Abstract
Reentrant visual processing is theorized to operate in cycles of extraction and confirmation. On first pass, extraction generates perceptual hypotheses, which are then tested by selectively re-sampling from lower levels in the visual hierarchy. It is possible to probe the information contained in these perceptual hypotheses by interrupting and then reinitiating a visual task. A response signature known as rapid resumption provides evidence of perceptual hypothesis confirmation. Here we present two sets of experiments exploring the dynamics and constraints on perceptual hypothesis confirmation. First, in a modified visual search task, we dissociate target detection and response selection by adding a level of response selection following target detection. This design permits testing whether rapid resumption reflects localizing or identifying visual targets. Second, we present convergent evidence using interrupted and resumed pattern matching in a same/different task. Observers are able to rapidly resume both ‘same’ and ‘different’ trials. Interestingly, in all target detection and same/different tasks, the temporal distribution of rapid responses appears to be largely unaffected by stimulus complexity and task difficulty. This strongly contrasts with responses made during constant presentation, and with responses made subsequent to an initial confirmation window. When observers fail to rapidly resume, a later peak of responses is sensitive to stimulus complexity and task difficulty. These findings suggest that the confirmation stage of perceptual processing is robustly consistent across task difficulty and stimulus complexity.