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

Poster A19, Wednesday, November 8, 10:30 – 11:45 am, Harborview and Loch Raven Ballrooms

ERP responses to active versus “passive” gap filling

Laura Snider1, Jon Sprouse1;1University of Connecticut

The goal of this project is to compare the ERP correlates of active gap filling – the processing of a gap that is predicted by the grammar, and “passive” gap filling – a gap that is not predicted by the grammar (and presumably identified using bottom-up information). The hope is that this information could be instrumental in resolving debates about the presence or absence of dependency constraints in various languages. To create syntactic contexts that allowed for the manipulation of active versus passive gap filling, we adapted the design of Wagers and Phillips 2009, and compared two types of multiple-gap constructions: across-the-board (ATB) constructions (“and” in 1), and parasitic gap (PG) constructions (“before” in 1). (2) are control conditions. (1) Target: Who was the announcer presenting __ energetically on the show {and/before} discussing __ cheerfully during the segment? (2) Control: Who was energetically presenting the guest on the show {and/before} discussing the guest cheerfully during the segment? PG and ATB constructions each have two gaps: one in the first clause, and one in the second clause. Wagers and Phillips demonstrated with reading times that the second gap in ATB and PG differ in terms of gap processing: the second gap of ATB is actively filled (because the gap is required by the grammar); the second gap of PG is not actively filled (because the gap is not required). The question is to what extent the ERP correlates of gap filling differ at the second gap for ATB and PG. We recorded continuous EEG over 32 channels, for 300 sentences (40 each of the 4 conditions, plus 140 fillers to eliminate processing biases), RSVP, 500ms SOA, 200ms ISI. The analysis includes 19 subjects with accuracy ratings over 70% for grammaticality judgments (following 25% of items). We believe the participants processed the two structures differently because we observed an increased frontal-central negativity between 300 and 500ms for PG versus ATB at the critical structural word (and vs before/during/after). However, we found no differences between ATB and PG constructions at the second gap. For the 500ms at the second verb, we found no effects for either ATB or PG target conditions versus their controls. For the 1000ms window containing the preposition and determiner, we found identical effects for ATB and PG targets versus controls: a more negative N1 peak for controls, a more positive P2 peak for controls, a posterior positivity for targets from 300 to 700ms, and a left anterior negativity for targets from 700 to 1000ms. We also compared ATB target to PG target conditions (without controls) at the adverb, and found no differences. Our results suggest that classic ERP responses do not distinguish active versus passive gap filling (at least for this sample size). This suggests that we need to look beyond classic ERP responses to study gap filling in detail. Future work will look at neuronal oscillations in this data set, and a follow-up study will look for differences in sustained anterior negativities in ATB and PG.

Topic Area: Grammar: Syntax

Back to Poster Schedule