Shifting Representational Infrastructures and Reconstituting Content to Democratize Access to the Math of Change & Variation: Impacts On Cognition,

Curriculum, Learning and Teaching1 [10/16/00 Draft] James J. Kaput Department of Mathematics University of Massachusetts-Dartmouth Jeremy Roschelle Center for Technology and Education SRI International ABSTRACT In order to set the basis for examining potential impacts on teacher education, this informal essay tracks the impact of (a) deep and historic shifts in representational infrastructure, from formal character string-based algebraic infrastructure towards visually definable and editable functions, (b) new dynamic change-visualization tools and learning environments that support direct links among mathematical notations, simulations, and support physical dataimport/export tools. In particular, we will examine how these ingredients affect (1) the nature of traditional mathematics of change content, (2) student thinking and learning of both old and new content, particularly by tapping more deeply into students' cognitive, linguistic and kinesthetic resources, (3) curriculum structure taken as given for centuries, and (4) appropriate pedagogies. Illustrations will be drawn from work in the authors' ongoing SimCalc Project, which builds and tests software simulations, related visualization tools, and

This material is based upon work supported by the National Science Foundation under Grant No. 9619102 &

0087771. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.

curriculum and teacher-support materials intended to render more learnable and teachable the ideas underlying calculus beginning in the early middle grades. We will reflect on how such technologies can change the experienced nature of the subject matter and alter assumptions regarding the appropriate structure of curriculum that have been unchallenged for centuries. We will also reflect upon teacher learning of newly reconstituted content, including learning in the newer classroom contexts of networked diverse platforms running parallel software. Introduction: The Larger Historical Perspective
Why History? We devote space to the historical basis of our current situation because it helps us understand the depth of existing curricular assumptions and their connections with other conditions of mathematics and science education that, while taken as given, deserve examination. This is especially the case in design of teacher education programs whose client-products will be teaching students who will be working well into the 22nd century. Put differently, design of teacher preparation needs to be more forward-oriented than almost any other educational design. But, of course, in its current forms, it is among the most conservative aspects of our educational system, optimized for stasis, not change. Fitness of aspiring teachers is almost always taken to be the extent to which they can fit into the existing system of education. And their preparatory educational experiences typically take the given curricular structures and teaching practices as the starting point, to be improved, but not fundamentally changed. As we hope to make apparent, the changes in representational infrastructures that provide newly visual and dynamic access to core mathematics and that are at the heart of our work involve foundational reconstitution of the content itself. Such a reconstitution forces re-examination of most target issues of this meeting as well as some that are not, e.g., assessment and accountability. The Shift From Static, Inert Media to Dynamic, Interactive Media The systems of knowledge that form the core of what was taught in schools and universities in the 20th century were built using some representational infrastructures that evolved (e.g., alphabetic and phonetic writing) and others that were somewhat more deliberately designed, mainly by and for a narrow intellectual elite (e.g., operative algebra). In all cases they were instantiated in and hence subject to the constraints of the static, inert media of the previous several millennia. But the computational medium is neither static nor inert, but rather, is dynamic and interactive, exploiting the great new advance of the 20th century, autonomously executable symbolic processes that is,

operations on symbol systems not requiring a human partner (Kaput & Shaffer, in press). We see three profound types of consequences: Type 1: The knowledge produced in static, inert media can become knowable and learnable in new ways by changing the medium in which the traditional notation systems in which it is carried are instantiated for example, creating hot links among dynamically changeable graphs equations and tables in mathematics. Type 2: New representational infrastructures become possible that enable the reconstitution of previously constructed knowledge through, for example, the new types of graphs and immediate connections between functions and simulations and/or physical data of the type developed and studied in the SimCalc Project to be described below. Type 3: The construction of new systems of knowledge employing new representational infrastructures for example, dynamical systems modeling or multi-agent modeling of Complex Systems with emergent behavior, each of which has multiple forms of notations and relationships with phenomena, as discussed among several of the researchers at this meeting. Tracing any of these complex consequences is a challenging endeavor, particularly since they overlap in substantive ways due to the inherent ambiguity in attempts to characterize knowledge apart from the means by which it is represented and used. Hence we will limit our discussion to a few cases close to our recent work in the SimCalc Project involving the Mathematics of Change & Variation (MCV), of which a subset concerns the ideas underlying Calculus. Thus we will be focusing on a Type 2 change. The Case of Calculus and Its Supporting Representational Infrastructures While the Greeks, most notably Archimedes, whose extraordinary computational ability compensated for the weaknesses of the available representational infrastructure in supporting quantitative computation, developed certain mainly geometric ideas and techniques, the Mathematics of Change and Variation leading to what came to be called Calculus evolved historically beginning with the work of the Scholastics in the 1300s through attempts to mathematize change in the world (reviewed in Kaput, 1994). The resulting body of theory and technique that emerged in the 17th and 18th centuries, cleaned up for logical hygiene in the 19th , is now institutionalized as a capstone course for secondary level students in many parts of the world, and especially in the United States. These ultimately successful attempts were undertaken by the intellectual giants of Western civilization, who also developed the representational infrastructure of algebra, including extensions to infinite series and coordinate graphs, as part of the task. Their work led to profoundly powerful understandings of the different ways quantities can vary, how

