Multiple-choice testing: Controlled and automatic influences of retrieval practice in an educational context

Abstract

Previous studies have shown that taking an initial multiple-choice (MC) test produced both automatic influences (i.e., those that operate quickly, without effort, and requiring few attentional resources) and controlled influences (influences that are slower, are applied more deliberately, sometimes oppose automatic processes, and require more attentional resources) on performance in a subsequent test. In this study, we examined the involvement of automatic and controlled processes on performance with MC questions that are related to earlier practice questions, but which have different correct answers. In Experiment 1, which was conducted online with MTurk, automatic influences tended to dominate responding despite using educational materials Scholastic Aptitude Test (SAT questions). Including repeated items in the final test (Experiments 1 and 4) and increasing the time lags between questions (Experiments 2 and 4) increased the automatic influence. However, in a genuine educational environment (university classroom), controlled influences tended to dominate responding instead, similar to what has been observed with cued recall (CR) final tests, but only when there are no repeated items. These controlled influences were enhanced by presenting the related questions back-to-back in the testing sequence (Experiment 2) but were unaffected by feedback on the initial test (Experiment 3). We conclude that performance on both MC and CR tests are affected by both automatic and controlled influences of retrieval practice, but that one type of influence will override the other depending on the presence of repeated items, the specific testing format, and examinees’ investment in scoring well.