Cobb 1999

This article was downloaded by: [University of Otago]
On: 04 January 2015, At: 15:00

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number:
1072954 Registered office: Mortimer House, 37-41 Mortimer Street,
London W1T 3JH, UK
Mathematical Thinking and

Learning
Publication details, including instructions
for authors and subscription information:
http://www.tandfonline.com/loi/hmtl20
Individual and
Collective Mathematical
Development: The Case of
Statistical Data Analysis
Paul Cobb
Published online: 18 Nov 2009.
To cite this article: Paul Cobb (1999) Individual and Collective Mathematical
Development: The Case of Statistical Data Analysis, Mathematical Thinking
and Learning, 1:1, 5-43, DOI: 10.1207/s15327833mtl0101_1
To link to this article: http://dx.doi.org/10.1207/s15327833mtl0101_1
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of

all the information (the “Content”) contained in the publications
on our platform. However, Taylor & Francis, our agents, and our
licensors make no representations or warranties whatsoever as to the
accuracy, completeness, or suitability for any purpose of the Content.
Any opinions and views expressed in this publication are the opinions
and views of the authors, and are not the views of or endorsed by
Taylor & Francis. The accuracy of the Content should not be relied
upon and should be independently verified with primary sources of
information. Taylor and Francis shall not be liable for any losses,
actions, claims, proceedings, demands, costs, expenses, damages,
and other liabilities whatsoever or howsoever caused arising directly
or indirectly in connection with, in relation to or arising out of the use
of the Content.
This article may be used for research, teaching, and private study
purposes. Any substantial or systematic reproduction, redistribution,
reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access
and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Downloaded by [University of Otago] at 15:00 04 January 2015
MATHEMATICAL THINKING AND LEARNING, ](I), 5-43
Copyright O 1999, Lawrence Erlbaum Associates, Inc.
Individual and Collective Mathematical

Development: The Case of Statistical
Data Analysis
Paul Cobb
Department of Teaching and Learning
Vanderbilt University
In the first part of this article,I clarify how we analyze students' mathematical reason-
ing as acts of participation in the mathematical practices established by the classroom
community. In doing so, I present episodes from a recently completed classroom
teaching experiment that focused on statistics. Against the background of this analy-
sis, I then broaden my focus in the final part of the article by developing the themes of
change, diversity, and equity.
In recent years, we have seen an increasing emphasis on the socially and culturally
situated nature of mathematical activity. l k s trend encompasses a range of theoret-
ical positions that include sociocultural theory, discourse theory, and symbolic
interactionism. There are, of course, significant differences among these various
perspectives that have, at times, been the subject of intense debate. However, rather
than highlight differences, I focus on a central notion that I believe cuts across these
positions and serves to differentiate them from purely psychological perspec-
tives-that of participation in communal practices. In developing this notion, I
ground the discussion in my own and my colleagues' work in classrooms. My im-
mediate goal is to clarify how we analyze students' mathematical reasoning as acts
of participation in the mathematical practices established by the classroom commu-
nity. In doing so, I present episodes from a recently completed classroom teaching
experiment that focused on statistics. Against the background of this analysis, I
then broaden my focus in the final part of the article by developing the themes of
change, diversity, and equity.
Requests for reprints should be sent to Paul Cobb, Vanderbilt University, Peabody College, Box 330,
Nashville, TN 37203. E-mail: cobbP@ctrvax.vanderbilt.edu
ORIENTATION: DEVELOPMENTAL RESEARCH AND
THE CLASSROOM MICROCULTURE
The type of research that my colleagues and I conduct involves classroom teaching
experiments of up to 1 year in duration (cf. P. Cobb, in press; Confrey & Lachance,
in press; Simon, in press; Yackel, 1995). In the course of these experiments, we
both develop sequences of instructional activities and analyze students' mathemat-
ical learning as it occurs in the social situation of the classroom. Research of this
type falls under the general heading of developmental research (Gravemeijer,
1994) in that it involves both instructional development and classroom-based re-

search. It should, therefore, not be confused with either child development research
or with research into the development of particular mathematical concepts. The ba-
sic developmental research cycle is shown in Figure 1. The first aspect of the cycle
involves developing instructional sequences as guided by a domain-specific in-
structional theory. In our case, we draw on the theory of realistic mathematics edu-
cation developed at the Freudenthal Institute (Gravemeijer, 1994; Streefland, 1991 ;
Treffers, 1987). Gravemeijer has written extensively about the process of instruc-
tional design in developmental research and clarifies that the designer initially con-
ducts an anticipatory thought experiment. In doing so, the designer envisions how
students' mathematical learning might proceed as the instructional sequence is en-
acted in the classroom, thereby developing conjectures about both (a) possible tra-
jectories for students' learning and (b) the means that might be used to support and
organize that learning. It is important to stress that the conjectures are tentative and
provisional, and they are tested and modified on a daily basis during the teaching
experiment. These adaptations and revisions are informed by an ongoing analysis
of classroom events, and it is here that the second aspect of the developmental re-
search cycle-classroom-based analyses+omes to the fore.
I~aUebnal CMs~oom-based
Development Analyses
(guded by domain-specific (guided by inorpretive

instructional design theory) frsmwork)
FIGURE 1 The developmental research cycle.

MATHEMATICAL DEVELOPMENT 7
Interpretations of classroom events reflect suppositions and assumptions about

learning, teaching, and mathematics as well as about the general relation between
individual activity and communal processes. In my own case, for example, my col-
leagues and I initially intended to analyze students' mathematical reasoning in
purely psychological terms when we began working intensively in classrooms 12
years ago. This is not to say that we ignored the role of social interaction in sup-
porting mathematical learning. The classroom sessions in the first teaching experi-
ment that we conducted, in fact, involved small-group work followed by
whole-class discussions of students' mathematical interpretations and solutions.

However, we treated social interaction and discourse as a catalyst for otherwise au-
tonomous mathematical development and did not view them as influencing the
products of learning-increasingly sophisticated mathematical ways of knowing.
Incidents that occurred at the beginning of the first teaching experiment, which
was conducted with 7-year-old students in a second-grade classroom in the United
States, led us to question our sole reliance on an individualistic, psychological ori-
entation. Briefly, the teacher with whom we collaborated expected her students to
engage in genuine discussions in which they explained and justified their mathe-
matical reasoning. However, as a consequence of their prior experiences in school,
the students assumed that their role was to infer the responses that the teacher had
in mind all along rather than to articulate their own interpretations. The teacher
coped with this conflict between her own and the students' expectations by initiat-
ing a process that we subsequently came to term the negotiation of classroom so-
cial norms (P. Cobb, Yackel, & Wood, 1989). Examples of social norms that
became explicit topics of discussion included explaining and justifying solutions,
attempting to make sense of explanations given by others, indicating agreement or
disagreement, and questioning alternatives in situations in which a conflict in in-
terpretations had become apparent. In general, an analysis that focuses on social
norms serves to delineate the classroom participation structure (Erickson, 1986;
Lampert, 1990).These norms therefore constitute a crucial aspect of the classroom
microculture that is continually regenerated by the teacher and students in the
course of their ongoing interactions.
As this brief summary makes clear, our interest in classroom social norms
and, more generally, in the classroom microculture did not arise as an end in it-
self. Instead, it emerged within the context of developmental research as we at-
tempted to further our agenda of supporting students' mathematical learning in
classrooms. It was within this context that we subsequently came to view one as-
pect of our analysis of classroom social norms as inadequate. In particular, we
came to realize that these norms are not specific to mathematics; rather, they ap-
ply to any subject matter area. For example, one might hope that students would
explain and justify their reasoning in science or history classes as well as in
mathematics. We attempted to address this limitation by shifting our focus to
normative aspects of students' activity that are specific to mathematics
8 COBB
(Lampert, 1990; Voigt, 1995; Yackel & Cobb, 1996). Examples of these
so-called sociomathematical norms include what counts as a different mathemat-
ical solution, a sophisticated mathematical solution, an efficient mathematical
solution, and an acceptable mathematical explanation.
As we have noted elsewhere (Yackel & Cobb, 1996), the analysis of
sociomathematical norms has proven useful in helping us understand the process
by which teachers can foster the development of intellectual autonomy in their
classrooms. This issue is particularly significant to us, given that the development
of student autonomy was an explicitly stated goal of our work in classrooms from
the outset. However, we originally characterized intellectual autonomy in individ-
ualistic terms and spoke of students' awareness of and willingness to draw on their
own intellectual capabilities when making mathematical decisions and judgments
(Kamii, 1985; Piaget, 1973). As part of the process of supporting the growth of au-
tonomy, the teachers with whom we have worked initiated and guided the devel-
opment of a community of validators in their classrooms, such that claims were
established by means of mathematical argumentation rather than by appealing to
the authority of the teacher or textbook. However, for this to occur, it was not suffi-
cient for the students merely to learn that they should make a wide range of mathe-
matical contributions. It was also essential that they become able to judge both
when it was appropriate to make a mathematical contribution and what constituted
an acceptable contribution. This required, among other things, that the students
could judge what counted as a different mathematical solution, an insightful math-
ematical solution, an efficient mathematical solution, and an acceptable mathe-
matical explanation. However, these are precisely the types of judgments that are
negotiated when establishing sociomathematical norms. We therefore conjectured
that students develop specifically mathematical beliefs and values that enable
them to act as increasingly autonomous members of classroom mathematical com-
munities as they participate in the negotiation of sociomathematical norms
(Yackel & Cobb, 1996).
It is apparent from this account that we revised our conception of the most indi-
vidualistic of notions-intellectual autonomy-as we worked in classrooms. At
the outset, we defined autonomy in purely psychological terms as a characteristic
of individual students' activity. However, as we developed the idea of
sociomathematical norms, we came to view autonomy as a characteristic of an in-
dividual's way of participating in a community. In particular, the development of
autonomy can be viewed as synonymous with the gradual movement from rela-
tively peripheral participation in classroom activities to more substantial participa-
tion, in which students increasingly rely on their own judgments rather than on
those of the teacher (cf. Forman, 1996; Lave & Wenger, 1991). The example of au-
tonomy is paradigmatic in this regard in that it illustrates the general shift we have
made in our theoretical orientation away from an initial psychological perspective
toward what we call an emergentperspective (P. Cobb & Yackel, 1996).
Thus far, I have described two aspects of the classroom microculture that we
have found useful to differentiate when conducting analyses that feed back to in-
form the ongoing instructional development effort. Our motivation for teasing out
a third aspect of the classroom microculture, classroom mathematical practices,
stems directly from our concerns as instructional designers. Recall that the ap-
proach we take to instructional design involves conducting an anticipatory thought
experiment in the course of which the designer develops conjectures about the pos-
sible course of students' mathematical learning. However, these conjectures can-
not encompass the anticipated mathematical learning of each and every student in
a class, given that there are significant qualitative differences in their mathematical
reasoning at any point in time. Descriptions of planned instructional approaches
written so as to imply that all students will reorganize their mathematical activity
in particular ways at particular points in an instructional sequence are, at best,
highly idealized. It is, however, feasible to view a hypothetical learning trajectory
as consisting of conjectures about the collective mathematical development of the
classroom community. This proposal, in turn, indicates the need for a theoretical
construct that allows us to talk explicitly about collective mathematical develop-
ment. The construct that my colleagues and I have found useful is that of class-
room mathematical practices that are established by the classroom community.
Described in these terms, a learning trajectory then consists of an envisioned se-
quence of classroom mathematical practices together with conjectures about the
means of supporting their evolution from prior practices.
As an initial illustration to clarify the notion of a classroom mathematical prac-
tice, consider the social norm of explaining and justifying interpretations. As I
have noted, this and other social norms deal with facets of the classroom participa-
tion structure that are not specific to mathematical activity. In contrast, the related
sociomathematical norms for argumentation deal with criteria that the teacher and
students establish in interaction for what counts as an acceptable mathematical ex-
planation and justification. For example, a criterion that became established dur-
ing a teaching experiment that focused on place-value numeration was that
explanations had to be clear, in the sense that the teacher and other students could
interpret them in terms of actions on numerical quantities rather than, for instance,
in terms of the mere manipulation of digits (Bowers, Cobb, & McClain, in press).
Because sociomathematical norms are concerned with the evolving criteria for
mathematical activity and discourse, they are not specific to any particular mathe-
matical idea. Thus, the criterion that mathematical explanations should be clear
could apply to elementary arithmetical word problems or to discussions about rela-
tively sophisticated mathematical ideas that involve proportional reasoning.
Classroom mathematical practices, in contrast, focus on the taken-as-shared ways
of reasoning, arguing, and symbolizing established while discussing particular
mathematical ideas. Consequently, if sociomathematical norms are specific to
mathematical activity, then mathematical practices are specific to particular math-
ematical ideas. In the case of the teaching experiment that focused on place-value
numeration, the analysis of mathematical practices focused on the specific argu-
ments and ways of reasoning about quantities that the teacher and students treated
as being clear and beyond further justification. In addition, the analysis described
how each mathematical practice identified in this way emerged as a reorganization
of prior practices, thereby providing an account of both the taken-as-shared under-
standing of place-value numeration that eventually was established in this particu-
lar classroom and the process by which it emerged.
It is apparent from this illustration that analyses of classroom mathematical

