See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/229573998

Students' mathematical modeling of motion

Article in Journal of Research in Science Teaching · February 2008

DOI: 10.1002/tea.20210

CITATIONS READS
23 1,215

2 authors:

Jill Marshall David Carrejo

University of Texas at Austin University of Texas at El Paso
83 PUBLICATIONS 1,499 CITATIONS 12 PUBLICATIONS 115 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by David Carrejo on 10 October 2017.

The user has requested enhancement of the downloaded file.

JOURNAL OF RESEARCH IN SCIENCE TEACHING VOL. 45, NO. 2, PP. 153–173 (2008)

Students’ Mathematical Modeling of Motion

Jill A. Marshall,1 David J. Carrejo2

1
The University of Texas, Science and Mathematics Education, 1 University Station,
D5705, Austin, Texas 78712-0382
2
University of Texas at El Paso, Department of Teacher Education EDUC 601,
500 West University Avenue, El Paso, Texas 79968

Received 27 October 2006; Accepted 5 March 2007

Abstract: We present results of an investigation of university students’ development of mathematical

models of motion in a physical science course for preservice teachers and graduate students in science and
mathematics education. Although some students were familiar with the standard concepts of position,
velocity, and acceleration from physics classes, most students had difficulty using these concepts to
characterize actual or hypothetical motions. Furthermore, some students developed their own nonstandard
method of describing accelerated motion in terms of changes in the average velocity, from the start of the
motion up to a given time. This is in contrast to the physics community’s use of the acceleration construct,
defined in terms of changes in the instantaneous velocity, to describe such motion. Although the change in
average velocity is not typically identified as an important construct in traditional physics texts, some
students found it intuitively appealing, and were able to use it successfully to describe and predict motion.
We conclude that by focusing on standard constructs, and ignoring possible intuitive ways that students
might view motion, standard kinematics instruction may miss an opportunity to maximize student
understanding. ß 2007 Wiley Periodicals, Inc. J Res Sci Teach 45: 153–173, 2008
Keywords: physical science; inquiry; college/university; classroom research

Most students come to the physics classroom with a wealth of experience in motion, both in
moving themselves and observing other objects as they move. Still, these opportunities to acquire
commonsense knowledge about motion do not always seem to help students in mastering the
standard concepts that the physics community uses to describe motion. Students are reported
to have difficulty, for example, distinguishing between position, velocity, and acceleration
(Hestenes, Wells, & Swackhamer, 1992; Trowbridge & McDermott, 1980), interpreting negative
velocities (Goldberg & Anderson, 1989), interpreting graphs of position and velocity versus
time (McDermott, Rosenquist, & van Zee, 1987; Nemirovsky & Rubin, 1992), understanding of
the vector nature of velocity (Shaffer & McDermott, 2005), and using velocity information to

Correspondence to: J.A. Marshall; E-mail: [email protected]

DOI 10.1002/tea.20210
Published online 6 August 2007 in Wiley InterScience (www.interscience.wiley.com).

ß 2007 Wiley Periodicals, Inc.

154 MARSHALL AND CARREJO

predict positions (Shternberg & Yerushalmy, 2003). The research overwhelmingly points toward
deficiencies in traditional lecture/lab instruction in helping students to develop a robust under-
standing. Other work has focused on more effective ways to help students develop a coherent
qualitative understanding of these standard conceptions (e.g., Nemirovsky & Rubin, 1992;
Rosenquist and McDermott, 1987).
There has been less emphasis on the intuitive approaches students might take toward
describing and predicting motion when the standard constructs are not explicitly referenced. In
one study that focused on the approaches students would take outside of an instructional setting,
Bowden, Dall’Alba, Martin, Laurillard, Marton, Masters, et al. (1992) asked university and high
school physics students to make predictions about the time for objects to travel certain distances
and the distances that would be traveled in certain amounts of time in situations involving moving
frames of reference. The questions did not mention velocity specifically, and some students
invoked other constructs, such as force or power, to solve the problems. Although this sometimes
resulted in correct predictions about the motion, the responses were judged as inadequate because
they did not contain all of the elements of an expert understanding as defined by the physics
community. The authors point out the importance of uncovering the different ways in which
students understand physical situations, but do not consider the value of these approaches in their
own right or as part of a learning trajectory toward the canonical understanding.
In mathematics education, one program that encourages students’ intuitive approaches
to representing velocity is the Dutch approach of Realistic Mathematics Education (RME)
(Gravemeijer & Doorman, 1999). Although standard mathematical formulations are the ultimate
goal, the RME approach is expected to offer students opportunities ‘‘to develop informal, highly
context-specific solution strategies . . . [that] may function as foothold inventions, or as catalysts
for curtailment, formalization or generalization (p. 117).’’
In physics education, the modeling method also elicits students’ intuitive approaches to
describing and predicting physical phenomena. This method is rooted in constructivism and
challenges students to develop their own models (Hestenes, 1992; Wells, Hestenes, & Swack-
hamer, 1995). In the case of motion, ‘‘each student must literally reinvent the Newtonian World
in his/her own mind to understand it’’ (Hestenes, 1992, p. 733). It does not do simply to present the
standard concepts of kinematics unless learners can make sense of them in terms of what they
already understand. Therefore, it is important to elicit learners’ existing ideas, particularly with
regard to a subject like motion, with which they have had a lifetime of personal experience. These
personal approaches will not necessarily mirror those of the established scientific community, and
in fact, may even be at variance with standard thinking. They are, however, the ‘‘foothold inven-
tions’’ of Gravemeijer and Doorman (1999), the necessary first steps students make in con-
structing an understanding of the standard models, either by building on these original ideas or
replacing them.
Hestenes (1992) stresses the importance of presenting the ‘‘rules of the modeling game’’ (the
axioms of Newtonian theory) explicitly to students, before they are asked to use these rules in a
particular context, that is, to develop and apply models. These axioms, if applied correctly,
constrain the models which can be developed. The goal is for students to operate within a
particular framework, that is, the Newtonian one, rather than to focus on their intuitive
approaches. However, it should also be pointed out that Hestenes (1992) argues that students are
not ready to understand and apply the Newtonian framework until after they have struggled and
come to grips with motion. Students must first relate their intuitive understanding of length
and time to the construct of position (relative to a reference frame) and intuitive understandings of
motion to the notion of a trajectory. The full Newtonian framework is not presented until after the
unit on kinematics, which is the subject of our study.
Journal of Research in Science Teaching. DOI 10.1002/tea
 STUDENTS’ MATHEMATICAL MODELING OF MOTION 155

Lesh and Doerr (2003) present a perspective on modeling that perhaps best aligns with that
which informed instruction in the course that formed the basis for our study, as well as the research
itself. This approach has much in common with that of Hestenes (1992), but focuses on
the ‘‘sharable, manipulatable, modifiable, and reusable conceptual tools (e.g., models) for
constructing, describing, explaining, manipulating, predicting or controlling mathematically
significant systems’’ developed by students as they address personally meaningful problems (Lesh
& Doerr, 2003, p. 3). In contrast to the foothold inventions of Gravemeijer and Doorman (1999),
these are highlighted as ‘‘mathematically significant sense making systems’’ in their own right
(Lesh & Doerr, 2003, p. 5), as opposed to simply necessary but insufficient first steps. Thus, we
allowed for the possibility that our students might develop and use models that are not typically
presented in physics courses, and that might even involve constructs that are not part of the
Newtonian framework as presented by Hestenes.
Although it is possible to generate many models that follow the rules of Newtonian theory,
students are generally presented only one or two and rarely allowed to generate them for
themselves. In introductory physics, the most commonly presented model for motion is:

xi ¼ x0 þ v0 ti þ 1=2 ati2 ð1Þ

where x0 is the position at the initial time, xi is the position at the time ti, v0 is the velocity at the
initial time, and a is the acceleration, which is assumed to be constant in this model. This model
also makes the assumption that the initial time, t0, is set equal to zero. In Equation 1 and all of what
follows, x, v, and a are actually vector quantities, and therefore should be represented in vector
notation. Introductory textbooks, however, generally focus first on the one-dimensional case and
use the magnitudes of these quantities, treating them essentially as scalars, in developing Equation
1 (e.g., Giancoli, 1998, p. 27). Equation 1 is typically justified based on a simpler one involving the
commonsense idea of average velocity:

v ¼ distance traveled=time of travel ¼ ðxi
x0 Þ=ðti
t0 Þ ð2Þ

Students are often familiar with the first equality in Equation 2 from their experiences with
motion. Even prior to instruction, students are often familiar enough with car travel to be able to
calculate an average speed given a distance traveled and a time interval of travel. The second
equality in Equation 2, however, necessitates a shift in perspective, from ‘‘distance traveled’’
(a length) to xi
x0 (a difference in positions), that may not be straightforward for all students.
Considerable mathematical manipulation is necessary to move from the commonsense idea of
average velocity to Equation 1, and for many students the mathematics is not compelling
conceptually. This leaves open the possibility that they leave physics coursework without a viable
model for motion that has conceptual significance for them. This is particularly problematic for
students who will go on to become science teachers at the K–12 level, where motion is likely to be
part of the integrated science or physical science curriculum, but the mathematics of Equation 1
may not be appropriate.
In contrast, the class we studied is designed for future and former teachers to take a deeper
look at the curriculum in the way that they might allow their own students to explore it. It was not a
traditional physics class. Therefore, we did not feel constrained to arrive at Equation 1. Still, we
considered it essential that the students leave with models that could be justified (and that they
themselves could justify) on the basis of physical evidence or experience. That is, it was our goal
that students eventually develop models that did not violate the Newtonian framework nor give
inaccurate predictions about motion.
Journal of Research in Science Teaching. DOI 10.1002/tea
156 MARSHALL AND CARREJO

Method

Investigational Approach
Following the modeling paradigm described above (Lesh & Doerr, 2003), we chose to
investigate how students would proceed directly from experiencing or observing a motion, to
gathering data about the motion, to generating a descriptive model from which predictions could
be made. Although we were open to the possibility that useful qualitative models might be
generated, the nature of the predictions we asked students to make (e.g., where an object would
be at a certain time) tended to evoke quantitative (mathematical) models. An understanding of the
steps students would take, in the absence of constraining guidance, toward mathematizing motion
as they encounter it in their everyday lives is needed to design appropriate instructional sequences,
in particular, to identify tasks that will elicit the development of target concepts. Such an under-
standing is particularly important in the case of students who will become, or already are, teachers.
The National Science Education Standards specifically call for K–12 teachers to engage students
in coming to their own understanding of physical phenomena and developing mathematical
models directly from data (NRC, 1996, p. 214). Precollege teachers need to explore ways to allow
their students to develop their intuitive understanding rather than simply presenting them with the
standard mathematical models that may have comprised the teachers’ formal experience with
motion in their science coursework. Finally, precollege teachers’ own understanding of motion,
and their ability to relate the mathematical formalisms of physics to real-world objects and events,
will constrain their ability to provide the requisite opportunities and guidance for their students
(McDermott, 1990). We therefore focused on a class intended for preservice teachers and for
graduate students who had been teachers. With the goal of investigating these students’ intuitive
approaches to describing and predicting motion, we took the following as research questions for
the study described here:

1. What constructs (foothold inventions) do students (preservice and former teachers)

develop in analyzing motion; how do these relate to the standard kinematical concepts of
average velocity, instantaneous velocity, and acceleration, and how useful are these
constructs in predicting motion?
2. How does the task students are given interact with the constructs they employ? In
particular, do these constructs differ when students create and observe a motion for
themselves, as opposed to when they are given data from a hypothetical motion?
3. How does previous formal instruction in physical science interact with the process of
students analyzing and making predictions about motion?

The second question was motivated by the National Science Education Standard’s
recommendation that students develop mathematical models of data they have gathered
themselves, as well as the emphasis in the research literature on establishing a connection between
real-world phenomena and mathematical abstractions. Our goal was to inform the design of
instructional tasks meeting this recommendation. The third question arose from previous work
with in-service physics and physical science teachers. In that work, familiarity with established
results from kinematics seemed to circumvent mathematical exploration of data (Carrejo, 2004).
Teachers who remembered the standard algebraic formalisms for motion from previous
coursework were generally not able to generate those relations from their own descriptions
of motions they had observed, and seemed unwilling or unable to pursue their own intuitive
models.
Journal of Research in Science Teaching. DOI 10.1002/tea
 STUDENTS’ MATHEMATICAL MODELING OF MOTION 157

Setting
Our study took place in a physical science course designed for preservice teachers and science
education graduate students. The course was a hands-on, inquiry-oriented learning experience
largely based on the Physics by Inquiry materials (McDermott, 1996). The motion unit involved a
series of activities developed by the first author and piloted in a previous study (Carrejo, 2004).
The first author was also the instructor for the course. The class met for 3 hours each week in a
room with lab tables. Students worked in groups of two to three at their own pace, with regular full-
class discussions. Groups were self-selected and varied from task to task, but generally were
diverse in terms of their backgrounds in teaching and content preparation. The unit on motion
comprised 4 weeks of the curriculum in the course.

Participants
The population consisted of 15 students enrolled in the course: 5 graduate students and
10 undergraduates. Informed consent was obtained from all participants. The five graduate
students all had previous teaching experience. Three had taught at the precollege level (one in
physics, one in chemistry, and one in math) and two at the college level (one as an instructor in
physical science and one as a TA in physics). Of the 10 undergraduates, seven were in teaching
certification programs (two in secondary science, three in elementary education, and two in youth
and community programs). Although the course was designed for preservice teachers and
graduate students in science and mathematics education, enrollment was not restricted, and three
of the students were not currently in certification programs and had no previous teaching
experience. One of these was a biology major, one a government major, and one a mathematics
major. Table 1 lists all the students involved in the study by pseudonyms and gives their degree
program and previous or planned teaching experience.

Instructional Sequence
The instructional sequence for the kinematics unit was a series of model-eliciting activities
(Lesh & Doerr, 2003). The sequence as planned consisted of three overarching activities. First,

Table 1
Study participants
Student Degree Program Teaching Experience/Plans
Joan PhD in Science & Math Ed Mathematics
Helen PhD in Science & Math Ed University physical science
John PhD in Science & Math Ed University physics (TA)
Mary PhD in Science & Math Ed High school chemistry
Dave Masters in Science & Math Ed High school physics
Cecelia Biology (Composite Science Certification) Preservice secondary
Steve Chemistry (Composite Science Certification) Preservice secondary
Veronica Education—Youth & Community Preservice youth and community educator
Olivia Education—Bilingual Preservice elementary
Linda Education—General Preservice elementary
Aaron Education—Youth & Community Preservice youth and community educator
Courtney Education—Bilingual Preservice elementary
Paul Mathematics —
Jimmy Biology —
Lee Liberal Arts—Government —

