Proceedings of the International Conference on Undergraduate Physics

Education (College Park, August 1996)

MODELING METHODOLOGY FOR PHYSICS TEACHERS

David Hestenes
Department of Physics and Astronomy
Arizona State University
Tempe, Arizona 85287-1504

Abstract: Scientific practice involves the construction, validation and application of

scientific models, so science instruction should be designed to engage students in making
and using models. Scientific models are coherent units of structured knowledge. They are
used to organize factual information into coherent wholes, often by the coordinated use of
general laws or principles. Therefore, the structure of scientific knowledge can be made
more explicit for students by organizing course content around a small number of basic
models. The ability of students to make and use models depends on the representational
tools at their command. Students learn transferable modeling skills by applying given
models to a variety of situations to describe, explain, or predict physical events or to
design experiments. These ideas have been incorporated into a methodology for physics
teaching and a course for training teachers.

1. Introduction
This paper describes theoretical underpinnings for a two-semester course
in the methodology of physics teaching designed especially for in-service and pre-
service high school teachers but valuable for regular physics majors as well.
Currently the course is being disseminated nationally in a series of summer
"Modeling Workshops" supported by the National Science Foundation. The first
two-summer series, held at university sites in Tempe and Chicago, was completed
in 1996. A second series begins in summer 1997 at three new sites, and a third
begins in 1998 at three more sites. Information about the workshops can be
obtained by email from <[email protected]> or by visiting our web site at
http://modeling.la.asu.edu/modeling.html. Incorporation of the courses into the
university physics curriculum is discussed by Hestenes and Jackson (7).
The Modeling Workshops have evolved over more than a decade from
educational research and the experience of exceptional physics teachers. The story
has been told in (16). This paper updates and elaborates the methodology, which
is still evolving. Unfortunately, a full understanding of the "Modeling Method"
cultivated in the Workshops cannot be acquired from a written description.
Participation in a Workshop is essential. This paper gives an overview of the
pedagogical framework with selective emphasis on a few important issues. For a
more complete account the references should be consulted.

2. Course Objectives and Scope

The main purpose of the Modeling Workshops is to empower teachers
with a robust teaching methodology. This includes the cultivation of teacher
abilities to critically analyze any given curriculum materials and organize valuable
parts into effective instructional units which make the underlying models explicit
tasks which require a strong pedagogical framework. These abilities are needed
to take advantage of accelerating changes in curriculum materials, driven on the
one hand by advances in educational research and by new computer technology
and software on the other.
Workshops I and II have different designs to promote complementary
pedagogical objectives. The main objective of Workshop I (on mechanics) is to
acquaint teachers with all aspects of the modeling method and develop skill in
implementing it. To that end, they are provided with a fairly complete set of
written curriculum materials to support instruction organized into coherent
modeling cycles (as described in (16)). The physical materials and experiments in
the curriculum are simple and quite standard, already available in any reasonably
equipped physics classroom. In due time the teachers come to understand that the
modeling method can be applied as well to more unusual, complex, and exciting
subjects. Since "teachers teach as they have been taught," workshops include
extensive practice in implementing the curriculum as intended for their classes.
They circulate through roles of student and instructor as they practice techniques
of guided inquiry and collaborative learning.
The design of the Second Workshop (on light, electricity, heat, ) is more
fluid and open-ended. The main objective is to give the teachers intensive
experience in using what they have learned about the modeling method to
construct and evaluate coherent instructional units of their own design. They must
first learn how to construct such units from standard curriculum materials, already
available or easily procured for their classes. For each of the major topics, small
groups of teachers study available curriculum materials. They are guided to
identify the underlying models and develop coherent instructional units which
engage students in using the models to structure their own understanding. The
units developed by the various groups are presented to the entire class for critique,
evaluation and discussion of how to organize the units into a complete curriculum.

Instructional materials developed and disseminated in the Workshops are

secondary to the greater goal of developing a flexible teaching methodology
receptive to new materials, especially in the nascent domain of physics education
software. This point deserves emphasis, because there is an unfortunate tendency
in some quarters to equate educational reform with the creation and distribution of
new materials. The Modeling Workshops aim to promote a community of reform-
minded teachers who will continue upgrading their own curriculum materials and
classroom practice who are motivated by the vision of a dynamic teacher
empowered with special skills, rather than by the static "Here's a set of materials,
use 'em!"

3. Where Physics Teaching Fails: Evaluating Instruction

Physicists appreciate the need for instruments which produce accurate and
reproducible measurements. Reif (1996) has called for the same in physics

