ABSTRACT
Whereas a good sensitivity measure is the difference between the transformed hit and false-alarm rates, a good measure of response bias is the sum of the same two quantities, and is called criterion location, c. In the decision space for the yes-no experiment, c describes the location of a criterion that divides the decision axis between values that lead to “yes” and “no” responses. Other related measures, such as the relative criterion and likelihood-ratio criterion, are equivalent when sensitivity is unvarying, but not when sensitivity changes across conditions. Criterion location has advantages, both logically and, in some cases, empirically. Using a criterion to partition the decision axis is an optimal response strategy. The optimal location of the criterion can be calculated if the performance goal is specified.