these differences in variation relate to the ways the quantities accumulate, and the fundamental connections between varying quantities and their accumulation. These efforts also gave rise to the eventual formalization of such basic mathematical ideas as function, series, limit, continuity, etc. (Boyer, 1959; Edwards, 1979). Over the past two+ centuries this communitys intellectual tools, methods and productsthe foundations of the science and technology that we utterly depend uponwere institutionalized as the structure and core content of school and university curricula in most industrialized countries and taken as the epistemological essence of mathematics (Bochner, 1966; Mahoney, 1980). The resulting historically privileged algebraic notation system for representing quantitative relationships affords quantitatively coherent transformations, combinations and comparisons of characterstrings, usually representing closed-form descriptions of functions or relations. Consequences for Todays Curricular Structure The algebraic techniques developed by the masters in the 17th and 18th centuries to model rates of change and accumulations of variable quantities have remained at the heart of the modal calculus course to this day. The requirements of that modal calculus course govern the prerequisite structure of much of students experience with earlier mathematicswhether or not those students are among the 10% or fewer who will study calculus. Importantly, however, the intellectual triumphs that yielded the web of concepts and technique at the heart of that course occurred largely without regard to learnability outside the community of intellectual elite involved. Even Leibniz, whose carefully crafted notations we utilize today, engaged only his peers in his notation design decisions (Edwards, 1979). Furthermore, mastery of the algebraic prerequisites became the measure by which academic success was defined. Mastery of these algebraic tools became the gateway to all that academic success offers, more often than not perpetuating social class structure that advantages some students above others in access to these prerequisites. The fact that the basic curricular structures set down in textbooks by L'Hopital, the Bernoulli's, Euler, and their contemporaries, have remained largely invariant through the 20th century is not merely a matter of inertia, because these structures served traditional purposes and populations extremely well. Indeed, this basic intellectual material is at the foundation of our civilization's scientific and technological infrastructure that we now regard as natural as the earth and sky. While its educational forms evolved into an almost sacred academic tradition (MacLane, 1984) as a capstone course for which much of the traditional quantitative curriculum could be regarded as preparation, the ambient societies, the nature of education, and the relations between education and the larger society, changed and continue to change profoundly.

More specifically, as the 20th Century came to a close, the received semiotic constraints have been overcome by the affordances of dynamic and interactive media, and socio-economic conditions have changed so that now the key ideas underlying calculus must be learned by the great majority of the population, not merely a technically-oriented elite drawn from a demographically narrow and mainly economically advantaged population. A third profound change, the Type 3 change, is a shift in the nature of mathematics and science towards the use of computationally intensive iterative and visual methods that enable entirely new forms of dynamical modeling of nonlinear and complex systems previously beyond the reach of classical analytic methodsa dramatic enlargement of the MCV that will continue in the new century (Kaput & Roschelle, 1998). Despite these profound historic changes, less than 10% of the school population actually completes the capstone course wherein the key MCV ideas are developed, the curriculum remains organized around preparation for this course written in the classic algebraic language, most reform efforts, including calculus reform (see below) continue to take most of these conditions as given, and the newer MCV is virtually unrepresented in school curricula. Since the institution of education is deeply connected internally and closely reflects the assumptions and structures of the larger society, these received givens of content and curriculum define expectations across all aspects of education: assessment of progress and competence of students, teachers, schools, districts, and even countries; they define teacher preparation, both pre- and in-service, as well as technology support of education for learning, for teacher development, for connections between school and other resources, and so on. These historically rooted expectations, built into the fabric of our society and ways of thinking about mathematics science, technology, and education, illustrate the context and challenge of deep reform. University Calculus Reform An Illustration of a Type 1 Reform In the United States these changes, especially technologically-driven changes, and resulting educational ill-fit with traditional forms, led to a major university-centered "Calculus Reform Movement" (Tucker, 1990). However, these reforms had two basic characteristics that our current work, described below, does not share: (1) they were university-centered, intending to reform the teaching of calculus at the university level without attention to K-12 curricula, and (2) they focused on the use of interactive technologies to facilitate the learnability and use of traditional notation systems, both to manipulate within systems as well as to link between representational systems, especially numeric, graphical and algebraic systems (the traditional "Big Three"). This reform effort is a good example of an effort that does not employ new representational infrastructures, but rather improves use and learnability of the inherited ones. Indeed, almost all functions in school

mathematics continue to be defined and identified as character-string algebraic objects, especially as closed form definitions of functionsbuilt into the technology via keyboard hardware and input.

SimCalc Representational Innovations A Type 2 Change

An Overview of SimCalc MathWorlds and Its Representations In order that the cognitive and design issues are understandable, we will provide an overview of selected aspects of the computer version of our software environment and how it is used in selected curricular contexts. Parallel versions are available for hand-held devices as well. Indeed, we regard the desktop computer versions as a supplement and complement to the more widely accessible versions on hand-held computers. We expect that more than 90% of users will employ the hand-held versions. Visually Constructing Functions: In MathWorlds, by choosing an appropriate icon from the vertical toolbar (shown on the right side of the various screens in the screen shots below) the student or teacher can easily construct a function by concatenating line segments. These can represent rates of change, such as velocity or acceleration (rate of change of velocity) or price, pay or tax rates, or they can represent total amounts, such as position (total displacement) or total amount of money spent, earned or paid in taxes, respectively. The domain variable often is taken to be time, but need not be, as would be the case in, say, a price-per-item rate, where the domain (independent) variable is number of items. For example, we could make a step-wise varying velocity function, where the function appears as discrete steps (constant velocity) as in Figure 1. We could also make up a velocity function using linearly changing segments (constant acceleration segments). The point is that the functions need not be described algebraically. Indeed, many of the functions we create are used to describe situations that would be very difficult to describe algebraically. In addition, however, MathWorlds can accept standard input of most standard algebraically defined and hence globally defined exponential and periodic functions, as well as direct drag-based graphical editing of such functions. Visually Enacted Actions On Functions: One of the great powers of traditional algebraic-like mathematical notations is their support for syntactically coherent actions on the notations representing the functions. One can change their form, compare one with another, combine them, and so on. This was the extraordinary leap that moved mathematics forward in an entirely new way in the 16th and 17th centuries (Bochner, 1966). MathWorlds provides a visual analog of certain actions on functions via direct click-and-drag editing of any segment. For example, a user can drag the top of a rectangular velocity segment as in Fig. 1 higher to make a faster velocity. Or a

