Friday, October 28, 2011

Improving memory - (retrieval) practice makes perfect

Today we would like to present you with the another (terrific) guest blogger. Joseph Williams is a graduate student in the cognitive psychology program at UC-Berkeley. Enjoy!

You’re about to read a 200-word science passage on sea otters so that you can successfully answer questions about it in a week’s time. What strategies would you use to study it? Which of these options would you choose? (a) reading it four times, (b) drawing out a concept map of all the key ideas, or (c) reading it, trying to recall it, reading it one more time, and trying to recall it one more time.

If you smugly chose the alluring quadruple study option or took a gamble on the newfangled concept map, it’s likely that a week from now your memory would be letting you down. A recent paper in Science by Karpicke & Blunt at Purdue University reports an experiment along these lines. Testing oneself or engaging in retrieval practice had the greatest benefit for being able to remember facts from the passage and for drawing inferences that required putting these facts together. But it seems so counterintuitive that testing yourself on information could be better than thoroughly studying material or building elaborate diagrams. Surely students would all be on the honor roll if only they put in that much effort!

This result may be surprising because it contradicts one of our common implicit assumptions about learning: that it primarily involves receiving or acquiring knowledge, and being asked to retrieve that knowledge is a simple matter of reading it out. This has been termed the “bucket model of the mind,” with new facts being added to a mental container. Under this view a test’s only function is assessment – seeing what’s in the bucket. In fact, when people were asked about how effective each study method would be, they gave retrieval practice the lowest evaluation – although it was the best.

The experiment described earlier provides evidence for the limits of that view by demonstrating the role that tests can play in learning. Psychological research on human memory emphasizes processes involved in encoding information, but also the key role of the processes involved in retrieval. Although there isn’t a clear consensus, one reason testing knowledge helps learning may be that it helps people lay down pathways and cues to information that allow them to access it later. Simply reading a text may not do that automatically, but trying to retrieve elements of that text does.

A skeptic might point out that the key issue is that the experiment wasn’t a fair test – isn’t a recall test just more similar to the later questions than the concept map? Why not test learning through drawing a concept map? A second experiment addressed this issue, having people draw a concept map or do study-test-study-test retrieval practice, then draw a concept map a week later. Remarkably, retrieval practice resulted in a more accurate concept map, than having drawn a concept map a week earlier! Practice in retrieving the relevant information made it more accessible and available for later use.

Research on these testing effects is a growing field, with experiments showing the benefits of pre-testing before any information is learned, testing over short and long time periods, and testing in the classroom. However, the generalization that “all testing promotes learning” probably outstrips the current evidence, and many interesting questions about the strengths and limits of testing remain to be answered. What practical contexts will testing be most effective in? For what kinds of knowledge will it work better or worse?

While awaiting these answers, we can profit from the research that has been done so far. Are there times you’ve fallen prey to the “bucket model” in your own learning, or trying to teach others? What are situations where you could benefit from using retrieval practice? A friend mentioned to me that after years of taking notes during academic talks and never remembering them, he abandoned the practice and made a point of not making notes until later that night. While counterintuitive, it aligns with the principle of retrieval practice.

What I find most fascinating about retrieval practice is that it makes me rethink how information is structured in my mind. Knowing something isn’t just a matter of having encoded it and placed it in the bucket. Truly knowing requires storing information so that it will be retrieved and accessed later ­– understanding its purpose, knowing when and where it will be needed, and practicing the act of accessing it in the right situations. That theory of knowledge sounds much more like a brain than a bucket.

The article: Karpicke JD, & Blunt JR (2011). Retrieval practice produces more learning than elaborative studying with concept mapping. Science (New York, N.Y.), 331 (6018), 772-5 PMID: 21252317

Joseph Jay Williams is a doctoral candidate at UC Berkeley in the area of cognitive psychology. He is broadly interested in how people learn and acquire new knowledge, and works on understanding how seeking and generating explanations promotes learning and generalization. He's also interested in how basic research in psychology connects to practical contexts such as education and health behavior change. You can read more about Joseph here:


  1. Great post. As a high school social studies teacher, the debate over testing recalled facts or application/synthesis is a big one. I hope you might dedicate a future post to the "so what?" question. Just because we have learned the best way to have facts stay in our brains, what is the use of those facts to our larger ability to apply what we are learning to our selves and our world. I know how to teach students how to best memorize isolated historical facts (names, dates, etc) but so what?

    To me, this the larger question of cognitive science is how to help students make meaning of the facts they have committed to memory.

  2. Wonderful, just what a blog it is! This blog has provided the helpful data to us continue the good work.
    Piracetam @ peaknootropics