Abstract
Memory for visual scenes is a constructive process that is prone to systematic distortions. These distortions can reveal the mechanisms by which the visual environment is encoded in the mind. One striking distortion is “boundary extension”, whereby observers mistakenly recall viewing a scene from farther away than actually observed—and recent work has revealed the existence of a surprising, complementary effect of “boundary contraction”. What memory processes drive these distortions? We hypothesized that these distortions are driven by normalization toward canonical viewing distances, pushing memory outward for close-up scenes and inward for distant scenes. We directly tested this hypothesis by exploiting image manipulations that selectively alter perceived distance while preserving other perceptual and semantic content. First, we created “fake miniatures” of distant scenes using a digital “tilt-shift” effect, which simulates the shallow depth-of-field of close-up images. For example, we made a distant railway scene appear to be a diorama with a toy train. If memory distortions are dependent on how far away a scene appears, then decreasing perceived distance should increase boundary extension. We embedded distant scenes and “close-up” (tilt-shifted) versions of the very same scenes in a boundary judgment paradigm: Participants reported whether a probe image was closer-up or farther-away than a briefly displayed (250ms) and masked target. As predicted, tilt-shift induced powerful increases in boundary extension at the image-level (Experiment 1)—an effect that could not be explained solely by the manipulation’s low-level properties, e.g., the addition of blur gradients or saturation (Experiment 2). A final experiment showed that the perceived-distance effect generalizes to a completely different image manipulation: a spherical distortion which reduces perspectival cues to distance. Taken together, our results reveal that perceived viewing distance plays a causal role in driving memory distortions for scene boundaries, suggesting that scene memories are biased toward canonical views.