Anne R. Teppo. Qualitative Research Methods in Mathematics Education PDF

ISSN 0883-9530
Journal
inesearch
Mathematics
Education
Qualitative
Research Methods
in
Mathematics Education
Monograph Number 9
National Council of Teachers of Mathematics

A MonographSeries of the National Council of Teachersof Mathematics
The JRME Monograph Series is published by the Series Editor

Editorial Panel as a supplement to the journal. Each DOUGLAS A. GROUWS,University of Iowa, Iowa
monographhas a single theme related to the learning City, IA 52242
or teaching of mathematics. To be considered for
publication, a manuscript should be (a) a set of Editorial Panel
reports of coordinated studies, (b) a set of articles LAURIE D. EDWARDS, University of California at
synthesizing a large body of research, (c) a single SantaCruz;GRAHAMA. JONES,IllinoisStateUniver-
treatise that examines a major research issue, or (d) a sity; DAVID KIRSHNER,Louisiana State University;
JUDIT MOSCHKOVICH,TERC, Cambridge,Massa-
reportof a single research study that is too lengthy to
be published as a journal article. chusetts; NEIL A. PATEMAN, University of Hawaii,
Chair; ANNIE SELDEN, Tennessee Technological
Any person wishing to submit a manuscript for University;JANESWAFFORD,Illinois State Universi-
considerationas a monographshould send four copies ty; JOHNVAN DE WALLE, VirginiaCommonwealth
of the complete manuscript to the monograph series University, Board Liaison; STEVEN R. WILLIAMS,
Brigham Young University; VICKI ZACK, St.
editor. Manuscripts should be no longer than 200
George's School, Montreal,Quebec
double-spaced, typewritten pages. The name,
affiliations, and qualifications of each contributing
authorshould be included with the manuscript.
Manuscriptsshouldbe sent to Era Yackel; Department

of Mathematics, Computer Science, and Statistics;
PurdueUniversity-Calumet; Hammond,IN 46323.
Qualitative Research Methods
in
Mathematics Education
edited by
Anne R. Teppo
MontanaState University-Bozeman
National Council of Teachers of Mathematics

Copyright ? 1998 by
THE NATIONAL COUNCIL OF TEACHERS OF MATHEMATICS, INC.
1906 Association Drive, Reston, Virginia 20191-1593
All rights reserved
Library of Congress Cataloging-in-Publication Data:
Qualitativeresearchmethodsin mathematicseducation/ edited by Anne

R. Teppo.
p. cm. - (Journalfor researchin mathematicseducation.
Monograph; no. 9)
Includesbibliographicalreferences.
ISBN 0-87353-459-X
1. Mathematics-Sstudy and teaching-Research. I. Teppo,Anne R.
II. Series.
QA11.Q32 1998
510'.71-dc21 98-36575
CIP
The publicationsof the NationalCouncil of Teachersof Mathematicspresenta

varietyof viewpoints.The views expressedor impliedin this publication,unlessoth-
erwise noted, should not be interpretedas official positions of the Council
Printed in the United States of America

Table of Contents
Authors .................................................... iv
Acknowledgments.. ........... ........................... v
Abstract .................................................... vi
Chapter1 Diverse Ways of Knowing
Anne R. Teppo .................................. 1
Chapter2 Towarda Definition for Research
Susan Pirie . . ....................... 17
Chapter3 The EpistemologicalBasis of QualitativeResearchin
MathematicsEducation:A PostmodernPerspective
Paul Ernest ...................................22
Chapter4 ObservingMathematicalProblemSolving Through
Task-BasedInterviews
GeraldA. Goldin ............................... 40
Chapter5 Phenomenography:Exploringthe Roots of Numeracy
Dagmar Neuman ............................... 63
Chapter6 WorkingTowardsa Design for QualitativeResearch
Susan Pirie ...................................79
Chapter7 Studyingthe ClassroomNegotiation of Meaning:
ComplementaryAccounts Methodology
David J. Clarke ................................98
Chapter8 The Centralityof the Researcher:Rigor in a
ConstructivistInquiryinto MathematicsTeaching
Barbara Jaworski ............................ 112
Chapter9 Using a Computerin Synthesis of QualitativeData
JudithMousley, Peter Sullivan,and Andrew Waywood . 128
Chapter10 Using Researchas a Stimulusfor Learning
Beatriz S. D'Ambrosio .... .............. 144
Chapter11 Where Do We Go From Here?
Susan Pirie .................................. 156
References .................................... ....... 164
iii
Authors
David J. Clarke DagmarNeuman

Associate Dean Departmentof Educationand
Faculty of Education EducationalResearch
University of Melbourne GoteborgUniversity
Parkville,Victoria, 3052, Goteborg,Sweden
Australia
Susan Pirie
Beatriz S. D'Ambrosio Professorof MathematicsEducation
Associate Professor Departmentof CurriculumStudies
School of Education University of British Columbia
IndianaUniversity Purdue Vancouver,BC V6T 1Z4
University Indianapolis
Indianapolis,IN 46202 Peter Sullivan
Assistant Professor
Paul Ernest Faculty of Education
Readerin MathematicsEducation AustralianCatholic University-
University of Exeter ChristCampus
Exeter EX1 2LU United Kingdon Oakleigh,Victoria, 3166, Australia
GeraldA. Goldin Anne R. Teppo

Professorof Education,Mathematics, AdjunctInstructor
and Physics Departmentof MathematicalSciences
Centerfor Mathematics,Science and MontanaState University-Bozeman
ComputerEducationand Bozeman, MT 59717
GraduateSchool of Education
RutgersUniversity AndrewWaywood
New Brunswick,NJ 08903 Lecturerin MathematicsEducation
AustralianCatholicUniversity-
BarbaraJaworski ChristCampus
University Lecturerin Educational Oakleigh,Victoria, 3166,
Studies Australia
OxfordCentrefor Mathematics
EducationResearch
Oxford0X2 6PY United Kingdom
JudithMousley
Senior Lecturer
Faculty of Education
Deakin University
Geelong, Victoria, 3217, Australia
iv
Acknowledgments
The editorwishes to thankthose who providedsupportand encouragementfor

the theme of this monographin the early stages of the project. Theirinvolvement
in the field of mathematics education and their individual research interests
helped clarify the natureof the product. Gratefulthanksare extendedto Deborah
Ball, Catherine Brown, Jere Confrey, Robert Davis, Kathleen Heid, Carolyn
Maher,TerezinhaNunes, Leslie Steffe, and Era Yackel. Thanksare also due to
the contributingauthorswhose ideas and continual input defined the focus and
producedthe end result.
v
Abstract
The chaptersin this monographdescribequalitativeresearchmethods used to

investigate students' and teachers' interactionswith school mathematics.Each
contributingauthoruses data from his or her own researchto illustratea partic-
ulartechniqueor aspect of researchdesign. The differentchapterspresenta wide
rangeof methods,representinga varietyof goals and perspectives.Ratherthan a
comprehensive reference manual, this monograph illustrates the diversity of
methods available for qualitativeresearchin mathematicseducation.
The monographbegins with a discussion of key elements thatcontributeto the
dynamic and evolving domain of mathematicseducationresearch.Background
informationis then providedthat relates to the philosophical and epistemologi-
cal assumptionsunderlyingall qualitativeresearch.In the chaptersthat follow,
actual studies present the contexts for discussions of researchdesign and tech-
niques. Issues of researchdesign include the importanceof making explicit the
underlyingtheoreticalassumptions;the selection of an appropriatemethodology;
the interpretative,intersubjectivenature of analysis; and the establishmentof
reliabilityand validity. Specific data collection techniquesinclude clinical inter-
views, stimulatedrecall interviews,open-endedsurveyquestions,and field notes
and video or audio taping to record classroom events. Methods of analysis
include participantvalidation,the categorizationof data throughconstantcom-
parison and software indexing and retrieval,phenomemographicanalysis, and
the identification of empirical examples of theoretical constructs. The mono-
graphends with a discussion of general issues, including the role of theory and
the establishmentof criteriafor judging the goodness of qualitativeresearch.
vi
1
Chapter 1
Diverse Ways of Knowing
Anne R. Teppo
The chapters in this monograph describe qualitative methods used in mathe-

matics education research. Rather than write a comprehensive manual, con-
tributing authors describe specific methods from their own studies to illustrate
the range of techniques used to investigate students' and teachers' interactions
with school mathematics. Each chapter focuses on only one aspect of the
author's research to provide a more in-depth discussion of that particular facet of
the overall design.
A goal of this book is to stimulate dialogue. Mathematics education research
supports a variety of methodological perspectives and goals of inquiry, which
makes communication across perspectives difficult and the need for dialogue
imperative. Creating dialogue is not easy. Fenstermacher and Richardson (1994)
suggest enjoying each speaker on his or her own ground. However, as Cobb
(1995) points out, this task requires a decentering of participants to allow them
to "appreciate the other's position ... even when it is difficult to argue for it from
their own perspective" ( p. 25). Such decentering involves developing a sensi-
tivity to the contextual meaning of others and regarding each research result as
the product of a particular line of inquiry that must be viewed in the context of
that inquiry (Bredo, 1994).
This chapter begins with an examination of the general context within which
qualitative research in mathematics education is placed and then discusses the
particular contexts within which the contributing authors situate each chapter. To
set the stage, key elements are described that contribute to the dynamic and
evolving domain of qualitative research in mathematics education, including the
acceptance of qualitative research as an important methodology in educational
inquiry, a broadening of perceptions of the nature of mathematics and mathe-
matics education, and a recognition of the complexity of classroom mathematics
teaching and learning. Against this broad background, the chapter concludes
with a summary of the contributions of the other authors and a brief discussion
of the ways in which these contributions illustrate the diversity of qualitative
research methods employed in mathematics education.
2 Diverse Waysof Knowing
QUALITATIVERESEARCH
Qualitative research has been described as a field of inquiry (Denzin and

Lincoln, 1994) thatcuts across disciplines and subjectmatter,finding application
in such areas as anthropology,sociology, psychology, sociolinguistics, political
science, and education.Qualitativeresearchfocuses on processes, meanings,and
the socially constitutednatureof reality and providesinsights into the phenome-
na being studiedthat cannotbe obtainedby othermeans. Denzin and Lincoln, in
their comprehensive Handbook of Qualitative Research (1994), offer the fol-
lowing generic definition that recognizes the cross-disciplinarynature of the
field and the fact that qualitativeresearchmeans many things to many people:
Qualitative researchis multimethod naturalistic
in focus,involvinganinterpretative,
to its
approach subject matter.
Thismeans that researchers
qualitative study thingsin
theirnaturalsettings,attempting to makesenseof, or interpret,phenomena in terms
of meaningspeoplebringto them.Qualitative researchinvolvesthestudieduse and
collectionof a varietyof empiricalmaterials-casestudy,personalexperience, intro-
spective,life story, interview,observational, and visual
historical,interactional,
texts-that describeroutineandproblematic momentsandmeaningsin individuals'
lives. (p. 2)
Qualitative Research in Education
Anthropologistsand sociologists have employed qualitativeresearchmethods

since the turnof the century,but only in the last 30 years has this type of research
become regarded as a legitimate method for educational inquiry (LeCompte,
Millroy, & Preissle, 1992). The road to acceptancehas been at times confronta-
tional.In a purposivelyprovocativearticle,Gage (1989) describedwhathe termed
the "paradigmwars"of educationalresearchin the 1980s (p. 10). Protagonists
adheredto one of three alternativeand competing research paradigms,which
Gage characterizedas the objective-quantitativeof the naturalsciences, the inter-
pretative-qualitativeof anthropology,and the critical-theoreticof sociology and
political science. These debates were conducted along disciplinary lines as
researcherswithinpsychological,anthropological,and social science orientations
arguedover the hegemony of their particularmethodologicalframework.Critics
of the scientific paradigmdecried educationalpsychology's long-held influence
on researchon teaching and learning (exemplified by process-productstudies)
and pushedfor increaseduse of qualitativestudies focused on personalmeanings
and socially constructedreality. (Ernest,Chapter3 this volume, presentsa com-
parisonof the threealternativeresearchparadigmsand theirunderlyingontologi-
cal and epistemologicalbeliefs.)
Emerging from the methodological debates of the 1980s, educational
researcherswithin the qualitativetraditionhave found "themselvesin the pecu-
liar position of having achieved orthodoxy"and have become partof the "dom-
inant methodological establishment"(LeCompte,Millroy, & Preissle, 1992, p.
xvi). The qualitative-quantitativedebate has been replaced with a recognition
AnneR. Teppo 3
that each paradigmoffers a differentway to focus researchon the complexities

of contemporaryeducation(Cizek, 1995).
Paradigmsare overarchingworldviews that representparticularbelief systems
aboutthe natureof knowledgeand how thatknowledgeis acquired.Because of the
differencesin theirfundamentalbeliefs, it is not easy to move betweenthem.Hence,
the characterizationof "competing"paradigms.Withina paradigm,however,it is
possibleto holddifferingperspectivesorpointsof view. As the qualitativeparadigm
has moved into mainstreameducationalresearch,its multipleperspectives,which
have long been a traditionwithinthis paradigm,have openedup an overwhelming
arrayof methodologiesfrom which to choose. Educationalresearchcurrentlysup-
ports "a diverse arrayof voices speakingfrom quite different,often contradictory
perspectivesand value commitments"(Donmoyer,1996, p. 19).
Diversity of Perspectives WithinMathematicsEducationResearch

The diversity of qualitative research perspectives employed in mathematics
education research, although not yet representative of the wide variation
employed across the entire educationalresearchfield, presents sufficient range
to elicit controversy. Recent commentaries (Cobb, 1994; Cobb, Jaworski, &
Presmeg, 1996; Cobb & Yackel, 1996; Greer, 1996; Bredo, 1994) address the
conflicting and complementaryperspectivesthat cognitive science, sociocultur-
al theory, and constructivismprovide for studyinglearningwithin the broadpar-
adigm of qualitativeresearch.Greer(1996, p. 182) characterizesthis situationas
creatinga "fermentof new ideas, liberalizationof methodology, and [an] open-
ness to concepts from many disciplines." The following brief discussion illus-
tratessome of the range of points of view currentlyavailablefrom which to con-
duct mathematics-educationinquiry.
Cobb (1994, 1995) describesthe applicabilityof the separate,but complemen-
tary, perspectives of constructivismand socioculturaltheory for mathematics-
education research.Learning,from the constructivistperspective, is seen as an
individual cognitive activity that involves the internalreorganizationof mental
schema. Although such learning takes place within a social setting, researchis
focused on the individual'smentalconstructions.In contrast,a socioculturalper-
spective views learningas the enculturationof an individualinto a communityof
practice, and the focus of inquiry is placed on the individual's participationin
social practice. These two perspectives provide complementaryways to frame
inquiry, with each perspective providing the figure against the backgroundfor
the other.Because each perspectivetells "halfthe story,"Cobb maintainsthatthe
selection of either the cognizing individual or the social interactionas the pri-
mary unit of analysis should be guided by the needs of any particulareducation-
al inquiry.
Cobb and Yackel (1996) have proposeda thirdperspectivethat uses the com-
plementarynatureof the constructivistand socioloculturalpoints of view. Their
social constructivistor "emergent"framework is used to study mathematical
learning as it occurs within the social contexts of the classroom. The develop-
ment of individualmeaning and the developmentof social meaningare taken as
being reflexively related in that neither can exist independentlyof the other.
Individual"constructionsare seen to occur as students participatein and con-
tributeto the practicesof the local community"(p. 185).
Researchmethods of educationalpsychology and cognitive science have also
been employed in mathematicseducation.The information-processingmodel of
humancognition used by cognitive psychologists in the 1960s and 1970s is now
recognized as inadequatefor capturingthe "complexity and richness of mathe-
matical activity" (Greer, 1996, p. 181). Recent focus on detailed studies of the
cognitive processes of individualsengaged in the performanceof everydaymath-
ematics (as opposed to academicmathematicaltasks) reflects a change in educa-
tional psychology's perspectiveon the study of the mind to one of situatedcog-
nition. Mind is regardedas an aspect of a given person-environmentinteraction.
Research from this perspective focuses on problems arising in the course of
everyday activities in which an individual's social and physical interactions
define the object of research(Bredo, 1994).
The precedingdiscussiontouches only briefly on the rangeof perspectivesand
relatedresearchmethodologiesthatis currentlybeing employed in mathematics-
education research. The disciplinary perspectives of sociology, anthropology,
and cognitive science and the theoreticalframeworkof constructivismpresent
multiple vantage points from which to launch inquiry into the complexity and
messiness of the classroom. In spite of this diversity, a common theme running
acrossthe differentperspectivesis the increasingimportancebeing given to local
context as the determinantof researchdesign. When the contributionsavailable
from each perspective are considered, the issue should not be which point of
view is better but which one is most useful and appropriatefor the problem at
hand. "Claimsthat [a particular]perspectivecapturesthe essence of people and
communities should be rejected for pragmaticjustifications that consider the
contextualrelevance and usefulness of a perspective"(Cobb, 1994, p. 13).
The wide range of frameworksof inquiry available for qualitativeresearch
reflects the diversity of the disciplines that have developed the various method-
ologies. Employing the techniques of a particularpoint of view involves more
than simply adoptinga set of researchpractices.Underlyingeach set are funda-
mental differencesin how one views the world, how the objects of study fit into
this view, and how knowledge about these objects can be acquired. Doing
researchalso means understandingthe underlyingways of thinkingimplicit in a
given perspective(Steffe & Wiegel, 1996).
A Need for Dialogue

