Schlatter Writing ConcepTests for a Multivariable Calculus Class

WRITING CONCEPTESTS
FOR A MULTIVARIABLE
CALCULUS CLASS
Mark D. Schlatter

ADDRESS : Department of Mathematics, Centenary College of Louisiana,

2911 Centenary Blvd., Shreveport LA 71104 USA . mschlat@centen-
ary. edu.

ABSTRACT: In a multivariable calculus course, students must master a

large number of concepts in order to successfully learn the material.
This paper will discuss one way of addressing this difficulty through
the use of ConcepTests, that is, multiple choice questions given in the
lecture that test understanding as opposed to calculation. In partic-
ular, we will look at various types of ConcepTests and the material
they can cover .

KEYWORDS : Multivariable calculus, ConcepTests, small group work .

INTRODUCTION
As I started preparing for Centenary's multivariable calculus course in the
fall of 2000, I was wondering how I could help student understanding. I
had taught the course the past two years and had incorporated MATLAB
programs to help the students visualize surfaces , curves, and vector fields.
Even with this help , a significant number of students had difficulties under-
standing multivariable and vector concepts, particularly in the latter half of
the course. Unfortunately, due to requirements on coverage of course ma-
terial (our syllabus covers almost all of the Harvard multivari able calculus
book [2]), I was not able to dramatically slow down the pace. I wanted to
find a way to gauge and improve student understanding and provide a solid
foundation in th e essential concepts.
Let me provide some background. First, our mul tivari abl e calculus class
is offered every fall with between 10 and 20 students (Centenary has 860

305
i'~iIilU) December 2002 Volume XII Number 4

undergraduates). These students include mathematics, physics, and engi-

neering majors as well as a few students from other disciplines. The first
time I taught the class (Fall 1998), we met about 20% of the time in a
computer lab equipped with MATLABj the second year (Fall 1999) we met
full-time in the lab. The class meets five days a week for 50 minutes a
day with a typical week consisting of four days of lecture and one day to
discuss homework. Students consistently had problems with the following
concepts: dot product, interpretations of the gradient, Lagrange multipli-
ers, integrating with cylindrical and spherical coordinates, and line and flux
integrals. While students could sometimes perform calculations with this
material, they often had troubles explaining the concepts or attacking un-
familiar problems .
After attending the Associated Colleges of the South's Teaching and
Learning Workshop and being introduced to ConcepTests, I decided to try
using them in my multivariable calculus course.

HISTORY OF CONCEPTESTS
ConcepTests were developed by Eric Mazur, a Harvard physics professor.
He had noticed that his introductory physics students could handle compu-
tational problems, but could not solve similar problems if the calculations
were removed and only the underlying concept assessed. In other words,
students appeared to be mistaking 'plug and chug' problem solving skills
for understanding.
To address this problem, Mazur started using ConcepTests - multiple
choice questions that students could answer in their heads if they correctly
understood the concepts. In lecture, a test would be presented, students
would vote on the correct answer and then break into small groups to discuss
their votes, and finally a revote would be taken. Students would therefore
engage the concepts in class through their votes and discussions with their
classmates. A typical lecture might consist of several tests, with less time
spent on examples. A longer discussion of the implementation and effec-
tiveness of ConcepTests can be found in [1] . In addition, Scott Pilzer in [3]
shows how these tests can be implemented in a first-year calculus course.

WRITING CONCEPTESTS
In my Fall 2000 multivariable calculus course, I used ConcepTests, with
about 20 minutes of each class period devoted to the tests. I will not be
discussing the specifics of implementing ConcepTests in the classroom - I

306
Schlatter Writing ConcepTests for a Multivariable Calculus Class

refer the reader to Scott Pilzer's article [2] for that information. Instead, I
want to focus on the types of tests I wrote for my class and the material
each type was best suited for. Roughly speaking, my tests fell into one of
five types: 1) visualization, 2) comparison, 3) translation, 4) theorem-using,
and 5) theorem-provoking.

