What the research says

What 435 Studies Say About Feedback

Feedback is one of the most studied ideas in education. A 2020 meta-analysis of 435 studies found a solid average effect, and a fourfold difference between the weakest and strongest kinds.

By the Giga Quizzes team4 min read
An orange megaphone against an orange wall
Photo: Oleg Laptev, Unsplash, Unsplash License, cropped.
In this post
  1. 1The average
  2. 2What kind of feedback
  3. 3What it changes
  4. 4Who gives it
  5. 5The limits
  6. 6Feedback after a quiz

Feedback has a reputation as one of the most powerful things a teacher can do. Much of that reputation rests on John Hattie’s syntheses, which estimated a very large effect. In 2020 Benedikt Wisniewski, Klaus Zierer and Hattie himself went back and did the job more carefully: a fresh meta-analysis of the original studies rather than a summary of earlier summaries. They found 435 studies with 994 effect sizes and more than 61,000 participants. Feedback came out well, but not uniformly well.

The average

Across all the studies, feedback improved learning by d = 0.48, a medium effect. That is lower than the 0.79 that Hattie and Helen Timperley had reported in 2007, and the authors say so plainly. The earlier figure combined other meta-analyses, which can double-count studies and mix very different things under one label.

More important than the average was the spread. The studies differed far more than chance would explain, and 17% of the original effects were negative: feedback that made learning worse. The authors conclude that feedback cannot be treated as a single consistent treatment. Some feedback is powerful. Some is not.

What kind of feedback

The clearest dividing line was how much information the feedback carried.

Type of feedbackWhat it tells the learnerEffect (d)
Reinforcement or punishmentWell done, or that was poor0.24
Corrective feedbackRight or wrong, and the correct answer0.46
High-information feedbackWhat was wrong, why, and how to improve0.99
Wisniewski, Zierer & Hattie (2020)

Praise and blame, the most common feedback in everyday life, had the smallest effect. Telling students whether they were right and what the right answer was did about twice as well. Feedback that also explained the task, the process and how to do better was about four times as effective as praise. In the authors’ words, feedback is more effective the more information it contains.

What it changes

Outcome measuredEffect (d)
Cognitive outcomes: knowledge and understanding0.51
Physical and motor skills0.63
Behaviour0.48, with very wide uncertainty
Motivation0.33
Wisniewski, Zierer & Hattie (2020)

Feedback did most for what students know and can do, and less for how motivated they feel. The motor-skills figure rests on few studies, so it should be read with care. The behaviour figure is so uncertain that it says little either way.

Who gives it

Most of the studies looked at feedback from teacher to student, with an effect close to the overall average, 0.47. A smaller group looked at feedback from students to teachers, such as surveys or quick polls about the lesson, at 0.35. Feedback between students showed the largest effect, 0.85, but it came from only 16 effect sizes, too few to lean on. Whether feedback was written or spoken made no significant difference.

The limits

A meta-analysis can only sort studies by the labels its authors give them, and one study’s “corrective feedback” may look quite different from another’s. Designs differed too: studies that only compared before and after found bigger effects than those with a proper control group, 0.63 against 0.42, and published journal articles found more than unpublished dissertations. And any single figure here is an average over ages, subjects and settings that differ enormously. The honest reading is that well-designed feedback reliably helps, and the size of the help depends on what the feedback actually says.

Feedback after a quiz

A quiz is a natural place for feedback, because every question has an answer and a reason. The meta-analysis suggests aiming for the top row of the table, not the bottom.

Three levels of feedback on a quiz question

  • Low: a cheer for a right answer and a buzzer for a wrong one. Fun, but on its own it teaches little.
  • Medium: right or wrong, and here is the correct answer.
  • High: the correct answer, why it is correct, and what to look for in a similar question next time.

High-information feedback, for one question

Question: Which gas do plants take in for photosynthesis? A student answers oxygen. Low-information feedback says: wrong, the answer is carbon dioxide. High-information feedback adds: plants take in carbon dioxide and give out oxygen, so oxygen is the product, not the ingredient; if you mixed them up, picture the leaf releasing bubbles of oxygen into the water.

The second version takes ten more seconds to write and to read. It names the mistake, explains why it happens and gives the student something to hold on to next time. That is the kind of feedback the strongest studies in the review were testing.

Sounds and celebrations are not the enemy; they keep a game lively. The point is that they work best on top of information, not instead of it. The study on explanation feedback shows the same pattern in a single experiment: an explanation made the difference on new questions, while the bare answer did not.

Feedback also runs the other way. The answers a class gives in a quiz are feedback to the teacher about what was understood, and the post on reading quiz results covers how to use them.

Sources

  1. Wisniewski, B., Zierer, K. & Hattie, J. (2020). The power of feedback revisited: A meta-analysis of educational feedback research. Frontiers in Psychology, 10, 3087.

More from the blog

All posts →