On this page
Most of us learned physics and maths by listening to a teacher explain at the board, then copying the solution. Is that the way students learn the most? In 2014 a team led by Scott Freeman tried to answer this with every study they could find. This post walks through what they did, the exact numbers they reported, what the study cannot tell you, and how to use it as a student or teacher in India.
What question did the study ask?
The paper, Active learning increases student performance in science, engineering, and mathematics, published in PNAS, set out to test one idea: that lecturing gives the most learning. The authors compared courses taught by "continuous exposition by the teacher" with courses that included at least some active learning, which they defined as engaging students "through activities and/or discussion in class, as opposed to passively listening to an expert".
Active learning in the studies covered a wide range: occasional group problem-solving, worksheets done in class, clicker questions with or without peer discussion, and fully redesigned studio courses. Its share of class time ranged from 10% to 100%.
Who and what was studied?
This was a meta-analysis, a study of studies. According to the full text on PubMed Central, the team hand-searched every issue of 55 STEM education journals from 1998 to 2010, searched seven databases and followed reference lists. They coded 642 papers and kept 225 studies that met fixed rules: the same course taught both ways, class time differing by no more than 30 minutes a week, and comparable exams.
The courses were undergraduate courses in subjects including physics, chemistry, biology, engineering, computer science, mathematics and statistics. Two outcomes were pooled:
Scores on identical or equivalent exams and concept inventories (158 studies).
Failure rates (67 studies).
What did it find?
The headline numbers, as reported in the paper:
Exam and concept-test scores rose by 0.47 standard deviations under active learning. The authors estimate this as about 6% on average exam scores.
Failure rates averaged 21.8% under active learning and 33.8% under traditional lecturing. Students in lecture courses were 1.5 times more likely to fail.
The benefit held across all STEM disciplines, for majors and non-majors, and for introductory and advanced courses.
Gains were larger on concept inventories than on instructor-written exams, and largest in classes of 50 or fewer, though medium and large classes also benefited.
Randomised studies (16 of them) gave an effect of 0.51, close to the overall figure, so weaker study designs were not inflating the result.
You can check the second line yourself: 33.8 minus 21.8 is 12 percentage points, and 33.8 divided by 21.8 is about 1.55, which is the "55% increase" in failure the paper describes. The authors also worked out that among the 29,300 students in the lecture groups, a 12-point lower failure rate would mean about 3,516 fewer failures (29,300 times 0.12).
What does it not show?
The authors list their own cautions, and they matter:
The teachers were volunteers. People who choose to try active learning may be unusually committed. The paper says it is "an open question" whether results would be as large if every teacher were required to do it.
It mixes many methods. "Active learning" here covers clickers, worksheets and full redesigns. The study tells you that moving away from pure lecture helps on average, not which method helps most.
It is about college courses. Every study was an undergraduate course. School classrooms, coaching batches and exam preparation in India were not studied.
It measures course exams, not entrance exams. No study here looked at JEE, NEET-UG or board results.
A later PNAS meta-analysis by Theobald and colleagues (2020) added an important detail: only high-intensity active learning narrowed achievement gaps for underrepresented students. A few minutes of discussion in an otherwise lecture-only class may not be enough.
What does this mean for students and teachers in India?
Most JEE and NEET-UG preparation is still lecture-heavy, and many school classes are too. You cannot always change how you are taught, but you can make your own study time active. That means doing the thinking yourself rather than watching it done:
Predict before you are told. When a teacher starts a derivation or a solved example, write what you think the next step is. Being wrong first and then corrected is part of how this works.
Use concept questions. The biggest gains in the meta-analysis were on concept inventories, which test understanding rather than formula recall. The Force Concept Inventory story shows the kind of questions involved.
Discuss, then check. In a study group, each person answers a conceptual question alone, then argues it out, as in Eric Mazur's peer instruction.
Teachers can start small. One worksheet or two concept questions per class is active learning by the paper's definition, though the Theobald result suggests more is better.
When you are working alone and get stuck on what to do next, the Fermi Problem Library has worked problems with a step-by-step solver that explains the reasoning rather than printing an answer, so you can attempt each step before you look.
How to read a study like this
When a result is given in "standard deviations", turn it into percentiles. An effect of 0.5 standard deviations moves an average student to roughly the 69th percentile; 0.2 moves them to about the 58th. Then ask who was studied (here, college students taught by volunteer teachers) before you assume it applies to your class.
Next step: in your next lecture, write down a prediction before every worked example, and see how many you get right by the end of the week.