Visualization Tests: These tests were designed to develop the students'

ability to think in three dimensions. One example is:
Example 1: The set of all points whose distance from the z-axis is 4
is the:
a) sphere of radius 4 centered on the z-axis
b) line parallel to the z-axis 4 units away from the origin
c) cylinder of radius 4 centered on the z-axis
d) plane z = 4

As is typical for ConcepTests, note that this problem is answerable

without symbolic calculation. In this example, I am seeing if students
clearly understand the freedom of three dimensional space. A common
answer on the first vote was b) - students saw some of the points that
were four units away from the z-axis, but not all of them. Another
common mistake was a) - students were ready to assume that any set
described as an equal distance from something was a sphere. In both
cases, the discussion between votes cleared up the misapprehensions
for most students.
I used visualization tests early in the semester, when the class focused
on the three-dimensional coordinate system and various cross-sections
of surfaces. Later on in the semester, I used this type of test to help
students visualize vector fields. Here is an example:
Example 2: Which of the following formulas will produce a vector
field where all vectors move away from the y axis?
a) F(x, y) = (x 3)i
b) F(x, y) = (x 2 )i
c) F(x, y) = (x 3)j
d) F(x, y) = (x 2 )j

In this case, I wanted students to be able to look at a vector field and

carry out some straightforward visualization. Students were quickly
able to rule out c) and d) based on direction, leaving most of the
discussion on the differences between a) and b).

307
 December 2002 Volume XII Number 4

Throughout the course, I introduced software to help the students

visualize shapes and fields. But before I showed students these tools, I
used visualization tests so that the students developed a solid intuition
and could use software to confirm their suspicions.
Comparison Tests: I used this type of test most often during the semes-
ter. These questions involved determining the sign of a quantity or its
relative magnitude. One example is:
Example 3: In which direction is the directional derivative of z =
x 2 + y2 at the point (2,3) most positive? (We are using i and j as the
unit vectors in the x and y directions.)
a)
b) -i - j
c) -i + j
d) i + j
I would use this test before giving the formula for computing direc-
tional derivatives with the dot product. Here I am asking the students
to combine several concepts: their visual picture of z = x 2 + y2 (a sur-
face we had looked at previously), their visual pictures of the four
vectors, and their understanding of the directional derivative. An ini-
tial vote on this test produced lots of guessing, but the discussion
period was very fruitful. Students would quickly review their mental
pictures of the surface and vectors and usually rule out b) and c) .
The discussion would then focus on how you might compare the di-
rectional derivative in the directions given by a) and d). Even if the
second vote was split between the two, the class was then primed to
look at a contour plot of the function and argue from there.
What I liked about these type of questions is that they pushed stu-
dents towards considering the different parameters that affect a scalar
quantity in multivariable calculus (e.g., the function and the direction
in the above example). Discussions between the votes were then es-
pecially helpful since teams of students could usually identify all the
necessary parameters. I used these types of tests extensively when
discussing vector arithmetic (including velocity and acceleration vec-
tors and dot product), partial and directional derivatives, the gradient
vector, and line and flux integrals.

Translation Tests: When I arrived at the chapter on integration, I was

not sure how to use ConcepTests. Of course, the concept of Rie-
mann sums over rectangles is important, but the bulk of the textbook

308
Schlatter Writing ConcepTests for a Multivari abl e Calculus Class

material covers techniques of int egrati on and int egration in different

coordinat e syst ems. Since we were using a compute r algebra syste m
in class, I decid ed to focus less on calcul ation and more on a spe-
cific pr oblem student s had in past semesters: translating a particular
int egral into a specific coordinate system. One exa mple is:
Example 4: Whi ch of the following is equival ent to

