The composition of abstract images : differences between artists and laypersons

The compositional abilities of artists have been widely discussed. Yet, there is not much experimental research on this topic. This study examines the difference in composition of abstract art between artists and laypersons. During the first experiment, artists and laypersons were asked to create abstract images out of 30 given abstract elements. This experimental design helped excluding semantic meaning and drawing ability as factors. We analyzed the created images for their statistical low-level properties. An interesting find was that artist created images that were significantly less self-similar (Pyramid of Histogram of Oriented Gradients [PHOG] Self-Similarity) and adhered more closely to the so-called Rule-of-Thirds. In a follow-up experiment, laypersons were asked to categorize the images from the first experiment into images created by artist and created by layperson. Strikingly, naive observers were able to distinguish between the groups significantly. Overall, our studied showed that there are differences in the creation of abstract art between laypersons and artists and artists creations can be recognized as a class by naive observers.