Divisive normalization versus inhibition during visual motion integration in humans

摘要

The amount of time required to discriminate the direction of motion of a stimulus depends on its size and contrast in apparently complex ways. For example, at high contrast, large gratings require more time than those that are smaller. This effect is referred to as spatial suppression, as it is believed that large gratings stimulate suppressive receptive field surrounds, thereby reducing neural responses to stimuli within the receptive field centers. While the neural processes underlying this behavior are not fully known, recent evidence suggests a significant role for neurons in area MT. It is also not clear what role inhibition may play in this perceptual suppression effect. We show how a computational model that includes divisive normalization provides a unifying explanation for this perceptual behavior. Further, responses in human MT complex (hMT+) measured with fMRI reflected key predictions of the model, suggesting that divisive normalization performed in this region underlies visual motion integration in humans. Next, we examined the role of inhibition during spatial suppression using two different methodological approaches. First, we quantified the concentration of GABA in a region of lateral occipital cortex including hMT+ using MR spectroscopy. We found that higher GABA concentrations in this hMT+ region were associated with reduced motion duration thresholds overall, but not with stronger spatial suppression. Second, the effect of GABA was potentiated pharmaceutically via oral administration of lorazepam, a benzodiazepine which acts as a positive allosteric modulator of the GABAA receptor. Contrary to our spectroscopy findings, motion duration thresholds were modestly increased by lorazepam in a manner that was accounted for by the normalization model. Our results point to a critical role for divisive normalization in mediating spatial suppression, as well as a complex role for GABA in determining neural responsiveness, but not in directly scaling the strength of spatial suppression.