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

 
Poster C43, Thursday, November 9, 10:00 – 11:15 am, Harborview and Loch Raven Ballrooms

Neural evidence for representationally-specific pre-activation: Evidence from Representational Similarity Analysis over time and space

Lin Wang1,2, Gina Kuperberg1,2, Ole Jensen3;1Department of Psychiatry and the Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Harvard Medical School, Charlestown, MA, USA, 2Department of Psychology, Tufts University, Medford, MA, USA, 3Centre for Human Brain Health, University of Birmingham, Birmingham, UK

Introduction: Previous studies have shown that people generate probabilistic predictions at multiple levels of linguistic representation during language comprehension. Here, we used MEG in combination with Representational Similarity Analysis (RSA) over time and space to determine whether representationally-specific pre-activation can be detected in the brain under experimental conditions that bias strongly towards lexico-semantic prediction. Stimuli: We constructed 120 pairs of highly-constraining sentence contexts. Each member of a context pair was distinct (contained different content words) but constrained strongly for the same sentence-final word (SFW), e.g. “In the hospital there is a newborn …”; “In the crib there is a sleeping …”, where both contexts constrain for “baby”. We measured MEG activity as 32 native Chinese participants read these sentences, presented word-by-word in Chinese. Although each participant saw both members of each context pair, they only saw the predicted SFW (e.g. “baby”) after one context; after the other, they encountered a plausible but unexpected SFW (e.g. “child”). Each word was presented for 200ms with a long interstimulus interval (ISI) of 800ms. Our analyses focused on anticipatory activity prior to the onset of the SFW. Analysis and Results: (1) RSA based on spatial similarity: At each time point within the anticipatory time window, we extracted the spatial pattern of neural activity across all sensors produced by all contexts. We first correlated these spatial patterns across contexts that constrained for the same SFW (within-pair correlations), and then across contexts that constrained for different SFWs (between-pair correlations). We found that, between 490-890ms before the onset of the SFW, the spatial similarity was significantly larger for the within-pair than the between-pair contexts. A cross-temporal matrix showed that this increase in spatial similarity was significant only along the diagonal (where the spatial similarity was calculated at corresponding time points), suggesting that representationally-specific pre-activation was dynamic rather than sustained in nature. (2) RSA based on temporal similarity: After projecting the sensor-level data to source space, we correlated the time series of neural activity at all voxels between 490ms and 890ms window prior to the onset of the SFW, across contexts that constrained for the same SFWs, and then across contexts that constrained for different SFWs. We found that the degree of temporal correlation (temporal synchrony) was greater for within-pair than between-pair contexts within the left anterior temporal region (with a medial temporal focus, but extending into the inferior, middle and superior temporal cortices). Conclusions: The brain can produce unique spatial patterns of activity associated with the pre-activation of specific lexico-semantic representations, prior to new bottom-up input becoming available. These patterns may reflect the mobilization of specific sets of highly distributed semantic representations across the cortex [1]. The anterior temporal region may play a role in instantiating lexico-semantic prediction [2], possibly by binding these highly distributed representations through temporal synchrony. [1] Patterson, et al (2007). Nature Review Neuroscience. 8(12), 976-987. [2] Lau et al., (2016). Cerebral Cortex. 26 (4): 1377-1387.

Topic Area: Meaning: Combinatorial Semantics

Back to Poster Schedule