Skip to content

Peer Teaching

Pattern · All patterns
Evidence · 31 claims (14 for, 15 mixed, 2 against) · 18 studies (11 quant-synthesis, 5 causal, 1 review, 1 design), q2–q4 · 10 of 18 report an effect size · 28 claims rest on one study

Description and scope

A reusable session routine for learners teaching one another material they have already met: each learner studies a part → prepares to teach it → teaches it to a partner who must respond (restate, answer, ask, attempt) → the partner corrects with a scripted move when the answer is wrong → roles swap within the session → each learner is checked alone. Run across weeks, the scripts thin out and learners take over the questioning. The intended change is in what each learner can do alone afterwards (immediate-performance on the taught concept, procedure or reading skill), for the learner who teaches and the learner taught.

It sits inside the Peer Learning principle, which owns the general relationship (a structured exchange of explanation and response between learners of similar standing raises individually measured achievement over studying alone) and its evidence. This page owns one configuration of it: a reciprocal, scripted, class-wide teaching session, of which Peer-Assisted Learning Strategies (PALS) is the most tested branded form in the wiki. Neighbouring configurations have their own pages: the Peer Instruction pattern (commit to an answer, discuss, revote), the Structured Peer Review pattern (comment and revise on drafts), and jigsaw-style expert groups in Fostering Communities of Learning and Cooperative Learning.

Two kinds of claim test parts of this pattern, and none tests the whole of it as specified here. Meta-analyses test peer tutoring against no tutoring (Peer Tutoring Improves Achievement [+S]), teaching against studying without expecting to teach (Learning By Teaching Improves Tutor Learning [+S]), and structured against unstructured tutoring (Structured Peer Tutoring Outperforms Unstructured [+M]). What Works Clearinghouse and evidence-rating reports test the scripted PALS routine against usual instruction, mostly in elementary reading, with small or nonsignificant effects that vary by outcome and population, and one null in mathematics. No claim tests the session lengths, the number of weeks, the point at which scripts are thinned, or the rule for when a learner moves on; those are labelled proposals below.

Inputs: establish the design brief

Field Record or elicit
Learner and starting observation Before pairing, give each learner one short task on the material alone (read a passage aloud, solve two items, explain one idea in a sentence). Record what each can do unaided, whether they can read the material, and who has met it before. Do not label a learner "strong" or "weak" from grades alone; preserve unknowns.
What is to be taught A concept, a procedure, a reading skill (fluency, decoding, comprehension), or an interpretation. Whether a wrong explanation can be spotted by the partner with a key or a criterion, or only by the teacher. Whether the material splits into parts each learner can own.
Context and activities Setting (classroom, online-instructor-led, online-self-paced, workplace-clinical), single-session or days-weeks, class size, how many adults are present to listen, whether pairs are same-age or cross-age, and whether a scripted programme (such as PALS) is already adopted.
Learner-valued goal Ask what learners want from teaching each other: to get faster help than waiting for the teacher, to practise explaining before a real audience, to help a friend, not to look slow. Record disagreement with the designer objective.
Designer objective Which capability the routine is for (decoding, fluency, comprehension, a calculation, a concept explanation), and whether the teacher's own learning (the learner who teaches) is a target as well as the taught learner's.
Outcome An individual check on new items or a new passage, scored for each learner apart from the pair's work; immediate or delayed; and whether a standardised measure or a test written for the material will be used, since effects differ between the two.

A request to "add peer teaching" does not specify these. Ask what each learner must do alone afterwards, and who will hear a wrong explanation, before choosing the pairs, the scripts or the dose.

Sequence and conditional policy

The sequence below makes the earlier page's one-line description concrete. Each step names its evidence status. The doses are untested proposals: no claim in the wiki sets a session length, a number of sessions per week or a programme length for this routine.

