Abstract
Empirical aesthetics has identified relationships between quantifiable characteristics of visual scenes and the aesthetic preferences observers show for them. We investigated the relationship between aesthetics and scene characteristics using a set of chromatic and spatial image statistics investigated in previous studies (e.g., Mather, 2020), but here defined in a biologically meaningful color space, as well as novel chromatic statistics aiming to capture the naturalness of the color distributions. In Experiment 1, 139 participants rated aesthetic preference of color calibrated natural scene images of objects in everyday contexts. Results showed that the image statistics accounted for 29% of the variance in group-level preference and identified the particular importance of a number of natural scene statistics such as fractal dimension and spectral slope. Experiment 2 investigated whether natural image statistics calculated from images from head-mounted GoPro cameras can predict preference for abstract art from the JenAesthetics data set (Amirshahi et al., 2014) to test the ‘matching-to-nature’ hypothesis (e.g., Nakauchi & Tamura, 2022). We found mixed evidence that artworks with image statistic profiles characteristic of natural scenes are more strongly preferred. Experiment 3 investigated inter-observer variance in the contribution of image statistics to preference. Seventy two participants rated color calibrated images of urban and rural scenes. We explored the relationship between the autism quotient scores of individuals (measuring traits of autism in the general population) and the contribution of image statistics to aesthetics. We found that the image statistics which most strongly predicted visual preference were different among those scoring high in autism traits compared to those scoring low in autism traits. The findings have implications for efficient coding models of vision, natural scene statistics, and individual differences in aesthetics.