Journal of Research in Science Teaching. DOI 10.1002/tea

158 MARSHALL AND CARREJO

students would be asked to work in groups to produce and describe any motion that could be
executed by a small wooden block. The description needed to be precise and accurate enough so
that others could reproduce the motion exactly from the description alone. We intentionally made
no reference to standard constructs (such as position or velocity) that students might have
encountered in previous instruction so as to elicit their intuitive ideas about what is important in
describing motion.
The second activity as planned was to create a motion with constant velocity, prove that the
velocity was indeed constant, and predict the position of the moving object at an arbitrary time.
The third activity as planned was to create a motion with constant acceleration, demonstrate that
the acceleration was indeed constant, and predict the position of the moving object at an
arbitrary time. The second and third tasks as actually given to the students were influenced by the
outcomes and discussions of the previous tasks. The tasks as implemented are described in the
Results section. After each task the class met as a whole so that each group could present and
justify their results, and the entire class could reflect on their validity as well as their relation to
the results of others. This provided opportunities for the ‘‘postmortem analysis’’ described
by Hestenes (1992, p. 746) to allow students to reflect on their work and that of others. After
the second and third tasks, students were presented with hypothetical data sets to analyze based on
their individual approaches to describing and predicting motion as they had been refined in the
group discussion, and the results of these analyses were also presented to the entire class and
discussed.
The activities were supplemented with a task interview that served as a summative assessment
for the motion unit as well as items on the final exam for the course. In the task interview, students
were asked to describe the motion of a ball moving on a ramp and to fill in missing data from a set of
position versus time data for which the velocity was changing in a consistent fashion. They were
asked to use a ‘‘think-aloud’’ protocol as they worked the two problems. The final exam problems
consisted of position versus time data sets for one-dimensional motion with both constant
and changing velocity. Students were asked to supply missing position data and provide their
reasoning.
Data Collection and Analysis
Groups were videotaped on a rotating basis during all of the activities. Full class discussions
were also videotaped. The task interviews at the end of the unit were audio taped. Student artifacts
were collected, including lab notebooks, worksheets, exams, and presentation materials that
individual groups used in explaining and justifying their thinking to the entire class. The audio and
videotapes were reviewed and portions that related to the development or presentation of a
construct that students were using to explain or predict motion were catalogued and in some
cases transcribed. Although it is not possible to determine the conceptual models in students’
minds explicitly, these written, drawn, and spoken artifacts give us the best indication of what
quantities students think are important in analyzing motion, what mathematical relationships
relate these quantities, and how one might use them to make predictions about motion. These are
the ‘‘visible components’’ of the students’ models (Doerr & Lesh, 2003, p. 9).
We then analyzed the student artifacts against the standard framework used in physics to
describe and predict motion, that is, the basic principles of Newtonian mechanics. This framework
subsumes kinematics, the study of motion, for example, as described in Hestenes (1992). It
acknowledges the constructs of position, (instantaneous) velocity, and the change in instantaneous
velocity, or acceleration, as key concepts. We chose to evaluate the constructs that seemed to be
important to students against these standard ones, not simply to rate them as being in accordance
with the standard ideas, but to see how the students’ ideas related to those used by physicists, how
Journal of Research in Science Teaching. DOI 10.1002/tea
 STUDENTS’ MATHEMATICAL MODELING OF MOTION 159

they might or might not lead to the standard constructs, and whether they might provide useful
alternative ways to think about motion in their own right.

Results

Task 1: Describe a Motion

For their initial task in the kinematics unit, students worked in groups of two or three to study a
motion of their choice executed by a small wooden block and describe the motion itself (as
opposed to the actions that created it) accurately and precisely enough that others could recreate it.
As expected, no group responded to this task with all three of the constructs from Hestenes’ (1992)
‘‘Zeroth’’ Law (the portion of his framework governing kinematics): position, reference frame,
and trajectory. This law states that ‘‘Each particle l has a definite position xi with respect to a given
reference frame’’ and ‘‘Motion of the particle is to be represented by a trajectory, xi(t)’’ (emphasis
original, p. 734).
Only one group, Lee, a history major, and Linda, a preservice elementary teacher, explicitly
associated a time with the motion, but they were not explicit about trajectory information as a
function of time. In discussing how they would describe their motion, Lee spontaneously
wondered whether another motion would have to have ‘‘the same . . . like, time period’’ in order for
it to be considered a replication of theirs. At first, Linda argued that it did not need to because,
‘‘we haven’t learned about velocity yet.’’ Lee retorted that he had learned about it in high school.
The final version of their description read, ‘‘The block slides on one of its faces a total distance of
70 centimeters in a ten second time span.’’ In the following transcript, they argue that only the
total time of travel is relevant.

Linda: We decided that the time span didn’t matter as long as it happened within that time
span.
Lee: It got from point A to point B ..
Linda: Yeah, as long as it got from point A to point B in that ten seconds.
Lee: and it traveled 70 cm in 10 seconds. . .
JM: Ten seconds later it’s at point B.
Linda: And if it went fast in some parts and slow in some parts, as long as in that
10 second time span it traveled that distance, it’s all. . .
JM: It’s all the same?
Linda: Yeah.

Likewise, Lee and Linda felt that only the total distance traveled, rather than an exact
trajectory, was an important part of describing a motion. At first, Dave, a former high school
physics teacher, stated that he was ‘‘not a huge fan’’ of Lee and Linda’s description. He elaborates
why in the following transcript.

Dave: So, if my motion was to go across the room . . . like, my starting point would be this
wall and my other point would be that wall and I’ve got 30 seconds to do it, I could
either go straight for 30 seconds and at a constant speed or zigzag all around this
room really quickly and you’re saying that’s exactly the same motion?
Lee: That’s incorrect, though. ‘Cause we specified a certain distance that you could
travel. If you started zigzagging you’d travel a greater distance instead of going
70 centimeters.

Journal of Research in Science Teaching. DOI 10.1002/tea

160 MARSHALL AND CARREJO

JM: So I’m just going to sort of make notes. So, we’ve got one argument that says
you’ve got the end points and the time—that’s enough. And, then, [Dave], how
would you describe your objection to that?
Dave: I think it’s something . . . it’s a hard, I don’t know, it’s hard to do in a little sentence
like that. For it to be the motion exactly repeated it’s gotta have the same speed for
the same time through the same path.

Here, Dave is essentially expressing the idea that Lee and Linda’s description of motion is
missing the trajectory component of Hestenes’ Zeroth Law. However, after some continued
argument, he backed away from this stance, as seen in the following transcript.

Dave: That’s not a bad idea. Now I’m starting to . . . that’s not . . . like, the fact that you still
traveled 70 centimeters, the same amount of time . . . Maybe that is the same motion.

Some groups did focus on a position or positions, as important elements in their original
descriptions of motion, but were generally not explicit about a reference frame for these positions.
For example, Steve, a future chemistry teacher, and Veronica, a preservice elementary teacher,
said that the ‘‘Block moved upwards almost vertically in the air until a height of 60 to 70 cm and
then it stopped and fell back to the starting point.’’ Dave’s group came closest to situating their
motion in a reference frame (the ‘‘surface’’) in their description: ‘‘Starting at rest, the base of a
block is one inch above a surface. And the face is parallel to the surface. The block falls
perpendicular to the surface until its first contact with the surface and any further motion after first
contact is irrelevant.’’
The idea of the block traveling ‘‘vertically’’ and ‘‘perpendicular to the surface’’ embodies the
notion of a trajectory, but for some group members, it seemed that this trajectory was simply there
to indicate a distance traveled. For example, Aaron, a preservice youth and community educator
from Dave’s group, argued that one could replicate the motion they had described anywhere in the
room, because the meter stick was only there to ensure that the block traveled 1 inch. His real focus
was on the distance traveled rather than the path.
Further, none of the groups who mentioned positions referred to time in their descriptions in
any way. For some, the trajectory may have been important, but it was clearly not a function
of time. Dave’s argument above, that to replicate a motion it would be necessary for the object to
‘‘have the same speed for the same time’’ came only after hearing Lee and Linda’s description.
After the group discussion, there was a general consensus that time, at least total time of travel, was
important. The students with the least physics training, Lee and Linda among them, convinced
some with more training, for example, Dave who had taught high school physics, that a distance
traveled and a time interval might indeed be the salient features of motion.

