Long-term effects of testing on the recall of nontested materials

Abstract

Testing, or memory retrieval, is a powerful way to enhance long-term retention of studied material. Recent studies have shown that testing can also benefit later retention of related but nontested material (a finding known as retrieval-induced facilitation, Chan, McDermott, & Roediger, 2006), but the long-term consequences of this benefit is unknown. In the current experiment three retention intervals—20 minutes, 24 hours, 7 days—were used to assess the effects of testing on subsequent recall of the nontested material. The results indicate that the magnitude of retrieval-induced facilitation, like that of the testing effect (i.e., the memorial benefit of testing on the tested material), increases with delay at the beginning (i.e., between 20 minutes and 24 hours) but asymptotes afterward (i.e., between 24 hours and 7 days). Theoretical and applied implications of this finding are discussed.

Keywords:

• Testing effect
• Education
• delay
• Retrieval-induced facilitation
• Retrieval-induced forgetting

Acknowledgements

The experiment was supported in part by a grant from the James S. McDonnell Foundation (220020041). I thank Moses Langley for his help with participant testing at Iowa State University.

Notes

¹To examine whether output interference affected results of the control questions, they were split into two halves based on presentation order. The recall probabilities for the first and second half questions were .55 and .53 for the 20-min condition, .41 and .43 for the 24-hr condition, and .24 and .24 for the 7-day condition. If output interference was evident, then participants should have performed more poorly on the second half questions than on the first half: No such evidence was found, all ts < 1.

²Based on their initial test results, the 24 top-performing participants in the 20-min condition yielded an initial test recall rate of .62, which was similar to those in the other delay conditions. Final recall probabilities for these participants were .57, .56, and .65 for the control, nontested-related, and tested items, respectively. Clearly, these final recall probabilities matched up well with those reported in the main text of the manuscript, suggesting that subject selection did not contribute to the findings of the final test.

³A few other studies have also reported long-term RIFO (Conroy & Salmon, 2005, 2006; Ford, Keating, & Patel, 2004; Migueles & Garcia-Bajos, 2007; Storm, Bjork, Bjork, & Nestojko, 2006). However, these studies have all used procedures that differed significantly from the typical retrieval practice paradigm. See Chan (2009) for a detailed review of these studies.