Ella Beschta Westfall¹ (), Berenice Benitez-Carlos¹, Frank Durgin¹; ¹Swarthmore College

The comparison of visual number is often measured using mixtures of black and white dots in two separate patches (to control for luminance). Because mixtures typically contain equal numbers of each color, it is unclear whether observers base their comparisons on a single subset, or on the collection as a whole. We investigated this in two experiments. In an initial experiment, we compared performance on blocks of trials where 50/50 mixtures of black and white dots were used, with performance on blocks where the percentage of black dots in each field was randomly assigned in the range from 30% to 70%. Across a total of 30 participants tested at three numerosities (14, 40, and 114) this manipulation did not significantly affect the precision of number discrimination. However, when we split the variable-percentage trials by the median ratio of the percentages of black dots in the two fields (RoP), we found overall evidence of significantly higher Weber fractions for trials where the RoP differed substantially between the two patches (Cohen’s D = 0.66). In a second experiment, we directly manipulated the RoP with 30 participants. Weber fractions were higher when there were large differences in RoP, than when the ratios of black and white dots were similar in the two dot fields, F(1, 29) = 24.2, p < .001, η2 = 0.14, Cohen’s D = 0.73. More importantly, despite that the black and white dots were of equal luminance contrast with the background, re-analysis confirmed that participant responses were strongly biased by the relative numbers of black dots in the patches, but not by the relative number of white dots. This was true across all conditions of the experiment. These observations show that color subsets play a significant role in number comparison tasks even when reliance on subsets impairs performance.

Acknowledgements: Swarthmore College