ABSTRACT
Three experiments explored the effect on judgment of prior context (signal-response events on the preceding trial) in two-choice discrimination without immediate feedback. Auditory frequency was the independent variable in Experiment 1 and visual intensity in Experiment 2. The results of the two experiments showed that following errors, as opposed to correct responses, there was: (1) a significant red uction in discriminability (d), which interacted with modality (frequency larger than intensity); and (2) a significant change in measured bias (β), away from the previous signal and toward the previous response, which was independent of modality. The two effects were significantly correlated only for frequency. A single theoretical account is thus inadequate for the changes in bias and the impairment of intensity judgment. Experiment 3 evaluated organismic state versus modality process contributions to the discriminability reduction and general versus specific accounts of the bias change, by interleaving auditory-frequency and visual-intensity signals on alternate trials. Presenting prior context in a different modality: (1) produced a small but reliable impairment of judgment following errors, which was independent of modality; and (2) eliminated the effect on bias. However, analysis of the same data conditional on two trials back (i.e., in the same modality) showed a reduction of discriminabili-ty that interacted with modality and a change in bias (as in Experiments 1 and 2). These results indicate that the reduction in discriminability following errors was multiply determined: for visual intensity (1) by an organismic state variable, such as attention or arousal; (2) by sequential variance; and in addition for 160auditory frequency (3) by a modality process variable involved in the judgment itself. Concerning bias, the findings argue against models based on generalized response bias and in favor of those that assume that bias is specific to the way in which judgments are made. Two of these models—memory trace and memory state—are used to interpret the frequency-specific component of reduced discriminability following errors.