practices account for the emergence of what traditionally is called mathematical
content in terms of successive reorganizations of communal processes. This ap-
proach might seem controversial given that, in mathematics education, we typi-
cally view the development of mathematical ideas and concepts as a matter of
individual learning. Thus, although it is now common to acknowledge seemingly
nonmathematical aspects of the classroom microculture such as social norms, we
typically think in terms of individual students' learning when we address issues
that relate directly to mathematical content. In addition, it might seem
counterintuitive to speak of the mathematical learning of the classroom commu-
nity given the diversity of individual students' reasoning at any point in time. To
address these concerns, I present episodes from a recently completed teaching ex-
periment that focused on statistics to illustrate how an analysis of classroom math-
ematical practices characterizes changes in collective mathematical activity while
taking into account the diversity in individual students' reasoning.
BACKGROUND TO THE TEACHING EXPERIMENT
The teaching experiment was carried out with 29 twelve-year-old students in a

seventh-grade classroom in the United States and involved 34 lessons conducted
over a 10-week period.' A member of the project staff served as the teacher for
the first 21 classroom sessions, and two members of the research team shared
the teaching responsibilities for the remaining 13 sessions. The overarching
mathematical idea that served to orient our instructional design effort was that of
distribution. We therefore wanted students to come to view data sets as entities
that are distributed within a space of possible values (Hancock, in press; Konold,
Pollatsek, Well, & Gagnon, 1996; Wilensky, 1997). Notions such as mean,
mode, median, skewness, spread-outness, and relative frequency then would
emerge as ways of describing how specific data sets are distributed within this
space of values. Furthermore, in this approach, various statistical representations
'Kay McClain, Koeno Graverneijer, Jose Cortina, Lynn Hodge, Maggie McGatha, Beth Petty, Carla
Richards, Michelle Stephan, and I conducted the teaching experiment.
or inscriptions would emerge as different ways of structuring distributions. For

example, students who use box plots flexibly to compare data sets are reasoning
about distributions that they have structured multiplicatively. Viewed in this
way, the development of increasingly sophisticated ways of structuring and or-
ganizing data is inextricably bound up with the development of increasingly so-
phisticated ways of inscribing data (Biehler, 1993; de Lange, van Reeuwijk,
Burrill, & Romberg, 1993; Lehrer & Romberg, 1996). In general, this focus on
distribution allowed us to frame our instructional intent as that of supporting the
gradual emergence of a single, multifaceted mathematical notion rather than a

collection of, at best, loosely related concepts and inscriptions.
In preparation for the teaching experiment, we surveyed the relevant research
literature and conducted a series of interviews and classroom performance assess-
ments with seventh graders in the same school in which we planned to work. A
broad distinction that emerged from these analyses was one between additive and
multiplicative reasoning about data (cf. Hare1 & Confrey, 1994;Thompson, 1994).
Briefly, the hallmark of additive reasoning about data is that students partition one
or more data sets in ways appropriate to the question or issue at hand and then rea-
son about the number of data points in the various parts of the data sets in
part-whole terms. This can be contrasted with multiplicative reasoning about data,
wherein students reason about the parts of a data set as proportions of the whole
data set. Our goal for the learning of the classroom community was that reasoning
about the distribution of data in multiplicative terms would become an established
mathematical practice that was beyond justification.
Thus far, in discussing distribution as a key mathematical idea and distinguish-
ing between additive and multiplicative reasoning about data, I have focused on
what is traditionally termed mathematical content. It is therefore important to
stress that our instructional focus also had a process aspect in that we attempted to
ensure that the instructional activities as realized in the classroom had the spirit of
genuine data analyses from the outset. To this end, we developed instructional ac-
tivities that involved univariate data sets and that involved either describing a sin-
gle data set for a particular purpose or comparing two or more data sets to make a
decision or judgment. The importance of attending to process as well as to content
in statistics becomes apparent once we acknowledge that anticipation is at the
heart of data analysis. Proficient analysts anticipate that certain ways of structuring
and inscribing data might reveal trends, patterns, and anomalies that bear on the
questions at hand. These anticipations, in turn, reflect a deep understanding of cen-
tral statistical ideas. For example, a student who decides that it might be productive
to inscribe data sets as box plots anticipates the possibility of structuring the data
sets multiplicatively. Similarly, a student who decides to create a scatter plot does
so to investigate the covariation of two sets of univariate measures. The challenge
as we formulated it, therefore, was to transcend what Dewey (1980) called the di-
chotomy between process and content by systematically supporting the emergence
12 COBB
of key statistical ideas while ensuring that the successive classroom mathematical
practices that emerged in the course of the teaching experiment were commensura-
ble with the activities of proficient data analysts. As Biehler and Steinbring (1991)
noted, an exploratory or investigative orientation is not merely a means of support-
ing learning but is instead central to data analysis and constitutes an instructional
goal in its own right.
The summary I have given of our instructional intent clarifies the potential end
points of the learning trajectory that we envisioned for the classroom community.
With regard to the starting points, the performance assessments that we conducted
with seventh-grade students prior to the teaching experiment indicated that data
analysis for them involved "doing something with the numbers," frequently by us-
ing methods derived from their prior instructional experiences with statistics in
school (McGatha, Cobb, & McClain, 1998). In other words, these students did not
view data as measures of particular aspects or features of situations that were
judged to be relevant when addressing a particular question or issue. An immedi-
ate goal at the beginning of the teaching experiment, therefore, was to ensure that
the first mathematical practices established in the classroom actually involved the
analysis of data. In the approach that we took, the teacher talked through the data
creation process with the students. This involved discussing the particular problem
or question under investigation, clarifying its significance, delineating relevant as-
pects of the situation that might be measured, and considering various ways of
measuring them. The data the students were to analyze were then introduced as re-
sulting from this process. We conjectured that, as a consequence of participating in
such discussions, the data would have a history for the students such that it was
grounded in the situation and reflected particular purposes and interests (cf.
Latour, 1987; Lehrer & Romberg, 1996; Roth & McGinn, 1998).
Beyond this general instructional strategy, we developed two computer-based
minitools for the students to use as integral aspects of the instructional sequence.*
Each minitool offered students several ways of structuring data. More important,
these options do not correspond to a variety of conventional inscriptions as is typi-
cally the case with commercially available data analysis tools. Instead, we drew on
the research literature to identify the various ways in which students structure data
when given the opportunity to conduct genuine analyses. Therefore, the tools were
designed to fit with taken-as-shared ways of reasoning at particular points in the
envisioned learning trajectory while serving as a means of supporting the reorgani-
zation of that reasoning, The students used these minitools in 27 of the 34 class-
room sessions. Typically, they worked at computers in pairs to conduct their
analyses, and then the teacher organized a whole-class discussion using a com-
puter projection system. I subsequently describe these two minitools when I pres-
2Koeno Gravemeijer, Michiel Doorman, Janet Bowers, and I developed the two minitools.
ent an analysis of two of the classroom mathematical practices that emerged

during the teaching experiment.
EMERGENCE OF THE FIRST MATHEMATICAL

PRACTICE
The first computer minitool was not introduced until the fifth classroom session.
The whole-class discussions in the preceding four sessions typically involved a se-
quence of separate reports in which different students described how they had com-
pleted the instructional activities. Furthermore, students did not appear to adjust
their explanations by taking into account the interpretations of the listening stu-
dents, and the listening students rarely asked clarifying questions on their own ini-
tiative. The classroom participation structure established at the beginning of the
teaching experiment, therefore, delimited the possibility of developing a
taken-as-shared basis for mathematical communication.It is also doubtful whether
most of the students were actually analyzing data, in that the numbers they manipu-
lated did not appear to signify measures of attributes of a situation about which a de-
cision was to be made. The analyses that many students reported involved methods
derived from their prior instructional experiences of doing statistics in sixth grade.
For example, an appreciable number of the students initially calculated the mean of
every data set irrespective of the question at hand. In general, the students' contri-
butions to these initial whole-class discussions appeared to reflect their prior partic-
ipation in the practices of traditional U.S. mathematics instruction.
A shift in the quality of classroom discourse occurred in subsequent discussions
when the students explained how they had used the first minitool to conduct their
analyses. This minitool was designed to provide students with a means of ordering,
partitioning, and otherwise organizing sets of up to 40 data points in a relatively
immediate way. When data are entered into the minitool, each individual data
point is inscribed as a horizontal bar, the length of which signifies the numeral
value of the data point. The students could select the color of each bar to be either
pink or green, enabling them to enter and compare two data sets. For example, Fig-
ure 2 shows data generated to compare how long two different brands of batteries
last. Each bar shows a single case, the life span of the tested batteries. The students
could sort the data by size and by color. In addition, they could hide either data set
and could also use what they called the value tool to find the value of any data point
by dragging a vertical red bar along the horizontal axis. Furthermore, they could
find the number of data points in any horizontal interval by using what they called
the range tool.
The initial data sets the students analyzed were chosen so that the measure-
ments had a sense of linearity and, thus, lent themselves to inscription as horizontal
bars (e.g., the bralung distance of cars, the length of time that batteries lasted, etc.).
Nonetheless, the students spoke almost exclusively of "pinks" and "greens" during
14 COBB
R a g e Tool
Value Tool I Count I10

