Testing drives learning.

Once you’ve tested on something correctly once, you will remember it forever. Right?

In a study on “The Critical Importance of Retrieval for Learning” by Karpicke and Roediger (2008), four different student groups are compared in order to figure out the importance of both repetition and testing for longer-term recall of learned facts.

Students are asked to memorize a list of 40 Swahili-English word pairs, and then tested on those pairs. After the first test, the four groups are then treated differently: The first group continues studying and testing on all word pairs. The second group continues studying all word pairs, but is only tested on those words that were not successfully recalled. As soon as one word pair is successfully recalled, it is dropped from all subsequent tests. The third group is tested on all word pairs in all tests, but word pairs that were successfully recalled in a test are dropped from subsequent studying. And for the last group, every successfully recalled word pair is dropped from all subsequent studying and testing.

The learning gain during the study period is very similar for all four groups, but interestingly the recall a week later is not.

The groups that were always tested on all word pairs, no matter whether the word pairs were studied until the end or dropped after successful retrieval, could recall about 80% of the word pairs one week later. The students in the other two groups, where word pairs were dropped from testing after successful retrieval, only recalled between 30 and 40% of word pairs correctly.

This basically shows that repeated studying does not have an effect once a word pair has been successfully recalled once, but that repeated testing even after a successful recall consolidates the learning. Testing drives learning, indeed.

These findings should probably have substantial implications on the way we teach – and on how we learn ourselves. The authors report that self-testing is rarely reported as a self-studying technique and that practicing retrieval is only ever a side benefit of students testing whether or not they have learned. And the findings are indeed contradicting the widely accepted conventional wisdom that repetition will improve retention of material. So at the very least, we should share the findings of this study with students and educators.

One way to include more testing in large classes are clickers and multiple choice questions, and the benefits of clickers on retention of material are discussed in the Marsh et a. (2007) paper discussed recently.

Another way would be to encourage students to not just repeatedly read a text when studying for an exam, but to ask themselves questions on details of the text to test what they remember and how well they understand it.

Come to think of it, there are really a lot of possibilities for including question-asking in classes. How are you going to do it?

4 thoughts on “Testing drives learning.

  1. This is helpful research to know about. Thanks for summarizing it and bringing it to our attention.

    I would add that “testing” drives learning in precisely the way that the scientific theory of fill-in-the-blank is “just a theory.” In other words, the precise research meanings of certain words like “testing” and “theory” are very different from the well-known common uses of those words. Which allows a lot of confusion for some people.

    So the title of your post, for instance, could be taken to suggest that any and all testing–presumably including the ubiquitous high stakes standardized tests that have come to be virtually synonymous with the word testing in the public sphere–benefits learning. But the kind of “testing” you go on to describe is an entirely different sort, one that helps students learn rather than merely attempts to measure what they’ve learned.

  2. True, very good point! I have to note, though, that I chose the title “testing drives learning” as a reference to work my office mate is currently doing, so that was a bit of an inside joke. Which was a poor choice, seeing that I am probably one out of two people in the whole world who can possibly get that joke ;-)

Leave a Reply