The fidelity of visual memory for scenes

Previous research investigating visual memory for scenes has focused predominantly on memory for intentionally encoded static scenes, which may not be reflective of the way in which we view scenes every day. Across seven experiments, a range of experimental paradigms were used to test the fidelity of visual memory for natural, everyday scenes, over different periods of time. This included experiments which used incidental encoding and dynamic scene stimuli. In two initial experiments (Chapter Two), participants first viewed static scene images incidentally, whereby they were unaware of the later memory test. An explicit recognition task was later given (a two-alternative forced choice task). In addition, an adapted version of the recent-probes task, an implicit memory task, where participants were again unaware their memory for earlier viewed scenes was being tested. Here, participants took part in a task in which previously viewed scenes were embedded in certain trials; participants’ responses to these old stimuli compared to novel stimuli were assessed. Findings from these experiments support the notion of a detailed memory trace for scenes that endures over several days, even when scenes are incidentally encoded, and not explicitly probed. The research that uses dynamic scene stimuli is then reported. To enable this, a stimuli set of dynamic scenes was created, and through two experiments, separate groups of participants rated these scenes on a number of attributes, including complexity, distinctiveness and attractiveness (Chapter Three). Using these novel stimuli, three experiments are then reported (Chapters Four and Five). Participants viewed short film episodes depicting everyday dynamic scenes, followed by a recognition test in which they were presented with selected static frames taken either from those film episodes, or highly similar foil images taken from the same scenes, but from a different time point to the extract participants viewed. Participants performed well and above chance when they were given a memory test immediately or after a longer delay period, but were reliably more accurate when tested immediately. Furthermore, to understand the detail in which the dynamic scenes were remembered across time, accuracy for static ‘target’ frames taken from the beginning, middle and end of scenes was compared, to test if certain parts of a scene are better remembered than others. Overall, accuracy was good for all target types, regardless of temporal position. When tested immediately, there was some evidence that accuracy was higher for target frames taken from the middle or end of the films compared to the beginning, which is consistent with outcomes for serial order recency effects in visual memory. The results are interpreted conceptually within the notion of a highly detailed memory ‘map’ or trace, preserving spatiotemporal detail from recent dynamic visual episodes. The experiments reported in this thesis demonstrate the fidelity with which natural, everyday scenes are remembered. This is the case even under circumstances of incidental encoding and retrieval, varying periods of delay between encoding and test, and when scene stimuli are static or dynamic.