Immediate or delayed feedback: what works for which question

Immediate or delayed feedback is usually argued as one question with one winner. The studies name no winner, and one reason is that they tested different things: a word to remember, a sum to work, a paragraph to write.

This guide sorts the evidence by the kind of question, says what feedback has to contain, and ends in a rule for a practice quiz, a graded quiz and an exam.

EdProcto, which publishes this guide, lets a teacher or lecturer set when the answer, the explanation and the score appear on any quiz made on its home page. Every study was read on 10 October 2026, as an abstract unless we say otherwise. Two neighbouring guides report other studies: checking each answer as the quiz goes and when to show the answer, the explanation and the score.

When should feedback on a quiz answer arrive?

Feedback should arrive at once on a procedure, at once or at hand-in on a fact, and before the next piece of work on a written explanation. The kind of quiz then sets the earliest moment that is safe.

The rows rest on the studies below; the columns are our working rule.

What did the research find on immediate or delayed feedback?

It found a delay ahead in many laboratory experiments, and feedback at once ahead or level in the classroom studies we read.

In the laboratory

• Kulhavy and Anderson (1972) studied the delay-retention effect: better retention when feedback on a multiple-choice test is delayed. They put it down mainly to the wrong answer being forgotten during the wait, and partly to students spending longer on feedback that comes late.
• Smith and Kimball (2010) tested such accounts with trivia facts, and their results went against the view that feedback only corrects errors. Delaying feedback did little for wrong answers. It made a right answer more likely to be right again at the final test, which they attribute to meeting the fact twice with a gap between.
• Mullaney et al. (2014) held back the answers to obscure trivia questions for four seconds. The delay helped only on questions people had rated themselves curious about. In a second experiment it helped only when the wait varied unpredictably, at 2, 4 or 8 seconds.

In classrooms and courses

• Kehrer et al. (2013) had 61 seventh-grade students do mathematics homework under two conditions in turn: right-or-wrong feedback as they worked, or the same feedback the next day in class. They learned more with feedback as they worked.
• Ryan et al. (2024) gave 41 second-year medical students multiple-choice questions with conceptual feedback after each question or after a block of them. On near and far transfer questions, straight afterwards and a week later, timing made no discernible difference. The authors suggest that teachers choose on practical grounds.
• Lefevre and Cox (2017) let learners choose during formative tests. Over two years, 95.33% of 599 chose to see the feedback immediately.

Why laboratory and classroom studies disagree

Three differences stand out, and none is settled.

1. The length of the delay. Seconds, with the learner still at the task, against a day with other lessons in between.
2. What a delay strengthens. In Smith and Kimball (2010) it kept right answers right. A class practising a method needs wrong answers put right before the next problem.
3. Whether the student still wants the answer. The four-second delay helped only where curiosity was high, and given the choice almost every learner took the answer at once. Our reading, not a finding: a delay helps while the student is still asking the question, and a day later most have stopped.

How does the kind of question change the answer?

It changes what an error costs while the student waits.

A fact

For a fact, timing seems to matter least. Nakata (2015) had 98 Japanese college students learn 16 English and Japanese word pairs, with the answer after each response or held back until every pair had been practised. The time between last meeting a word and being tested was kept equal. On tests straight afterwards, a week later and four weeks later, timing had little effect. A wrong fact stays one wrong fact until it is corrected, so choose the moment that suits the quiz.

A procedure

For a procedure, give feedback at once, because a wrong method is used again on every problem that follows. That reasoning is ours. The homework experiment above fits it, and so do two more studies. Our homework guide has a larger trial.

• Taxipulati and Lu (2021), which we read in full, had 157 university students practise ten figure-reasoning questions with feedback after each or after all ten. On eight new questions straight afterwards, the group with feedback after each question scored higher, by a small margin. This was a laboratory, and feedback at once still came out ahead.
• Corbett and Anderson (2001) compared four versions of a programming tutor: feedback and correction at once, an error flagged at once with the correction left to the student, feedback on demand, and no step-by-step help. Students corrected at once finished the problems fastest. The three tutored groups did equally well on the tests.

