Pan and Sana (2021) Pretesting vs. posttesting: Errorful generation, prequestions, and retrieval practice

Abstract

The use of practice tests to enhance learning, or test-enhanced learning, ranks among the most effective of all pedagogical techniques. We investigated the relative efficacy of pretesting (i.e., errorful generation) and posttesting (i.e., retrieval practice), two of the most prominent practice test types in the literature to date. Pretesting involves taking tests before to-be-learned information is studied, whereas posttesting involves taking tests after information is studied. In five experiments (combined n = 1,573), participants studied expository text passages, each paired with a pretest or a posttest. The tests involved multiple-choice (Experiments 1-5) or cued recall format (Experiments 2-4) and were administered with or without correct answer feedback (Experiments 3-4). On a criterial test administered 5 minutes or 48 hours later, both test types enhanced memory relative to a no-test control, but pretesting yielded higher overall scores. That advantage held across test formats, in the presence or absence of feedback, at different retention intervals, and appeared to stem from enhanced processing of text passage content (Experiment 5). Thus, although the benefits of posttesting are more well-established in the literature, pretesting is highly competitive with posttesting and can yield similar, if not greater, pedagogical benefits. These findings have important implications for the incorporation of practice tests in education and training contexts.

Guessing as a learning intervention: A meta-analytic review of the prequestion effect

“…Most experimenters have explored the PE using in-person domains, specifically, by having participants partake in research laboratories. Recently, researchers have started to explore the PE using online domains (e.g., MTurk) as technology allows for a wider range of data collection opportunities (Janelli & Lipnevich, 2021;Pan, Schmitt, et al, 2020;Sana, 2021aSana, , 2021b). However, because few studies have been conducted outside of the research laboratory (k = 1 for classroom domains; k = 4 for online domains), this moderator analysis was not conducted.…”