The burgeoninglist of qualitativemethods employed in educationalresearch
has helped createa field in which there is little consensus concerningthe follow-
ing questions: What is educationalresearch?What should scholarly discourse
Anne R. Teppo 5
look like? What role should research play in education? (Donmoyer, 1996;
Fenstermacher& Richardson,1994; Lester,Kehle, & Birgisson, 1996). The exis-
tence of a proliferationof approachescan be viewed as daunting.Alternatively,
it can be taken as a sign that the field of educationalresearchis alive and well.
If we adopt the latter view, then dialogue and informed critique are needed to
maintainthe field's health in the face of diversity. (Goldin, Chapter4 this vol-
ume, presents an example of how the explicitation of methodology facilitates
dialogue.) Differences in points of view can be used as a mechanism for
progress.
It is by the very processof "misunderstanding"others-that is, interpretingtheir
claims and beliefs in slightlydifferenttermsthanthey do themselves-thatthe
processof communication actuallymovesforwardto new understandings. ... We
needto be similarenoughto makedialoguepossible,butwe alsoneedto be differ-
entenoughto makeit worthwhile. (Burbules& Rice, 1991,p. 409)
However, given the currentdiversityof perspectivesand competingparadigms
in educationalresearch,consensus may not always be possible. The process of
debateis worthwhileonly to the extent thatthose engaged undergosome change
in opinion-at the least, enablingthose who disagree to gain greaterinsight into
their own positions. What is importantis to encourage "healthy confusion"
(Fenstermacher& Richardson, 1994, p. 54)-to engage in open discussion; to
allow new and interesting, along with old, voices to be heard; and to make
explicit one's assumptionsabout one's research and related educationalgoals.
"Thereare as many worlds as ways to describe them" (Eisner, 1993, p. 6), and
we should celebrate the multiplicity of voices ratherthan seek synthesis into a
single perspective.
AN OVERVIEWOF THE FIELD OF MATHEMATICSEDUCATION
Disciplinary Perspective
The context of inquiryis made up of a complex web in which our underlying
beliefs are carriedout within the local dynamicsof a particularinvestigation.Not
only is the individual situatedwithin a given researchsetting, he or she is also
situatedwithin a particularepistemological,cultural,and genderedframeworkof
beliefs and values that both facilitate and constrainhow we perceive the world
and what we select for study within it.
Eisner (1993), using an historical perspective, illustrateshow particularper-
ceptions can influence the framingand examinationof educationalpractice.
Howwe answerthequestionof whetherhistoryis thetexthistorians writeorthepast
historianswriteaboutis crucialto ourownviewof whathistoryis and,therefore,to
whatis relevantfor helpingstudentsunderstand it. If historyis text,thentext must
continueto be centralto the teachingof history:To understand historyone has to
understand text.Butif historyis thepastaboutwhichhistorians write,thenanyform
of representationthatsheds light on the past is relevant,indeeda useful,way to
understand history.(p. 9)
Mathematicseducators face a similar problem in considering their field of

inquiry.Underlyingassumptionsconcerningthe natureand role of mathematics
in society and how it is to be taughtand learnedinfluenceboth educationalprac-
tice and research(Dossey, 1992). A recognitionof the changingface of 20th cen-
tury mathematics,the emergence of computersand powerful handheldcalcula-
tors, and the changing needs of contemporarysociety raise questions about the
appropriatenessof specific educationalcontent. Emerging theories of learning
suggest new approachesto traditionalclassrooms and question old assumptions
aboutthe kinds of mathematicalknowledge thatbest promoteunderstanding(R.
B. Davis, 1994). One role of researchin mathematicseducationis to illuminate
these issues and provide an information base for their resolution (Research
Advisory Committee, 1997). A great deal has been written about these issues
(e.g., R. B. Davis, 1994; Dossey, 1992). The following discussionrepresentsonly
the tip of an importantdisciplinaryiceberg thatbearscarefulscrutinyelsewhere.
The Nature of Mathematics
Mathematicshas been describedas the science of abstractpatternsand char-

acterized by its usefulness for organizing mental and empirical structures
(Devlin, 1997; Steen, 1990; van Oers, 1996). This characterizationonly hints at
the complex natureof a field thatis valued by some for its intellectualbeauty,by
othersfor its utilitarianapplications,andby still othersfor its emancipatoryprop-
erties in an increasinglytechnologicalworld. Mathematicscan be regardedas the
productof intellectualabstractionor as the processes that produce such a prod-
uct. It can be regardedas a static entity or as a fallible, creative, evolving activ-
ity that permeatesmany aspects of our daily lives.
The subject of mathematicsis multifaceted. Steen (1990) describes a set of
diverse perspectives that "illustratethe complexity of structuresthat support
mathematics"(p. 4). These perspectives,or "deepideas thatnourishthe growing
branchesof mathematics"(p. 3), can be thoughtof as (a) specific mathematical
structures,such as numbersor shapes;(b) mathematicalattributes,such as linear
or periodic;(c) actions, such as representor prove; (d) abstractions,such as sym-
bols or equivalence; (e) attitudes, such as wonder or beauty; (f) mathematical
behaviors,such as motion or iteration;or (g) mathematicaldichotomies, such as
discrete versus continuous.
Anothercharacteristicof the subjectis thatmany mathematicalentities exhib-
it a process/productduality reflecting an "interplay of form with content"
(Freudenthal,1991, p. 10). Symbolic representationsof such entities can be per-
ceived either as mathematicalprocesses or as the productsof these processes.
Mathematicalthoughtis characterizedby the ability to generalizedetail (process
or form) into structure(productor content)and to createnew conceptualentities
from an abstractionof this structure(Sfard, 1991; Tall, 1991; Teppo & Esty,
1994, 1995). This movementfrom form to contentmakes it possible to deal with
complexity by reducingdetail throughabstraction(Devlin, 1997; Dryfus, 1991).
Anne R. Teppo 7
In addition, the use of some form of external representationmakes it possible

thinkaboutand manipulatethe concepts underconsideration,which leads to new
levels of thought. The search for, and development of, adequate symbols to
express new levels of conceptualizationare a characteristicof the generative
powers of mathematics(van Oers, 1996).
The natureof mathematicsis also reflectedthroughthe activity and disposition
of its practitioners.Doing mathematics,which is differentfrom the mathematics
presentedin professionaljournals, includes, among other actions, exploring sit-
uations,searchingfor patterns,inventing strategies,using intuition,generalizing,
and abstracting-creating mathematics.Coupled with such activity is the devel-
opment of a mathematicaldisposition-a particularway of perceiving and doing
that enables a person to navigate his or her way around the domain. Breiter
(1997) describes this as "a diffuse kind of knowledge or competence that makes
performancepossible" (p. 3). Cobb and Yackel (1996) identify a set of reflex-
ively related beliefs, values, and "sociomathematicalnorms" that, developed
throughclassroominteractions,enables studentsto engage in autonomousmath-
ematical practice.
Mathematicsexists as an aim in itself of concernto practicingmathematicians,
as a tool applied in the pursuitof other disciplines or in everyday life, and as a
social and culturalartifact.As an aim in itself, it is self-generating,serving as "a
forceful motor for its own long-term development"(Freudenthal,1991, p. 3).
Withinthis area,mathematicalobjects serve as the raw materialfrom which other
mathematicalobjects are created.As a tool, it is used to mathematizeaspects of
humanexperiencethatexist outside the realm of mathematics(P. J. Davis, 1993;
Restivo, 1992). When employed in the course of everyday activities, such as in
buildinga house or selling goods, the termethnomathematicshas been developed
to emphasizethe culturalembeddednessof the tool (Nunes, 1992).
Our understandingof mathematicsis groundedin practice that is implicitly
social. In his sociology of mathematics,Restivo (1992) arguesthatmathematical
objects do not exist "independentof the flux of history and culture"(p. 3). These
objects are a humancreationthatare "embeddedin and ... embody world views"
(p. 103). It is necessary to think in terms of a "culturalconception of mathemat-
ics" (Radford,1997, p. 28). Mathematicsis also groundedin common sense and
stems from the systematization,organization,and reorganizationof contextually
bound experiences (Freudenthal,1991). "The effects of culture and society are
fundamentalto the way in which we come to know" (Radford,1997, p. 29).
The Nature of MathematicsEducation

It is importantfor researchersengaged in the processes of "giving meaning to
educationalevents" to situate these meanings against one's "assumptionsabout
the teaching and schooling that underlie those events" (Gitlin, 1990, p. 460).
How one thinksaboutmathematicseducation,how one defines what it means for
students to know and do mathematicsin school, is, in turn, affected by one's
views about the nature of mathematics,one's underlyingepistemological per-

spective, and one's educationalgoals.
P. J. Davis (1993) argues that today's world is characterizedby the large
degree to which mathematizationspermeateour daily lives, both in the humanis-
tic areas and in the sciences. Everyone uses mathematics at some level.
Consequently,it is the role of educationto enablecitizens to become awareof and
assess these mathematizations-to develop "mathematicalestreet smarts' that
enable [them]to formjudgmentsin the absence of technicalexpertise"(p. 192).
Breiter (1997) addresses the role of education from a more individual per-
spective, advocatingthe developmentof mathematicaldisposition-an "intuitive,
perception-likeunderstanding... that makes lifelong learningin mathematicsa
possibility"(pp. 3, 5). Renz (1997) offers a similarvision for mathematicsedu-
cation, stating a minimal list of competencies requiredfor an uncertainfuture.
Students should know how to solve problems by asking others; should be able
to communicateby speaking,writing, and drawing;shouldbe awareof the exis-
tence of multiple solutions; and should understandthat most problems do not
have definite solutions. Cobb and Yackel (1996) introduceadditionaldisposi-
tional attributesfrom their social constructivistperspective. In particular,math-
ematics education should foster the development of sociomathematicalnorms
including the developmentof studentautonomyand the ability to judge mathe-
matical solutions on the basis of theirdifferences, sophistication,efficiency, and
acceptability.A common theme runningthroughthe differenteducationalcrite-
ria listed here is that what students believe and think about mathematics is
importantfor succeeding both in and out of school and for facilitating future
learning.
Mathematicseducation can be defined as formal schooling to distinguish it
from the ethnomathematicsof everydaylearning.The operationalizationof such
educationis then to take mathematicalknowledge, which was originally devel-
oped to be used ratherthantaught,and transformit into a teachableform (Greer,
1996). This transformationprocess must considernot only what knowledge is to
be learned,but also the natureof the knowledge and the kinds of experiencesthat
studentsare to develop (R. B. Davis, 1994). The focus of the transformational
process is on the developmentof appropriateeducationaltasks within effective
learningenvironments.
The Netherlandshas developed a programof realistic mathematicseducation
that employs the processes of horizontal and vertical mathematicizingas the
"teachableform" (Freudenthal,1991; Treffers, 1991). Conceptualdevelopment
proceeds from informal, context-bound experiences to mathematical for-
malisms. A given situation is horizontally mathematizedby students into a
model; through vertical mathematization,this model is transformed,again by
students, into formal mathematicalstructure.The emphasis in realistic mathe-
matics education is to use contextual situations that connect with children's
existing methods of working and that promote natural,furthergeneralizations
and abstractions. Education is seen as the process of guided reinvention in
Anne R. Teppo 9
which the learnerreinvents "mathematisingratherthan mathematics;abstract-

ing ratherthan abstractions;... algorithmisingratherthan algorithms;verbalis-
ing ratherthan language"(Freudenthal,1991, p. 49).
Learning mathematics is recognized as a social and cultural activity. Our
schools serve as one of the places in which studentsare introducedto the "mean-
ing of culturally approved mathematical signs, symbols, and techniques"
(Crawford,1996, p. 145). The question of whetherthis process is enculturation
(the assimilationof an existing tradition)or acculturation(the process of inter-
culturalborrowingto create a new and blended culture) is one of perspective.
When a mathematicsclassroomis examinedfromthe externalvantagepoint of an
educationalsystem, thatis, outside the cultureof the classroom,the processes of
education are seen as the enculturationof students as they interactwith more
knowledgeableothers.Fromthe point of view withinthe cultureof the classroom,
the differencesin beliefs and values of the participantssupporta view of accul-
turationin which the intersubjectivemeanings of the students and teacher are
negotiated in the process of constitutinga classroom mathematicscommunity
(Cobb, Jaworski,& Presmeg, 1996; Cobb & Yackel, 1996). Learningenviron-
ments can be thoughtof as being constructedby individualsin activityratherthan
as existing independentlyof the participants(Saxe & Bermudez, 1996).
Richards(1996) describesa type of mathematicsclassroomthatfostersaccultur-
ationinto a sharedcommunitythroughthe use of whathe calls, "inquirymath."In
such a classroom,the studentsand teacherparticipatein mathematicaldiscussion
andact mathematically-askingquestions,solvingproblemsthatareproblematicto
the solvers,posingconjectures,andlisteningto mathematicalarguments.The class-
room atmosphereallows takingrisks and makingmistakes,and the teacheris able
to trulylistento the studentsbecauseof his or herunderstanding of "thelargermath-
ematical picture that provides a context for the students' questions" (p. 74).
Mathematicalcommunicationand the negotiationof meaningtake place at a level
at which thereare two sides, andeach is able to listen to the other.
The preceding characterizationsillustrate the diversity of perspectives from
which school mathematicscan be viewed. Each perspectivehighlights a differ-
ent aspect of the complex reality of mathematicslearningwithin a socially and
culturallyconstitutedenvironment.As R. B. Davis (1994) points out in his dis-
cussion of what mathematicsstudents should learn, "people no more agree on
what they most value in an act of mathematicalproblemsolving than they do in
paintings or poems or symphonies. I know people who can't see why anyone
would want to listen to Bach" (p. 25).
Mathematics Education Research

