Poster C52, Sunday, March 25, 1:00-3:00 pm, Exhibit Hall C
Alpha- and theta-band time-frequency representations in free reading of stories using EEG and EM coregistration
Max Cantor1, John Trueswell2, Albert Kim1; 1University of Colorado Boulder, 2University of Pennsylvania
We used coregistration of eye movements (EM) and electroencephalography (EEG) to investigate the neurocognitive mechanisms of naturalistic story reading. We examined how semantic fit between a word and its context influenced neural oscillatory activity during, and immediately prior to, fixation on the word. Contextual fit for a word was quantified as the summed cosine distance between each word and the 10 closest content words in its left context, within a space that represented each word as a vector derived from co-occurrence statistics in a large corpus (using the Global Vectors for Word Representation, or GloVe system). For each content word fixated by readers, we created a time-frequency representation (TFR) of brain activity surrounding the first fixation on that word. We compared brain activity to words with low vs. high contextual fit. Contextual fit influenced word-related brain activity in three main ways. First, ~250 to 50 ms prior to fixation, alpha band activity (8-12Hz) was reduced for low- relative to high-fit contexts. We suggest that this reflects event-related desynchronization (ERD) due to greater anticipatory processing in low-fit contexts, which are higher in uncertainty and trigger attempts to predict the upcoming word. Second, in a period ~200-600ms after fixation onset, we observed increased alpha band activity for low-fit words. This may correspond to inhibitory activity related to the competitive dynamics of word recognition. Third, ~300-700ms post-fixation, theta band activity (~5-7 Hz) was greater for high than low fit words. This may reflect memory retrieval operations related to integrating a word with context.
Topic Area: LANGUAGE: Semantic