By Clay Shumate
Peer tutoring is a structured arrangement in which one student teaches another specific content, with assigned roles, a script and a set schedule. Across 145 studies it averages six months of additional progress, and secondary scores better than primary. The best-designed secondary trial found nothing for the students being tutored — while their tutors gained.
Key Takeaways
- The headline average is good and the best secondary trial is not. The Education Endowment Foundation rates peer tutoring at +6 months from 145 studies on high-security evidence, with secondary at +7 months. Its own randomised secondary trial of 2,736 pupils found −1 month and no effect on reading.
- The tutor is the one who reliably learns. In a randomised secondary study, the older tutors gained 0.24 of a standard deviation while the younger students they tutored lost 0.07, and the difference between the two groups was statistically significant.
- At secondary, pair within the grade. The only review restricted to grades 7–12 found same-age pairs at d = 0.61 against cross-age at d = 0.21, a significant gap — and its authors flag that this contradicts earlier reviews.
- Most of the famous numbers are elementary. The effect size usually quoted as “the” peer tutoring figure comes from a meta-analysis of elementary students, and the largest modern cross-age synthesis is kindergarten and elementary, where mathematics came in at −0.02 and writing at 0.01.
- It is cheap, which changes the calculation. Two randomised trials costed the approach at £10.50 and £8.25 per pupil. Adult one-to-one tutoring in the same country ran £167 to £180 per pupil for fifteen hours.
What is peer tutoring?
Peer tutoring is an arrangement where students work in pairs or small groups to give each other explicit teaching support. The defining feature is the asymmetry of role, not the closeness of the desks: one student is teaching, one is learning, and both know which they are doing right now.

One clarification before anything else, because the phrase is used two ways. This article is about instructional peer tutoring inside a class or a timetabled block. It is not about the volunteer tutoring club that students join for service hours. Those programmes can be worthwhile, but nothing in the research below was measured on them.
The terminology is settled and worth using precisely. Same-age tutoring pairs students from the same grade. Cross-age tutoring puts an older student with a younger one. Reciprocal tutoring has the pair swap roles; fixed-role tutoring does not. Classwide peer tutoring runs the whole room in pairs at once, usually with points and short turns in each role.
It is also not the same thing as cooperative learning and its five elements, which is peers working together on a shared task rather than one instructing the other. A secondary review of peer-mediated instruction draws that line cleanly: tutoring builds a tutoring relationship, cooperative work builds a cooperative one. The EEF reviews them separately too — collaborative learning sits at +5 months on low-security evidence from 212 studies, against peer tutoring’s +6 months on high-security evidence from 145.
| Approach | What it is | EEF impact | Evidence strength | Cost |
|---|---|---|---|---|
| Peer tutoring | One student explicitly teaches another, with roles and structure | +6 months (145 studies) | High | Very low |
| Collaborative learning | Students work together on a shared task | +5 months (212 studies) | Low | Not rated here |
| One-to-one tuition | A teacher, assistant or other adult works with one pupil | +5 months (123 studies) | Moderate | Moderate |
Does peer tutoring actually work in grades 6–12?
The averages say yes and the best secondary trial says no, and both are published by the same organisation. That tension is the honest answer. Any article that gives you the +6 months without the null trial, or the null trial without the +6 months, is not reporting the evidence.
Start with the average. The EEF Toolkit puts peer tutoring at six months of additional progress over a year, rates the evidence high from 145 studies, rates the cost very low, and reports secondary at +7 months against primary at +5. Lower attainers gain more than higher attainers. That is a strong entry by Toolkit standards.
Then read the trial. EEF’s Paired Reading evaluation put older secondary tutors with younger ones for reading across 10 secondary schools, randomising 2,736 pupils in 120 classes at a cost of £10.50 per pupil. The younger group came out at −0.02, confidence interval −0.15 to 0.11. The older group came out at −0.06, confidence interval −0.14 to 0.02. Both round to minus one month. There was no evidence of impact on overall reading ability, sentence completion or passage comprehension, and none for pupils eligible for free school meals.
