Abstract
Memory is obviously critical for visual preference, but how precisely is yet unclear. Two seemingly inconsistent principles, Novelty and Familiarity, have been proposed. We reported(VSS'07) a surprising segregation of these principles across object categories: familiarity preference monotonously increased in faces(FC) whereas novelty preference increased in natural scenes(NS), over repeated 2AFC preference judgment between a new and an old stimulus. Geometric figures(GF) did not show any bias. While the segregation may indicate intrinsic biological differences among object categories, it may also be due to differences in similarity/variability within category between FC and NS. To examine it, we ran two experiments. First, we reused the same FC and NS stimuli but presented them upside down in the same 2AFC preference task, to see if the critical factor is orientation-specific and holistic. The results faithfully duplicated the pattern of the original results, i.e. Familiarity in FC and novelty in NS. Second, we directly manipulated variability within object category; i.e. to introduce more variability within FC by including cartoons, and natural face-like patterns, etc., and more similarity within NS. As a result, the familiarity preference disappeared in FC. In NS, however, the same increasing bias towards novelty preference persisted. In addition, we ran two sets of GF; one with high similarity within, and the other with low similarity. We found an increasing bias towards familiarity preference in the high similarity set, but no such bias in the low. The overall pattern of results is consistent with a notion that increasing similarity in local features shifts preference from novel to familiar stimuli, in close relation to the averageness hypothesis of attractiveness and the efficient coding theory of perception. The persistence of novelty preference in NC with more similarity is an exception that requires more investigation.
This project is supported by the Japanese Science and Technology Agency.