyer 2001, for stem syncretism). All these
devices were proposed in the belief that any pair of paradigm cells is as likely as any other to have
identical exponents.I show that syncretism is frequently predictable when phonology is factored in: cells that are on the
surface strictly identical, without syntactic reason, are the ones that whose underlying
representations single them out as being very similar, in some local respect. Familiar data showing
this is the obligatory identity between English past tense and past participles when both end in t or d
(sat/sat; rid/rid; built/built; burnt/burnt; said/said). There is variation, but no paradigm internal
contrast between forms like burnt~burned or lit~lighted. Pairs not ending in t/d tend to be distinct in
strong verbs (wrote/written; swelled/swollen; dove/dived) showing that its the local similarity
between final alveolar stops that triggers syncretism. What we need then is a mechanism that
generates global identity between forms that are underlyingly distinct but similar, i.e. hypothetical
inputs like burnt/burned.Most of this talk is about the Latin syncretism that inspired the morphome (Aronoff 1994; Matthews
1974) and the related hypothesis of an autonomous morphological component. The stem of many
Latin deverbal derivatives (e.g. agent nouns like laudator praiser) is always identical to that of the
passive-perfect participle (laudatus praised). I show that local similarity is the critical factor here
too, in ways reminiscent of the English case. The Latin pattern is more revealing because stem
syncretism violates here an otherwise general condition linking the referent of a deverbal derivative
to the argument structure of verb forms containing the same stem as the derivative: if the latter are
passive, the former must refer to a passive subject. The full picture suggests exactly the opposite of
Aronoffs conclusions: syncretism is not arbitrary, and cannot be analyzed in an autonomous
component. Rather, it involves an interaction of phonological constraints seeking identity of similar
forms with conflicting preferences on the proper exponence of syntactic structures.