Task 2: Describe and Predict a ‘‘Constant Motion’’

The basic constructs of distance traveled and time of travel continued to arise throughout the
course, indicating that these were indeed meaningful and intuitive for some of students. In the
general discussion of the first task, students agreed that it would be easier to describe motion if
the object moved ‘‘at a steady pace’’ because ‘‘the more things you keep constant throughout the
situation, the less there is to look at.’’ This then became the second task: to create and describe a
‘‘constant’’ motion. Students were given latitude to decide exactly what this meant for themselves,
although in later discussion students agreed that in constant motion the object in question would
always cover the same distance in the same (arbitrary) amount of time. All the groups attempted to
Journal of Research in Science Teaching. DOI 10.1002/tea
 STUDENTS’ MATHEMATICAL MODELING OF MOTION 161

create either a one-dimensional motion with constant velocity or a periodic motion with constant
speed. Afterward they were given hypothetical data describing one-dimensional motion to ana-
lyze using the procedures they had developed.
Two groups, of two students each, decided to work together to characterize the motion of a
bowling ball rolling down the hall outside the classroom. They constructed a short ramp and used it
to launch the ball rolling down the hall and described the consequent motion. One student released
the ball from the ramp and another was posted at the end of the hall to stop it. The two remaining
members stationed themselves together at a given distance from the end of the ramp and timed how
long it took the ball to reach them. The times the two observers recorded were averaged. The timers
increased their distance from the ramp with each successive roll.
In describing the motion, these two groups calculated an average speed for each roll by
dividing the distance the ball traveled in that run by the time interval that it took the ball to roll from
the bottom of the ramp to the observers. These speeds should have been equal had the velocity of
the ball been constant, but there was a clear trend toward lower average speeds as the distances
became longer. This group treated the average speeds as measurements at different times within
the same motion, and averaged them together. The process is described in the following transcript
segment. (Steve is a preservice secondary science teacher.)

Steve: We had six times for different distances and so I took each distance divided by
the time given from each of these guys. That gave me a speed and I did that for all
the data. So, I had six times for six distances and I averaged those together . . . I
averaged all the speeds together and I got an average speed of 155 cm per second.
So what I did to test to see if we could use that to predict distance, I have a
theoretical time, a theoretical distance, so after 1 second it should be 155
centimeters. . .
JM: How do you justify that? How did you arrive at this theoretical time?
Steve: If we . . . if we’re able to use our average speed as a constant speed, we should be
able to use our average speed to predict the distance after so many seconds. . .

During their presentation to the class, John, with Bachelors and Masters degrees in physics,
illustrated their procedure qualitatively (see Figure 1). What they called the speed for each roll was
the slope of the line segment from the origin to the point (ti,xi). They then averaged the six speeds
they had calculated together as if they were independent, calculated for sequential intervals rather
than overlapping intervals. Thus, they were associating an average velocity (speed) over the
interval from 0 to ti with the single clock reading ti, rather than the entire time interval. The fact that

Figure 1. Graphical representation of the students’ velocity construct. The image on the left is a still frame
of the illustration John drew on the chalkboard during a whole class discussion to illustrate his group’s method
of finding a characteristic velocity. The image on the right is a word-processor representation of his drawing.

Journal of Research in Science Teaching. DOI 10.1002/tea

162 MARSHALL AND CARREJO

their data consisted of times of travel paired with distances from the end of the ramp, each obtained
on a separate roll, may have steered this group toward thinking in this way. In previous work,
in-service teachers employed this same process to describe a motion, by coincidence also that of a
bowling ball rolling down a hall (Carrejo, 2004).

Task 3: Describe and Predict a Changing Motion

The third task was to describe and predict a motion that was not constant. Again, motion that
was ‘‘not constant’’ was defined by class consensus as motion in which the object did not cover the
same distance in equal periods of time. Although students were free to select any such motion, all
the groups ultimately ended up studying the motion of a cart moving down a ramp for repeatability
and ease of data acquisition. A strip of paper was attached to the cart and threaded through a
ticker timer, a device that struck the paper at fixed time intervals, leaving a carbon mark. Thus,
students had a permanent visual record corresponding to the location of the cart at a series of times.
After they analyzed the motion of the cart, students were asked to make predictions about hypo-
thetical motions from sample data sets with changing velocities.
Students with more formal physics experience focused on the construct of instantaneous
velocity and acceleration (change in instantaneous velocity per unit time), on which the model of
motion presented in Equation 1 is based. This is the model most commonly presented in intro-
ductory college or university physics classes. Students ran into trouble, however, when they tried
to create a model of their own using these constructs.
Helen, a university physical science instructor, and Olivia, who had recently taken a physics
course, analyzed their ticker timer data using an Excel spreadsheet to calculate a change in
position and divide by the elapsed time between each set of tick marks to create an average velocity
for each successive interval. Although these were average velocities, they were averages over
the smallest time intervals available to them, and Helen and Olivia essentially treated them as
instantaneous velocities, the limit for infinitely small intervals.
They took the difference between these successive velocities and divided by the time interval
again to calculate an average acceleration for each pair of successive intervals.
a ¼ ðviþ1
vi Þ=ðtiþ1
ti Þ ð3Þ

These showed considerable variation at the start and end of the motion, but oscillated
around a constant value in the central portion, so they then decided that they would take one
velocity from an interval at the start of the consistent motion and one from the end, take the
difference, and divide by the time between the two to determine an average acceleration for that
portion of the motion, again treating the velocities for small intervals as proxies for instantaneous
velocities. They then created a mathematical model d ¼ at2 using the spreadsheet to predict the
position, labeled d, at any time t, given their average acceleration, a. Helen justified the process as
follows.

Helen: We also took an average acceleration like you did, knowing that our value was
gonna be off because it was rather erratic in the data collection. And I’m going to
say, OK, so this is the acceleration in meters per second squared [points to the first
column in her Excel display, which is 0.76 m/s2 all the way down]. So if I take
the time squared times that value [the average acceleration] it should predict where
I’m at at any given point of time, right?
JM: How do you know that?

Journal of Research in Science Teaching. DOI 10.1002/tea

 STUDENTS’ MATHEMATICAL MODELING OF MOTION 163

Helen: [Long pause] Acceleration is a change in velocity over a change of time, and
velocity is position, change in position, over some time increment so if I rearrange
that to solve it for position I’m gonna get initial velocity plus acceleration times
time squared.

Here, Helen speaks of two different constructs recognized in the standard physics model.
When she says, ‘‘Acceleration is a change in velocity,’’ she is referring to the instantaneous
velocity. In the case of the cart rolling down the ramp, in which the initial velocity is zero, the
instantaneous velocity, v(ti), is indeed the acceleration, a, times the time, ti, for which the position
is being predicted:

vðti Þ ¼ ati þ v0 ¼ ati ð4Þ

On the other hand, the change in position over a time interval (xi
x0) is the average velocity,

v, over that time interval, times the time interval, ti
t0. Assuming that the initial position and time
are zero, Equation 2 gives:

xi ðti Þ ¼
vti : ð5Þ

In treating v(ti), the instantaneous velocity at a particular clock reading ti, as
v (associated with
a time interval), Helen and Olivia’s equation overpredicts the position by a factor of 2. The
discrepancy with their data was not obvious to them because their predictions were for much
longer times than their data. Thus, they were unable to use the instantaneous velocity construct,
v(ti) successfully.
Another group of students used the instantaneous velocity construct to find an average
acceleration for the motion successfully, but they then used the acceleration in a three-step
approach to finding the position at an arbitrary time, rather than using it directly in a model like
Equation 1. Steve and John, working with Jimmy, averaged a series of successive, nearly
instantaneous, velocities for small intervals to calculate an average acceleration over the entire
motion, much as Helen and Olivia did at first. They then used this average acceleration to predict
an instantaneous velocity at an arbitrary time, ti (Equation 4). At this point they averaged v(ti) and
the initial velocity (which, again, was zero for the cart on the ramp),
v (Dt) ¼ 1/2 (0 þ vf), to
calculate an average velocity for the entire time interval from 0 up to ti. They then successfully
used Equation 5 to predict the position, xi(ti), at the time, ti.
Although they did not at first discuss it, the poster that they used in presenting their work
showed an equation that mirrored Equation 1, in addition to the approach using the average
velocity. When other students asked about this equation, Steve, a chemistry major who had
recently completed a university physics course, went through a laborious process of combining
Equations 4 and 5 with the assumption that
v (Dt) is just the average of the initial velocity and the
final velocity vf, that is,
v (Dt) ¼ 1/2 (v0 þ vf), to show that that the result is algebraically equivalent
to Equation 1. This resulted in more confusion for the other students. John, who held a BS and an
MS in physics and was seeking a PhD in Science Education, then attempted to derive Equation 1
for the simplified case of motion starting from rest, but part way through admitted, ‘‘I’m confused
now.’’ Lee, with no college physics coursework, then responded, ‘‘If you’re confused, we’re laying
the ditch.’’
Note that Steve and John’s method parallels the justification found in most physics texts for
developing Equation 1. Dave, who had taught physics at the high school level, had not been able to
develop a method for predicting position based on his own data, but clearly recognized Equation 1
and reverted to using it as soon as John’s group introduced it. In presenting his approach to
Journal of Research in Science Teaching. DOI 10.1002/tea
164 MARSHALL AND CARREJO

analyzing a hypothetical set of position versus time data later, he invoked Equation 1 and described
the effect of his previous instruction in telling fashion:

Dave: And again, it’s this confusing equation that people don’t like. But it’s what I was
brought up on and I can’t run away from it. I don’t know. . .
Linda: You were brought up on that, eh?
Dave: Yeah, it was like my [pacifier].
Linda: Here’s your bottle and here’s your equation.

