Deepasri Prasad¹, Adam Steel^1,2, Caroline E. Robertson¹; ¹Dartmouth College, ²University of Illinois Urbana-Champaign

Our classic understanding of visual recall proposes that category-selective areas are activated during perception and similarly re-activated when recalling their preferred stimulus (O’Craven & Kanwisher, 2000). Recent work has complicated this "sensory reinstatement" hypothesis, demonstrating variations like topographic shifts between encoding and recall (scenes; Steel et al., 2021) and intermixed neuronal subpopulations selective for encoding and recognition (faces; Quian Quiroga et al., 2023; Chen et al., 2024). Given this variety, are there common principles or characteristic differences across how category-selective areas implement sensory reinstatement? Here, we examine the activation patterns during perception and recall in a range of category-selective visual areas using fMRI. Participants (N=25) performed visual recall and perception tasks during the scan. For recall, participants vividly recalled self-generated personally familiar faces, scenes, objects, and body parts. For perception, participants viewed dynamic videos from the same categories. Using functional areas defined from independent atlases (parahippocampal (PPA), occipital (OPA) place areas; occipital (OFA), fusiform face areas (FFA1, FFA2); extrastriate body subregions (ITG, LOS, MTG)), we identified category-specific regions of interest in both tasks (contrast: category of interest>all other categories). All areas showed some degree of sensory reinstatement, with stronger overall univariate activity for their preferred category in both perception and recall. For almost all areas, the most selective voxels for perception were similarly the most selective for recall. Interestingly, in PPA, OPA, and OFA, the top-most selective voxels differed between the tasks. A significant anterior shift in recall versus perceptual activity drove this effect in PPA and OPA only (all p-values<0.001). OFA activity showed no systematic shift between recall and perception. Instead, OFA contained intermixed subpopulations of mnemonic- and perceptually-selective voxels. Together, this indicates that no single implementation of sensory reinstatement is common across all visual categories; instead, the reinstatement method might reflect unique mnemonic demands of each visual category.

Acknowledgements: NSF Career Award 2144700