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

Poster D11, Thursday, November 9, 6:15 – 7:30 pm, Harborview and Loch Raven Ballrooms

EEG correlates of covert dependency formation in Mandarin wh-questions

Chia-Wen Lo1, Jonathan Brennan1;1University of Michigan

Event-related potential (ERP) components are sensitive to the processes underlying how questions are understood. For example, English wh-questions elicit an increased negativity over left-anterior sensors (“LAN”) between the sentence-initial wh-word and the gap site where it is interpreted (Kluender & Kutas 1993) while a P600 and/or a LAN has been observed at the gap site itself. Kaan et al. (2000) suggest that while the P600 reflects syntactic integration of the sentence-initial wh-word with the rest of the question, the earlier LAN may reflect maintenance of an incomplete wh-dependency in working memory. Work with Japanese extends these efforts to languages with so-called “covert” dependencies where the question word is in situ rather than sentence initial, and is followed by a sentence-final question particle. Ueno and Kluender (2009) find that, like English, there is an anterior negativity (right lateralized) in Japanese but no P600. These results suggest that memory maintenance demands are similar both for “overt” dependencies, like English, and covert dependencies, like Japanese. Mandarin Chinese also has covert wh-questions, though it lacks a sentence-final question particle. We test the prediction that, despite surface differences, wh-questions in Mandarin will show similar memory maintenance demands as indexed by an anterior negativity. Methods: N=34 native speakers of Mandarin Chinese read Chinese question and declarative sentences word-by-word during EEG recording. Target sentences (100 per condition plus fillers) were (1) direct questions (English approximation: “J thinks M saw which paper” = “Which paper does J think M saw __?”, (2) indirect questions (“J wonders M saw which paper” = “J wonders which paper M saw __”), or (3) yes/no questions (“Did J think M saw this paper?”). Only the indirect questions with “wonder”-type verbs provide a cue to form a dependency prior to encountering the question word itself. This condition is where we predict an anterior negativity for memory maintenance. Data were manually cleaned of artifacts, filtered from 0.1-30 Hz, baseline corrected, and conditions were compared across 0-1 s using a non-parametric permutation test. Simulations indicate that N=34 with ~100 trials per condition yields power > 0.9 to detect effects as small as d = 0.1. Results: We find a left-lateralized anterior negativity at the in-situ wh-word for indirect and direct questions as compared to yes/no questions (120-250 ms, p < 0.05). This effect is consistent with memory mechanisms engaged when forming a covert long-distance dependency triggered by the wh-word (e.g. Kluender & Kutas 1993), but these data cannot rule out lower-level sensitivity to wh-words versus a deictic pronoun (“these”). Consistent with prior work with Japanese covert dependencies, we do not find a P600 at the gap site, but in contrast we also do not find an anterior negativity between the cue “wonder” verb and the wh-word in indirect questions. The absence of such a negativity may suggest earlier reports reflect storage of the wh-word, specifically, and not more general active maintenance mechanisms.

Topic Area: Grammar: Syntax

Back to Poster Schedule