The multiple perspectivesfrom which the natureof mathematicsis now being
considered and the variations in processes and contexts that are increasingly
being used to characterizeschool mathematicsreflect a paradigmshift from a
modernistto a postmodernistworldview. This shift representsa reconceptual-
ization of the natureof knowledge from a single and externalreality to a set of

multiple and subjectiverealities. Currenttrendsin researchin mathematicsedu-
cation reflect a similarparadigmshift from an emphasis on scientific or quanti-
tative studies to the use of qualitative,interpretativemethodologies (see Ernest,
Chapter3, this volume). "Researcherstoday are looking at aspects of mathemat-
ical learningin ways thatwere, if not unthoughtof, at least not common 25 years
ago" (Kieran, 1994, p. 583).
The goals of mathematicseducation researchreflect the diversity and com-
plexity of its subjectsof inquiry.At a generallevel, the goal of researchis to pro-
duce new knowledge. In a field that is still in the process of defining itself, this
knowledge provides an importantbase for progress in mathematicseducation
(Silver & Kilpatrick, 1994). How one characterizessuch progress, however,
depends on one's perspective.The field of mathematicseducationencompasses
the study and implementationof classroom instruction(in all its complexity) as
well as the formulationof theories of the developmentof mathematicalunder-
standing(in all its variations).
From a practicalperspective, there is a need for researchto addressthe con-
cerns of practicingeducatorsin the field and to presentresults that are accessi-
ble to them and are of immediateuse (Kennedy, 1997). The questions and data
from such studies also provide importantexamples of exemplary practice and
resources to stimulate reflection on relevant issues (Research Advisory
Committee, 1994; Renz, 1997). Researchis also needed to facilitatethe creation,
implementation,and evaluation of innovative curriculathat can move mathe-
matics programsbeyond simply making improvementsto those that currently
exist. In addition,researchersshould considerwhat role they can, or should,play
in the development of informed public discussion concerning educational
progress(R. B. Davis, 1992; ResearchAdvisory Committee, 1997).
Researchon practicecan also be used as an emancipatoryvehicle. The process-
es of inquiry,when conductedby practicingteachers,give them voice and power
and enable them to pursuelines of inquirydirectlyapplicableto their needs. The
teacher-researchermovement changes the natureof the traditionalrelationship
between the researcherand those studied by recognizing the value of teachers'
personalknowledge and providingthem with a vehicle for effecting change and
settingeducationalpolicy (Brown, 1997; Gitlin, 1990; Richardson,1994).
The incorporationof qualitativemethodologies into mathematicseducation
researchhas madeit possibleto investigatethe teachingandlearningof mathemat-
ics at new anddifferentlevels of complexityandfrommultipleperspectives.A goal
of much of this researchis to investigatethe processesof coming to know mathe-
matics both from the perspectiveof the cognizing individualand from within the
socioculturalinteractionsof the classroom.Such researchaids in the development
of explanatorymodels of whatconstitutesmathematicallearning(Steffe, 1996).
Researchhas also been used to expand our own conceptions of mathematics
and our perceptionsof what is possible in mathematicseducation.Attendingto
students'expressionsof theirways of thinking,doing, and describing,or "student
Anne R. Teppo 11
voice," promotesreconceptualizationsof one's own mathematicalunderstanding

and promotes diversity of mathematicalperspective (Confrey, 1995b). Careful
analysis of videotapes of groups of studentsworking throughproblematicsitua-
tions reveals students'abilities to develop powerfulmathematicalways of think-
ing (Maher& Martino,1996).
Multiple challenges face the field of mathematics education research. The
availability of diverse perspectives for launching inquiry will requirea greater
acceptanceof alternativeresearchmethodologies and a greatercommitmentfor
researchersto communicatewith one anotheracross divergentpoints of view. As
we recognize and accept the complexity of the questions we wish to ask and the
phenomenonwe wish to study, new methods will need to be identified that can
capturethis complexity in meaningful ways-requiring a "new set of explana-
tions and a new set of tools" (Schoenfeld, 1994, p. 703).
The dynamic natureof the field of mathematicseducationresearchis reflect-
ed in an interplaybetween emerginglines of inquiryand evolving researchmeth-
ods. At the same time, the growthin diversityof perspectivesmakes it more dif-
ficult to establish standardsof scholarly inquiryon which all participantsin the
field can agree. As Silver and Kilpatrick(1994) point out,
It maybe neitherpossiblenordesirableto forgea singleperspective outof themany
thatarefoundin ourfield.... [Whatis neededis a] spiritof greateropenness,toler-
ance,andrespectfor the workandideasof thosecolleagueswho shareneitherour
culturenorourtradition.(p. 763)
DIVERSE WAYS OF KNOWING

A study of the field of mathematicseducationresearchor, in the case of this
monograph,of a small partof one corer of the field, is a useful endeavor.Bishop
(1992, p. 720), commentingon its internationalscope, notes "theimmensely dis-
parateand complex [natureof the] field" and recommendsthatthose who are try-
ing to understandmathematicseducationand the natureof its researchcompare
andcontrasttheirworkwith others.This book providesa forumfor such an exam-
ination. It is not intendedto serve as a referencemanual.Rather,the intent is to
illustratethe diversityof methodsavailablefor researchers.
The purposeof this monographis primarilyto examine the processes by which
knowledge is generated (researchmethods). In so doing, that which is known
(reportedresults) becomes more clear. Thus, we seek to extend our knowledge
of mathematicseducation by examining the ways in which knowledge in the
field is created. We can increase our understandingof what we know if we
understandhow we come to know it.
The chaptersuse informationfrom actualstudies to illustratedifferentcompo-
nents of qualitativeresearchdesign, presentinga wide rangeof methodsand rep-
resenting a variety of goals and perspectives. Insteadof focusing on the results
of theirresearch,the authorsselect one facet of theirdesign and describein detail
how it contributesto the final product.
Chapters2 and 3 orient the readertowarda critical examinationof the mono-

graph. In Chapter2, Pirie raises the question "Whatmakes this research?"in
relationto the field of mathematicseducationand invites the readerto become a
participantin the process of creatinga definition. In Chapter3, Ernestprovides
background information related to the philosophical and epistemological
assumptionsunderlyingall qualitativeresearch.He describes the constructivist
theory of learning that undergirdsmuch of the researchreportedhere and con-
trastsquantitative,qualitative,and criticalresearchparadigms.
Chapters4 and 6 raise key issues regardingresearchdesign. Goldin (Chapter
4) discusses the importance of making explicit the theoretical assumptions
underlyingthe selection of the researchtask and the methods of data collection
and analysis-describing how protocolsfor in-depthclinical interviewsare used
to study individual children's problem-solvingskills. Pirie (Chapter6) focuses
on the decision-makingsteps, requiredfor robustresearchdesign, that she used
to meet her goal of developing theory to describe how pupil-pupil discussion
facilitates mathematicalunderstanding.
The other chapterspresenta wide range of approachesand focus on different
aspects of the total researchdesign. Neuman (Chapter5), using a phenomeno-
graphicperspective,describes the model that she developed to characterizethe
variationin the ways thatbeginning school childrenexperiencedaspects of sub-
traction.She presentsa detailedanalysis of cognitive constraintsthat make some
problems easy and others difficult in her discussion of the word problems she
selected for her clinical interviews.Clarke(Chapter7) describesa techniquefor
creatingintegrateddatasets consistingof transcriptsof classroomvideotapes,stu-
dents' interpretationsof episodes on these videotapes, and observerfield notes.
His analysesof these datasets illustratethe use of multipleperspectivesfor inves-
tigatingwhat it meansto "cometo know"somethingin a mathematicsclassroom.
Jaworski(Chapter8) describeshow she assignedsignificanceto classroomevents
in her study characterizingan "investigativeapproach"to teaching. She offers
examples of the detailedreportsused to recordnot only each incidentof interest,
but its classroom context, the interpretationsof this incident by herself and the
teacher,and the relationsof the incidentto her underlyingtheoreticalframework.
Mousley, Sullivan, and Waywood (Chapter9) presentinformationon the use
of the computerprogramNUD-ISTto analyze open-endedresponses to a large-
scale survey. The purposeof this analysis was to identify featuresthat members
of the mathematicseducationcommunitybelieved were desirablecomponentsof
a qualitymathematicslesson. D'Ambrosio (Chapter10) describesaspectsof pro-
fessional-developmentprogramsthatfocused on developing teacher-researchers.
She presents the steps used to move a group of preservice students toward an
understandingof the natureof qualitativeresearchand provides examples of in-
service teachers'reflectionson theiruse of researchwithintheirown classrooms.
The sequencing of chaptersrepresentsa continuous transitionof focus from
student to teacher. Goldin and Neuman report on the individual student's
meanings of mathematicalconcepts. Pirie and Clarke shift to a focus on the
Anne R. Teppo 13
development of meaning within a social context, the former within a small

group of students and the latter within the larger context of the whole class-
room. Jaworski and Mousley et al. reporton methods designed to characterize
the classroom experience from the teacher's perspective. D'Ambrosio moves
back to an individual focus, but one that is centered on the teacher. Finally,
Pirie (Chapter11) examines common themes runningthroughthe chaptersthat
address importantissues, including the role of theory and the establishmentof
criteriafor judging the goodness of qualitativeresearch.
ChapterPerspectives
To understandthe contributionsthatthe authorsmake,it is importantto place the
informationreportedin each chapterinto its properperspective.A useful analyti-
cal frameworkfor readingthe chaptersincludesidentifyingthe specific goals of the
researchandthe assumptionsthatunderlieeach work,the objectsof studyandtheir
situatedcontexts,the natureof the primarydatacollected, the unit of analysisand
the type of understandingsoughtby the analysis,and the applicabilityof the find-
ings to educationalpractice.The following comparisonsillustratehow this frame-
work can illuminatethe diverseperspectivesof the reportedresearchmethods.
Goldin and Neuman provide contrastingpurposes of researchand objects of
study within similarcontexts. Both researchersuse clinical interviews involving
carefully constructed mathematicaltasks. Although the primary data in each
study consist of children's observed behavior,the objects of study are very dif-
ferent. Neuman's findings are framed in terms of descriptionsof the variations
in the ways that childrenexperience a given mathematicalphenomenon,where-
as Goldin's findings relate to the growth of children's complex, internalrepre-
sentationalcapabilities. Each chapterpresents a different aspect of a complete
clinical study-Goldin discusses principles of interview design, and Neuman
focuses on how a phenomenonis experiencedand depictedin a model and on cri-
teria for establishingthe reliabilityand validity of this model.
The researchof Pirie, Clarke,and Jaworskiillustratethe ways in which choic-
es of methodology and analysis are drivenby the natureof the researchquestion
and the underlyingtheoreticalperspective.The primarydata in all three studies
were collected in classrooms, yet the objects of each study and the goals of
research were different. Pirie audiotapedthe conversationsof small groups of
studentsto develop a theoryexplaininghow classroomdiscussion facilitates stu-
dents' mathematicalunderstanding.Clarkeused videotapesof classroomlessons
and students' interpretationsof episodes in these tapes to investigate how the
process of "comingto know"was developed by classroomparticipants.Jaworski
used an ethnographicapproachto collect data centeredon teachers' interactions
with studentsto characterizean investigative approachto teaching.
The chaptersby Jawarskiand by Mousley, Sullivan, and Waywood illustrate
the subjective, interpretivenatureof qualitativeanalysis. Using descriptionsof
her interpretationprocesses, Jaworskihelps the readerunderstandthe meanings
she assigned to observedclassroombehaviorand her reasons for attributingsig-

nificance to particularincidents. Mousley et al. emphasize the subjectivenature
of the interpretationsthey made as they organized and categorized the survey
responses. They discuss the subjective role that language plays, both in their
respondents'descriptionsof mathematicslessons and in the researchers'catego-
rizationsof these responses.
D'Ambrosio highlightsthe constructivistnatureof the researchact. As preser-
vice studentsand in-service teachersparticipatedin the design and implementa-
tion of small-scale studies, they constructedtheir own understandingsof the
nature of reflective practice and, in the process, became more empowered,
autonomousdecision makers.
Unlike the otherauthorswho focus on aspectsof researchdesign, Neumanpre-
sents a partialdescriptionof her findings. This informationillustratesthe central
componentof phenomenographicresearch-the developmentof a model depict-
ing the variationsin ways that a collection of individuals experiences a given
phenomenon.A discussion of a proposed set of criteriafor claiming reliability
and descriptionsof ways to establish validity are illustratedin the context of the
author'sphenomenographicmodel.
Each chapter can also be comparativelypositioned within a set of contexts
common to researchin mathematicseducation.These contexts include the use of
mathematicaltasks, the mathematicsclassroom and its participants,the study of
the constructionof knowledge that is individually or socially derived, or both,
and the researcher'srole in data collection.
The chaptersdiffer on how mathematicaltopics areused-either as foreground
or as background for the investigation. Neuman's research and one of
D'Ambrosio's studies investigatestudents'understandingof specific mathemati-
cal content. In contrast,Goldin, Pirie, Clarke,and Jaworskistudy specific types
of mathematicalbehavior that exist across a range of mathematicalsituations.
Although not always the explicit object of study, the natureof the mathematical
task in these four chaptersis an integralpartof the context of each investigation.
Goldin, Neuman, and D'Ambrosio illustratethe importanceof in-depthanalysis
of mathematicalstructurein the design of researchtasks. In contrast,Mousley et
al. use an implicit mathematicalcontext.Respondentsto their surveyare askedto
imagine a qualitymathematicslesson andthento list characteristicsof this lesson.
Anothertheme runningthrougheach chapteris the way the classroomcontext
is used. At one extreme,Goldin and Neumansituatetheirresearchwithinclinical
interviews that use mathematicaltasks designed specifically for the interviews.
The classroom context appears only implicitly in Goldin's recognition of the
experiences that his subjects bring to the interviews. In contrast,Pirie, Clarke,
Jaworski,and D'Ambrosiopurposefullyincorporatethe realitiesof school math-
ematicsinto theirresearch,studyingmathematicalbehaviorthatis an integralpart
and direct consequence of this context. Mousley et al. make the mathematics
classroomthe explicit, but indirect,object of investigation,using others' impres-
sions of this context ratherthandirectobservationsas theirprimarydata.
Anne R. Teppo 15
The reportedresearchalso variesas to the authors'perspectiveson the sourceof

knowledgeconstruction.Goldin situatesknowledgeand reasoningsquarelyin the
head of each subject,regardingknowledgeconstructionas the buildingof internal
representationsby the individualsthroughinteractionwith a structuredenviron-
ment. Neuman, while studying individualchildren,focuses her researchon the
experientialrelationshipbetween subject and object. D'Ambrosio describes the
growthof understandingof the researchprocess by preservicestudentsas a con-
structiveactivity. In contrast,Pirie and Jaworskirepresenta social constructivist
orientationinvestigatingthe ways in which students'knowledge is constitutedin
the process of classroom interactions.Jaworskipresentsboth radical and social
constructiveperspectiveswith regardsto students'learningas she discusses how
her point of view changed from the formerto the latterduringthe course of her
research.Clarke,in his study of classroominteractionsaugmentedby individual
studentinterviews,considers a symbiotic relationshipin which the two perspec-
tives of individualand social constructionare mutuallydependentand supportive.
Several authorscharacterizethe knowledge derived from their researchactiv-
ity in terms of social constructions.In the interpretationof her data, Jaworski
describes the development of intersubjectivemeanings that occur between her-
self and the teachers involved in observed classroom incidents. Mousley et al.
comment on how theoryrelatedto qualitymathematicsteachingdeveloped from
sharedunderstandingwithin their group of researchers.
The chaptersalso illustratethe range of roles that the researchersassumed in
the collection of their data. Pirie describes how her selection of an unobtrusive
position in datacollection was determinedby the goals of her research-electing
to audiotapeinteractions to capturestudents' discussions that were unaffected
by any adultintervention.At the otherextreme,Jaworski,as a participantobserv-
er, actively interactedwith the teachers and students she studied. The clinical
interviewersin Goldin's study also directly interactedwith the children under
study. They employed carefully established protocols, however, to minimize
variationacross the interviews.
D'Ambrosio's use of a teacher'svoice to illustratethe developmentof teacher-
researcheractivities reinforcesthe vital role that researchmust play in the field
of mathematicseducation. Qualitativeresearch,with its ability to capturemore
of the reality and complexity of classroom experience, provides an appropriate
vehicle for investigating important issues as well as for narrowing the gap
between researchand practice.
A common threadlinking the chaptersin this book is the interpretativenature
of qualitativeresearchmethods.Interpretationis a necessarycomponentof tech-
niques that are designed to study, within complex learning environments,the
meanings thatparticipantsmake of their experiences and aspects of humancog-
nition that can only be inferredfrom overt behavior. Qualitativemethods lend
themselves well to this type of inquiry.
The chaptersalso illustratethe aim of qualitativeresearchdiscussed by Ernest,
whereby the particularis used to illuminate the general. Each authorpresents
concrete instances that "suggest, evoke, and illustrate" situations that exist
beyond the immediatecontext of each study. The specifics of the differenttech-
niques also serve as exemplars, and through a look at the particularsof the
research studies reportedhere, qualitativeresearchmethodology is illuminated
more generallyby this monograph.
It is recommendedthat this book be regardedas a whole. Insteadof providing
discrete descriptions of research techniques that can be taken separately, the
chapters, taken together, enrich the reader's understandingof each individual
contribution.Comparingand contrastingthe reportedinformationnot only com-
plement and extend one's understandingbut provide new windows on the field
of mathematicseducation. The chapters also show a glimpse of the power of
qualitativemethods, developed in other human sciences fields and modified to
fit new needs, to uncoverhithertoinaccessible,but importantaspects of the com-
plex reality of mathematicsteaching and learning.
17
Chapter 2
Toward a Definition for Research
Susan Pirie
Before the readerplunges into the remainingchaptersof this book, it is appo-

