Speaker: Dan Duncan
Title: Spell-Out as the locus of syntactic variation
Abstract: Syntactic variation, which I take to mean variation between structurally different forms with identical meanings, is a probabilistic phenomenon like lexical, phonological etc., variation. An important research question, then, is where and how in the grammar such variation is implemented. In order to hold to some basic assumptions, namely, that syntax is built from the bottom-up, and a single derivation is sent to Spell-Out, two approaches have emerged. The competing grammars approach (Kroch 1994) suggests probabilistic selection between multiple grammars, which then yields differing derivations. The probabilistic grammar approach (for example, Burnett 2015) suggests that operations such as Merge are applied probabilistically. As Embick (2008) notes, these approaches yield identical surface patterns of variation. In this talk, I’ll go a step further and suggest they make an identical prediction: if variation is implemented at a decision point (Wallenberg 2013), subsequent operations cannot condition syntactic variation. In other words, material higher than the variable in the derivation cannot condition such variation. I test this shared prediction in a study of embedded passives in Pittsburgh English, which can take two structurally different forms that are identical in meaning (see Edelstein 2014):
1. The car needs washed.
2. The car needs to be washed.
A variationist analysis of >500 tokens of embedded passives shows that the selection of embedded passive is in fact conditioned by material higher in the derivation. This suggests that variation is implemented in the derivation later than previous approaches claim. I’ll argue for an approach in which multiple derivations are selected from probabilistically in Spell-Out. Effectively, I claim that syntactic variation is derived post-syntactically.