Abstract
The relationship between the intensity of the light reflected from an achromatic object and the perceived lightness of the object appears depends on visual context. It has been difficult to relate this complexity to the activity of individual neurons, because neurons respond in varied ways to stimulus intensity. Can a population decoding approach clarify the neural underpinnings of perceived lightness? We employed stimuli derived from Adelsons checker-shadow illusion, such that probe disks presented on checkerboard images were seen to lie either within a shadowed region (shadow condition) or within a luminance-matched region without a shadow (no shadow condition). We evaluated whether similar changes in the lightness of the probe across the two conditions were revealed by human psychophysics and by lightness estimates obtained via decoding of the responses of several dozen cortical neurons in either area V4 or V1 of rhesus monkeys. Our psychophysical experiments confirmed and quantified the checker-shadow illusion for our stimuli: probes with the same intensity were perceived to be lighter in the shadow condition than in the no shadow condition. When we decoded the intensity of the probes using population responses of V4 neurons to the same stimuli, we found that the decoded intensities for the shadow condition were consistently higher than those for the no shadow condition. Moreover, there was a quantitative match between the psychophysical and V4 neural effects. In contrast, although probe intensity could be decoded from the population of V1 neurons as accurately as from the V4 neurons, the V1 decoding differences were not in agreement with the perceptual illusion. This result suggests that the checker-shadow illusion arises at least in part from cortical computations. More generally, our data support the notion that decoding the responses of neural populations can shed light on the neural correlates of complex perceptual phenomena.
Meeting abstract presented at VSS 2014