site to ask, even if initially the possible response might appearto be obvious,
"Whatdoes the title of this book, QualitativeResearchMethodsin Mathematics
Education,mean?"If we unpackthe wording a little, two notions bear scrutiny:
"qualitativemethods"and "researchin mathematicseducation."The first is per-
haps a predictablefocus for an early chapterin a book such as this; the second,
however, is a more fundamentalconcern.
Mathematicseducationis congruentwith neithermathematicsnor education.
Mathematicseducationis an emergingdiscipline, no longer in its infancy yet not
fully adult.We are in the formativeadolescentyears when it behooves us to seek
to establish our identity as a legitimate, independent,academic community.To
do this we must, among many other tasks, addressthe notions of "research"and
"methods"as they apply to the field within which we work. It is time for us to
put aside the debate that tries to uphold or refute the supremacyof quantitative
over qualitative methods. Neither has merit in itself. The appropriatenessof
methods and methodologies espoused by researcherscan be consideredonly in
the light of the intentionsof the specific researchbeing undertaken.
As the discipline of mathematicseducationcomes to be more clearly defined,
our prime concern should be the business of deciding what constitutethe appro-
priateareasfor inquiry.The process of defining the natureof acceptableresearch
in the field of mathematicseducationis not a task that can be undertakenlight-
ly. The roots of such researchlie below the surfaceof a whole spectrumof cul-
tures.They are fed by the backgroundsof those who undertakethe researchand
the historicalprecedentsof the environmentsfrom which they come. It is imper-
ative that we, as a community,addressthe questions of what we consider legiti-
mate researchin the field of mathematicseducationand what we termacceptable
results of such research.If we do not do this for ourselves, we will continue to
be judged by the criteriaof other disciplines.
It is certainly inappropriatefor our researchto be evaluated solely from the
standpointof scientific proof. The appeal of the scientific paradigmlies both in
its appearanceof certaintyand in its common acceptability,which stem from a
long traditionof establishedpractice.The methodsdevised within this paradigm
have evolved over time and have been shapedand mathematicallydeveloped so
that generally acceptablecriteriafor evaluationof the resultsexist. We must not
18 Towardsa Definitionfor Research
be seduced by this history. Blind applicationof scientific methods will not nec-
essarily produceresearchresults of interest or value to the mathematicseduca-
tion community. Notions of representativity,replicability, and generalizability
are fundamentalto quantitativeresearchbut not necessarily to work in all areas
of mathematics-education.On the one hand, we cannot ignore the affective and
socially influentialdomains surroundingthe teachersand studentswe study. On
the other hand, we are not engaged solely in anthropologicalor sociological
study. Ourinterestslie in the realmof mathematicseducation,and we cannotdis-
regardthe influence and peculiarnatureof the subjectmatter,namely, the math-
ematics, on the teaching and learningthat concernus.
Teppo, in the introductionto this book, alludes to the numerousfields from
which mathematicseducationresearchhas in the past drawnits techniquesand
methods.It is rightthatwe shouldhave done so, but we need to be awarethatwe
are borrowingfrom anotherfield of concern and, if necessary, adaptand make
these methods more precisely our own. Diversity is essential as we seek the
emergenceof the discipline of mathematicseducation,andimaginationandinno-
vative approachesare needed as we attemptto explore the natureof the field
within which we work. Innovation,however, must not be at the expense of rigor;
a tension must be preservedbetween novelty and acceptability.If we are to have
externalcredibilityand if our researchis to be seen as of value to the largercom-
munity outside mathematicseducation, we need to begin to seriously consider
the closer definitionof researchacceptableto our own community.
As a preludeto this defining process, we need to articulatefor one anotherthe
ways in which we have come to adoptthe methodswe areindividuallyusing. We
need to clarify for the rest of our community the cultures from which we are
coming and to make explicit the perspectives from which we are viewing the
problems we tackle. We should not feel a need to define ourselves in terms
appropriateto some otherdiscipline, but we must be clear to ourselves what it is
that we are and what it is that we do as researchersin mathematicseducation.
Only then can we expect those outside the field to recognize the legitimacy of
our work. Honesty and openness are needed in our disclosureof how we choose
our methods so that self-critical appraisaltakes place alongside external scruti-
ny. We cannot,of course, be complaisantin our isolation, defining ourselves and
ignoringthe concerns and perspectivesof others.The externalcriticismsneed to
be addressed,particularlythe criticisms of our uses of qualitativemethods. For
instance, consider the case-study method. The issues of validity and reliability
cannotbe tossed aside as "irrelevantto case study"but must be examinedfor rel-
evance in the particularcircumstancesand to the particularquestionsthat we are
considering.Questions of concern to the academic community at large need to
be openly debatedby mathematicseducators-but from the perspectiveof their
own researchparadigms.
Any discussionconcerningitself with researchmethodsneeds first to examine
the questionsthatsuch researchis expected to answeror illuminate.Whatareour
questions in mathematicseducation?What are the issues we wish to examine?
Susan Pirie 19
More fundamentally,what do we consider to be researchin our discipline? We

do not yet have any coherentresponse to these questions I have posed.
At this point the readeris invited to activelyjoin in the debate.The list thatfol-
lows presentsvarious scenariosin which no attempthas been made to represent
the wealth of currentand past work in mathematicseducation.They are offered
as a provocativestartingplace for discussion. Where would you draw the divid-
ing lines between, say, researchand personalinterest?The question you need to
answer is not "What are you interested in knowing more about?";research
should encompass a wider perspectivethanpurely personaldevelopment,yet its
effects need not be generalizableto the whole community.The question is not
"Whatwould be useful to you in yourteaching?"I contendthatwritingtextbooks
and creating materials properly lie in the domain of curriculumdevelopment.
However valuable and revolutionarythese tools might be, their productiondoes
not constitute research,althoughthey may, of course, evolve as a consequence
of researchfindings. The question that you are asked to addresshere is "What
should the mathematicseducation community as a whole accept as legitimate
research?"As you attemptto grapplewith this problem,examine your personal
backgroundand cultureto trace the influences that they have on your decisions
because here lies the crux of the matter.Ourpersonalhistoriesareunique,yet we
must find and depict common groundfor the acceptablebasis of researchwithin
our discipline. Considerthe following scenarios:
* A teacherreads a variety of journal articles on teaching fractionsand, on the
basis of this reading,carefullyplans a sequence of lessons. From the teacher's
perspectivethe lessons go well and the pupils seem to have a good graspof the
concept of fractionsby the end of the set of lessons.
* A classroom teacher, unhappywith the lack of understandingof some of her
pupils, recordsherself teaching a particularlesson, transcribesit, and careful-
ly examines her interactionswith various children with a view to improving
her own communicationskills. As a resultof this examination,she changes her
behavior in class and the pupils appear,to her, to have a better grasp of the
concept being taught.
* A mathematics educator from outside the classroom takes a transcriptof
recordedclassroom interactionsand examines it against a backgroundknowl-
edge of existing, reportedresearchfindings in classroom communicationwith
a view to siting the teacher's behavior in the wider perspective of classroom
practice. The teacher has been chosen as the subject of study because the
investigator has previously noticed particularmathematicalpractices among
some of the pupils in this teacher's class.
* A large,internationalsampleof childrenis testedon a particularrangeof topics
with a view to rankingcountriesby the mathematicalabilityof theirchildren.
* The data gatheredabove are examined with a view to exploring the impact of
culturalenvironmentson mathematicsachievement.
20 Towardsa Definitionfor Research
* On the basis of existing findings relatingto errorsand difficulties that pupils

have with a particularmathematicaltopic, a piece of computer software is
written specifically to offer appropriateand precise remediation to pupils
strugglingwith this topic. The softwaredeveloped is used in a varietyof class-
rooms with a view to comparingthe effects on pupils with different mathe-
maticalbackgroundsand problems.
* The findings above relatingto the softwareare comparedwith the achievement
of similar pupils who have not been exposed to the computerwith a view to
determiningthe efficacy of the remediation.
* All these datarelatingto the softwareare takenas a base for theorizingon why
the novel approachdid (or did not) resultin increasedunderstandingof the tar-
geted topic.
* On the basis of broad, accumulated,personalexperience, a mathematicsedu-
catorputs forwarda theory aboutthe structureof the learningof novice teach-
ers with a view to influencing generalteacher-trainingprograms.
* ... and so on....
Some of the previous scenariosyou may have been able to categorizeinstant-

ly as within or outside your acceptableboundaryfor legitimatemathematicsedu-
cationresearch.As suggestedearlier,pause and reflect on why you could do this.
Can you begin to define your boundarieswith a degree of precision? It is pre-
cisely this defining process that we, as practitionerswithin the discipline of
mathematics education, have to undertake. Answers will not be produced
overnight,but without active debate they will also not evolve over time.
There is yet a furtherquestion to consider in our endeavor to define mathe-
matics educationresearch.Can the resultsof the inquiry,in fact, be used to deter-
mine the acceptabilityof the work as "research"?If so, then methodology and
methods will play a very big partin the debate.My contentionis that acceptable
researchlies somewhereon a continuumfrom "lying on one's back in the grass
gazing at the sky while thinkingaboutthe generalnotion of arithmetic"to "test-
ing all 8-year-oldchildrenin the world on theirability to computeaccuratelythe
answers to all possible additionproblemsinvolving two numberswith two dig-
its." The acceptableintervalbetween these extremes will be governed, as stated
earlierin part,by the purposeof the inquiry,but not entirely. If as a result of my
sky gazing, I conclude that arithmeticis all about understandingnumbers,am I
doing research?What if insteadI producea theoreticalstructurefor the learning
of numbertheory that will revolutionizehow arithmeticis taught?Have I been
doing research?If as a result of my two-digit additiontest, I am able to confirm
thatthereis a wide rangeof abilityamong 8-year-oldsacrossthe world, can I jus-
tify my work as research?But suppose thatI also notice the unpredictedfact that
a very high proportionof children in only one specific nation have very low
accuracyscores on all computationsinvolving the number7. Does thisjustify my
actions as research?No single person can dictate the answers to these questions
Susan Pirie 21
for the whole community, but mathematicseducators need to participatein a

continuing,vigorous discussion of what constitutesresearch.
The choice of researchmethodsis a very personaldecision, althoughit will be
on this choice that the acceptabilityof the results will largely depend.Unlike the
questions posed earlier in this chapter for which there are as yet no definitive
answers,those that follow must be, and can only be, answeredby the individual.
Responses are needed that apply to specific researchquestions. There are not,
and can never be, prescriptiveanswers applicableto all mathematicseducation
research. In this area, the researcher is personally responsible for providing
appropriateanswers.
* When is it appropriateto attemptto remain objective and approachour data
throughstatisticalmethods?When is the researcher'spersonalknowledge of
the social setting relevant?
* When do we want to take a broad overview, and when is in-depthinterview-
ing more likely to yield the desiredinformation?When is a representativesam-
ple and when is a specific individual the more valuable unit of scrutiny?Do
we seek insight or do we want generalizability?
* When are we working with existing theory, and when are we hoping to build
new theory? When can we predefine the coding of our data for analysis, and
when do we prefer to allow a taxonomy to emerge from the data as they are
gathered?
* When and how do we take accountof the dependencyof our study on notions,
such as understandingand emotions, thatarenever directlyaccessible but must
come to us mediatedor filtered throughlanguageand behavior?
This book is not intendedto be a how-to manual,nor is it a polemic advocat-
ing the use of qualitative research. The research questions must always come
first. One does not set out to do qualitativeresearch;one sets out to advancethe
knowledge or understandingof some portionof the field of mathematicseduca-
tion and then searches for the most effective way of achieving this goal. The
remaining chapters in this book illuminate the ways that different researchers
have addressed some of the previous questions and provide insights into their
decision making.
22
Chapter 3
The Epistemological Basis of

Qualitative Research in
Mathematics Education:
A Postmodern Perspective
Paul Ernest
In the past decade or so, a new paradigm widely referred to as the qualitative
research paradigm has begun to dominate research in mathematics education.
Although its roots go back a long way, in mathematics education this paradigm
emerged in Piagetian-style research based on clinical interview methods. There
were anticipations that fit with the qualitative research paradigm, such as
William Brownell's studies of understanding and problem solving in the 1930s
and 1940s (see Noddings, 1994). However, only lately has research of this type
become widely accepted and commonplace in the leading journals in the field.
The emergence and growth of the qualitative research paradigm in mathemat-
ics education represent an important shift in style in a still young field of inquiry.
This development raises a host of issues about the nature of significant research
questions, research methods, styles of research reporting, and the possible impact
of such research on the teaching and learning of mathematics. Many of these
issues are difficult to address in the abstract and are best demonstrated through
concrete exemplars, as consistency with the epistemology of the qualitative
research paradigm also requires. This monograph presents many such examples.
The qualitative research paradigm has a deep philosophical significance. The
aim of this chapter is to address some of the general epistemological and foun-
dational issues and implications concerning qualitative research in mathematics
education and to relate the paradigm to broader developments in 20th-century
thought-especially postmodernist thought. In this chapter, I sketch the philo-
sophical background of the qualitative research paradigm and relate it to current
developments in the philosophy of mathematics. I survey the epistemological
foundations of this paradigm and its relationship with constructivist and social
theories of learning and their implications for mathematics education research.
After elucidating some of the theoretical assumptions and characteristics of the
qualitative paradigm, I contrast it with two other educational research perspec-
tives: the scientific paradigm and the critical theoretic paradigm.
A central aim of the chapter is to distinguish research methodology from meth-
ods. The qualitative research paradigm provides a methodology, that is, a general
Paul Ernest 23
theoreticalperspectiveon knowledge and research,that allows specific methods,

instruments,and techniquesto be selected for particularprojects.Maintainingthis
distinctionis vital to ensurethatqualitativeresearchis conductedthoughtfullyand
to preventit frombecoming formulaicand recipe driven.
PHILOSOPHICALBACKGROUND
From Modernismto Postmodernism

The beginnings of modernismare often attributedto Descartes's (1637/1955)
seminal contributionsto epistemology. He imagined a logical master plan that
could provideindubitablefoundationsfor all knowledge. This plan was modeled
on the geometry of Euclid, "the only [true] science ... bestow[ed] on
[hu]mankind"(Hobbes, 1651/1962, p. 77). Descartes's epistemology came to be
known as rationalism,and his powerful rationalvision was to dominate many
areas of knowledge, including philosophy, physics, and mathematics. The
approachof rationalism(and modernism) is to recast knowledge by using the
logical structureof axiomatic geometry as a model to demonstratethe indu-
bitability of the knowledge claims involved. This model is central to what
Lyotard(1984) terms the metanarrativeof modernism:the style of narrativeof
the overarchingphilosophicaldiscourseused to legitimate scientific knowledge.
Early 20th-centuryphilosophy of mathematicsalso reapplied this model to
mathematicsitself in the quest for absolutely certainfoundationsfor mathemat-
ical knowledge. However, the failureof the prescriptiveprogramsof the logicist,
intuitianist,and formalist schools in the philosophy of mathematicsto achieve
this certaintyis well documented(Ernest, 1991, 1997; Kline, 1980; Tiles, 1991).
The applications of the scientific and rationalist legacy of modernism in the
physical sciences and social and managementsciences have continuedunabated
despite technical setbackslike these. Science seeks to build unified abstractthe-
ories to explain the phenomenaof the world and to predictfutureregularitiesand
outcomes. The absolute space-time frameworkof Newtonian mechanics may
have given way to relativisticand quantummodels of the universe,but the laws
of science continueto be used to make powerful and widespreadpredictions.For
a while, the modernistperspectivedominatedother subjects, such as positivistic
philosophy and behaviorist psychology. It also had an impact on research in
mathematicseducationboth throughthese human sciences and directly through
the influence of mathematicsitself. The rationalplanningmodel of management,
itself a developmentof modernism,continuesto be appliedin the governmentof
educationdespite radicalcritiques(Stenhouse, 1975).
However, there have been significant developments in the culturaland intel-
lectual spheres that reject the assumptionsof modernism.Modernism's legacy,
the rigid barriersbetween adjacentfields of inquiry,is dissolving, and an increas-
ing numberof interdisciplinaryfields of study are developing. The influence of
logical positivism, logical empiricism, and linguistic analysis on anglophone
24 TheEpistemologicalBasis of QualitativeResearch
philosophy is beginning to wane. Instead, contemporaryphilosophy is now

