Color Distribution Rarity in Naturalistic Scenes as a Possible Integrative Descriptor of Preference for Abstract Paintings
Ikuto Hanada, Takehiro NagaiABSTRACT
Color preferences may be shaped by the statistical regularities of naturalistic scenes. To examine this possibility, we investigated whether preference for abstract paintings is related to the rarity of the color distributions of the naturalistic images used as donors. As an indirect measure of this rarity, we used a variational‐autoencoder‐based donor‐image rarity index. We specifically tested a restricted matching‐to‐nature hypothesis, which predicts higher preference for more typical color distributions within the range of color distributions observed in naturalistic images. Abstract painting stimuli were generated by mapping color distributions from naturalistic images onto several base images, and preference was measured using Thurstone's pairwise comparison method. Contrary to the prediction of the restricted matching‐to‐nature hypothesis, the results showed a moderate positive correlation between the donor‐image rarity index and preference, indicating that painting stimuli with higher donor‐image rarity indices tended to be preferred. The donor‐image rarity index was also associated with greater variance in lightness, chroma, and hue, and these statistics were likewise positively associated with preference. These findings provide indirect support for the possibility that the rarity of color distributions within the naturalistic‐image range captures variance shared by multiple color‐distribution statistics relevant to abstract painting preference.