2
teaching. He contends that we need operationally specified performance
standards for comparing the effectiveness of different teaching methods. This
calls for development and calibration of a battery of instruments for evaluating
instruction in a variety of ways. The Force Concept Inventory (FCI) is one such
instrument with proven value (8).
The FCI was developed to assess the effectiveness of mechanics courses in
meeting a minimal performance standard: to teach students to reliably
discriminate between the applicability of scientific concepts and naive
alternatives in common physical situations. The FCI systematically probes student
abilities to make such discriminations with respect to six fundamental aspects of
the Newtonian force concept.
Including the recent survey by Hake (3), we now have FCI data on more
than 12,000 students and 100 physics courses in high schools, colleges and
universities through the United States. This large data base presents a highly
consistent picture, showing that the FCI provides statistically reliable and
discriminating measures of minimal performance in mechanics. The results
strongly support the following general conclusions:
Before physics instruction, students hold naive beliefs about
mechanics which are incompatible with Newtonian concepts in most
respects.
Such beliefs are a major determinant of student performance in
introductory physics.
Traditional (lecture-demonstration) physics instruction induces only a
small change in the beliefs. This result is largely independent of the
instructors knowledge, experience and teaching style.
Much greater changes in student beliefs can be induced with
instructional methods derived from educational research.
These conclusions can be quantified. From pre/post course FCI scores of 14
traditional courses, Hake (3) found a mean normalized gain of 22%, with a largest
gain of 32%. In contrast, for 41 courses using non-traditional teaching methods,
he found a mean gain of 52%, with a largest gain of 69%. The two means differ
by several standard deviations a highly significant result. This shows that
traditional instruction fails badly in meeting a minimal performance standard for
mechanics. Moreover, the failure cannot be attributed to inadequacies of the
students, for the data show that alternative methods of instruction can do much
better. A major problem in physics education research for more than a decade has
been to identify essential conditions for learning Newtonian physics and
therefrom devise more effective teaching methods.
To understand the significance of the low FCI scores for traditional
instruction, consider the results in Box 1 which have been extracted from FCI
data. These results (which can be supported by a much more extensive analysis)
show that most students complete introductory physics, even at the university
level, with a seriously deficient understanding of the Newtonian force concept.
The consequences of this fact are devastating. Unaware that their own ideas

3
about force differ drastically from those of the teacher, most students
systematically misunderstand what they hear and read in traditional introductory
physics. Consequently, they cannot understand why they fail at problem solving,
and they are forced to resort to rote methods for learning meaningless formulas
and procedures. The result is frustration, humiliation and student turnoff!
Fortunately, educational research has shown us how to do better. Indeed,
we have documented that several teachers using the modeling method have
achieved FCI gains within and even surpassing the highest performance range in
Hakes data set. Details will be published elsewhere.

4. Understanding Physics: A Question of Structure

One reason for the failure of traditional instruction is that it overlooks the
crucial influence of students personal beliefs on what they learn. In the traditional
setting naive student beliefs about physics are labeled as misconceptions and are
likely to be summarily dismissed as unworthy of consideration. However, the FCI
data show that students are not easily induced to discard their misconceptions in
favor of Newtonian concepts. Indeed, physics teachers and educational
researchers who are aware of this problem have expended considerable effort in
designing and testing teaching methods to deal with specific misconceptions.
Although their outcomes have been decidedly better than the traditional ones,
success has been limited and the methods can be criticized as excessively time
and labor intensive. Many have concluded that student beliefs are so "deep-
seated" that heavy instructional costs to unseat them are unavoidable. However,
documented success with the modeling method (16) suggest that an indirect
treatment of misconceptions is likely to be most efficient and effective. In any
case, an optimal solution of the problem will surely require an understanding of
the cognitive factors involved.
Our analysis of the cognitive factors begins with the observation that the
student "misconceptions" in Box 1 are framed in terms of metaphors. Cognitive
scientists (10) have identified metaphors as a fundamental tool of human thought,
which we use so frequently and automatically that we seldom notice the
metaphors unless they are called to our attention. Metaphors are used to structure
our experience and thereby make it meaningful. In particular, metaphors help us
make sense of new experience (target domain) by mapping it onto structure of
familiar experience (source domain). Johnson (9) argues that our bodily
experience, structurally represented by mental image schemas, is the source for
our strongest metaphors. For example, we use prepositions to construct a rich
system of spatial metaphors, like "more is up," which are grounded in such
schemas. Johnson argues further that the metaphorical use of image schemas is
pervasive in our understanding of abstract ideas, including mathematics. Even the
idea of "deduction" derives from the spatial concept of "following a path."
This analysis suggests that the source domain for the metaphors in Box 1
is indeed deep-seated, but that the mapping to the target domain is more
compliant and may adjust fairly quickly to make sense of new experience under

4
 Newtonian Concepts vs. Naive Beliefs Results
First Law vs. Motion requires force (~60%)
(Impetus Principle)
Second Law vs. Force is action (~40%)
(No Passive forces)
Third Law vs. Force is war (~90%)
(Dominance Principle)
BOX 1. Major beliefs about forces. The percentages are rough estimates of
typical post-instruction results for traditional instruction in University (calculus-
based) Physics at a large state university. They indicate the fraction of students
who fail to consistently discriminate an appropriate use of Newtons Laws from
naive alternatives.

appropriate conditions. In other works, misconceptions about mechanics are just

cases of misplaced metaphors which will be corrected automatically in response
to appropriate experience. The problem of instruction is therefore to arrange the
conditions for such experience. This perspective provides a rationale for one of
the most systematic methods for treating misconceptions. Clement and Camp (2 )
engage students in constructing a series of "bridging analogies" which directs
their attention to commonalties in a variety of physical situations.
It has been suggested that student misconceptions about mechanics are
fundamentally different than misconceptions about electricity and magnetism,
because students have so much more experience with moving objects. Our present
perspective on metaphors suggests otherwise. Students automatically use
metaphors to structure experience in any scientific domain. The case of mechanics
shows that when they do this casually the metaphors are likely to be misplaced
even in domains where they have extensive experience. A major objective of
teaching should therefore be to help students "straighten out" their metaphors.
This readjustment of mappings onto their personal image schemas can be
regarded as developing physical intuition.
Physics teachers know the importance of "physical intuition," but they
should be aware that their intuitions (metaphorical mappings) are not the same as
their students. Therefore, when they express their intuition about force by
describing it as "a push or a pull," they are likely to evoke in students a
metaphorical association of force with human action, one of the major
misconceptions in Box 1.
As explained in the following sections, the modeling method approaches
the problem of restructuring students intuitions by engaging them in explicit
construction and manipulation of externally structured representations. In the case
of mechanics (6), we have found it advisable to engage students in explicit
comparisons of the three major misconceptions in Box 1 with their Newtonian
alternatives. When these three are adequately treated, many other misconceptions
about mechanics fall away with them.

