Abstract
How a scene is represented is inextricably tied to its spatial and temporal context. When we navigate in the world, we build expectations about one view leading to the next. Previous work has found representations of visual scenes in scene-selective cortex (parahippocampal place area, PPA) depend on temporal context. Specifically, repetition suppression for repeated scenes was greater when they appeared after the same previously encountered temporal context than for scenes appearing after a novel context. In this study, we ask whether representation of a scene's context is tied to the level of the exemplar, or whether it is encoded more flexibly, at the abstract level of the scene category. To test this, we use the repetition suppression paradigm. During fMRI, participants (ages 18-27, n = 18) viewed a stream of scenes while performing an indoor/outdoor decision task. The temporal context was generated by repeating 'triplets' of scenes (e.g., ForestA-FieldA-KitchenA). There were three temporal context conditions—Novel: third item repeated preceded by novel items (e.g., OfficeA-HighwayA-KitchenA), Repeated-Exemplar: all three exemplars of a triplet repeated (e.g., ForestA-FieldA-KitchenA), or Repeated-Category: third item repeated preceded by new exemplars of repeated categories (e.g., ForestB-FieldB-KitchenA). We found significant repetition suppression for Repeated-Exemplar and Novel conditions in posterior PPA, but not anterior PPA. Moreover, we found significant repetition enhancement for the Repeated-Category condition in retrosplenial cortex (RSC). We propose that the repetition enhancement in RSC reflects the prediction error arising from a mismatch between prediction and stimulus presentation in the Repeated-Category condition. Current results suggest that neural response of scene-selective cortex is sensitive to temporal predictions made at multiple levels, and especially to the mismatch of the prediction and the stimulus presented at any level. Such sensitivity may help distinguish scenes with similar views from different places that have different navigational context.
Meeting abstract presented at VSS 2016