In contrast, some students independently invented a construct that is not generally recognized
in physics instruction as they were analyzing the hypothetical data sets in the follow-on activity to
the cart on the ramp. These students were able to generate a procedure that not only made sense to
them, but gave accurate predictions. At each clock reading, ti, they calculated what they called the
velocity at that time by taking the distance traveled since the starting time and dividing it by
the time interval from 0 to ti. Thus, they were actually calculating an average velocity,
v, over
the time interval from 0 to ti (as per Equation 2), but they were associating it with the single clock
reading, ti. Examining the data from 0 to t1, 0 to t2, 0 to t3, and so on, they found a pattern of regular
increases in these (overlapping) average velocities and were able to use that pattern to predict the
average velocity at any given time. Multiplying the average velocity between 0 and ti by the time
interval from 0 up to ti gave them the distance traveled up to that time and an accurate position if
they adjusted the data so that the position at the starting time was zero (as per Equation 5).
The approach was first introduced to the class when Lee presented his analysis of the data
set shown in Figure 2. The hand-written numbers are Lee’s annotations and the first two typed
columns are the data as originally given. Students were asked to predict where the cart would be at
7 seconds. The artifact indicates that Lee at first attempted a successive interval approach, filling in
the third column with the change in position for each, successive, 1-second interval, as directed by
the label provided for the column. He then divided this local change in position by the total time of
travel for the first two entries, yielding 1.5 and 2.25, which are written by the changes in position
in the first two data rows of the third column, and then discarded this approach for one that
apparently had more meaning for him. As shown in the fourth column, he began to use the
overlapping average velocity, from 0 up to each time for which he made a calculation. In explai-
ning his work to the class, Lee described his procedure as follows.

Lee: OK. OK. I’ll just, I’m just going to go over how I did it. OK. From the positions, I
found the, found the velocity at each second, for . . . given the position. So at
1 second, my position is 1.5 cm, and so to find the velocity, that’s 1.5 over t, which
is time, which is 1, so that’s 1.5. So that was my first velocity I found and then I just
went through and did each one. It was at 6 cm at 2 seconds, so that’s a velocity of 3,
and 4.5, 6, 7.5 [writes these numbers in a column on the board]
JM: And those are all cm per second?
Lee: Yes. Nine. And then . . . And so, and after looking at this I’m getting an increasing
velocity of 1.5 cm a second for every second. So, from there, to predict the
seventh second, all I did was take my time, which is 7, multiply that by my 1.5, and
that gives me, what, 10.5, and then I took that, which is my velocity at the
seventh second, and I multiplied that by my time, which is 7, and that gave me 73.5.

Lee first finds the change in overlapping average velocity. Then, paralleling Equation 4 with
the initial (overlapping average) velocity set equal to 0, he calculates the overlapping average
velocity at 7 seconds by multiplying the change in overlapping average velocity of ‘‘1.5 cm
Journal of Research in Science Teaching. DOI 10.1002/tea
 STUDENTS’ MATHEMATICAL MODELING OF MOTION 165

a second for every second’’ (analogous to the acceleration, a, in Equation 4) by the time
(7 seconds), to predict an average velocity ‘‘at’’ 7 seconds of 10.5 [cm/s].

v ¼ 1:5½cm a second for every second  7 seconds ¼ 10:5½cm=s ð6Þ

Then, following Equation 5, he multiplies the ‘‘velocity at the seventh second’’ by the time
interval (7 seconds) to predict that the cart will be at a position of 73.5 cm at the 7-second clock
reading

xi ðti Þ ¼
vti ¼ 10:5 ½cm=s  7 seconds ¼ 73:5 ½cm ð7Þ

Note that Lee has done more than simply finding a pattern in the given data. To see the pattern
of an increase of ‘‘1.5 cm a second for every second,’’ it was necessary for him to construct the
series of what he calls velocities first. These velocities were, in fact, the total distance traveled
divided by the total time interval of travel up to each point in time, the same construct Lee and
Linda had originally chosen to describe motion. The worksheet lists ‘‘position,’’ rather than
distance, and Lee uses the term position. The two are numerically equal, however, for this data set,
because the cart is at a position of 0 cm when the time is 0 seconds. It is likely that Lee was actually
thinking of the number in terms of a distance (length) rather than a position (point), corresponding
to their intuitive view of how motion should be described. He does not appear to be engaging in the
commonly cited misconception of simply using the position in place of the distance traveled (e.g.,
Kirkpatrick & Francis, 2004, p. 21).
Lee and Linda and other novice students appreciated the fact that the number given as
‘‘position’’ might not be the actual distance traveled in the time interval represented by the
given time, that is, the object might have had a nonzero starting position. In prior work they had
been careful to ‘‘back it back to zero’’ as she described it, to find the position at zero seconds, and
make an adjustment so that the number associated with each time was actually the distance
traveled between 0 and ti in cases where the data indicated a nonzero starting position. It should be
noted, however, that in the case where Lee first reported employing the change in overlapping
velocities, that is, the exercise shown in Figure 2, the starting position would have indeed been
0 cm as predicted by both Lee’s and the standard mathematical models.

Figure 2. Worksheet artifact showing Lee’s approach to predicting the position at 7 seconds for the object
whose position and time data are given. The fourth (unlabeled) column shows his calculation of the
overlapping average velocity.

Journal of Research in Science Teaching. DOI 10.1002/tea

166 MARSHALL AND CARREJO

On the other hand, Lee identifies the average velocity with a specific point in time, not with
a time period. He refers to the ‘‘velocity at each second’’ and ‘‘my velocity at the seventh second’’
[emphasis added]. Some confusion then arose as the more advanced students had calculated the
standard acceleration construct typically used in physics instruction (change in instantaneous
velocities per unit time) as 3 m/s2, rather than 1.5, as Lee did, but still predicted the same positions
as Lee did.

John: You’re getting the right answer . . . I’m trying to figure out how you are getting the
right answer using 1.5 instead of 3.
Lee: Oh, it blew my mind ‘cause I asked them what they got and all of our answers
matched, but I was using 1.5.
Paul: He took the time and the position and just did position divided by time at that time.
So 13.5 divided by 3 is 4.5.
Lee: Yeah, this is, 6 divided by 2 is three, 13.5 divided by 3 is 4.5.
JM: Ah, so he’s finding the average velocities for the whole period, not between . . . OK I
was saying between 1 and 2, but you’re actually saying between 0 and 2. Got it. So
this is why his way works, because he’s taking the average over the whole thing.

Note that the instructor (JM), despite many years of teaching physics at the university level,
was unfamiliar with the approach and had been confused about what Lee was doing. She still does
not have it quite right, as she refers to the average over the whole thing, although Lee is clear that,
in his mind, the property he has calculated is associated with a particular point in time (clock
reading). Lee then explained how he would predict the position using this approach.

