Abstract
There has been an ongoing debate in the visual search literature on whether or not visual search has memory. One manipulation to test if memory is used in visual search has been to randomize the location of stimuli in an image every 111ms, which prevents observers from tracking the locations of previously inspected items. Horowitz and Wolfe (1998) used this method and found no significant differences in search slopes between static and random conditions, leading to the conclusion that visual search has no memory. Here we revisit this claim. We reason that memory in search should only be necessary for a search where there is little guidance. Thus search may appear memoryless when the search task allows for adequate guidance, but search may rely on memory in more difficult search tasks when guidance is ineffective. In Experiment 1 we replicated Horowitz and Wolfe’s findings when observers searched for a T among Ls. But when we made the task a search for the same T among offset Ls observers in the random presentation condition were very close to chance performance, making it difficult to interpret search slopes. However, error rates suggest that presentation type (static v. random) interacted with stimulus type (easy v. hard) suggesting a role for memory with the harder search. In Experiment 2 we sought to increase overall accuracy to avoid chance performance. We decreased the set sizes in Experiment 2 from 8, 12, and 16, to 4, 6, and 8, while increasing the stimulus presentation duration from 111 to 160ms. Again we found that poorer accuracy in the difficult stimuli condition is moderated by the presentation type. The data suggest that memory is not necessary in searches where guidance to the target is efficient, but memory is necessary for high performance in searches with limited guidance.
Meeting abstract presented at VSS 2015