Transforming Physics Education

established by experiment, and they

By using the tools of physics in their teaching, instructors use systematic concept-based prob-
can move students from mindless memorization to lem-solving approaches that are ap-
understanding and appreciation. plicable to a wide variety of situations.
Most people (“novices”) see physics
more as isolated pieces of information
Carl Wieman and Katherine Perkins handed down by some authority and
unrelated to the real world. To novices,
he science community needs to change science educa- “learning” physics simply means memorization of infor-
T tion to make it effective and relevant for a much larger
fraction of the student population than in the past. This
mation and of problem-solving recipes that apply to highly
specific situations.2
need is the result of significant changes in the environment
and society over the past several decades. First, society now Research on traditional instruction
faces critical global-scale issues that are fundamentally We now examine how well traditional instruction does at
technical in nature—for example, climate change, genetic getting the average student to think like an expert. Tradi-
modification, and energy supply. Only a far more scientifi- tional science instruction is used in the overwhelming ma-
cally and technically literate citizenry can make wise deci- jority of college physics courses and has familiar character-
sions on such issues. Second, modern economies are so istics. Most of the class time involves the teacher lecturing
heavily based on technology that having a better under- to students; assignments are typically back-of-the-chapter-
standing of science and technology and better technical type homework problems with short quantitative answers,
problem-solving skills will enhance a person’s career aspi- and grades are largely based on exams containing similar
rations almost independent of occupation. Furthermore, a problems. Over the past couple of decades, physics educa-
modern economy can thrive only if it has a workforce with tion researchers have studied the effectiveness of such prac-
high-level technical understanding and skills. tices. (For reviews with useful citations, see references 3–5
As a community, we must now ask ourselves, “How and the article by Edward Redish and Richard Steinberg,
successfully are we educating all students in science?” This PHYSICS TODAY, January 1999, page 24). In this section, we
objective is very different from in the past, when the goal present representative examples of research on three quite
of science education was primarily to train only the tiny different but important aspects of learning: conceptual un-
fraction of the population that would become future scien- derstanding, transfer of information, and basic beliefs about
tists. The new, broader educational need does not elimi- physics.
nate the need to educate future generations of scientists. The first aspect of learning, conceptual understand-
However, improving science education for all students is ing, has been extensively studied3,4 and is particularly rel-
likely to produce more and better-educated scientists and evant because the great strength of physics is that a few
engineers. This claim is supported by data showing that fundamental concepts can explain a vast range of phe-
the fraction of students who complete a physical science nomena. Most studies have looked at students’ learning of
major in college is determined more by the students’ abil- basic physics concepts in traditional introductory physics
ity to tolerate traditional physical science instruction than courses. The results are remarkably consistent. We will
by their ability to do science.1 discuss two examples, one from mechanics and one from
For a variety of reasons, the physics community electricity.
should and can take the lead in providing an effective and Physics education researchers have developed several
relevant science education for all students. Moreover, this carefully constructed tests that explore student under-
is in their enlightened self-interest. A better-educated cit- standing of the basic concepts of force and motion. These
izenry would better appreciate the value of supporting tests have been administered at the beginning and end of
physics research. many, many courses across the country. The oldest and
But what specifically do we mean by effective physics best-known test is the Force Concepts Inventory (FCI).6
instruction? It is instruction that changes the way stu- Figure 1 shows a sample question from the FCI and re-
dents think about physics and physics problem solving and sults compiled by Richard Hake from data on 62 courses
causes them to think more like experts—practicing physi- (14 traditional).7 As shown in the figure, students receiv-
cists.2 Experts see the content of physics as a coherent ing traditional instruction master, on average, less than
structure of general concepts that describe nature and are 30% of the concepts that they did not already know at the
start of the class. The result is largely independent of lec-
Carl Wieman, corecipient of the Nobel Prize in Physics in 2001 turer quality, class size, or institution.
and the Carnegie–CASE US University Professor of the Year in Eric Mazur, a highly renowned teacher at Harvard
2004, is a Distinguished Professor of Physics at the University of University, has studied students’ understanding of con-
Colorado in Boulder. Katherine Perkins is an assistant professor cepts in electricity. Motivated by FCI results, Mazur gave
of physics attendant rank at the University of Colorado. his students an exam with a series of paired problems8 such