user can drag the right edge of the rectangular segment to the right to give the motion a longer duration. Students can also construct a function (or extend an existing one) by adding more segments to the graph. Thus operations on the representation have clear and simple qualitative interpretations. For example, Fig. 1 shows a velocity graph that controls the elevator on the left side of the screen, which will travel at 3 floors/sec for 2 seconds. As indicated in Fig. 4 a linear or piecewise parabolic position function can be constructed using a single piecewise linear velocity segment (where, say, a velocity segment can have zero slope, yielding a linear position graph). In this case, Baby Ducky is controlled by linear position segments (constant velocity) and Momma Ducky is controlled by parabolic position segments (linear velocity). Functions Defined by Sampled Data: MathWorlds provides a range of other function types to complement piecewise or algebraically defined functions. A sampled function type supports continuously varying positions, velocities, or accelerations. These data points can be entered directly with the mouse (by sketching the desired curve, ala Stroup, 1996), from Microcomputerbased Laboratory (MBL) data collection gear (Mokros & Tinker, 1987; Thornton, 1992), or by importing mathematical data from another software package. Motion can also be controlled in realtime through the use of a mouse-driven "velocity-meter" or "accelerator-meter." A typical scenario is pictured in Fig. 3, where one vehicle has its motion given in advance and the second vehicle is controlled by one of the meters in real time. The task might involve following behind the given vehicle at a specified distance, for example. Furthermore, the given motion might be described via a position vs. time graph while the student's feedback on the car that she is controlling might be in terms of a velocity vs. time graph. Here, in Fig. 3 by using the controller on the left to drive the "VW Bug" with a concave up velocity graph, the student is enacting a typically confusing situation involving two cars that begin side-by-side but where one has a concave up velocity graph and the other is to have a concave down velocity graph which crosses the first at a certain point in time. Well-documented student expectations assume that the cars will be adjacent when their velocity graphs are adjacent. By "driving" in such situations and many variations on them, the students come to see not only that this adjacency is not the case, but could never be the case. Fig. 5 illustrates how a sampled function from a motion sensor can drive an actor in a simulationthe "Froggie Dude" character in the bottom of the picture. A student has created a motion physically by moving in front of the motion-sensor, an MBL activity. This data has been uploaded to MathWorlds, and attached to Froggie Dude. Then the student created a series of "Clown" characters and synthetic motions for each using piecewise linear functions. In effect, the student is "leading his own Clown Parade." Note that Fig. 5 shows the parade in progress, so only the first part of the graphs is revealed. For orientation to the different kinds of data and notation

connections possible, see Figure 8, where some notations and phenomena are identified as Inside the computer, and others as Outside. Functions Bidirectionally Linked to Phenomena: Throughout, functions drive motions and other phenomena. And, the other way around. Historically, mathematics has been used to model situations that are apart from the mathematics, where processes of abstraction and idealization are used to mathematize the situation, usually in an iterative way. Simulations provide immediate and controllable connections between the mathematics and cybernetically defined phenomena. The ability to import physical data and integrate the data into simulations tightens by orders of magnitude the connection between the mathematics and the phenomena, both experientially and temporally. Indeed, the time for feedback cycles of phenomenon-adjustment and mathematics adjustment is decreased by orders of magnitude. And the kinesthetic connections between physical actions and immediately visible changes in the model simultaneously opens up new channels for feedback and conceptual change. Shown in Fig. 8 are situations developed by Nemirovsky and colleagues at TERC that reverse the data-import-enaction sequence whereby a student creates a function on the computer and this function, in turn, drives a physical device, such as a car on a track (shown) or a pump filling a tank (Nemirovsky, Kaput & Roschelle, 1998). Hot Links Between Functions and their Integrals (accumulations) or Derivatives (rates of change): These connections, formalized and systematized by Newton and Leibniz, are related by what has traditionally been called the Fundamental Theorem(s) of Calculus. In effect, they say (roughly) that if one starts with a rate description of a varying quantity and forms the accumulation of that quantity (e.g., start with velocity and determine the position), then the rate of change of the accumulated quantity is the same as the original varying quantity and vice-versa. This extraordinarily powerful relationship is at the heart of the power of calculus as a mathematical discipline, as reflected in its title. We have already noted how the traditional curriculum puts calculus as a capstone course at the end of a series of algebraic prerequisites. The SimCalc Project begins with this relationship and builds it into activities and our representations from the very beginning. Hence we built in a link between the two descriptions to serve activities at the outset, where a construction is first done in, say, the rate mode, and then it is revisited in the totals mode, or vice-versa. That is, we frequently treat these two descriptions as providing a second opinion on each major idea, and often put students in the position of controlling one type of graph while either the computer or another student controls the other type. Therefore, instead of treating determining derivatives or integrals as two uni-directional processes, we treat the two kinds of descriptions as a basic relationship. This is possible because of the simultaneous presence and immediacy of the connection afforded by the two kinds of graphs built into the learning