Lee: The formula I was using is just like uh . . . t times 1.5 times t. [Writes t(t 1.5) on the
board] And that gives me all my . . .
Dave: That’s 1/2 a t2.
Paul: Yeah that’s exactly what that is. 1/2 a t2.
JM: And you figured that out just from looking at those numbers?
Lee: Yeah.
Dave: It’s like Good Will Hunting or something. [laughter]

Lee’s formula is simple and direct, a simple combination of an analog to Equation 4 (with the
change in average velocity per unit time, D
v=Dt, replacing the acceleration, a, so that it calculates

v rather than v(ti) and with the initial velocity, v0 equal to zero) and Equation 5. The more
experienced students realize that it is equivalent to the 1/2 at2 of Equation 1, with the initial
position and initial velocity set to 0. Lee’s model maps to:
½position ¼ ‘‘tð1:5 tÞ00 ! tðtD
v=DtÞ ¼ D
v t2 ð8Þ

The reference to Good Will Hunting, a movie in which an untrained janitor is able to solve
difficult problems that have perplexed the math faculty at the school where he works, indicates that
the more experienced students have recognized the value of his self-generated approach.
Mary, a former high school chemistry teacher with little formal instruction in physics,
independently took the overlapping velocities approach to problems involving accelerated
motion. She had been absent due to work travel and had not seen Lee’s presentation, but presented
the same approach in response to this same task when she returned. She attributed her success on
this problem to lack of formal knowledge, alleging, ‘‘It’s ‘cause I didn’t know any physics. I had to
do it conceptually.’’
Linda, the elementary education major who had worked with Lee earlier, and Paul, a
mathematics major, both of whom had little formal background in physics, spontaneously
Journal of Research in Science Teaching. DOI 10.1002/tea
 STUDENTS’ MATHEMATICAL MODELING OF MOTION 167

Figure 3. Task assigned to students after they had developed their mathematical model of motion with
changing velocity.

employed overlapping average velocities as they worked together to analyze the sample data set
shown in Figure 3. She described what she had done as taking ‘‘distance over time to equal the
velocity,’’ although the data were labeled as positions. In this case they found that the (over-
lapping) average velocities changed by 1 m/s with each added second and argued that
the ‘‘velocity’’ at zero seconds must be 4 m/s. Thus, they were able to determine that the average
velocity between 0 and any time, t, was ‘‘t þ 4 [m/s].’’ This parallels Equation 4, with the change
in overlapping average velocity (corresponding to a) equal to 1 m/s2 and the initial (overlapping
average) velocity set equal to 4 m/s. They multiplied that average velocity by the time, t, to get the
distance traveled (as per Equation 5), which in this case was equivalent to the position. It was the
first time either of them had been able to mathematize their descriptions of motion with changing
velocity successfully.
Students with more formal instruction in physics also appreciated the approach. Cecelia, a
preservice secondary science teacher, used overlapping average velocities after seeing Mary
present the method, but reverted back to using velocities calculated over successive intervals
for her final task interview. Jimmy, a biology major with recent college physics experience, used
overlapping average velocities in his final task interview. Even Dave, who said he was ‘‘brought
up on Equation 1,’’ admitted, ‘‘I like [Lee’s way], but I would have never done it [Lee’s] way, but I
like it.’’
It is important to note that none of these students used the overlapping velocity construct
concept consistently, as an indication of a coherent conceptual framework, throughout all the
activities. None of the four students who used it to describe the motion of the bowling ball went on
to use it in describing their data for the cart on the ramp. In fact, none of the students used this
approach to analyze the motion of the cart, and some of those who adopted it for the follow-on
activities with hypothetical data abandoned it in the final task interview.

Discussion

What Constructs Do Students Develop in Analyzing Motion?

Beginning with their original unconstrained descriptions of motion, students in this study
who had not experienced formal physics instruction at the university level paired the ideas of
distance traveled and time interval of travel to describe a motion. This notion of an average
Journal of Research in Science Teaching. DOI 10.1002/tea
168 MARSHALL AND CARREJO

velocity in its most primitive sense, as a distance (rather than a change in positions) per time
interval, was the foothold invention on which these students based their ensuing descriptions
of motion with constant velocity and constant acceleration. Their focus remained on a scalar
quantity, the distance traveled, rather than on a (vector) difference in positions, indicating that this
was a strong intuitive construct for them. Recall that Lee and Linda, novices in college physics, led
Dave, a former high school physics teacher, to consider this construct as well. Hestenes (1992)
argues that his Zeroth Law, invoking a definitive position (xi) for every particle, as the ‘‘theoretical
foundation for measurements of length’’ (p. 734), but perhaps these results indicate that length
measurements might form the foundation for the position construct, rather than the other way
around, at least for some students. Recall that some of our students invoked positions only in order
to describe a length.
Only the novice students considered it necessary for time to enter into a description of motion
at all. Students with more formal experience in physics focused on positions, although these did
not seem to be established against an absolute reference frame and may have been included only to
indicate a distance traveled. These students initially invoked the notion of a trajectory, but not one
that depended on time.
In analyzing ‘‘constant’’ motion, some of our students also developed a construct analogous
to the standard construct of average velocity (distance traveled divided by the time of travel). An
important difference is that these students viewed the average as associated with a given clock
reading, ti, and calculated the average as the total distance traveled from 0 up to ti divided by the
total time of travel in every case. When the velocity is constant, the average velocity is often
presented as the distance over time for the entire motion, but the physics community views the
average velocity as something that can be calculated over any arbitrary time interval and associates
the average with the entire time interval, not a clock reading. If the velocity is varying, either
randomly or systematically, then the standard procedure is to calculate average velocities over
successive nonoverlapping intervals to average them or look for a trend in them. We note that some
of our most experienced students employed the overlapping average velocity construct to make
predictions in a case where they were attempting to create and analyze a motion for which
the velocity was intended to be constant.
The use of successive intervals leads, in the limit of infinitely short time intervals, to the
standard construct of instantaneous velocity, the time derivative of the position, dx/dt. The time
derivative of the instantaneous velocity, the acceleration, d2x/dt2, is the basis for the second
of Hestenes’ (1992) dynamical laws. It is important to note that our novice students’ way of
looking at motion as a series of overlapping average velocities does not lead to the standard
conception. That would clearly pose a limitation in exploiting the overlapping averages velocity
construct to build toward the standard model. These students’ notion of velocity was such that one
could never take the limit of it for small time intervals and get an instantaneous velocity. The time
interval was always from the start of motion up to the time in question; it could not shrink to a small
interval around ti. Because they had no true notion of position, xi(ti), the velocity as the time
derivative of the position did not exist for them.
Some of our students did focus on the standard construct of acceleration, although it was
always treated as constant in the models they developed (as it is in Equation 1). Those who used it
directly to predict positions were not able to do so successfully, but those who used a three-step
process to calculate a distance traveled and final position at an arbitrary time, ti, were able to make
sense of their data and justify their predictions. The latter were, in fact, resorting to the notion of
average velocity over the entirety of the motion in question to make their predictions.
Finally, our novice students were successful in making predictions about motion with
changing velocity by identifying a systematic change over time in their overlapping average
Journal of Research in Science Teaching. DOI 10.1002/tea
 STUDENTS’ MATHEMATICAL MODELING OF MOTION 169

velocity construct. If they had not been careful to make sure the initial position was zero, so that the
distance was numerically equal to the position, the change in overlapping average velocities over
the time interval would not have been constant and would only asymptotically have approached a
value of 1/2 the standard acceleration. In that case, it is unclear how our students would have
proceeded, because they were always careful to ‘‘back it back to zero,’’ as Linda said. That is, in
cases where the initial position was not zero in the problem as posed, they simply relabeled
the times. In cases where the starting position is zero, the change in overlapping velocities is
constant if the acceleration is constant, as can be seen by combining Equation 1 and Equation 2,
with x0 ¼ 0 and t0 ¼ 0:

v ¼ ðxi
x0 Þ=ðti
t0 Þ ¼ xi =ti ¼ ðv0 ti þ 1=2 a ti2 Þ=ti ¼ 1=2 a ti þ v0 ð9Þ

Considering a time interval between clock reading t1 and clock reading t2, the change in
overlapping average velocity per unit time is simply:

