Abstract
Experiments using rapid presentation of large numbers of picture have demonstrated impressive recognition for pictures that have been shown for less than a second. What are the visual processes that support successful picture memory? We hypothesized that observers encode two separate components: gist and/or texture information, which might be extracted without attention, and some recognized objects requiring selective attention. METHODS: We compared memory for scenes, objectless textures and for scenes and textures shuffled in checkerboard fashion. Shuffling should disrupt object perception in scenes with little effect on texture perception. We used a concurrent visual search task to divert selective attention from the pictures. We hypothesized that this would block object recognition, impairing memory for scenes but not textures. Os viewed scenes and textures in single and dual task conditions. Each condition consisted of 32 training trials followed by 32 test trials. In single task conditions, Os viewed 32 pictures for 500 ms each, followed by 32 test trials where they classified pictures as old or new. In dual task conditions, Os either performed the visual search task or an auditory control task during initial viewing of the pictures. RESULTS: Unsurprisingly, memory for scenes was better than memory for textures or shuffled scenes in single task conditions. Memory for all stimuli was impaired by concurrent visual search, falsifying the hypothesis that texture memory would be independent of attention. Visual search imposed a greater cost than concurrent tone monitoring, suggesting a specific effect of visual selective attention on visual memory as opposed to a more general dual-task cost. CONCLUSION: These results indicate that successful picture memory requires the involvement of selective visual attention during encoding. They cast doubt on the notion that some aspects of picture processing can entirely escape the “bottleneck” of attention.