Skip to content

Study techniques and their evidence

Feedback and learning from mistakes

How learning works · Published · Evidence: Strong · About this guide

Do mistakes help you learn? Yes, when the correct answer follows. In experiments, trying to answer, failing and then seeing the answer led to better memory than only reading it, while wrong answers left uncorrected could stick [1][2][3]. Learning from mistakes therefore depends on checking the answers. A mistake you were sure about is often the easiest to fix, though it can return once the correction is forgotten [4][5].

A quiz does two jobs. Trying to recall an answer strengthens the memory of it. Checking the answer afterwards (feedback) tells you which answers to keep and which to replace. This article is about the second job: what feedback adds, when mistakes are useful, and why feedback does not help every time.

What does checking your answers add to a quiz?

Testing yourself can help you remember even if nobody tells you whether you were right. But a large review that pooled 159 comparisons from 61 laboratory studies found a bigger benefit in studies where people saw the material again after the test: there, testing beat restudying (going over the same material again) by nearly twice as much as in studies without feedback [6]. In studies without feedback where people got half or fewer of the answers right the first time, the test gave no reliable advantage over restudying. When a test is hard, feedback appears to be what makes it pay off. These are comparisons between different studies, though, which also differed in other ways, and seeing the material again is itself extra study.

Feedback matters most where you went wrong. In one experiment, 258 people learned pairs of Luganda and English words and were tested several times. Showing the correct answer after a wrong response greatly increased what they remembered a week later. After a right answer, feedback made little difference in that study [7]. But in experiments with multiple-choice general-knowledge questions, for answers people got right but were unsure of, seeing the correct answer roughly doubled how many they still knew later [8]. The two studies used different materials and do not fully agree. Even so, a lucky guess is not yet knowledge, so a right answer you were unsure of is worth checking too.

Can a multiple-choice quiz teach you something wrong?

Multiple-choice questions show you wrong answers on purpose, and that has a cost. In a study by Roediger and Marsh, students took a multiple-choice test and then a later test where they had to come up with their own answers. On that later test they sometimes gave a wrong option they had seen earlier, even though they were warned not to guess. The more wrong options a question had, the more often this happened [2]. Researchers sometimes call this the negative suggestion effect. Taking the test still helped overall, but part of what people "learned" was false.

Feedback largely removes this cost. In a follow-up study, 72 students took a multiple-choice test on short history passages, two thirds of them after reading the passages. Each student met three conditions, on different passages: no feedback, feedback after each question, and feedback at the end of the test. The feedback showed the question, the answer they had chosen and the correct answer. A week later they took a test that made them produce answers themselves, answering every question even if they had to guess [9]:

  • Correct answers: about 31% with no feedback, 45% with feedback after each question and 56% with feedback at the end of the test. For questions they had not been tested on, it was 14%.
  • Wrong options repeated from the multiple-choice test: about 24% of answers with no feedback, against 14–15% with feedback. That is about the same as for questions the students had not been tested on (16%).
Bar chart: feedback raised correct answers from 31% to 45–56% and cut repeated wrong options from 24% to 14–15%.

Feedback after a multiple-choice test: correct answers and repeated wrong options a week later

Measured data from Butler and Roediger (2008): a week after a multiple-choice test on short history passages, 72 undergraduates answered every question of a test where they had to produce the answers themselves. Feedback raised correct answers from 31% to 45% (after each question) and 56% (at the end of the test), and cut the answers that repeated a wrong option from 24% to 15% and 14%, about the level of questions never tested (16%). Each student met every condition, on different passages; group averages from one study, not a prediction for any one learner.

Data: Butler & Roediger (2008), Memory & Cognition 36(3), Tables 2 and 3. [9]

Data table
Feedback after a multiple-choice test: correct answers and repeated wrong options a week later
ConditionCorrect answersAnswers repeating a wrong option
Not tested14%16%
No feedback31%24%
After each question45%15%
At the end of the test56%14%
  • Every question had to be answered, guessing if needed (forced report). The values are averages over the number of options and over how much of the passages the students had read beforehand.
  • The feedback showed the question, the answer chosen and the correct answer. "At the end of the test" meant minutes later, not days.
  • When the students could then leave out answers they did not want to keep, wrong options were repeated in 14% of answers without feedback and in 8% and 7% with it (questions not tested: 5%).

When the students could then leave out the answers they did not want to keep, feedback still roughly halved the repeated wrong options, though not quite down to the level of the untested questions [9]. Feedback also made people's confidence more accurate. Without it, they were more confident than their scores justified [9].