36 November 2005 Physics Today © 2005 American Institute of Physics, S-0031-9228-0511-020-5

 0.5 Figure 1. Students master rela-
FRACTION OF COURSES 1 (A) (B) (C) tively few concepts in physics
(D) courses using traditional in-
0.4
struction. The histogram
(E) shows, for the Force Concepts
2 Inventory (FCI), the average
0.3
normalized learning gain—that
is, the fraction of the concepts
0.2 that students learned that they
did not already know at the
0.1
start of the course. Results from
14 traditional courses are in
red, and results from 48
0.0 courses using a wide variety of
0.08

0.12

0.16

0.20

0.24

0.28

0.32

0.36

0.40

0.44

0.48

0.52

0.56

0.60

0.64

0.68

0.72
interactive-engagement tech-
niques are shown in green. Su-
FRACTION OF CONCEPTS LEARNED perimposed on the histogram
are data (blue arrows) from
two large lecture courses that
as those shown in figure 2. His and similar data show that nized, cognitive princi- use well-tested research-based
students are able to correctly answer traditional test ques- ples about how people practices.16 The inset shows a
tions and complete traditional courses without under- learn.10 figure accompanying a typical
standing the basic physics concepts or learning the useful Our third topic is FCI question: Students are
concept-based problem-solving approaches of physicists. research on students’ asked which path the ball will
We next examine a second aspect of learning, simple general beliefs about follow upon exiting the tube.
transfer of information and ideas from teacher to student physics and problem (Adapted from ref. 7.)
in a traditional physics lecture. The following example is solving in physics. Re-
from data collected in our own introductory physics class search groups including our own have studied these beliefs
for non-science majors. After explaining the physics of through extensive interviews and well-tested surveys.11
sound in our usual incredibly engaging and lucid fashion, These surveys measure where students’ thinking lies on
we brought a violin into class. We explained how, in ac- the expert–novice scale discussed above, and how their
cordance with the physics we had just explained, the views are changed by taking a physics course. The surveys
strings do not move enough air to create the sound from have now been given to many thousands of students at the
the violin. Rather, the strings cause the back of the violin beginning and end of introductory physics courses at many
to move via the soundpost, and thus it is the back of the different institutions. After instruction, students, on aver-
violin that actually produces the sound that is heard. Fif- age, are found to be less expert-like in their thinking than
teen minutes later, we asked the students the multiple before. They see physics as less connected to the real world,
choice question shown in figure 3, “The sound you hear less interesting, and more as something to be memorized
from a violin is produced mostly by . . .” As illustrated in without understanding. This is true in almost all courses,
the figure, only 10% gave the correct answer. We have seen including those with teaching practices that have sub-
that this 10% level of retention after 15 minutes is typical stantially improved conceptual mastery. If it is any conso-
for a nonobvious or counterintuitive fact that is presented lation to physics teachers, we have measured similar re-
in a lecture, even when the audience is primarily physics sults from introductory chemistry courses.
faculty and graduate students. The examples we have discussed are just a few from
When we have asked physics teachers to predict the a large body of research on the effectiveness of the tradi-
student responses to the violin question, nearly all of them tional approach to teaching physics. The definitive con-
greatly overestimate the fraction of students who answer clusion is that no matter how “good” the teacher, typical
correctly. Many physics faculty go so far as to simply re- students in a traditionally taught course are learning by
fuse to believe the data. For readers who may share their rote, memorizing facts and recipes for problem solving;
they are not gaining a true understanding. Equally unfor-
skepticism, we briefly mention two other studies. Redish
tunate is that in spite of the best efforts of teachers, typi-
had students interviewed just as they came out of his lec-
cal students are also learning that physics is boring and
ture.4 The interviewer simply asked the students, “What
irrelevant to understanding the world around them.
was the lecture about?” The students were unable to recall
anything beyond the general topic. In a more structured A better approach
study,9 Zdeslav Hrepic and coworkers gave 18 students six Is there a way to teach physics that does not produce such
elementary questions on the physics of sound. Immedi- dismal results for the typical student? Our answer, and
ately after attempting to answer the questions, the stu- that of many others doing research in physics education, is
dents were told that they were to get the answers to the unequivocally yes. Many of the same methods that have
six questions from watching a 14-minute commercially worked so well for advancing physics research also improve
produced videotaped presentation given by a nationally physics education. These methods include basing teaching
renowned physics lecturer. For most of the six questions, practices and principles on research and data rather than
no more than one student was able to learn the correct an- on tradition or anecdote; using new technology tools effec-
swer from the lecture, even under these highly optimized tively; and disseminating and copying proven results. Con-
conditions! siderable evidence shows that this approach works. Classes
When presented with these data, teachers often ask, using research-based teaching practices have shown dra-
“Does this mean that all lectures are bad?” The brief an- matic increases in retention of information, doubling of
swer is no, but to be effective, lectures must be carefully scores on the FCI and other conceptual tests, and elimina-
designed according to established, but not widely recog- tion of negative shifts in beliefs about physics.

