
Active Learning vs. Lecture: What a 225-Study Meta-Analysis Found and What It Means for Your Class
Does active learning beat lecture? In a 2014 meta-analysis of undergraduate STEM courses, Scott Freeman and colleagues (PNAS) found exam and concept-inventory scores 0.47 standard deviations higher under active learning (158 studies), and lecture students about 1.5 times as likely to fail (67 studies). Average failure rates were 33.8% under lecturing and 21.8% under active learning.
Those 225 studies came from college courses, mostly science and math, and were completed or published before 2010, so the step to a high school classroom is an inference. Freeman's team defined active learning broadly, as engaging students through in-class activities and discussion. A later update from the same research group, a meta-analysis from a separate team on the humanities and social sciences, and a randomized Harvard physics comparison add detail on how big the effect is and how students experience it.
Making the call in a live simulator
Freeman's definition is broad enough that a lesson counts as active when students are doing something with the content, and a later analysis from his group links more class time spent on activities to stronger results. In Rapunzl, students make buy and sell decisions in a real-time simulator with a simulated $10,000 portfolio and then deal with what happens next, instead of sitting through a talk about how investing works.
The 48 modules use project-based learning throughout, which puts students in the position of producing work. That fits the broad definition above. The studies tested college courses, not Rapunzl, so they say nothing about this particular tool.
What did Freeman and colleagues actually find?
The authors searched for studies comparing undergraduate STEM students taught by traditional lecturing with students taught by active learning. They called their meta-analysis the largest of undergraduate STEM education to that date. The headline numbers are easy to blur together, so the table keeps the counts attached.
| Finding | The number | What it means |
|---|---|---|
| Exam and concept-inventory performance | +0.47 SD (158 studies) | About half a standard deviation, roughly the gap between an average student and one at the 68th percentile |
| Average exam scores | About 6% higher | The authors describe this as raising average grades by about half a letter |
| Odds ratio for failing under lecture | 1.95 (67 studies) | In the authors' words, lecture students were 1.5 times more likely to fail |
| Average failure rate in those 67 comparisons | 33.8% lecture, 21.8% active | About one in three students failed the lecture sections and a little over one in five failed the active ones |
An odds ratio of 1.95 does not mean lecture students were 1.95 times as likely to fail. The paper's own translation is 1.5 times. Odds ratios run larger than plain ratios of rates when the outcome is common. The failure rates are sample averages from these comparisons, not a general rate for lecture courses.
The authors also report that results held across STEM disciplines, were largest on concept inventories, and were largest in classes of 50 or fewer. Publication bias did not explain them, and the exam-score effect held across differences in how well studies controlled for student and instructor differences. The source is Freeman et al. (2014). The authors note that the failure-rate studies lacked randomized designs, so that result rests on comparisons of existing course sections.
Do newer studies reach the same answer?
Two later meta-analyses point the same way. One comes from a separate team. The other comes from Freeman's own group, so it is an update of earlier work and not independent confirmation.
| Study | Meta-analysis scope | Students | Result for assessment scores |
|---|---|---|---|
| Kozanitis & Nenciovici (2022), Higher Education | 104 studies (k = 111, N = 15,896) | College students in humanities and social sciences | 0.489 SD higher under active instruction; larger benefits in economics, sociology, psychology, language and education, and in smaller classes |
| Xu et al. (2026), Education Sciences | 134 studies of undergraduate STEM courses published 2010 to 2016 | Undergraduate STEM students | About half an SD (g = 0.519); mid-intensity (g = 0.583) and high-intensity (g = 0.593) courses beat low-intensity ones (g = 0.246) |
An effect of 0.489 SD is roughly the distance from the average student to one at the 69th percentile. The Xu analysis updates an earlier meta-analysis from the same research group, and Freeman is among its authors.
For a teacher, the intensity result is the useful part. Intensity was judged by how much class time the studies reported students being actively engaged in course activities. The authors caution that intensity may be tangled up with other aspects of course redesign, so read it as "more time on task goes with stronger results," not as a dial that causes them.
Why can active learning feel like it is working less?
In a Harvard physics study by Louis Deslauriers and colleagues (2019), 149 students in two introductory courses were randomly assigned to active instruction or to a lecture from an experienced, highly rated instructor, then switched for a second class meeting on a different topic. Content and handouts were identical. Students did better after the active lesson, yet rated their own learning lower than after the lecture.
The authors suggest that students read the extra mental effort as a sign they are learning less. The study measured ratings over two class meetings, and it did not measure whether students resisted active learning over a semester. For a high school teacher, the practical reading is that some students may rate the format lower early on, and that their ratings are not a good guide to what they learned.
Where this research stops
The age range comes first. Freeman's synthesis covers undergraduate STEM, and the two newer meta-analyses cover college students. None is a high school study. We do not review K-12 studies of inquiry or project-based teaching here; our post on the value of project-based learning takes up that angle.
Active learning is also a broad label. Freeman's team notes that the interventions varied widely in intensity and implementation, from occasional group problem-solving to studio or workshop course designs, and that they could not test how intensity or type related to results. Xu and colleagues later did, with the confounding caveat above.
Using active learning is not the same as using it well. In Andrews et al. (2011), researchers studied introductory biology courses at institutions randomly selected from a list of prominent colleges and universities (instructors then had to agree to take part). Instructors' reports of how much active learning they used, which students' own reports backed up, did not predict students' gains on natural selection. The study was correlational and covered one topic. The authors suggest that some activities may have asked students only to recall information rather than to reason.
Finally, the comparison group is traditional lecturing. These studies compare whole course designs, so they do not test whether a short explanation inside an active class helps or hurts.
What this means in a high school classroom
- Make students choose, act and defend. A prompt such as "pick one of these two options, then justify it with evidence from today's data" is active learning. A worksheet that asks for recall is not.
- Cap the talking. When you explain something, follow it with a task where students apply the idea, so that most of the period is students working.
- Tell students in advance that it will feel harder. The Harvard result suggests the effort gets misread, so say out loud that struggling with a problem is part of learning.
- Check learning with a quick, ungraded question at the end, not with how students say they feel.
- Start with one recurring routine, such as a daily decision question, which is easier to do well than a full redesign.
Frequently Asked
Questions
In the studies reviewed here, on average, yes. Freeman et al. found exam and concept-inventory scores 0.47 SD higher across 158 undergraduate STEM studies, and a 2022 meta-analysis found 0.489 SD in humanities and social sciences. The evidence is from college courses, not high school.
Freeman et al. found effects across class sizes, with the largest in classes of 50 or fewer. Kozanitis and Nenciovici also found larger benefits in smaller groups (20 students or fewer) in humanities and social sciences.
The strongest evidence is from undergraduate courses, so these studies give no direct answer. The principle is plausible for high school, and the practical advice is to try it and measure the result in your own class.
These studies do not say so. They compare traditional lecturing with courses where students spend class time on activities and discussion. A shorter explanation followed by students applying the idea fits that pattern.
Turning a finance unit into one where students make and defend decisions is worth talking through with our team: request a free Rapunzl demo to see the curriculum, the simulator and the educator tools.
By Clarissa Collins, Curriculum Designer at Rapunzl.
Get the free Rapunzl Starter Unit
Module 1, ready to teach: a slide deck, a quiz, and the answer key. Free.











