On the fundamental processes of recognition memory

Most of the major trends in recognition-memory data from studies using words and pictures can be predicted with a small number of core processes, suggesting these processes are fundamentally important. The case for this assertion is supported by results and modeling of three long-term recognition memory studies, using a new design, and showing the core processes generalize well between different task types, list lengths and stimulus types. In most conditions memory was tested with two items: both or neither from the list, or one from the list. In Experiment 1, in two-alternative forced choice blocks the more likely old item was to be chosen; in four-way classification blocks the two items were to be classified as both old, both new, or one old and one new. Experiment 2 required judgment of the item more likely new. Experiment 3 added blocks of single item old-new testing. The 288 distinct conditions giving probabilities of correct and error responses were predicted very accurately by the REM model of Shiffrin & Steyvers (1997). REM was applied to every condition with the same decision criterion and with two of its original parameter values. This ability to generalize and account for such a large number of conditions is especially surprising because REM is missing many components known to play an important role in recognition memory. It suggests that the few components it does include are fundamental and important enough to produce a good approximation to most results from recognition memory studies.