environment. Here all the usual relationships explored in calculus courses through the algebraic medium as procedures (e.g., taking the derivative) that yield products (the derivative function) that are then graphable and comparable with the original function can be dealt with as a side-byside relationship where each is treated as a description (or driver of) the same phenomenon! Hence one can work with slopes of position graphs whose values are heights of corresponding velocity graphs and where each drives the same motion. Furthermore, since they are hot-linked and (if we choose to configure the system to do so), a dragged change to one is immediately and visually reflected in its counterpart and is immediately reflected in the phenomenon at hand. Hence dragging a velocity segment up and down changes the slope of the corresponding position graph up and down, respectively, and the actor in the motion simulation moves faster or slower, respectively, during that segment when the simulation is run. See Figure 10, where two functions are available of each type. We often provide a target function in one mode, say a position function controlling object A, and the students task is to match or otherwise interact with that given motion and description by working with the velocity function for object B. For example, B might follow A, or two actors might need to exchange places in a certain way as illustrated in Figure 10 and explored further in the following lesson-scenario. A Lesson-Scenario Clown & Dude Switching Positions & Eventually Dancing (See Fig. 10) : For concreteness, consider the following, where, in earlier parts of the lesson from which this piece is taken, the students were involved in creating graphs to move Clown and Dude around, switching places at constant speed, coming together and then returning to their original positions, and so on. (Only step-wise constant velocities have been made available here, although other function types could have been.) The Challenge: Clown and Dude are to switch their positions so that they pass by each other to the left of the midpoint between them and stop at exactly the same time. First, walk their motions. Now make a position graph for Clown and a velocity graph for Dude so that they can do this.. The student needs to construct graphs similar to #1 and #2. We have also shown the respective corresponding velocity and position graphs, #3 and #4, which can be revealed and discussed later. Note that the velocity and position graphs are hot-linked, so changes in the height of a velocity segment are immediately reflected in the slope of the corresponding position segment, and vice-versa. Importantly, the activity requires interpretations of positive and negative velocities, and hence signed number arithmetic, as well as the representation of simultaneous position. Later activities in the lesson involve a story-line where Dude is patrolling the area (periodic motion) and Clown gets interested in Dude, follows him at a fixed distance, harasses him, and eventually, they dancewhere the student, of course, is responsible for making the dance.

Determining Mean Values: Fig. 2 shows two velocity graphs, each controlling one of the two elevators (graphs are color-coded to match the elevator that they control). The downward-stepping, but positive, velocity graph typically leads to a conflict with expectations, because most students associate it with a downward motion. However, by constructing it and observing the associated motion (often with many deliberate repetitions and variations), the conflicts lead to new and deeper understandings of both graphs and motion. The second graph in Fig. 2 provides constant velocity and is shown in the midst of being adjusted to satisfy the constraint of "getting to the same floor at exactly the same time." This amounts to constructing the mean value of, or the average velocity of, the other elevator which has the variable velocity. This in turn reduces to finding a constant velocity segment with the same area under it as does the staircase graph. In this case the total area is 15 and the number of seconds of the "trip" is 5, so the mean value is a whole number, namely, 3. It is possible to configure MathWorlds so that all segment endpoints have whole number coordinates - this is denoted and experienced as "snap-to-grid" because, as dragging occurs, the pointer jumps from point to point in the discrete coordinate system. Note that if we had provided 6 steps instead of 5, the constraint of getting to the same floor at exactly the same time (from the same starting-floor) could not be satisfied with a whole number constant velocity, hence could not be reached with "snap-to-grid" turned on. The standard Mean Value Theorem, of course, asserts that if a function is continuous over an interval, then its mean value will exist and will intersect that function in that interval. But, of course, the step-wise varying function is not continuous, and so the Mean value Theorem conclusion would fail as it would if 6 steps were used. However, if we had used imported data from a students physical motion, as in Figure 6, then her velocity would necessarily equal her average velocity at one or more times in the interval. We have developed activities involving a second student walking in parallel whose responsibility is to walk at an estimated average speed of her partner. Then the differences between same-velocity and same-position begin to become apparent. Additional activities involve the two students in importing their motion data into the computer (or calculator) serially and replaying them simultaneously, where the velocity-position distinction becomes even more apparent due to the availability of the respective velocity and position graphs alongside the cybernetically replayed motion. Note how the dual perspectives illustrated in Figure 9 show two different views of the average value situation. In the left-hand graph, we see the connection as a matter of equal areas under respective velocity graphs. In the right-hand graph, we see it through position graphs as a matter of getting to the same place at the same time, one with variable velocity and the other with constant velocity.

Putting Phenomena At the Center, Especially Motion: Underlying all the above illustrations and worth making explicit is the theme of putting phenomena at the center of the enterprise. This is partially served by the graphical approach to piecewise linear functions, which allows richer relations with students' experience of motion. Consider the problem of defining a function that represents the motion of an elevator that will pick up and drop off passengers in a building. While such a function is very difficult to formulate algebraically, it is relatively easy to directly drag hotspots on piecewise linear velocity segments to create an appropriate function. Similarly, defining motion-functions for two characters who are dancing would be extremely cumbersome to do algebraically - cumbersome for younger students in entirely unproductive ways. (Exercise: Write out an algebraic description of the position functions driving Momma and Baby Duckies depicted in Fig. 4). Equally important to drawing upon children's resources is providing opportunities to make necessary distinctions in places where prior knowledge may be poorly differentiated. A classic example is the distinction between slowing down and moving downward (between "going down and slowing down") forced by the step-graph in Figure 2 (Moschovich, 1996). More generally, children have great difficulty distinguishing how much from how fast, (Stroup, 1996). By combining the above capabilities, an enormous variety of activities is possible, few of which have been available to students in ordinary classrooms previously. Before turning to their a few cognitive considerations and curriculum implications, we will summarize the bigger representational picture, since it is as the heart of all the other issues.