http://www.physicstoday.org November 2005 Physics Today 37

 Figure 3. Counterintuitive facts are not retained by lec-
ture students. Fifteen minutes after being explicitly told
that it is the back of the violin that produces the sound,
students were given the boxed multiple-choice ques-
tion. The histogram of their responses shows that only
10% answered correctly.

load—the less effectively it can process anything12 (see fig-

ure 4). Any additional cognitive load, no matter what form
it takes, will limit people’s abilities to mentally process and
learn new ideas. This is one of the most well-established
and widely violated principles in education, including by
many education researchers in their presentations.
Figure 2. Paired problems compare students’ ability to Cognitive load has important implications for both
calculate quantitative answers with their conceptual un- classroom teaching and technical talks. To maximize
derstanding. (a) Students were asked, “For the circuit learning, instructors must minimize cognitive load by lim-
shown, calculate (a) the current in the 2-W resistor and iting the amount of material presented, having a clear or-
(b) the potential difference between points P and Q.” ganizational structure to the presentation, linking new
The average score of 69% on the question indicates material to ideas that the audience already knows, and
that most of them were able to calculate the currents avoiding unfamiliar technical terminology and interesting
and voltages in this moderately complex DC circuit. little digressions.
(b) Those same students performed much worse (aver- Expert competence5,12 is a primary goal of education
age score of 49%) when asked to explain what happens and is another area in which research has provided useful
qualitatively to, for instance, the brightness of these insights. Expert competence has been found to have
light bulbs and the current drawn from the battery roughly two parts: factual knowledge and an organiza-
when you close the switch S—questions that seem far tional structure that allows the expert to effectively re-
simpler to any physicist. The message is that students
trieve and apply those facts. Organizing physics ideas
can answer traditional test questions without really un-
derstanding basic physics concepts or mastering con- around general concepts is part of building such a struc-
cept-based problem-solving approaches. (Adapted from ture. If students do not have a suitable organizational
ref. 8, with permission of the publisher.) structure, simply pouring additional facts on them may ac-
tually deter learning.
To move a student toward expert competence, the in-
structor must focus on the development of the student’s
Research on learning has provided results that both mental organizational structure by addressing the “why”
explain many of the disappointing results of traditional in- and not just the “what” of the subject. These mental struc-
struction and provide guidance as to how to improve. We tures are a new element of a student’s thinking. As such,
they must be constructed on the foundation of students’
present three examples here, chosen in part because they
prior thinking and experience.5,12 This prior thinking may
are relatively easy to use throughout the standard cur-
be wrong or incorrectly applied, and hence must be ex-
riculum and classroom setting. Numerous other examples, plicitly examined and adequately addressed before further
including many about specific physics topics, are given in progress is possible. The physics education research liter-
references 3–5. ature can help instructors recognize and deal with partic-
Cognitive research shows that the amount of new ma- ular widespread and deeply ingrained misconceptions.3,4
terial presented in a typical class is far more than a typi- In summary, expert competence is likely to develop only if
cal person can process or learn. People’s brains function in the student is actively thinking and the instructor can
a way somewhat analogous to a personal computer with suitably monitor and guide that thinking.
very limited random-access memory. The more things the Our final example of useful research concerns stu-
brain is given to process at the same time—the cognitive dents’ beliefs. Students’ beliefs about physics and how it is