makingcontact with those Europeantraditionsthat regardknowledge as histori-
cally and culturally situated and not as objective and existing solely in some
Platonicor otherdisembodiedrealmof pureideas. Partof this move is the recog-
nition that knowledge, money, and power do not circulatein differentand non-
intersectingrealms, thus challenging the Cartesiandualism of mind and body.
Instead, there is a growing acceptance among some, at least, that knowledge,
money, and power are all materially embodied and that all form an intercon-
nected partof the humanworld we inhabit.This insight is partof the emerging
perspective of postmodernismin philosophy and cultural theory, according to
which a numberof knowledge fields are being simultaneouslyreconceptualized
as distributedand concretelybased practices.These fields include the following:
* Philosophicalpostmodernism(e.g., Lyotard, 1984; Rorty, 1979): Grandlogi-
cal ("top-down") metanarratives,like Descartes's rationalism, are being
replaced by locally distributed("bottom-up")knowledge practices in which
knowledge is produced, shared, and warranted in "local," institutionally
groundedlinguistic practices.The academic communityof mathematicsedu-
cators may be regardedas one or more "local"communitiesof this type.
* Poststructuralism:Self and knowledge are being reconceptualizedas distrib-
uted over a number of different discursive practices (Foucault, 1972;
Henriques,Urwin, Venn, & Walkerdine,1984).
* Wittgensteinianepistemology:Meaningand knowledge areregardedas situat-
ed in habituallyconductedor changing "languagegames"embeddedin social
"formsof life" (Wittengenstein1953).
* Tacit and personalknowledge (e.g., Ryle, 1949; Polanyi, 1958): These forms
play an essential role in humanand scientific knowing but are not expressible
in explicit propositionalform, contraryto the ideals of logical, rational,scien-
tific knowledge.
* Cognitive science and the philosophy of mind (e.g., Gardner, 1983, 1987;
Minsky, 1986): Mind is understoodto be modular, with local knowledges,
skills, and agencies in place of a single controllingintelligence.
* Social psychology: New emphaseson situatedlearningprioritizecontext over
individual minds (Gergen, 1985; HarrE, 1979; Lave & Wenger, 1991) and
stressthe formativeimportof discourse(HarrE& Gillett, 1994; Shotter,1993).
* Sociology and philosophy of science: There is a new emphasison the histori-
cal, laboratory,and rhetoricalpracticesof scientists instead of on overarching
theories of method (Feyerabend,1975; Kuhn, 1970; Simons, 1989; Woolgar,
1988).
* Philosophy of mathematics: This embodies a shift of emphasis onto the
methodologiesand practicesof mathematiciansaway from the logical theories
of mathematicalknowledge and truth(Kitcher, 1984; Lakatos, 1976).
* Social epistemology (Fuller, 1988; Toulmin, 1972), semiotics (Eco, 1984), and
Paul Ernest 25
feminist epistemology (Harding,1991): Parallel"bottom-up"developmentsin

epistemology have been taking place.
These examples illustratethe epistemological shift in which knowledge fields
are reconceptualizedas comprisingmulticenteredhumanpractices.Withinthese
decenteredpractices,knowing cannot be divorced from the concrete particulars
known. These range from exemplaryproblemsolutions and a knowledge of lab-
oratorypracticein science (Kuhn, 1970; Woolgar, 1988) througha knowledge of
particularlinguistic practicesand speech acts (Austin, 1962). This means thatthe
natureof knowledge is being reconceptualized,which is importantfor education
where the selection, recontextualization,and communicationof knowledge, as
well as the assessment of its acquisition,are centralactivities.
The Philosophy of Mathematics
Accompanyingthe emergence of these new of conceptions of knowledge and

humanknowing is a new traditionin the philosophyof mathematicsthathas been
gaining momentum.This movement has variously been termedpostmodernist
(Tiles, 1991), maverick(Kitcher& Aspray, 1988), andfallibilist (Ernest, 1991,
1997; Lakatos, 1976). This traditionis primarilynaturalisticand concerned to
describe mathematicsas an extant knowledge field, including the practices of
mathematicians,past and present.It is quasi-empiricistand fallibilist in its epis-
temology (Kitcher, 1984; Lakatos, 1976). A numberof additionalphilosophers
and mathematicianscan be identified as contributingto this tradition,including
Wittgenstein (1953, 1978), Putnam (1975), Wang (1974), Davis and Hersh
(1980, 1988), and Tymoczko (1986). Also, a growing numberof researchersare
drawing on other disciplines to account for mathematicsin terms of social and
culturalpractices.Their aim is to documentmathematicsas a social institutionin
the past and presentand as an element of all humancultures,both urbanand trib-
al. The outcome is an overlappingset of vistas that illustratethe various human
aspects of mathematicsand thattogetherchallenge the traditionalmodernistcon-
ception of mathematics as objective, superhuman, and value-free. These
researchersand their disciplinaryperspectivesinclude Bloor (1976) and Restivo
(1992) in sociology; Wilder (1981) and Livingston (1986) in culturalstudies and
ethnomethodology;Rotman (1993) in semiotics; Aspray and Kitcher (1988),
Joseph (1991), Kline (1980), and Gillies (1992) in the history of mathematics;
and Ascher (1991), D'Ambrosio (1985), Gerdes(1996), and Zaslavsky (1973) in
ethnomathematics.Thus the fallibilist, postmodernisttraditionin the philosophy
of mathematicsrepresentsthe convergence of several multidisciplinaryperspec-
tives (see Ernest, 1994a).
Withinthe philosophyof mathematicsandwithinotherfields thattheorizeabout
mathematicsis a move to reconceptualizeaccountsof mathematicsto accommo-
date greater plurality and diversity, including external, social dimensions of
mathematics-its history, applications,and uses. There is also a widely shared
commitmentto a multidisciplinaryaccount of mathematicsthat accommodates
ethnomathematics,mathematics-education studies,and feminist and multicultural

critiques.This commitmentis importantbecause if mathematicsis conceived as
inseparablefrom humancontextsandpractices,then social implicationsfor math-
ematics education follow, enabling notions of accessibility, equity, and social
accountabilityto be applied to the discipline of mathematics.The outcome is a
demystificationof mathematics,to the benefitof the disciplineandmathematicians
and also to students,teachers,and otherusers of mathematicsin society.
Postmodernism and the Qualitative Research Paradigm
Postmodernismis a portmanteauterm used to denote a varietyof perspectives

of different strengthsand persuasions.Most mathematicseducatorswould not
wish to subscribeto all variantsand formulationsandcould not because of incon-
sistencies between different versions. However, the sharedfeature of different
versions of philosophicalpostmoderism is the rejectionboth of foundationalism
(the quest for indubitablefoundationsfor knowledge) and of the associated log-
ical metanarrativesof certaintyfor mathematical,scientific, and other forms of
knowledge.
The philosophicalpostmodernismof Rorty (1979) and othersrejectsthe grand
design of modernism,which is based on one big idea of a logical order,built up
from clear and simple ideas and explicitly stated postulates. It rejects the
Cartesian epistemology with its overconfident logic-centered metanarrative.
Postmodernismis instead polycentric, pluralistic,and more connected to tradi-
tion. It values the concrete,the local, what is given in, and sharedthrough,local
practices.This postmodernistdecenteringof knowledge has a powerful affinity
with the qualitativeresearchparadigm,with its emphasis on the concrete, the
particular,the case study, and human-basedknowing. Thus postmodernismpro-
vides an epistemological foundationfor the qualitativeresearchparadigm.But
note that it is just one of many possible supportingnarratives.It is not another
unique metanarrativeinvoked to justify a knowledge field.
The essential function of a researchparadigmis to supportand facilitate the
generationof knowledge, and, of course, one of its componentsis epistemology.
From the epistemological point of view, two standpointscan be distinguished:
the absolutistand fallibilist perspectives(Ernest, 1991). The differencebetween
these standpointsis profound.An absolutistepistemologyviews "truth"as some-
thing that can be attained-that aspects of the world or thought can be under-
stood completely or at least known with certainty.Such a view is associatedwith
philosophical modernism and with some versions of the scientific paradigm.
Fallibilist approachesto research, which include the qualitativeresearchpara-
digm, fit with postmodernismand do not regardthe world as somethingthat can
be known with any certainty. Followers of this approachsee the relationship
between the knower and the known as problematicand accept that no certain
knowledge is attainableby humans(e.g., Guba & Lincoln, 1989). This humility
with regard to epistemology, knowledge, and the results of the methods
PaulErnest 27
employed in the researchprocess resonateswith developmentsin philosophy,the

humanities,and social sciences. It means thatneitherthe qualitativeresearchpar-
adigm nor its methodology can be employed mechanically in the quest for
knowledge. Every such approachis fraughtwith epistemologicaldifficulties and
standsin need of justification. Such an approachcan result in only a partialand
imperfect knowledge of events-the boundariesand claims of that knowledge
standin need of qualificationandjustification.Perhapsit is only because the sci-
entific research paradigm has been the dominant paradigm in educational
research,whereas positivism was the dominantparadigmin scientific research,
that it has not had to justify its methodologicalapproachesto the same extent as
has the qualitativeresearchparadigm.
EPISTEMOLOGY
EpistemologicalFoundations
The qualitativeresearch paradigmis the product of a number of significant
epistemological shifts for which postmodernismprovides a philosophical foun-
dationand support.These include shiftingemphasisto humanknowing from dis-
embodied knowledge and to knowledge of concrete practices and particulars
from universalgeneralizationsand laws. I now expand on these importantshifts
and delineatethe basis and sources of this paradigm.Historically,the qualitative
research paradigmhas epistemological foundations that are at least as old as
those of modernism.Where one locates the beginnings of a traditionis almost
arbitrary.I could startwith the pre-SocraticGreekphilosopherProtagorasof the
5th centuryBCE,who wrote, "Of all things the measureis Man, of the things that
are, thatthey are, of the things that are not, thatthey are not" (Freeman,1956, p.
125). This emphasizes the human features and limitations of knowing that are
centralto the qualitativeresearchparadigm.
A more recent startingpoint, not long afterDescartes's seminal contributionto
modernism,is in the work of Vico. Vico (1710/1858) argues that we can know
rationally only what we ourselves have made and that other forms of knowing,
such as knowledge of persons, are of a different,more humankind. For these lat-
ter forms of knowledge, "we must seek aid from our imaginationto explain them
and, like painters,form humanimages of them"(Vico, 1744/1961, p. 168). Thus,
Vico claims that there are two forms of knowing. The first emphasizesthe ratio-
nal. In the second, the emphasis is on the concrete, analogical, and particular
aspects of knowing that are typical of the qualitativeresearchparadigm.
The notionthattherearetwo fundamentallydifferentways of knowingwas fur-
ther and seminally elaboratedby Dilthey, one of the chief foundersof moder
hermeneutics(the study of interpretation,which originatedwith biblical exege-
sis). He distinguished the method of understandingfor the human sciences
(Verstehen)fromthatof the physical sciences (Erklaren).Verstehenis the method
of understandingnecessary to grasp the subjectiveconsciousness of participants
in some meaningfulactivity or context. Erklaren,in contrast,is the method of

seeking causal explanationsin studyingnaturalphenomena.
This distinction was developed in the hermeneutictradition as a means to
extend knowledge and understandingbeyond the limits of the scientific research
paradigmand to recognize the essential role of interpretation(Blaikie, 1993).
Verstehenhas been furtherelaboratedas an epistemological concept by Weber,
Sch,tz, and othersthis centuryin the social sciences: "Itsgoal is to find out what
the actor 'means' in his action, in contrastto the meaning which this action has
for the actor's partneror a neutralobserver"(Schutz, 1970, p. 9). This is the cen-
tral epistemologicalconcept of the qualitativeresearchparadigm.
Perhapsthe most importantcriticalcontributionto hermeneuticsand the philo-
sophical foundationsof social theory is that of Habermas(1971). Habermasis
the leading exponent of the CriticalTheory of the FrankfurtSchool, and his dis-
tinctionamongthreeknowledge-constitutinginterestsis often takenas providing
the basic distinctionbetween educationalresearchparadigms(Bassey, 1990-91;
Carr& Kemmis, 1986; Schubert,1986). He distinguishesthe scientific ("quan-
titative")paradigmand the qualitative("interpretative") paradigm;in additionto
these two manifestationsof traditionalways of knowing, he distinguishes the
critical theoretic research paradigm,which some theorists subsume under the
qualitativeparadigm.
Thus in the history of epistemology and the human sciences, two different
ways of knowing have long been distinguished.These correspondto the scien-
tific perspectiveand that of the qualitativeparadigm.
CONSTRUCTIVISTTHEORIESOF LEARNING
One of the centralcomponentsof the qualitativeresearchparadigmis the con-

structivistperspectiveon learning.This is taken to include the differentvariants
of constructivism, although radical and social constructivism will be distin-
guished later.It is primarilythe influence of JeanPiaget thathas establishedcon-
structivismas a centraltheoreticalperspectiveon learningin mathematicsedu-
cation. However, constructivismoffers more than an account of learning and
includes a fully fledged epistemology and a researchmethodology. (An episte-
mology includes a theory of public knowledge and its justification as well as a
theory of individual knowing. Note that some proponents of constructivism
regardthe position as postepistemologicalbecause they reject foundationalism
and epistemology as they are traditionallyformulated[Noddings, 1990].) The
constructivistperspective has had a profound impact on research on the psy-
chology of mathematicseducationin the past decade or two and also underpins
many recent developmentsin teaching.
Piaget's methodology centers on the use of the clinical interview. In this pro-
cedure, an individual subject is requiredto performcertain carefully designed
tasks in front of, and with promptingand probingfrom, an interviewer.A series
of sessions are likely to be needed for the researcherto develop and test his or
Paul Ernest 29
her model of the subject'sunderstandingconcerningeven the narrowestof math-

ematical topics. Piaget's clinical interview method is a seminal contributionto
qualitativeresearchmethodology in mathematicseducationbecause it supplies
in-depthinformationon which to construean individual'sthinkingand cognitive
processing. With its accompanyingmethodologicalassumptions,it is among the
most widely used approachesof today (Steffe, 1991b; Steffe & Gale, 1995).
Piaget's epistemology has its roots in a biological metaphor, according to
which the evolving organismmust adaptto its environmentin orderto survive.
Likewise, the developing human intelligence also undergoes a process of adap-
tation in order to fit with its circumstancesand remain viable. Indeed, Piaget
claims that the human intelligence is orderingthe very world it experiences in
organizing its own cognitive structures."L'intelligence organise le monde en
s'organisantelle-meme" (Piaget, 1937, cited in von Glasersfeld, 1989, p. 162).
Ernst von Glasersfeld (1989, 1995) and colleagues have extended Piaget's
epistemology significantly in developing radical constructivismbased on two
principles:(a) knowledge is not passively received but is actively built up by the
cognizing subject, and (b) the function of cognition is adaptive and serves the
organizationof the experientialworld, not the discovery of ontological reality.
Accordingto the first principle,knowledge is not transferreddirectlyfrom the
environmentor otherpersonsinto the mind of the knoweror learner.Instead,any
new knowledge has to be actively constructedfrom pre-existingmental objects
within the mind of the learnerin response to stimuli or triggersin the knower's
experiential(or psychic) world to satisfy the needs and wants of the learnerher-
self or himself. As Kieren and Pirie (1991) argue, knowledge constructionis
based on a recursive restructuringof personal knowledge in the light of the
knower's construings of mathematicalexperiences. Consequently, individual
learners construct unique and idiosyncratic personal knowledge even when
exposed to identical stimuli. As Kilpatrick(1987) and others have made clear,
the acceptanceof this principleor variantsof it is very widespreadamong math-
ematics educators,psychologists, and cognitive scientists.
The second principle states that all knowledge is constructedand can reveal
nothing certainaboutthe world nor any other domain.This includes mathemati-
cal knowledge and parallels developments in fallibilist philosophies of mathe-
matics. This assumptionis much more radicalbecause it amountsto a rejection
of scientific realism. It implies not only that all our constructionswill fall short
in attemptingto describe aspects of externalreality, whetherthis be the physical
world or learners'understandingsof it, but thatthis realityis essentiallyunknow-
able. As Kilpatrick(1987) points out, this is an unpalatableconsequenceto many
researcherswho believe both that learnersconstructtheirown meaningsand that
we inhabita knowable externalreality, thus accepting the first but rejectingthe
second principleof radicalconstructivism.
Constructivismhas introducedan importantsense of awarenessof epistemo-
logical limitationsinto researchin mathematicseducation.As the postmodernist
philosopher Rorty (1979) puts it, human knowledge can never mirrornature.
Von Glasersfeld's (1989, 1995) radical constructivistformulation is that our