://///////A Green - Pink
FIGURE 2 The first computer minitool
the first whole-class discussion in which the minitool was used and did not offer
conclusions with respect to the question at hand. It therefore seemed that they were
describing differences in two sets of numbers inscribed as colored bars rather than
analyzing data. However, the teacher was able to initiate a shift in the discourse
during this session such that the students began to speak about the bars as attributes
of individual cases that had been measured. This shift continued during the second
discussion conducted with the minitool, when the students explained how they had
analyzed the data shown in Figure 2. The green bars showed the data for a brand of
battery called "Always Ready," and the pink bars showed the data for a brand
called "Tough Cell." The first student who gave an explanation directed the
teacher to use the range tool to bound the 10 highest values (see Figure 2).
Casey: And I was saying, see like there's 7 green that last longer.
Teacher: OK, the greens are the Always Ready, so let's make sure we keep
up with which set is which. OK?
Casey: OK, the Always Ready are more consistent with the 7 right there,
and then 7 of the Tough ones are like further back, I was just saying
'cause like 7 out of ten of the greens were the longest, and like.. .
Ken: Good point.
Janice: I understand.
Teacher: You understand? OK, Janice, I'm not sure I do, so could you say it
for me?
Janice: She's saying that out of 10 of the batteries that lasted the longest, 7
of them are green, and that's the most number, so the Always
Ready batteries are better because more of those batteries lasted
longer.
Although Casey spoke of "the greens," her comment that they lasted longer sug-
gests that each bar signified how long one of the batteries lasted. Janice certainly
understood Casey's explanation in these terms, and in revoicing it, both stated an
explicit conclusion ("the Always Ready batteries are better") and justified it by
summarizing the results of Casey's analysis ("because more of those batteries
lasted longer"). In doing so, she contributed to the gradual emergence of an initial
practice of data analysis.
As the episode continued, another student, James, challenged Casey's analy-
sis by arguing that four of the pink bars (Tough Cell) were "almost in that area,
and then if you put all those in you would have seven [rather than three pinks]."
As James described features of the inscription, it is impossible to know whether
the bars carried the significance of data for him. However, the teacher inter-
preted his challenge as calling into question the way in which Casey had orga-
nized the data.
Teacher: So maybe, Casey, you can explain to us why you chose 10, that
would be really helpful.
Casey: All right, because there's 10 of the Always Ready and there's 10 of
the Tough Cell, there's 20, and half of 20 is 10.
Teacher: And why would it be helpful for us to know about the top 10, why
did you choose that, why did you choose 10 instead of 12?
Casey: Because I was trying to go with the half.
Significantly, Casey's justification for the way she organized the data, and thus her
method for comparing the two types of batteries, did not make reference to the
question at hand, that of comparing the two brands. It is also noteworthy that, with
the possible exception of James, none of the students asked her for such a justifica-
tion. This issue of justifying analyses became increasingly explicit as the discus-
sion continued.
The next student to explain his reasoning, Brad, directed the teacher to place the
value tool at 80 hr (see Figure 2).
Brad: See, there's still green ones [Always Ready] behind 80, but all of
the Tough Cell is above 80. I would rather have a consistent battery
16 COBB
that I know will get me over 80 hours than one that you just try to
guess.
Teacher: Why were you picking 80?
Brad: Because most of the Tough Cell batteries are all over 80.
Possibly as a consequence of the questions that the teacher had asked Casey, Brad
justified the method he had used without prompting. Furthermore, in doing so, he
interpreted a feature of the inscription ("There's still green ones behind 8 0 ) as in-
dicating a difference in the two brands of batteries that he considered significant,

namely, whether the batteries of a particular brand would last consistently at least
80 hr. In this respect, his explanation involved a significant advance when com-
pared with those that Casey and Janice had given.
Later in the discussion, Jennifer compared Casey's and Brad's analyses di-
rectly.
Jennifer: Even though 7 of the 10 longest-lasting batteries are Always Ready

ones, the two lowest are also Always Ready, and if you were using
those batteries for something important, then you might end up
with one of those bad batteries.
Significantly, Jennifer justified her preference for the statistic that Brad used by fo-
cusing on the pragmatic consequencesof the two analyses. The obligation ofjustify-
ing particular ways of organizing the data with respect to the practical issue at hand
gradually became taken-as-sharedduring theremainder of the session. For example,
toward the end of the discussion, one of the students observed the following:
Barry: The other thing is that I think you also need to know something
about that or whatever you're using them [the batteries] for.
Teacher: You bet.
Barry: Like, if you're using them for something real important and you're
only going to have like one or two batteries, then I think you need to
go with the most constant thing. But if you're going like, "Oh well,
I just have a lot of batteries here to use," then you need to have most
of the highest.
In making this comment, Barry explicitly clarified the situations in which the quali-
ties of the two brands assessed by the two criteria (consistency vs. most of the high-
est) would be relevant.
It is important to note that, in the latter part of the discussion, 4 students volun-
teered that they had changed their judgments as a consequence of others' argu-
ments. For example, Sally explained:
Sally: When you first look at the chart that you gave us like, oh, Tough
Cell has more, 'cause look at all the high ones and it didn't hardly
have any low ones. But when you compare them, they're a whole
lot closer than what you think.
The chart that Sally referred to was a numerical table from which the students had
entered the data into the minitool. Both Sally's comments and those of the other 3
students indicate that they experienced the discussion as an investigation in the
course of which they had developed insights into the issue at hand, that of the rela-
tive merits of the two brands of batteries. In this respect, the discussion had the spirit
of a genuine data analysis, even though the data sets were small and the methods the
students proposed were relatively elementary.
The characteristics of data sets that emerged as significant in this discussion
and in the subsequent classroom sessions in which the first minitool was used in-
cluded the range and maximum and minimum values, the number of data points
above or below acertain value or within a specified interval, and the median and its
relation to the mean. The arguments that the teacher and students developed as
they reasoned with the minitool, however, were generally additive rather than
multiplicative in nature. In the first sample episode, for example, Casey, Janice,
and the teacher jointly developed an argument that focused on how many of the 10
batteries that lasted the longest were of each brand. In doing so, they compared two
data sets that they had structured in part-whole terms. This argument can be con-
trasted with one that focuses on the proportion of each data set that is among the 10
highest values. An argument of this type would involve comparing two data sets
that have been structured multiplicatively. Crucially, such an argument is con-
cerned with the relative amount of the data in each set that is above a certain value
and, thus, with how each data set is distributed. Although additive reasoning is suf-
ficient when comparing data sets with equal numbers of data points, the students
failed to make arguments that involved reasoning about data proportionally when
they experienced difficulties while comparing unequal data sets. This indicates
that data sets were constituted in public classroom discourse as collections of data
points rather than as distributions. The mathematical practice that emerged as the
students used the first minitool might therefore be described as that of exploring
qualitative characteristics of collections of data points. It is important to stress that
these characteristics were treated as features of the situation from which the data
were generated. For example, it was taken-as-shared in the sample episodes that
the qualitative characteristic that the Tough Cell data were more "bunched up" in-
dicated greater consistency. Participation in this first practice of data analysis
therefore involved the fusion of inscriptions and the situations inscribed such that
to use the minitool to structure data was to organize the inscribed situation (cf.
Nemirovsky & Monk, in press).
18 COBB
EMERGENCE OF THE SECOND MATHEMATICAL

PRACTICE
The students first used the second of the two computer minitools during the 22nd
session of the teaching experiment. This tool was designed to allow students to ana-
lyze one or two data sets of up to 400 data points. Individual data points were in-
scribed as dots located on a horizontal axis of values (see Figure 3). The tool pro-
vides students with a variety of options for structuring data sets. The first, called
"Create Your Own Groups," involved dragging vertical bars along the axis to parti-
tion the data set into groups of points. The number of points in each group was
shown on the screen and adjusted automatically as a bar was dragged along the axis.
The remaining four options were:
Partitioning the data into groups of a specified size (e.g., 10 data points in
each group).
Partitioning the data into groups with a specified interval width.
Partitioning the data into two equal groups.
Partitioning the data into four equal groups.
The students also could hide the data, leaving only the axes and the vertical par-
tition bars visible.
45 48 50 53 55 58 60 63 65 68 70
After Speed Trap
l...e
0.0
...e
- 0
o....
* o n 0.0.
0.0.e.
...........
l
0.. ..
0 .
45 48 50 53 55 58 60 63 65 68 70
Before Speed Trap
FIGURE 3 The second computer rninitool

From the point of view of instructional design, the students' reasoning with this
tool can be viewed as a progression from their activity with the first tool. For exam-
ple, thedots at theend of the bars in the first tool, in effect, have been collapsed down
onto the axis. The teacher, in fact, introduced the new line plot inscription by first
showing a data set inscribed as horizontal bars, then removed the bars to leave only
the dots, and finally transposed the dots onto the axis. In addition, theact ofpartition-
ing a set of data points into groups also had a history in students' prior use of the first
minitool. Our general instructional intent when designing the second minitool was
to build on the students' participation in the first mathematical practice by support-

ing the emergence of increasingly sophisticated ways of structuring data, particu-
larly those that involvemultiplicative reasoning. It is for this reason that the palateof
options offered by the minitool does not correspond to a range of conventional in-
scriptions. Wedid, however, take intoaccounttheneed for studentseventually touse
conventional inscriptions in powerful ways. In this regard, two of the five op-
tions-fixed interval width and four equal groups-are precursors to two important
types of conventional inscriptions, histograms and box plots, respectively.
One of our initial concerns when the students began to use the second minitool
was to ensure that the inscriptions did signify data sets that had been generated by
measuring attributes of a situation rather than simply numbers on a line. As it tran-
spired, the students' activity with this minitool did appear to involve reasoning
with data from the outset. The practice of data analysis that emerged as the stu-
dents used this tool can be illustrated by focusing on episodes from three
whole-class discussions. In each of these discussions, two members of the project
staff shared the teaching responsibilities. The first of these discussions occurred in
the 26th classroom session and focused on the question of whether the introduction
of a police speed trap in a zone with a 50 miles per hr speed limit had slowed down
the traffic speed and thus reduced accidents. The data the students analyzed are
shown in Figure 3. The bottom graph shows the speeds of 60 cars before the speed
trap was introduced, and the top graph shows the speeds of 60 cars after the speed
trap had been in use for some time.
To begin the discussion, one of the teachers asked Janice to read the report she
had written of her analysis.
Janice: If you look at the graphs and look at them like hills, then for the be-
fore group, the speeds are spread out and more than 55, and if you
look at the after graph, then more people are bunched up close to
the speed limit, which means that the majority of the people slowed
down close to the speed limit.
This was the first occasion in public classroom discourse in which a student de-
scribed a data set in global, qualitative terms by referring to its shape. One of the
20 COBB
teachers legitimized Janice's interpretation and indicated that it was particularly

