Whamit!

The Weekly Newsletter of MIT Linguistics

Syntax Square 10/6 - Nick Longenbaugh

Speaker: Nick Longenbaugh (MIT)
Title: The processing of long-distance dependencies in Niuean (joint work with Masha Polinsky)
Time: Tuesday 10/6, 10-11am
Place: 32-D461

It is well documented that nominative-acccusative alignment coincides with a strong subject-preference in long-distance dependency formation, both in terms of processing (in particular, subject gaps in relative clauses are processed more easily than other types of gaps) and accesibility for extraction (there appears to be an implicational universal that the availability of relativization with a gap entails the availability of subject relativization with a gap Keenan & Comrie 1977). This subject preference does not, however, carry over uniformly to languages with ergative-absolutive alignment; in some morphologically ergative languages, the ergative subject cannot extract with a gap at the extraction site, a phenomenon termed `syntactic ergativity’. In this talk, I explore the viability of a processing-based explanation of syntactic ergativity. Much as has been proposed for various island phenomena (Kluender 1998, 2004), the extraction of ergative arguments may simply be more taxing on the parser than corresponding absolutive extraction. If this is true, following Hawkins (2004, 2014), syntactically ergative languages could then be taken to differ from their morphologically ergative counterparts in their tolerance for difficult structure, eliminating the less efficient, more difficult ergative extraction. To test this account, I explore the processing of relative clauses in Niuean, a morphologically, but not syntactically, ergative language. Niuean is an ideal test case, given that it is closely related to the syntactically ergative Tongan, and thus might be expected to show an obvious bias against ergative extraction (a bias that Tongan turns into a categorical restriction). I present novel experimental data showing that ergative subject gaps in Niuean RCs do not impose any additional processing difficulty as compared to the processing of absolutive object gaps, thus calling into question the viability of the processing account in this domain.