Skip to content

Peer Instruction

Pattern · All patterns
Evidence · 21 claims (9 for, 12 mixed) · 28 studies (9 causal, 5 quant-synthesis, 5 review, 4 associational, 3 theoretical, 2 qualitative), q1–q4 · 6 of 28 report an effect size · 15 claims rest on one study

Description and scope

A reusable policy for using a conceptual question to elicit an individual committed answer, have learners discuss their reasoning with peers, collect a second answer and close with explanation: question → individual vote → peer discussion → revote → explanation. It instantiates the Peer Discussion principle. The study configurations below are evidence; the response-dependent policy (when to discuss, when to explain, when to move on) is an untested design proposal. Use it to collect better observations of reasoning, not to certify that a class has understood because a revote converged.

Inputs: establish the design brief

Field Record or elicit
Learner and starting observation Each learner's individual answer and, where possible, confidence and a one-line reason; the class distribution; task-specific prior instruction. Preserve unknowns rather than labelling the class "novice".
Context and activities Setting (classroom, online-instructor-led), class and group size, response mode (clickers, cards, digital poll, written), time per cycle, whether the distribution is shown before discussion, how groups are formed, language and access conditions, and what the instructor explains afterwards.
Learner-valued goal Ask what learners want from the discussion and why. Record disagreement with the designer's objective; do not assume that participation means agreement.
Designer objective The concept or principle targeted, the misconception or alternative the question is built to separate, and whether the aim is answering this kind of question, explaining it, or applying it to an unfamiliar case.
Outcome An individual answer on a new isomorphic question, a justification rubric, and the horizon (immediate in-class, or delayed); any course exam or concept inventory used, and how closely it matches the questions practised.

A request to "make lectures more interactive" does not specify these inputs. Ask for them before prescribing a cycle length, a correct-answer threshold for discussion, or an expected gain. Check that the question has one defensible answer and that its distractors represent reasoning learners actually hold.

Sequence and conditional policy

  1. Pose the question and collect individual answers. No discussion before the first vote. Record the distribution and, if feasible, confidence and reasons. Withhold the correct answer.
  2. Decide whether to discuss. Use the distribution as one observation, not a rule: a split class suggests reasoning worth exchanging; near-unanimity in either direction may call for explanation or a harder question. Any threshold is a local choice; the evidence here does not set one.
  3. Peer discussion of reasoning. Ask learners to convince a neighbour of their answer by giving a reason, not to agree on a letter. Circulate and sample what groups say.
  4. Revote, then an isomorphic check. Collect a second individual answer. Where the aim is understanding, follow with a new isomorphic question answered alone, since the revote cannot separate revision from copying.
  5. Explain and reobserve. Explain why the stronger reasoning is stronger, address the distractors chosen, and recheck at the stated horizon. If observations disagree with the expectation, revise the question, grouping or explanation rather than the interpretation.
Observation Discriminating probe Proposed action, with its uncertainty
Individual answers split across options Ask a sample of learners for their reasons before discussion. Run peer discussion, then revote and an isomorphic question. If reasons show the question is ambiguous rather than diagnostic, fix the question; the split itself does not say which.
Most answer correctly on the first vote Ask for a justification or pose an isomorphic question with changed surface features. If justifications hold, explain briefly and move on; if they do not, the correct answers may be guesses or recognition, and a harder or changed question may be needed.
Few answer correctly on the first vote Ask what learners took the question to be asking; check whether the needed concept was ever taught. Consider explanation or a simpler bridging question before discussion. Discussion among learners who share one misconception may reinforce it; this is a proposal, not a tested branch.
Revote converges, isomorphic question does not improve Compare who changed answers with who spoke first or most confidently in each group. Treat the revote gain as possible copying; add instructor explanation and another isomorphic check rather than declaring understanding.
Some groups talk little or one member dominates Ask privately or in writing for each learner's reasoning; vary group composition on the next question. Adjust grouping, roles or response mode. Neither silence nor talk volume alone establishes understanding.

Choosing configurations from evidence

  • For a check that separates revision from copying, the isomorphic-question design [+M] supports following the revote with a new isomorphic question answered alone. In the recorded comparison, peer discussion followed by instructor explanation exceeded either alone in undergraduate genetics courses; that is one quasi-experimental classroom study with no standardized effect size, and its "similar gains" detail could not be confirmed from the abstract. It supports keeping the explanation step, not dropping it.
  • For the expectation over a course, active learning in undergraduate STEM [+M] reports 0.47 SD higher exam and concept-inventory scores than traditional lecturing across 158 studies, largest on concept inventories and in small classes, and gap narrowing only under high-intensity active learning. Peer instruction is one of many formats pooled; the synthesis does not isolate it, and the entries have not been confirmed against their sources.
  • For what the vote alone may contribute, classroom quizzing [+W] is reported (second-hand, via a review chapter) to improve classroom exams without any discussion step. A design that drops discussion is a different configuration, and gains from the full cycle should not be attributed to discussion alone.
  • For what the cycle may not change, content gains without reasoning gains [~W] records normalized gains of about 0.38–0.42 on content instruments but 0.06 on a scientific-reasoning test under an active-engagement pedagogy, in an associational study with no comparison group, not yet checked. Do not expect concept-question practice to build general reasoning ability unless that is taught and measured.
  • For grouping, poor group leadership [~W] is reported (second-hand) to have prevented effective discussion in cooperative conceptual-change groups in community-college chemistry. Watch who closes the discussion.

Do not convert a normalized gain into a d, rank these studies by effect labels, or predict delayed retention from in-class revote changes. A precise abstention names the missing comparison, target or observation and the next way to resolve it.