valued by drawing the "hills" on the projected data. Both teachers then capitalized
on Janice's contribution in the remainder of the discussion, treating other students'
analyses as attempts to describe qualitative differences in the data sets in quantita-
tive terms. For example, Karen explained that she had organized the data sets by us-
ing a fixed interval width of 5.
Karen: Like, on the first one [before the speed trap was introduced], most
people are from 50 to 60, that's where most people were on the
graph.
One of the teachers checked whether other students agreed with her interpretation.
Karen then continued:
Karen: And then on the top one [after the speed trap was introduced], most
people were between 50 and 55, because, um, lots of people slowed
down .. . so like more people were between 50 and 55.
The same teacher then recast Karen's analysis as a way of characterizing the global
shift of which Janice had spoken. As a consequence of this revoicing, it gradually
became taken-as-shared that the intent of an analysis was to identify global trends
or patterns in data that were significant with respect to the issue under investigation.
The history of this development can be traced to the first mathematical practice, in
which, it will be recalled, the ways that collections of data points were organized
had to be justified with respect to the question at hand.
A second illustrative episode that serves to clarify the nature of the emerging
mathematical practice occurred in the next classroom session. The students'
charge was to evaluate a special diet program that was designed to reduce the cho-
lesterol levels of people who are susceptible to heart problems. In developing the
situation, one of the teachers and the students talked through the data creation pro-
cess that involved measuring the cholesterol levels of 60 people before and after
they had followed the dizt for 1 month. The data the students analyzed is shown in
Figure 4. The bottom graph shows the cholesterol levels before the treatment, and
the top graph shows the cholesterol levels after the treatment.
The first pair of students to describe their analysis, Sally and Madeline, ex-
plained that they had partitioned each data set into four groups of equal size.
Throughout the discussion, the data were hidden so that only the axes and the parti-
tion lines were visible (see Figure 5). Their argument was that, even though the
ranges of the two data sets were the same, the range of the middle two groups (or
quartiles) was lower after the treatment, indicating that the diet was successful in
lowering cholesterol. This argument indicates that, for Sally and Madeline, the two
graphs revealed a global difference in the way that the two sets of data were distrib-
After Program
....
.
0 . e.
l 0 .
W.. - 0 .
0 . . .OW...-.*.
l..I.- .........).W..*
100 130 160 190 220 250 280 310 340 370 400
Before Program
FIGURE 4 The cholesterol data.
After
Before
FIGURE 5 The cholesterol data organized into four equal groups.

uted. However, because the data were not visible, students had to infer how the
data might be distributed from the graphs to understand their argument. Most had
difficulty in doing so, and their analysis became the focus of a protracted ex-
change.
During their initial attempts to explain their reasoning, Madeline used the term
"middle section" to refer to the middle two groups. One of the teachers asked the
students if they knew what she meant by this term and subsequently established
with them that, because one fourth of the data (or 15 data points) were in each
group, half the data-or 30 data points-were in the middle section of each graph.
Against this background, Valarie asked Sally and Madeline for clarification.
Valarie: What exactly were you talking about, the middle thing I didn't un-
derstand, I got everything else.
Sally: The middle half.
Madeline: There's four sections and we're talking about the two middle sec-
tions.
Sally: The middle half . .. the second and third fourths.
Teacher 1 : Valarie, do you understand?
Valarie: No, now I understand the middle thing, but I don't understand how
they used it in the problem.
In asking for further elaboration, Valarie indicated that, although she understood
how Sally and Madeline had structured the data, she did not understand why they
had done so. She therefore requested that they explain why the way that they had or-
ganized the data was relevant to the question at hand-that of assessing the effec-
tiveness of the diet program. Significantly, although Sally and Madeline previously
had explained that the middle section was lower on the after-treatment graph, they
had not spoken explicitly about global differences in the way that the two sets of
data were distributed when they described the results of their analysis.
As the discussion continued, James attempted to explain Sally and
Madeline's reasoning, but he too spoke about groups or sections without inter-
preting them in terms of global, qualitative differences in the data sets. Later,
Casey echoed Valarie in asking Sally and Melissa why they had focused on the
middle parts of the graphs. The questions that both girls asked appeared to re-
flect the assumption that explanations should be interpretable in global, qualita-
tive terms. In requesting clarification, they therefore contributed to the continual
regeneration of data analysis as a practice that involved investigating trends or
patterns in data that were considered relevant with regard to the issue at hand. In
response to these questions, Sally and one of the teachers finally developed an
explanation that made explicit reference to broad patterns in the way the data
were distributed. They first established that the ranges of the two data sets were
the same, and then Sally continued:
Sally: The range is the same, but like the median is what is different, like
the median right here [points to the after-treatment graph], it means
these [data points] move lower in the bottom half closer to the bot-
tom than in here [points to the before-treatment graph].
In this explanation, Sally attempted to clarify what the difference in the medians
meant in terms of how the data were distributed. The teacher then capitalized on her
contribution:
Teacher 2: This is another way to think about it ... so we agree that before and
after they are in about the same place [places his hands on the high-
est and lowest values of each distribution], they're in there some-
where, the range is the same. So what you're trying to focus on is
where between the lowest and the highest they are before and after.
Are they all up at one end, or have they all moved down to the other
end?
In describing the intent of Sally and Madeline's analysis as that of investigating

how the data were distributed, the teacher was attempting to orient other students'
efforts to make sense of the graph. Although it is doubtful that all the students came
to interpret these particular graphs in this way by the end of the episode, a third il-
lustrative episode indicates that this interpretive stance did become taken-as-shared
during subsequent classroom sessions.
The third illustrative episode occurred a week later during the 30th classroom
session. The students had compared two treatment protocols for AIDS patients by
analyzing the T-cell counts of people who had received one of the two protocols.
Their task was to assess whether a new experimental protocol in which 46 people
had enrolled was more successful in raising T-cell counts than a standard protocol
in which 186people had enrolled. The data the students analyzed are shown in Fig-
ure 6. The computer minitool was not used during the subsequent discussion. In-
stead, the discussion focused on the reports that the students had written of their
analyses.
The inscription from the first report that was discussed showed global differ-
ences in the way the two sets of data were distributed (see Figure 7). The students
judged this report to be adequate and made a number of comments.
Janice: I think it's an adequate way of showing the information because

you can see where the ranges were and where the majority of the
numbers were.
David: What do you mean by majority of the numbers?
Teacher 1: David doesn't know what you mean by the majority of the num-
bers.
Experimental
200 270 340 410 480 550 620 690 760 830 900
Standard
FIGURE 6 The AIDS protocol data.
Numbers
(highest)
1
I 1
I
200 72 1 850
lowest highest
Numbers
I
(highest) I
I 1
I I I
230 257 552 792
lowest highest
FIGURE 7 First analysis of the AIDS protocol data

Janice: Where most of the numbers were.
Teacher 1: Sharon, can you help?
Sharon: What she's talking about, I think what she's saying, like when you
say where the majority of the numbers were, where the point is, like
you see where it goes up.
Teacher 1: I do see where it goes up [indicates the "hill" on the lower diagram].
Sharon: Yeah, right in there, that's where the majority of it is.
Teacher 1 : OK, David?
David: The highest range of the numbers?

Sharon: Yes.
Teacher 1: The highest range?
Several
students: No.
Teacher 1: Valarie.
Valarie: Out of however many people were tested, that's where most of
those people fitted in, in between that range.
Teacher 1: [Pointing to lower and upper bounds of one of the "hills"] You
mean this range here?
Valarie: Yes.
It is evident from this exchange that, when the students spoke about "the major-
ity" or "most of the people," they were talking about data organized
multiplicatively as qualitative proportions (P. Thompson, personal communica-
tion, September 1997). Janice first had introduced the term "the majority" during
the discussion of the speed trap data when she had described hills in the data. A
concern with global patterns in the way that data are distributed, in fact, assumes
that the data are structured multiplicatively. In describing hills, Janice was reason-
ing about qualitative relative frequencies. However, this notion of the majority of
the data did not become an explicit topic of conversation until the students ana-
lyzed data sets with unequal numbers of data points.
During the remainder of the discussion, the teachers attempted to guide the
gradual refinement of the taken-as-shared notion of qualitative proportionality.
For example, the third report discussed read as shown in Figure 8. One of the
teachers clarified with the students during the subsequent exchange that the writ-
ers of the report had chosen the statistic of the number of patients with T-cell
counts above 525 because the majority of the data points in the old treatment
were below this value, and the majority in the new treatment were above it.
Thus, they had developed a method for describing a global difference in two dis-
tributions. One of the students then suggested drawing graphs to show the re-
sults of the analysis (see Figure 9).
One of the teachers then made the following argument that reflected an additive
interpretation of the graphs.
26 COBB
OLD PROGRAM N E W PROGRAM

200 - 525 200 - 525
130 9
525 - 850 525 - 850
56 37
FIGURE 8 Third analysis of the AIDS protocol data

FIGURE 9 Graphs developed from the third analysis of the AIDS protocol data
Teacher 2: Could you just argue that this shows really convincingly that the
old treatment was better, right, because there were 56 scores above
525,56 people with T cell counts above 525, and here [points to the
graph on the right] there's only 37 above, so the old one just had to
be better, there's more people, I mean there's 19 more people in
there, so that's the better one, surely.
The initial arguments the students made when rejecting this claim involved reason-
ing in terms of qualitative proportions. However, Ken made the following pro-
posal:
Ken: I've got a suggestion. I don't know how to do it [inaudible]. Is there

a way to make 130 and 56 compare to the 9 and 37, I don't know
how.
Teacher 2: I'll tell you. How many of you have studied percentages?
In the ensuing exchange, several students calculated the percentages of data

points above the T-cell count of 525 in each distribution. As the discussion con-
tinued, it seemed to be taken-as-shared that the results of these calculations pro-
vided a way of describing global differences in the two distributions in
quantitative terms.
In both the remainder of this session and in the final three sessions of the teach-
ing experiment, discussions continued to focus on reasoning about data
multiplicatively. Interviews conducted with the students shortly after the teaching
experiment was completed indicate that most could readily interpret graphs of un-
equal data sets organized into equal interval widths, an analogue of histograms,
and into four equal groups, an analogue of box plots, in terms of global characteris-
tics of distributions. The classroom mathematical practice that had emerged as
they developed these competencies can be described as that of exploring qualita-
tive characteristics of distributions. Participation in this practice involved reason-