  1. Check the starting point and form pairs (one session before the routine starts, 10 minutes). Use the starting task to pair learners so that each pair has one learner a little further along, and plan to swap who leads. Untested proposal; the tested programmes the claims describe pair learners for turn-taking, and peer-assisted learning [+M] showed larger effects with more student autonomy and self-referenced evaluation, which the pairing does not settle.
  2. Teach the roles once before relying on them (one 20–30 minute session, plus a 5-minute reminder in the next two). Give each role a card: the learner teaching states the step, gives an example and asks the partner to do it; the partner restates, attempts and asks one "why". Model one exchange in front of the class with a deliberate error and the correction move (one fixed phrase that stops the error and asks for a retry, or the material's own correction). Basis: Structured Peer Tutoring Outperforms Unstructured [+M] (structured programmes, with roles and trained tutors, produced larger achievement effects than unstructured ones; unstructured tutoring still helped). The session length is an untested proposal.
  3. Prepare to teach (5 minutes). Each learner studies their part and writes or marks what they will teach: the main idea or step, one example, and the question they expect. Basis: Learning By Teaching Improves Tutor Learning [+S]: preparing to teach alone gave a smaller benefit than preparing and then teaching. Minutes are an untested proposal.
  4. Teach with a partner who must respond (8–10 minutes per role). The learner teaching does not read from notes; the partner attempts the task or answers, and the learner teaching corrects with the scripted move. For reading, the old page's tested session shape is partner reading with retelling, then "paragraph shrinking" (stating the main idea of each paragraph in a few words), then "prediction relay" (predict, read, check). For mathematics it is the coach modelling self-talk questions, then practice. Basis: the meta-analysis in Learning By Teaching Improves Tutor Learning [+S] found interactive teaching (responding to a learner) better than non-interactive; the PALS session shapes are those tested in the reading and mathematics trials under step evidence below. Minutes are an untested proposal.
  5. Swap roles inside the session. The learner who was taught now teaches the same or the next part. Untested proposal in itself: the tutoring claims discuss reciprocal roles but no claim compares swapping with fixed roles. The PALS routines the claims test use turn-taking.
  6. Listen to a few pairs each round and correct for everyone. The teacher hears two or three pairs per round, notes any wrong explanation, and ends the round by giving the correct model to the whole class. Untested proposal for how many pairs; structured programmes in the meta-analysis included monitoring among the features that define "structured".
  7. Check each learner alone (5 minutes). A new passage, two new items, or a one-sentence explanation in writing, without the partner. The pair's answer shows what the pair can do together, not what either can do alone. Basis: the Peer Learning principle owns the individual check; In the meeting-standards study, PALS classrooms used more peer-mediated and less teacher-led instruction [+W] is a fidelity check showing the routine displaced teacher-led time, which is why the individual check matters.
  8. Across sessions (untested proposal for an 8–12 week arc, three 30–35 minute sessions a week): weeks 1–2 with full role cards, the modelled exchange revisited, and the teacher sampling pairs every round; weeks 3–6 with the partner's prompts cut to one card ("restate, try it, ask why"), learners choosing which part to teach, and harder material; weeks 7 onward with learners running the session unprompted and the teacher reading only the individual checks and sampling one pair. In the mathematics peer-tutoring meta-analysis, programme duration and time of day moderated the effect while session length and frequency did not (Peer Tutoring Improves Achievement [+S]; the direction of the duration effect is not recorded on the claim page). The one tested dose recorded on a claim page is daily 20-minute sessions for three months (37 sessions, 12.3 hours), in a cross-age trial that found no effect on oral language proficiency (cross-age tutoring and oral proficiency [~M]).
  9. When a learner moves on (untested proposal). A learner who teaches their part correctly, answers the partner's question and passes the individual check in two consecutive sessions moves to harder material or becomes the one who teaches a learner meeting it first. A learner who fails the individual check twice in a row is taken out of the pair for that part and taught by an adult before rejoining. For first-grade readers who had not responded to class-wide PALS, one-to-one adult tutoring did as well or better (below).
Observation Discriminating probe Proposed action, with its uncertainty
Pair answers correctly; one learner's individual check is wrong Ask that learner to teach the step back to the teacher; give a matched item with and then without the partner present. If they can do it with the partner only, the partner carried the reasoning: swap roles so this learner teaches next. If they cannot do it either way, the material is not yet learned: reteach before pairing again. Untested.
Both learners give the same wrong explanation Check each learner's starting task: did both start with the error? If both started with it, it came from earlier instruction: teach it to the whole class. If only the learner teaching had it, they passed it on: check the teaching learner's part before the next session. Untested.
The learner teaching reads from the card or text; the partner only agrees Ask the partner a "what if" question the text does not answer. If the partner cannot answer, the exchange is recitation: require the partner's attempt before the next step, or give harder material to the learner teaching. Untested.
Gains on the taught passages or items but not on a new passage or a standardised test Compare the individual check on taught and untaught material. Effects on tests close to the taught material can be much larger than on standardised tests (reciprocal teaching [~M]); widen the material before claiming a general gain.
A learner keeps failing the individual check after several weeks of class-wide sessions Is the failure on the same skill each time, and does the learner read the material accurately? Treat as non-response to the routine: move to adult small-group or one-to-one teaching for that skill (nonresponders with PALS against adult tutoring [-M]).

