Skip to content

Active Learning

Principle · All principles
Evidence · 43 claims (20 for, 22 mixed, 1 against) · 41 studies (11 causal, 8 quant-synthesis, 8 associational, 6 review, 4 theoretical, 3 design, 1 qualitative), q1–q4 · 6 of 41 report an effect size · 36 claims rest on one study

Conditional relationship

For learners meeting new material in a course, replacing part of the time they would spend only listening with tasks in which they generate, explain, predict, compare or apply the content, with feedback on what they produce, is expected to raise later course performance. The best-recorded case is undergraduate STEM courses (postsecondary, classroom, term-plus), with active-learning sections compared against traditional lecturing on course exams and concept inventories (course-assessment, general-achievement, conceptual-understanding). The relationship is about what learners are asked to do with the content, not about visible activity: a task that can be completed without thinking with the content is not expected to help.

Four conditions bound it. Learners with little prior knowledge of the target (novice) are not expected to benefit from activity that leaves them to discover the essential content unaided. The kind of engagement matters: generating beyond the material is predicted to help more than merely manipulating it, though that ordering has not held in every setting. The comparisons behind the average usually changed several course features at once, so the average belongs to "an active-learning course", not to any single activity within it. And the horizon recorded is the end of a course; durable retention and far transfer are not established here. This is a bounded expectation for groups under stated conditions, not a forecast for one learner.

Two converted pages hold narrower relationships inside this one. Retrieval Practice is one active task, recalling from memory, with its own delayed-retention evidence; Peer Discussion is another, discussing reasoning after an individual commitment, with an immediate isomorphic-question outcome. This page owns the general claim that changing what learners do in class changes what they learn, and the qualifications that apply to every such task. The wiki holds no converted pattern for active learning as a whole.

Fitting the design to a situation

Ask for four facts when the brief omits them: how much new the topic is to these learners, what share of class time can be given to activity, which outcome counts (course exam, concept inventory, information recall, general reasoning), and how large the class is.

