Abstract
The present research investigates the effect of confidence on perceivers' ability to make visual discriminations. Confidence is affected by the difficulty of the discrimination at hand. Here we show that it also depends on the perceivers' confidence during preceding discriminations; i.e., confidence is state-dependent. The evidence for state-dependence takes the form of confidence hysteresis: How well 4–6 year old children and adults perform in a two-alternative forced-choice numerical discrimination task depends on whether they have had a prior history of easier discriminations or of harder ones. Participants had to identify whether a set of blue or yellow dots was larger by number; in these tasks, the numerical ratio of the two sets determines the difficulty of the judgment, and lower ratios (e.g., 11:10 dots) result in poorer performance, and, presumably, lower confidence, that higher ratios (e.g., 22:10 dots). We manipulated confidence by placing participants in either an Easy-to-Hard condition, where all the easiest trials were at the start, or in an Hard-to-Easy condition, where all the hardest trials were at the start. Additionally, in the first experiment, we gave participants accurate feedback after every trial. Both children and adults performed significantly better on the Easy-to-Hard compared to the Hard-to-Easy condition, suggesting that confidence hysteresis modulated their performance. The effect was especially pronounced for the Hard-to-Easy children, whose performance was so poor that it resembled the discrimination abilities of 9-month old infants; it is as if the children had “given up”, despite having the ability to discriminate the two sets. In the second experiment, we removed external feedback. The adults still showed a significant difference between the two conditions, but the children did not, suggesting a developmental change in the effect of feedback on perceptual confidence.