1) AN ELEVATOR AT 3 FLOORS/SEC FOR 2 SECONDS

2) THE MEAN VALUE OF THE S TAIRCASE

3) DRIVING TOY C ARS

4) BABY C ATCHES UP TO MOMMA

5) MBL DUDE LEADS A C LOWN P ARADE

6) MIXING KINESTHETIC EXPERIENCE WITH S IMULATIONS

Figure 7 Hot Connection Between Functions and Derivatives or Integrals

Inside

Outside

Notational

"Big Three" & Rate-Totals LBM MBL or mouse

Off-line Notations

Target

Simulations

mouse

Physical Entities (Devices or people)

Figure 8 Multiple Connections Between Phenomena and Models

Figure 9 Math Functions Driving Physical Systems (LBM)

Figure 10 Switching Positions Using Velocity & Position Functions

Summary of SimCalc Representational Changes We summarize the core web of five representational innovations employed by the SimCalc Project, all of which require a computational medium for their realization. The fifth, not discussed above, is mentioned for completeness, but has not been a sustained focus of our work to date. (1) Definition and direct manipulation of graphically defined functions, especially piecewisedefined functions, with or without algebraic descriptions. Included is snap-to-grid control, whereby the allowed values can be constrained as neededto integers, for example, allowing a new balance between complexity and computational tractability whereby key relationships traditionally requiring difficult prerequisites can be explored using whole number arithmetic and simple geometry. This allows sufficient variation to model interesting situations, avoid the degeneracy of constant rates of change, while postponing (but not ignoring!) the messiness and conceptual challenges of continuous change. (2) Direct connections between the above representational innovations and simulations, especially motion simulations, to allow immediate construction and execution of a wide variety of variation phenomena, which puts phenomena at the center of the representation experience, reflecting the purposes for which traditional representations were designed initially, and enabling orders of magnitude tightening of the feedback loop between model and phenomenon.

(3) Direct, hot-linked connections between graphically editable functions and their derivatives or integrals. Traditionally, connections between descriptions of rates of change (e.g., velocities) and accumulations (positions) are usually mediated through the algebraic symbol system as sequential procedures employing derivative and integral formulas but need not be. (4) Importing physical motion-data via MBL/CBL and reenacting it in simulations, and exporting function-generated data to define LBM (Line Becomes Motion), which involves driving physical phenomena, including cars on tracks, using functions defined via the above methods as well as algebraically. (5) We also employ hybrid physical/cybernetic devices embodying dynamical systems, whose inner workings are visible and open to examination and control with rich feedback, and whose quantitative behavior is symbolized with real-time graphs generated on a computer screen. The result of using this array of functionality, particularly in combination and over an extended period of time, is a qualitative transformation in the mathematical experience of change and variation. However, short term, in less than a minute, using either rate or totals descriptions of the quantities involved, or even a mix of them, a student as early as 6th 8th grade can construct and examine a variety of interesting change phenomena that relate to direct experience of daily phenomena. And in more extended investigations, newly intimate connections among physical, linguistic, kinesthetic, cognitive, and symbolic experience become possible. Importantly, taken together, these are not merely a series of software functionalities and curriculum activities, but amount to a reconstitution of the key ideas. Hence we are not merely treating the underlying ideas of calculus in a new way, treating them as the focus of school mathematics beginning in the early grades and rooting them in children's everyday experience, especially their kinesthetic experience, but we are reformulating them in an epistemic way. We continue to address such familiar fundamentals as variable rates of changing quantities, the accumulation of those quantities, the connections between rates and accumulations, and approximations, but they are experienced in profoundly different ways, and are related to each other in new ways. These approaches are not intended to eliminate the need for eventual use of formal notations for some students, and perhaps some formal notations for all students. Rather, they are intended to provide a substantial mathematical experience for the 90% of students in the US who do not have access to the Mathematics of Change & Variation (MCV), including the ideas underlying Calculus, and provide a conceptual foundation for the 510% of the population who need to learn more formal Calculus. Finally, these strategies are intended to lead into the mathematics of dynamical systems and its use in modeling nonlinear phenomena of the sort that is growing dramatically in

importance in our new century (Cohen & Stewart, 1994; Hall, 1992; Kaput & Roschelle, 1998; Stewart, 1990).

A Few Cognitive Considerations

We sought to ground the design of learning activities in a thorough understanding of the experiences, resources, and skills students can bring to the MCV. We initially examined attempts by the Scholastics to mathematize change before algebra was available (Claggett, 1968; Kaput, 1994), and took into account the large literature on students' difficulties with kinematics (McDermott, et al., 1987) and graphs (Leinhardt, et al. 1990). Our aim was to build the ideas to which the more formal algebraic notations conceptually refer, the ideas that they are "about." These key underlying ideas of rate of change, accumulation, the connections between variable rates and accumulation, and approximation, all have forms sensible to young students from diverse populations. We work with students ranging in age from 6 and 7 years to university students. Following the historical lead and recognizing that the language and metaphors of motion are used quite generally to describe change and variation, we focused (although not exclusively) on mathematizing linear motion, particularly by controlling motion simulations in familiar or fanciful situations: elevators, people walking or dancing, cars, duckies on a pond, boats in a river, spacevehicles, and so on (see Figure A). Research at TERC and elsewhere (e.g., the Shell Centre in Nottingham, England during the 1980s) has uncovered the important roles of physical motion in understanding mathematical representations (Nemirovsky et al., in press; Nemirovsky & Noble, in press). In studying their own movement, students confront subtle relations among their kinesthetic sense of motion, interpretations of other objects' motions, and graphical, tabular and even algebraic notations. Our starting criteria were to begin with students' intuitive experience with speed and motion, minimize computational complexity, and yet maintain sufficient variation to avoid the conceptual degeneracy of constant velocity and linear functions (Stroup, 1996). These criteria led to extensive use of piecewise constant velocity functions as shown in Figures (12). Furthermore, we wanted to support direct graphical manipulation of these velocity functions - after all, defining and manipulating piecewise constant functions algebraically is a very cumbersome process, and the vertical arrow in 2 indicates a dragging action to change the height of the velocity graph segment to which it is attached. Yet another major source of design consideration supporting piecewise defined functions, is also based in the work of our colleagues at TERC, who found that children spontaneously engage in "interval analysis" to understand the graphical behavior of a complex mathematical function. Without explicit instruction students parse a graph into intervals based on their understanding of

