Predictable Object Motion is Extrapolated to Support Visual Working Memory for Surface Features

Space and time support visual working memory (VWM) by providing incidental reference frames for task-relevant information. While this has been studied with stationary items, natural scenes also contain moving objects, whose positions change over time, often in a predictable manner. We investigated if predictable item motion is leveraged to facilitate VWM for surface features. In a dynamic change-detection task, participants memorised the colours of three disks moving at constant speed in different directions before disappearing. After a retention interval, the disks reappeared (a) at the movement endpoint locations where they had disappeared (b) at positions spatiotemporally congruent with their previous motion (where they would have been had they continued their movement), (c) at positions with a temporal offset (consistent with a change in speed while out of view) or (d) at positions with a spatial offset (consistent with a change in movement direction). Performance decreased with increasing temporal or spatial offsets relative to congruent positions, indicating that the memorised items' positions were remapped to their anticipated future locations. This updating of positions in spatiotemporal reference frames, however, only occurs if motion extrapolation allows for reliable predictions of where occluded items will reappear. In a task context with unreliable motion patterns, the congruency effect diminished over time and performance instead increased at the movement endpoints. A second experiment confirmed this influence of motion reliability. Thus, predictable motion is extrapolated to update spatiotemporal reference frames in VWM, supporting memory for surface features and thereby contributing to visually guided behavior in dynamic environments.