A successful visual system extracts meaning from stimuli that are spread over space and time. To do so requires integrating and segregating features at multiple scales. In this talk I describe a method to measure how sites in visual cortex pool information across the spatial extent of the image (spatial windows) and across time (temporal windows). The method is applied to signals from functional magnetic resonance imaging (fMRI) and intracranial electrical recordings in patients. One set of studies reveals a compressive nonlinearity in spatial summation. This effect becomes increasingly more pronounced in higher visual areas. The compressive spatial summation we observe has implications for a wide range of visual computations, including tolerance for changes in the position and size of objects. A second set of studies examines temporal windows across the visual pathways. As with space, we find that temporal signals are combined subadditively, and that the subadditivity increases in higher visual areas. The response patterns we observe in both space and time are explained in terms of a model with a small number of simple, interpretable computations. The parallels in the responses to space and time suggest that for the two domains, cortex uses a similar processing strategy to achieve higher-level and increasingly invariant representations of the visual world.
- [PDF]Wandell BA, Winawer J. (2015) Computational modeling and population receptive fields. Trends in Cognitive Sciences, 19:349-357. PMC4484758 | doi:10.1016/j.tics.2015.03.009
- [bioRxiv]Zhou JY, Benson NC, Kay KN, Winawer J. Compressive temporal summation in human visual cortex. (Currently under peer review)