38 November 2005 Physics Today http://www.physicstoday.org

learned are important.1,11 They affect motivation, ap-
proaches to learning and problem solving, and, not sur-
prisingly, choice of major. As we noted earlier, teaching
practices influence students’ beliefs, usually by making
them more novice-like. Presenting mechanics in terms of
general concepts and the motion of abstract items such as
blocks on frictionless ramps can inadvertently teach many
students that these principles do not apply to real-world
objects. Assigning problems that are graded strictly on a
final number, or that can be done by plugging the correct
numbers into a given procedure or formula, can teach stu-
dents that solving physics problems is only about memo-
rization and coming up with a correct number—reasoning
and seeing if the answer makes sense are irrelevant. The
good news is that courses with rather modest changes to
explicitly address student beliefs have avoided the usual
negative shifts.11 Those changes include introducing the
physics ideas in terms of real-world situations or devices
with which the students are familiar; recasting homework
and exam problems into a form in which the answer is of
some obvious utility rather than an abstract number; and
making reasoning, sense-making, and reflecting explicit
parts of in-class activities, homework, and exams.
New educational technology
Utilizing principles established by educational research
can greatly improve physics education. Technology can
make it easier to incorporate these principles into in-
struction. For example, online surveys and student–
faculty e-mail are rather simple ways to enhance commu-
nication, thereby helping faculty understand and better
guide student thinking. Here we will discuss a couple of
more novel technologies—personal electronic response
systems and interactive simulations. These technologies
Figure 4. Students in lecture are apt to suffer from cognitive
are relatively simple and inexpensive, and we have found
overload.
them to be pedagogically powerful and easy to incorporate
into the standard curriculum.
A variety of commercial vendors sell personal elec- a series of about six clicker questions that cover the key
tronic response systems, or “clickers” as they’re usually learning goals for that day. Although multiple-choice ques-
known to our students. The various systems are all based tions may seem limiting, they can be surprisingly good at
on a similar idea. Each student owns a clicker and uses it generating the desired student engagement and guiding
to answer multiple-choice questions asked during class. A student thinking. They work particularly well if the pos-
computer records each student’s answer. After all the re- sible answers embody common confusions or difficult
sponses are in, the system displays the answers in a histo- ideas. Useful clicker questions and valuable guidance on
gram like that given in figure 3. Software grades the re- writing effective questions are now available.8,13
sponses and allows the instructor to later examine each It is important to actively encourage students to talk to
student’s answer. A clicker system for a classroom of about each other about the questions. We do this, sometimes after
200 seats requires several receivers, a computer, and a pro- they have answered individually, by requiring our groups to
jector; the total cost is about $5000. come to a consensus answer, enter it with their clickers, and
If used properly, clickers can have a profound impact be prepared to offer reasons for their choice. Those peer dis-
on the students’ educational experience. The value of the cussions are the times when most students are doing the
clicker is that it provides a way to quickly get an answer primary processing of new ideas and problem-solving ap-
for which the student is accountable, and that answer is proaches. Critiquing each other’s ideas to arrive at a con-
anonymous to the student’s peers. While the clickers pro- sensus answer also enormously improves their ability to
vide some measure of what students are thinking, it is the carry on scientific discourse. Finally, the discussion helps
specifics of the implementation—the change in the class- them to learn to evaluate and test their own understand-
room dynamic, the questions posed, and how they are fol- ing. Experts have the ability to monitor and test their own
lowed up—that determines the learning experience. These thinking on an ongoing basis by asking questions like “Does
specifics need to be guided by an understanding of how this make sense?” and “How can I test this?” However, it is
people learn. Instructors must also make sure their stu- very difficult for students to learn this skill without some
dents understand how and why the clickers are being used. amount of social interaction and feedback. The student dis-
If students perceive clickers merely as a way to give more cussions in our classes are inspired by the peer instruction
tests, rather than as a method to improve engagement and technique popularized by Mazur.8 The clickers and consen-
communication, the clickers will be resented. sus groups just provide a way to enhance the process, par-
We have found that the biggest impact of clickers ticularly for the less active or less assertive students.
comes when they are used with a combination of practices A major value of clickers is how they can enhance com-
that others have developed. We randomly assign students munication in the classroom. The sometimes painful feed-
to groups the first day of class (typically three or four stu- back provided to the instructor by histograms like figure 3
dents in adjacent seats). Each lecture is designed around is the most obvious. However, there are other, more valuable

