Retrieval Practice: Limiting Learning by ‘Teaching to the Test’ or Generalizing Knowledge by Facilitating Memory Recall

Cognitive Science has been promising as a "disruptive technology" for education. Just as social networking sites have reshaped our interactions, the application of modern scientific understanding of cognitive processes may lead to qualitative improvement in students' learning (e.g., LearningScientists.org). Among the techniques suggested for improving learning is "retrieval practice" (e.g., RetrievalPractice.org). Rather than only recalling from memory on tests, students can be asked to retrieve knowledge from memory as a strategy during learning - prior to assessment. There are now dozens of published peer-reviewed studies of the technique.

Most cognitive research creates tightly controlled 'classroom' analogs with experimental manipulation of the precise material to be learned, retrieved, and assessed (e.g., Uner & Roediger, 2017). In contrast, I conducted action research in upper-level college classes. I compared student performance on test items that were identical across semesters. To test the value of retrieval practice, I created practice tests for students to complete individually without notes, then discuss in small groups, and correct through class-wide discussion.

One concern about retrieval practice is that it may be nothing more than "teaching to the test" so improved test performance does not really show increased learning. Studies published to date defend retrieval practice. For example, in one study students recalled a definition and an example of the concept, and this led to greater ability to identify another example on the assessment (Finn et al., 2017). However, I worry that when the assessment so closely matches the practice, it is not genuine evidence of deeper learning. In my study, I purposely chose to include only half of the class concepts on the practice test to assess if students self-awareness during the practice test would encourage further studying. On the practice test, I made half of the items closed responses (e.g., matching) and half open-ended responses (e.g., short answer) to examine if the kind of retrieval matters. I also made half the items nearly the exact same use of the concept as the actual test item, and the other half required generalization of the concept.

I tested the hypotheses using inferential statistics and propose to showcase results with bar graphs that are understandable to educated laypersons without statistical expertise. Results showed that retrieval practice led students to have higher grades on tests, but the benefit was limited to only those concepts that the practice test covered. The practice test did not encourage better studying. I subsequently examined only items on the test that corresponded with items on the practice tests. Students performed better on the test if retrieval practice nearly exactly resembled the test regardless of it the practice test item was closed-ended or opened-ended. The practice test did not lead students to generalize understanding of concepts.

This study suggests retrieval practice aids students in doing better on tests. For this reason I intend to further examine its use, especially to "level the playing field" for students from under-privileged backgrounds who may have missed the opportunity to learn study strategies prior to college. Retrieval practice may best be combined with additional pedagogical techniques that are necessary to foster the deeper learning hoped for in college.