A companion trial in elementary mathematics landed in the same place: 82 schools, 6,472 pupils, effects of 0.01 and 0.02, at £8.25 per pupil for the year. EEF published both, called them robust given the design and the number of schools, and said the findings should not be dismissed. Its own candidate explanations — all explicitly unconfirmed — were that cross-age tutoring may work less well in these schools or age groups, that implementation may have been too thin, and that business as usual may already contain enough peer work to shrink what a new scheme adds.
The genuinely secondary evidence supports something narrower than “it works.” A meta-analysis of 26 single-case studies with 938 participants found no significant difference between elementary and secondary results, with the secondary subset at a Tau-U of 0.74 across 10 studies and 368 participants. A review of 42 secondary mathematics studies found a mean of d = 0.38, with 5 negative.

One caution about effect sizes in this literature, from inside it. In that secondary mathematics review, studies using a pretest, a posttest and a control group averaged d = 0.31, while studies with weaker designs averaged d = 0.60. The authors recommend control-group designs precisely because the alternative overestimates. When you see a big number for a classroom practice, ask what it was compared against.
Who gains more, the tutor or the student being tutored?
The tutor, on the clearest secondary evidence available. In a randomised study across three secondary schools, the older students acting as tutors gained 0.24 of a standard deviation in reading comprehension while the younger students they tutored came in at −0.07, and the gap between the two groups was significant.
That study ran 295 students in 12 classes, paired reading, twelve weeks at thirty minutes a week, measured on an age-standardised reading comprehension test. Its authors are straight about the limits: classes rather than individuals were randomised, the sample is small, and the paper reports its own overall effect inconsistently. The tutor-versus-tutee contrast is the part that holds.
It does not stand alone. An updated meta-analysis of tutors’ own achievement across 16 articles reports a weighted mean of 0.43, and lists tutors drawn from secondary school among the conditions associated with stronger results. The classic 1982 synthesis of 65 tutoring evaluations reported that tutors gained better understanding of the subject and more positive attitudes toward it, similar to the students they tutored.
While we are here, kill the other claim. That same 1982 synthesis found tutoring had little or no effect on the self-esteem of either tutors or tutees, and the elementary mathematics trial found no quantitative evidence of impact on attitudes toward school, reading or maths. Teachers in that trial perceived confidence gains, and the evaluators flagged that perception as needing study rather than reporting it as a result. If you are selling peer tutoring to a department on confidence, you are selling something nobody has measured.
There is a reason the tutor result is the one I find persuasive, and it is not just the statistics. Explaining something to a person who is entitled to be confused is a harder intellectual task than answering a question on a worksheet. It is also the version of responsibility this site keeps arguing for: giving a teenager a real role with real consequences teaches the teenager. The evidence says that part works even in the trials where the tutee gains did not show up.
Should you pair students same-age or across grades?
At secondary, pair within the grade. The one review restricted to grades 7–12 found same-age pairs at d = 0.61 and cross-age pairs at d = 0.21, a significant difference, and its authors suggest same-age pairs benefit from trust that already exists between classmates.
Those authors also do the right thing and flag that their finding conflicts with earlier reviews, which generally treated the two configurations as equally effective. One older British paper in the set argues the opposite outright and favours cross-age. So this is a lean, not a law.
What makes the lean worth following is where the cross-age evidence comes from. The strongest cross-age results are elementary: a district-wide randomised trial across 129 elementary schools reported effects of about 0.2, and a modern meta-analysis of cross-age tutoring in kindergarten and elementary settings found composite reading at g = 0.18 from 16 cohorts and 8,251 students — with mathematics at −0.02, writing at 0.01 and fluency at 0.11, none of them significant. The one large cross-age trial run in secondary schools is the null above.
If you do go across grades, the Toolkit guidance is specific: keep the age gap under about three years, do not match pairs by attainment, and pitch the work so it stretches the younger student while staying comfortable for the older one.
Should students swap roles or keep fixed tutor and tutee roles?
Either is defensible, and nobody has shown a difference at secondary. The grades 7–12 review put fixed roles at d = 0.33 and reciprocal roles at d = 0.50, but at p = 0.30 that is not a finding. A second mathematics meta-analysis also reported roles as a non-significant moderator.