http://www.physicstoday.org November 2005 Physics Today 39

 also shows what cannot normally be seen—
electrons that flow around the circuit with
their velocity proportional to current, im-
mediately responding to any changes in
circuit parameters. Our studies14 have
found this simulation helps students un-
derstand the basic concepts of electric cur-
rent and voltage and, when substituted for
an equivalent lab with real components,
improves how well students can build and
explain real-life circuits.
Many physicists find it quite mysteri-
ous and somewhat disturbing that care-
fully developed simulations are more edu-
cationally effective than real hardware.
Both the efficacy of simulations and the
physicists’ discomfort can be understood by
recognizing the difference between how the
beginning student and the expert instruc-
tor perceive the same situation. These per-
ceptual differences are readily apparent in
our testing of simulations and in other re-
search on the effectiveness of lecture
demonstrations.15
Figure 5. The circuit construction kit is a simulation A real-life demonstration or lab includes enormous
that allows students to build virtual circuits containing amounts of peripheral information that the expert in-
a number of different elements. A pedagogically useful structor filters out without even thinking about it. The stu-
feature of the simulation is that it displays motion of the dent has not learned what can be filtered out, and so all
electrons, shown here in blue. Students who work with this other information produces confusion and a much
the interactive simulation are better able to understand heavier cognitive load. The student’s attention is often on
and build real-life circuits. The simulation and many things the instructor doesn’t even notice, because they are
others are available at the Physics Education Technol- irrelevant. For example, in a real circuits lab, inexperi-
ogy website, ref. 14. enced students will often spend considerable time and con-
cern on the significance of the different colors of the plas-
tic insulation on the wires.
A carefully designed computer simulation can main-
forms of feedback. By circulating through the classroom and
tain connections to real life but make the student’s per-
literally listening in on the consensus-group discussions, the
ception of what is happening match those of experts. This
instructor can quickly learn particular points of student un-
is done by enhancing certain features, hiding others, ad-
derstanding and confusion. Then in the follow-up lecture or justing time scales, and so on, until the desired student
whole-class discussion, the instructor can directly target perception is achieved. Simulations also can provide visual
those specific items of confusion. Perhaps even more impor- representations that explicitly show the models that ex-
tant than the feedback to the instructor is the feedback pro- perts use to understand phenomena that are not directly
vided to the students through the histograms and peer dis- visible, such as the motion of electrons. It is likely that
cussions. Students become much more invested in their own both features are important in explaining the observed
learning. One manifestation of this change is that we now benefits of simulations.
receive many more substantive questions, and they are The educational importance of recognizing and deal-
asked by a much broader distribution of students; 10–15 ing with differences between student and expert thinking
questions per class period is typical. goes well beyond the use of simulations. An apt metaphor
Clickers can also be useful in other ways. For in- is that of the student and the expert instructor separated
stance, we use them to quickly survey the range of stu- by the mental equivalent of a canyon; the function of teach-
dent backgrounds and to quiz students at the start of class ing is to guide the student along the path that leads safely
to check that they’ve done the assigned reading. Ensur- and effectively across the canyon to the nirvana of expert-
ing background reading considerably facilitates useful in- like thinking. Guidance that ignores the student’s starting
class discussions. point or that is interpreted differently than intended usu-
ally just sends the student over a cliff. But education re-
The reality of virtual physics search, careful measurement, and new technology make it
Interactive simulations that run through a regular Web possible to guide most students safely along the path to-
browser can be highly effective. Using an existing simula- ward a true understanding and appreciation of physics.
tion also often takes less preparation time than more tra-
ditional materials. Our research group has created and We are pleased to acknowledge the valuable input from all the
studied the effectiveness of about 45 simulations.14 We members of the University of Colorado at Boulder physics ed-
have explored their use in lectures, as part of homework ucation research group.
problems, and as laboratory replacements or enhance-
ments. Figure 5 shows our circuit construction kit simu- References
lation. This simulation allows one to build arbitrary cir- 1. E. Seymour, N. Hewitt, Talking About Leaving: Why Under-
cuits involving lifelike resistors, light bulbs, wires, graduates Leave the Sciences, Westview Press, Boulder, CO
batteries, and switches; measure voltages and currents (1997); K. Perkins et al., http://www.colorado.edu/physics/
with realistic meters; and see light bulbs lighting up. It EducationIssues/PERC2005Formatted_submitted.pdf.

