The test-retest reliability of the retrieval practice effect.

Abstract

Retrieving information from memory, compared with different control conditions, improves subsequent retention of that information. Given the ubiquitous demonstration of this retrieval practice effect, researchers have asked whether such benefit is moderated by individual-difference variables. An implicit assumption in individual-difference research is that the retrieval practice effect at the participant level is reliable across participants. In this study, we tested this reliability assumption. In two sessions, 54 participants studied foreign-native word pairs, repeatedly restudied half of the word pairs, repeatedly retrieval-practised the other half, and, finally, took a final test for all pairs. Different word pairs were used in each session. We replicated the retrieval practice effect at the group level in Sessions 1 (d = 0.54, 95% confidence interval [CI] = [0.27, 0.80]) and 2 (d = 0.79 [0.53, 1.05]). In addition, we found that the retrieval practice effect at the participant level was reliable over a 1-week span both for absolute agreement, intraclass correlation coefficient (ICC) = .33 [.12, .51], and for consistency estimates, ICC = .35 [.14, .53]. The result bridges a gap in the literature of individual differences on the retrieval practice effect. We suggest that future studies identify whether, and under which experimental conditions, the retrieval practice effect at the participant level may show even greater reliability estimates than the ones reported here. The finding may also pave the way for studies assessing whether the magnitude of the retrieval practice effect is domain-general or paradigm-specific.