Abstract Detail


Talk Sessions

Poster Sessions

All Presentations


Emerging Representational Geometry for Objects Predicts Reaction Time for Categorization

52.25, Tuesday, 20-May, 10:45 am - 12:30 pm, Talk Room 2
Session: Object recognition: Neural mechanisms 2

J. Brendan Ritchie1,2,4, David Tovar3, Thomas Carlson1,2; 1Department of Cognitive Science, Macquarie University, 2ARC Centre of Excellence in Cognition and its Disorders, Macquarie University, 3School of Medicine, Vanderbilt University, 4Department of Philosophy, University of Maryland, College Park

Recent human and primate neurophysiological studies have characterized a representational geometry for visual objects in inferior temporal cortex (ITC), in which individual exemplars are both discriminable and cluster based on object category (e.g. faces and body parts). Two outstanding questions are: when does this representational geometry emerge; and how does the structure of the geometry relate to behavior? Recent findings have made important progress on these questions. First, multivariate pattern analyses in conjunction with Magnetoencephalography (MEG decoding) have described the emerging representational geometry of objects in high temporal resolution. And second, using human fMRI, it has been shown that reaction times (RT) for object categorization are predicted by the structure of ITC’s representational geometry. We build on these findings and show RTs for object categorization can be predicted by the geometry of the brain’s representation of objects shortly after the presentation of a visual stimulus. In the present study, participants categorized as quickly and accurately as possible images of 12 animate and 12 inanimate object exemplars. Time-resolved MEG decoding was used to reconstruct the brain’s representational geometry of the exemplars on a moment-to-moment basis (20 ms resolution). Using linear discriminate analysis, for each time period we computed a discriminate boundary through the representational geometry that separated animate and inanimate exemplars, and then calculated the distance of individual exemplar representations from the boundary. Classical signal detection theory (SDT) predicts a negative correlation between distance from the boundary and RT. In accordance with the prediction of SDT, and previous findings using fMRI, we found that distance from the decision boundary negatively correlated with RTs beginning at 200 post-stimulus onset. Our results reveal when the structure of the brain’s emerging representational geometry for objects predicts behavioral RTs, supporting the contention that “representing” is a core aspect of decision-making for categorization.

< Back