On this page
In 1990, Eric Mazur had been teaching introductory physics at Harvard for seven years. His lectures were polished, his student evaluations were high, and his students solved complicated problems. Then he tested whether they understood force, and found that his success "was a complete illusion, a house of cards," as he told Harvard Magazine.
What he built next, peer instruction, became a widely studied change to the physics lecture. Here is what happened, what the evidence shows, and how you can use the same idea in a study group or on your own.
What made Mazur change his lectures?
The trigger was a simple conceptual test on force by Arizona State professor David Hestenes, written in everyday language. Students who could recite Newton's third law and use it in numerical problems still said, for example, that a heavy truck exerts a larger force than a light car in a collision.
When Mazur gave the test to his own class, one student asked: "How should I answer these questions—according to what you taught me, or how I usually think about these things?" His students "did well on textbook-style problems," he said, but had not grasped the basic ideas, Harvard Magazine reported.
The breakthrough came by accident. Mazur tried twice to explain one question and the class stayed confused. "Then I did something I had never done in my teaching career," he recalled. "I said, 'Why don't you discuss it with each other?'" Within three minutes, the students had worked it out. His explanation: a fellow student "is more likely to reach them than Professor Mazur," because, as he puts it, "You're a student and you've only recently learned this, so you still know where you got hung up."
How does peer instruction work?
Crouch and Mazur (2001), in the American Journal of Physics, describe the cycle:
The teacher gives a short presentation on one central point.
A conceptual question follows, called a ConcepTest.
Students get one or two minutes to answer on their own and report it.
Students discuss with neighbours for two to four minutes, trying to convince each other with reasons.
Students answer again; the teacher explains and moves on.
What did ten years of data show?
Crouch and Mazur's paper reports a decade of Harvard introductory physics courses. On the Force Concept Inventory, the average normalized gain (the share of possible improvement students actually made) was 0.25 in 1990 with traditional lectures and 0.49 in 1991, the first year of peer instruction. It kept rising in later years.
Problem-solving did not suffer. Average scores on the Mechanics Baseline Test rose from 66% in 1990 to 72% in 1991 and 79% in 1997. When the 1985 final exam, made entirely of quantitative problems, was given again in 1991, the mean rose from 63% to 69%.
Discussion moved students toward the right answer. Across all ConcepTests in Fall 1997, nearly half of the correct answers were reached after discussion, and students switched from correct to incorrect only 6% of the time.
A fair caveat: these are comparisons across years at one university, with improvements added along the way, not a randomized trial. The pattern matches larger surveys, such as Richard Hake's 1998 study of 6,542 students, which found average gains of 0.23 in traditional courses and 0.48 in interactive-engagement courses. Our post on Hake's study has the details.
How to use peer instruction without a teacher
Two to four friends preparing for JEE, NEET or board exams can run it:
Choose conceptual questions, not plug-in ones. "A ball is thrown straight up. What is its acceleration at the top?" works better than "find the maximum height."
Vote first, silently. Everyone writes an answer and a one-line reason before anyone speaks, so the strongest student doesn't decide for the group.
Then argue for three minutes. Reasons only: "the book says so" doesn't count.
Vote again, then check. Look up the answer only after the second vote.
Adjust difficulty. If everyone gets it right alone, pick harder questions; if nobody does, review the concept first.
Studying alone? Write your answer and reason, then argue for a different answer as strongly as you can before checking. For group structure and roles, see how to run a study group that helps.
Where Fermi fits
Peer instruction works because you must commit to an answer and defend it. If you study alone, Fermi's AI tutor works through problems with you, decides whether to explain, ask or show, and checks understanding as it goes. If you want a teacher and a class for JEE Main and Advanced, Fermi Testprep pairs an AI tutor with live classes and adaptive mocks.
Try one ConcepTest round with a friend this week and count how many answers change.
