Purchase this article with an account.
Vincent de Gardelle, Pascal Mamassian; Contributions of signal strength and reliability to performance and confidence. Journal of Vision 2012;12(9):162. doi: 10.1167/12.9.162.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
In perceptual decision tasks, participants have to choose between several interpretations of noisy sensory input, and their performances reflect different types of uncertainty about the stimulus. For instance, in a motion discrimination task where participants have to decide whether a random dot motion stimulus moves primarily on one side or the other of a reference direction, performance is affected independently by the mean and the standard deviation of the distribution of dot motion directions. Here, these two parameters may correspond to two different sources of uncertainty: the latter is about how precise the estimate of the net motion direction is (signal reliability), while the former is about the relative position between this estimate and the midpoint separating the two choices (signal strength). In the present study, our goal is to assess how these two forms of uncertainty impact on observers’ confidence about their choices, and whether confidence judgments are following expected performance, as predicted by signal detection theory, or rather are more sensitive to one particular source of uncertainty. Using adaptive methods, we kept observers’ accuracy constant while varying conjointly the mean and variability parameters, and we collected confidence judgments in both an objective and a subjective manner. This approach allows us to assess 1) the similarity between subjective and objective measures of confidence judgments, and 2) whether confidence follows preferentially signal strength, signal reliability, or rather the expected performance resulting from both quantities. Preliminary results suggest that observers feel more confident in responding to stimuli that lead to better performance, as expected, but also that observers tend to prefer stimuli that provide greater signal reliability, even when the influence of performance is accounted for. On the methodological side, it appears that the objective measure of confidence is also a better predictor of choice accuracy than the subjective measure.
Meeting abstract presented at VSS 2012
This PDF is available to Subscribers Only