the events that the graph represents (Nemirovsky, 1994; Monk & Nemirovsky, 1994), where the intervals correspond to identifiable, separable sub-events. Within this framework students understand curved pieces of graphs as signifying behaviors of objects or properties of events, rather than as sets of ordered pairs in a kind of perceptual subitizing of quantifiable events into naturally occurring, pre-quantitatively understood chunks. They also readily constructed more flexible and richer schemes as they made sense of increasingly complex situations and constructed rich mathematical narratives that tell the story of a graph over time (Nemirovsky, 1996). These well-documented student resources directly influenced our focus on piecewise defined and editable functions.

Curriculum Integration Issues: Opportunities and Constraints

Using the MCV to Organize, Contextualize and Energize the Traditional Core Quantitative Curriculum An additive approach to curriculum change is impossiblethe curriculum is already overstuffed. Further, in a standards/accountability environment, one cannot take liberties with the content that is subject to high-stakes assessment. Hence, to complement offerings of alternative curricular materials (available for those who can afford to take alternative approaches), we have taken a transformative approach to curriculum integration. The intent is to enhance the learnability of traditional, but often difficult ideas such as rate, ratio, proportion, variable, slope, linearity, function, simultaneous equations, average, signed numbers and areas, periodicity, linear change (and hence quadratic accumulation), interpretation of graphsall of which appear briefly in the above examples. Our aim has been to organize these ideas in the service of understanding the key ideas of the MCV that gave rise to them historically. This in turn means that the students are simultaneously learning the basic ideas underlying calculus: the different kinds of variation, relations between rates of change of varying quantities and how they accumulate, continuity and approximation. Thus, in reference to the Lesson-Scenario above(see also Figure 10), while the students are making the two characters in the simulation exchange places while crossing to the left of the center, they are needing to deal with signed (positive and negative) areas, the idea of variable rate, simultaneity, and, if one character is driven by a velocity graph and the other by a position graph, they need to coordinate the relationship between the two kinds of descriptionswhich, as noted earlier, is the idea at the heart of the Fundamental Theorems of Calculus.

Furthermore, in engaging in activities that mix physical, simulated and imported motion-data, the students also develop heuristic skills crucial for life and work in the 21st Century: modeling, simulation, the differences between physically and cybernetically generated data, how assumptions play out in models and simulations, etc. At the same time, the use of dynamic simulations contextualizes and energizes these ideas: students are learning the ideas in the context of deepening their understanding of some phenomenon or as they try to design some dynamic event, such as a dance, a catch-up situation where a car on a ramp meets traffic on a highway, an elevator trip to satisfy some constraint, and so on. In addition, we do not limit attention to the Mathematics of Motion because most of the MCV ideas apply much more generally, and indeed, apply even more naturally to piecewise defined functions: consider all sorts of rates with naturally occurring steps, such as tax rates, phone rates, royalty rates, etc. We expect that the representationally enabled curricular innovations will gradually infiltrate the mainstream in the next decade. For the newer MCV involving system dynamics, etc., rather than curriculum design, we have concentrated on understanding what students bring to our dynamical systems exemplars, what kinds of knowledge, representations, and actions are needed to make sense of such systems, and how that knowledge itself is transformed by experiences with such systems.

Teacher Learning of Content and Understanding of Curricular Changes

We have begun to uncover commonalties and subtleties of teacher learning, having done dozens of both pre- and in-service workshops for hundreds of teachers of lengths ranging from 2 hours to 25 hours for teachers ranging from elementary teachers to college level teacher-educators. We have seen that the complexities of the MCV include the needs for deeper understanding of concepts such as rate and ratio (especially middle school teachers, whose understanding tends to be very superficial and formulaic), greater fluency with the range of representational media now possible, and concomitant understanding of the links between and among notations and phenomena (Bowers & Doerr, 1998). In addition to understanding mathematical interactions and experiences that were by and large not part of their own mathematical education, teachers also need to understand learners' conceptual development and hot to create the alternative pedagogical strategies that exploit our tools (Doerr, & Bowers, 1999). For example, to build concepts of rate we have developed sequences of activities directed towards both teachers and students using our ability to provide discretized traces of motion (moving objects drop marks for any specified step-time) (Nickerson, & Bowers, 1999). These can also become the bases for reformulating approaches to algebra,

especially linear functions and interpreting slope as rate of change. See especially Nickerson, et al.(2000). Reflecting the historical dependence on character strings described at the outset, among high school and college level teachers, we sometimes see a reluctance to treat our materials as mathematically seriousin particular, the unfamiliar graphical mathematics of piecewise constant velocity functions, and their two-way connections with polygonal position graphs, is seen as secondary to derivative and integral formulas that apply to globally defined algebraic functions, which embody the real math. Since the fundamentally graphical approach to the MCV is usually unfamiliar to teachers, we continue to design activity sequences for teachers that build new understandings about the relationship between the derivative and the integral. For example, asking such questions as why does a vertical translation in the velocity graph change the position graph, but not conversely? often reveals a new insight into the +C of the familiar integration formulas. Most teachers come to realize that there is much more to this mathematics than derivative and integral formulas, just as slope is much more than rise over run. Indeed, this mathematics is what the formula mathematics is about.

