Wednesday, February 27, 2008

Tests Produce Learning

Tests do more than just measure learning. Tests are learning events. That is, testing forces retrieval of incompletely learned material and that very act of retrieval helps to make the learning more permanent. Testing, and not actual studying, is the key factor on whether or not learning is consolidated into longer term memory.

A recent experiment by J. D.Karpicke and H. L. Roediger at Washington University in St. Louis, examined the role that retrieval had on the ability to recall that same material after a delay of a week. In the experiment, college students were to learn a list of 40 foreign language vocabulary word pairs, which were manipulated so that the pairs either remained in the list (were repeatedly studied) or were dropped from the list once they were recalled. It is like studying flash cards: one way is to keep studying all the cards over and over again; the other way is to drop out a card from the stack every time you correctly recalled what was on the other side of the card. In the experiment, after a fixed period of study time, students were tested over either the entire list or a partial list of only the pairs that had not been dropped. Four study and test periods alternated back-to-back. Students were also asked to predict how many pairs they would be able to remember a week later, and their predictions were compared with actual results on a final test a week later.

The initial learning took about 3-4 trials to master the list, and was not significantly affected by the strategy used (rehearsing the entire list or dropping items out as they were recalled). On average, the students predicted that they would be able to remember about half of the list on a test that was to be given a week later. However, actual recall a week later varied considerably depending on learning conditions. On the final test, students remembered about 80% of the word pairs if they had been tested on all the word pairs, no matter whether they had been studied multiple times with all of them in the list or if they dropped correctly recalled words from the list in later study trials. However, recall was only about 30% correct when correctly identified words were dropped from subsequent tests, even though all words were studied repeatedly. In other words, it was the repeated testing, not the studying, that was the key factor in successful longer-term memory.

So, what is the practical application? When using flash cards, for example, you need to follow each study session (whether or not you drop cards from the stack because you know them), with a formal test over all the cards. Then, repeat the process several times, with study and test epochs back-to-back. Can we extend this principle of frequent testing to other kinds of learning strategies? Probably. But there are no formal experiments.

Let us speculate on the case of trying to remember names of people at a party. You might study the name of each person by using it in conversation or associating the name with some feature of the person's anatomy or personality. Then, silently quiz yourself, looking at the person and asking yourself to recall the person's name. Then, repeat the study-and-test process several times. You would have to keep number of people low (say four to six), because you may not have many opportunities to hear the name repeated other than your own repeating it in conversation. In most practical learning situations, you will not be given repeat tests immediately after each study session, so you must simulate that with self-tests.

Why does forced recall, as during testing, promote consolidation? It probably relates to other recent discoveries showing that each time something is recalled the memory is re-consolidated. If the same information is consolidated again and again, the memory is presumably reinforced.

The failure of students to predict how well they would remember is consistent with my 40 years experience as a professor. Students are frequently surprised to discover after an examination that they did not know the material as well as they thought they did. Tests not only reveal what you know and don't know, they serve to increase how much you eventually learn. If I were still teaching, I would give more tests. And I would encourage students to use self-testing as a routine learning strategy, something that one study revealed to be a seldom-used strategy. The repeated self-tests should include all the study material and not drop out the material that the student thinks is already mastered.

Source: Karpicke, Jeffrey D., and Roedinger, Henry L. III. 2008. The critical importance of retrieval for learning. Science. 319: 966-968.

1 comment:

  1. Great Blog!

    Found it interesting that dropping remembered cards from the deck is just as effective as the full deck until initially mastered. Seems more efficient to drop, since it takes less time for the same result.

    However, the testing phenomenon you mention can be explained by the forgetting curve (discovered over 120 years ago). http://en.wikipedia.org/wiki/Forgetting_curve

    Perhaps you studied it in Psychology 101 and forgot ;-)

    Thanks for a great blog.

    ReplyDelete

Please contribute your ideas. This blog is all about making learning easier for everyone.