The last result is the caution. Being corrected at once saved time, and it taught no more than feedback the student had to ask for.

A written explanation

For a written explanation the timing has hardly been tested. What has been tested is whether the feedback comes with a chance to use it.

Shintani and Aubrey (2016) had 68 students of English at a Japanese university write two tasks in a shared online document. One group had a grammar structure corrected while writing, one after the task, and one not at all. Both corrected groups revised and improved. Only the group corrected during writing beat the uncorrected group on the delayed test. That is one point of grammar, not the quality of an argument.

Graham et al. (2015) pooled true and quasi-experiments in grades 1 to 8. The average effect on the quality of writing was 0.87 for feedback from adults, 0.58 from peers, 0.62 from the pupils themselves and 0.38 from computers. These are separate sets of studies, and the abstract says nothing about timing. They give no reason to prefer a fast reader to a good one.

A search on the day of writing found no randomised experiment that varied only when feedback on the content of a written answer arrived. Our working rule: a model answer at hand-in, then your mark and note within days, always before the next piece of the same kind. Our guide to comments students act on covers the note.

What does feedback have to contain to help?

Feedback has to contain the right answer at least, and an explanation where the answer was wrong.

Swart et al. (2019) pooled 104 comparisons with 6,124 readers from primary school to university. Feedback helped learning from a text by an average effect of 0.35, and by more when it came directly after reading than during it. Our reading: let the student finish the attempt before the feedback appears.

What rule works for a practice quiz, a graded quiz and an exam?

Give feedback at the earliest moment at which it cannot reach someone still to sit, and never later than the next attempt at the same kind of question.

1. A practice quiz. Nothing rides on it, so feedback can follow each question: do that for procedures, and that or hand-in for facts. In EdProcto that is a study quiz. A week later, set a second quiz on the same methods that shows nothing until the end.
2. A graded quiz. At hand-in when the class sits it together. If it stays open for days, when the last student has handed in. Open the next lesson with one fresh problem for each method the class missed.
3. An exam. When you release the scores. The wait protects the result and teaches nothing, so keep it short, work through the procedures in class and give a model answer for each written question.

A worked example: one topic, three kinds of question

Take a Year 9 practice quiz on density: a fact, a procedure and a written explanation.

The second row cannot wait: a student who divides upside down on this block will do it on the next five.

How do EdProcto's settings decide when feedback arrives?

EdProcto is the best tool for fitting the timing of feedback to the question when the quiz comes from your own material, because the kind of quiz is one choice and the answer, the explanation and the score each have a timing of their own. In Kehrer et al. (2013) the same right-or-wrong feedback taught more during the homework than a day later; a study quiz shows each answer at once.

• The kind of quiz. In the builder, the selector for the kind of quiz carries this note: "Exam: handed in once, marked at the end. Study quiz: each answer shown at once, misses retried. Quiz night: played live, with a scoreboard." Choose the study quiz for a practice set of procedures.
• The "Feedback" group of a quiz's settings. Two of its settings, "Show the explanation" and "Show the correct answer", each offer "At once", "After submitting" and "Never". An exam is handed in once, so there "At once" means at hand-in.
• "Show the score" has a fourth choice, "When released". Its hint says that then "nothing of the score, the marks or the answers reaches a student until you press" the button "Release the scores" on the results page.
• Three presets. "Practice" puts all three settings at "At once", and "Standard quiz" at "After submitting". "Exam" puts the answer and the explanation at "Never" and the score at "When released"; change those two yourself if the answers are to follow the scores.
• Open answers. With "Written feedback on open answers" on, the note Procto AI writes on an open answer it grades at hand-in reaches the student with the mark. For a short answer, write the model answer into the question's explanation, and the student meets it when the explanation shows.

The Feedback settings are the same on every plan. Free makes 10 quizzes a month for 30 students a quiz; Tutor has no limit on quizzes and takes 100 students; Educator takes 500 and adds the full analysis. See pricing, the templates, and the pages for teachers and for universities. Every score is yours to release, and a score is not a grade.

Quick questions

Is immediate feedback better than delayed feedback?