If the brief says… Then change… Basis
An undergraduate STEM course taught mainly by lecture, judged on the course exam and pass rate Replace part of each lecture with tasks learners do in class (questions, problems, predictions, explanations) with feedback on what they produce; expect higher exam scores and fewer failing grades on average Active learning improves exam performance [+S]; Lecture students are 1.5 times more likely to fail [+M]
The course has wide gaps in pass rates or exam scores between student groups, and narrowing them is a goal Make the activity high-intensity and highly structured (regular graded preparation and in-class work every session), a large share of class time, not an occasional break in lecture; apply it to the whole class rather than to students singled out; track the gap per section, because results varied widely between courses Active learning narrows achievement gaps [+S] (two course studies: a disadvantaged-student gap in biology halved, a gender gap in physics closed at the highest dose); Only high-intensity active learning narrows gaps [+M]; Gap effects vary widely across studies [~M]; Class size, level and discipline did not explain the variation [~W] (null moderators, which the authors say are underpowered)
The class is a large lecture (well over 50 students) Still convert time to activity, using formats that scale (individual responses, pair discussion, polls); expect a smaller average gain than in classes of 50 or fewer Benefits are greatest in classes of 50 or fewer [~M]
The course uses in-class response questions (clickers, polls) Follow each targeted question with a short written explanation of the same question after class, and read the explanations rather than the votes: presence for the question and a completed explanation each predicted the related exam item, an explanation showing understanding predicted it best, and for the lowest-scoring quarter of the class presence alone showed no exam benefit Clicker participation and homework completion each predict exam performance [+M]; Explanations showing understanding predict exam performance [+M]; No benefit for the lowest quartile [~M] (all one observational study of two introductory biology sections; not randomised)
Teaching moves to video or online Keep the learners' activity at least as interactive as what the video replaces (questions that need an answer, discussion, co-viewing with a peer); a video less interactive than the class it replaces showed no benefit. Do not count on embedded self-explanation prompts alone when viewing time is fixed Video helps only when its interactivity matches or exceeds the comparison [~M]; Self-explanation prompts in video did not help at controlled time [~M]
Learners are in primary or secondary school The course-level comparisons above are undergraduate. One associational study across Chicago elementary schools found students of teachers reporting frequent interactive instruction gained about 5% more in a year than the city average in reading and mathematics; treat this as a lead, keep the guidance for novices, and check learning with an independent probe Interactive instruction associated with achievement gains in Chicago [+W] (teacher-reported instruction; not a comparison of formats)
The activity asks learners to predict or answer before being told Ask for a committed answer (and, where quick, how sure they are) before explaining, then give immediate corrective feedback; do not steer away from confident wrong answers, since errors made with high confidence are corrected more durably once feedback arrives High-confidence errors are retained better after correction [~M] (a review and one experiment, read from abstracts; mostly memory for answers, not course concepts)
Learners are meeting the topic for the first time (novice) Give a worked example or short explanation before the task, and scaffolds and feedback during it; do not leave learners to discover the key content unaided Minimal guidance is less effective for novices [~M]
The outcome that counts is conceptual understanding, measured by a concept inventory Aim activities at the concepts and misconceptions the inventory targets; expect a larger gain there than on instructor-written exams Gains are larger on concept inventories [+M]; Peer Instruction raised Force Concept Inventory scores at Harvard [+M] (one course compared across years, with other refinements added over time)
The goal is general scientific reasoning, not course content Teach and assess the reasoning explicitly; content-focused active engagement alone left reasoning-test gains near zero in one course series Content gains but no reasoning gains [~W]
The goal is mainly to transmit information (terms, facts, a procedure to recognise) A clear explanation may do as well; spend the active time on applying, comparing and explaining rather than on receiving the information Discussion promotes more active thinking; lecture is no worse for information [~W]
The brief asks which activity to use Prefer tasks that make learners generate beyond the material (explain, predict, compare, justify) and, in pairs, build on each other's ideas, over highlighting or copying Constructive learning beats active and passive [~W]; Self-explanation improves understanding [+M] (immediate tests; delayed and classroom evidence is thin)
The instructors are new to designing active tasks Give them ready-made generative and interactive tasks and review lesson plans with them; teachers trained in the framework rarely produced such tasks on their own Teachers struggled to design constructive and interactive activities [~M]
Learners will rate sessions, or the redesign will be judged by course evaluations Tell learners that effortful sessions can feel less productive than a smooth lecture, and judge learning with an independent probe rather than ratings Learners in active classes felt they learned less [~W]
The redesign will be evaluated against the old course Record every feature that changed with the format (quizzes, homework, home study time, instructor) and hold what you can constant; otherwise the gain cannot be credited to the activities. Expect the "active" version to keep a lecture component and to add quizzes, homework and home preparation unless you decide otherwise Most comparisons differ on more than one feature [~M]; Time on task is often unequal [~M] (in 44% of 59 studies, always with more home time for the active condition); Active conditions more often add quizzes, homework and exam preparation [~M] (quizzes in 36% of active against 6% of lecture conditions); Most active conditions keep a substantial lecture component [~M] (80% of 88 conditions)
Some learners are neurodivergent or find group work hard Use fully structured tasks with defined roles and some instructor involvement, in person where possible, and avoid unannounced public presenting Comfort highest for fully structured assignments [~W]; Both groups prefer some instructor involvement [~W]

Observation, state and explanation

Record what a learner does with a stated task under stated conditions: whether they produce an explanation, a prediction, a solution or a comparison, what support was available, and what they could do afterwards on an independent probe. "Engaged" and "passive" are inferences, not observations. Talking, writing, clicking a response or moving materials are behaviours; whether the learner is generating inferences beyond the material is a hypothesis about processing that none of these behaviours establishes on its own. Learner expertise is task-specific, as elsewhere: a student new to this topic is a novice for it whatever their course level.

The rows below are proposals for reading responses during an active task. They have not been tested with learners.