5
 MODELING METHOD Synopsis
The Modeling Method aims to correct many weaknesses of the traditional lecture-
demonstration method, including the fragmentation of knowledge, student passivity, and the
persistence of naive beliefs about the physical world.

What to teach: Model-centered instructional objectives

To engage students in understanding the physical world by constructing and using scientific
models to describe, to explain, to predict, to design and control physical phenomena.

To provide students with basic conceptual tools for modeling physical objects and processes,
especially mathematical, graphical and diagrammatic representations.

To familiarize students with a small set of basic models as the content core of physics.

To develop insight into the structure of scientific knowledge by examining how models fit
into theories.

To show how scientific knowledge is validated by engaging students in evaluating scientific

models through comparison with empirical data.

To develop skill in all aspects of modeling as the procedural core of scientific knowledge.

How to teach: Student-centered instructional design

Instruction is organized into modeling cycles which engage students in all phases of model
development, evaluation and application in concrete situations thus promoting an
integrated understanding of modeling processes and acquisition of coordinated modeling
skills.

The teacher sets the stage for student activities, typically with a demonstration and class
discussion to establish common understanding of a question to be asked of nature. Then, in
small groups, students collaborate in planning and conducting experiments to answer or
clarify the question.

Students are required to present and justify their conclusions in oral and/or written form,
including a formulation of models for the phenomena in question and evaluation of the
models by comparison with data.

Technical terms and representational tools are introduced by the teacher as they are needed
to sharpen models, facilitate modeling activities and improve the quality of discourse.

The teacher is prepared with a definite agenda for student progress and guides student
inquiry and discussion in that direction with "Socratic" questioning and remarks.

The teacher is equipped with a taxonomy of typical student misconceptions to be addressed

as students are induced to articulate, analyze and justify their personal beliefs.

BOX 2.

6
 5. Teaching for Structured Knowledge: The Modeling Method
Physics can be characterized as a complex network of models interrelated
by a system of theoretical principles. Models are units of structured knowledge
used to represent observable patterns in physical phenomena. Accordingly,
"physical understanding" is a complex set of modeling skills, that is, cognitive
skills for making and using models. The primary objective of physics teaching
should therefore be to develop student modeling skills for making sense of their
own physical experience and evaluating information reported by others. The
Modeling Method of instruction (Box 2) has been developed to meet this
objective.
To implement the instructional objectives in Box 1 effectively, teachers
need to be equipped with clear ideas about models and modeling along with many
examples of both. These matters are addressed in the next two sections. Teachers
also need strategies and tactics for managing classroom activities. That is
discussed in section 8, but it takes Workshop experience to fully appreciate the
skills involved.
Systematic implementation of the modeling method requires some
restructuring of the physics curriculum. To make the structure and content of
physics as explicit as possible, course material is organized around a small
number of "basic models." For the mechanics portion of our high school physics
course we use the five single particle models listed in Box 3. These are subdivided
into five kinematical models to clarify the distinction between kinematics and
dynamics as well as between descriptive and causal models. Instruction is
designed to make students intimately familiar with the structure and use of the
basic models. That includes using them to construct and analyze more complex
models. It provides them with a set of clear examples grounded in personal
experience from which they can develop general concepts of models and
modeling in physics.
6. Models
The word "model" is used so frequently and expressively by physicists,
especially on the research frontier, that it obviously refers to something important.
However, some use the term so loosely as not to discriminate between models and

Basic Particle Models in Newtonian Mechanics

Kinematical Models Causal Models
Constant velocity Free Particle: Fi = 0
Constant acceleration Constant force: Fi = constant
Simple Harmonic Oscillator (SHO) Linear binding force: Fi = - k r
Uniform circular motion (UCM) Central force (with constant | r | )
Collision p = I Impulsive force

BOX 3.

7
theories (6). To capture the concept for the purpose of instruction, we provide it
with a precise definition: A model (in physics) is a representation of structure in a
physical system and/or its properties. The system may consist of one or more
material objects or massless entities such as light. Unlike a theory, a model refers
to an individual system, though that individual may be an exemplar for a whole
class of similar things.
The concept of model is predicated on a philosophy of scientific realism
holding that the universe is populated by things with definite physical properties.
Physicists learn about things by investigating their properties empirically and
framing their conclusions in terms of models and theories. They understand
physical things through representations in the structure of validated models.
Although scientific realism has been challenged by some philosophers, it is
doubtful that physics research makes sense without it.
To complete our definition of model we still need to clarify the concepts
of "structure" and "representation." Models in every domain of physics are
characterized by the four types of structure listed in Box 4. Some commentary is
needed to illuminate the structural types and the various representations used to
specify them in a model. Examples are given in Box 5. It is especially important
to note (in Box 5) how representation of structure is distributed across several
different diagrammatic and mathematical representations. Thus, complete

Model Specification
A model is a representation of structure in a physical system and/or its
properties. It describes (or specifies) four types of structure, each with
internal and external components:
(a) systemic structure specifies
composition (internal parts of the system)
environment (external agents linked to the system)
connections (external and internal causal links)
(b) geometric structure specifies
position with respect to a reference frame (external geometry)
configuration (geometric relations among the parts)
(c) temporal structure specifies change in state variables (system
properties)
descriptive models represent change by explicit functions of time
causal models specify change by differential equations with
interaction laws
(d) interaction structure specifies interaction laws expressing
interactions among causal links, usually as function of state variables

BOX 4.

