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

Poster D9, Thursday, November 9, 6:15 – 7:30 pm, Harborview and Loch Raven Ballrooms

Investigating task-modulated syntactic prediction with MEG

Phoebe Gaston1, Chia-Hsuan Liao1, William Matchin2, Ellen Lau1;1University of Maryland, College Park, 2University of California San Diego

INTRODUCTION: Prior work has argued that sentence processing is predictive, to an extent potentially modulated by demands of the task and context. This may account for variability in reported effects of syntactic structure in regions such as left inferior frontal gyrus (IFG) and posterior superior temporal sulcus (pSTS), if observed activity is in fact reflective of syntactic prediction rather than structure-building. The goal of this study was to directly investigate the relationship between effects of syntactic structure and task demands for prediction, using a novel paradigm explicitly encouraging or discouraging structural prediction while keeping stimuli constant. METHODS: We visually presented determiner phrases (DPs) in sentence contexts (These wolves chase many ducks), in two-phrase list contexts (these wolves many ducks), and in mixed contexts where DPs were equally likely to begin a sentence or a phrase list. Participants were informed of the context before each block began, and tasked with detection of agreement violations. Because linguistic input at the first DP has not yet begun to differ between conditions, effects of syntactic structure should reflect only the cost of syntactic prediction and the extent to which it is encouraged by the context. The sentence trials in mixed vs. sentence blocks also allow comparison of the neural response throughout the trial to identical stimuli in contexts that had encouraged structural prediction to a greater or lesser extent. Neural data were recorded using MEG (N = 21), and permutation cluster tests were performed on the time-courses of activation in four regions of interest (ROIs): pars triangularis and pars orbitalis in left IFG (IFG-tri, IFG-orb), left pSTS, and left temporal pole. Spatiotemporal cluster tests were then performed in a broader language-related area encompassing left temporal and inferior parietal lobes and left IFG. RESULTS/DISCUSSION: Surprisingly, initial results demonstrated the reverse of the expected pattern for prediction effects, raising the possibility that this version of the paradigm engenders other unintended differences between conditions. 796-980 ms. after the onset of the first DP in the left IFG-orb ROI, we found increased neural activity during phrase blocks, which do not encourage structural prediction, relative to sentence blocks, which do. This apparent facilitation for sentences could be due to shifts of attention in anticipation of upcoming structure, but may also follow from the necessarily higher proportion of within-DP agreement violations in the phrase blocks than in the sentence or mixed blocks. A potentially related effect in left anterior temporal lobe, from 696-888 ms., showed increased activity for the second DP in a phrase list when it occurred during phrase-only relative to mixed blocks. We also found more activity for the second DP in a phrase list than the second DP in a sentence, across the temporal lobe from 636-1000 ms. This is likely due to increased semantic predictability of the second DP in sentences. These results indicate the need for further development of a promising window onto predictive structure-building, encouraging careful accounting of both task demands in neurolinguistic paradigms and the role of probabilistic structural predictions in language comprehension.

Topic Area: Grammar: Syntax

Back to Poster Schedule