Extension of the dual-memory model of test-enhanced learning to distributions and individual differences

Abstract

The dual-memory model of test-enhanced learning (Rickard & Pan, 2018, Psychonomic Bulletin & Review, 25[3], 847–869) provides empirically supported quantitative predictions about multiple core phenomena for the case of cued recall. That model has been evaluated to date only for mean proportion correct. However, it also makes predictions about the distribution over subjects for both test-condition proportion correct and testing-effect magnitude. As a consequence, it makes predictions about aggregate individual difference effects on learning through testing. The current paper evaluates those and other predictions, focusing on a data set of 509 subjects aggregated over multiple experiments that were conducted in my laboratory. Results show that the distribution predictions hold to a close approximation for materials ranging from paired associates to history facts, and for retention intervals ranging from 1 to 7 days. The distribution analyses also allow for a novel assessment of whether accuracy on a training test with feedback is a determinant of testing-effect magnitude, and the results suggest constraints on alternative models. Limitations and prospects are discussed.