Effects of stimulus congruence on hypotheses held in a concept identification task

Abstract

This investigation compares overt judgments about tenable hypotheses to choices in a concept identification task, as a function of stimulus similarity on successive trials. The most frequent error in processing information was the failure to eliminate a dimension when both hypotheses associated with it had been excluded by previous feedback, an error which ranked very high in Berger's study as well. Relatively more of these errors occurred with low congruence of successive stimuli than with high congruence, where congruence is defined as the number of hypotheses which would lead to correct responses on both of two successive trials. Relatively fewer premature exclusions of a dimension occurred with low congruence than high congruence. The high incidence of hypothesis processing errors following correct responses contradicts Restle's all-elements at a time model. The frequency with which all hypotheses held on a trial are inconsistent with the choice response on the next trial contradicts hypothesis theory in general. However, a l-element local consistency version of Restle's model gives almost the same predictions as the same model with passive states precluding a direct relation between hypotheses and choices on early trials.