Abstract
Visual confidence refers to the ability to predict one's own performance. Thus for confidence to be useful, it must be well calibrated with performance. How then does confidence change when percepts are modified due to adaptation (aftereffects). Observers were instructed to estimate their confidence across 2 different perceptual tasks. In an orientation discrimination task, a 500ms series of Gabor adapters (10deg, 2cyc.deg-1, 50% contrast, random phase every 50ms), oriented -20, 0 or +20deg from vertical, was followed by a 100ms blank before a 100ms test target, oriented in 7 values from -2.1 to +2.1deg. Similarly, in a color discrimination task a 500ms adapter patch (10deg, high contrast black edge) with -0.75, 0 or +0.75 (green, gray or red) of the monitor maximum gamut along the "red/green" axis of the DKL colorspace was followed by a 100ms blank, then a 100ms test patch with a gamut ranging in 7 values from -0.35 to +0.35. Following a pair of orientation and color stimuli, whose order varied from block to block, observers were instructed to report in which of their 2 prior answers they felt more confident (most likely to be correct). Observers successfully switched between the two tasks and were able to make a confidence judgment across the two different visual modalities. Clear negative aftereffects were observed in both tasks (means > 1SD). Response times and confidence judgments were shifted by the adapters by about the same amount as the psychometric functions. Normal cumulative, probability density and inverse of the probability density functions were fitted to respectively performance, response times and confidence. Correlations on fit parameters normalized by SD were high (>0.7), with a regression slope of about 1, suggesting that confidence is mediated by the same mechanism that mediates perceptual reports and response times.
Meeting abstract presented at VSS 2016