knowing can at best pragmaticallyfit the world and can never epistemologically
match or mirrorit.
Constructivismhas had a profoundimpact on researchin mathematicseduca-
tion and probablyconstitutesthe majorimpetus for the recent shift towardmore
qualitativeresearch.However, the constructivistepistemology requiresthat the
methodology be used with humility and caution. Although we may tentatively
come to know the knowledge of othersby interpretingtheirlanguageand actions
through our own conceptual constructs, we must acknowledge that the others
have realities that are independentof ours. Indeed,these realities of others along
with our own realities are what we strive to understandin qualitativeresearch,
but we may never take these realities as fixed (Steffe & Gale, 1995).
The researcheras constructorof knowledge has to be included in the discus-
sion of research. The researchercannot be viewed as external to the object
known in this type of researchin mathematicseducation. "When we speak of
cognition, education,problemsolving, mathematicsor learningand teaching,we
must take care to recognize the role of the observerin the descriptionand analy-
sis of the problem"(Confrey, 1995a, p. 196).
Contributions From Social Perspectives
A number of other traditions and developments have helped introduce the

qualitative research paradigmto social science research and, subsequently,to
researchin mathematicseducation.In particular,these include theoreticaldevel-
opments in sociology, symbolic interactionism,ethnomethodology,and social
theory (see, e.g., Garfinkel, 1967; Goffman,1971;Mead, 1934; Schutz, 1972).
These perspectives are centrallyconcerned with the social constructionof per-
sons, interpersonalrelationships,and the types of interpersonalnegotiationthat
underpineveryday roles and functionings, such as those of the teacher in the
classroom. Berger and Luckmann(1966), building on these theoreticalperspec-
tives, elaboratedthe theory that our knowledge and perceptions of reality are
socially constructedand thatwe are socialized in our upbringingto shareaspects
of the conventionalview. Applicationsof these theories have a direct impact on
researchin mathematicseducationincludingresearchwith a constructivistflavor
(e.g., Bauersfeld, 1994) and without one (e.g., Bishop, 1985, 1988; Eisenhart,
1988). This impact is both theoretical,in terms of the underpinningepistemolo-
gy and overall framework,and methodological.
Researchin mathematicseducationis also drawingon othersocially orientated
traditionswithin psychology (Ernest, 1994b). Some of these use the theories of
Vygotsky (1978) and othermorerecentactivityof socioculturaltheorists,such as
Lave and Wenger (1991). For example, Bartolini-Bussi(1994) drawson activity
theory, whereas Saxe (1991) combines anthropologicaland socioculturalper-
spectives. There are also influences from more radical social theoreticdevelop-
psychologyof Henriqueset al. (1984), founded
ments,such as the poststructuralist
Paul Ernest 31
on the work of Foucault (1972) and applied to mathematics education by

Walkerdine(1988, 1989) and others.
One outcomeof the impactof sociallyorientatedperspectiveshas been the emer-
gence of various forms of social constructivismin mathematicseducation.This
includes the anthropologically derived position of Bishop (1985, 1988),
Bauersfeld's(1992) interactionistversion of constructivism,and Ernest's (1991,
1997) coordinationof the social growthof collectivemathematicalknowledgewith
the individual'sconstructionof personalknowledge.What these social construc-
tivist ideas shareis the notionthatthe social domainaffectsthe developingindivid-
ual in some crucialformativeway andthatthe individualconstructsor appropriates
her or his meaningsin responseto her or his experiencesin social contexts.
Some recent researchin mathematicseducation that incorporatesa construc-
tivist perspectivealso commonly attemptsto coordinatea social dimensionwith
the constructivistperspectiveon learning(Cobb, 1989; Richards,1991; Steffe &
Tzur, 1994). Currentlya controversyexists among supportersof various forms
of constructivism and sociocultural learning perspectives (Ernest, 1994b;
Lerman, 1996; Steffe & Gale, 1995). Cobb (1994) reviews contributionsto
socioculturaland constructivistperspectives and suggests that the main differ-
ences between them are over the location of mind, whetherin the head or in the
individual-in-social-action,and the concomitantview of mathematicallearning,
whetherconstitutedby active cognitive reorganizationor by enculturationinto a
communityof practice.The associatedpublisheddialogue elucidatessome of the
key termsand ideas involved but also reveals unresolvedtensions between com-
peting perspectives.(The dialogue is publishedin EducationalResearcher 1994,
23(7), pp. 4-23 and 1995, 24(7), pp. 23-28.)
Implicationsfor Mathematics-EducationResearch
The emergence of constructivismin research in mathematicseducation has
foregroundeda new set of researchemphases that are centralto the qualitative
researchparadigm.It is importantat this stage to recognize some of them. These
include attachingimportanceto-
* attendingto the previous constructionsthat learnersbring with them;
* attendingto the social contexts of learning;
* questioning the status of knowledge, including mathematicalknowledge and
logic, and the learner'ssubjectiveknowledge;
* proceeding cautiously with regardto methodological approaches,since there
is no "royalroad"to knowledge or "truth";
* attendingto the beliefs and conceptions of knowledge of the learner,teacher,
and researcher,as well as theircognitions, goals, metacognitions,and strategic
self-regulativeactivity;
* attending to language, discussion, collaboration, negotiation, and shared
meanings in the personalconstructionof knowledge.
These emphasescombine to indicatethat constructivistresearchin mathemat-

ics education needs to consider the learner as a whole person-the complex
social context of the learner,teacher, and researcher-and the constitutive and
self-implicatedrole of the researcherin research,whateverthe focus. These are
all importantfeatures of the qualitativeresearchparadigmin action in mathe-
matics education.
However, a note of cautionshouldbe soundedin attributingthese emphasesto
constructivism.If constructivismhad never emerged in psychology or mathe-
matics educationresearch,it is likely that all these emphases and the qualitative
researchparadigmin educationwould still have emergedfrom anthropologyand
sociology. Althoughconstructivismis importantin researchin mathematicsedu-
cation, especially for those with a backgroundin psychology, it is by no means
the sole source of the insights we gain from it. Most of the same insights are
equally available to those drawing on anthropology, sociology, and eth-
nomethodology.We must thereforerejectany myth of origins thatpromotescon-
structivismas an essential partof the qualitativeresearchparadigm.
EDUCATIONALRESEARCHPARADIGMS
I have been using the termparadigm to describe the overall frameworkwith-

in which qualitativeresearchtakes place. This draws on Kuhn's (1970) philo-
sophical analysis of science as having normaland revolutionaryphases. During
normalphases, there is a single acceptedparadigmwithin the scientific commu-
nity (e.g., Newton's mechanicsor Darwin's theory of evolution). Duringa revo-
lutionaryphase, several paradigmscompete, and their supportersare usually so
immersedin their own paradigmthatthey find it difficult to relocate themselves
within anothereven when their own has been refuted.Kuhn's claim is stronger,
namely, that competing paradigms are incommensurable, that is, mutually
incomprehensible.But thereis controversyin the philosophyof science literature
over this claim (Lakatos& Musgrave, 1970).
With Kuhn's conception,researchis usually understoodto take place within a
recognized or unconsciouslyassumedoverall theoreticalresearchperspectiveor
paradigm.In education,and in the social sciences in general,are found multiple
researchparadigms,each with its own assumptionsabout knowledge and com-
ing to know (epistemology),aboutthe world and existence (ontology), and about
how knowledge is obtained (methodology). Following the work of Habermas
(1971), a numberof educationalresearchersdistinguishthree main educational
researchparadigms:the qualitative(or interpretative),the scientific, and the crit-
ical-theoreticresearchparadigm(Bassey, 1990-91; Schubert,1986). It shouldbe
mentioned that there is some controversyover whether Habermas'sdistinction
between the interpretative(i.e., qualitative)and critical theoreticresearchpara-
digms is as strongas he contends (this controversyis discussed later).
Habermas argues that underpinningevery knowledge-seeking enterprise is
a particulartype of interest or desire at work, even in the case of science. He
Paul Ernest 33
distinguishesthreetypes of interestthatunderliethe quest for knowledge:to pre-

dict and control the phenomena under study (the technical interest), to under-
stand and make sense of them (the practicalinterest),and to achieve social jus-
tice throughthis understanding(the emancipatoryinterest).These correspondto
the interestsunderlyingthree educationalresearchparadigms:the desire to pre-
dict and control educationalprocesses throughknowledge (the scientific para-
digm); the desire to understandeducational phenomena, including individual
sense making (the qualitativeparadigm);and the desire to change education-
and through it, society-for the better (the critical theoretic paradigm).
Correspondingto these interests are the intended outcomes of the three para-
digms, respectively: objective knowledge, scientific generalizations,and truths;
subjective understanding,personal truths, and illuminating studies of unique
individuals;social changes and improvedsocial institutionsand conditions.
Bassey (1990-91), Ernest (1994c), and Schubert(1986) offer a discussion of
these paradigms from an educational perspective. They have been discussed
specifically in the context of mathematicseducation by Dunne and Johnston
(1992), Ernest (1994d), and Galbraith (1991). An outline of the qualitative
researchparadigmis now given, followed by brief descriptionsof the two other
paradigms.
The QualitativeResearch Paradigm

The qualitativeresearchparadigmdevelopedfrom the methodologyof sociolo-
gy and social science research,includinganthropologyand ethnomethodology.It
is primarilyconcernedwith humanunderstanding,interpretation, intersubjectivity,
lived truth(i.e., truthin humanterms),and so on. It takes fromethnomethodology
a concernto recordphenomenain termsof participantunderstandings.It uses var-
ious ethnographic,case study, and largely qualitative methods and forms of
inquiry,and it attemptsto overcomethe weaknessesof subjectivitythroughtrian-
gulatingmultipleviewpoints.Much attentionhas been paid in the literatureto the
problem of how qualitativeresearchfindings can be validated(e.g., Lincoln &
Guba, 1985).
In mathematicseducation research,the qualitativeresearchparadigmcan be
seen in the work of many researchers.A seminal early use of the researchpara-
digm is that of Erlwanger(1973) in his celebratedcase study of a single child's
learning(Benny). In the two decades since, a wide varietyof qualitativeresearch
has been published that presents, for example, in-depth knowledge of student
learningof mathematicaltopics, problem-solvingproceduresand strategies,and
teachers' beliefs.
One of the special features of the qualitativeresearchparadigmis its use of
the case study. Traditionally,scientific inquiryhas been concernedwith repeat-
able (replicable) circumstancesthat can be described by general laws. All the
particularsof the world are unique, but sharedfeaturesand resemblancesallow
generalizationsto be made, although always with a degree of uncertaintyand
unreliability(Popper, 1959). Once general laws have been derived, the scientif-
ic researchparadigmadoptsa top-downperspective,using the generalto deduce
predictionsabout particularinstances or observations.
The qualitativeresearchparadigmworks in an opposite directionand explores
the unique features and circumstancessurroundinga particularcase. However,
the aim is not to celebratethe uniquenessand oddity of a case. It is to explore the
richness of a particularthat may serve as an exemplarof somethingmore gener-
al. Kuhn (1970) has arguedthat even in the physical sciences, much use is made
of particular,exemplaryproblem solutions that serve as general models of rea-
soning and problemsolving.
Researchin the qualitativeparadigmbuilds up a rich descriptionof the case
understudy. Geertz(1973) calls it a thick description.Since a case typically con-
cerns human beings and their interrelationshipsand contexts, this description
allow a readerto understandthe case throughidentification,empathy,or a sense
of entry into the lived reality. Thus the kind of truthinvolved can be regardedas
akin to that of the novelist: the truthderived from identificationwith, and living
through, a story with the richness and complex interrelationshipsof social,
humanlife.
However, a case is meant to be illustrativeand generative. The particularis
intendedto illustratethe general-not with the precision of the exact sciences,
but suggestively as an illustrationof a more general and complex truth.The aim
is, as Blake wrote in his Auguries of Innocence, "to see a world in a grain of
sand"-to illuminate the general through the particular.Thus research in the
qualitativeparadigmadoptsa bottom-upperspective,using a particularand con-
crete instanceto suggest, evoke, and illustrate,if not describe,the generalcase.
Because of its renunciationof certainty,the issue of reflexivity arises for the
qualitativeresearchparadigm.The paradigmincorporatesan epistemology that
rejects the disembodied viewpoint of positivism that takes for granted the
assumption that it gazes on a fully knowable and separate objective reality.
Instead,in the qualitativeresearchparadigm,the researcheruses herself or him-
self (and her or his conceptualframework)as a researchinstrumentand should
incorporatereflections on the implications of using this "instrument,"with its
limitations,in any accountof the research.
The qualitativeresearchparadigmis referredto undera wide varietyof names,
includinginterpretative(andinterpretive),naturalistic,and alternativeparadigms
research.Some researcherspreferto avoid the name "qualitativeresearchpara-
digm"because althoughit is in widespreaduse, there is a risk of confusion with
qualitativeresearchmethods. In fact, the qualitativeresearchparadigmcan use
quantitativeas well as qualitativemethodsand data,just as the scientific research
paradigmin educationcan also use qualitativemethods as well as quantitative.
Quantitativedata and methods can be used within the qualitativeresearchpara-
digm, as and when appropriate(paradoxicalas this might seem), because of the
importantdifference between methodand methodologyin educationalresearch.
Methods are particulardata-gatheringor analysis techniques. For example,
PaulErnest 35
mathematics achievement tests or Likert-type attitude questionnairesprovide

datathatare typically analyzedby statisticalor quantitativemethods.In contrast,
the transcriptsof "thinkaloud" protocols during problem solving or videos of
classrooms or videos of personal interviews provide data that are typically ana-
lyzed by qualitativeresearchmethods. Educational-research methodsare specif-
ic and concrete approaches.In contrast,educational-researchmethodologyis a
theory of methods-the underlyingtheoreticalframeworkand the set of episte-
mological (and ontological) assumptionsthat determine a way of viewing the
world and, hence, that underpinthe choice of researchmethods. In this broad
sense, an educational-researchmethodologywith all its assumptionscorresponds
to an educational-researchparadigm.
Despite the theoreticalpossibility of other pairings,the qualitativeand scien-
tific research paradigmsdo generally tend to use qualitative and quantitative
methods, respectively.Typically, duringresearchin the qualitativeresearchpar-
adigm, the categories of analysis are generated,at least in part,duringthe analy-
sis of qualitativedata, whereas prechosen categories are applied in the analysis
of data in the scientific research paradigm(Strauss, 1987). This is one of the
majoroperationaldifferencesbetween the researchmethodologies.
One reason for qualitativeresearchmethods to be employed within the scien-
tific researchparadigmarises when researchersin mathematicseducation find
themselves in transition between the paradigms. Mathematicaltraining often
implantsthe assumptionsof the scientific researchparadigm.Thus, the first use
of qualitativemethods can be against the hidden backdropof some or all of the
assumptionsof the scientific researchparadigm.
The ScientificResearch Paradigm

The scientific researchparadigmis also called the positivistic, neo-positivist,
or experimentalresearchparadigmin education.It originateswith the scientific
method as employed in the physical sciences, in experimentalpsychology, and
so on. It is concerned with objectivity, prediction,replicability,and the discov-
ery of scientific generalizationsdescribing the class of phenomenain question.
The forms of inquiry used include survey, comparative experimental, quasi-
experimental,and so on. There is often an emphasis on quantitativedata, but
qualitativedatacan also be used, as and when appropriate.Whatis centralto the
scientific research paradigmis the search for generalizationspredictingfuture
educationaloutcomes. Thus process-productresearchin mathematicsteachingis
typical work in the scientific paradigm.It examines correlationsbetween teach-
ing practices and student learning outcomes and seeks to empirically validate
relationshipsbetween them (e.g., Good, Grouws, & Ebmeier, 1983). Similarly,
the constructionof empirical learning hierarchiesby the CSMS project (Hart,
1981), the comparisonof instructionalprograms(e.g., Charles& Lester, 1984),
the quantificationof teaching behaviors (e.g., Cooney & Henderson, 1972), and
the evaluation of aptitude-treatmentinteractions (e.g., McLeod, Carpenter,
McCorack, & Skvarcius, 1978) fall within the scientific research paradigm.
Althoughit may be controversialto make this claim, in my view Piaget also used
qualitativeresearchmethods (clinical interviews) to advance his theory of cog-
nitive stages. This latter use, with its age-stage measures and predictions,lies
squarelywithin the scientific researchparadigm.
The scientific researchparadigmhas many of the advantagesassociated with
the physical andbiological sciences. When successful, it resultsin replicableand
objective generalizations.These have the strengthsof being rigorouslyscientifi-
cally tested. The paradigmalso has the strengthsof clarity,precision,rigor, stan-
dardization,and generalizability.It is also, in theory, universally applicable.
However, the weakness of this paradigmis that it involves simplifying the phe-
nomena described, and its application is too often based on unquestioned
assumptions.All personsand humansituationsand contexts are unique and indi-
vidual, but the scientific researchparadigmtreatswhole classes of individualsor
events as identical, or at least indistinguishable,except in terms of a range of
selected variables. Thus, this approachcan often be insensitive to contextual
variationsand individualdifferences,althoughin theory it can always be refined
to accommodateomittedaspects. Some of the epistemologicalassumptionsasso-
ciated with this paradigmare questionable,too. For often it is associatedwith an
absolutistepistemologyand a Newtonian-scientificontology. However,these are
defensible perspectives,even if they are sometimes uncongenialto those work-
ing in the qualitativeresearchparadigm.
The Critical TheoreticResearch Paradigm

