Abstract
Vul and Pashler (2008) reported that the average of two guesses from the same observer was more accurate when guesses were made weeks apart versus when they were made in immediate succession. As these results suggest that an observer's guesses become more independent with a greater delay, we wondered whether the same may be true for multiple samples of a visual percept. To this end, we investigated whether averaging multiple glances improves pattern recognition. In an n-back paradigm, observers estimated the density of a briefly presented texture display, which repeated one-, two-, three-, four-, or at a random number of trials after its initial presentation. To measure the degree of dependence within response pairs, we computed the average of two density judgments of the same display as a function of the number of intervening trials. Overall, the average of two judgments of the same stimulus became more accurate as judgments occurred farther apart in time, indicating that multiple perceptual samples of a pattern become more independent, and thus more useful, with greater intervening delay. Given that participants were not asked to explicitly report whether a display was repeated, but only to report the perceived density of the pattern on each trial, and generally reported not noticing that displays repeated when questioned after the experiment, this paradigm may provide a new, implicit measure of visual short term-memory decay.