Abstract
Recent neuroimaging studies have shown that brain activity generated during mental imagery in early visual cortex encodes the same low-level visual features that are encoded during perception. This finding was obtained by fitting voxel-wise encoding models to activity generated during perception, and then using those models to predict activity generated during imagery. The success of this analysis lends plausibility to the concept of an imagery receptive field (iRF): a region of the mental visual field in which imagined stimuli evokes real brain activity. To determine if iRFs are in fact instantiated in early visual cortex, we measured whole-brain BOLD activity as participants viewed and then imagined previously memorized objects at 8 distinct locations in the visual field. We then fitted voxel-wise encoding models to the activity generated during mental imagery. These models assigned an iRF size and center to each voxel. In general, the iRF sizes obtained by this analysis were larger and more scattered than their visual counterparts. To visualize the iRFs, their centers were mapped to visual field space and colored according to the amount of activity generated during perception of each of the 8 object locations. This visualization revealed that iRFs do indeed characterize the relationship between activity and imagined stimulus location, and that the relationship is largely conserved during perception. To validate the iRFs we performed a classic population-vector style analysis of population activity. Specifically, we performed a vector summation of the iRF centers scaled by brain activity. For both perception and imagery, the length of the resulting population vectors were significantly greater than could be obtained if the receptive field centers were unable to predict measured brain activity (p< 0.01, randomization test). These preliminary results demonstrate the feasibility and potential power of iRF for exploring the functional role of imagery in vision.
Meeting abstract presented at VSS 2016