Do mistakes help you learn?

Many people study as if every error were a setback. Experiments suggest that trying and failing need not be wasted effort, as long as the right answer follows. Kornell, Hays and Bjork asked people questions they could not answer: made-up trivia ("What peace treaty ended the Calumet War?") or a request to guess a word's loosely related partner. When people tried to answer first and were then shown the answer, they remembered it better later than when they simply read the question and answer together [1].

The benefit has limits. With word pairs, it has appeared when the cue and the answer are related in meaning. When they were unrelated, guessing first did not help [10]. Other experiments, with unfamiliar words and foreign vocabulary, did find a benefit, although the guesses were unlikely to have anything to do with the answers [11]. These results do not yet add up to a simple rule about which material gains. A review of this research on errorful learning concludes that making errors and then getting corrective feedback tends to help learning. It names the corrective feedback, including working out the reasoning that led to the mistake, as crucial [3].

Is immediate or delayed feedback better?

In the multiple-choice study above, feedback at the end of the test led to more correct answers a week later than feedback after each question: about 56% against 45% [9]. The end of the test was only minutes later, though, so this one study does not show that waiting longer is better.

How soon the answer has to come seems to depend on the material. With related word pairs, a failed guess helped when the answer was shown straight away, but mostly not when it came later [12]. One later study did find a benefit with a delayed answer, when the picture that had been on screen during the guess appeared again with the answer [13]. With meaningful trivia questions, a failed guess still helped when the answer came up to a day later [14]. The article on guessing before you learn looks at trying questions before a lesson.

The hypercorrection effect: are confident mistakes easier to fix?

You might expect a mistake you were sure about to be the hardest to shake. Butterfield and Metcalfe found the opposite. People answered general-knowledge questions, rated how confident they were in each answer and were then shown the correct one. On a retest, the errors made with high confidence were the most likely to have been corrected [4]. This is known as the hypercorrection effect, and the review cited above reports the same pattern [3].

The correction can fade, though. In a later study, the effect was still there when the retest came a week later, but people corrected fewer of their errors overall. When they had forgotten the correction, a high-confidence error was more likely to come back than a low-confidence one [5]. The study did not test any way to prevent this. A reasonable step, though an untested one, is to check a confident mistake again on a later day rather than treating one correction as the end of it.

Can feedback make things worse?

Yes, sometimes. A large analysis pooled 607 results from feedback studies, many of them outside the classroom [15][16]. On average feedback improved performance, but in over a third of cases performance got worse. Feedback tended to work less well the more it drew attention away from the task and towards the person [15].

Hattie and Timperley's review of feedback in education offers a framework for thinking about this [16]. In their model, useful feedback answers three questions: Where am I going? How am I going? Where to next? Feedback can work at four levels:

  • the task: was this answer right?
  • the process: how did I go about it?
  • self-regulation: how am I checking my own work?
  • the self: "you're so clever."

They argue that praise of the person is the least effective, because it carries little information about what to do next. They also note that corrective feedback does more when it fixes a misunderstanding. If the knowledge is missing altogether, more teaching helps more than feedback.

How can you use this when you study?

  • Look at every answer, not only the score. A score tells you how you did. The answers tell you what to change.
  • For each miss, find out why. Say or write the correct answer and what led you to the wrong one [3].
  • Check the right answers you guessed, too. Seeing the correct answer helped people keep correct answers they were unsure of [8].
  • Note how sure you were before you check. This makes confident mistakes easy to spot. The studies used confidence ratings to measure the hypercorrection effect; they were not designed to show that rating your confidence improves memory by itself.
  • Come back to your misses on a later day, especially the ones you were sure about. A corrected confident mistake can return once the correction is forgotten [5]; the study did not test whether a second look prevents that.
  • Look for the next step, not a verdict on yourself. Feedback aimed at the task tends to help more than feedback aimed at the person [15][16].

What the research does not say

  • It does not say that mistakes help on their own. Every benefit of errors described here came with the correct answer shown afterwards [1][3]. Without feedback, wrong answers can stick [2][9].
  • It does not say that guessing first helps with any material. With unrelated word pairs, a wrong guess did not help [10].
  • It does not say that feedback helps in every case. In one large analysis it made performance worse in over a third of cases [15].
  • It does not settle when feedback should come. In one study, feedback at the end of the test beat feedback after each question, but that meant minutes later, not days [9]. For failed guesses, the timing results differ with the material and the setup [12][13][14].
  • It does not tell you how much you personally will gain. Most of the experiments described here were run in laboratories with university students, using trivia, word pairs, vocabulary or short passages. The review of learning from errors states its conclusion for neurologically typical students [3]. Hattie and Timperley's model is a way to think about feedback, not a measured effect.
  • It does not say that a corrected mistake is gone for good. A confident error that was corrected can come back once the correction is forgotten [5].