8
specification (and understanding) of a model requires coordination of multiple
representations
(a) Systemic Structure
Every complete model includes a system schema specifying the
composition, environment and connections of the system (Box 5). Though system
schemas can be specified verbally, for systems of any complexity tabular and/or
diagrammatic representations are more informative. Diagrammatic techniques for
this purpose have evolved in all the sciences without recognition of their common
cognitive function. For example, a standard electrical circuit diagram is a system
schema. It specifies constituent parts of the system and electrical connectivity
(causal links) among them. Similarly, chemists have developed elaborate
representational tools for constructing molecular system schemas, and in biology,
for example, animal digestive and circulatory systems can hardly be understood
without diagrams.
Strangely enough, diagrams are seldom used to represent system schemas
in elementary mechanics, perhaps because the systemic structure of a simple
mechanical system is deemed to be so obvious to physicists. It is not so obvious to
students, however. A common source for student failure to solve mechanics
problems can be traced to their inability to identify agents of force on an object
(5) in other words, their inability to construct (in their heads) appropriate
system schemas.
To construct a system schema from a given physical situation, or even
from an artificially simple situation described in a "word problem," can be
deceptively difficult; for it requires a judicious choice of system, identification of
relevant properties and suppression of irrelevant information. It is actually a
complex skill requiring extensive modeling to develop to a high level. Having
students routinely construct and discuss diagrammatic system schemas like those
in Box 5 may help them develop that skill. (This is a new idea that has not yet
been tested in practice.) The schema in Box 5 specifies a system composed of two
objects connected by a string (S) and interacting with three external agents:
table(T), earth (E) and pulley (P). Note that the closed dotted line serves to
separate the system from its environment.
System schemas should not be confused with maps, which
represent other kinds of information about systems. The schema in Box 5 may be
regarded as implicit in the situation map (in the same Box), so it can be extracted
(or abstracted) from the map, though this involves a choice in identifying the
system of interest. When system structure is easily "read off" a map, a schema
diagram may be superfluous. Nevertheless, its construction may serve as a check
that something obvious has not been overlooked, or that everyone is talking about
the same system.
(b) Geometric Structure: Geometric Models
Spatial location and configuration represent geometric properties
possessed by all systems, though some models (for example, electrical circuit

9
 Representations of structure in a model
for the modified Atwood machine
P
System Schema

1 S 2

Situation Map T E

m1 Subsystem Schema

1 S S 2
m2

T E E

Motion Map Geometric Structure

.
0
. v1

x1
L x 2 = x1 + L
> 2 = v1
== v

Interaction N
x2 . v2
> a2 = a 1
==

Interaction Laws
Map
f . T1 Internal: T1 = T2
External: f =
W1
T2
. W1
W2
N
= m1 g
= m 2g
Temporal Structure W2
(Equations of Motion)
For single particle subsystems: T1 - N = m 1a 1
m 2 g - T2 = m2 a2
N = m1 g
For entire 2 particle system: m 2g - m1 g = ( m 1 + m 2 ) a 1

BOX 5. (see text for explanation)

models) ignore geometric properties. On the other hand, purely geometric models
(such as maps and architectural plans) ignore interactions and temporal structure.
Newtonian theory presumes that geometric properties of all objects can be
modeled within a 3D Euclidean space. This has been called the Zeroth Law of
physics (6). Relativity theory requires a profound modification of the Zeroth Law,
but that will not concern us here.

10
 Temporal Structure in the
Constant Acceleration Model

Motion Map
v
v0 a
r

Parametric equation: r = v0 t + _12 a t 2

Temporal structure: 2-time (2-point) relations

2r 2r
v0 v + v0 =
t t
v v0 = a t
v at

BOX 6.
Mathematical representations of geometric structure employ vectors or
coordinate systems, but these are invariably coordinated with maps of diagrams to
(i) designate an interpretation for the mathematical symbols, and (ii) help the user
visualize the structure as a whole. Note that the geometric structure in Box 5 is
distributed across several representations. Geometric structure relevant to the
model has been extracted from the situation map and represented in a motion
map, wherein the objects are represented as particles and their positions (x1, x2) on
the path of motion are designated. The positions are related to the length L of the
string. For an inextensible string (implicitly a model of interaction structure in the
string) it follows immediately that the velocities and accelerations of the particles
are equal.
(c) Temporal Structure: Process Models
State variables are descriptors representing basic properties that change
with time. The state of a system at a particular time is described by the values of
its state variables at that time. A model of temporal structure provides a
comparison of states at different times. We can distinguish two kinds of model for
temporal structure: descriptive and causal.

