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

Poster A21, Wednesday, November 8, 10:30 – 11:45 am, Harborview and Loch Raven Ballrooms

EEG tracking of grammatical structures with different cloze probabilities in connected speech

Adria Rofes1,2, Giovanni Di Liberto1, Emily Teoh1,3, Robert Coen4, Sonja Kotz5, Edmund Lalor1,3, Brian Lawlor1,4, Paul Dockree1;1Trinity College Dublin, Ireland, 2Johns Hopkins University, USA, 3Rochester University, USA, 4St James’s Hospital, Ireland, 5Maastricht University, Netherlands

Introduction: The processing of grammatical structures in connected speech has been tracked with MEG (Ding et al. 2015, 2016). However, it is unknown whether the indices are only a marker of grammatical processing or can be modulated by semantic processing. This is a relevant issue as some theories argue that grammar happens first (Frazier & Rayner, 1982) while other theories that grammar and semantics occur in parallel (MacDonald et al., 1994). First, we study whether Ding et al.’s results can be replicated with EEG. Second, we assess whether the indices are modulated by semantic cloze probability in young and older healthy individuals. Methods: EEG signals were recorded from 16 healthy young adults (age 18-30) and 11 healthy older adults (age 61-72) using a 128-channel system. All participants heard sequences of monosyllabic words and were asked to indicate whether or not the words made sentences. Individual words were presented at a frequency of 2.5Hz and were independently synthesized to avoid prosodic cues. The experiment contained 32 trials repeated in 3 separate runs. Each trial, consisting of 12 sentences, tested one semantic probability condition (high cloze= top chefs cook steak; low cloze= top chefs buy steak; no cloze= #top chefs jump steak) or a condition with 48 scrambled words (e.g., chefs steak top cook). To build the high and low probability conditions we asked 79 healthy participants (age 18-69) to write the first and second verb that occurred as a likely completion of a sentence missing a verb (top chefs ___ steak). A frequency analysis was conducted on the EEG data following Ding et al. Fourier transforms were derived for individual conditions to extract cortical indices of sentence, phrase, and word structure. We also studied the topographical distribution of these indices. Results: Mean percentage of correct behavioral responses was above chance level in young (86, sd=11) and older adults (68, sd=12). For young adults EEG tracks sentence structure (indexed by a frequency domain peak at 0.625Hz for sentences vs scrambled words) as well as semantic processing (differences between semantic probabilities). The topographies show a rhythmic component in central-posterior areas that we associate to a N400. For older adults EEG tracks sentence structure similarly to younger adults. The index is modulated by cloze probability (higher magnitude for higher cloze probability). However, the data does not show the central-posterior component. Conclusion: EEG tracks brain processes that correlate with sentence comprehension. Cloze probability modulates indices of sentence processing. This indicates (1) the signal is a marker of grammatical and semantic processing, in line with theories of parallel processing; (2) the paradigm has the potential to detect grammatical and semantic problems in people with neurological etiologies. Further analyses and pilot results on people with semantic impairments will be presented.

Topic Area: Grammar: Syntax

Back to Poster Schedule