Fitting the design to a situation

The three facts that most change the decision:

  • Whether the learner teaching knows the part well enough to correct the partner. Teaching material one has studied helps the one teaching; teaching what one has not met passes on errors, and the correction move needs a key. If unstated, ask: has each learner studied this part, and is there an answer key or criterion the partner can check against?
  • What outcome must change, and how it will be measured. The PALS evidence is positive for early reading skills (letter sounds, first-grade literacy), mixed for comprehension and fluency, and null for mathematics in learners with learning disabilities. If unstated, ask: which skill, measured how (a test of the taught material or a standardised test), and how long afterwards?
  • Which learners, and whether they have already not responded to class-wide teaching. Effects in the claims differ for kindergarten, first grade, learners with learning disabilities, English learners and older learners, and for learners who had not responded, adult tutoring did as well. If unstated, ask: who are the learners, and has a class-wide routine already been tried with them?
If the brief says… Then change… Basis
Learners are novice and meeting the material for the first time Do not ask them to teach unstudied material. Teach it to the class first (or study a worked example together), then run the routine on it with an answer key on each card, so the partner can check. Untested proposal; the Peer Learning principle carries the same caution
Some learners are well ahead (advanced) or the school can use older pupils Use cross-level or cross-age pairs: the learner ahead teaches with a role card that makes them question rather than tell, and gets harder material to prepare. Check the taught learner alone, not only the pair. Older peers as tutors support gains for older and younger students [+W] (a narrative policy review, no data printed); limit: cross-age tutoring showed no effect on oral proficiency [~M]
Kindergarten or first grade, early reading (letter sounds, decoding) Use a short scripted routine on sounds and words with turn-taking and a simple correction move, three or more times a week, and train teachers with a workshop before starting. K-PALS on Rapid Letter Sounds [+M] (2,959 kindergarten pupils in 67 schools; significant for workshop, booster and helper conditions alike); PALS on first-grade literacy [+M] (three studies, 1,501 pupils, average effect +0.23); fluency and comprehension effects in first grade were not significant (three domains [~M])
Learners with learning disabilities, in reading Keep the routine for reading, but do not rely on it alone: check each learner's comprehension and fluency individually every two weeks (proposal) and add adult-led small-group teaching for those not moving. Comprehension, learners with learning disabilities [~M] (two trials, 40 and 20 pupils; effect sizes 0.60 and 0.91, neither significant); fluency [~M]; significant for the full sample, not the disability subsets [~M]
The goal is mathematics for learners with learning disabilities Do not expect the peer routine to carry it; keep explicit teacher instruction as the main route and use peer practice only for fluency on steps already taught, with an answer key. PALS showed no discernible effects on mathematics for learners with learning disabilities [~M] (one trial, 40 pupils in grades 2–4); the mathematics peer-tutoring meta-analysis in Peer Tutoring Improves Achievement [+S] is across learners generally
Learners have already not responded to class-wide teaching (nonresponders) Prefer one-to-one or small-group teaching by an adult for the skill they missed; keep the peer routine for practice, not as the intervention. Nonresponders: PALS against adult tutoring, comprehension [-M] (41 first graders; effect −0.31, not significant); alphabetics [~M] and fluency [~M] showed no difference
English learners Run the routine on reading with roles in English and allow explanation in the first language where content is the goal; check oral language separately, since pairing alone may not move it. PALS for English learners [~M] (99 pupils in grades 3–6, combined effect 0.31, no single measure significant); Peer Tutoring and Response Groups on English language development [+M] (three trials); limit: oral proficiency null [~M]
Upper elementary or older learners, reading comprehension Use the comprehension activities (retell, shrink the paragraph to its main idea, predict and check) with learners leading in turn, after the teacher has modelled each. Measure on a new passage, not only the taught ones. PALS comprehension, Fuchs et al. 1997 [~M] (120 pupils, average age 9.8 despite the claim's "adolescent" label; 0.48, not significant after clustering); Reciprocal Teaching Improves Reading Comprehension [~M] (16 studies; much larger on experimenter-made than standardised tests)
