Can multiple-choice testing potentiate new learning for text passages? A meta-cognitive approach to understanding the forward testing effect

A burgeoning area of research has begun to examine how retrieval practice can influence future learning that occurs after a test. In general, the extant literature has demonstrated a forward testing effect, in which prior testing enhances new learning. However, there is not a consensus as to the mechanism that leads to this phenomenon. In the present dissertation, I propose a metacognitive account, in which testing is purported to benefit subsequent learning by leading learners to better attend to and encode material. In particular, individuals who are tested gain valuable information about the nature and difficulty of upcoming tests, which helps guide their strategy use. Under this account, more difficult retrieval (e.g., recall) should lead to a greater metacognitive benefit than easier retrieval (e.g., multiple-choice). In Experiment 1, I compared prior cued-recall, prior multiple-choice, or no prior testing on performance for a criterial test of a text passage. Furthermore, I examined how the match between initial and criterial tests might determine whether testing influences new learning. In fact, prior cued-recall testing enhanced learning to a greater degree than prior multiple-choice testing (relative to no prior testing) regardless of criterial test format. Reading times for each text passage provided preliminary evidence for a metacognitive benefit of testing. Whereas reading times fell across the passages for individuals who were not tested, reading times remained stable in both testing conditions. In Experiments 2 and 3, I aimed to further investigate the metacognitive mechanism underlying the forward testing effect by requiring explicit judgments of learning (JOLs) prior to each testing (or non-testing) episode. Surprisingly, when JOLs were required, there was no forward testing effect observed in Experiment 2 (which included prior cued-recall, multiple-choice, or no-testing) or in Experiment 3 (which included a more difficult prior multiple-choice condition). In both Experiments 2 and 3, reading times remained stable across passages in each condition, although JOLs indicated less confidence in predicted performance in the tested conditions than in those who were not tested prior to the criterial test