Active Learning Improves Exam Performance¶

Claim · All claims
Evidence · 3 studies · 2 quant-synthesis r2–r3, 1 theoretical r2 · q2–q4 · i1–i2
Currency · newest study 2026 · checked against its sources 2026-10-20 (automated, 3 of 3 entries) · not reviewed by a person

Students taught through active-learning approaches (e.g., in-class problem-solving, discussion, peer instruction) tend to perform better on examinations than students taught through traditional lecture alone.

Subclaims¶

q4 i2 Across 158 studies in undergraduate STEM courses, exam and concept-inventory scores were 0.47 SD higher under active learning than under traditional lecturing; the result held across STEM disciplines and class sizes, with the largest effects in small classes and on concept inventories. → Freeman et al. 2014

q3 i1 In 15 STEM studies (9,238 students), active learning narrowed the exam-score gap between underrepresented/low-income and other students by 33% (from −0.62 to −0.42 SD), but only high-intensity active learning narrowed gaps, so how much class time is active matters. → Theobald et al. 2020

q4 i2 Across 158 independent comparisons, student performance on identical or equivalent examinations, concept inventories, and other assessments increased by a weighted standardized mean difference of 0.47 under active learning versus lecturing. → Freeman S 2014

q2 i? Environments that proactively engage students show demonstratively higher learning gains than traditional one-way lecture environments, per the consensus study the article cites. → Finkelstein 2026

Evidence¶

Freeman et al. 2014¶

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. doi:10.1073/pnas.1319030111

q4 · well-powered meta-analysis · i2 · medium effect, 0.47 SD · n=225 studies · quant-synthesis · r2

A meta-analysis of 225 studies that compared undergraduate STEM courses taught by traditional lecturing with the same courses taught with active learning. On examinations and concept inventories (158 studies), students under active learning scored 0.47 standard deviations higher, about 6% on average exam scores. Students under traditional lecturing were 1.5 times more likely to fail (odds ratio 1.95, 67 studies). The effect held across STEM disciplines and all class sizes, was largest in small classes (50 students or fewer) and on concept inventories rather than course exams, and did not appear to be explained by publication bias or by weaker study controls.

Theobald et al. 2020¶

Theobald, E. J., Hill, M. J., Tran, E., Agrawal, S., Arroyo, E. N., Behling, S., Chambwe, N., Cintrón, D. L., Cooper, J. D., Dunster, G., Grummer, J. A., Hennessey, K., Hsiao, J., Iranon, N., Jones, L., Jordt, H., Keller, M., Lacey, M. E., Littlefield, C. E., … 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. doi:10.1073/pnas.1916903117

q3 · meta-analysis (Bayesian regression over pooled student data) · i1 · small effect, exam gap narrowed by 0.20 SD (−0.62 to −0.42) · n=15 studies (9,238 students) for exam scores · quant-synthesis · r3

A meta-analysis of published and unpublished STEM course comparisons of traditional lecturing and active learning in which instructor differences were controlled (the same instructor in both, instructors drawn at random from a pool, or both conditions team-taught by four or more), looking at students from underrepresented minority and low-income groups against their classmates. On identical or equivalent exams, the gap was −0.62 SD under lecturing and −0.42 SD under active learning, a 33% reduction; gaps in passing rates narrowed by 45% (26 studies, 44,606 students). The results varied widely between studies, and only classes that used high-intensity active learning (a larger share of class time in activities) narrowed the gaps. This qualifies the claim: the benefit on exams is not uniform and depends on how much active learning is actually done.

Finkelstein 2026¶

Finkelstein, N. (2026). A principled way to think about AI in education: guidance for educators and policy makers on action based on goals, models of human learning, and use of technologies. Draft manuscript, University of Colorado, Boulder. https://arxiv.org/abs/2510.01467. doi:10.48550/arxiv.2510.01467

q2 · i? · theoretical · r2

The article reports this as a summary of a recent National Academies consensus study on principles of educational activity design, not as its own empirical analysis. It states that "environments that proactively engage students show demonstratively higher learning gains" than lecture. No effect size or sample is printed in the article.

"Decades of evidence and foundational theories of learning note that learning is an active process (not the absorption of information), and environments that proactively engage students show demonstratively higher learning gains than our traditional, one-way, lecture environments."

Studies bearing on this claim¶

Does replacing part of lecture with active learning raise exam scores and pass rates? This table gathers the studies behind this claim and the 15 narrower claims that bear on it: 12 distinct studies (7 causal, 2 quant-synthesis, 1 theoretical, 1 associational, 1 review), q2–q4, 5 of 12 with an effect size coded. One row per study, codes as each entry records them; no effect is averaged and no verdict is drawn. Built by scripts/synthesize_claim.py from eval/synthesis/; a row that looks wrong is fixed on the claim it comes from.

Discussion¶

Scope of the claim. "Active learning" here refers to instructional designs that engage students in doing things and thinking about what they are doing during class — solving problems, discussing cases, answering questions, working in groups — rather than passively receiving content. The claim concerns exam performance, not attitudes or self-reported engagement, which can diverge from measured learning gains.

Moderators to expect. The size of the benefit likely depends on class size, discipline, the depth of active engagement (structured tasks vs. unstructured activity), and how well active elements are aligned with assessment — see Constructive Alignment. Active formats that merely add activity without cognitive engagement (e.g., unstructured group work with no accountability) may show little or no benefit. The expertise reversal effect suggests that for advanced learners, some scaffolding-heavy active formats can lose their advantage.

