Abstract
Visual adaptation has traditionally been viewed as passive, but more recent models propose that it depends upon active inference of environmental change. To test this theory, we measured adaptation to environments that were more or less variable, under the assumption that variability will slow inference and hence adaptation. Apparent contrast of a vertical grating was measured following adaptation in high or low contrast conditions. Adapting gratings (1.5 cyc/deg, 5.4 deg) were presented for 4 sec on one side of fixation, and their contrast was updated every 200 msec from a distribution of contrasts. In low variance sessions, narrow uniform distributions centered on 0.1875 and 0.7125 were used for the low and high contrast conditions respectively. In high variance sessions, the distributions were the weighted sum of three narrow uniform distributions, centered on contrasts of 0.1, 0.45, and 0.8. For the low contrast condition, 0.1 was weighted highly, and the overall distribution mean was 0.1875. For the high contrast condition, 0.8 was weighted highly, and the overall distribution mean was 0.7125. The adapting sequence was followed by a test patch (1.5 cyc/deg, 4.5 deg, 25% contrast) presented along with a "match" patch on the unadapted side of fixation for 200 msec. Subjects judged which appeared higher contrast, and the match contrast was adjusted using a staircase procedure. 60 trial blocks with low contrast adapters alternated with blocks using high contrast adapters. For each block, adaptation rate was estimated by an exponential fit to the staircased match contrasts. The rate of adaptation when the adaptor changed from low to high contrast was slower for the high variance session than the low variance session (p < 0.01). This suggests that adaptation is slower when environmental changes are more difficult to detect, consistent with adaptation being controlled by an inference process.
Meeting abstract presented at VSS 2017