Studies in subjective probability V: Chance vs. structure in visual patterns

Student subjects were instructed to produce maximally non-random, aesthetically attractive and random appearing dot patterns of two different levels of complexity (3 and 5 dots). When the patterns were grouped according to similarity, random patterns were found to belong to greater equivalence groups than non-random patterns. They were also estimated to be far more probable to arise from a chance process, regardless of number of elements. The aesthetically attractive patterns were placed between random and non-random patterns in terms of probability estimates. With increasing pattern complexity, patterns produced according to different instructions became increasingly distinguishable. It is argued that there may be an optimal complexity level for discovering "structure" or unique configurations in chance patterns (cf. the constellations of stars).