The criticaltheoreticparadigmhas developed out of the CriticalTheoryof the
FrankfurtSchool, especially the work of JurgenHabermas(1971). The central
featureof this position is the desire not just to understandor to find out, but to
engage in social critique and to promote social and institutional change to
improve or reformaspects of social life. In education,this often involves work-
ing on social justice issues, such as redressinggender, class, or racial inequali-
ties. To this end, it often involves participantengagementand validation.One of
the best known discussions of this approachapplied to educationalresearchis
that of Carrand Kemmis (1986). As in this reference,the criticaltheoreticpara-
digm is often closely associatedwith actionresearch,which is popularamongthe
"teacher-as-researcher" movement,with teachersworkingto change theirteach-
ing or school situations to improve classroom learning. In my view, action
research,however, too often balks at addressingoppressionin society to fit com-
fortably under the critical theoretic paradigm.Such projects as Paolo Freire's
(1972) work emancipating Brazilian peasants through literacy, although not
explicitly critical theoretic, serves as an excellent example of this type of
research.Likewise, in mathematicseducation, the paradigmis reflected in the
work of Gerdes (1985) in Mozambique and such researchersas Mellin-Olsen
(1987) and Skovsmose (1985, 1994) in Scandinavia.
Paul Ernest 37
The critical theoreticresearchparadigmis explicitly concernedwith improv-

ing some context, situation,or institution.Most othereducationalresearchis also
concernedwith improvingschooling in some way or other,but this improvement
is usually more of an indirectconsequence of the inquiry.The paradigmhas the
advantageof specifying this goal explicitly and not being concernedwith trying
to leave undisturbedthe situation being investigated. The disadvantageof this
paradigmis that hidden institutionalsources of resistance to change, such as
teacher and pupil ideologies, institutionalstructures,and so on, may often pre-
vent progress.If there is no progressand there is little of the knowledge that the
other two educationalresearchparadigmsseek to establish, then the danger is
that there may be no worthwhileoutcome for the energy and time invested.
A philosophicalcriticism of the critical theoreticresearchparadigmis that its
intendedoutcome (emancipatorysocial change)is of a differentcategoryfromthe
intendedknowledge outcomes of the scientific and qualitativeresearchparadigm
(Blaikie, 1993). The rationalbasis for emancipatoryknowledge qua knowledge
has not been made explicit (Carr& Kemmis, 1986). However,Habermas's(1981)
projectin recent years has been to meet this challengeby developing a theory of
communicativeaction that unifies knowledge with emancipatoryaction.
Comparingthe Research Paradigms

Given that multiple educationalresearchparadigmsexist, it is worth compar-
ing them briefly. Table 3.1 shows a simplified summaryand comparisonof the
three major researchparadigmsby using some of the factors previously men-
tioned (based on Bassey, 1990-91; Ernest, 1994c; and Schubert,1986).
Table 3.1
SimplifiedSummaryand Comparisonof the ThreeMain Paradigms
Component Paradigm
Scientific Qualitative Criticaltheoretic
Ontology Scientific realism (objects Subjectivereality Persons in society
in physical space) (personalmeanings) and social institutions
Epistemology Absolutist, objective Personal,constructed Socially constructed
knowledge or socially constructed knowledge
knowledge
Methodology Mainly quantitative Mainly qualitativecase Mainly critical action
and experimental, studies of particular researchon social
involving many subjects individualsand contexts institutions
and contexts
Intendedoutcome Applicable knowledge Illuminativesubjective Interventionfor social
and generalizations understandings reform,social justice
Interest To comprehendand im- To understandand make Social justice,
prove (throughprediction sense of the world emancipation
and control) the world
One of the major epistemological differences among the paradigms con-

cerns what is problematized. The scientific research paradigm locates uncer-
tainty exclusively in the immediate object of inquiry, such as the teaching and
learning of mathematics in a particular classroom. This paradigm does not
require any reflexivity concerning the researcher's constitutive role in knowl-
edge and meaning making. There are of course objectified requirements to
attempt to remove distortions introduced by the researcher in the process of
inquiry, such as the concern to establish the validity and reliability of the
research instrumentsused. In contrast, the other two paradigmsdo not regard
the world and its events as something that can be known with any certainty.
They problematize the relationship between the knower and the known and
adopt a position of humility with regardto epistemology, knowledge, and the
results of the methods employed in research. This means that neither of these
two research paradigms or methodologies should be employed mechanically
in the quest for knowledge but that every application stands in need of justi-
fication. A fallibilist epistemology requires the recognition of the limits of
knowledge claims at every level of educational research. A note of caution
should be added. Sometimes, critical theoretic researchpresupposes that it has
a privileged viewpoint delivering reliable knowledge about the social situa-
tion it seeks to change.
The threeparadigmsrepresentclustersor general styles of approachto educa-
tionalresearch,characterizedin termsof theirtypes of basic assumptions.Within
each paradigm(and the fit may be loose in parts), it is possible to have a wide
variety of approaches.Some of the disciplined approachesthat fit more or less
within the qualitative research paradigm are the phenomenological, eth-
nomethodological, psychoanalytic, and hermeneutic approaches. Perhaps in
eitherthe qualitativeor the criticaltheoreticparadigms(andperhapsoverlapping
with both) are the social constructivist,poststructuralist,andfeminist standpoints
(Harding,1987). Dunne and Johnston(1992) relateall threeeducationalresearch
paradigmsto gender issues in research in mathematicsand science education.
Ernest (1994a, 1994b) includes contributions representing many of these
approachesto researchin mathematicseducation.
It has been suggested by some scholarsthat the distinctionbetween the quali-
tative and the critical theoreticresearchparadigmsis not as clear-cutas the pre-
ceding account suggests. After all, it is largely based on Habermas'sdistinction
in defining a thirdresearchparadigm,namelythe criticaltheoreticone. Certainly
the possibility of overlap between the qualitativeand the critical theoreticpara-
digms should be countenanced,and indeed some examples of researchin math-
ematics education are hard to locate within just one of the paradigms (e.g.,
Walkerdine's[1988] poststructuralistapproach).Some other researcherssome-
times distinguishonly two majorparadigms,the scientific and the interpretative
(i.e., qualitative)researchparadigms(e.g., Lincoln & Guba, 1985), with the lat-
ter incorporatingthe critical theoretic paradigm.HarrEand Gillett (1994) also
contrast only two research paradigms in contemporary psychology, the
Newtonian (scientific) and the discursive (qualitative)paradigms,thus reducing
the distinctionto a dichotomy.
Paul Ernest 39
Finally, it should be acknowledgedthat issues of philosophy and epistemolo-

gy concerningeducationalresearchparadigmsare controversial.For ratherthan
acknowledgingthat multiple valid paradigmsand sets of assumptionsunderpin
research,each with differentstrengthsand aims, some researchershave preferred
to fight for their own paradigmas the sole valid one. Scientific researchpara-
digm supportershave arguedthatthey own the sole routeto objectivityand truth.
Supportersof the qualitativeresearchparadigmhave argued that the quest for
objectivity and truthis futile and that only they can offer valid understanding.
Critical theoretic researchparadigmsupportershave arguedthat the others are
victims of "false consciousness"and that only they can reveal the ideologically
induced distortionsin educationand society.
Thus there is no consensus about which educational research paradigm is
"true"or "correct."There are only proponentsof one or anotherparadigmand
those who arguethat all have some validity, as I do here. Fromthe point of view
of the fallibilist epistemology underpinningthe qualitativeresearchparadigm,
"correctness"is not possible, anyway. Instead,it is importantto be aware of the
strengthsand weaknesses of scientific, qualitative,and criticalapproachesand to
be able to question the epistemological assumptionsthat are made in each of
them. Gage (1989) has writtenof the paradigmwars waged among supportersof
the three paradigmsin the educationalresearchcommunityin the United States.
His recommendationis that educationalresearchparadigmsare tools that should
serve our practicalends in educationand that the best policy is to acknowledge
their multiplicitywhile judging them by their fruits.
40
Chapter 4
Observing Mathematical Problem

Solving through Task-Based
Interviews
GeraldA. Goldin
Over a period of 2 decades, mathematicseducationhas evolved to stress con-

ceptual understanding,higher-level problem-solvingprocesses, and children's
internalconstructionsof mathematicalmeaningsin place of, or in additionto, pro-
cedural and algorithmic learning (Davis, Maher, & Noddings, 1990; von
Glasersfeld, 1991). With this trend, the structuredclinical interview has found
greateracceptanceas a researchmethod.It lends itself well to the qualitativestudy
and descriptionof mathematicallearningand problemsolving withoutthe exclu-
sive relianceon countsof correctanswersassociatedwith pencil-and-papertests.
In general, such structuredinterviews are used in researchfor the twin pur-
poses of (a) observing the mathematicalbehavior of children or adults, usually
in an exploratoryproblem-solvingcontext, and (b) drawinginferences from the
observationsto allow somethingto be said about the problem solver's possible
meanings,knowledge structures,cognitive processes, affect, or changes in these
in the course of the interview.
For me, structuredinterviews are especially attractiveas a means of joining
researchwith educationalpractice.Reformsin school mathematicsin the United
Statesendeavor(among othergoals) to foster the discovery of patternsand ways
of reasoning about them and to develop skill in constructingoriginal, nonstan-
dard solution methods. Guided explorationsby children and small-groupprob-
lem solving are encouraged.These goals supplement(if they do not actuallysup-
plant) more "traditional"teacher-centered,direct instructionemphasizingmas-
tery of standardizedmathematicalrepresentations,rules, and procedures.In the
This chapter expands on talks presented at the 16th Annual Conference of the
MathematicsEducationResearchGroupof Australasia(MERGA-16,July 1993, Brisbane,
Australia), and at the 17th Annual Conference of the InternationalGroup for the
Psychology of Mathematics Education (PME-17, August 1993, Tsukuba, Japan). The
research described was partially supportedby a grant from the U.S. National Science
Foundation (NSF), "A Three-Year Longitudinal Study of Children's Development of
MathematicalKnowledge,"directedby RobertB. Davis and CarolynA. Maherat Rutgers
University. Opinions and conclusions expressed are those of the authorand do not neces-
sarily reflect the views of the NSF or the projectdirectors.
GeraldA. Goldin 41
reformedcontext it becomes increasingly importantto be able to describe and

assess the longitudinal mathematicaldevelopment of individual children. We
need to find ways of observing that permit valid inferences about the deeper
understandingsthat the new emphases try to develop (Lesh & Lamon, 1992).
Thus, task-basedinterviewshave importanceboth as researchinstrumentsand as
potentialresearch-basedtools for assessmentand evaluation.They offer the pos-
sibility of obtaininginformationfrom studentsthat bears directly on classroom
goals and can help answer researchquestions central to the educationalreform
process: What long-termconsequences are innovative teaching methods having
for children's mathematical development? What powerful problem-solving
processes (if any) are studentslearningin "reform"classrooms?What cognitive
representationalstructuresare they developing? Are all children developing
these, or only some? What are the affective consequences of reform? What
beliefs about mathematicsare childrenacquiring?
The main purposeof the chapteris to discuss some of the scientific underpin-
nings of task-basedinterviewmethodologyin the study of mathematicalproblem
solving. I touch on a set of issues having to do with the reproducibility,compa-
rability, and generalizabilityof researchfindings. The importanceof having an
explicit theoretical perspective when structuringan interview is discussed, as
well as the fact that choices made duringinterview design can result in foresee-
able consequences-for instance, obtaining some informationat the expense of
otherinformation.I try to be awarethroughoutthe chapterof constraintsandlim-
itations imposed by the social and psychological contexts of interviews as well
as of the interplayamong task variables,contextualfactors, observedbehaviors,
and cognitions inferredby the researcher.
The main points are illustratedwith reference to five structured,individual
interviews,designed aroundmathematicalproblem-solvingtasks for the purpos-
es of a longitudinalstudy. These provide concrete examples relatedto the cen-
tral questions. The views described here helped shape the development of the
scriptsfor these interviewsand were in theirturnconsiderablyinfluencedby that
process. What we learned in developing the interview scripts, carryingout the
interviews, and interpretingthe results influenced some principles of interview
design and constructionthat are suggested for considerationby the mathematics
educationresearchcommunity.
QUESTIONSRAISED BY TASK-BASED INTERVIEWRESEARCH

Whetherwe regardtask-basedinterviewsas researchinstrumentsor as assess-
ment tools, their use to observe and draw inferences from mathematicalbehav-
ior raises fundamentalquestions. It is my view that future research studies
involving clinical interviews would benefit greatly by giving explicit, advance
considerationto the following questions:
1. In what sense do the interviews permitgenuinely scientific investigations?
By this I mean to inquire about the implications of the task-based interview
42 ObservingMathematicalProblemSolving
methodologyemployed for (a) the examination,analysis, and communicationto

othersof the measurementprocess, (b) the replicabilityof results(fromone inter-
view to anotherwith the same subject,from one populationto anotherwith sim-
ilar characteristics,from the currentstudy to other studies, and so forth), (c) the
comparabilityof outcomes across studies that may employ different interview
instruments,and most important,(d) the eventualgeneralizabilityof thefindings
that are obtainedfrom the observationsmade.
2. Whatrole does theoryplay in structuringthe interviews?To what extent are
the observationsmade duringan interviewcontingenton the tacit or explicit theo-
retical assumptionsthat underlie the interview questions and procedures?How
does theoryguide the choice of questionsin the interview?How does it guide the
contingenciesthat are plannedfor? How does it allow for unplannedcontingen-
cies? How are we to draw inferencesabout cognition, affect, or both, from our
observations?Whatis the interplayamongtaskvariables(the characteristicsof the
problemson which task-basedinterviewsare based), observedbehaviors,and the
inferenceswe can draw?How shouldwe come to modify, substantiallyrevise, or
even discardourtheorieson the basis of the empiricaloutcomesof the interviews?
3. Whatconstraintsor limitationsare imposed by the social, cultural,and psy-
chological contexts of the interviews?How may the student'sexpectations,pre-
sumptions,apprehensions,and intentionsinteractwith mathematicalcognitions
and affect (and with task variables)to influence the interview outcomes?
The intent in raising these questions is to begin the discussion from a scientif-
ic perspective,offering illustrativeexamples from the currentstudy, and to pro-
pose some preliminaryand partialanswers-in the context of that study-that
may be more generally applicable.My goal is to frame some generalprinciples
of interview design and constructionthat may be appropriatefor the mathemat-
ics educationresearchcommunityto adopt. For example, it may be possible to
characterizethe trade-offsthat take place as questions are selected for incorpo-
ration in an interview script and, through explicit principles, to optimize the
informationgatheredin a task-basedinterview.
The ideas advancedhere have their origins in earlier studies of mathematical
problemsolving and in discussions aboutobservation,measurement,and assess-
ment (Bodner & Goldin, 1991a, 1991b; Cobb, 1986; DeBellis & Goldin, 1991;
Goldin, 1982, 1985, 1986, 1992a; Goldin & Landis, 1985, 1986; Goldin &
McClintock, 1980; Hart, 1986). But they are immediatelyinstigatedby a series
of task-basedinterviews that a group of us at RutgersUniversity createdin the
context of a longitudinalstudy of individualelementaryschool children'smath-
ematicaldevelopment.Five scriptswere written,and used from 1992 to 1994, as
the basis for a series of individualproblem-solvinginterviewswith children.The
next section describesthese briefly. I then returnto explore aspects of the scien-
tific natureof task-basedinterviews and addressthe role of theory and the role
of context. The chapterconcludes with commentsconcerningprinciplesof inter-
view design and construction.
GeraldA. Goldin 43
AN EXPLORATORYLONGITUDINALSTUDY
In a study whose outcomes are still being analyzed,the mathematicaldevelop-

