Abstract
Symmetry makes stimuli less complex. Symmetric arrangements are also generally preferred to asymmetric ones. However, it is unclear whether these relationships depend on global mirror symmetry, or local parts symmetry. Here we investigate the roles of both local and global symmetry in subjective complexity and pleasure judgements of natural scenes. We collected subjective ratings of complexity, aesthetic pleasure, and interest for 720 scene images – approximately 50 ratings per image – using Amazon Mechanical Turk. We then calculated average complexity, pleasure, and interest ratings for each image, as well as several measures of local and global symmetry. Global symmetry measures were obtained by creating an axis of symmetry at every column (vertical) and row (horizontal) of the image, and correlating the rows/columns of pixels on one side of the symmetry axis to those on the other side, weighted by the proportion of rows/columns included in the correlation. Local symmetry measures were computed by converting each photograph into a line drawing and calculating the parallelism (ribbon symmetry) and distance (separation) between contours. To investigate the relationship between symmetry and participants’ ratings, we ran a canonical correlation analysis using ten symmetry measures as predictors of the three subjective rating measures. The full model explained 35.1% of the variance shared between features and ratings. The first canonical root revealed that local symmetry measures, and vertical global symmetry, are negatively related to complexity ratings (i.e., symmetry reduces complexity). The second canonical root showed that horizontal global symmetry, and large variations in global symmetry within one image, are positively related to pleasure and interest. Our work suggests that global and local symmetry are distinct in their influence on people’s judgements of pleasure and complexity: complexity is affected mostly by local aspects of the scene, while pleasure is influenced more by the scene as a whole.