Purchase this article with an account.
Jason Haberman, Jordan Suchow, George Alvarez; The visual system adapts to average orientation. Journal of Vision 2011;11(11):881. doi: 10.1167/11.11.881.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The visual system efficiently represents summary statistical information in multiple domains. For example, we readily extract the average orientation of a set of lines, the global direction of a collection of moving dots, and the average expression of a crowd of faces. In a series of matched behavioral and fMRI experiments, we explored the neural representation of ensemble coding using an adaptation paradigm. Observers viewed sets of gabors which were jittered around a global orientation (e.g., ±22.5 deg from horizontal). In one experiment, observers adapted to a set of gabors and judged whether a test set was presented to the left or right of fixation. The test set was either identical to the adapting stimulus, or different in one of two ways: (1) each gabor changed orientation and the global orientation changed, or (2) each gabor changed orientation by the same amount while the global orientation remained the same (“ensemble same”). For each of the three conditions, we measured the elevation in contrast threshold after adaptation as a function of eccentricity. As expected, elevation in contrast threshold was greatest when the test stimulus was identical to the adapting stimulus, and least when the test stimulus was different both locally and globally. Critically, the higher the eccentricity, the more closely the ensemble same condition matched the identical condition, suggesting adaptation to the average orientation. In a complementary fMRI experiment, we measured BOLD response to the same three test patterns using an adaptation paradigm. We found adaptation (i.e., reduced signal) in lateral occipital areas when the ensemble was globally the same. Remarkably, observers perceived and adapted to the average even though an exemplar of the average orientation was never present in the set. Taken together, these results suggest the existence of neural mechanisms tuned to the statistical properties of a set.
This PDF is available to Subscribers Only