Education

[Paragraph 1] For most of the twentieth century, textbook exercises in mathematics were organized in blocks: a lesson introduced a procedure, and the problem set that followed asked students to apply that same procedure a dozen times in a row. The arrangement has obvious appeal. Students who work through a block of similar problems make visibly fewer errors as they proceed, and both they and their teachers read that improvement as evidence that learning is taking place.

[Paragraph 2] Researchers who study memory have long suspected that this evidence is misleading. When every problem in a set calls for the same procedure, students never have to decide which procedure a problem requires; the heading at the top of the page has already decided for them. What improves during blocked practice, on this view, is the execution of a strategy, not the ability to choose one—and choosing is precisely what an unlabeled examination demands.

[Paragraph 3] Interleaved practice, in which problems requiring different procedures are shuffled together, forces that choice on every item. At Harwell Teachers' College, instructors gave two groups of first-year statistics students identical problems over six weeks, arranged for one group in blocks and for the other in an interleaved sequence. During the practice weeks, the interleaved group scored lower on homework and rated the material as harder. On a cumulative test given a month after the course ended, however, the pattern reversed: the interleaved group outperformed the blocked group, and its advantage was largest on problems whose surface features resembled those of one procedure but whose structure called for another.

[Paragraph 4] The finding does not mean that blocked practice has no place. Students meeting a procedure for the first time may need several consecutive examples simply to grasp what the procedure is. The difficulty lies elsewhere. Because interleaving depresses performance in the short term, learners and instructors who judge a method by how its practice sessions go are likely to abandon the approach that serves them better—a tendency that may persist even among those who are familiar with the research.