Optimizing Resistance to Forgetting

To examine the question of how study should be scheduled to promote the retention of skills and knowledge over the long-term, the team completed a total of 8 studies (they had originally planned to complete only 4). The first study examined spacing effects and retention over a one-year period. The team found that spacing effects are seen at a year delay, but that the optimal spacing gap is different for longer time periods. The team also completed a second series of studies looking at the effects on spacing gaps for retention of foreign vocabulary, facts, and names of visual objects, with test delays of up to 6 months. When an optimal gap was used, final recall was improved by up to 150%. These gap effects were nonlinear, and indicated that cumulative reviewers are enormously important for promoting retention over meaningful time periods. The team also examined effects of spacing on acquiring arithmetic skills (multiple-digit multiplication problems) and found that training on a larger set of practice problems reduced initial mastery of automaticity, but led to faster responses and more retrieval in a test session presented at a later point in time. The team also examined different patterns of presenting mathematics practice problems (such as homework sets at the end of a chapter), and found that typical practices seen in textbooks (e.g., massing all problems on a given topic in one practice set, versus distributing problems on a topic across multiple practice sets). In three separate studies completed both with college students and elementary school students, the found that spacing practice improves test performance compared to massing practice, and that interleaving practice problem types as opposed to blocking the practice problems led to substantial improvement in final test score in addition to benefits of the spacing effect. Finally, the team examined the spacing effect on the ability to identify benign or cancerous skin lesions, and did not find the spacing effect. Instead, they found very little forgetting across all conditions, even after a year. The reasons for this are not understood, however, it does suggest that the spacing effect may well depend upon the type of content to be learned (or perhaps, the personal significance of the to-be-learned content).

The second basic empirical question that the team proposed to examine was to come pursue when testing is a useful mechanism to facilitate learning. The team completed 4 studies which addressed this question. First, the team examined if there were any facilitative effects of testing beyond improving the retention of tested items. They found that testing did indeed enhance more than only the retrieved item. They also examined whether testing can enhance the learning of visuospatial information in maps. They found that testing indeed improved map learning as compared to conventional study. Tests of this effect were a continuing focus in their subsequent IES grant.

The third empirical question asked if self-discovery of insights or generalizations is an effective way to retard forgetting. The completed two studies examining this question – which they frame as an examination of the utility of discovery learning. In one study, participants were asked to learn which parts, when combined, made the fasted “alien racer.” Subjects learned by using a computer simulation – discovery subjects constructed racers and raced cars to answer the question; presentation subjects watched already made cars race. In these studies, the subjects in the discovery condition outperformed the subjects in the presentation condition.

The fourth question was to see if they could determine which forms and the timing of feedback were best to retard forgetting. In the one study completed to address this question, the team found that supplying the correct response after an incorrect response was critical to performance, but that feedback after correct responses made little difference. They also found that delayed feedback was also quite beneficial. This suggests that corrective feedback should always be given.

The team also developed a classroom-based intervention (review session) built upon these principles for a U.S. History course, and evaluated them with high-school students. Using a within subjects design, they examined whether facts that were reviewed by testing followed by feedback, or restudied at one of two intervals, or not reviewed at all, were retained differentially. They found that facts reviewed through testing were retained significantly better than facts reviewed through re-study, and nearly twice as well as those given no review. They also found facts reviewed by testing at the 16 week interval were best retained.

Finally, the team also developed a computer-aided instruction component incorporating these and other principles designed to improve students’ retention of textbook information. Preliminary data indicates that students who used the hierarchical testing procedure they developed during study outperformed those who reread the chapter. They are continuing to gather additional data, and are working on designing variations that may improve its effectiveness.