40 November 2005 Physics Today http://www.physicstoday.org

 2. D. Hammer, Cogn. Instr. 15, 485 (1997).
3. L. McDermott, E. Redish, Am. J. Phys. 67, 755 (1999).
4. E. Redish, Teaching Physics with the Physics Suite, Wiley,
Hoboken, NJ (2003).
5. J. Bransford, A. Brown, R. Cocking, eds., How People Learn:
Brain, Mind, Experience, and School, National Academy
Press, Washington, DC (2000).
6. D. Hestenes, M. Wells, G. Swackhamer, Phys. Teach. 30, 141
(1992); see ref. 4 for a compilation of other useful concept
surveys.
7. R. Hake, Am. J. Phys. 66, 64 (1998).
8. E. Mazur, Peer Instruction: A User’s Manual, Prentice Hall,
Upper Saddle River, NJ (1997).
9. Z. Hrepic, D. Zollman, N. Rebello, in Proc. 2003 Physics Ed-
ucation Research Conference, J. Marx, K. Cummings, S.
Franklin, eds., American Institute of Physics, Melville, NY
(2004), p. 189.
10. D. Schwartz, J. Bransford, Cogn. Instr. 16, 475 (1998).
11. See the discussion of the Maryland Physics Expectations
(MPEX) survey in ref. 4, and the articles by W. Adams et al.
and by K. Perkins et al. in Proc. 2004 Physics Education Re-
search Conference, J. Marx, P. Heron, S. Franklin, eds.,
American Institute of Physics, Melville, NY (2005), p. 45 and
p. 61, respectively.
12. R. Mayer, Learning and Instruction, Merrill, Upper Saddle
River, NJ (2003).
13. I. Beatty et al., http://umperg.physics.umass.edu/library/
umperg-2005-01/entirePaper.
14. Those simulations, along with copies of research articles, are
available at the Physics Education Technology website
http://phet.colorado.edu.
15. C. Crouch et al., Am. J. Phys. 72, 835 (2004); W.-M. Roth et
al., J. Res. Sci. Teach. 34, 509 (1997).
16. C. Crouch, E. Mazur, Am. J. Phys. 69, 970 (2001); S. Pollock,
in Proc. 2004 Physics Education Research Conference, J.
Marx, P. Heron, S. Franklin, eds., American Institute of
Physics, Melville, NY (2005), p. 137. 䊏

See www.pt.ims.ca/6088-17 See www.pt.ims.ca/6088-31