ing about data multiplicatively while using the computer minitool to identify
global patterns and to describe them in quantitative terms. The transition from the
first to the second mathematical practices involved a shift in the nature of discus-
sions, such that the focus was on ways of organizing data that were relevant to the
purpose at hand, rather on the practical decision or judgment per se. For example,
during the discussion of the battery data, the students developed data-based argu-
ments for why the batteries of one of the brands were superior. In contrast, the stu-
dents agreed that the new treatment for AIDS patients was better than the standard
treatment at the beginning of the discussion. The focus was instead on different
ways of describing global differences in the two data sets. It might therefore be
said that participation in the second mathematical practice involved analyzing data
from a mathematical point of view. Taking this characterization of the second
practice one step further, Konold et al. (1996) argued that a focus on the rate of oc-
currence of some set of data values within a range of values is at the heart of what
they termed a statistical perspective. As participation in the second practice in-
volved a concern for the proportion of data within various ranges of values, the stu-
dents appeared to be well on the way toward developing this statistical perspective.
CHANGE
My overall purpose in presenting the sample episodes has been to illustrate a theo-
retical approach that involves analyzing the mathematical learning of the class-
room community. The discussion of the two classroom mathematical practices
documents how the taken-as-shared ways of reasoning and arguing about data
changed in the course of the teaching experiments. It is important to stress that the
account I have given does not focus on the mathematical development of any par-
ticular student. Instead, I have been concerned with changes in public mathematical
activity and discourse. As Voigt (1995) observed, the taken-as-shared meanings
and understandings inherent in classroom mathematical practices constitute a se-
mantic domain in their own right and should not be equated with an overlap in indi-
vidual meanings. The latter focus is essentially individualistic in that it is concerned
with a relation between the mathematical interpretations of individual members of
28 COBB
the classroom community. In contrast, the approach I have illustrated takes the
cIassroom community itself rather than the individuals that compose it as the unit of
analysis and delineates changes in collective meanings and practices.
Turning now to clarify the general notion of a classroom mathematical practice,
it should be apparent from the analysis of the sample episodes that the use of tools
and symbols is integral to both the mathematical practices and the reasoning of the
students who participate in them (cf. Dorfler, 1993; Kaput, 1991; Pea, 1993). For
example, when the students explained their analyses of the battery data near the
beginning of the teaching experiment, an act of moving the value tool or the range
tool to a particular location was an act of structuring collections of data points.
Similar comments can be made about the students' use of the second minitool. For
example, when Janice spoke of hills in the speed trap data, she was describing the
shape of data that had been inscribed as a line plot. The notion of data sets as distri-
butions, rather than collections of data points, emerged and became
taken-as-shared as she and the other students reasoned with line plots. In all likeli-
hood, this central mathematical idea would not have arisen had data been inscribed
differently. The analysis of the sample episodes is therefore consistent with the ba-
sic Vygotskian insight that the tools students use profoundly influence both the
process of mathematical development and its products, increasingly sophisticated
mathematical ways of reasoning (Meira, 1995; Saxe, 1991; van Oers, 1996;
Wertsch, 1994).
A second aspect of classroom mathematical practices that complements the em-
phasis on tool use is that of argumentation. Norms or standards of mathematical ar-
gumentation were established relatively early in the teaching experiment. I can
best substantiate this claim by following Krummheuer (1995) and Yackel (1997)
in using Toulmin's (1969) scheme of conclusion, data, warrant, and backing. In
this scheme, Toulmin referred to the support one might give for a conclusion as
data. In the case of the analysis of the battery data, for example, a student might
merely point to the two data sets and state the conclusion that one of the brands of
batteries is superior. In doing so, the student treats the conclusion as a self-evident
consequence of the data. If questioned, the student would be obliged to give a war-
rant that explains why the data support the conclusion. For example, Casey justi-
fied her conclusion that the Always Ready batteries were more consistent by
explaining that she had focused on the 10 batteries that lasted the longest and noted
that 7 of them were Always Ready batteries. In giving this warrant, Casey ex-
plained how she had structured and interpreted the data sets. In Toulmin's scheme,
the warrant can be questioned, and it is then necessary to give a backing that indi-
cates why the warrant should be accepted as having authority. Casey was, in fact,
challenged by the teacher, who asked her why she had chosen to focus on the 10
batteries that lasted the longest. The backing that Casey gave, namely that 10 was
half of the data set of 20 points, was delegitimized as the episode progressed. In-
stead, it gradually became taken-as-shared in the remainder of this episode and in
subsequent sessions that a particular way of structuring data had to be justified by

explaining why it was relevant to the question or issue at hand. The backing that
Casey gave proved to be unacceptable because she did not explain why focusing
on the 10 longest-lasting batteries was an appropriate way of comparing the two
brands. In contrast, Brad's explanation that he wanted a battery that he knew
would last at least 80 hr was accepted as giving authority to his approach of parti-
tioning the data sets into values above and below 80 hr.
The scheme of argumentation I have outlined is summarized in Figure 10. Be-
cause this scheme also captures the structure of argumentation established when
the students used the second minitool and participated in the second mathematical
practice, it constitutes a sociomathematical norm that cuts across specific prac-
tices. The distinction between argumentation as an aspect of the two practices con-
cerns the nature of the data about which arguments were developed. Participation
in the first mathematical practice involved developing arguments about collec-
tions of data points, whereas, in the second practice, the arguments were about dis-
tributions. The scheme of argumentation shown in Figure 10 is, in fact, quite
general and applies to data analysis more broadly. This becomes apparent when we
note that, in structuring and interpreting data, the students created methods that
17-n Conclusion
Explain how the data

have been structured
and interpreted
Explain why this way of

structuring the data is
appropriate with respect
to the question at hand
FIGURE 10 Scheme for argumentation that emerged during the teaching experiment.
30 COBB
Conclusion
Warrant:
Explain the statistics
Backing:
Explain why the statistics
are appropriate with
respect to the question at
FIGURE 11 General scheme of argumentation for data analysis
served the function of statistics. The compatibility between the general scheme of
argumentation shown in Figure 1 1 and the norms for argumentation established
during the teaching experiment indicates that the students were being inducted into
what might be termed an authentic data analysis point of view.
As Yackel (1997) demonstrated, Toulmin's (1969) scheme has important
methodological implications for the analysis of classroom mathematical practices.
To illustrate this point, recall that, when Janice first introduced the notion of hills
in the data, she had to explain her interpretation. In doing so, she gave a warrant for
her interpretation of the data sets as distributions rather than collections of data
points. In contrast, the legitimacy of a hills interpretation was not questioned when
the students discussed analyses of the AIDS protocol data. It is precisely this lack
of a need for a warrant that serves to indicate that the interpretation of data sets as
distributions was taken-as-shared (Yackel, 1997). In general, an analysis of the
evolution of mathematical practices focuses as much on what no longer needs to be
said and done as it does on what the teacher and students actually say and do.
At the beginning of this article, I justified my focus on the learning of the class-
room community by referring to the concerns and interests of developmental re-
search. In doing so, I argued that the conjectures that instructional designers
develop when formulating hypothetical learning trajectories are about the learning
of the group rather than of any particular student. Viewed in these terms, an analy-
sis of the evolution of classroom mathematical practices documents the actual
learning trajectory of a classroom community. This, in turn, implies that the theo-
retical notion of a classroom mathematical practice encompasses the two major as-
pects of developmental research-instructional design and classroom-based
analyses. Analyses of the type I have illustrated are therefore cast in such a way
that they can readily feed back to inform the ongoing instructional design effort. In
the case of the statistics teaching experiment, for example, we are currently collab-
orating with a group of teachers to revise the instructional sequence.

It should be clear from the sample analysis that an approach of this type takes
what are traditionally called issues of mathematical content seriously. For exam-
ple, the contrast between the two mathematical practices is characterized, at least
in part, by the distinction between additive and multiplicative reasoning about
data. However, this approach also calls into question the metaphor of mathematics
as content. The content metaphor entails the notion that mathematics is placed i n
the container of the curriculum, which then serves as the primary vehicle for mak-
ing it accessible to students. In contrast, the approach I have illustrated character-
izes what is traditionally called mathematical content in emergent terms. For
example, the mathematical idea of distribution was seen to emerge as the collec-
tive practices of the classroom community evolved. This theoretical orientation
clearly involves a significant paradigm shift in how we think about both mathe-
matics and the means by which we might support students' induction into its prac-
tices. However, this approach does have the merit of being compatible with the
view of mathematics as a socially and culturally situated activity (cf. Bauersfeld,
1992; John-Steiner, 1995; Lave, 1993; Sfard, in press).
This shift from the content metaphor to the emergence metaphor immediately
brings issues of teachers' professional development to the fore. In this regard, it is
important to observe that an analysis of the type that I have illustrated delineates a
learning trajectory that culminates with overarching mathematical ideas that are
the goal of an instructional sequence. The analysis therefore provides a justifica-
tion for the instructional sequence that is cast in terms of (a) the collective develop-
ment of particular mathematical ideas, and (b) the means of supporting that
development. Such a justification, it should be noted, is not tied to specific instruc-
tional activities. Instead, the instructional activities used in, for example, a teach-
ing experiment, illustrate one concrete enactment of the sequence. My colleagues
and I conjecture that sequences justified in this manner might constitute an impor-
tant means of supporting the development of professional teaching communities.
When a sequence is justified solely in terms of traditional experimental data,
teachers know that the sequence proved effective elsewhere, but they do not have
the opportunity to develop an understanding that would enable them to adapt the
sequence to their own situations. In contrast, the type of justification derived from
an analysis of classroom mathematical practices offers the possibility that teachers
will be able to adapt, test, and modify the sequence in their own classrooms.
This conjecture about the potential role of instructional sequences is consis-
tent with the view of implementation as idea-driven adaptation. In addition, the
conjecture finds support in Ball and Cohen's (1996) argument that re-
search-based instructional sequences can constitute important resources for
teachers' as well as students' learning (see also Gearhart et al., 1994). At the
time of writing, my colleagues and I were just beginning to investigate the via-
bility of this conjecture in collaboration with a group of teachers. Our overall
goal is to support the development of a professional teaching community that

learns from its collective experience by analyzing, adapting, testing, and refining
pedagogical ideas and processes that have led to the improvement of students'
mathematical learning in other settings. These communal norms and practices
can be thought of as the envisioned end points of a pedagogical learning trajec-
tory, in which we and the teachers are joint participants. Part of the challenge we
currently are attempting to address is that of developing the means of enabling
the teachers to reconstruct justifications for particular instructional sequences
by, to some extent, living through the design process.
In concluding this discussion of collective mathematical learning, it is worth
noting that analyses that focus on communal mathematical practices in no way
deny individual initiative and creativity. In the introductory section of this article, I
discussed intellectual autonomy and argued that it can be viewed as a particular
way of participating effectively in communal practice in which individuals rely on
their own judgments. The main point of my argument was that, although we do not
have to give up the notion of autonomy when we view individual reasoning as an
act of participation in communal practices, we do need to reconceptualize it. A
similar argument can be made about creativity. In the course of the seventh-grade
teaching experiment, the students certainly made creative contributions. For ex-
ample, we did not anticipate, when we designed the first computer minitool, that
students would use the range tool to partition collections of data points. From our
point of view, the first data analyses in which they used the range tool in this way
were creative. Similarly, Janice's introduction of the hill metaphor to talk about
global patterns in data was novel, as was Sally and Madeline's argument in which
they focused on the middle section of data that had been structured into four equal
groups. In the viewpoint I have outlined, these contributions are not seen as in-
stances of the unbridled, purely individualistic creativity of the type that is some-
times glorified in mathematics education. Instead, each contribution is viewed as
an act of participating in and contributing to the evolution of communal mathemat-
ical practices. More generally, creative acts cany with them the history of partici-
pation in previously established practices (cf. Hicks, 1996; Shotter, 1995). Viewed
in this way, creativity is social through and through. Rather than being characteris-
tic of a purely individual act, it is characteristic of a relation between individual ac-
tivity and the communal practices in which the individual participates. This
conceptualization is nondeterministic in that it does not in any way trivialize or
denigrate students' creativity. It does, however, challenge romantic views of cre-
ativity by locating it in a social context, one that teachers and students jointly con-
stitute as they establish the practices in which they participate.
DIVERSITY
Thus far, in focusing on collective practices, I have emphasized the taken-as-shared