Adults in work settings (staff teaching each other a procedure or case) Each colleague prepares one procedure or case, teaches it to another who must perform it or decide a variant, then they swap; the check is a short case decision made alone. Untested; carried from Learning By Teaching Improves Tutor Learning [+S], whose studies are not described on the claim page by age or setting
classroom with one teacher for 25–30 learners, or scarce facilitator time (volunteers, a busy shift, one tutor dropping in) Put the structure in the role cards and answer keys, not the adult: the adult samples two or three pairs a round, corrects the commonest error for everyone, and reads only the individual checks. Structured Peer Tutoring Outperforms Unstructured [+M]; the sampling rule is an untested proposal
A single session only, or a fixed short budget (for example 30 minutes) One cycle: 3 minutes to study the part, 5 to prepare, two 7-minute teaching turns with roles swapped, 5 for the individual check, 3 for the teacher's correction to the class. Report only immediate performance. Untested proposal; the tested contrast (teaching against studying) sets no duration (Learning By Teaching Improves Tutor Learning [+S])
Mixed modes: some learners remote, some in the room Pair remote with remote and room with room, or give mixed pairs a shared document and an audio channel; time the teaching turns on a visible clock, since an untimed peer step is the one dropped. Untested proposal
online-self-paced or phone-only, no live session Each learner records a 1–2 minute voice note or short text teaching one part; an assigned partner must reply within two days with an attempt and one question; the first learner answers. The individual check stays a short quiz the learner takes alone. Untested proposal
Learners with low confidence, low literacy, or anxious about speaking Pair with a trusted partner, let them rehearse their part privately first, allow teaching by showing or pointing, give sentence starters on the card, and keep the first weeks unassessed. Untested proposal; teachers and pupils reported favourable impressions of PALS [+M] (surveys, no learning measure)
The brief names its own routine, script, rubric or categories (its own check-in, correction prompt, PALS manual) Keep it as given and run the teaching session on it; change only one thing the evidence says matters: train the roles once with a modelled error, or add the partner's attempt before the next step, or add one individual check. Untested proposal; training from Structured Peer Tutoring Outperforms Unstructured [+M]
A team or points competition is part of the brief Keep points tied to each learner's own check if the brief wants them; drop the announcement of a winning team if teachers find it costly. Teachers were least favourable towards PALS team components [~M] (announcing the winning team, the point structure); group rewards were linked to larger effects in peer-assisted learning [~M]
Material is private or sensitive (personal histories, health, incidents at work) The learner controls what is shared and with whom: teach the concept on a neutral or invented example; each learner applies it to their own case privately. Untested proposal
A whole-task performance the brief wants kept whole (a presentation, a clinical encounter, a lesson taught) The partner watches the whole performance and then asks and corrects against the brief's criteria (one strength, one change, one question), rather than splitting it into parts to teach. Untested proposal
Learners will perform for a real audience (clients, a family, a public event) Use the partner as the rehearsal audience: each learner teaches or performs to a partner who asks what that audience would ask, before the real event. Untested proposal
High stakes, or grades depend on the work Keep peer correction formative; grade only the individual check or product; record which explanations the teacher corrected. Untested proposal

Eleven of the twenty rows cite a claim. The rows on early reading, learning disabilities, mathematics, nonresponders, English learners and older readers rest on trials or reviews of peer tutoring against an alternative, mostly from one programme (PALS) in US elementary schools; the cross-age, adult, single-session and large-class rows carry a neighbouring finding to settings it did not test, as each Basis cell says; the rest are untested proposals.

