Determining the Psychometric Properties of the Retrieval-Induced Forgetting Procedure

Abstract

Repeatedly retrieving information from memory has been shown to induce forgetting of related, un-retrieved information below baseline, an effect termed retrieval-induced forgetting (RIF; Anderson, Bjork, & Bjork, 1994). In the current research, stability and alternate-forms reliability estimates of RIF scores were evaluated through correlations of five RIF tasks using two sets of equated category-word pairs and one set of facts in sentence format. Convergent and discriminant validity estimates were evaluated through correlation of RIF scores and scores on the Cognitive Failures Questionnaire (CFQ) and the Social Desirability Scale–17 (SDS-17), respectively. Analysis indicated that although RIF was obtained on all four tasks, stability reliability was obtained for only the sets of materials that participants completed twice, with no evidence for alternate forms reliability. Stability reliability for the category-word pair RIF task that participants completed twice, two-weeks apart, accounted for 17.6% of the variance in scores, r (50) = .42, p = .003. The facts RIF task was completed again approximately one month following the initial administration and stability reliability was also obtained using these materials, r (18) = .51, p = .032, accounting for 27% of the variance in scores. Evidence of discriminant validity was found through non-significant correlations between the RIF tasks and scores on the SDS-17, however evidence of convergent validity was not obtained when correlating CFQ and RIF scores. The results suggest that variations in the degree of forgetting observed from RIF tasks may largely depend on the type of materials used. Implications for theory and research regarding RIF are discussed.