« Hints Dropped on Direction for U.S. Education Research | Main | Study: Time Changes How Teachers See Students—Literally »

The More You Test, the More You Learn


The best part of the research conference that the federal Institute of Education Sciences holds each year is the poster sessions. If you've never been to one, poster sessions are like science fairs for grown-ups. Scholars post neat, readable summaries of their work on display boards, and visitors stroll up and down the aisles, browsing and asking questions.

In this case, though, the presenters—more than 100 in all—were all researchers whose studies are being financed by the institute. One of the first projects to catch my eye this year was a study that tries to puzzle out how to schedule quizzes for maximum learning power. In 2006, I wrote about an earlier iteration of this project, which found that taking a test can improve students' learning retention. That's probably not a welcome message for the anti-testing crowd, but you can read more about it in this EdWeek article.

This time around, the research team from Washington University in St. Louis is trying to determine how many and what kind of quizzes produce the best results. To find out, the researchers tested eight possible combinations of pre-tests, post-tests, reviews, and no tests at all on students in four 8th grade science classrooms in Illinois.

What they learned was that the more quizzes students took, the more information they retained. The most successful regimen was for teachers to give one test before the teacher's lesson, a second quiz immediately after the lesson or the next day, and a review quiz the day before students were scheduled to take the unit exam. Under that approach, students answered an average of 85 percent of the items correctly. When students were given no tests at all prior to the unit exam, they got the right answers only 64 percent of the time, on average.

The researchers aren't done yet, though. Pooja K. Agarwal, a graduate student working on the project, said the plan is to design an "ideal" instructional combination, which would include tests with open-ended response questions, quick feedback to students on the right answers, and frequent quizzes. The researchers will test that routine in middle schools in the fall.

For more on these results, contact the lead researcher, Henry L. Roediger III, at Washington University.


As someone likely to often be cast in the "anti-testing crowd," I am not bothered by this finding. That is because my organization, FairTest, has always supported high-quality assessment, including formative assessment, of which this would appear to be a variant. (http://www.fairtest.org).

Note this from the earlier article to which Debra links: "Mr. Roediger’s colleagues have also found that the learning benefits are greatest when the tests are composed of short-answer rather than multiple-choice questions. That’s a point on which several of the studies converge: Learning seems to stick when students are forced to generate their own answers to questions or their own definitions of words and concepts."

Most current 'benchmark' etc tests are multiple choice. That is not what is effective (as FairTest has long pointed out, deducing from cognitive science and long educator experience - nice to have this specific research).

The real policy questions, then, are: when will districts and states stop imposing dumbed-down recall tests? Will the next ESEA support high-quality assessment (not merely tougher-to-pass tests) - formative and summative? Will banks of useful tasks and items be prepared (by teachers, mainly) so teachers can use the ones they want at the time they need them (for formative, summative and accountability purposes)? Waht are the range of high-quality assessments that could be useful here - these researchers are focusing on short-answer items? Will that format be as useful for students involved in project-based learning, or will teachers need to adapt the insight that if you ask students to think and demonstrate their work and thinking they will learn more?

Here's my objection:

Some kinds of lessons are easily quizzable and testable. Some are much harder to assess. I have no doubt that this kind of model produces better scores on resulting exams, but I think that it reinforced the primacy of content that is easy to assess over content that is more fundamental or more important. (OK, some fundamental and important content IS easy to test, but not all of it is. And I'd suggest that most of it is not.)

The issue here is that all of the tests and quizzes need to be reviewed, corrected and/or graded. That inevitably (??) leads to items that are easier/faster to review/correct/grade.

If this model leads to a shallower or more simplistic curriculum, it does more harm than good, even if it raises test scores.

Comments are now closed for this post.

Follow This Blog


Most Viewed on Education Week



Recent Comments

  • ceolaf: Here's my objection: Some kinds of lessons are easily quizzable read more
  • Monty Neill: As someone likely to often be cast in the "anti-testing read more