On the role of color in the aesthetic appreciation of paintings
Poster Presentation: Saturday, May 17, 2025, 2:45 – 6:45 pm, Pavilion
Session: Color, Light and Materials: Adaptation, constancy and cognition
Schedule of Events | Search Abstracts | Symposia | Talk Sessions | Poster Sessions
Ombretta Strafforello1, Johan Wagemans1; 1KU Leuven, Department of Brain & Cognition
Why do we like some paintings more than others? Empirical aesthetics studies investigated factors like content, style, composition, harmony, order and complexity. The role of color, however, has received much less attention, although it is often discussed in art history and theoretical aesthetics. Here we explored the role of color in aesthetic perception through a data-driven approach. We computed statistics on the pixel distribution and on the dominant color palette of paintings, extracted via Modified Median Cut Quantization. We analyzed the average pixel brightness and saturation and the amount of color variation per painting, calculated as the average distance among the palette colors in RGB space. We used three datasets — LAPIS, JenAesthetics and BAID — comprising 11,790, 1,628 and 60,337 paintings labelled with aesthetic scores. LAPIS and JenAesthetics contain art style and subject matter annotations, enabling analysis across different painting categories. Our results show that top-rated paintings contain significantly more color variations than bottom-rated paintings in all datasets, indicating that the color diversity of a painting contributes to its liking. This partly relates to style: abstract paintings are rated lower than figurative paintings and also contain less color variations, as in the case of color field and minimalist art. We found a weak association (Pearson correlation) between the amount of color variations and liking in LAPIS’ abstract paintings (0.266). Similar associations were found in JenAesthetics’ figurative styles where color is a key feature, like Impressionism (0.308) and Rococo (0.330), in paintings of buildings (0.565), landscapes (0.317) and one-person portraits (0.206). However, no association emerges in paintings where spatial composition or shape dominates, such as portraits with many people and nudes. Finally, the amount of brightness variations correlates more weakly with liking (0.203 in LAPIS, 0.183 in JenAesthetics), suggesting that the effects of color variation cannot be reduced to brightness.