Abstract
The perception of stimuli depends, in part, on the properties of stimuli seen in the past. We examined the characteristics of observers’ responses based on the stimuli they were shown and the responses they gave going back several minutes, Critically, we manipulated the statistics of the stimuli to be in a position to model the dynamics of these history effects on perception. Observers were presented with gratings in one of five orientations: two easy to discriminate (supra-threshold) ‘far-left’ and ‘far-right’, two intermediate (near-threshold) ‘middle-left’ and ‘middle-right’, and one ambiguous ‘middle’. Far-left and far-right orientations were seen most frequently, and observers were not asked to respond to these ‘standard’ stimuli. When presented with one of the three orientations between the left and right standards observers had to indicate whether the orientation was closer to, or farther from, the randomly selected standard orientation which preceded it. In three contiguous conditions of 960 trials, left and right standard stimuli were presented in equal proportions, then biased to one side, then balanced. Not surprisingly, the data show a negative correlation between the orientation bias of the stimuli in the second condition and observers’ mean responses, a negative aftereffect. While this negative correlation dominates in the near past, the data also suggest a weaker, positive correlation with stimuli seen in the more remote past (1000 trials). Critically, the aftereffect observed in the second condition disappears before the end of that condition, which is a prediction of the predictive adaptation model (Chopin and Mamassian, Current Biology, 2012). These results confirm earlier findings highlighting the contribution of remote past history on perception Adaptation appears to be predictive in that the statistics of the remote past are used to estimate a ‘norm’ against which present stimuli are compared.
Meeting abstract presented at VSS 2015