Abstract
Could memory for unseen layout beyond the edges of a view (boundary extension; BE) facilitate the integration of successive views? For this to be the case, BE would 1) need to occur following a retention interval as brief as a saccade, and 2) survive a gaze shift. On each trial, 3 photographs of unrelated scenes were presented for 325 ms each in an RSVP sequence. They were followed by a masked retention interval (42, 100, or 250 ms; between-groups variable) and repetition of a scene from serial position 1, 2, or 3. The repeated picture remained on the screen while the observer rated the view as the same, more close-up, or more wide-angle than the target (5-point scale). In Experiment 1 (N=108), targets were either close-up or wide-angle views, and test pictures were either the same view or the alternate view. The rating pattern diagnostic of BE occurred across all conditions. Most important, close-ups yielded BE at all masked intervals and at all serial positions. After a retention interval as brief as 42 ms, observers tended to rate the same close-up as being more close-up than before, indicating that memory for the target included extended boundaries. In Experiment 2 (N=31), we recorded observers eye movements (EyeLink II). All targets and test pictures were identical close-ups. Following a 42 ms mask, test pictures appeared on either the same side of the screen as the RSVP sequence or the opposite side, necessitating a gaze shift. BE occurred in both cases for all target serial positions (mean retention interval on trials with a gaze shift = 346 ms). Spatial extrapolation of layout occurred rapidly for all scenes in dynamic presentations, even across gaze shifts. BE is likely available in transsaccadic memory to facilitate view integration during normal visual scanning.
This research was supported by NIMH Grant MH54688.