11
 A descriptive model gives the state variables as explicit functions of time.
For example, all the kinematic models listed in Box 3 are descriptive models. For
the constant acceleration model, temporal structure of the state variables (x and v)
is specified explicitly by vector-valued functions of time in Box 6. It is also given
by a motion map, which depicts the values of kinematical quantities along the
path of the particle.
A causal model specifies a change in state variables by equations of
change, which determine the change as a consequence of interactions. Causal
models are explanatory models, where the cause explains the change of state.
Newtons Second Law F = ma is the prototypical prescription for an equation of
change, where the F represents cause and the a represents effect. It becomes a
determinate equation of motion (change of place) when force laws are used to
specify F as a function of the state variables (x and v).
A process model is a model for change in a single property (energy, for
example) without reference to the underlying systems that possess that property.
This can lead to error, as in the pseudoscience Energetics, which holds that energy
is the primal stuff of which everything is made. Process models are therefore
incomplete models, so to understand them students must know how to extend
them to complete models. For example, in a process model of energy storage and
transfer, students must be able to identify the carriers of the energy and the
systems in which it is stored.
Motion (represented as change of position) is the most fundamental kind
of change, because time itself is measured by standardized motion in an object
called a clock. Relativity theory reduces temporal structure to geometric structure
by fusing space and time into a single model of spacetime structure. However,
that will not concern us here.
(d) Interaction Structure: Interaction Models
Interactions between systems are usually modeled by expressing the
interaction variables as functions of state variables for the systems. Such
"interaction laws" are needed to determine equations of change for the systems. In
Newtonian mechanics, where the interaction variables are forces, the interactions
are modeled by force laws, such as Hookes law or Newtons law of gravitation.
Alternatively, interactions can be modeled by potential energy functions. Other
kinds of interaction involve the exchange of mass, charge or other physical
quantities.
In Box 5 the interaction structure is distributed across two kinds of
representation. First, the forces on each object are represented by force diagrams
on an "interaction map." Second, magnitudes of the forces are specified by a set of
interaction (force) laws. Note that the directions of the forces are specified by
arrows on the map and not by any mathematical symbols or relations. This one
more example of mixed representations employed in specifying a model.

12
 7. Modeling and Modeling Tools
Scientific understanding emerges from making and using models, in short,
from modeling. To learn science, therefore, students must engage in all aspects of
modeling. Modeling instruction is designed to promote that. To see what such a
design involves, in this section we discuss three major modeling modes: model
construction, analysis and validation. This has implications for problem solving as
well as for the design of modeling tools and software to enable modeling.
Modeling in science is a complex process, requiring coordination of
several kinds of modeling activities. Most, if not all, of the procedural knowledge
implicit in scientific practice can be classified into some mode of modeling.
Indeed, modeling constitutes the heart of the scientific method (6).
It has often been argued that instruction in the scientific method should be
the primary purpose of science education, so students can learn to "think like a
scientist." Unfortunately, implementation of this objective in science curricula has
foundered on the lack of an adequate characterization of "scientific method."
Indeed, as researchers in the history of science and cognitive science document
the complexities of scientific practice, some have argued that there is no such
thing as the scientific method that there is not one method but many.
Modeling theory provides a new perspective on the methodology of
science. A detailed analysis of modeling modes provides a rich characterization of
the scientific method as a coherent system of procedural knowledge for
constructing and using models in any scientific domain. The complexity of
scientific practice is then seen to arise from specialization of some modeling tools
to limited domains and the special skills required to use such tools.
A clear explication of the various modeling modes makes it possible to
design instructional sequences to promote the acquisition of coordinated modeling
skills. We address that issue in the next section.
Model construction
Modeling begins with a physical situation which we wish to understand.
Understanding comes from making, analyzing and evaluating a model for the
situation. The situation may be presented to us in any form; for example, as a lab
experiment or a textbook problem. Whatever the situation, we may need to
disregard some information as irrelevant or introduce additional information (e.g.
relevant physical knowledge). From examining the situation, our first problem is
to come up with (i.e. adduce) a suitable model in any way we can. Sometimes we
can simply adapt a familiar model to the new situation; otherwise the problem is
more difficult. In any case, our understanding of what constitutes a model (Box 4)
is a sure guide for what to do.
On examining the situation the first thing to do is "Identify the system!"
This is old advice known to every experienced science teacher, but it neglects to
say what comes next. Box 5 tells us that we should next represent the system we
have identified in a system schema. Then we will have specified precisely what it
is that we intend to model.
Old advice tells us next to "Identify the variables!" Strictly speaking, this

13
 Model Construction
adduce
Situation Model

represent
Identify the system System schema

Identify the variables:

Geometric properties Geometric structure
Interactions Interaction structure
Variable properties State variables
Temporal structure

BOX 7.
advice should be phrased "Identify the properties of interest and specify variables
to represent them." More specifically, as indicated in Box 7, the advice should be
expanded to "Examine the geometry, interactions and variable properties of the
system in order to represent their structure faithfully in the model." As explained
below, the quality of the resulting model depends, in part, on the quality of the
available modeling tools.
An example of the detailed model resulting from this model construction
process is laid out in Box 5 for the situation presented there.
Model analysis
When a model has been constructed it must be analyzed to understand its
structure and implications. The extraction of information or implications from a
model is called model-based inference. All scientific predictions are inferences
from models. An ordinary city map is a descriptive model, and using it to
determine a route between two places is model-based inference. This can be
regarded as a non-propositional form of deductive inference. The analysis of
explanatory models can be very involved, even when the formulation of the model
is quite simple. Some physicists devote their careers to model analysis.
For an example of model-based inference in mechanics, consider the
constant acceleration model in Box 6, where the temporal structure is represented
explicitly. Every question that can be asked of the model can be answered by
inference from the two-time(two-point) parallelogram or the equivalent pair of
vectorial equations. Details are given by Winans (17), so we can concentrate on
general observations here. The representation in Box 6 should be compared with
the coordinate-based representations in most textbooks. Not only is it more
compact, but it is a better starting point for inferences.