Sources

Every source the guide cites

  1. [1] Kornell, N., Hays, M. J., & Bjork, R. A. (2009). Unsuccessful retrieval attempts enhance subsequent learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 35(4), 989–998. experiment · doi.org/10.1037/a0015729
  2. [2] Roediger, H. L., III, & Marsh, E. J. (2005). The positive and negative consequences of multiple-choice testing. Journal of Experimental Psychology: Learning, Memory, and Cognition, 31(5), 1155–1159. experiment · doi.org/10.1037/0278-7393.31.5.1155
  3. [3] Metcalfe, J. (2017). Learning from errors. Annual Review of Psychology, 68, 465–489. review · doi.org/10.1146/annurev-psych-010416-044022
  4. [4] Butterfield, B., & Metcalfe, J. (2001). Errors committed with high confidence are hypercorrected. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27(6), 1491–1494. experiment · doi.org/10.1037/0278-7393.27.6.1491
  5. [5] Butler, A. C., Fazio, L. K., & Marsh, E. J. (2011). The hypercorrection effect persists over a week, but high-confidence errors return. Psychonomic Bulletin & Review, 18(6), 1238–1244. experiment · doi.org/10.3758/s13423-011-0173-y
  6. [6] Rowland, C. A. (2014). The effect of testing versus restudy on retention: A meta-analytic review of the testing effect. Psychological Bulletin, 140(6), 1432–1463. meta-analysis · doi.org/10.1037/a0037559
  7. [7] Pashler, H., Cepeda, N. J., Wixted, J. T., & Rohrer, D. (2005). When does feedback facilitate learning of words? Journal of Experimental Psychology: Learning, Memory, and Cognition, 31(1), 3–8. experiment · doi.org/10.1037/0278-7393.31.1.3
  8. [8] Butler, A. C., Karpicke, J. D., & Roediger, H. L., III (2008). Correcting a metacognitive error: Feedback increases retention of low-confidence correct responses. Journal of Experimental Psychology: Learning, Memory, and Cognition, 34(4), 918–928. experiment · doi.org/10.1037/0278-7393.34.4.918
  9. [9] Butler, A. C., & Roediger, H. L., III (2008). Feedback enhances the positive effects and reduces the negative effects of multiple-choice testing. Memory & Cognition, 36(3), 604–616. experiment · doi.org/10.3758/MC.36.3.604
  10. [10] Huelser, B. J., & Metcalfe, J. (2012). Making related errors facilitates learning, but learners do not know it. Memory & Cognition, 40(4), 514–527. experiment · doi.org/10.3758/s13421-011-0167-z
  11. [11] Potts, R., & Shanks, D. R. (2014). The benefit of generating errors during learning. Journal of Experimental Psychology: General, 143(2), 644–667. experiment · doi.org/10.1037/a0033194
  12. [12] Hays, M. J., Kornell, N., & Bjork, R. A. (2013). When and why a failed test potentiates the effectiveness of subsequent study. Journal of Experimental Psychology: Learning, Memory, and Cognition, 39(1), 290–296. experiment · doi.org/10.1037/a0028468
  13. [13] Zawadzka, K., Zaborowska, O., Butowska, E., Piątkowski, K., & Hanczakowski, M. (2023). Guessing can benefit memory for related word pairs even when feedback is delayed. Memory & Cognition, 51(5), 1235–1248. experiment · doi.org/10.3758/s13421-022-01385-0
  14. [14] Kornell, N. (2014). Attempting to answer a meaningful question enhances subsequent learning even when feedback is delayed. Journal of Experimental Psychology: Learning, Memory, and Cognition, 40(1), 106–114. experiment · doi.org/10.1037/a0033699
  15. [15] Kluger, A. N., & DeNisi, A. (1996). The effects of feedback interventions on performance: A historical review, a meta-analysis, and a preliminary feedback intervention theory. Psychological Bulletin, 119(2), 254–284. meta-analysis · doi.org/10.1037/0033-2909.119.2.254
  16. [16] Hattie, J., & Timperley, H. (2007). The power of feedback. Review of Educational Research, 77(1), 81–112. review · doi.org/10.3102/003465430298487

This guide is general information about learning research for students, not medical advice.