Response Plausible explanations Observation that could change the interpretation
Completes the activity sheet or group task, but cannot explain or apply the idea on an individual probe The task could be completed without thinking with the content (busywork); a peer did the work; the learner understood during the task but did not retain it Ask for an individual explanation on a matched item immediately after the task, and record who produced each part of the group output. If immediate explanation succeeds and later recall fails, the problem is retention, not the task.
Stalls or produces mostly wrong output on a generative task Missing prior knowledge needed to generate anything useful (novice); task demand too high for working memory; unclear task instructions Give a short worked example or explanation of the same relation, then a matched generative task. Success after the example points to missing knowledge or load; continued failure with clear instructions points elsewhere.
Generates confidently, but the output is wrong A misconception the task has surfaced; a correct idea poorly expressed; generation aimed at the wrong features of the task Ask the learner to predict an outcome the misconception and the correct idea disagree on, then give corrective feedback and re-probe on a new case.
Answers an in-class question correctly but writes an incomplete or wrong explanation of it afterwards Guessed or followed peers' votes; understood but cannot yet put it into words; understood in class and forgot Treat the explanation, not the vote, as the observation: in one introductory biology course an explanation showing understanding predicted the related exam item across the whole class, while presence for the question did not help the lowest-scoring quarter (explanation quality predicts exam performance [+M]). Ask for the explanation on a new isomorphic question to separate words from understanding.
Section scores rise after a switch to an active-learning format The activities; extra quizzes, homework or preparation time that came with the new format; a different instructor or exam Record every feature that changed with the format, including home time and assessments. Only a comparison that holds the other features constant can attribute the gain to the activities.
Learner reports the active session as less useful than lecture Dislike of effort or exposure to peers; a real mismatch with what the learner values; feeling of fluency from lecture that is not matched by performance Compare self-report with an independent probe on the same content, and ask what the learner wanted from the session.

These probes can shift confidence among explanations; one contrast does not identify a cause, and a probe is itself a learning event. Record what remains unresolved.

Evidence and qualifications

  • Active learning improves exam performance [+S]: the central comparison. A meta-analysis of undergraduate STEM courses found exam and concept-inventory scores 0.47 SD higher under active learning than under traditional lecturing (158 studies), and failure under lecturing 1.5 times as likely (odds ratio 1.95, 67 studies); the effect held across disciplines and class sizes and was largest in small classes and on concept inventories. A second meta-analysis, of comparisons in which the same instructor taught both ways, found the exam gap for underrepresented and low-income students narrowed from −0.62 to −0.42 SD, varying widely between studies, and only in classes using high-intensity active learning. So the expected benefit depends on how much class time is active. Course-end outcomes only; one of the two entries passes the source check (abstract), the other could not be confirmed.
  • Most active learning comparison studies differ on more than one course feature [~M]: a limit on attribution. A review of 59 college science comparison studies found only 11.9% in which exactly one feature differed between conditions; active-learning conditions averaged 5 of 14 course features. The average effect above therefore describes course redesigns, not a single activity. Descriptive counts from one review, not yet checked against its source. Its neighbours, recorded from the same review, say most active-learning conditions still contained a substantial lecture component and that active conditions often asked for more home time (listed under Further evidence below).
  • Minimal guidance is less effective for novices than explicit instruction [~S]: a qualification on guidance. Across 580 comparisons, unassisted discovery did worse than explicit instruction (d = 0.38 favouring explicit instruction), while discovery enhanced with feedback, worked examples, scaffolding or elicited explanations did better than other instruction. The claim page says plainly that the deficit belongs to unguided discovery, not to active learning as a whole. For novices, "active" has to mean guided activity. Partly checked (1 of 3 entries pass on abstract).
  • Constructive learning beats active and passive learning [~M]: a qualification on the kind of engagement. A theoretical review with an embedded lab study reports learning rising from passive to active to constructive to interactive engagement; a large multilevel field study in university classrooms supported only the end points (passive lowest, interactive highest) and did not confirm that constructive activities outperform active ones, on perceived learning and affective outcomes. Neither reports a standardised effect size. The direction is a working hypothesis, not a settled ordering.
  • Discussion promotes more active thinking than lecture, though lecture is no worse for transmitting information [~M]: a qualification on the objective. Reviews summarised on the claim page find lecture about as effective as alternatives for acquiring information, while an observation study of students' reported thoughts found more problem-solving and synthesis during discussion (8.3% against 1% of class time). Where the objective is information acquisition alone, the expected advantage of active formats is not established by this evidence. Not yet checked against its sources; two of its three entries have no source text available.
  • Active learning narrows achievement gaps for disadvantaged and underrepresented students [+S]: the equity case, at course level. In introductory biology, adding several highly structured active-learning components raised everyone's performance and cut the gap for students in a programme for educationally and economically disadvantaged students (baseline failure about 22% against 10%) to roughly half; in introductory physics, a "high dose" of active engagement closed the gender gap. Both are quasi-experimental course comparisons with no effect size printed for the gap in the entries read, and both used a whole-course redesign rather than targeting students. Theobald et al.'s meta-analysis (in the first bullet) adds that only high-intensity active learning narrowed gaps, and its null moderator results for class size, level and discipline are underpowered by the authors' own account.
  • Self-explanation improves understanding [+M] and Self-explanation prompts in an instructional video do not improve retention or comprehension when learning time is controlled [~M]: the best-recorded generative task, and a limit on it. A meta-analysis of 64 reports found prompting learners to self-explain improved outcomes by g = .55 across task types, subjects and levels; a mathematics meta-analysis found the gains on immediate tests, with much more limited evidence after a delay or in classrooms. A controlled-time experiment with video (N = 128) found specific, inference-demanding prompts gave no benefit, with Bayes factors favouring no difference. So a generative prompt is a reasonable default task but not a guaranteed gain once time is equated.
  • Video benefits learning only when its interactivity matches or exceeds the comparison condition [~M]: the same relationship seen from the other side. Across higher-education studies replacing teaching with video, video showed no benefit when the comparison condition was more interactive (g = −0.07), a moderate one when interactivity was equal (g = 0.33) and a larger one when video came with more interaction (g = 0.62, k = 8). What the learner is asked to do, not the medium, carries the effect.
  • Clicker participation and homework completion each predict exam performance [+M]: one observational study of 346 introductory biology students. Presence for a targeted in-class question and completing an at-home written explanation each predicted the related exam question, without interaction; the lowest-scoring quarter showed no benefit from presence alone (lowest quartile [~M]). Associational: students who attend and complete homework differ in other ways.

