Abstract
A fundamental question remains open: How "constructed" is everyday perception in complex scenes? Evidence supports opposing views, while implicating interaction between bottom-up and top-down perception. Neisser's (1976) perceptual cycles framework integrates the opposing views, while positing a critical role for top-down schemata over time: Once activated (often by stimulation), a schema (or attentional set) determines which stimuli are most likely to be perceived. Little research tests this idea directly, beyond inattention studies where (e.g.) a gorilla is missed once. Here, we measure the power of an active schema to inhibit continuing conscious perception of known, visible, repeated, object-based events. Observers first learned the search tasks and the animated object-events (distractor events and color- or motion- feature target events), demonstrating 100% accuracy during training. Testing then began, with multiple objects alive contemporaneously, forming a continuous event stream; performance asymptoted at 80%. A schema (set) was activated by presenting only one target type for 2 blocks of trials. The schema was predicted to cause efficient perception of that event type ("set targets"), but not the other event type ("other targets"). The other targets were added in the next, critical block but (Experiment 1) only in outer regions, away from central attention. Accuracy remained high for the set targets, but was drastically lower for other targets (48% lower hit rate), consistent with predictions. When the other targets were added in the critical block at the center of attention (Experiment 2), performance dropped drastically for all target types (46% lower), implying set needed to be changed. Performance gradually returned to asymptote over 2 blocks in both experiments, as set was re-instantiated. These are strong schema effects on conscious perception, with object-events that are known, practiced, repeated, and highly visible.
Meeting abstract presented at VSS 2014