ways of reasoning, arguing, and using tools that are established by a classroom
community. It therefore is important to acknowledge that students participate in
any particular mathematical practice in a variety of qualitatively different ways.
Recall, for example, that a number of students had difficulty in understanding Sally
and Madeline's reasoning when they explained that they had structured the choles-
terol data into four equal groups. It appeared that, at that point in the teaching exper-
iment, these students needed to know the actual location of individual data points to
infer global patterns from graphs and, thus, reason multiplicatively about data sets
as distributions. In contrast, Sally, Madeline, and a number of other students in-
ferred global patterns in the distribution of the data directly from the graphs of four
equal groups. In doing so, they seemed to structure data into quantitative rather than
qualitative proportions. Consequently, whereas the first group of students inter-
preted the graphs by reasoning from the data points to distributions, the second
group of students reasoned directly about the distributions. There were, therefore,
significant differences in the ways in which the two groups of students participated
in the second mathematical practice.
It might be thought that this diversity in students' thinking is specific to statistics,
a domain that might appear to lend itself to alternative interpretations. Therefore, I
should note that students also participated in classroom mathematical practices in a
variety of different ways in previous teaching experiments that focused on elemen-
tary addition and subtraction (Gravemeijer, Cobb, Bowers, & Whitenack, in press),
place-value numeration (Bowers et al., in press), and linear measurement (McClain,
Cobb, Gravemeijer, & Estes, in press; Stephan, 1998). The need to clarify the rela-
tion between individual students' reasoning and the collective practices in which
they participate is therefore a pressing one. As a first step, imagine as a thought ex-
periment that we had interviewed not only the students in the teaching experiment
classroom but also thosefrom another seventh-grade classroom in the same school. I
have no doubt that, if we shuffled the video recordings of these interviews, the reader
could almost unerringly identify from which classroom each student had come. It is
precisely this contrast in the mathematical reasoning of two groups of students that is
accounted for by their participation in the differing mathematical practices estab-
lished in the two classrooms.
34 COBB
To continue the thought experiment, imagine now that we focus only on the stu-
dents in the teaching experiment classroom. The contrast is then not between one
group of students as compared to another; instead, what is being contrasted is the
reasoning among students who have participated in the same classroom mathemat-
ical practices. In focusing on differences in individual students' reasoning in this
manner, we have adopted a psychological orientation of the type that is so promi-
nent in mathematics education. This perspective is of value and complements a so-
cial perspective by bringing the diversity of students' reasoning to the fore.
However, it is, by itself, inadequate for the purposes of developmental research in
that it also blinds us to the taken-as-shared basis for mathematical communication

established by the classroom community. Were we to adopt only this perspective,
we would, like the proverbial fish, be oblivious to the water of communal prac-
tices. The challenge, therefore, is not that of choosing between social and psycho-
logical perspectives on mathematical activity but, instead, to develop ways of
coordinating the two perspectives.
In the viewpoint that has emerged from my own and my colleagues' work in
classrooms, the relation between the two perspectives is taken to be reflexive. This
is an extremely strong relation and does not merely mean that individual students'
reasoning and the practices in which they participate are interdependent. Instead, it
implies that one literally does not exist without the other (Mehan &Wood, 1975).
Thus, when adopting a psychological perspective, one analyzes individual stu-
dents' reasoning as they participate in the practices of the classroom community.
Conversely, when adopting a social perspective, one focuses on communal prac-
tices that are continually generated by and do not exist apart from the activities of
the participating individuals. The coordination at issue is therefore not that be-
tween individual students and the classroom community viewed as separate,
sharply defined entities. Instead, the coordination is between two alternative ways
of looking at and making sense of what is going on in classrooms. In other words,
we are coordinating different ways in which we can interpret classroom events.
What, from one perspective, are seen as the norms and practices of a single class-
room community is, from the other perspective, seen as the reasoning of a collec-
tion of individuals who mutually adapt to each others' actions. Whitson (1997)
emphasized this point when he proposed that we think of ourselves as viewing hu-
man processes in the classroom, with the realization that these processes can be de-
scribed in either social or psychological terms. In my view, both perspectives are
relevant to the concerns and interests of classroom-based developmental research.
I already have hinted at the fact that this theoretical orientation has grown out of
and remains deeply rooted in our attempts to support students' mathematical de-
velopment while working in classrooms. I can best illustrate the way in which the-
ory is grounded in the reality of the classroom by returning to the statistics teaching
experiment. As we have seen, the students frequently worked in pairs at the com-
puters and then explained their analyses in a subsequent whole-class discussion.
While the students were working at the computers, the teacher and a second mem-
ber of the project staff circulated around the classroom to gain a sense of the di-
verse ways in which the students were organizing the data. Toward the end of the
small-group work, the teacher and project staff member then conferred briefly to
plan the whole-class discussion. In doing so, they routinely focused on the qualita-
tive differences in students' analyses in order to develop conjectures about mathe-
matically significant issues that might emerge as topics of conversation. The intent
was to capitalize on the students' reasoning by identifying data analyses that, when
compared and contrasted, might give rise to substantive mathematical discussions.
The episodes I have presented from the seventh-grade teaching experiment exem-
plify such discussions. In the first episode, for example, the issue of justifying the
statistic used with respect to the question at hand emerged from the contrast be-
tween Casey's and Brad's analyses. In the next episode that I presented, the two
teachers decided to ask Janice to explain her hills metaphor at the beginning of a
discussion because they conjectured that it might bring to the fore the issue of in-
terpreting data in terms of qualitative proportions. In the final episode, the analyses
of the AIDS protocol data that were discussed were sequenced so that an initial
analysis that focused on hills in the data might provide the students with a point of
reference when interpreting a subsequent analysis in which the two data sets were
partitioned at a particular value.
In this opportunistic approach to instructional planning, students' diverse ways
of participating in communal practices are a key resource upon which the teacher
attempts to capitalize. The mathematically significant issues that become topics of
conversation emerge from this diversity with the teacher's guidance. In reorganiz-
ing their thinking while participating in these discussions, students contribute to
the evolution of the classroom mathematical practices. In the hands of a skillful
teacher, the diversity in students' reasoning is, in many respects, the primary mo-
tor of the collective mathematical learning of the classroom community. In the last
analysis, it is this realization that convinces my colleagues and me of the need to
coordinate a social perspective on communal practices with a psychological per-
spective that takes into account the students' diverse ways of participating in them.
As a final point, it is important to note that the viewpoint I have outlined has two
major ethical implications. The first is that all students must have a way to partici-
pate in the mathematical practices of the classroom community. In a very real
sense, students who cannot participate in these practices are no longer members of
the classroom community from a mathematical point of view. This situation is
highly detrimental given that to learn is to participate in and contribute to the evo-
lution of communal practices. Students who are excluded are deprived not merely
of learning opportunities but of the very possibility of growing mathematically.
One of our primary concerns when conducting a teaching experiment is therefore
to ensure that all students are "in the game." To this end, we adjust the classroom
participation structure, classroom discourse, and instructional activities on the ba-
36 COBB
sis of ongoing observations of individual students' activity. In doing so, we once

again find ourselves coordinating psychological and social perspectives and con-
tend that an approach of this type is necessary, if not sufficient, when addressing
concerns of equity at the microlevel of classroom action and interaction.
The second ethical implication is closely related to the first and concerns the
view one takes of students whose ways of participating in particular classroom
practices are less sophisticated than those of other students. For example, by the
end of the statistics teaching experiment, the majority of the students routinely
developed arguments by inferring global patterns directly from complex graphs
of data they had not themselves analyzed. However, a number of students devel-
oped arguments that indicated that they had interpreted the graphs in
less-sophisticated ways. In the theoretical orientation I have presented, these dif-
fering interpretations are not viewed as cognitive characteristics of the individ-
ual students but as characteristics of their ways of participating in communal
mathematical practices. In other words, the differences in the students' reason-
ing are seen to be socially situated and to reflect the history of their prior partici-
pation in particular practices. As a consequence, my colleagues and I do not take
a cognitive deficit view of the students who made less sophisticated interpreta-
tions. Instead, our reflections on the teaching experiment have focused on the
evolving mathematical practices that constituted the immediate social situation
of their mathematical development as well as on the nature of their participation
in those practices. In doing so, we have treated academic success and failure in
the classroom as neither a property of individual students nor a property of the
instruction they receive. Instead, we have cast it as a relation between individual
students and the practices that they and the teacher coconstruct in the course of
their ongoing interactions. In the last analysis, the ethical dimension of this per-
spective on success and failure in school is perhaps the most important reason
for adopting a viewpoint that brings the diversity of students' reasoning to the
fore while seeing that diversity as socially situated.
EQUITY
Throughout this article, I have focused on the issues of change and diversity as they
relate to the concerns of instructional design at the classroom level. It therefore is
important to acknowledge that the teaching experiment we conducted did not take
place in a social vacuum. Instead, the classroom in which we worked was itself lo-
cated within the sociopolitical setting of one particular school and community and
ultimately within the activity system that constitutes schooling in the United States.
At this broader level, the work of several scholars has made us aware that schooling
involves a number of taken-for-granted policies and practices that foster inequity
due to race, gender, class, and economic status (Apple, 1995;Zevenbergen, 1996).
Furthermore, as Lave (1996) observed, schooling as a social institution involves an

