Abstract
What is the nature of the information transformation that occurs between V1 and V2? Receptive fields (RFs) are larger in V2, in humans (Smith et al. Cereb. Cort. 2001, 11, 1182) as in other primates. Is there simply a biased projection from V1 to V2 that favours neurons sensitive to low spatial frequencies, as has been suggested? Or does the increase in RF size in V2 reflect integration of information across space? These alternatives have quite different implications for the function of V2. Human observers were presented with a stimulus consisting of a slowly expanding, centrally fixated ring of band-pass spatially filtered 2D dynamic noise. This was intended to isolate a subset of neurons with a particular spatial frequency sensitivity. Functional MRI (1.5T GE LX/Nvi) was used to estimate the average RF size of this neuron subset in human V1 and V2, based on the duration of the response elicited as the stimulus passes through the receptive fields of the neurons in each voxel (Smith et al. op. cit.). If information is passed from V1 to V2 with no major change in RF properties (i.e. V2 just contains a higher proportion of low-spatial-frequency neurons), then a given band-pass filtered stimulus should activate neurons with the same RF size in both areas. The size difference we have previously observed with broadband stimuli should therefore disappear. Results: In both V1 and V2, mean estimated RF size decreased as the stimulus spatial frequency was increased, confirming that neurons in both areas are tuned for spatial frequency. Receptive fields were larger in V2 than V1, even with band-pass filtered stimuli. This suggests that receptive fields in human V2 are generated by integrating the outputs of a group of V1 neurons that have a common spatial frequency sensitivity but slightly different RF locations.
Supported by The Wellcome Trust