Abstract
Most research on texture perception has focused on segmentation and classification. Recent efforts however are now underway to investigate human affective interpretation and construction of predictive computational models (Kim et. al., 2006; Thumfart, et. al., 2012). Much of this work has attempted to explain how lower-order statistics get mapped onto higher-order psychological variables. In the current study we manipulate a very basic texture property, namely density, to examine its effect on aesthetic judgment. Our patterns were generated using a square grid. Individual squares in this grid could either be filled in as black or unfilled and left white. There were ten different density conditions, ranging from a 0.1 occupation probability (10% of the cells were filled randomly) to a 1.0 cell occupation probability (all cells filled). Participants rated each pattern in terms of their perceived attractiveness using a seven-point scale. In a second experiment we manipulated the shape of the density function, varying the drop in fill probability from the center of the grid outwards. In the first experiment there was a statistically significant effect of density F(9, 216) = 65.7, p <0.1. Average attractiveness ratings increased as density increased, peaking at a 0.7 occupation probability. Participants preferred more dense patterns in experiment two t(35) = 4.9, p <0.5. Those with less dispersion, i.e., with a higher slope value for the density function, were ranked as significantly more attractive. The results are interpreted according to an "edge of chaos" model where peak attraction occurs near the border between patterns that are ordered and those that are chaotic. Many complex systems in nature exist in this regime and preference for patterns of this sort may reflect an innate propensity toward naturalistic stimuli.
Meeting abstract presented at VSS 2013