D
v=Dt ¼ ð
v1
v2 Þ=ðt1
t2 Þ ¼ ð1=2 a t1
1=2 a t2 Þ=ðt1
t2 Þ ¼ 1=2 a ð10Þ

As Paul noted, the change in overlapping average velocity is numerically equivalent to 1/2
the standard acceleration. Thus, if students recognize, or are led to recognize, a quadratic depen-
dence of the position on time, the coefficient of the t2 term has a simple physical meaning: it is
the change in the overlapping average velocity per unit time. Even experienced students
like Helen have difficulty seeing why that coefficient should be 1/2 the standard acceleration. In
this case, students were able to use the construct they developed for themselves much more readily
to make predictions about motion.
Clearly, the overlapping average velocity approach has limitations. It does not lead to the
standard concept of the derivative and thus does not support the codevelopment of calculus and
physics, often a goal of instruction in physics. The fact that the overlapping average velocity
(distance traveled over time of travel) is not well defined at time zero did not seem bother our
students. They simply took it to be 0 in cases where an object started at a position of 0 at 0 seconds.
When it did not, they made an appropriate adjustment.
If they had failed to convert positions to distances traveled by accounting for the starting
position, there would not have been a consistent pattern in the overlapping average velocities; the
change in the overlapping averages in that case would not be constant but would have approached a
limit of the correct value (the value for adjusted data). Because our students were diligent about
making an adjustment for the initial position, we cannot say how they would have reacted in a
situation where they could not do so. Based on their willingness to identify rough trends in variable
data, they might have identified the limiting value of the change in overlapping velocity as what the
data were ‘‘trying to do,’’ and used that value to determine a viable model for the motion. At
a minimum, failure to find a constant change in the overlapping average velocities would
likely have alerted them to a problem with the procedure.

How Does the Task Interact with the Constructs Students Employ?
The task in which students were engaged clearly made a difference in the models produced by
our students. Consider the two groups of students who used overlapping average velocities to
describe and predict the motion of a bowling ball rolling down a hall. The limitation of only being
Journal of Research in Science Teaching. DOI 10.1002/tea
170 MARSHALL AND CARREJO

able to measure the time at one distance per roll (allowing for a redundant measurement) may have
led these students along this path. Had they been strongly inclined toward calculating velocities
over successive intervals, they could have stationed their two observers at different positions, but
would not have been able to average the two values together. The fact that some of the in-service
teachers who were engaged in the same activity also resorted to the overlapping average, even with
10 pairs of observers to take data in each roll (Carrejo, 2004), indicates that this is a task that is
likely to elicit that construct in students for which it makes intuitive sense.
On the other hand, our students did not calculate overlapping average velocities with the
ticker timer data. Even Lee and Linda attempted to interpret those data in terms of an average
velocity for successive intervals. They may have been led in this direction by the fact that it was
easier to measure the distances between successive tick marks rather than to sum them to get
distances traveled. This task might have steered students toward the idea of calculating velocities
for successive intervals, but most of our students were not successful in it. Only Steve, John, and
Jimmy were successful in making predictions about their ticker timer data. A task with ticker timer
data might be used to steer students toward the three-step approach these three used. Involving as it
did the average velocity over the entire motion rather than overlapping average velocities, this
approach might be a bridge between the strong intuitive focus on average velocity, and the
standard construct of instantaneous velocity.
In the follow-on task, however, the data from hypothetical motions with constant, nonzero
acceleration steered several groups of students to use the overlapping average velocities, despite
the fact that the data were labeled as position and time. In the exercise shown in Figure 2, there was
a column explicitly indicating that students should consider changes in position rather than
distance traveled. Still, our novice students interpreted this as the change in position from 0 up to ti,
that is, the complete distance traveled, rather than the distance traveled between two successive
data records. The presentation of the data in this task, where position (equal to distance traveled)
was presented adjacent to time of travel, may have steered our students toward using these intuitive
constructs to create a model of motion. The fact that the overlapping velocity construct arose as
the students analyzed these ‘‘canned’’ data may also have been due to the fact that, in the first cases
presented, the data followed an exact pattern, as opposed to the ticker timer data. However, even in
the case of data showing statistical variations, presented in another follow-on task, Mary and other
students were able to identify a pattern in the overlapping velocities and make valid predictions,
although none had exploited this construct in analyzing the ticker timer data.
Although Hestenes (1992) warns against equating such a pattern or functional relationship in
data with a meaningful model (p. 738), it was clear that these students were interpreting the
patterns in the data in terms of constructs that were meaningful to them. They were able to describe
what they were doing in terms of velocity as total distance traveled over total time interval of
travel. Further, he argues that ‘‘pattern recognition skills are essential to understanding physics’’
(p. 745) and tasks should be designed to develop these skills in students.
Finally, we should note that these tasks were not designed specifically to challenge the
overlapping velocity construct in cases where it could not be readily applied to the data, especially
those with nonzero starting points. Had the goal been to move them from their intuitive approaches
toward the standard approach, such cases might have been explicitly discussed to encourage
students to refine their ideas. Previous tasks, hypothetical data sets displaying constant or near-
constant velocity, had led students to develop the technique of ‘‘backing it [the position] back to
zero,’’ as Linda put it. The previous tasks were designed with an eye toward developing the
standard model, getting students to recognize the initial position as a component of the model
explicitly. However, it would be possible not to challenge students in this regard unless and until
the overlapping model arose.
Journal of Research in Science Teaching. DOI 10.1002/tea
 STUDENTS’ MATHEMATICAL MODELING OF MOTION 171

How Does Previous Instruction Affect Students’ Analyzing and Making

Predictions about Motion?
As in previous work (Carrejo, 2004), there were instances in which previous instruction seemed
to hamper our students’ ability to generate reasonable models for themselves. For example, it is
unlikely that Helen would have confused the instantaneous and average velocity constructs the way
she did had she been given time and tasks in previous instruction that allowed her to develop these
constructs on her own. Instead, she simply resorted to dimensional (unit) analysis as a justification
for her model, as many students are instructed to do in physics coursework. This approach is valid
only if the constructs involved have actual meaning for students.
Lee’s claim that the time interval that elapsed during a motion should be included in a
description of that motion may also be traced to previous instruction. As the only students who
included a reference to time in their original description of motion, Lee argued to his partner,
Linda, that what he called ‘‘velocity’’ was important because he had learned about it in high
school. Later Linda recalled the phrase, ‘‘distance equals rate times time’’ to describe their
procedure for getting from the average velocity to the distance traveled. This is a common mantra
in precollege instruction and the rate in question is indeed the average velocity over the entire
motion. Thus, our students’ comfort with describing motion in terms of averaged velocities may
also be to some extent the result of prior instruction, although it is arguably closer to the intuitive
approaches students might use in the absence of instruction, based on the results of our first task.