So choose on other grounds. Reciprocal roles mean everybody tutors, which removes the labelling problem that fixed roles create and keeps the tutor benefit available to the whole class rather than to the stronger half. Fixed roles are simpler to run and easier to script when the content is genuinely beyond the weaker student.
The one thing not to do is let the configuration encode a verdict on who is good at school. A room where the same students are permanently the tutors is a room where you have published a ranking. If fixed roles are right for the content, rotate the content so the roles move — the way the jigsaw method makes every student the expert on something.
How long should a peer tutoring block run, and how often?
Shorter than you would guess, and the sources disagree. EEF recommends sessions four to five times a week in blocks of four to ten weeks. The secondary mathematics review found the opposite: programmes of eight weeks or less beat longer ones, and so did sessions of thirty minutes or less.
That review also found weekly frequency trending downward — one session a week averaged d = 0.48, five averaged d = 0.13 — though not significantly, so it is a hint rather than a result. The single-case meta-analysis found dosage did not moderate outcomes at all, above and below its median of 480 total minutes.
What the actual trials delivered is the most useful guide. The null secondary trial ran thirty-minute periods with twenty minutes of reading over sixteen weeks. The elementary maths null ran twenty minutes a week in sixteen-week blocks. The positive secondary tutor result came from thirty minutes a week for twelve weeks. On that evidence a short, tight block is the sensible bet, and a programme that runs all year because nobody stopped it is the least supported option on the board.
One more piece of Toolkit guidance that matters more than the timetable: peer tutoring works best for reviewing and consolidating learning, not for introducing new material. Use it the week after you taught something, not the week you teach it.
How much training do student tutors need?
Enough that they have a script and know what to do when their partner is stuck. EEF calls training for staff and tutors essential. One meta-analysis lists shorter, less frequent training among the conditions tied to stronger results, so the sources conflict on dose while agreeing it cannot be zero.
The training detail from the null trial is the part worth staring at. Teachers got about a day and a half. Pupil training varied badly across the schools: some used only the first lesson, some trained only the older pupils, and one school did no advance pupil training at all. That is not a trial of a well-implemented programme, which is exactly why EEF listed implementation among its explanations rather than declaring the approach dead.
What the training should contain is better specified. EEF points at questioning frames, modelling, handling misconceptions, giving feedback and evaluating progress. An older review adds the blunt reason it matters: untrained tutors may resort to threats and put-downs, and younger tutors need more structure and closer supervision than older ones. If a department or a district is going to adopt this, that training is the budget line — not the materials, and certainly not the pairing spreadsheet.
From a student’s side, the training is also where the dignity is. Being handed a partner and told to help them is an instruction to improvise in public; being handed three questions to ask and a sentence to say when neither of you knows is a job you can actually do. Script the hard moment, not the easy one.
What if the tutor teaches it wrong?
Nobody has measured it, and that gap should be stated rather than papered over. None of the studies here tracked tutor accuracy or error spread as an outcome. What the literature offers instead is a structural answer: keep peer tutoring on review material, which is the Toolkit’s own advice.
Two related findings are the closest thing to a warning. In the elementary maths trial, teachers reported that pupils with additional language needs, special educational needs or lower prior attainment struggled with the programme and that the teachers felt ill-equipped to support them — the evaluators concluded it may need tailoring before being tested again. And the secondary mathematics review found a larger effect for students without learning disabilities, 0.62 against 0.23, which cuts against the Toolkit’s own expectation that lower attainers and pupils with special educational needs gain most. Two credible sources, opposite directions. Say so and plan for both.
Practically, the mitigations are cheap. Tutor on content you have already taught and marked. Give the tutor the answer key for the set they are working through. Circulate and listen to explanations rather than checking that pairs are on task, because an on-task pair confidently teaching a wrong method looks exactly like success from the front of the room.
How do you set up a peer tutoring block?
Six steps, and the unusual one is the last. Keep it to review content, keep the block short, pair inside the grade, script the hard moment, and tell students plainly what the arrangement is and why they are in the seat they are in.
- Pick review content, not new content. A unit you have already taught and assessed. The tutor needs to have met the material before, not be meeting it a page ahead of their partner.
