Fig. 1: Inter-individual differences in aesthetic evaluations.
a Examples of stimuli representing images used by Germine et al.32 and Sutherland et al.45. The authors originally investigated the etiology of individual preferences for faces (dashed box) but also collected data on abstract images and scenes. Examples of abstract images and images of a scene were obtained from ref. 31. Images of faces were created as an example of the images of faces. Individuals depicted in the images gave informed consent for publication. The exact images used by both authors are not shown (but see “Methods” for details). b Multilevel modelling for twins’ (one per pair only) ratings of abstract images and images of scenes and faces (results for all twins are shown in Supplementary Fig. 3). The variance coefficients quantify the amount of variance in ratings due to the image (hence shared among individuals), individual features (therefore unique to the individual), the exposure of the images (hence due to the repetition), and their interactions. Error bars represent 95% confidence intervals (CI). c The pair-level and individual-level metrics were used to quantify the three aesthetic phenotypes: pairwise agreement, taste-typicality and evaluation-bias. Pairwise agreement reflects the agreement within a MZ (light green), DZ (orange) or unrelated pseudo-randomised (UR; turquoise) pair; Taste-typicality reflects how similar an individual set of ratings is in comparison to the average ratings of everyone else but their twin. Evaluation-bias reflects the overall aesthetic value evoked by a domain of visual images. d Results from the PCA over the individual ratings for abstract images. PC1 and PC2 individual scores are plotted in the bottom panel. Colour represents the taste-typicality (left) and evaluation-bias (right) scores.