14
 For example, a standard textbook problem is to calculate projectile range
from given initial conditions. The textbooks derive a "range formula" which most
students try to memorize instead of recognizing that the method of derivation is
more important than the result. Indeed, the range formula applies only to
horizontal displacements, and calculating the range to targets which are not on the
horizontal is unnecessarily complicated by coordinate methods. Winans (17)
shows how to get the general solution by graphical or trigonometric methods
which are accessible to high school students. Essentially, it is the geometric
method of "solving a triangle" for sides or angles, or rather, "solving a rectangle"
(which is composed of several triangles). Actually, the most efficient way to do
this is by using vector algebra, though that is not ordinarily appropriate for high
school students, except in a second year Advanced Placement course. For
example, from the vectorial equations for the parallelogram in Box 6 we can use
the "inner product" to obtain ( v + v 0 ) ( v v 0 ) = (2r / t ) (at ) , hence
v 2 v 0 2 = 2r a . This formula relates sides of the parallelogram to the lengths
and included angle of the diagonals. It illustrates the fact that the ease of inference
depends on the representational and computational tools at ones disposal.
When vector multiplication is not available, the vectorial equations (in
Box 6) must be decomposed into components in order to use elementary algebra
to "solve them." Even so, the vectorial equations are useful because of their direct
relation to the motion map and the two-time parallelogram, and these display the
geometric structure of the model without irrelevant features of a coordinate
system. Coordinate methods are overused in introductory physics. The only
introductory textbook that consistently puts vectors ahead of coordinates is
Alonso and Finn (1), but it is for calculus-based physics. A more advanced
textbook by Hestenes (4) shows how to do all of classical mechanics without
coordinates, but this requires new and more powerful rules for multiplying
vectors.
Model validation
Validation is the process of assessing the adequacy of a model to represent
a particular system and situation. It involves comparing the structure and
predictions of the model to empirical data about the structure and behavior of the
system it represents. If a match is confirmed to some level of accuracy the model
is said to be validated to that degree. A perfect match is not to be expected,
because every model is an incomplete representation, that is, there are always
some features of the system which are not represented by the model. Moreover,
the experimental apparatus for measuring a system's properties always has a
limited accuracy. A full assessment of a model's validity will include an
assessment of what is neglected in the model as well as the accuracy of the
representation.

Problem solving
Traditional physics courses lay heavy emphasis on problem solving. This
has the undesirable consequence of directing student attention to problems and

15
their solutions as units of scientific knowledge. Modeling theory tells us that these
are the wrong units; the correct units are models. Problem solving is important,
but it should be subservient to modeling. Indeed, the various modeling modes
raise a variety of problems, so problems can be classified according to their roles
in modeling processes. It tells us that most physics problems are solved by
constructing or selecting a model from which the answer to the problem is
extracted by model-based inference. In a profound sense the model provides the
solution to the problem. In fact, a single model solves many problems, and a
handful of models suffice to solve most of the problems in introductory physics
(e.g. Box 3). Thus, an emphasis on models and modeling simplifies the problem
of organizing a physics course into understandable units.
Educational research has established that students main difficulty in
problem solving is in the preliminary "qualitative analysis" of a problem. We
maintain that this qualitative analysis is best described as "setting up a model"
from which the solution to the problem can be inferred. Besides providing a
universal approach to problem solving, this view directs attention to the crucial
role of representational tools in problem solving, or better, in modeling.
Reif (12) observes that even physics professors often fail to display good
technique in solving elementary problems, and he recommends a prescriptive
approach to teach optimal methods. We would modify his recommendations to
prescribing a modeling approach to problem solving with emphasis on skill in
using representational tools, to which we direct our attention next.

Modeling tools
Since models are representations of structure, we need modeling tools to
construct the representations. Teachers are familiar with standard representational
tools verbal, mathematical, graphical and others. However they need to pay
greater attention to the critical role of representational type and structure in
developing physical understanding. It should be recognized that one's ability to
model, and so to understand, depends on the tools at one's disposal. Therefore,
teachers must be prepared to introduce new tools when students are ready to make
good use of them. Instruction must be designed to develop student skills in
wielding the tools.
We have seen that specialized modeling tools have been developed to
represent different kinds of structure in models: system schema for system
structure, motion maps for temporal structure, force diagrams for interaction
structure. Motion maps and force diagrams have proved their value in years of
modeling instruction and in the classrooms of other physics teachers. However,
the instructor must take care to see that students learn to use these tools properly,
for students, and, as we have seen in our Workshops, most teachers do not
immediately see their value and use (5).
Among other modeling tools that have not been mentioned already, energy
diagrams and bar charts are especially valuable for representing interaction
structure, along with state transition diagrams for temporal structure. Graphs, of
course are valuable for revealing structure in data by representing it in geometric

16
form where it can be apprehended as geometric structure. In fact the very notion
of structure is derived from geometry. However, graphs are sometimes used where
other tools would do better. For example, motion maps are often more informative
than kinematic graphs, though many textbooks do not exploit them systematically,
and students frequently get the two confused.
It is obvious that progress in experimental science depends heavily on the
development of more powerful instruments, from electron microscopes to radio
telescopes. It is seldom noted, though, that progress in theoretical science is
equally dependent on the development of more powerful modeling tools. It is not
sufficient to call on mathematics to supply the tools. Each science needs new tools
tailored to its own tasks. This is most evident in the development of specialized
maps and diagrams peculiar to each science.
The heavy dependence of scientific practice on the available tools is not
reflected in the typical science curriculum, which fails to discriminate not only
between models and theories, but also between facts, models and the tools that
produce them. No wonder that students come to view science as a motley
collection of facts.
The modeling method aims to change that view by making models and
modeling tools as explicit as possible in the curriculum, and by empowering
students to understand and apply science by using these tools. The modern
personal computer has immense potential to facilitate these objectives. It is a
representational medium of unprecedented capacity and flexibility. Modeling
theory can serve as a guide to exploiting this medium in the future, a guide for the
design of computer modeling tools with unprecedented power. Some tools can be
created easily by adapting existing software. Others will have to be designed
expressly to facilitate particular modeling tasks. As major developments in this
direction are to be expected in the near future, a brief overview may help us to
exploit them.
From our discussion of modeling modes it is evident that computer
modeling tools fall naturally into two classes:
(a) Tools for constructing and analyzing models. Of course, standard
verbal, mathematical, graphical and diagrammatic tools have already been
implemented in software. More significant are the possibilities for new
representational modes that cannot be realized without computers. One that is
already receiving a lot of attention is computer simulation. Another is the
extension of graphs to 2D and 3D visualization of functional relations. Computers
can make diagrams as easy to manipulate as mathematical symbols. They can also
generate smart symbols and equations, which carry mathematical syntax along
with them. These last two possibilities have seldom been noted, let alone
exploited.
(b) Model validation tools are used to compare the predictions of given
models to empirical data. These include data representation such as spread sheets,
graphs and maps, which are universally recognized as superior to old manual
representations. These general purpose tools have been further specialized into
more powerful tools with already proven pedagogical value in physics instruction;