- Set the block before you start it. Four to eight weeks, one or two sessions a week, thirty minutes or less. Write the end date on the board so it ends on purpose rather than drifting.
- Pair inside the grade, and not by rank order. Same-age pairing has the better secondary evidence. Avoid the arrangement where the pairing publishes a league table of who is good at school.
- Give both students the same one-page script. Three questions the tutor asks, the sentence to say when both are stuck, and the rule that the tutor explains rather than does it for them.
- Train the hard moment and hand over the key. Spend one session on what to do when the partner is wrong or silent, and give the tutor the answers for the set they are working through.
- Tell students the evidence, including the awkward part. Say that the research is clearest on the tutor learning from teaching, that being tutored is not a verdict on anyone, and that the roles will move. Teenagers can handle the real reason; they cannot handle being managed.

One addition to step four that comes from thinking about the seat rather than the plan: let the student being tutored pick the set they want help with. Being assigned a tutor is easy to read as a public verdict on your ability, and the fastest way to stop that is to put the tutee in charge of the agenda. They choose which three problems or which paragraph; the tutor runs the script on those. Same structure, completely different meaning from the chair.
Families will ask, and they are entitled to a straight answer: no, their child is not being taught by another student instead of by you. Peer tutoring sits on content you have already taught, for a fixed block, with the teacher in the room and the answer key in the tutor’s hand. The honest version of the pitch is that the child will also spend half the block doing the explaining, because that is the part the research is clearest about.
Two notes on the cost of running it. The time goes into training and materials rather than into the sessions themselves, which is why EEF rates the cost very low; in the null trial, 20 of 52 classes found the time inside existing lessons rather than adding a slot. And if you want a record of what each student contributed without turning the block into a grading project, having students rate their own contribution to group work is less work than scoring pairs yourself.
Is peer tutoring worth it in grades 6–12? Clay’s take
Yes, with the claim narrowed to what the evidence supports. Run it so the tutor role rotates through everybody, because the tutor gain is the finding that keeps replicating at secondary. Treat gains for the student being tutored as a hope rather than a promise.
That is a smaller claim than the Toolkit’s +6 months, and I think it is the defensible one. The strongest secondary results in this whole literature are the ones about the person doing the explaining, and the biggest, best-designed secondary trial failed to find anything for the person being explained to. If the practice were expensive, that pattern would be a reason to drop it. At eight to ten pounds a pupil, against sixteen times that for adult tutoring, it is a reason to be honest about which outcome you are buying.
What also stands out is how often the structure is doing the work rather than the pairing. Every version of this that produced something had a script, a role, a short block and somebody checking. The versions that produced nothing had pupil training that varied from one lesson to none at all. “Help each other” is not an intervention, and a 90-second pair routine run properly will beat a term of unstructured helping.
What should you do next?
Pick one unit you have already taught and run a four-week block on it. Same-grade pairs, one thirty-minute session a week, a one-page script, and the tutor role swapping halfway through so every student does the explaining at least once.
Measure the thing the evidence says you will get. Give a short quiz on the review content before and after, and look at the scores of the students who did the explaining. That is the outcome with the best secondary support behind it, and it is the one most likely to be worth your four weeks.
Frequently Asked Questions
What is peer tutoring?
Peer tutoring is a structured arrangement in which students work in pairs or small groups to give each other explicit teaching support, with one student teaching and one learning at any moment. It differs from group work in that the roles are asymmetric and assigned. Same-age pairing uses students in the same grade, cross-age pairing puts an older student with a younger one.
Is peer tutoring effective for high school students?
The averages are good and the best secondary trial is not. The Education Endowment Foundation rates peer tutoring at six months of additional progress from 145 studies, with secondary ahead of primary. Its own randomised trial of 2,736 secondary pupils found no effect on reading. A meta-analysis of single-case studies found secondary results no weaker than elementary ones.
Do tutors learn more than the students they tutor?
At secondary, that is what the clearest evidence shows. In a randomised study of 295 students, the older tutors gained 0.24 of a standard deviation in reading comprehension while the younger students they tutored came in at minus 0.07, and the difference was significant. A meta-analysis of tutors’ own achievement across 16 articles reports a weighted mean of 0.43.
