Contributed Article  June 2009
Author: Dr. Eric Mazur
Chairman of Turning Technologies' Instructional Strategies Advisory Committee

I have been teaching an introductory physics course for engineering and science majors at Harvard University since 1984. Until 1990, I taught a conventional course consisting of lectures enlivened by demonstrations. My students did well on what I considered difficult problems and the evaluations I received from them were very positive. I thought that I was an effective teacher until I tested my students conceptual understanding. The results were shocking.
On a midterm examination in the spring of 1991, I decided to pair conceptual qualitative questions with traditional quantitative problems on the same concept shown in Figure 1. The first question is purely conceptual and requires only a knowledge of the fundamentals of simple circuits. Question 5 requires setting up and solving two equations. Figure 2 shows the lack of correlation between the scores on the conceptual and conventional problems in Figure 1.

Figure 1 Conceptual (top) and conventional question (bottom) on the subject of dc circuits. These questions were given on a written examination in 1991.

Although 52% of the scores lie on the broad diagonal band, indicating that these students achieved roughly equal scores on both questions (±3 points), 39% of the
students did substantially worse on the conceptual question. Conversely, only 9% of the students did worse on the conventional quantitative question. It became apparent that many students were simply memorizing algorithms without understanding the underlying physics.
After this experience I explored new approaches to teaching. In particular, I looked for ways to focus attention on the underlying concepts without sacrificing students’ ability to solve problems. The result is Peer Instruction, a method that teaches conventional underpinnings and leads to better student performance on conventional problems.
Conventional lectures simply reinforce students’ feelings that the most important step in mastering the material is memorizing formulas. It is difficult to provide adequate opportunity for students to critically think through the arguments being developed. The traditional lecture is nearly always delivered in a monotone directly from the textbook. It provides little incentive to attend class.

Figure 2 Correlation between conceptual and conventional problem scores from Figure 1. The radius of each datapoint is a measure of the number of students represented by that point.

Peer Instruction uses the textbook and the lecture in different roles. Preclass reading assignments from the book first introduce the material. Next, lectures elaborate on the reading, address potential difficulties, deepen understanding, build confidence, and add additional examples. Finally, the book serves as a reference and a study guide.
The basic goals of Peer Instruction are to exploit student interaction during lectures. Instead of presenting the level of detail covered in the textbook or lecture notes, lectures consist of a number of short presentations on key points, each followed by a ConcepTestshort conceptual questions on the subject being discussed. The students are first given time to formulate answers and then asked to discuss their answers with each other. This process (a) forces the students to think through the arguments being developed, and (b) provides them (as well as the teacher) with a way to assess their understanding of the concept.
The lecture generally follows this format:
 I lecture on a subject for 710 minutes emphasizing the concepts and the ideas behind the proof while avoiding equations and derivations.
 I read a conceptual question to the students, making sure there are no misunderstandings about it.
 I tell them that they have one minute to answer.
 Students record individual answers. While answers can be tabulated through the show of hands or scanning forms, there are unique advantages to using student response systems such as TurningPoint by Turning Technologies. Student responses are anonymous, encouraging honest participation. Students do not have the opportunity to mime the responses of over achieving students. Such a system also provides instructors with immediate accurate data analysis and allows for numerical responses in addition to multiple choice responses.
 I view the results from the poll without displaying the results to the students. If more than 30% and less than 70% responded correctly, I give the students one to two minutes to convince their neighbors of the correct answer. I always participate with a few groups of students in the animated discussions that follow. If more than 70% of the students get the question correct, I give a brief explanation an move on to the next concept. If less than 30% of the students get the question correct, I revisit the concept and repoll.
 After student discussion I repoll the question and show the results of both polls.
 I finish by giving a two minute explanation of the correct answer.
 I move on to the next concept and repeat the process. Typically, I will cover three to four keypoints in an hourlong lecture.
The convinceyourneighbor discussions systematically increase both the percentage of correct answers and the confidence of the students. Figure 3 illustrates the improvement in correct responses and confidence for all questions during a given semester. Close to one third of the class correctly revised their initially incorrect answer, while only 3% changed from correct to incorrect. The overall results show significant learning gains.

Figure 3 How answers were revised after convinceyourneighbors discussion.

In this new lecturing format, the ConcepTests take about one third of each lecture period, leaving less time for straight lecturing. I do not cover all the material covered in the textbook and lecture notes. I require students to read the textbook and my lecture notes before coming to class.