inherent contradiction between the functions of universal socialization on the one
hand and those of reproducing the unequal distribution of particular ways of know-
ing as cultural capital on the other hand. It was, in fact, with these global, structural
analyses of schooling in mind that I said that the theoretical orientation I have pre-
sented is necessary but not sufficient when addressing issues of equity. My col-
leagues and I have argued elsewhere that this viewpoint on classroom activities and
events must itself be complemented by a strong sociocultural perspective that
places the classroom in a broader sociopolitical context (P. Cobb & Yackel, 1996).
I therefore anticipate further insights from sociocultural analyses that, although not
necessarily specific to mathematics education, cast into sharp relief social policies
and practices that foster inequity.
At the more local level of the school and community in which we conducted the
teaching experiment, the norms and practices established in the project classroom
were potentially in conflict with those that the students experienced throughout the
remainder of the school day. Furthermore, the students came from a number of dif-
ferent communities within the city in which the school was located and, therefore,
had participated in a diverse range of out-of-school practices, some of which may
have been inconsistent with the microculture established in the project classroom
(cf. Ladson-Billings, 1995; Moll, 1997; Secada, 1992; Warren & Rosebery, 1995).
In an initial attempt to address these concerns, two members of the project staff
followed several of the students throughout the school day to develop an under-
standing both of general school norms and of the groupings that had been consti-
tuted within the student body. In addition, a member of the project staff has
analyzed classroom interactions to identify possible inconsistencies between the
classroom microculture and students' home cultures as well as to examine how the
students perceived themselves and other members of groups within the classroom.
The purpose of these explorations is to delineate issues whose investigations will
contribute to our understanding of equity as it relates specifically to teaching and
learning mathematics with understanding.
In addition to these concerns that take us beyond the classroom, issues of equity
come to the fore when we restrict our focus to instructional design. In the case of
the seventh-grade teaching experiment, a question that we had to address was why
statistics should be taught in school. Two general types of justifications can be
found in the literature. The first refers to developments in the discipline, many of
which have been fueled by the use of computers as exploratory tools. The meta-
phor that emerges from these justifications is that of students as apprentice re-
search statisticians. A second type of justification refers to the increasingly
prominent role of statistical reasoning in both work-related activities and informed
citizenship. The emphasis in this rationale is on social utility, and the image that
emerges for students' roles is as that of consumers of analysis techniques devel-
oped by others.
38 COBB
In contrast to these two common rationales, we find a third justification to be far

more compelling. Briefly, the increasing use of computers, not just within the dis-
cipline but in society in general, has placed an increasing premium on quantitative
reasoning in general and on statistical reasoning in particular. There is much talk
of preparing students for the "information age" but without fully appreciating that
the information in this new era is largely quantitative in nature. This shift has dra-
matic implications for the discourse of public policy and, thus, for democratic par-
ticipation and power (G. W. Cobb, 1997). It is already apparent that debates about
public policy issues tend to involve reasoning with data. In this discourse, policy
decisions are justified by presenting arguments based on the analysis of data. In

many respects, this discourse is increasingly becoming the language of power in
the public policy arena. Inability to participate in this discourse results in de facto
disenfranchisement that spawns alienation from, and cynicism about, the political
process. Cast in these terms, statistical literacy that involves reasoning with data in
relatively sophisticated ways bears directly on both equity and participatory de-
mocracy. The image that emerges for the students' role is then not that ofjunior re-
search statistician or utilitarian consumer of standard techniques. Instead, it is of
students as increasingly substantial participants in the discourse of public policy.
The important competencies for this participation are those of developing and
critiquing data-based arguments.
I should stress that the rationale I have given for the importance of statistics in
students' mathematics education is concerned with overall instructional goals. It
does not in itself imply that a particular instructional approach such as one involv-
ing investigations should be taken. Nonetheless, our decision to focus on the corn-
petencies of developing and critiquing data-based arguments did lead us to make
an important design decision when planning the teaching experiment. In particu-
lar, we ruled out an open-ended project approach in which students investigate is-
sues of personal interest by generating data and instead developed instructional
activities in which the students analyzed data sets created by others. However, we
also were aware that data do not speak for themselves but instead are the product of
a sequence of interpretive decisions and judgments (Latour, 1987; Roth, 1997).
For example, data embody assumptions both about which aspects of the situation
under investigation are relevant with regard to the issue at hand and about how
they should be measured. We therefore anticipated that the students would not ini-
tially be able to "look through data to the situation from which they were gener-
ated. It was for this reason that we developed the approach of talking through the
data-creation process so that the data might have a history for the students. It is ap-
parent from the sample episodes I have presented that this approach worked rea-
sonably well. As the teaching experiment progressed, the students, in fact,
assumed increasing responsibility for asking questions that related to the data cre-
ation process. Furthermore, although most of the classroom discussions focused
on analyses that the students had conducted, in the last few classroom sessions
they developed arguments on the basis of graphs created by others. In the course of
this transition, the students were developing the very competencies that make in-
creasingly substantial participation in public policy discourse possible.
In terms of the broader literature on equity, the approach we have taken to sta-
tistics instruction is broadly compatible with Delpit's (1988) admonition that stu-
dents should be taught explicitly what she calls the culture ofpower. Our approach
also makes contact with the equity pedagogy of Banks and Banks (1995), which
aims to help students from diverse cultural backgrounds develop the ways of
knowing needed to participate effectively within and maintain a just, democratic
society. Thus, although we look for further inspiration from scholars whose work
focuses on global, structural characteristics of schooling and society, we also con-
tend that a concern for equity is critical when considering issues traditionally ad-
dressed by mathematics educators. In particular, it is essential that we scrutinize
the overall goals we have for students' mathematics education and examine
whether they can be justified in terms of participation in a democratic society. I
will be more satisfied if our work in the area of statistics can serve as a useful ex-
ample in this respect.
ACKNOWLEDGMENTS
The analysis reported in this article was supported by National Science Foundation
(NSF) Grant REC 9604982 and by Office of Educational Research and Improve-
ment (OERI) Grant R305A60007. The opinions expressed here do not necessarily
reflect the view of either the NSF or OERI. An abridged version of this article was
presented as a plenary address at the Twenty-Second Meeting of the International
Group for the Psychology of Mathematics Education, Stellenbosch, South Africa,
July 1998. I thank Pat Thompson for the many conversations we have had about re-
lations among multiplicative reasoning, quantitative reasoning, and understanding
statistics conceptually. In addition, I am grateful to Tony Brown and Lyn English
for helpful comments on a previous draft of this article.
REFERENCES
Apple, M . (1995). Taking power seriously. In W. Secada E. Fennema, & L. Adajian (Eds.),New direc-
tionsfor equity in mathematicseducation (pp. 329-348). New York:Cambridge University Press.
Ball, D. L., & Cohen, D. K. (1996). Reform by the book: What is--or might be-the role of curriculum
materials in teacher learning and instructional reform? Educational Researcher, 25(9). 6 - 8 , 14.
Banks, C., & Banks, 1. (1995). Equity pedagogy: An essential component of multicultural education.
Theory Into Practice, 34, 152-158.
Bauersfeld, H. (1992). Integrating theories for mathematicseducation. For the Learning c$Marhemat-
ics, 12, 19-28.
40 COBB
Biehler, R. (1993). Software tools and mathematics education: The case of statistics. In C. Keitel & K.
Ruthven (Eds.), Learning from coniputers: Mathematics education and technology (pp. 68-100).
Berlin, Germany: Springer-Verlag.
Biehler, R., & Steinbring, H. (1 991). Entdeckende Statistik, Strenget-und-Blatter, Boxplots: Konzepte,
Begrundungen und Enfahrungen eines Untenichtsversuches [Explorations in statistics,
stem-and-leaf, boxplots: Concepts, justifications and experience in a teaching experiment]. Der
Mathematikunterricht, 37(6), 5-32.
Bowers. J., Cobb, P., & McClain, K. (in press). The evolution of mathematical practices: A case study.
Cognition and Instruction.
Cobb, G. W. (1997). Moreliteracy is notenough. InL. A. Steen (Ed.), Whynumberscount: Quantitative
literacy for tomorrow's America (pp. 75-90). New York: College Entrance Examination Board.
Cobb, P. (in press). Conducting classroom teaching experiments in collaboration with teachers. In R.
Lesh & E. Kelly (Eds.), New methodologies in mathematics and science education. Mahwah, NJ:
Lawrence Erlbaum Associates, Inc.
Cobb, P., & Yackel, E. (1996). Constructivist, emergent, and sociocultural perspectives in the context of
developmental research. Educational Psychologist, 31, 175-190.
Cobb, P., Yackel, E., &Wood, T. (1989). Young children's emotional acts while doing mathematical
problem solving. In D. B. McLeod & V. M. Adams (Eds.), Affect and mathematical problem solv-
ing: A new perspective (pp. 1 17-148). New York: Springer-Verlag.
Confrey, J., & Lachance, A. (in press). A research design model for conjecture-driven teaching experi-
ments. In R. Lesh & E. Kelly (Eds.), New methodologies in mathematics and science education.
Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
de Lange, I., van Reeuwijk, M., Bunill, G., & Romberg, T. (1993). Learning and testing mathematics in
context. The case: Dara visualization. Madison: University of Wisconsin, National Center for Re-
search in Mathematical Sciences Education.
Delpit, L. (1988). The silenced dialogue: Power and pedagogy in educating other people's children.
Harvard Educational Review, 58,280-298.
Dewey, J. (1980). Democracy and education. In J. A. Boydston (Ed.), John Dewey: The middle works,
1899-1924 (Vol. 9, pp. 1-370). Carbondale, IL: Southern Illinois University Press.
Dorfler, W. (1993). Computer use and views of the mind. In C. Keitel & K. Ruthven (Eds.), Learning
from computers: Mathematics education and technology (pp. 159-186). Berlin, Germany:
Springer-Verlag.
Erickson, F. (1986). Qualitative methods in research on teaching. In M. C. Wittrock (Ed.), The hand-
book of research on teaching (3rd ed., pp. 119-161). New York: Macmillan.
Forman, E. (1996). Forms of participation in classroom practice: Implications for learning mathematics.
In P. Nesher, L. Steffe, P. Cobb, G. Goldin, & B. Greer (Eds.), Theories of mathematical learning
(pp. 115-130). Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
Gearhart, M., Saxe, G. B., Stipek, 0. J., Bennett, T., Hakansson, S., & Rhine, S. (1994, April). Inte-
grating assessment and instruction in elementary mathematics: A frameworkfor staffdevelopment.
Paper presented at the annual meeting of the American Educational Research Association, New Or-
l e i s , LA.
Gravemeijer, K. E. P. (1994). Developing realistic mathematics education. Utrecht, The Netherlands:
Cd-O Press.
Gravemeijer, K.E. P., Cobb, P., Bowers, J., & Whitenack, J. (in press). Symbolizing, modeling, and in-
structional design. In P. Cobb, E. Yackel, & K. McClain (Eds.), Symbolizing andcommunicating in
mathematics classrooms: Perspectives on discourse, tools, and instructional design. Mahwah, NJ:
Lawrence Erlbaum Associates, Inc.
Hancock, C. (in press). The medium and the curriculum: Reflection on transparent tools and tacit mathe-
matics. In A. DiSessa, C. Hoyles. R. Noss, & L. Edwards (Eds.), Computers andexploratory learn-
ing. Heidelberg, Germany: Springer-Verlag.
Harel, G., & Confrey, J. (Eds.). (1994). The development of multiplicative reasoning in the learning of
mathematics. Albany: State University of New York Press.
Hicks, D. (1996). Contextual inquiries: A discourse-oriented study of classroom learning. In D. Hicks
(Ed.), Discourse, learning, andschooling (pp. 104-141). New York: Cambridge University Press.
John-Steiner, V. (1995). Spontaneous and scientific concepts in mathematics: A Vygotskian approach.
In L. Meira & D. Camher (Eds.). Proceedings of the 19th International Conference of the Interna-
tionul Groupfor the Psychology ofMathematics Education (Vol. l , pp. 30-44). Recife, Brazil: Pro-
gram Committee of 19th PME Conference.
Kamii, C. K. (1985). Young children reinvent arithmetic. New York: Teachers College Press.
Kaput, J. J. (1991). Notations and representations as mediators of constructive processes. In E. von
Glasersfeld (Ed.), Constructivism in mathematics education (pp. 53-74). Dordrecht, The Nether-
lands: Kluwer.
Konold, C., Pollatsek, A., Well, A,, & Gagnon, A. (1996, July). Students' analyzing data: Research cfl
critical barriers. Paper presented at the Roundtable Conference of the International Association for
Statistics Education, Granada Spain.
Krummheuer, G. (1995). The ethnography of argumentation. In P. Cobb & H. Bauersfeld (Eds.), The
emergence of mathematical meaning: Interaction in classroom cultures (pp. 229-269). Hillsdale,
NJ: Lawrence Erlbaum Associates, Inc.
Ladson-Billings. G. (1995). Making mathematics meaningful in multicultural contexts. In W. Secada,
E. Fennema, & L. Adajian (Eds.). New directions ,for equity in mathematics education (pp.
126-145). New York: Cambridge University Press.
Lampert, M. (1990). When the problem is not the question and the solution is not the answer: Mathemat-
ical knowing and teaching. American Educutional Research Journal, 27, 2 9 4 3 .
Latour, B. (1987). Science in action. Cambridge, MA: Harvard University Press.
Lave, J. (1993). Word problems: A microcosm of theories of learning. In P. Light & G. Butterworth
(Eds.), Context and cognition: Ways oflearning and knowing (pp. 74-92). Hillsdale, NJ: Lawrence
Erlbaum Associates, Inc.
Lave, J. (1996). Teaching, as learning, in practice. Mind, Culture, and Activity, 3, 149-164.
Lave, J., & Wenger, E. (1991). Situated learning: Legitimate peripheral participation. Cambridge,
England: Cambridge University Press.
Lehrer, R., & Romberg, T. (1996). Exploring children's data modeling. Cognition and Instruction, 14,
69-108.
McClain, K.. Cobb, P., Gravemeijer, K., & Estes, B. (in press). Developing mathematical reasoning
within the context of measurement. In L. Stiff (Ed.), Developing mathematical reasoning (1999
Yearbook of the National Council of Teachers of Mathematics). Reston, VA: National Council of
Teachers of Mathematics.
McGatha, M., Cobb, P., & McClain, K. (1998, April). An analysis cflstudents' statistical understand-
ings. Paper presented at the annual meeting of the American Educational Research Association, San
Diego, CA.
Mehan, H., & Wood, H. (1975). The reality ofethnomethodology. New York: Wiley.
Meira, L. (1995). The microevolution of mathematical representations in children's activity. Cognition
and Instruction, 13, 269-3 13.
Moll, L. (1997). The creation of mediating settings. Mind, Culture, m d Activity, 4, 191-199.
Nemirovsky, R., &Monk, S. (in press). "If you look at it the other way .. .": An exploration into the na-
ture of symbolizing. In P. Cobb, E. Yackel, & K. McClain (Eds.), Symbolizing and communicating
in mathematics classrooms: Perspectives on discourse, tools, and instructional design. Mahwah,
NJ: Lawrence Erlbaum Associates, Inc.
Pea, R. (1993). Practices of distributed intelligence and designs foreducation. In G. Salomon (Ed.), Dis-
tributed cognitions (pp. 47-87). New York: Cambridge University Press.
Piaget, J. (1973). To understand is to invent. New York: Grossman.
42 COBB
Roth, W. -M. (1997). Where is the context in contextual word problems?: Mathematical practices and
products in grade 8 students' answers to story problems. Cognition and Instruction, 14, 487-527.
Roth, W. -M., & McGinn, M. K. (1998). Inscriptions: Towards a theory of representing as social prac-
tice. Review of Educational Research, 68, 3 5 4 0 .
Saxe, G. B. (1991). Culture and cognitive development: Studies in mathematical understanding.
Hillsdale. NJ: Lawrence Erlbaum Associates, Inc.
Secada, W. G. (1992). Race, ethnicity, social class, language, and achievement in mathematics. In D.
Grouws (Ed.), Handbook of research on mathematics teaching and [earning (pp. 623460). New
York: Macmillan.
Sfard, A. (in press). Symbolizing mathematical reality into being. In P. Cobb, E. Yackel, & K. McClain
(Eds.), Symbolizing and communicating in mathematics classrooms: Perspectives on discourse.
tools, and instructional design. Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
Shotter. J. (1995). In dialogue: Social constructionism and radical constructivism. In L. P. Steffe & J.
Gale (Eds.), Constructivismin education (pp. 41-56). Hillsdale. NJ: Lawrence Erlbaum Associates,
Inc.
Simon, M. A. (in press). Research on mathematics teacher development: The teacher development ex-
periment. In R. Lesh & E. Kelly (Eds.), New methodologies in mathematicsand science education.
Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
Stephan, M. ( 1998). Supporting the development of one first-grade classroom's conceptions of mea-
surement: Analyzing students' learning in social context. Unpublished doctoral dissertation,
Vanderbilt University, Nashville, TN.
Streefland, L. (1991). Fractions in realistic mathematics education: A paradrgm of developmental re-
search. Dordrecht, The Netherlands: Kluwer.
Thompson, P. (1994). The development of the concept of speed and its relationship to concepts of rate.
In G. Hare1 & J. Confrey (Eds.), The development of multiplicative reasoning in the learning of
mathematics (pp. 181-287). Albany: State University of New York Press.
Toulmin, S. (1969). The uses of argument. Cambridge, England: Cambridge University Press.
Treffers, A. (1987). Three dimensions: A model of goal and theory description in mathematics instruc-
tion-The Wiskobas Project. Dordrecht, The Netherlands: Reidel.
van Oers, B. (1996). Learning mathematics as meaningful activity. In P. Nesher, L. Steffe, P. Cobb, G.
Goldin, & B. Greer (Eds.), Theories of mathematical learning(pp.91-1 14). Mahwah, NJ: Lawrence
Erlbaum Associates, Inc.
Voigt, J. (1995). Thematic patterns of interaction and sociomathematical norms. In P. Cobb & H.
Bauersfeld (Eds.), The emergence of mathematical meaning: Interaction in classroom cultures (pp.
163-201). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.
Warren, B., & Rosebery, A. S. (1995). Equity in the future tense: Redefining relationships among teach-
ers, students, and science in linguistic minority classrooms. In W. Secada, E. Fennema, &L. Adajian
(Eds.), New directionsfor equity in mathematics education (pp. 298-328). New York: Cambridge
University Press.
Wertsch, J. V. (1994). Struggling with the past: Some dynamics of historical representation. In M.
Carretero & J. Voss (Eds.), Cognitive and instructionalprocesses in history and the social sciences
(pp. 323-338). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.
Whitson, J. A. (1997). Cognition as a semiotic process: Grounding, mediation, and critical reflective
transcendence. In D. Kirshner & J. A. Whitson (Eds.), Social, semiotic, and neurological perspec-
tives (pp. 97-149). Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
Wilensky, U. (1997). What is normal anyway? Therapy for epistemological anxiety. Educational
Studies in Mathematics. 33, 171-202.
Yackel, E. (1995). The classroom teaching experiment. Unpublished manuscript, Purdue University
Calumet. Department of Mathematical Sciences. Hamrnond, IN.
Yackel, E. (1997, April). Explanation as an interactive accomplishment: A case study of one sec-
ond-grade mathematics classroom. Paper presented at the annual meeting of the American Educa-
tional Research Association, Chicago.
Yackel, E., & Cobb, P. (1996). Sociomathematical norms, argumentation,and autonomy in mathemat-
ics. Journal for Research in Mathematics Education, 27, 458-477.
Zevenbergen, R. (1996). Constructivism as liberal bourgeois discourse. Educational Studies in Mathe-
matics, 3I, 95-1 14.