J 5 13 J ~ x dy dz dx?
- 5 0 - .j25-x 2

a) Jo" J03J~,2 cos( e) dz dr de

b) J; J~ J; ,2cos(e) dz dr de
c) J02" J; J~ r cos( e) dz dr de
d) J~" J; J~ r 2cos(e) dz dr de
Not e that this test does not focus on an underlying concept in the
sa me way t he above test s do. However , student s st ill benefited from
the voting and discussion phases. One definite adva ntage was that th e
class had to visua lize five different int egrals in two different coordina te
systems . Student s quickly picked up the differences between the limits
of th e four answers and what spaces they describ ed. Answers c) and
d) were both given to make sure students rememb ered to multiply by
a fact or of r .
In addition to th e sections on int egration, I wrot e Conc epTest s like
the below for parametric curves and surfaces and the parametrization
of line integrals. Here is an example:
Example 5: Whi ch of the following is equivalent to the line int egral
of F (x , y) on the line segment from (1, 1) to (3, 4)?
a) J01 F (1 + 2t , 1 + 3t ) dt
b) J01 F (1 + 2t , 1 + 3t ) . (2i + 3j) dt
c) J01 F (3, 4) . (2i + 3j ) dt
d) J01F (1 + t , 1 + t ) . (2i + 3j) dt
Again, the focus is not so much on an underlying concept as it is on
students recognizin g how to find the line integral of a parametrization.
Here, st udent discussion would focus on the differences between b) and
d).
Wi th tr an slation test s I found th ere to be a higher risk of having a
lar ge portion of the class making random guesses on the first vote .

309
 December 2002 Volume XII Number 4

These tests have more of an 'either you get it or you don't' quality
than the others I developed for the class. I still valued them , however,
because I was able to see where students were having problems during
the discussion time.

Theorem-Using Tests: With these types of tests, I assessed whether the

students knew how and when to apply a theorem. One example is:
Example 6: Which of the following facts about the vector field
F(r) = r (where r is a position vector) is implied by Stoke's The-
orem?
a) The line integral from (0,0,0) to (1,1,1) is equal to ~ .
b) F(r) = r has positive divergence everywhere.
c) The line integral on any closed curve is zero.
d) The curl of F(r) = r is non-zero.

I did not want the students to carry out a specific calculation using
Stoke's Theorem, but to understand its consequences. Here I am
looking for students to combine their mental picture of the vector
field (we had discussed it earlier in class) with their understanding
that Stoke's Theorem concerns the curl of a vector field. Once they
were able from their picture to see that the curl of F was zero, they
were able to move to c) as a correct answer. (One of the reasons d) is
included is to prompt students to move in that direction.)
I also used theorem-using tests extensively when we covered optimiza-
tion and the classification of critical points. In both cases, I was able
to help the students understand the power and limitations of the the-
orems we used .

Theorem-Provoking Tests: Occasionally throughout the course, I would

use ConcepTests not as a way of assessing student understanding of
covered material, but to ready them for new material. One example
is:
Example 7: The plot in Figure 1 shows the gradient vectors for
a (hidden) function f(x, y) and a linear constraint . Which point is
closest to the global min of f(x, y) on this constraint?
a) A
b) B
c) C
d) D

310
 Figure 1. Plot for Example 7.

My goal here was to help the students see that at a global minimum on
a constraint, the gradients of the objective and constraint functions
are parallel. I wanted the students to use their conception of what
the gradient tells them to prepare them for the method of Lagrange
multipliers.
I also used tests like this when we discussed normals to surfaces in
preparation for parametrizing flux integrals. For those sections which
were almost purely computational, this type of test gave me the op-
portunity to engage the students before we got to the symbolic ma-
nipulation.

I wrote a total of 89 ConcepTests for my multivariable calculus class,

covering almost all of the topics in the Harvard multivariable calculus book.
After using these in my Fall 2000 multivariable course, I refined the tests and
used them again in my Fall 2001 course. You can download the collection
at http://personal.centenary.edu/-mschlat/conceptests.pdf.

REACTION

When I started using the ConcepTests I discovered how much they enhanced
student feedback and my understanding of the students' abilities. The use
of ConcepTests resulted in an active class that was not afraid to ask ques-
tions or make comments. In my experience, the discussion period between
votes primes the students for further discussion in class by allowing them
to focus on a carefully defined question. I saw a greater range of student

