Stochastic independence in the recognition/identification paradigm

Abstract

Experiments in which subjects study a set of items and then take two successive tests, a recognition test followed by a test of implicit retrieval of the same items (e.g. Tulving, Schacter, & Stark, 1982), have typically yielded data showing stochastic independence between the tests. Ostergaard (1992) has suggested that the procedure for assessing such independence should include an evaluation of the data against a model that assumes maximal dependence between the components of the test attributable to the study episode. In applying his Maximum Memory Dependence model to the published data, however, Ostergaard inexplicably used test-primed data for baseline measures, rather than non-primed data, and erroneously concluded that there was no clear evidence for independence. In this article, we show that when the standard baseline measures are used, the results of many experiments demonstrate stochastic independence.