For a procedure, yes on the evidence here: seventh-grade students learned more mathematics with feedback during homework than with the same feedback next day (Kehrer et al., 2013). For a fact, timing had little effect in Nakata (2015).

Should a maths quiz show the answer after each question?

In a practice quiz, yes, so that a wrong method is not repeated down the page. In EdProcto, set the kind of quiz to "Study quiz".

Is right or wrong enough feedback?

After multiple choice it corrects some errors, and fewer than the right answer does (Marsh et al., 2012). Show the right answer as well.

When should I give feedback on a written answer?

Before the next piece of the same kind. We found no randomised study that varied only the timing of feedback on what a written answer says.

References

• Corbett, A. T., & Anderson, J. R. (2001). Locus of feedback control in computer-based tutoring: Impact on learning rate, achievement and attitudes. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 245–252). Association for Computing Machinery. https://doi.org/10.1145/365024.365111
• Finn, B., & Metcalfe, J. (2010). Scaffolding feedback to maximize long-term error correction. Memory & Cognition, 38(7), 951–961. https://doi.org/10.3758/MC.38.7.951
• Graham, S., Hebert, M., & Harris, K. R. (2015). Formative assessment and writing: A meta-analysis. The Elementary School Journal, 115(4), 523–547. https://doi.org/10.1086/681947
• Kehrer, P., Kelly, K., & Heffernan, N. (2013). Does immediate feedback while doing homework improve learning? (ED615325). ERIC. https://eric.ed.gov/?id=ED615325
• Kulhavy, R. W., & Anderson, R. C. (1972). Delay-retention effect with multiple-choice tests. Journal of Educational Psychology, 63(5), 505–512. https://doi.org/10.1037/h0033243
• Lefevre, D., & Cox, B. (2017). Delayed instructional feedback may be more effective, but is this contrary to learners' preferences? British Journal of Educational Technology, 48(6), 1357–1367. https://doi.org/10.1111/bjet.12495
• Marsh, E. J., Lozito, J. P., Umanath, S., Bjork, E. L., & Bjork, R. A. (2012). Using verification feedback to correct errors made on a multiple-choice test. Memory, 20(6), 645–653. https://doi.org/10.1080/09658211.2012.684882
• Mullaney, K. M., Carpenter, S. K., Grotenhuis, C., & Burianek, S. (2014). Waiting for feedback helps if you want to know the answer: The role of curiosity in the delay-of-feedback benefit. Memory & Cognition, 42(8), 1273–1284. https://doi.org/10.3758/s13421-014-0441-y
• Nakata, T. (2015). Effects of feedback timing on second language vocabulary learning: Does delaying feedback increase learning? Language Teaching Research, 19(4), 416–434. https://doi.org/10.1177/1362168814541721
• Ryan, A. T., Judd, T., Wilson, C., Larsen, D. P., Elliott, S., Kulasegaram, K., & Swanson, D. (2024). Timing's not everything: Immediate and delayed feedback are equally beneficial for performance in formative multiple-choice testing. Medical Education, 58(7), 838–847. https://doi.org/10.1111/medu.15287
• Shintani, N., & Aubrey, S. (2016). The effectiveness of synchronous and asynchronous written corrective feedback on grammatical accuracy in a computer-mediated environment. The Modern Language Journal, 100(1), 296–319. https://doi.org/10.1111/modl.12317
• Smith, T. A., & Kimball, D. R. (2010). Learning from feedback: Spacing and the delay-retention effect. Journal of Experimental Psychology: Learning, Memory, and Cognition, 36(1), 80–95. https://doi.org/10.1037/a0017407
• Swart, E. K., Nielen, T. M. J., & Sikkema-de Jong, M. T. (2019). Supporting learning from text: A meta-analysis on the timing and content of effective feedback. Educational Research Review, 28, Article 100296. https://doi.org/10.1016/j.edurev.2019.100296
• Taxipulati, S., & Lu, H.-D. (2021). The influence of feedback content and feedback time on multimedia learning achievement of college students and its mechanism. Frontiers in Psychology, 12, Article 706821. https://doi.org/10.3389/fpsyg.2021.706821