Retain the learners, setting, comparator, which course features changed, the instrument and its alignment, and the horizon when transporting these observations. A course-level average does not forecast an individual's response, and a nonsignificant difference between engagement modes does not establish that they are equivalent.

The evidence by question

Each question below links a general claim that gathers the studies behind it into one table, one row per distinct study, with each study's design and effect as recorded and no averaged verdict (eval/synthesis/active-learning.json, scripts/synthesize_claim.py). The count is of distinct studies across the claim and the narrower claims that bear on it.

Further evidence, not yet read against this model

Claims read against the model on 2026-10-19 that do not change it. Each says why.

Objective and learner-valued goal

Ask what the learner wants from the session and why, separately from the designer's objective. A learner may value hearing an expert explain the material clearly, or may value being able to use it on a problem they care about; a designer may want conceptual understanding that shows on a concept inventory. Agreement, divergence and uncertainty are observations. Do not infer understanding from participation, and do not treat a learner's preference for lecture as a deficit to be corrected; it may be a real goal, or a feeling of fluency that an independent probe can test.

For example, a first-year chemistry student may value passing the end-of-term exam and see in-class problem sets as time taken from "covering the material", while the instructor's objective is that students can explain why a reaction rate changes with temperature on an unfamiliar case. Attendance and completion of the in-class sheets cannot establish either aim. An individual explanation on a new case, the exam the student is aiming at, and the student's own account of what the problem sets were for can.

What would revise this model?

The expectation should weaken if comparisons that hold course features other than the activities constant (time, assessments, quizzes, instructor) show no advantage for active-learning sections on aligned outcomes. If gains are explained by the extra quizzes, homework or home time that came with the format, credit those features rather than activity. If constructive and interactive tasks keep failing to outperform merely active ones in authentic classrooms, drop the engagement ordering and keep only the end points. If novices given guided activity do no better than novices given clear explanation followed by practice, the claim reduces to the value of practice. If generative prompts keep showing no gain once learning time is held equal, as in one controlled-time video experiment, part of the credit given to generative tasks belongs to the extra time they take. Do not protect the model by relabelling every failure as "not really active learning" after the fact.

A course-end exam advantage, a narrowed achievement gap, durable retention, transfer to unfamiliar problems and changes in reasoning ability are separate claims. The present evidence does not establish an optimal share of active class time, a best activity type, or an effect for any individual learner, and it rests largely on undergraduate STEM courses: the wiki records only one associational study for school-age learners and none that compares an active-learning course with lecture for them.

