Judgmental Bias in Perception as a Function of Learned Relationships between Attributes

Abstract

An hypothesis was tested which is relevant to judgmental bias in the perception of valued objects as well as to other multi-attribute objects: when one attribute is more easily discriminable than another attribute, and the magnitudes of both attributes have previously covaried in an object series, then the judgments of a magnitude of a less discriminable attribute are biased in the direction of a more easily discriminable attribute of an object. When size was the more easily discriminable attribute, the bias was found in two of the three brightness judgments. The hypothesis was not supported when brightness was the more easily discriminable attribute. The discrepancy between predictions and results is accounted for by (1) the inadequate learning of covariation of size and brightness because of the low correlation of the values, and (2) the fact that only size was closely related to a numerical scale, a relationship that may be a determinant of the amount of bias of one attribute scale upon the judgments of another attribute of an object series. A formulation by Wiener accurately predicts that bias is greatest in the judgments of one attribute when it occurs with extreme magnitudes of another attribute. The paradigm of this experiment, with certain modifications, can be extended to account for a number of other psychophysical and social-psychological judgments, such as coin-size estimation effects and halo effects.