Fitting the design to a situation

Ask for three facts when the brief omits them: whether a bank of conceptual questions with diagnostic distractors exists, how answers will be collected (clickers, cards, poll), and what follows class (homework, exam items matched to the questions).

If the brief says… Then change… Basis
A lecture course with a polling system and conceptual content Run the full cycle: individual vote, discussion of reasons, revote, isomorphic question alone, instructor explanation Peer discussion improves conceptual understanding [+M]
Time pressure tempts the instructor to skip either discussion or explanation Keep both; the recorded comparison found the combination ahead of either alone Peer discussion improves conceptual understanding [+M]
The exam has higher-order items Write questions at application or analysis level rather than recall Higher-order clicker questions improve lower-order exam performance [+M]
The lowest-performing students are a concern Add a short take-home reflection asking learners to explain the discussed question; completing it went with better performance on related exam items Follow-up homework modestly boosts exam performance [+W]; Clicker participation does not help the lowest quartile [~W]
Many learners switch to the correct answer on the revote Treat switchers as not yet secure; give an isomorphic item later, since only correct-correct students outperformed on related exam items Correct-correct pattern predicts higher exam performance [~W]
The brief wants clicker questions without discussion, to save time Expect some benefit from answering alone, but do not credit it to discussion; it is a different design Classroom quizzing improves exam performance [+W]
The goal includes general scientific reasoning Add explicit reasoning instruction and a reasoning measure; concept-question practice did not move a reasoning test in one course series Content gains but no reasoning gains [~W]
Learners are pre-service teachers or others who may see clickers as a game Say what the discussion is for; learners liked the routine and credited learning to the discussion rather than the device, but rated fun above learning Learning attributed to discussion, not technology [+W]; Students saw fun and variation as the main impact [~W]
Groups form freely and one member tends to decide Ask each member for a reason before the revote and vary groups; poor group leadership has been reported to cut discussion short Poor group leadership undermined discussion [~W]
Fewer than about a third answer correctly on the first vote Explain or pose a simpler bridging question before discussion, rather than discussing a shared misconception untested proposal
Nearly everyone answers correctly on the first vote Explain briefly and move to a harder or changed question untested proposal
The class runs live online (online-instructor-led) Use the poll for the votes and breakout rooms of two or three for discussion, with the isomorphic item answered alone in the poll untested proposal

The evidence by question

Further evidence, not yet read against this model

Claims this page cited before it was rewritten as a conditional model. They are evidence about the relationship, but none has yet been re-read against the model above; each says how far its own sources have been checked.

Illustrative design instance and observation record

Invented illustration, not a tested lesson. In an introductory physics lecture, the instructor poses a question on whether a heavier and a lighter ball dropped together land at the same time. Individual answers split between two options. A learner says she wants to be able to explain her answer to a younger sibling; the instructor's objective is reasoning about acceleration that transfers to an inclined-plane case. After discussion and revote, learners answer an isomorphic question about two carts on a ramp alone, then the instructor explains and asks for a one-line justification; the same kind of question returns on a quiz the following week. These are local design choices; the evidence above does not establish their optimality.

Record: question and intended distinction → individual answer, confidence and reason → distribution shown or withheld → group composition and sampled talk → revote → isomorphic individual answer → candidate interpretations (revision, copying, convergence) and their basis → instructor explanation given → next observation at the stated horizon. Preserve the learner's stated purpose and any change in it. An agent may summarize this record but must not fabricate missing inputs or assign numerical probabilities of understanding without a calibrated model.

Elements and limits

ConcepTest, conceptual questioning, misconception probes, peer discussion, reassessment, feedback and debrief. Principles the cycle draws on: Peer Discussion, Formative Assessment, Immediate Feedback and Purposeful Reflection.

This pattern is scoped to conceptual questions with a defensible answer. Open-ended discussion, procedural fluency, first exposure to unfamiliar material and attitude change need different configurations and evidence. The policy supports observation and design reasoning; its branches, any discussion threshold and its effect on delayed or transfer outcomes remain to be tested with learners.

Claims

Examples

  • Physics learners debating force or motion concept questions before repolling.
  • Medical learners comparing diagnostic reasoning on a conceptual clinical prompt.
  • Math or engineering classes using concept checks before moving into longer problem work.

Key Sources

  • Mazur, E. (1997). Peer instruction: A user's manual. Prentice Hall.
  • Crouch, C. H., & Mazur, E. (2001). Peer instruction: Ten years of experience and results. American Journal of Physics, 69(9), 970-977. https://doi.org/10.1119/1.1374249
  • Hake, R. R. (1998). Interactive-engagement versus traditional methods: A six-thousand-student survey of mechanics test data for introductory physics courses. American Journal of Physics, 66(1), 64-74. doi:10.1119/1.18809
  • Deslauriers, L., Schelew, E., & Wieman, C. (2011). Improved learning in a large-enrollment physics class. Science, 332(6031), 862-864. doi:10.1126/science.1201783
  • Arduini-Van Hoose, N. (2020). Flipped classroom. In Educational psychology. Retrieved from https://edpsych.pressbooks.sunycreate.cloud. CC BY-NC-SA 4.0.
Page metadata
Field Value
type pattern
id peer-instruction
title Peer Instruction
description A reusable question, individual vote, peer discussion, revote and explanation policy for conceptual questions, with an isomorphic individual check to separate revised reasoning from copying.
status review
grain_size lesson
author Eric Mazur
canonical True
generated by claude/unspecified on 2026-10-01
trust unverified — no one has confirmed this page's content
sources 1 — crouch-2001

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