Principles of cognitive science hold the promise of helping children to study more effectively, yet they do not always make successful transitions from the laboratory to applied settings and have rarely been tested in such settings. For example, self-generation of answers to questions should help children to remember. But what if children cannot generate anything? And what if they make an error? Do these deviations from the laboratory norm of perfect generation hurt, and, if so, do they hurt enough that one should, in practice, spurn generation? Can feedback compensate, or are errors catastrophic? The studies reviewed here address three interlocking questions in an effort to better implement a computer-based study program to help children learn: (1) Does generation help? (2) Do errors hurt if they are corrected? And (3) what is the effect of feedback? The answers to these questions are: Yes, generation helps; no, surprisingly, errors that are corrected do not hurt; and, finally, feedback is beneficial in verbal learning. These answers may help put cognitive scientists in a better position to put their well-established principles in the service of children's learning.
In the 40 years since the cognitive revolution took hold in American psychology, many advances have been made in our understanding of how people think and remember, as well as in what people know about what they knowthat is, their metacognition. These advances have come about largely through a plethora of exciting laboratory experiments. As a result, there is a variety of memory phenomena that, in the laboratory, are both highly replicable and well understood. Deep encoding, spaced practice, test-enhanced learning, the generation effect, the encoding specificity principle, the effects of difficult retrieval, the beneficial effects of spacing, judgments of learning, and over- and underconfidence biases are all buzzwords within cognitive psychology that have implications for enhancing learning in a classroom situation. Our increasing understanding of people's metacognitive biases and distortions may be particularly important. But applying these principles in a real-world setting frequently gives rise to unexpected issues that may modulate their effectiveness.
One of the most clearly applicable sets of findings are the biases and illusions in people's metacognition. People frequently believe they have learned something when in fact they have not (Bjork, 1999; Metcalfe, 1998). Given such metacognitive failures, learners may be in a poor position to remedy their own faulty study habits. Furthermore, there are many real-world situations in which the task itself destroys our metacognition. For example, when people study for a foreign language vocabulary test, the fact that the textbook presents the to-be-remembered word side by side with its translation would seem to be innocuous. But it produces exactly the condition, shown by Kelley and Jacoby (1996), that evokes overconfidence. Because people's choice of whether to study or not is directly related to whether they think they know or not (Metcalfe & Finn, hi press), they will decline study because they think, wrongly, that they already know.
One way to circumvent metacognitive illusions is to encourage people to test themselves, which has benefits both because it allows for more accurate metacognition (Dunlosky, Hertzog, Kennedy, & Thiede, 2005) and because the effects of the test are themselves beneficial (Glover, 1989; Hogan & Kintsch, 1971; Roediger & Karpicke, 2006), perhaps because the person generates the answers actively. But without the help of a teacher, a tutor, a parent, or a computer, self-testing can be logistically problematic, if the learner thinks of doing it at all. Educators and psychologists need to devise ways to induce people to do this rather than allow them to automatically fall into a dysfunctional, illusory metacognitive state. On a related note, generating the answer rather than simply reading it or having it presented gives rise to a well-documented beneficial memory effect (Hirshman & Bjork, 1988; Slamecka & Graf, 1978). …