Remembering colors: Bias and variability
56.421, Tuesday, May 14, 2:45 - 6:45 pm, Orchid Ballroom
Jessica McLaren1, Jeremy Bell2, Sarah Allred2; 1Rutgers--The State University of New Jersey. Graduate Program in Psychology, 2Rutgers--The State University of New Jersey. Department of Psychology
INTRODUCTION: Remembering colors to match previously seen objects is a subjectively difficult experience. Memory is often blamed; however, real world memory tasks often require matching colors across changes in visual environment. Thus, such tasks require color constancy as well as memory. Are mistakes in such tasks caused by memory or failures of constancy? TASK: 120 observers matched the color of 16 real, painted cubes to a 1022 chip palette in 8 different conditions: (1) baseline, where matches were made adjacent to the palette; (2) across an illuminant shift; (3) cube embedded in an approximately color-opponent surround; (4) across a ten-minute memory delay (M); (5) (8) each permutation of the above conditions. We compared the bias and variability of color matches in each condition. VARIABILITY: Compared to baseline, memory and illuminant conditions, but not the surround condition, showed significantly more variability. BIAS: Consistent with previous results, the illuminant change elicited a significant bias (imperfect constancy, ~80%) in the direction of the chromaticity of the illuminant. In contrast to flat, coplanar surfaces, embedding cubes in a surround elicited no significant biases (nearly perfect constancy). The empirical data on biases in color memory are contradictory: we found that memory was not biased towards greater saturation, nor towards prototypical hues, although memory matches showed small but significant biases in an apparently unsystematic way. This apparent memory bias (magnitude and variability between cubes) could be largely accounted for by modeling memory as unbiased but more variable than perception and taking into account the gamut and discretization of the matching palette. INTERACTIONS. When illumination, surround and memory were combined, neither variability nor bias were larger than with the illuminant shift alone. CONCLUSION: In a real world color memory task, errors could be largely attributed to failures of constancy, rather than failures of memory.