Decision criteria in dual discrimination tasks estimated using external-noise methods

Abstract

According to classical signal detection theory (SDT), in simple detection or discrimination tasks, observers use a decision parameter based on their noisy internal response to set a boundary between "yes" and "no" responses. Experimental paradigms where performance is limited by internal noise cannot be used to provide an unambiguous measure of the decision criterion and its variability. Here, unidimensional external noise is used to estimate a criterion and its variability in stimulus space. Within this paradigm, the criterion is defined as the stimulus value separating the two response alternatives. This paradigm allows the assessment of interactions between criteria assigned to different targets in dual tasks. Previous studies suggested that observers' criteria interacted or even collapsed to one (hence, nonoptimal) criterion. An alternative interpretation of those results is that observers equated their false alarm (FA) rates. The external-noise method enables the confrontation of the two hypotheses. It is shown that the variability of observers' criterion in stimulus space is about 1.6 times their measured sensory threshold, suggesting that the presence of external noise increases decision uncertainty. Observers' stimulus criterion settings are close to SDT predictions in single tasks, but not in dual tasks where the two criteria tend to "attract" each other. Observers maintain distinct FA rates even when SDT predicts equal rates. Observers trained in psychophysics or provided with basic notions of SDT exemplified with the present experimental design manage to better separate their criteria in some conditions.