Claims

Every claim this page cites is in the model above or under Further evidence (re-read 2026-10-19). - Student self-determination was infrequent and more common during Active, Constructive, and Interactive tasks than Passive tasks [~W] - Predominantly lecture-based courses are associated with less engagement, while class discussion and group work time are associated with more [+W] - The authors conclude that students in Just-in-Time Teaching courses better understand Newton's Third Law after instruction than students in traditional lecture courses [+W] - Simulation games are more effective when they actively rather than passively convey course material [+W] - About two-thirds of Freeman et al. (2014) articles reported results favoring only the active learning group, but most comparisons were active learning versus lecture with weak internal validity controls [~M]

Examples

Key Sources

  • Freeman, S., Eddy, S. L., McDonough, M., Smith, M. K., Okoroafor, N., Jordt, H., & Wenderoth, M. P. (2014). Active learning increases student performance in science, engineering, and mathematics. Proceedings of the National Academy of Sciences, 111(23), 8410-8415. https://doi.org/10.1073/pnas.1319030111
  • Chi, M. T. H., & Wylie, R. (2014). The ICAP framework: Linking cognitive engagement to active learning outcomes. Educational Psychologist, 49(4), 219-243. https://doi.org/10.1080/00461520.2014.965823
  • Biesta, Gert J. J. (1997). George Herbert Mead's Lectures on Philosophy of Education at the University of Chicago (1910-1911). Paper presented at the Annual Meeting of the American Educational Research Association, Chicago, IL. https://eric.ed.gov/?id=ED422199
  • Benedetti R., Mariotti E., Montalbano V., Porri A. (2011). Active and cooperative learning paths in the Pigelleto's Summer School of Physics. FFP12, Udine. https://www.sif.it
  • Theobald, E. J., Hill, M. J., Tran, E., Agrawal, S., Arroyo, E. N., Behling, S., … Freeman, S. (2020). Active learning narrows achievement gaps for underrepresented students in undergraduate science, technology, engineering, and math. Proceedings of the National Academy of Sciences, 117(12), 6476–6483. https://doi.org/10.1073/pnas.1916903117
  • Haak, D. C., HilleRisLambers, J., Pitre, E., & Freeman, S. (2011). Increased structure and active learning reduce the achievement gap in introductory biology. Science, 332(6034), 1213–1216. https://doi.org/10.1126/science.1204820
  • Lorenzo, M., Crouch, C. H., & Mazur, E. (2006). Reducing the gender gap in the physics classroom. American Journal of Physics, 74(2), 118–122. https://doi.org/10.1119/1.2162549
  • Bisra, K., Liu, Q., Nesbit, J. C., Salimi, F., & Winne, P. H. (2018). Inducing self-explanation: A meta-analysis. Educational Psychology Review, 30(3), 703–725. https://doi.org/10.1007/s10648-018-9434-x
  • Beege, M., & Ploetzner, R. (2025). Learning from interactive video: the influence of self-explanations, navigation, and cognitive load. Instructional Science. https://doi.org/10.1007/s11251-024-09693-5
  • Noetel, M., Griffith, S., Delaney, O., Sanders, T., Parker, P., del Pozo Cruz, B., & Lonsdale, C. (2021). Video improves learning in higher education: A systematic review. Review of Educational Research, 91(2), 204–236. https://doi.org/10.3102/0034654321990713
  • Kirkwood-Watts, D. L., Bremers, E. K., Robinson, E. A., Brazeal, K. R., & Couch, B. A. (2022). Revisiting clickers: In-class questions followed by at-home reflections are associated with higher student performance on related exam questions. Journal of Microbiology and Biology Education. https://doi.org/10.1128/jmbe.00038-22
  • Martella, A. M., Lovett, M. C., & Ramsay, L. (2021). Implementing active learning: A critical examination of sources of variation in active learning college science courses. Journal on Excellence in College Teaching, 32(1), 67–96. https://jlect.com/
  • Metcalfe, J. (2017). Learning from errors. Annual Review of Psychology, 68, 465–489. https://doi.org/10.1146/annurev-psych-010416-044022
  • Smith, J. B., Lee, V. E., & Newmann, F. M. (2001). Instruction and achievement in Chicago elementary schools. Consortium on Chicago School Research. https://consortium.uchicago.edu/publications/instruction-and-achievement-chicago-elementary-schools
Page metadata
Field Value
type principle
id active-learning
aliases education-around-child-response, paradox-puzzle-before-theory-lecture
title Active Learning
description For learners meeting new material in a course, class time spent generating, explaining or applying the content with feedback, rather than only listening, is expected to raise course-end exam and concept-inventory scores, qualified by guidance for novices, the kind of engagement and what else changed with the format.
status review
canonical True
generated by claude/unspecified on 2026-10-19
trust unverified — no one has confirmed this page's content
sources 12 — freeman-2014, chi-2014, biesta-1997, benedetti-r-2011, theobald-2020, haak-2011, lorenzo-2006, bisra-2018, beege-2025, noetel-2021, kirkwood-watts-2022, metcalfe-2017

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