17
for example, Microcomputer-Based Laboratory (MBL) tools that combine data
acquisition with real time graphing. Such tools have already been incorporated
into the modeling method.

8. Planning and Managing Instruction

The main objective of modeling instruction is to develop student skills in
making and using models to understand the physical world. The fairly abstract
analysis of structure in models and modeling activities in preceding sections is an
essential prelude to the design of effective instruction. For we must know the
structure of knowledge in order to teach it. The design of instruction, however,
involves many other factors, especially an understanding about how students
learn.
The modeling view is that students learn best from activities that engage
them in actively constructing and using structured representations to make sense
of their own experience and communicate with others. To optimize learning, the
activities must be carefully planned and managed by the teacher. This requires
considerable knowledge and skill on the part of the teacher which our Workshops
are designed to cultivate. Here we can only sketch some of the main
considerations. A list of instructional guidelines is given in Box 2.
The instructional modeling cycle (Box 2) is designed to teach scientific
inquiry. It is similar to other teaching methods in presenting to students
interesting physical phenomena to investigate collaboratively. It differs in the
structuring and guiding of activities by the teacher. From the beginning of a cycle
the teacher engages the student in giving an accurate description of the
phenomena in question identifying the relevant systems, components and
variables. Students are guided to an investigation of structure in the phenomena,
and, as needed, the teacher introduces appropriate tools for representing that
structure. For example, one of the best ways to learn kinematics is by using MBL
tools to investigate motion. These tools produce real time graphical
representations, which facilitate immediate comparison of the graphs with the
motion. The teachers job is to help students see such representations as models
of the motion and understand how they can be converted to reproducible (best fit)
mathematical models. It is well established that, without suitable teacher
guidance, the technological benefits of MBL are lost on the students (Hake, 3).
Likewise, teacher guidance is essential throughout the modeling cycle. Gradually,
over several cycles, the students are led to see that the outcome of every
investigation is a model which must be clearly delineated and evaluated. They
gradually internalize the steps in conducting an investigation (procedural
knowledge about science), so they become progressively more independent of the
teacher. Soon they are ready for deeper insights into physics, as comes from using
Newtons laws to model just about anything.
A typical modeling cycle is at least two weeks long with about a week for
each phase. The mechanics course is organized into a sequence of modeling
cycles built around the models listed in Box 3. There are two phases in the

18
modeling cycle: development and deployment. The development phase is devoted
to constructing and evaluating a model for some physical system. Typically it
involves the design and execution of an experiment followed by an oral
presentation and critique of results. The deployment phase involves applying the
model in a variety of new situations to explain, predict, design or control. Problem
solving is incorporated in this phase, often with class discussion and analysis.
The instructional modeling cycle is not intended to duplicate cycles of
model building and testing in research, but it does incorporate the various modes
of modeling in a systematic way that reflects the procedural structure of scientific
inquiry. The modeling cycle is thus a plan for organizing instruction into
structurally coherent units. It is a flexible plan that can be adapted to various
instructional objectives. For example, It is not necessary to begin every cycle with
an experimental investigation. Alternatively, the cycle could begin with the
presentation of a model to be analyzed, or with a design problem. The beginning,
however, is always a situation to be attacked with modeling skills and tools.
A detailed example of a modeling cycle is described in Wells et. al. (16).
That cycle began with an investigation of the modified Atwood machine. As
described there, the students do not get involved in modeling the Atwood machine
with the sophistication and detail of Box 5. However, sometime after that cycle
when the students have been introduced to the whole set of Newtons laws, and
have expanded their repertoire of modeling tools, it would be a good idea to
revisit the Atwood machine with an analysis of the model structure as laid out in
Box 5. This illustrates how to deepen student understanding by using new
conceptual tools for a richer analysis of previous experience.
The most critical element in successful implementation of the modeling
method is the skill of teacher in managing classroom discourse. The modeling
method aims to promote skill in scientific discourse by engaging the students
repeatedly in discussions and presentations about modeling real systems. The
expert teacher moderates student discourse carefully to insure that its quality, as
judged by scientific standards, continues to improve. The first standard governs
the quality of descriptions. Students learn to replace vague descriptive terms of
ordinary language with precisely defined scientific terms like system, interaction,
velocity and force. And to articulate reasons for doing so. They learn to
coordinate these terms with increasing skill in generating coherent scientific
descriptions, including the specification of system schema and descriptive
variables. Further, standards of scientific discourse govern the quality of
inference, explanation and argumentation. Such standards are implicit in discourse
within the scientific community, but they are seldom articulated explicitly, and
there is no consensus as to what they are.
A primary objective of scientific discourse is to formulate and evaluate
scientific claims. This requires (i) models for precise formulation of claims, (ii)
methods for analyzing the claims, and (iii) data for evaluating the claims. It is a
process of constructing arguments and evaluating evidence to produce justified