Cobb 1999

Uploaded by

Copyright:

Available Formats

Cobb 1999

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cobb 1999

Uploaded by

Copyright:

Available Formats

This article was downloaded by: [University of Otago]

On: 04 January 2015, At: 15:00

Mathematical Thinking and

To link to this article: http://dx.doi.org/10.1207/s15327833mtl0101_1

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of

Individual and Collective Mathematical

1994) in that it involves both instructional development and classroom-based re-

(guded by domain-specific (guided by inorpretive

FIGURE 1 The developmental research cycle.

Interpretations of classroom events reflect suppositions and assumptions about

whole-class discussions of students' mathematical interpretations and solutions.

It is apparent from this illustration that analyses of classroom mathematical

BACKGROUND TO THE TEACHING EXPERIMENT

The teaching experiment was carried out with 29 twelve-year-old students in a

or inscriptions would emerge as different ways of structuring distributions. For

gradual emergence of a single, multifaceted mathematical notion rather than a

ent an analysis of two of the classroom mathematical practices that emerged

EMERGENCE OF THE FIRST MATHEMATICAL

Value Tool I Count I10

://///////A Green - Pink

FIGURE 2 The first computer minitool

dicating a difference in the two brands of batteries that he considered significant,

Jennifer: Even though 7 of the 10 longest-lasting batteries are Always Ready

EMERGENCE OF THE SECOND MATHEMATICAL

After Speed Trap

Before Speed Trap

FIGURE 3 The second computer rninitool

to build on the students' participation in the first mathematical practice by support-

teachers legitimized Janice's interpretation and indicated that it was particularly

FIGURE 4 The cholesterol data.

FIGURE 5 The cholesterol data organized into four equal groups.

In describing the intent of Sally and Madeline's analysis as that of investigating

Janice: I think it's an adequate way of showing the information because

FIGURE 6 The AIDS protocol data.

FIGURE 7 First analysis of the AIDS protocol data

David: The highest range of the numbers?

OLD PROGRAM N E W PROGRAM

FIGURE 8 Third analysis of the AIDS protocol data

Ken: I've got a suggestion. I don't know how to do it [inaudible]. Is there

In the ensuing exchange, several students calculated the percentages of data

tive characteristics of distributions. Participation in this practice involved reason-

subsequent sessions that a particular way of structuring data had to be justified by

Explain how the data

Explain why this way of

Explain the statistics

FIGURE 11 General scheme of argumentation for data analysis

orating with a group of teachers to revise the instructional sequence.

goal is to support the development of a professional teaching community that

Thus far, in focusing on collective practices, I have emphasized the taken-as-shared

that it also blinds us to the taken-as-shared basis for mathematical communication

sis of ongoing observations of individual students' activity. In doing so, we once

Furthermore, as Lave (1996) observed, schooling as a social institution involves an

In contrast to these two common rationales, we find a third justification to be far

decisions are justified by presenting arguments based on the analysis of data. In

You might also like