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

Poster A20, Wednesday, November 8, 10:30 – 11:45 am, Harborview and Loch Raven Ballrooms

Tracking the dynamics of wh- dependency resolution inside and outside of islands: An ERP investigation

Lauren Covey1, Alison Gabriele1, Robert Fiorentino1;1University of Kansas

Wh-dependencies involve a long-distance relationship between a fronted wh-word (who) and the position in the sentence where it originated, called a gap site. Researchers have suggested that the processing of wh-dependencies involves at least two distinct processes, a predictive process in which the parser searches for a potential gap site upon encountering the wh-element, and an integrative process, when the dependency is successfully resolved at the gap site. The broader literature has linked these qualitatively different processes to distinct components: N400 for prediction (Federmeier, 2007; Lau et al., 2013; Michel, 2014; Van Berkum et al., 2005), and P600 for syntactic integration (Felser et al., 2003; Gouvea et al., 2010; Kaan et al., 2000; Phillips et al., 2005). Previous ERP studies have examined these processes independently, and few studies have tracked the dynamics of wh-dependency resolution across the sentence. The present study examines both prediction and integration and investigates whether these processes are indexed by unique components, focusing on the processing of wh-dependencies at three critical regions across the sentence: two associated with prediction and one with integration. The study also examines the extent to which prediction is grammatically constrained, investigating whether the parser avoids predicting a gap in positions that are not licensed by the grammar. To test gap prediction, we examine filled-gap effects, which occur when the parser predicts a gap in a position that is already filled with lexical material (Clifton & Frazier, 1989; Hestvik et al., 2007, 2012; Lee, 2004; Stowe, 1986). We also investigate whether the parser avoids predicting a gap in a position that is not licensed by the grammar, called ‘islands’ (e.g., Dave Campbell in 1c vs. 1d) (Johnson et al. 2016; Stowe, 1986). [Examples: (1a-b) Jamie wondered if/who the editor interviewed Dave Campbell with __ from the department. (Non-Island). (1c-d) Jamie wondered if/who the editor [that interviewed Dave Campbell] kissed __ after the meeting. (Island).] Effects of gap prediction were examined at a filled pre-verbal subject position (the editor in 1a/b and 1c/d) and a filled post-verbal object position (Dave Campbell), both inside (1c/d) and outside (1a/b) of an island. Dependency resolution was examined at the actual gap site (1a/1b with; 1c/1d kissed). Results from 26 native English speakers were analyzed using linear mixed-effects models for the following time windows: 300-500ms for N400 and 500-900ms for P600. At the subject position, a significant N400 emerged, with wh-extraction sentences yielding larger negativities than no-extraction sentences. For the object position, a significant N400 emerged only for the non-island conditions (1a/1b); no significant effects emerged for the island conditions (1c/1d). At the actual gap site, significant P600s emerged in both conditions, with wh-extraction sentences yielding increased positivities compared to no-extraction sentences. The fact that filled-gap effects (N400) emerged only at grammatically licit positions suggests that gap prediction is grammatically guided. P600s at the actual gap site indicate that the wh-dependency was ultimately resolved, even in sentences with islands. The study provides evidence for distinct neurophysiological responses associated with gap prediction and dependency completion during wh-dependency resolution.

Topic Area: Grammar: Syntax

Back to Poster Schedule