The New Issue of Multiple & Networked Hardware & Software Platforms in Classrooms
Integration of Hand-helds and Larger Computers Given the rapidly evolving universe of hand-helds and networks, any plans for technology use in teacher education need to examine how to engage prospective and in-service teachers in optimizing synergy between hand-held and larger computers, especially where each student has access to a hand-held device capable of running some version of parallel computer software. Teacher s need to be able to utilize a desktop or laptop with classroom display capable of running such software as Java MathWorlds in conjunction with a version running on a popular platform such as the TI-83+, and where classroom connectivity could range from currently available TI GraphLink 1-1 data passing between any 2 devices to a full wireless classroom network, and where the hand-held varies from the decidedly lo-tech but almost universal TI-83+ to wirelessly networked Palm-like devices. We have developed a full, document-oriented Flash ROM software system for the TI-83+ and a core set of activities embodying the curriculum ideas described above that parallels the computer software to the extent possible given the processing and screen constraints (96 by 64 pixels!) . The parallelism is evident in the Calculator MathWorlds screens below in Figure 11. (We have also developed a prototype version of MathWorlds for a PalmPilot.)

Most user interaction is through the softkeys that appear across the bottom of the screen which are controlled by the hardkeys immediately beneath them. The left-most screen depicts the Animation mode, with two elevators on the left controlled respectively by the staircase and constant velocity functions to their right. The right-most screen shows a horizontal motion world with both position and velocity functions displayed (hot-linked, as with the computer software). The middle screen depicts the Function-Edit mode, which shows a hot-spot on the constant-velocity graph. The user adjusts the height and extent of a graph segment via the 4 calculator cursor keys (not shown), and can add or delete segments via the softkeys. Other features allow the user to scale the graph and animation views, display labels, enter functions in text-input mode, generate time-position output data, and so onvery much in parallel with Java MathWorlds, but without the benefits of a direct-manipulation interface.

Figure 11 MathWorlds for the TI-83+ Studies of Classroom Interactions We now ask a critical question: How many of our activity-snippets above can be done in this environment? The answer is almost all of them. Indeed, our core MCV curricula for prealgebra, algebra and precalculus can be executed with this system. Another question: Why sacrifice all the power and visual capability of computers? The small device supports only 2 objects, limited scale, and only schematic one-dimensional motion worlds, and the computer software supports motions along user-defined paths as well as 2-dimensional change enabling richer and more complex activities. But hand-helds offer continual classroom availability, low cost (about a 5th the cost of a computer lab to equip a class including one computer and display) and portability. Hence a rich activity introduced on the teachers computer/display can be followed-up by individual or small-group activity, including homework, on the hand-helds. Increasingly rich interactions are possible as connectivity increases between a teachers computer and a classroom of hand-helds. For example, a teacher can download sets of documents for homework or quizzes, and more interestingly, the students can upload their solution-documents as

well as other data, which can then be aggregated in a variety of ways on the teachers computer. For example, groups of students can act out or choreograph a collective motion, say a dance, collectively, and then sit down to plan the coordination of their individual motions as mathematical functions that they will produce on their hand-held. They then upload their individual functions to the teachers computer where the serially produced motions are aggregated into a simultaneously executed dance to be viewed by the entire class! This amounts to a netwroked version of the Marching Parade activity depicted in Figure 5. Variations of this kind of aggregation activity can use CBL input as well, and a wide variety of other aggregation and target activities is possible. In early prototype testing , we found subtle perceptual carryovers from the computer to the calculator environments that may provide guidance on how to exploit the visual detail possible on the computer screen to compensate for limited screens of hand-helds. For example, despite the hard to read grid of the calculator screen, the students, who were often presented activities using graph printouts based on the computer screens, seemed to treat the calculator screen as having visual attributes that were present only on the computer software. These kinds of potentially important phenomena need to be studied and documented in more detail, as do potential interference effects across the different environments. We are currently pursuing research with several private sector partners, including Texas Instruments and Palm, to examine the affordances and constraints of networked mathematics classrooms employing mixes of hardware and software platforms. Of particular concern are issues of implementability and teacher knowledge, content knowledge as well as pedagogical knowledge and how these interact with the various technological options available. These results will have a direct bearing upon the design of pre- and in-service experiences for teachers.

References
Various versions of MathWorlds for computers and calculators , along with other articles and materials, can be downloaded from the SimCalc web site, http://www.simcalc.umassd.edu Bochner, S. (1966). The role of mathematics in the rise of science. Princeton, NJ: Princeton University Press. Bowers, J. S. & Doerr, H. M. (1998, October). Investigating teachers insights into the mathematics of change. In S. Berenson, K. Dawkins, M. Blanton, W. Coulombe, J. Kolb, K. Norwood, & L. Stiff (Eds.), Proceedings of the Twentieth Annual Meeting of the North American Chapter of the International Group for the Psychology of Mathematics Education (Vol. 2, pp. 789-795), North Carolina: North Carolina State University. Boyer, C. (1959). The history of calculus and its historical development. New York: Dover Publications.

