Abstract
We sometimes remember more than what we have seen: in the Boundary Extension effect we remember a wider view of a scene than the original image. We examined temporal boundary extension (TBE) in action perception: do observers remember seeing more frames of a movie than was shown? In Experiment 1, observers viewed short YouTube movies of solo actions, human-object interactions, or human-human interactions. After a mask, a test clip was shown from either long before the start of the first movie, right before the start, right at the start, halfway through middle, right before the end, right after the end, or long after the end. Observers reported their confidence in whether or not the test clip was seen in the first movie. We found more false memories for test clips of single actors for events before the movie started, and more false memories for clips of human interactions of events from after the movie ended. Memories of events were extended to the past for solo actions, but extended to the future for interactive activities. As a control condition, the same movies and test clips were also shown backward in time, which revealed no difference between the movie types, suggesting that extension was not due to local motion prediction. In Experiment 2, we used point-light actors without scene backgrounds. Videos depicted either two people performing one of four interactions or only showed one of the two individuals in action. As in Exp. 1, backward TBE to the past was observed for solo actions, while forward TBE to the future for two-person actions. These results suggest that action representations encoded in memory are extrapolated beyond the visual input, and the temporal direction of such extrapolation depends on the degree of action prediction evoked by the stimuli.
Acknowledgement: NSF grant BCS-1655300 to H.L.