On this page
Have you ever left a class thinking "that was so clear, I understood everything", then failed to solve a single problem at home? A 2019 experiment in Harvard's introductory physics course measured exactly this gap. It is one of the cleanest studies of why good learning often does not feel good. Here is what it did, what it found and how to use it.
What question did the study ask?
Louis Deslauriers, Logan McCarty, Kelly Miller, Kristina Callaghan and Greg Kestin published Measuring actual learning versus feeling of learning in response to being actively engaged in the classroom in PNAS. Earlier research, including Deslauriers' own 2011 Science study with Carl Wieman, had shown that active classes produce more learning. Yet many students say they prefer lectures. The team asked: if you switch active learning on and off, with everything else identical, how do students' feelings about their learning compare with what they actually learned?
Who was studied, and how?
According to the full paper, the study ran in two semesters of a calculus-based introductory mechanics course at Harvard, with 65 and 92 students enrolled. Of the 157, 149 completed the study.
The design was a crossover:
In week 12, students were randomly split into two groups in two rooms.
One class covered static equilibrium, the next covered fluids.
Group A got the active version of statics and the lecture version of fluids; group B got the reverse.
Both versions used identical slides, handouts and worked solutions. In the lecture version, the instructor solved the examples fluently while students filled in answers. In the active version, students worked the same problems in small groups first, then got the full solution.
At the end of each class, students rated their feeling of learning on a five-point scale (for example, "I feel like I learned a great deal from this class") and took a 12-question multiple-choice test written by an author who had not seen the class materials.
Because each student experienced both styles, each one acted as their own comparison.
What did it find?
Actual learning: students scored 0.46 standard deviations higher on the tests after active classes.
Feeling of learning: after active classes they rated their learning 0.56 standard deviations lower.
The pattern held for both topics and both instructors.
Students who saw the instructor as fluent and effective reported higher feelings of learning, whichever style was used. Fluency drove the feeling, not the amount learned.
Students with stronger prior physics knowledge had feelings that tracked their real learning more closely.
In follow-up interviews with 17 students, 15 described the active classes as disjointed and lacking flow, and 14 mentioned frustration and confusion. All but one found the lecture more enjoyable and easier to follow. After seeing the results, 14 of the 17 said it would change how they study.
The team also tried a fix. In a later semester, the instructor gave a 20-minute talk early in the course explaining this result. By the end, over 65% of students said their feelings about active learning had improved, and 75% said the early talk helped.
What does the study not show?
It was short. Each comparison was a single 90-minute class, tested at the end of the same class. It does not measure long-term retention.
The students were unusual. These were Harvard students with strong preparation. The authors argue less-prepared students may misjudge their learning even more, but that was not tested here.
The active class was not silent work. About half of the active class time was still the instructor giving short explanations after each group activity. It was a mix, not pure self-study.
Feelings were self-reported on a few survey items.
What does it mean for students in India?
Coaching classes and video lectures in India are often extremely fluent. A teacher who solves a JEE problem in three clean steps can leave you feeling you could do it too. This study says that feeling is not good evidence. Some practical rules:
Judge your learning by a test, not a feeling. After a lecture, close the notes and solve one problem cold. Our post on judgments of learning explains why your confidence is a poor guide.
Try before you watch. Attempt a problem for ten minutes before watching the solution video. The confusion you feel is part of the process.
Expect active study to feel worse. If group study or self-testing feels slower and messier than rereading, that matches what this experiment found. See why rereading creates an illusion of fluency.
Teachers can explain the effect. The 20-minute talk in this study helped students accept active methods. A short explanation at the start of term may do the same.
When you do get stuck during an active session, Fermi's AI tutor notices where a student is stuck and decides whether to explain, ask or show, so the struggle stays productive rather than turning into a dead end.
How to read a study like this
Look for what was held constant. Here, the slides, handouts, worked solutions, topic and instructor were all balanced, and only the teaching style changed. That makes the result much stronger than studies comparing two different teachers or two different years. Then check how outcomes were measured: an end-of-class test is a short-term measure, so the claim is about learning that day, not months later.
Next step: after your next lecture or video, write a one-line rating of how well you understood it, then solve one problem without notes and compare the two.
