The Replicability of the Negative Testing Effect: Differences Across Participant Populations
Although memory retrieval often enhances subsequent memory, Peterson and Mulligan (2013) reported conditions under which retrieval produces poorer subsequent recall—the negative testing effect. The item-specific–relational account proposes that the effect occurs when retrieval disrupts interitem organizational processing relative to the restudy condition. Rawson et al. (2015), in contrast, failed to replicate the negative testing effect despite repeated high-powered attempts. This article examines the discrepant results, ruling out differences in procedures, and concludes that differences in participant population produced the varying outcome. Specifically, participants from the University of North Carolina (UNC) and Kent State University (KSU) completed the same version of the negative-testing paradigm and were assessed on several measures of cognitive ability (working memory capacity, Raven’s progressive matrices, and SAT or ACT score). For the UNC sample, free recall scores and the amount of category clustering (a measure of organizational processing) was greater in the restudy than retrieval condition (i.e., the negative testing effect was found); for the KSU sample, there was no difference on either measure. Furthermore, in the restudy condition, recall and clustering was greater for UNC than KSU students whereas in the retrieval condition, there was no effect of site on either measure. As expected, measures of cognitive ability were greater for the UNC than KSU sample. The results indicate that the negative testing effect is replicable but is subject to limitation related to the participant population. An analysis in terms of the relationship between cognitive ability and memory predicted this pattern of results.