Maximizing Learning from Tests: Examining the Relationship between Practice Test Difficulty and Final Test Difficulty

Tests can be used not only to assess learning, but also to enhance it. After an initial learning phase, practicing to-be-remembered information by taking a practice test often produces better subsequent recall, even if no feedback on performance is given, than not taking a practice test or using the same amount of time to restudy the information. The present research focused on how to design practice tests to maximize their benefits. Specifically, we examine a novel prediction derived from the distribution-based bifurcation model of testing (Halamish & Bjork, 2011, Kornell, Bjork, & Garcia, 2011), namely that in order to maximize learning via testing, the difficulty of the practice test should be matched to the difficulty of the final test. To examine this prediction, participants in the current research studied paired associates and practiced them via testing, before taking a final, criterion test. We orthogonally manipulated practice test difficulty—in terms of the retention interval, and final test difficulty—in terms of test format (cued vs. free recall). Consistent with our prediction, analysis of final test performance yielded an interaction between practice and final test difficulty. Performance on the easy final test was higher following an easy practice test than following a difficult practice test. Performance on the difficult final test, however, was comparable following easy and difficult practice tests. The results point at the conditions under which the benefits of testing can be enhanced, and, more generally, point at the usefulness of distributionbased models in understanding the benefits of testing.