1. PERCEIVED BEAUTY OF RANDOM DENSITY PATTERNS
Jay Friedenberg, Emile Umile and Amber Bailey
Department of Psychology. Manhattan College. Riverdale, NY 10471
BACKGROUND
40% Filled Density Pattern
From Experiment 1
Results
There was a significant effect of density on response ratings as measured by a one-way ANOVA F(9, 260) = 65.7, p < 0.1. Ratings increased and peaked at an approximate 70% fill level then tapered downward.
We obtained two derived measures of complexity. One is contour pattern length, calculated as the sum of the edge lengths for all the black cells. The correlation between edge length and ratings was significant, r(48) = .47, p < A second is number of parts. Filled regions were considered to be parts if they were joined by at least one side but not by a corner. The correlation between parts and ratings was not significant.
EXPERIMENT TWO
Discussion
The results of the second experiment confirm that textures with intermediary fill levels are considered most beautiful. We also replicated the finding that contour length and not number of parts, both calculated as before, are psychologically important attributes. The correlation between ratings and edge length was significant, the correlation between ratings and number of parts was not.
One way to increase complexity is to simply increase the number of elements. If observers prefer grids with more elements than ratings should be higher for these patterns.
The relation between complexity and aesthetic judgement has been researched for decades. Early studies using polygon-like forms found peak preference at intermediary levels of complexity based on features like number of sides, angularity and symmetry (Roberts, 2007).
Far less attention has been devoted to the study of texture patterns with aggregate elements although there is some recent work modelling aesthetic texture perception (Thumfart, et. al., 2011). In these two studies we ask: What makes texture patterns complex and how is this complexity related to perceived beauty?
Stimuli and Method
The grids in this experiment now measured 15 X 15 cm square with a total of 225 elements as opposed to 100 in Experiment 1. Element size was maintained at 1 square cm. but overall pattern size increased accordingly.
Fifteen undergraduates participated. The filling process and conditions were otherwise identical to those in the previous study.
CONCLUSIONS
70% Filled Density Pattern
From Experiment 1
In comparing Experiment 1 and 2 we see that observers are not preferring an absolute number of elements. The peak in the first study was at 70 elements while in the second it was at around 113. Their ratings therefore seem invariant to element number. The raters instead seem to be making relative judgments, scaling their responses to the range and type of patterns seen in a session.
The results suggest that two few small elements at low densities and a few large parts at high densities are both too simple, whereas those in the middle are of greater complexity and interest. An analysis of edge length vs. density shows a middle peak, similar to the ratings data suggesting that maximal complexity is what is preferred. This is consonant with results we have obtained in our lab using polygons.
Average beauty ratings for density patterns
in Experiment 1. Error bars are ± one standard
error of the mean.
PURPOSE
Results
The results of the ANOVA for density and ratings was significant, F(9, 140) = 83.2, p < 0.1. The peak of the response function was now lower at a density of 0.5, where half the grid cells were filled.
In two experiments we vary the density of square grid patterns. If observers prefer moderate levels of complexity peak preference should be somewhere near the middle of the range.
Element size and pattern size was held constant in each experiment while the total number of elements was increased. If participants prefer an absolute number of elements the peak should correspond to this number in both studies.
Another estimate of complexity is edge or contour length. We calculated the edge length of these patterns to see if ratings correlated with length.
Average beauty ratings for density patterns
in Experiment 2. Error bars are ± one standard
error of the mean.
Method
Grids were presented at the center of a computer monitor. Twenty-five undergraduate students rated how attractive they were using a rating scale that ran from 1 – 7. They responded by pushing one of the corresponding keys on the computer keypad. Pattern duration was response-terminated. They were allowed as much time as needed to make their judgments and instructed to perceive the black-filled areas as figure and the white portions as ground.
There were 10 different versions of the patterns at each density level, with 100 trials in a block and a total of 400 trials in an experiment session. Order of presentation within the four blocks was randomized.
REFERENCES
EXPERIMENT ONE
Roberts, M. N. (2007). Complexity and aesthetic preference for diverse visual stimuli. Doctoral dissertation. Department of Psychology. Universitat de les Illes Balears, Brazil.
Thufart, S. Jacobs, R., Lughofer, E., Eitzinger, C., Cornelissen, F. W., Groissboeck, W. & Richter, R. (2011). Modeling human aesthetic perception of visual textures. ACM Transactions on Applied Perception. 8(4), 1-29.
Stimuli
10 X 10 cm square grids containing 100 one cm square cells were randomly filled in increments of 10% ranging from 10% - 100%. Filled cells were black. Empty cells were white. Patterns were presented against a white background. No grid lines were present.
Discussion
Participants preferred patterns that were on average 70% filled. They also liked grids with the greatest contour lengths. If edge length is a proxy for complexity then preference was for the most complex shapes and not ones of moderate complexity.