Claggett, M. (1968). Nicole Oresme and the medieval geometry of qualities and motions, Chapters I & II. Madison, WI: University of Wisconsin Press. Cohen, J. & Stewart, I. (1994). The collapse of chaos: Discovering simplicity in a complex world. New York: Viking Books. Confrey, J. (1991). Function Probe [software]. Ithaca, NY: Department of Education, Cornell University. Doerr, H. M. & Bowers, J. S. (1999). Revealing pre-service teachers thinking about functions through concept mapping. In F. Hitt & M. Santos (Eds.), Proceedings of the Twenty First Annual Meeting of the North American Chapter of the International Group for the Psychology of Education, (Vol. 1, pp. 364-369). Duckworth, E. (1991). Twenty-four, forty-two, and I love you: Keeping it complex. Harvard Educational Review, 61(1), 124. Edwards, C. (1979). The historical development of the calculus. New York: Springer-Verlag, Inc. Hall, N. (1992). Exploring chaos: A guide to the new science of disorder. New York: Norton. Kaput, J. (1994). Democratizing access to calculus: New routes using old roots. In A. Schoenfeld, (Ed.), Mathematical thinking and problem solving. Hillsdale, NJ: Erlbaum. Kaput, J., & Roschelle, J. (1998). The mathematics of change and variation from a millennial perspective: New content, new context. In C. Hoyles, C. Morgan, & G. Woodhouse (Eds.), Mathematics for a new millennium (pp. 155170). London: Springer-Verlag. Kaput, J. & Roschelle, J. (1996). Connecting the connectivity and the component revolutions to deep curriculum reform (http://www.ed.gov/Technology/Futures/kaput.html). Washington, DC: Department of Education. Kaput, J. & Shaffer, D. (in press). On the development of human representational competence from an evolutionary point of view: From episodic to virtual culture. In K. Gravemeijer, R. Lehrer, B.van Oers, & L. Verschaffel (Eds.), Symbolizing, modeling and tool use in mathematics education. London: Kluwer. Leinhardt, G., Zaslavsky, O., & Stein, M. (1990). Functions, graphs, and graphing: Tasks, learning, and teaching. Review of Educational Research , 60, 164. MacLane, S. (1984). Calculus is a discipline. The College Mathematics Journal, 15(5), 375. Mahoney, M. (1980). The beginnings of algebraic thought in the seventeenth century. In S. Gankroger (Ed.), Descartes: Philosophy, mathematics and physics. Sussex, England: Harvester Press. McDermott, L., Rosenquist, M., & Zee, E. (1987). Student difficulties in connecting graphs and physics: Examples from kinematics. American Journal of Physics, 55, 503513. Mokros, J.R. & Tinker, R.F. (1987). The impact of microcomputer-based labs on childrens ability to interpret graphs. Journal of research in science teaching, 24(4), 369383. Monk, S.; Nemirovsky, R. (1994). The case of Dan: Student construction of a functional situation through visual attributes. In E. Dubinsky, J. Kaput & A. Schoenfeld (Eds.), Research in collegiate mathematics education (Vol. 4, pp. 139168. Providence, RI: American Mathematical Society Moschovich, J. N. (1996). Moving up and getting steeper: Negotiating shared descriptions of linear graphs. Journal of the Learning Sciences, 5(3), 239277. Nemirovsky, R. (1996). Mathematical narratives. In N. Bednarz, C. Kieran, & L. Lee (Eds.), Approaches to algebra: Perspectives for research and teaching (pp. 197223). Dordrecht, The Netherlands: Kluwer. Nemirovsky, R. (1994). On ways of symbolizing: The case of Laura and velocity sign. The Journal of Mathematical Behavior. 13, 389422 Nemirovsky, R., & Noble, T. (in press). Mathematical visualization and the place where we live. Educational Studies of Mathematics. Nemirovsky, R., Kaput, J. & Roschelle, J. (1998). Enlarging mathematical activity from modeling phenomena to generating phenomena. Proceedings of the 22nd Psychology of

Mathematics Education Conference in Stellenbosch, (Vol 3., 287-294). Stellenbosch, South Africa. Nemirovsky, R., Tierney, C., & Wright, T. (in press). Body motion and graphing. Cognition and Instruction. Nickerson, S., Nydam, C. & Bowers, J. S. (2000). Linking algebraic concepts and contexts: Every picture tells a story. Mathematics Teaching in the Middle School 6(2), 92-98. Nickerson, S. & Bowers, J. (1999, October). Investigating students developing conceptions of rate. In F. Hitt & M. Santos (Eds.), Proceedings of the Twenty First Annual Meeting of the North American Chapter of the International Group for the Psychology of Education, (Vol. 1, p. 410). Roschelle, J. (1992). Learning by collaborating: Convergent conceptual change. Journal of the Learning Sciences, 2(3), 235276. Roschelle, J. & Kaput, J. (1996). Educational software architecture and systemic impact: The promise of component software. Journal of Educational Computing Research, 14(3), 217228. Roschelle, J., Kaput, J, & Stroup, W. (2000). SimCalc: Accelerating students engagement with the mathematics of change. In M. Jacobson & R. Kozma (Eds.) Educational technology and mathematics and science for the 21st century (pp. 4775). Hillsdale, NJ: Erlbaum. Stewart, I. (1990). Change. In L. Steen (Ed.), On the shoulders of giants: New approaches to numeracy (pp. 183219). Washington, DC: National Academy Press. Stroup, W. (1996). Embodying a nominalist constructivism: Making graphical sense of learning the calculus of how much and how fast. Unpublished doctoral dissertation. Cambridge, MA: Harvard University. Thornton, R. (1992). Enhancing and evaluating students' learning of motion concepts. In A. Tiberghien & H. Mandl (Eds.), Physics and learning environments [NATO Science Series]. New York: Springer-Verlag. Tucker, T. (1990). Priming the calculus pump: Innovations and resources. Washington, DC: MAA.