Abstract
Forty distinct scenes of furnished rooms were generated on a computer and three “snapshots” of each room were taken from different viewpoints in a horizontal arc. When a single scene was presented for .5 s followed by a mask and immediately tested for recognition, participants could correctly identify the old (previously viewed) room when presented old and new rooms and also could pick out the original viewpoint when presented old and new viewpoints of the room. When subjects viewed two successive rooms followed by a pair of 2-alt forced-choice tests, accuracy for old/new room judgements remained high (95%; chance=50%), whereas accuracy for old/new viewpoint judgements of a single room dropped to 71% for 30 degree viewpoint changes and 77% for 60 degree changes. In a second experiment in which only viewpoint changes were tested the results were similar. Thus, memory for the viewpoint from which a spatial layout is seen is much less robust than memory for the content of the scene.
The research was supported by NIMH grant MH47432. We thank Winston Chang for research assistance.