Choosing configurations from evidence

Do not rank peer teaching above adult small-group or one-to-one teaching from these claims: the one direct comparison (nonresponders) did not favour the peers, and the review that ranks methods puts the two side by side. Do not read a pair's correct answer, or gains on the taught passages, as each learner's learning. Many of the PALS claims draw on the same few trials (Fuchs et al. 1997 and Saenz et al. 2005 appear under more than one claim), so count trials, not claims. A precise abstention names the missing comparison (a scripted peer-teaching routine against an equal-time individual or adult-led alternative, at a stated horizon) and the local observation that stands in for it: each learner's individual check, on new material, before and after.

The evidence by question

Further evidence, not yet read against this model

Illustrative design instance and observation record

Invented illustration, not a tested lesson. A second-grade teacher with 26 pupils and no classroom assistant wants pupils to read more fluently and retell what they read; four pupils have learning disabilities in reading and three are English learners. The table changed the default: pairs are formed from a one-minute read-aloud each pupil does alone, the roles are taught in one 30-minute session with the teacher modelling a misread word and the correction move, and sessions run three times a week for 35 minutes over ten weeks (partner reading with retelling, then naming the main idea of each paragraph, then predict-and-check). The teacher samples three pairs per session and reads a one-minute individual read-aloud with retell from each pupil every second Friday. The pupils with learning disabilities also get a 20-minute adult-led small group twice a week, because the routine's effects for them were not significant; English learners may retell in their first language. A pupil says she wants to read to her younger brother; the designer's objective is fluency and retell on a new passage. They agree she will take a book home in week 6 and report back. The claims cover elementary readers in US trials, not this class; the doses and the extra small group are local choices to observe.

Record: material and parts → starting task per learner and pairing → roles taught and how → sessions held (date, minutes, activities) → pairs sampled, errors heard and corrections given → each learner's individual check on new material, by date → candidate interpretations (partner carried it, error passed on, recitation, not yet readable, non-response) and their basis → changes made (swap, harder material, adult small group) → reobservation at the stated horizon. Preserve each learner's stated purpose and any change in it. An agent may summarise this record but must not fabricate missing inputs or assign numerical probabilities of mastery without a calibrated model.

Elements and limits

Peer teaching, peer discussion, peer explanation and reciprocal teaching. Strategies that run parts of the routine: Peer Tutoring, Reciprocal Peer Tutoring, Learning by Teaching, Peer-Assisted Learning Strategies and training students before peer-tutoring roles. Principle: Peer Learning.

This pattern is scoped to learners teaching one another material they have met, in a scripted, reciprocal session with an individual check. It does not cover commit-and-discuss on conceptual questions (Peer Instruction), peer review of drafts (Structured Peer Review) or group projects. Whether it does better than an equal-time individual or adult-led alternative, for which learners, with what dose and at what horizon, is not shown in the wiki; in early reading it has the most support, in mathematics for learners with learning disabilities none, and for learners who have not responded to class-wide teaching an adult did as well.

Claims

Examples

  • Peer-Assisted Learning Strategies (PALS) — the scripted, reciprocal programme whose trials supply most of this page's claims
  • Paired reading in a primary class — pupils take turns reading aloud and retelling, correcting misread words with a set phrase, then each reads a new passage alone

Key Sources

  • Fiorella, L., & Mayer, R. E. (2013). The relative benefits of learning by teaching and teaching expectancy. Contemporary Educational Psychology, 38(4), 281-288. doi:10.1016/j.cedpsych.2013.06.001
Page metadata
Field Value
type pattern
id peer-teaching
title Peer Teaching
description A reusable session routine in which paired learners who have studied the material take turns preparing and teaching it to a partner who must respond, correct errors with a scripted move, swap roles and are then checked alone; meta-analyses test peer tutoring and teaching against no tutoring or studying, WWC reviews of the scripted PALS routine find early-reading gains and small, often nonsignificant or null effects elsewhere, and no claim tests the doses, role swaps or hand-over points.
status review
grain_size lesson
canonical True
generated by claude/unspecified on 2026-10-15
trust unverified — no one has confirmed this page's content

Frontmatter is the page's machine-readable record — see the Schema & Guide.