Conclusions and Implications

Woolnough (2000) emphasizes that students must see ‘‘links between the mathematical
processes they are using and the physics they are studying’’ (p. 259). It was clear from these
students’ approaches to the task of describing and predicting motion that the link was weak at best
between the standard equations of motion in the physics curriculum, Equation 1 in particular,
and actual motions as represented by data. Even participants with considerable experience with
Equation 1 were hard pressed to describe how it would arise from data or from readily identifiable
properties of the motions they created and observed, and yet they felt, as Dave said, they could
not ‘‘run away from it.’’ This confirms previous results indicating that traditional instruction, as
provided by most university physics classes, is likely to leave future teachers employing a model
of motion for which they themselves do not have a robust understanding (Carrejo, 2004).
This finding implies that much more must be done than simply presenting the algebraic
arguments connecting students’ intuitive understanding, that is, of average velocity, to the
standard forms, that is, Equation 1, as is generally done in textbooks. In the words of Lehrer,
Schauble, Carpenter, and Penner (2000), ‘‘Rather than transmitting inscriptions (even powerful
ones), we may do better by taking a longer term perspective, concentrating instead on trying
to help our students understand over the long term what their inscriptions are good for: They allow
you to plan and compose what would otherwise not be seen, and to keep a trace of what otherwise
would disappear. All of these functions are indispensable to good argument in mathematics and
science, and, more generally, to coming to know the world’’ (p. 359, emphasis added). Ideally,
students would be given time to develop their own models of motion and tasks that steered them
toward evaluating and refining them.
These results confirm that students can indeed generate productive models of motion for
themselves, particularly if they are not encumbered by standard constructs from the curriculum
that have no intuitive meaning for them. The success some of our novice students had with the
approach they invented themselves, using overlapping average velocities, makes it worth consi-
dering the value of this approach either as a stepping stone toward the standard approach or as a
Journal of Research in Science Teaching. DOI 10.1002/tea
172 MARSHALL AND CARREJO

model in its own right, despite its limitations. The approach allowed students to see the quadratic
dependence of the position on time in cases of uniform nonzero acceleration as arising in a
meaningful way directly from data. Lee was able to explain why he was multiplying by the time
twice in his model. This is in contrast to simply fitting the data with a calculator or recognizing the
quadratic dependence from a graph, without being able to relate these to properties of the motion,
either directly observed or as seen in data.
For younger precollege students who are trying to deal mathematically with motion for the
first time, the model that Lee and other students employed might be not only meaningful, but also
appropriate, if it empowers students to make sense of the ubiquitous phenomenon of motion.
Thinking in terms of a ratio of distance to time is indeed an important step, even if students will
ultimately need to move beyond it. It is important that precollege teachers be ready to recognize
and support this kind of thinking.
As noted above, Hestenes (1992) argues that motion is a subject from which to build the
Newtonian framework; not the other way around. Thus, it is also important to investigate how the
overlapping velocity construct might lead toward a more sophisticated understanding, noting that
it lacks critical elements of the Newtonian framework. First, it is a strictly scalar construct; it does
not involve the (vector) difference in positions. Second, it does not readily lead to the notion
of instantaneous velocity; it is not amenable to taking the limit for vanishingly small times nor,
consequently, to the development of important calculus concepts such as the derivative. Both have
implications for the kinds of tasks that might steer students from the overlapping construct to the
standard framework, based on (vector) positions as functions of time and their derivatives.
Follow-up work involving motion in two dimensions would force students to consider the first
issue. In the two-dimensional case, it is not generally possible to render the position and distance
traveled numerically equivalent by a simple adjustment of the starting position. The difference
between predicting the object’s position (where it is) and how far it has traveled becomes critical
and might lead to revision of the students’ models. The tendency of some of our students to invoke
positions as indicators of distances traveled in the original task could also be exploited in this
regard. The second issue may be more difficult to address, and would require additional investi-
gation if it is the goal to do more than simply confront students with the inadequacy of the
overlapping construct.
This work is clearly exploratory in nature. The small number of subjects involved limits the
generalizability of the results. Still, the robustness of the overlapping average velocity approach
within this population, the fact that it occurred independently within our students and has been
identified in other populations (Carrejo, 2004), makes it highly likely that this is a common
intuitive way for students to think about motion. The general usefulness of the overlapping average
velocity construct, either as a foothold on the path toward the standard models or as part of a model
of motion in its own right, is yet to be determined but merits further research. At a minimum, our
results imply that instructors in physics and physical science courses should be aware that many of
their novice students may, in fact, be viewing the velocity at a given point in time as the total
distance traveled up to that time divided by the total time interval of travel, despite the presentation
of the instantaneous velocity construct in the curriculum.

References
Bowden, J., Dall’Alba, G., Martin, E., Laurillard, D., Marton, F., Masters, G., Ramsden,
P., Stephanou, A., & Walsh, E. (1992). Displacement, velocity and frames of reference:
Phenomenographic studies of students’ understanding and some implications for teaching and
assessment. American Journal of Physics, 60, 262–269.
Journal of Research in Science Teaching. DOI 10.1002/tea
 STUDENTS’ MATHEMATICAL MODELING OF MOTION 173

Carrejo, D. (2004). Mathematical modeling and kinematics: A study of emerging themes and
their implications for learning mathematics through an inquiry-based approach. Unpublished
doctoral dissertation. University of Texas at Austin.
Doerr, H., & Lesh, R. (Eds.), (2003). Beyond constructivism: Models and modeling
perspectives on mathematics problem solving, learning, and teaching. Mahwah, NJ: Erlbaum.
Giancoli, D.C. (1998). Physics. (5th ed.). Upper Saddle River, NJ: Prentice Hall.
Goldberg, F.M., & Anderson, J.H. (1989). Student difficulties with graphical representations
of negative values of velocity. The Physics Teacher, 27, 254–260.
Gravemeijer, K., & Doorman, M. (1999). Context problems in realistic mathematics
education: A calculus course as an example. Educational Studies in Mathematics, 39, 111–129.
Hestenes, D., Wells, M., & Swackhamer, G. (1992). Force Concept Inventory. Physics
Teacher, 30, 141–158.
Hestenes, D. (1992). Modeling games in the Newtonian World. American Journal of Physics,
60, 732–748.
Kirkpatrick, L.D., & Francis, G.E. (2004). Physics: Aworld view. (5th ed.). Philadelphia, PA:
Saunders College Publishing.
Lehrer, R., Schauble, L., Carpenter, S., & Penner, D. (2000). The interrelated development of
inscriptions and conceptual understanding. In P. Cobb, E. Yackel, & K. McClain (Eds.),
Symbolizing and communicating in mathematics classrooms: Perspectives on discourse, tools,
and instructional design (pp. 325–360). Mahwah, NJ: Erlbaum.
Lesh, R., & Doerr, H.M. (2003). Foundations of a models and modeling perspective on
mathematics teaching, learning, and problem solving. In H.M. Doerr & R. Lesh (Eds.), Beyond
constructivism: Models and modeling perspectives on mathematics problem solving, learning,
and teaching (pp. 3–33). Mahwah, NJ: Erlbaum.
McDermott, L.C. (1990). A perspective on teacher preparation in physics and other sciences:
The need for special science courses for teachers. American Journal of Physics, 58, 734–742.
McDermott, L.C. (1996). Physics by inquiry. New York: John Wiley & Sons.
McDermott, L.C., Rosenquist, M.L., & Zee, E.H. (1987). Student difficulties in connecting
graphs and physics: Examples from kinematics. American Journal of Physics, 55, 503–513.
Nemirovsky, R., & Rubin, A. (1992). Students’ tendency to assume resemblances between a
function and its derivative. Cambridge, MA: TERC.
NRC. (1996). National science education standards. Washington, DC: National Academy
Press.
Rosenquist, M.L., & McDermott, L.C. (1987). A conceptual approach to teaching
kinematics. American Journal of Physics, 55, 407–415.
Shaffer, P.S., & McDermott, L.C. (2005). A research-based approach to improving student
understanding of the vector nature of kinematical concepts. American Journal of Physics, 73,
921–931.
Shternberg, B., & Yerushalmy, M. (2003). Models of functions and models of situations: On
design of modeling-based learning environments. In H.M. Doerr & R. Lesh (Eds.), Beyond
constructivism: Models and modeling perspectives on mathematics problem solving, learning,
and teaching (pp. 479–498). Mahwah, NJ: Erlbaum.
Trowbridge, D.E., & McDermott, L.C. (1980). Investigation of student understanding of the
concept of velocity in one dimension. American Journal of Physics, 48, 1020–1028.
Wells, M., Hestenes, D., & Swackhamer, G. (1995). A modeling method for high school
physics instruction. American Journal of Physics, 63, 606–619.
Woolnough, J. (2000). How do students learn to apply their mathematical knowledge to
interpret graphs in physics? Research in Science Education, 30, 259–267.
Journal of Research in Science Teaching. DOI 10.1002/tea

View publication stats