Should you pair students same-age or across grades?
At secondary, pair within the grade. The only review restricted to grades 7 to 12 found same-age pairs at d = 0.61 against cross-age at d = 0.21, a significant gap, though the authors note this conflicts with earlier reviews. The strongest cross-age evidence comes from elementary schools, and the one large secondary cross-age trial found nothing.
How long should a peer tutoring programme run?
Keep it short. The Toolkit suggests blocks of four to ten weeks with frequent sessions, while the secondary mathematics review found programmes of eight weeks or less outperformed longer ones and sessions of thirty minutes or less outperformed longer ones. The one positive secondary tutor result came from thirty minutes a week across twelve weeks.
How much training do student tutors need?
More than none, and the sources disagree about how much more. EEF calls training for staff and tutors essential and points at questioning frames, modelling, handling misconceptions and giving feedback. In the null secondary trial, pupil training ranged from a single lesson to none at all in one school, which is part of why the evaluators flagged implementation.
Is peer tutoring the same as tutoring?
No. Adult one-to-one tuition is a separate approach with its own evidence base, rated by EEF at five months from 123 studies with moderate cost. A review of 96 randomised evaluations of adult tutoring found pooled effects of 0.37 standard deviations, with teacher-delivered tutoring at 0.50 and volunteer-delivered at 0.21. Peer tutoring costs a fraction as much.
Does peer tutoring improve confidence or self-esteem?
There is no good evidence that it does. The classic synthesis of 65 tutoring evaluations found little or no effect on the self-esteem of tutors or tutees, and a randomised elementary mathematics trial found no quantitative evidence of impact on attitudes toward school, reading or maths. Teachers in that trial perceived confidence gains, which the evaluators flagged as needing further study.
Sources
- Education Endowment Foundation. Peer tutoring. Teaching and Learning Toolkit. Toolkit entry. Source of the +6 months average, the high evidence rating, the 145 studies, the very low cost rating, the secondary +7 against primary +5 split, the pairing and age-gap guidance, the training guidance, the review-not-new-material point, and EEF’s own warning that evaluations by organisations connected to an approach tend to report larger impacts.
- NatCen Social Research for the Education Endowment Foundation. Paired Reading: Evaluation report and executive summary. Evaluation report (PDF). Source of the secondary null: 10 schools, 2,736 pupils randomised in 120 classes, −0.02 for the younger group (CI −0.15 to 0.11) and −0.06 for the older (CI −0.14 to 0.02), minus one month each, no effect for pupils eligible for free school meals, £10.50 per pupil, about a day and a half of teacher training, and the variation in pupil training between schools.
- Education Endowment Foundation. Durham Shared Maths Project: Evaluation report and executive summary. Evaluation report (PDF). Source of the elementary mathematics null: 82 schools, 6,472 pupils, effects of 0.01 and 0.02, £8.25 per pupil per year, no quantitative evidence of impact on attitudes, and the finding that teachers felt ill-equipped to support pupils with additional language needs, special educational needs or lower prior attainment.
- Coleman, R., & Henderson, P. Peer tutoring — interpreting new evidence. Education Endowment Foundation. Blog post. Source of EEF’s assessment that both trials appear robust and should not be dismissed, and of its three unconfirmed explanations for the null results.
- Thurston, A., Cockerill, M., & Chiang, T.-H. Assessing the Differential Effects of Peer Tutoring for Tutors and Tutees. Education Sciences, 11(3). Full article. Source of the tutor gain of 0.24 against the tutee −0.07, the significant difference between the two groups, and the design: 295 students in 12 classes across three secondary schools, twelve weeks at thirty minutes a week. The paper reports its overall effect inconsistently, so only the group figures are quoted here.
- Alegre, F., Moliner, L., Maroto, A., & Lorenzo-Valentin, G. Peer tutoring and mathematics in secondary education: literature review, effect sizes, moderators, and implications for practice. Heliyon, 5. Full article (PDF). Source of the grades 7–12 figures: 42 studies, mean d = 0.38 with 5 negative, same-age 0.61 against cross-age 0.21, fixed 0.33 against reciprocal 0.50 at p = 0.30, the eight-week and thirty-minute findings, the weekly-frequency trend, the 0.62 against 0.23 learning-disability split, and the control-group design caveat of 0.31 against 0.60.
