Normative representation of objects and scenes: Evidence from predictable biases in visual perception and memory
Konkle & Oliva
Objects and scenes can be viewed at almost any distance, and so can subtend any visual angle. However, we propose that for a given visual stimulus there exists an input size that maximizes visual information needed for a given task. We term this size the “norm”, and present initial evidence suggesting that (1) perception is implicitly sensitive to the norm, and (2) memory for visual information is biased toward this norm. First, observers were presented with a picture of an isolated object (experiment 1a) or a scene (experiment 1b) and were told to adjust the visual angle of the object or the viewing distance of the scene until the image was at the "right place" to see the object or scene. Despite the subjectivity of the task, the selected size for each stimulus was remarkably consistent across observers. Second, the norms from experiment 1 were used in a boundary adjustment task. In the classical boundary extension paradigm, subjects tend to remember pictures of scenes as farther away than they were originally viewed. However, if memory is biased toward a norm point, then a view of an object should show both boundary extension and boundary compression, depending on where it is relative to the norm. Observers were presented with pictures of objects (experiment 2a) or scenes (experiment 2b) taken from different distances and displayed for 5 seconds each. Afterwards, participants adjusted the size of the object or scene so as to match the size viewed at learning. The remembered objects and scenes showed a systematic bias towards the normative size. Finally, using a change detection paradigm, we show that a change towards the norm is harder to detect then a change away from the norm, providing converging evidence that memory is biased toward a perceptual norm.
Konkle, T., & Oliva, A. (2007). Normative representation of objects and scenes: Evidence from predictable biases in visual perception and memory. Journal of Vision. 7(9): 1049; doi:10.1167/7.9.1049