You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

Poster D5, Thursday, November 9, 6:15 – 7:30 pm, Harborview and Loch Raven Ballrooms

ERP responses to two types of subject island violations and constructions with substantially similar processing dynamics

Jayeon Park1, Jon Sprouse1;1University of Connecticut

The goal of this experiment is to expand the set of ERP responses to complex syntactic phenomena, in order to contribute to the expansion of the empirical base of syntax beyond offline acceptability judgments. To that end, we compared two types of subject island violations to two constructions that share some of the same sentence processing dynamics. Though these results don't militate for or against any specific syntactic or processing theories yet, the results could be leveraged in the future to adjudicate less clear cases in the syntactic literature (e.g, Korean island effects). Our first comparison is between standard subject island violations in (1), and yes-no questions with a missing word in (2). (1) *What did [the article about __ ] annoy the governor? (2) *Did [the article about __ ] annoy the governor? Neville et al. 1991 recorded ERPs to (1), and found a more positive P2 peak and a P600/SPS at the verb after the illicit gap. Given that the active filling strategy is suspended within complex subjects (Stowe 1985), it seems plausible that the parser initially interprets the preposition-verb sequence in (1) as representing a missing word rather than a wh-gap. This suggests that a minimal comparison would be with a yes-no question with a missing word as in (2). Any unique effects in (1) versus (2) could represent properties of wh-dependency processing and subject island constraints. Our second comparison was between the non-finite subject island in (3), and a classic filled-gap paradigm as in (4). (3) *What did [the attempt to clean __ ] immediately shatter the beaker after the experiment? (4) What did the surgeon clean the scalpel carefully and thoroughly with __ at the hospital? The rationale behind this comparison is that the locus of ungrammaticality in (3) is not the gap, because the sentence would be grammatical if there were a gap downstream at the matrix object position; therefore the ungrammaticality arises at the filled-potential-gap after the matrix verb (Endgahl 1983, Phillips 2006). We compared ERP responses (N=24, RSVP, 500ms SOA, 200ms ISI) to (1-4) with matched grammatical controls. For both subject islands and missing words, we found increased P2 responses and late P600/SPSs. We also observed a marginal gradient negativity in the 300-500ms window (missing-word > subject-island > control), and a significant late anterior negativity for the missing word violation. For the classic filled-gap effect, we found a sustained anterior negativity (SAN) that is smaller in the words before the filled-gap location, and increases for the words after the filled-gap location. To our knowledge, this is the first report of an increasing SAN (cf. Phillips et al. 2005), which could be indexing the re-starting of the wh-dependency after the incorrect gap-filling at the verb. We observed no similar effect for non-finite subject islands. Instead, we observed two phasic posterior negativities after the "filled-gap". Though much work remains in linking these effects to specific processes, these results suggest that ERP responses to island constraints are qualitatively distinct from violations with similar processing profiles.

Topic Area: Grammar: Syntax

Back to Poster Schedule