Interaction with cognitive load. Poorly designed active tasks can impose extraneous load on novices; pairing active practice with worked examples or other cognitive load management supports is one way to protect the benefit.

Open questions. Which active-learning formats produce the largest gains, how benefits scale with implementation fidelity, and whether effects hold equally for high-stakes versus low-stakes exams remain to be established from the evidence to be added.

Related Claims¶

• Case-based learning improves exam performance. — a specific active format with direct exam-performance evidence
• Comparing contrasting cases improve learning. — structured comparison activity as an active-learning mechanism
• Activation improves learning. — activating prior knowledge during active tasks supports encoding
• Cognitive disequilibrium motivates conceptual change. — productive struggle as a driver of active-learning gains
• Small-group learning improves undergraduate STEM achievement — related
• Hake's survey found interactive-engagement courses achieved average normalized gains about two standard deviations greater than traditional courses, with overall gain–pretest correlation of +0.02 — a narrower finding that bears on this claim
• Active learning narrows achievement gaps for disadvantaged and underrepresented students — a narrower finding that bears on this claim
• The authors of one physics-course study attribute its technology-supported gains to the student-centred, active-engagement pedagogy the technology was used within — a narrower finding that bears on this claim
• HOCS clicker questions during learning improve performance on LOCS exam questions more than LOCS clicker questions do — a narrower finding that bears on this claim
• Students in courses with guided note-taking and other active learning activities reported less difficulty with motivation and engagement after the shift online — a narrower finding that bears on this claim
• QR faculty on several CSU campuses are incorporating active-learning and problem-solving pedagogies into redesigned entry-level math courses — a narrower finding that bears on this claim
• A carefully engineered AI tutor produced more than twice the learning of a strong active-learning class, in less time (Kestin et al., 2025, as reported) — related
• Students rated the manual annotation activity as more effective than traditional lectures for understanding bias and reported increased motivation — a narrower finding that bears on this claim
• AI tutors explicitly designed according to educational theory outperformed traditional active learning in rigorous academic settings — related
• Effects of active learning on achievement gaps vary widely across studies, with a minority of studies showing larger benefits for non-minoritized students — related
• Active learning reduces the examination-score achievement gap between minoritized-group and non-minoritized students in undergraduate STEM by 33% on average — related
• Class size, course level, and discipline did not explain significant variation in how achievement gaps changed under active learning — related
• Active learning's performance benefits hold across all STEM disciplines — related
• Active learning gains are larger on concept inventories than on instructor-written course examinations — possibly the same claim (merge candidate)
• Active learning benefits students in all class sizes, with the greatest effects in classes of 50 or fewer students — possibly the same claim (merge candidate)
• Students in traditional lecture STEM courses are 1.5 times more likely to fail than students in active-learning courses — related
• CLASP students achieved an average normalized FCI gain of 0.39, above the traditional-course range and in the middle of the interactive-engagement range, despite only about a third of one quarter on mechanics — related
• The review reports that formative assessment produces significant and often substantial learning gains, citing Black and Wiliam — related
• Discussion promotes more active thinking than lecture, though lecture is no worse for transmitting information — related
• On Newton's Third Law questions specifically, Just-in-Time Teaching students gained 50.8%±4.1% while traditional lecture students gained only 6.6%±5.2% — a narrower finding that bears on this claim
• Predominantly lecture-based courses are associated with less engagement, while class discussion and group work time are associated with more — related
• High school students taught with self-study followed by discussion achieved higher immediate retention and near-transfer scores than those taught with lecture followed by discussion — a narrower finding that bears on this claim
• Active learning narrows the passing-rate gap between minoritized-group and non-minoritized students in undergraduate STEM by 45% on average — related
• The active-learning advantage is robust to publication bias and to variation in methodological rigor of the included studies — related
• Introductory physics students taught with Just-in-Time Teaching showed larger overall Force Concept Inventory gains (37.6%±2.0%) than students in traditional lecture courses (17.9%±2.5%) — a narrower finding that bears on this claim
• In Harvard's introductory physics course, average FCI posttest scores and normalized gain rose sharply in the year it switched from traditional lectures to Peer Instruction — a narrower finding that bears on this claim
• 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 — related
• Changing activities during lessons improves student attention during the activity and subsequent lecture segments — related
• Differences in AI-supported instruction effects across STEM disciplines are not statistically significant — related
• AI-PBL students achieve high normalized learning gains (g = 0.73) in computational thinking and academic achievement while control students record low gains (g ≤ 0.20) — related
• Test questions reinforced by Warm Up questions or Cooperative Learning activities showed higher average normalized gains (~52%, ~60%) than questions discussed in class only (~15%) — a narrower finding that bears on this claim
• Experiential instruction produces larger immediate pre-post gains than instructional lecture in AI engagement and metacognitive knowledge of cognition — a narrower finding that bears on this claim
• Dosage of instruction was equal between comparison groups in fewer than a third of active learning studies and unspecified in most — a narrower finding that bears on this claim
• In physics-based courses for non-science majors taught with active-engagement pedagogy, normalized gains were 0.38–0.42 on content tests but 0.06 on a scientific-reasoning test — a narrower finding that bears on this claim
• High-intensity active learning narrows achievement gaps far more than low-intensity active learning; intensity of in-class active engagement is positively correlated with narrower gaps — related
• Students in a technology-enhanced modeling physics class learned significantly more than students in an otherwise identical no-technology class — a narrower finding that bears on this claim
• Active learning in an entry-level biology course closed performance gaps between underrepresented and non-underrepresented minority students and increased all students' self-efficacy — related