Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster A in Poster Session A: Tuesday, August 4, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms

Representational dynamics in inferotemporal cortex depend on image manifold scale

Ammar I Marvi¹, Jacob S. Prince¹, George A. Alvarez¹, Talia Konkle¹; ¹Harvard University

Presenter: Ammar I Marvi

Understanding how natural images are represented in IT cortex has long been a central goal of visual neuroscience, with work increasingly focused on uncovering the principles that govern their extended temporal dynamics. Given views of cortical tuning as a structured map of image space, one source of insight into a region’s dynamics may lie in the image manifold itself: probes with broadly sampled image sets emphasize global structure, whereas preferred image subsets concentrate on local sectors of image space. Using time-resolved macaque electrophysiology from the Triple-N dataset, we compared representational geometry across global and local manifold scales in 35 functionally distinct patches of IT cortex. At global scale, IT representations evolved smoothly over time. In contrast, local-scale geometry showed sharper temporal reconfiguration ~100–250 msec after image onset. Our results suggest that IT subregions have relatively stable and smooth tuning in global image space, whereas the local image code shows structured and often dramatic reconfiguration over time.

Topic Area: Computational Models of Vision & Visual Cortex