Abstract
A great deal of previous research has highlighted a range of incorrect beliefs that humans have about random processes. For example, when humans encounter sequences of outcomes that are generated randomly, Gambler’s Fallacy is commonly observed. If a particular outcome has been observed less frequently in the recent past, humans tend to believe that it is more likely to occur in the future. While typically not an explicit focus of perceptual learning studies, temporally independent processes are nearly always used to generate stimuli. As such, the presence of biases about temporal sequencing could disrupt our ability to measure true perceptual sensitivity. In other words, if participants viewing an oriented Gabor are more likely to say that it was counter-clockwise relative to a reference orientation if the previous two outcomes were clockwise, and vice versa, unless these temporal dependencies in choice were modelled, this would appear as “noise” in analyses of perceptual sensitivity. To examine these issues, we manipulated temporal dependencies in sequences of perceptual learning stimuli (Gabor orientation discrimination task: clockwise/counterclockwise from a reference angle). Three conditions with different temporal dependencies were utilized in a within-subject design (N=23): truly independent, “sticky” (current trial more likely to be the same as that on the previous trial), and “switchy” (current trial more likely to be the opposite of previous trial). Preliminary analyses not only suggest clear temporal dependencies in choice behavior, but more importantly, choices in the truly independent condition appeared most similar to the “switchy” condition, as would be predicted by Gambler’s Fallacy. Interestingly, these preliminary analyses also suggest the bias is “post-decision,” rather than the result of an optimal integration process (i.e., where in an optimal integration process, the degree to which the temporal bias impacts choices would be maximal for orientations at which perceptual uncertainty was maximal).