The reliability of criterion shifting in recognition memory is task dependent.

We explored 2 research questions: whether criterion shifts within different recognition tasks are reliable across testing sessions and whether such shifts correlate across the different tasks. In Experiment 1, we established consistent group-level criterion shifting based on expected strength of target items in the test. False-alarm rates were higher when expected strength was weak as opposed to strong, even when expected strength cues were unblocked at test. Test-retest reliabilities in this strength-cuing environment were good. In Experiment 2, we manipulated either the probability (high or low) or expected memory strength (strong or weak) of target items and had people perform both tasks on each of 2 days of testing. Varying the probability of target items consistently produced criterion shifts, as did manipulating target memory strength. Regarding individual differences, shifting in a given test context predicted shifting on a second day in that same test context. However, one's tendency to shift a decision criterion on 1 type of test context did not predict one's tendency to do so in the other. The extent to which people shifted their decision criterion in recognition memory testing was largely dependent on the type of test.