311
 December 2002 Volum e XII Number 4

participat ion (as compa red to my ot her classes) when it came t ime for st u-
dent s to explain t heir votes . Indeed , some of the weaker students were the
most vocal participants.
I benefited most as a teacher from walking around t he class roo m during
t he vot e discussions. (Our computer lab has an ope n area where it is easy
to form gro ups and move between th em. ) I was able to hear st udents'
discussion and act as an advocate for different points of view. Quite ofte n ,
I would give a group a leadin g question based on their discussion. W hen we
took t he revote, I frequently discussed the reasonin g different gro ups had
used . All t his meant I was better ab le to focus t he material following the
ConcepTest - in fact , it was not uncom mon for me to change my lecture if
a ConcepTes t had proven too difficult or controversial. This contact wit h
t he groups also meant that very early in the semest er I was ab le to gauge
indiv idu al st udent abilities.
T wo pieces of evidence at the end of the Fall 2000 semester pointed to the
effect iveness of the tests. First, I received some of the best written student
evaluat ions in my career, with several st udent s specifically stat ing how the
ConcepTests had helped. Second, of the four times I have taught this course,
t he Fall 2000 semester class was the most successful in keeping st udent
interest. I had fewer students who stopp ed coming to class , stopped t urn ing
in homework , or had lar ge drops in exam scores than in ot her semesters.
W hen I t aught t he class in fall of 1998, I had 13 st ude nts initially enrolled,
12 who too k t he final , and 9 wit h a grade of C or above . In fall of 1999,
t here were 19 initially enro lled , 18 who too k th e final , and 14 wit h a C
or above. In t he fall of 2000 when I used the ConcepTests , t her e were 20
init ially enro lled, 20 who too k the final , and 18 wit h a C or above. These
classes are not directl y comparable - my exa ms in fall of 2000 did include
ConcepTests while pr evious classes did not - bu t I did notice fewer st udents
who 'gave up ' t hroughout the semester.
W hen I repeated t he use of ConcepTests in th e fall of 2001, I had 12
students initially enro lled with 9 who too k the final. All of them earn ed a
C or above. While there was a higher dro p ra t e (all t hree dr ops came in the
first half of the semester ), those who staye d in the class fully pa rt icipated
throughout the semester. There was, however, a problem wit h using Con-
cepTests with a class that small. T here were fewer viewp oint s expressed and
less chance of a correct answer percolating through the discussion period.
At t he sa me t ime , the benefits t o t he instructor were st ill pr esent, and the
discussion t ime was ofte n fruit ful for st ude nts .
Finally, one of t he most int eresti ng pieces of feedback I got was from a
st udent who reported t hat when he started st udyi ng for one of my exams,

312
Schlatter Writing ConcepTests for a Multivariable Calculus Class

the first thing he did was to go through all the ConcepTests. His reason
behind this was not only to prepare for the ConcepTests on the exam,
but because reviewing the tests helped him to go back in time to the day
he learned the material. Apparently, the ConcepTests provided him with
'mental landmarks' in the course.

FUTURE PLANS

My first use of ConcepTests in the Fall 2000 semester was primarily intended
as 3, proof of concept - would the tests work in a mathematics classroom and
would student understanding be improved? The answer to the first question
was a definitive yes. Given my experience with both the Fall 2000 and Fall
2001 classes, the answer to the second question is a qualified yes. While
I have not used a common instrument to compare student understanding
between those using ConcepTests and those not, the students who have
taken the ConcepTests show a greater comfort in talking about and using
the material on a conceptual basis. In addition, the use of ConcepTests
appears to prevent students from falling behind or losing interest in the
class. My future plans are to see how ConcepTests can be used in other
classes, especially our college algebra course.

ACKNOWLEDGEMENTS

I would like to thank the Associated Colleges of the South's Teaching and
Learning Workshop, where I was introduced to ConcepTests by Duane Pon-
tius from Birmingham-Southern College and was given the encouragement
to experiment with my teaching.

REFERENCES

1. Mazur, Eric. 1996. Peer Instruction: A User's Manual. New Jersey:

Prentice-Hall.
2. McCallum, William G., Deborah Hughes-Hallett, Andrew M. Gleason
et al. 1997. Multivariable Calculus. New York: John Wiley & Sons.
3. Pilzer, Scott. 2001. Peer Instruction in Physics and Mathematics.
PRIMUS. 11(2): 185-192.

313
 Decemb er 2002 Volume XII Number 4

BIOGRAPHICAL SKETCH
Mark D. Schlatter received his PhD in math em atics from the University of
Ca liforn ia at Berkeley. Originally specializing in mathematical logic with a
focus on mod el theory, he has since br an ched out to nonnegative matrix t he-
ory, the mathematics of art, and curriculum development. Aft er three yea rs
as a Visiting Assistant P rofessor at Truman State University in Kirksville
MO , he is now an Assist ant P rofessor at Cente nary College.

314