- Bowman-Perrott, L., Davis, H., Vannest, K., Williams, L., Greenwood, C., & Parker, R. Academic Benefits of Peer Tutoring: A Meta-Analytic Review of Single-Case Research. School Psychology Review, 42(1). Full article (PDF). Source of the overall Tau-U of 0.75 across 26 single-case studies and 938 participants in grades 1–12, the secondary subset of 10 studies and 368 participants at 0.74 against elementary 0.69 with no significant difference, and the finding that dosage did not moderate outcomes. Tau-U is a non-overlap index and is not comparable to a Cohen’s d.
- Leung, K. C. C. An updated meta-analysis on the effect of peer tutoring on tutors’ achievement. School Psychology International, 40(2). Abstract record. Source of the weighted mean of 0.43 across 16 articles for tutors’ achievement, and of the listed conditions including tutors from secondary school, structured tutoring, same-age non-reciprocal pairing, and shorter and less frequent tutor training. No confidence interval is reported, and only the abstract was readable.
- Shenderovich, Y., Thurston, A., & Miller, S. Cross-age tutoring in kindergarten and elementary school settings: a systematic review and meta-analysis. Accepted manuscript (PDF). Source of the cross-age figures — composite reading g = 0.18 from 16 cohorts and 8,251 students, mathematics −0.02, writing 0.01, fluency 0.11 — and of the Cohen, Kulik and Kulik effect sizes quoted through this paper rather than from the 1982 original.
- Cohen, P. A., Kulik, J. A., & Kulik, C.-L. C. Educational Outcomes of Tutoring: A Meta-Analysis of Findings. American Educational Research Journal, 19(2). Bibliographic record with abstract. Source of the findings that tutors gained better subject understanding and more positive attitudes similar to tutees, and that tutoring had little or no effect on the self-esteem of either group, across 65 evaluations. The record carries no effect sizes; those are quoted above through Shenderovich and colleagues.
- Tymms, P., Merrell, C., Thurston, A., Andor, J., Topping, K., & Miller, D. Improving attainment across a whole district: school reform through peer tutoring in a randomized controlled trial. School Effectiveness and School Improvement, 22(3). Repository record with abstract. Source of the district-wide elementary trial across 129 schools reporting cross-age effect sizes of about 0.2.
- Mahoney, M. W. M. Peer-mediated instruction and intervention to support the academic achievement of secondary students with autism spectrum disorder: a systematic review of the literature. The Journal of Special Education Apprenticeship, 12(1). Full article (PDF). Source of the distinction between a tutoring relationship and a cooperative one, the description of classwide peer tutoring with short turns in each role, and the observation that most of the available studies sit in elementary schools rather than secondary ones.
- Gaustad, J. Peer and Cross-Age Tutoring. ERIC Digest No. 79, ERIC Clearinghouse on Educational Management. ED354608. Digest (PDF). Source of the same-age and cross-age terminology, the warning that untrained tutors may resort to threats and put-downs, and the point that younger tutors need more structure and closer supervision.
- Education Endowment Foundation. Collaborative learning approaches and One to one tuition. Collaborative learning entry. Source of the +5 months from 212 studies on low-security evidence for collaborative learning, the +5 months from 123 studies on moderate evidence for adult one-to-one tuition, and the £167 to £180 per pupil figure for fifteen-hour tutoring blocks under the National Tutoring Programme.
- Nickow, A. J., Oreopoulos, P., & Quan, V. The Transformative Potential of Tutoring for PreK–12 Learning Outcomes: Lessons from Randomized Evaluations. Abdul Latif Jameel Poverty Action Lab. Evidence review (PDF). Source of the adult-tutoring contrast: pooled 0.37 standard deviations across 96 randomized evaluations, with teacher-delivered tutoring at 0.50, paraprofessional at 0.40 and volunteer at 0.21. This is adult tutoring, used here only as the comparison case.



















