Kunlin Jin¹ (), Diane M. Beck^1,2; ¹Department of Psychology, University of Illinois Urbana-Champaign, ²Beckman Institute, University of Illinois Urbana-Champaign

Observers form aesthetic impressions within a fraction of a second, and processing fluency accounts propose that these immediate evaluations depend on how easily an image is processed (Reber et al., 2004). Two factors known to shape fluency are real world statistical regularity, which includes features such as scene prototypicality and canonical object viewpoints (Beck et al., 2024) and image complexity (Forsythe et al., 2011). Although both factors have been linked to aesthetic preference, their relative contributions during rapid judgments remain unclear. We examined whether statistical regularity predicts attractiveness ratings more strongly than two complementary measures of visual complexity: fractal dimension and edge density. Participants rated the attractiveness of scenes (N = 33) and objects (N = 28) on a 1–5 scale. Scene images were from four categories and were either highly representative of their category (statistically regular) or poor representatives (statistically irregular). Object images were either canonical views of the object (statistically regular) or non-canonical views (statistically irregular). Initial paired t tests showed that statistically regular images received higher attractiveness rating than irregular images for both scenes, t(32) = 9.44, p < .001, d = 0.64, and objects, t(27) = 7.42, p < .001, d = 1.24. We then modeled all images with a linear mixed effects approach including statistical regularity and the visual complexity factors as fixed effects. Statistical regularity emerged as the strongest predictor of attractiveness, with good exemplars rated higher than bad exemplars. Beyond this dominant effect, images that were more fractally complex received higher ratings, whereas images with greater edge density received lower ratings. These findings demonstrate that rapid aesthetic evaluations of our real-world scenes and objects were shaped primarily by statistical regularity, with image complexity exerting additional but weaker influences.