19
belief. Developing student skills in this process may be the most important
objective of science instruction.
The first step in promoting classroom discourse is to establish a climate of
openness and respect in which every student feels encouraged to participate.
Special measures may be required to engage reluctant or reticent students.
Personal opinions must be solicited so that they can be examined and evaluated
objectively and publicly. Participation is enhanced when students have a personal
stake in the course and outcome of the discourse. Strong feedback from peers and
teacher induces students to clarify, defend or modify their beliefs.
Effective communication requires shared meaning. Section 2 reviewed
conclusive evidence that students and teacher do not attach the same meanings to
technical terms like force at the beginning of a physics course and often at the
end. Meanings (of words, equations, diagrams) are constructed from situated use
by each student individually. Therefore, to establish a common understanding of
important terms and representations their meanings must be negotiated in open
discourse. This is essential to give every student access to the intellectual content
of the course.
There are many modes of discourse, including whole class and small
group discussions, reviews and summaries, prepared presentations and critiques.
The teacher decides what mode is most conducive to the objectives of the day.
The teacher establishes the subject of discourse, guides it toward the objective and
seeks to bring it to a satisfying closure.
The objective of discourse may be to formulate or analyze some target
model or representation. Discourse is most effective when it is centered on some
artifact and/or melded with some activity. This is called situated discourse. An
excellent example is given by Roth (14). He challenged his class to a tug of war
using a block and tackle (giving him the mechanical advantage, of course). The
students became greatly excited and animated. He easily induced them to sketch
and debate a flurry of diagrams to explain his advantage until they reached a
consensus on the explanation. This consensus could not have been achieved
without the public representation of structural relations in the diagrams. This is a
good example of how the quality of discourse and the structure of arguments
depend on the representational tools. Scientific argumentation arises naturally
when the students have the modeling resources. This must be engineered by the
teacher.
9. Conclusions
The National Research Council has recently published National Standards
for K-12 science course content and for the professional development of teachers
(11). The program described in this paper is fully in accord with those standards
and exceeds them in some respects.
The main message of this paper is that:
Informed teacher planning and guidance is the key to effective
instruction. (A good physics course cannot be canned in a textbook
or curriculum materials.)

20
 To learn the structure of physics, students must be continually engaged
in making and using representations of structure (models) for specific
physical systems and processes. Public presentation, examination and
critique of the models is the key to consolidating understanding.
The course described in this paper has emerged from a program of physics
education research and active collaboration with high school teachers. Such
collaboration is essential for effective implementation of research findings and a
powerful incentive for new research. Consequently, the course is under continual
development. Indeed, some of the specifics described in this paper have not yet
been implemented, and more is in the works.

Acknowledgment: I am indebted to my collaborators on the Modeling Workshop Project Larry

Dukerich, Ibrahim Halloun, Jane Jackson and Gregg Swackhamer for stimulating and
contributing many ideas in this paper, and especially for making the project a success. This work
was supported by a grant from the National Science Foundation.

References
(1) Alonso, M. and Finn, E., Physics, Reading Addison-Wesley, 1992.
(2) Clement, J. & Camp, C., Preconceptions in Mechanics, Dubuque, Kendall/Hunt, 1994.
(3) Hake, R., Interactive-engagement vs traditional methods: A six-thousand-student survey of
mechanics test data for introductory physics courses. Am. J. Phys. (submitted).
(4) Hestenes, D., New Foundations for Classical Mechanics, Dordrecht/Boston, Kluwer, 1986,
3rd printing with corrections, 1992 (to be republished by STAR Publ., Cambridge, 1996).
(5) Hestenes, D., Toward a Modeling Theory of Physics Instruction. Am. J. Phys. 55: 440454,
1987.
(6) Hestenes, D., Modeling Games in the Newtonian World, Am. J. Phys. 60: 732748, 1992.
(7) Hestenes, D., and Jackson, J. , These Proceedings (1996).
(8) Hestenes, D., Wells, M. & Swackhamer, G., Force Concept Inventory, The Physics Teacher
30: 141158, 1992.
(9) Johnson, M., The body in the mind, Chicago, U. Chicago Press, 1987.
(10) Lakoff and Johnson, Metaphors we live by, Chicago, U. Chicago Press, 1980.
(11) National Research Council, National Science Education Standards, Wash. DC, National
Academy Press, 1996.
(12) Reif, F. , Understanding and teaching important scientific thought processes. Am. J. Phys. 63:
1732, 1995.
(13) Reif, F. , Standards and measurements in physics Why not in physics education?
Am. J. Phys. 64: 687688, 1996.
(14) Roth, W-M. , Drawing Diagrams, Waving Hands & Talking Pulleys: Learning to Theorize
Simple Machines. Interactive Learning Environments (to be published, 1996).
(15) Simon, H. & Larkin, J., Why a Diagram is (Sometimes) Worth Ten Thousand Words.
Cognitive Science 11: 6599, 1987.
(16)Wells, M., Hestenes, D. & Swackhamer, G., A modeling method for high school physics
instruction, Am. J. Phys. 63: 606619, 1995.
(17)Winans, J., Trajectories without trigonometry, The Physics Teacher (Sept.) 341343, 1971.

21