Representational Geometries of Perception and Working Memory: A Pilot Study
This study aims to compare the neural representational geometry of visual perception and visual working memory using human fMRI. In our pilot experiment, observers viewed a face-scene blended image (sample) and selectively attended to either the face or scene aspects of the sample. They then judged whether the attended aspects were the same as or different from that of the subsequent test image. In three out of four observers, we found that the coding axis, which distinguishes faces and scenes, was rotated when perceptual content transforms into working memory, not only in the visual cortex but also in some higher association cortical areas. In these regions, exclusive combinations (i.e., perceived face and memorized scene vs. perceived scene and memorized face) were successfully classified by linear decoders, indicating that perception and working memory are represented in partially distinct subspaces. Such specific coding may help prevent confusion between perception and memory. Meanwhile, however, cross-decoding of visual contents across perception and working memory was also robust, indicating common coding components between them. Such common coding may help to explain why even though we generally have no difficulty in distinguishing working memory content from perceptual content, there is yet a sense in which their contents are qualitatively similar. Hereby, we pre-register the hypotheses, experimental design, analysis plan, and stopping rule for data collection before examining whether these preliminary results can be replicated in a larger sample.