Schedule of Events | Symposia

Sketchpad Series

Evaluating the impact of channel density on Representational Similarity Analysis of prediction-related effects in language

Poster Session E - Monday, March 31, 2025, 2:30 – 4:30 pm EDT, Back Bay Ballroom/Republic Ballroom

Aya Gomaa¹ (gomaa2@illinois.edu), Ryan J. Hubbard^1,2, Kara D. Federmeier^1,3,4; ¹University of Illinois at Urbana-Champaign, USA, ²State University of New York at Albany, USA, ³Beckman Institute for Advanced Science and Technology, University of Illinois Urbana-Champaign, USA, ⁴Program in Neuroscience, University of Illinois Urbana-Champaign, USA

Predictive processing plays a key role in many aspects of human cognition, including language comprehension. However, to directly capture pre-activation, neural activity must be measured preceding the anticipated stimulus. Representational Similarity Analysis (RSA) combined with EEG offers a promising approach by correlating neural representations across the scalp to identify similarities between stimuli, allowing for the detection of pre-activation signals. Hubbard & Federmeier (2021) used RSA to quantify the correlation between neural signals measured just before and after the presentation of a potentially predictable word, finding a similarity signal that was sensitive to contextual constraint. Given the increased use of RSA with EEG to study prediction in language, it is important to ascertain how the number of EEG channels (spatial density) might influence the power of the observed similarity signal. To address this, we are building on Hubbard and Federmeier (2021), recording EEG to sentences varying in constraint. Using RSA to compare EEG activity patterns elicited by pre-final and final words, we are replicating the finding that peak similarity is observed in an early window (100–230 ms) after word presentation and is modulated by sentence constraint, indicating pre-activation of the final word. Critically, we are evaluating the effect of varying the number of EEG channels (from 10–60) used for the RSA analysis. Initial findings suggest that similarity measures increase with more channels, but it is unclear if higher density improves power to detect condition-related differences. This work will ultimately provide valuable insight for any researcher using the RSA technique.

Topic Area: LANGUAGE: Other