Retrieving and transferring knowledge: Effects of test item variation on transfer in an authentic learning environment

Abstract

IntroductionLaboratory research has found superiority of transfer to new questions when prior ones asking about the same underlying problem (or construct) were varied in nature as compared to being constant. The current study examines this matter in a college Introduction to Methods in Psychology course. MethodsStudents (N = 287) with varying aptitudes as measured by college GPA and admissions test scores were taught and then repeatedly tested on a set of constructs (e.g., identifying a particular source of confounding). For half of those constructs, the four ‘mid-term’ exams repeatedly asked the same question (Constant condition). For the other half, each of those exams asked about the constructs via a new question (Varied condition). The final exam included both repeated and novel items testing each construct. We utilized explanatory item response models to examine whether students' performance on the final exam (the probability of a particular student correctly answering a particular item) depends on the type of question (Constant or Varied) previously asked about that construct and on aptitude. ResultsStudents with lower aptitude performed differentially worse on novel final exam items in the Varied condition. Evidence for the superiority of transfer items in the Varied over the Constant case was observed, though not as strongly as predicted.