Guide · Learning science

What is the testing effect?

The testing effect is the finding that retrieving information from memory, as you do on a test, improves later retention of it more than spending the same time studying it again. Retrieval practice is the name for using the effect on purpose, by quizzing yourself as a way to learn.

Experiments on it go back more than a century, and hundreds of studies have found it. It also has limits that popular summaries tend to skip.

Who discovered the testing effect?

No single person did. Edwina Abbott published an early monograph in 1909 on the part recall plays in learning, based on work in the psychology laboratory at the University of Illinois. Later reviewers note that her sample was very small.

The first large study was Arthur Gates's Recitation as a Factor in Memorizing, published in 1917. He had children from first to eighth grade learn nonsense syllables or short biographies taken from Who's Who in America. Each group spent a different share of the learning period, from none to 90 percent, looking away from the page and reciting to themselves. Recitation helped in almost every condition. Two exceptions matter. First graders showed no benefit, and with the biographies the benefit leveled off once recitation took more than about 60 percent of the time.

In 1939 Herbert Spitzer tested 3,605 sixth graders, the entire sixth grade of 91 schools in nine Iowa cities. They read a 600-word article and took 25-question multiple-choice tests on different schedules over 63 days. Scores on a first test fell the longer that test was delayed, tracing a classic forgetting curve. Students tested soon after reading kept far more on later tests, and a first test delayed to day 21 came too late to help.

Then interest faded. A 2006 review by Henry Roediger and Jeffrey Karpicke notes that research on testing and learning seemed to disappear around 1940.

What did Roediger and Karpicke find in 2006?

The revival is usually dated to a 2006 paper in Psychological Science by the same two authors, then at Washington University in St. Louis. Their material was two prose passages of 256 and 275 words, on the sun and on sea otters. Scores are the share of each passage's 30 idea units that students wrote down on a final recall test.

Experiment 1 had 120 students. After reading a passage for seven minutes, they either read it for seven more minutes or spent seven minutes writing down all they could recall. Experiment 2 had 180 students and three conditions. Students had four study periods, or three study periods and one recall test, or one study period and three recall tests. Nobody received feedback in either experiment.

ConditionTested after 5 minutesTested after 2 daysTested after 1 week
Experiment 1, study then study81%54%42%
Experiment 1, study then test75%68%56%
Experiment 2, four study periods83%Not tested40%
Experiment 2, three study periods and one test78%Not tested56%
Experiment 2, one study period and three tests71%Not tested61%

What the results show

Three details are often left out. Restudying won on the immediate test in both experiments, so this is an effect on delayed retention. The group with four study periods read the passage about 14 times and the group with three tests read it about 3 times. And the repeated-study group was the most confident that it would remember the passage a week later.

The authors also measured forgetting. Over the week, the repeated-study group lost 52 percent of what it could recall after five minutes. The repeated-test group lost 14 percent.

Why does retrieval practice work?

Nobody has a settled answer. These are the main proposals, and each is a theory.

  • Retrieval effort. A harder successful retrieval strengthens a memory more than an easy one. This fits Robert Bjork's idea of desirable difficulties, conditions that slow learning now and improve retention later.
  • Elaborative retrieval. Searching memory activates related information, which leaves more routes back to the answer. Shana Carpenter tested this account in 2009.
  • Episodic context. Karpicke, Melissa Lehman and William Aue propose that each retrieval updates a memory with the context in which you recalled it, which guides later memory search. A 2014 experiment from the same lab favored this account over elaboration.
  • Bifurcation. Nate Kornell, Robert Bjork and Michael Garcia argued in 2011 that a test gives recalled items a large boost and unrecalled items none, while restudy gives every item a small boost. That would explain why restudy wins on an immediate test and testing wins later.

Which theory has the most support?

A 2014 meta-analysis by Christopher Rowland in Psychological Bulletin found support for effortful processing and only limited support for the elaboration account. He treated the bifurcation model as a useful framework for the pattern of results. The question remains open.

Limits and boundary conditions

  • Timing of the final test. Minutes after study, restudying does as well or better. The advantage of testing appears after days.
  • Failed retrieval. A test you mostly fail teaches little unless feedback follows. Spitzer's late-tested groups show the cost of waiting too long.
  • Test format. Rowland's meta-analysis found larger benefits from recall tests than from recognition tests.
  • Age and preparation. Gates's first graders did not benefit, and his data suggest that some study has to come before recitation helps.
  • Complex material. Tamara van Gog and John Sweller argued in 2015 that the effect shrinks or disappears as material becomes more complex. Karpicke and Aue replied in the same journal that the null results came from the retrieval tasks used and that the effect holds for complex material. The dispute is not settled.
  • Size in real classrooms. A 2021 meta-analysis by Chunliang Yang and colleagues pooled 222 classroom studies with 48,478 students and found a medium effect, g = 0.499, that varied with feedback, test format and the number of tests.

How to use the testing effect

Close the source and produce the answer before you check it. Then do it again after a gap. Our guide to active recall covers the practical methods, and spaced repetition covers the timing.

Two rules follow from the limits above. Test yourself soon enough that you can still retrieve most of the material, and always check your answers. A low-stakes way to feel the effect is to take a quiz on a subject you read about last month and see what is still there.

Where Generalist fits

Generalist: Micro Learning uses retrieval in two places. Each lesson checks understanding with questions, and questions from lessons you completed return later in short review sessions. The app covers 244 topics in 16 subjects, including mind and psychology.

Common questions

What is the testing effect in psychology?

The testing effect is the finding that taking a memory test on material improves long-term retention of it more than restudying for the same amount of time. It shows up most clearly on tests given days or weeks later, and it has been studied since the early 1900s.

What is retrieval practice?

Retrieval practice is the study strategy built on the testing effect. Instead of rereading, you recall the material from memory, using flashcards, practice questions or a blank page, and then check your answer. The act of retrieving is what strengthens the memory.

Is the testing effect the same as active recall?

They describe the same thing from different angles. The testing effect is the research finding. Retrieval practice is what researchers call the strategy. Active recall is the name students and study guides use for that strategy.

Does the testing effect work without feedback?

Yes, if you can retrieve most of the material. Roediger and Karpicke's 2006 students received no feedback and still remembered more a week later than students who restudied. Feedback matters most for items you fail to retrieve, because a failed attempt alone does not correct the gap.

Does the testing effect help with cramming?

Not on a test minutes away. In the 2006 experiments, students who restudied did slightly better on a test five minutes after studying. The benefit of testing appeared at two days and held at one week, so it pays off for material you need to keep.

Sources