Abstract
Training on a visual task can, in some cases, produce changes in neural response in early visual cortex. But it remains unknown whether these effects reflect receptive field changes of early visual neurons or instead reflect feedback from later visual areas. We addressed this issue by testing whether learning can affect the earliest visual evoked potential (VEP) component, termed C1. Five subjects were trained to detect a low contrast sinusoidal grating of fixed diagonal orientation in a 2AFC task for over 28 days. Training took place at a single location in a selected quadrant of the visual field, and locations at the same eccentricity in the three other quadrants were used as controls to test the specificity of learning. Horizontal gratings in all four quadrants were also used as controls. Before and after the training, subjects' detection thresholds for all eight stimuli (4 locations × 2 orientations) were measured. In separate sessions, VEPs in response to briefly presented high contrast versions of the gratings were also measured. To control attention, subjects performed a demanding RSVP letter task at fixation while EEG data were acquired. Behavioral thresholds reliably decreased for the trained pattern. For stimuli at the trained location, amplitudes of the C1 component of the VEP (80–100 msec post-stimulus) were 22% greater for horizontal patterns than diagonals prior to training. Following training on diagonals, amplitudes were 19% greater for diagonals. Training produced only a small change in the difference between horizontal and diagonal amplitudes for patterns presented at the untrained locations ([[lt]]6%). Because the C1 is likely associated with the initial activity in primary visual cortex, these results suggest that visual perceptual learning can increase neural response through local receptive field changes rather than through feedback from higher areas.
Supported by NIHRO1EB007920 and a grant from UMN/IEM.