Abstract
Visually guided behavior often requires perception, working memory, and long-term memory. For example, selecting the ripest tomato requires an observer to compare the color of a tomato in view to the memory of recently seen tomatoes and the stored memory of ripe tomato color. In several studies, we manipulated both perceptual demands (changes in illumination or surrounding color) and memory demands while observers judged the color of test objects, and we measured the precision and bias of color representation. As expected, both perceptual uncertainty and temporal delays decreased precision; temporal delays also elicited biases for some colors, but not others. The pattern of results in both bias and precision was broadly consistent with a framework in which ideal observers combine noisy sensory information with prior information to arrive at decisions about color. In this framework, memory can be thought of as adding sensory uncertainty. Biases emerge because this increase in sensory uncertainty amplifies the effect of prior information and task constraints.