Teaching for transfer: A retrieval-based intervention, and a putative tool to gauge learning outcomes

Abstract

The phenomenon of transfer—our ability to perform novel tasks by generalizing from past experiences—has long captivated theorists and practitioners. As educators it is essential for us to understand what types of learning best promote transfer and to structure our curricula accordingly. With that goal in mind, this dissertation outlines two lines of research. For the first line of research I adopted an experimental approach in the domain of problem solving, examining a training technique whereby the learner solves practice problems for the same principle in dissimilar contexts as opposed to highly similar contexts. The key finding was that contextual variability improved transfer outcomes when a set of training problems were solved spaced in time (akin to a closed-book test), but not when prior training problems and their solutions remained visible throughout training (akin to an open-book test). This finding suggests that contextual variability during training can be beneficial because it forces the learner to more effortfully recall what they have learned in the past. For the second line of research I then adopted a correlational approach, investigating a ubiquitous self-report inventory, the Study Process Questionnaire (SPQ), which is meant to quantify student learning approaches to predict educational outcomes. However, the SPQ’s predictive validity has recently been challenged because deep learning and its corresponding outcomes remain poorly defined. To tackle this measurement issue, my colleagues and I operationally defined outcome measures in real university courses to tap more precisely into transfer of learning. Across several studies we found limited evidence for the SPQ’s ability to predict transfer outcomes, leading us to suggest that educators and researchers should be more cautious about using this self-report inventory to characterize student learning.