The visual system represents overall statistical, not individual properties of sets. It has been suggested that these summary representations evolved as a complimentary strategy to focused attention, circumventing the visual system’s capacity limits. We propose that a primary role of global statistical representations is to build and maintain a stable spatiotemporal context. If summary representations are the basis of spatiotemporal stability, then they should occur in multiple spatial reference frames in order to maintain stability over time and across gaze and head shifts. To test this idea, we used mean size, which is an adaptable attribute encoded over a single visual dimension in a qualitatively different manner than the sizes of individual objects. We examined whether mean size was adaptable across retinotopic, spatiotopic, hemispheric, and monocular frames of reference, as well as whether representations of mean size are built in different temporal windows within different spatial reference frames. In three experiments, we adapted observers to patches of small and large sized dots in opposite regions of the display (Left/Right or Top/Bottom), and tested their perceptions of the sizes of single test dots presented in the adapted regions. We observed a negative adaptation aftereffect, such that participants perceived a test dot presented in the area adapted to large dots as being smaller than the same sized dot presented in the region adapted to small dots (and vice versa) in retinotopic and hemispheric, but not spatiotopic, coordinates. This aftereffect also transferred between eyes. Finally, varying the duration of serially presented streams of adapting dots, we found that retinotopic, hemispheric, and binocular representations emerged during different temporal windows. Overall, our results suggest that mean size is represented across multiple spatiotemporal frames of reference, suggesting that such global statistics may play a fundamental role in maintaining perceptual stability. Meeting abstract presented at VSS 2013
Read full abstract