Abstract
Visual working memory holds task-relevant visual information for very brief intervals. Yet, previous work also has shown that subjects can store in long-term memory details of items presented briefly only once. Here, we examined whether visual working memory representations of simple displays during a change detection task persist in long-term memory. In two experiments, subjects performed a standard color change detection task. Interspersed between blocks of 6–10 change detection trials were set recall trials, in which subjects were shown a set of colored squares and were asked to report whether they had seen the set before. Half of set recall trials showed a set previously seen within the first four trials of the change detection block, while the other half showed new arrays. In Experiment 1 (Set Size 3; n=50), subjects performed near ceiling on change detection trials (91.83% correct), and also performed significantly above chance on set recall trials (59.17% correct; t(49)=6.6; p< 0.001; BF(10)=680,927.3), suggesting they maintained long-term representations of arrays presented during change detection trials, even across multiple intervening and perceptually similar change detection trials. In ongoing Experiment 2 (n=23), we increased set size to 5 and found that subjects were again successful on both change detection (78.07% correct) and the critical set recall trials (56.39% correct; t(22)=2.44; p=0.014; BF(10)=3.47). Change detection accuracy differed significantly between Experiments 1 and 2 (p< 0.001), but there was no difference in set recall performance across the two experiments (p=0.59). Regardless of set size, accuracy on change detection trials predicted performance on set recall trials, with a .73% increase in set recall accuracy for every 1% increase in change detection accuracy. Together, these results suggest that, representations of the content and configuration of sets of simple, single-feature items persist in long-term memory even after these representations are no longer task-relevant.