ment of an initial group of 22 childrenwas observed for approximately3 years.
At the outset, in the 1991-92 school year, subjectswere 8 to 10 years old. They
were then in the thirdand fourthgradesin a cross-sectionof New Jersey's public
schools: two urbanschools (5 thirdgradersand 4 fourthgraders);one school in a
predominantlyblue-collar, "workingclass" community (7 fourth graders);and
one school in a suburban,"uppermiddle class" district(6 third graders).These
schools, and the children'steachers,were participatingin an intensive, construc-
tivist-orientedmathematicsteacherdevelopment-mathematicseducationreform
partnershipcalled MaPS (MathematicsProjects in Schools), sponsored by the
Rutgers Center for Mathematics, Science, and Computer Education and the
GraduateSchool of Educationand directedby CarolynA. Maherand RobertB.
Davis. In fact, one reason for initiating the longitudinalstudy-for which data
sources includedvideotapesof the children'sindividualproblemsolving, as well
as their small-groupmathematicalactivity inside and outside class-was to be
able to assess some of the project's outcomes in relationto individualchildren's
mathematicalunderstandingsas they grew over time.
One component of this study consisted of a series of task-based,individual
interviewswith each child over a partof the 3 years, conductedunderthe direc-
tion of the author (DeBellis & Goldin, 1993; Goldin, 1993; Goldin, DeBellis,
DeWindt-King,Passantino,& Zang, 1993). Five interviews were designed and
administeredbetween spring 1992 and Spring 1994, with the goals of observing
complex, individualmathematicalproblem-solvingbehaviorin detail and draw-
ing inferences from the observationsabout the children's thinkingand develop-
ment. Thus, this componentof the study was, from a scientific standpoint,main-
ly exploratoryand descriptive-subjects were not a randomsample from a larger
population,and no general hypotheses were being explicitly tested. Rather,we
hoped to describeindividualmathematicaldevelopmentin as much detail as pos-
sible, focusing not on standard,discreteskills or algorithmicproblemsolving, but
on the growth of complex, internalrepresentationalcapabilities. Tied to these
goals, the interview design included several steps: (a) planning in relation to
mathematicalcontent and structure,anticipatedobservations,and inferences-
discussed furtherin the next two sections; (b) creatingan interviewscript,and its
critiqueby the researchgroupin a graduateseminar;(c) pilot testing the scriptin
a differentschool, with childrennot partof the longitudinalstudy, and revising it
on the basis of the pilot test; and (d) training and rehearsingwith clinicians,
includingpracticesessions. Initiallywe hoped thathalf or more of the 22 children
would remain in the study for the full term; originally six interviews were
planned,but fundingconstraintslimitedus to five. As it turnedout, 19 of the orig-
inal group of childrenparticipatedin all five interviews.
The interviewsthemselves were designed to take less than one class period. In
every interview, alternativeembodimentsfor externalrepresentationwere given
to the child:paperandpencil, markers,cards,chips or othermanipulatives,paper

cutouts, a hand calculator,and so on, in accordancewith the task. The questions
within an interviewtendedto increasein difficulty, so thateach child began with
a level of comfort,but even mathematicallyadvancedchildrenencounteredsome
questions that were challenging before the interview ended. Free problem solv-
ing was encouragedwherever possible, with (specified) hints given or sugges-
tions made only afterthe child had the opportunityto respondspontaneously.All
responses were accepted by the clinician (with occasional exceptions, specified
in advance);thus "wrong"and "correct"answerswere treatedsimilarly.Follow-
up questions by the clinician were asked without an overt indicationof the cor-
rectnessof earlierresponses.Two video camerasoperatedsimultaneouslyduring
each interview-one focusing on the clinician and the child or the child's face,
the second focusing on the work the studentwas doing (workingwith paperand
pencil or handlingmanipulatives);in Interview3, a thirdcameraalso provideda
close-up of the child's facial expressions. An observer made notes during the
interview. Subsequentlythe videotapes were transcribed,viewed, and analyzed.
What follows is a capsule descriptionof each interview script. The full texts of
the interviews are availablefor researchpurposesfrom the RutgersCenter.
Task-Based Interview 1
The first interview script (55 pages, about 45 minutes) was written during
1991-92 and administeredin May and June 1992. The task, based on a high
school-level problem of the National Assessment of Educational Progress,
involves laying out for the child threecards,one at a time (see Figure4.1): "Here
is the first card, here is the second card, and here is the thirdcard."
Figure 4.1. The first threecards presentedin Task-BasedInterview 1.
The cards are drawnfrom a stack in an envelope, so the child may infer from
the context that there is a deck larger than the few cards shown and (possibly,
tacitly) may also infer thatthere is a patternpresent.After a brief pause to allow
a spontaneousresponse, the child is asked,
* "Whatdo you think would be on the next card?"
The materialsplaced aheadof time on the table areblankindex cards(the same
size as those with dots), felt-tipped markersof different colors, round red and
black chips (checkers),a pad of paper,and a pencil. The child can use any items.
GeraldA. Goldin 45
A series of exploratoryquestionsfollows, with contingenciesbased on the nature

of the responses and special emphasis on exploring the child's patternconstruc-
tion and use of externalrepresentations.After a complete, coherentresponse to
the first question has been elicited, the child is similarly asked the following
questions in slow succession:
* "Whatcard do you think would follow that one?"
* "Do you think this patternkeeps going?"
* "How would you figure out what the 10th card would look like?"
* "Here's a card [showing 17 dots in the chevron, or inverted V, pattern].Can
you make the card that comes before it?"
* "How many dots would be on the 50th card?"
The script is written so that for each main question, explorationproceeds in
four stages: (a) posing the question ("free"problemsolving) with sufficienttime
for the child to respond and only nondirectivefollow-up questions (e.g., "Can
you tell me more about that?); (b) heuristic suggestions if the response is not
spontaneous(e.g., "Can you show me by using some of these materials?");(c)
guideduse of heuristicsuggestions,againto the extent thatthe requesteddescrip-
tion or behaviordoes not occur spontaneously(e.g., "Do you see a patternin the
cards?");and (d) exploratory(metacognitive)questions (e.g., "Do you think you
could explain how you thought about the problem?").The clinician's goal is
always to elicit (a) a complete, coherent verbal reason for the child's response
and (b) a coherentexternalrepresentationconstructedby the child, before going
to the next question (for the question about the 50th card, an externalrepresen-
tation is not required). A complete, coherent reason means one based on a
describedor modeled pattern,but this patternis not requiredto be "canonical"
(i.e., to have the 4th card drawn with 7 dots in the chevron pattern) for the
response or externalrepresentationto be consideredcomplete and coherent.
This "nonroutine"task embodies an additive structure in an arithmetic
sequence represented through a geometric arrangementof dots. It provides
opportunitiesfor the child to detect numericalor visual patterns,or both; to use
visual, manipulative, and symbolic representations; and to demonstrate
reversibilityof thinking.
Task-BasedInterview2
The design for the second interview script (38 pages, up to about 55 minutes)
was completed in fall 1992. The scriptwas used in individualinterviews admin-
istered during winter 1993 with the same children (then in fourth and fifth
grades).As in the first interview, materials(a pad, a pencil, markers,and check-
ers) are placed ahead of time on the table in front of the child. First some pre-
liminaryquestions are asked with the intent of exploring the child's imaginative
and visual processes:The child describeswhethershe or he is right-or left-hand-
ed. Then the child is asked to imagine a pumpkin,to describe it, to manipulate
the image in various ways (including cutting the pumpkinin half), to spell the
wordpumpkin,to spell it backward,and to talk aboutthese activities. A series of
mathematicalquestions follows. For each, the follow-up includes (where appro-
priate):"Canyou help me understandthatbetter?"or "Arethere any otherways
to take (one half) (one third)?"or both questions.
* "Whenyou think of one half, what comes to mind?"
* "Whenyou think of one third,what comes to mind?"
* "Supposeyou had 12 apples. How would you take (one half) (one third)?"
* [Next cutouts are presented in succession: a square, a circle, and a 6-petal
flower. For each, the child is asked] "How would you take (one half) (one
third)?"
* [Circle cutouts are presentedto the child, first with (one half) (one third)(one
sixth) representedconventionally(as in a pie graph),then with the same frac-
tions representedunconventionally(the part representingthe fraction at the
center of the circle). In each case the child is asked] "Canthis card be under-
stood to represent(one half) (one third)?(Why?) (Why not?)"
* [A 3-by-4 arrayconsisting of 12 circles and 6-petal flowers is now presented.]
"How would you take (one half) (one third)?"
* The child is also asked to write and interpretthe usual notation for the frac-
tions one half and one third.
Next a solid wooden cube is shown. Some preliminaryquestions are asked
aboutits characteristics(numberof faces, edges, and comers). The child, guided
as necessarytowardunderstandingwhat these mean, is then asked to thinkabout
cuttingthe cube in various ways:
* "Now think about cutting this cube in half. What would the two halves look
like?"
* "Supposewe painted the cube red and then cut it the same way. How many
faces are paintedred, for the smallerpieces you told me about?"
Similar questions follow about cutting a series of up to five additionalcubes,
dependingon the time available.These cubes are markedwith lines at designat-
ed vertical or horizontalpositions, or both, which results in mutuallycongruent
pieces that are respectively 1/3, 1/4, 1/8, 1/9, and 1/27 the volume of the original
cube. The scriptcontainsnumeroussuggestedexploratoryquestionsand a series
of retrospectivequestions at two differentpoints. This interview thus provides
opportunitiesfor the childrento express a variety of conceptualunderstandings
relatedto one half and one third,in many differentembodimentsin both two- and
three-space dimensions. A multiplicative structureis embodied in cutting the
solid wooden cube across different dimensions, and special emphasis is placed
on exploring visualizationby the child.
GeraldA. Goldin 47
Task-Based Interview #3
The thirdinterview script(28 pages, about50 minutes)was completedin May

1993 and administeredduring May and June of that year. It begins with some
introductoryquestions designed to elicit some of the child's affect in relationto
mathematicalproblem solving: "Could you think back to the first time you
rememberdoing mathematics?What do you remember?""Whatis the earliest
you rememberdoing math in school?" "Did your (parents)(brothersor sisters)
ever do mathematics with you? Did they like to do mathematics?""Do you
rememberdoing puzzles or playing games at home? What games did you play?"
"Did you ever see or do mathematicson TV?" "Do you rememberdoing mathe-
matics with friends?"Each question is followed up at the clinician's option; for
example, "When did that happen?How old were you then? Could you tell me
more about what happened?"In all instances, the child is asked, "How did you
feel about that?"or "How did you feel when that happened?"and, if not yet
described,"Did you enjoy it? Was there anythingyou didn't like about it? How
do you feel about mathematicsnow? Do you think this ... has anything to do
with how you feel about mathematicsnow?"
The child is also asked, "Do you think you are good at solving problems?"
"Whatdo you thinkmakes someone a good problemsolver?""Whodo you think
solves problemsbest in your class? Why do you think (name) is a good problem
solver?"
Two different sets of problems are then presentedsuccessively: (a) cutting a
birthdaycake (withoutor with frosting)to shareequally amongtwo or threechil-
dren and (b) moving colored jelly beans back and forth between two jars. Both
problemsembody symmetryand coordinationof conditions-the first in the con-
text of volume and area, the second in a numerical context. Emphasis is on
exploring the child's affect as well as his or her metacognitionsabout the two
tasks. Materialson the table are a ruler;markers;pencils; a pad of blank paper;
scissors; sheets of graphpaper;a spool of string and a length of cut string;con-
struction paper, Styrofoam shapes with rectangular,circular, and triangular
bases; andjelly beans.
The main birthdaycake questions were the following:
* "Whichwould be easier, to cut a birthdaycake into three equal pieces or four
equal pieces? Why? Could you explain that to me?"
* "Does the shape of the cake matter?"
* "Supposethe cake has icing on the top and on the sides. (Four) (three)people
are at the birthdayparty.How would you cut the cake so thateach person gets
an equal amountof cake and an equal amountof icing?"
After various explorations,broughtto a close when 25 minutes have passed
since the startof the interview, the child is encouragedto retrospectwith addi-
tional questions. Then two transparentglass baby-food jars with twist-off lids,
each filled nearly to the top with jelly beans, are presentedto the child. One has
100 orangejelly beans and is labeled "ORANGE";the otherhas 100 greenjelly

beans, and is labeled "GREEN."
* "The next problem is about jelly beans. This jar has 100 green jelly beans
[points to the green jar], and this jar has 100 orange jelly beans" [points].
Suppose you take 10 greenjelly beans from the greenjar and put them into the
orangejar [points] and mix them up [pretendsto transferthe jelly beans, but
does not do it]. Then supposeyou take 10 jelly beans from this mixtureandput
them back into the green jar [pretends].Which jar would have more of the
other color jelly beans in it? Would there be more green jelly beans in the
orangejar, or would therebe more orangejelly beans in the greenjar?"
If the child does not do so spontaneously,he or she is first encouragedto try
the experimentand, if necessary, is guided to do so as follows:
* "Could you show me how to do it with the jelly beans? Let's try the
experiment ..."
* "Will it always come out that way? Why do you think so?"
After the studenthas expressed a firm conclusion, the clinician asks follow-up
questions and a final set of retrospectivequestions focusing on affect as well as
on cognition.
Interviews 4 and 5 returnto selected mathematicalideas from the first two

interviews. Interview 4 (41 pages, up to about 55 minutes) again explores the
child's strategic and heuristic thinking in the context of sequences of cards, in
close parallel with Interview 1. Materials this time include a hand calculator.
Four problems, depicted in Figures 4.2a-d, are presentedin succession, in the
formatof Interview 1: "Hereis the first card,here is the second card,and here is
the thirdcard."After a brief pause to allow a spontaneousresponse or detection
of a patternin Problem 1, the child is asked, "Whatdo you thinkwould be on the
next card?"and questionsareposed as in Interview 1. After severalquestions,or
after 15 minutes, Problem2 is presented(see Figure4.2b), withoutthe clinician
removing the cards of Problem 1 that were discussed. After furtherexploratory
questions, Problem 3 is posed (Figure 4.2c), and after additionalquestions the
child is given Problem4 (Figure4.2d).
The key follow-up questions in all four problems are similar to those in
Interview 1. Once the child has given both an external representation(for
Problem 1 only, a good verbal descriptionis accepted) and a coherent reason,
the clinician moves to the next problem. As usual, suggestions are made only
when the child reaches an impasse. If the child does not spontaneouslydetect
relationshipsbetween problems, the clinician asks about this after Problem 2.
During the final retrospective,the first 3 cards of Interview 1 (Figure4.1), with
which the children engaged a year and a half earlier, are laid out. Gesturingto
GeraldA. Goldin 49
1 1+3 1+3+5 * :: :
a. b.
Thefirstthreecards Thefirstthreecards
presentedin Problem1 presentedin Problem2
*:o0 ?0
... . * ....-
000
*00 *0000i
c. d.
Thefirstthreecards Thefirstthreecards
presentedin Problem3 presentedin Problem4
Figure 4.2. The sequence of tasks in Task-BasedInterview4.
all the cards, the clinician asks if the child sees any way to relate today's cards
to the previous cards.
Interview5 (27 pages, up to about55 minutes) also returnsto selected mathe-

matical ideas from the earlierinterviews,particularlyfractionsrelatedto 1/2, 1/3,
and 1/4 as exploredin Interviews2 and 3. Materialsgiven to the studentinclude
scissors; a 12-inch ruler marked in both inches and centimeters; an 18-inch
length of white curling ribbon;papercircles, squares,and triangles;a pile of red
and white plastic chips; a calculator;paperand pencils; and a solid piece of wood
the approximateshape and size of a stick of butter,measuring1"x 1"x 5". The
interviewbegins with open-endedquestionsaboutfractions:"Whenyou thinkof
a fraction,what comes to mind?""Canyou tell me more about that?""Canyou
show me what you mean?""Have you studied fractionsin school yet?" "What
(else) have you studiedaboutthem?""Do you like fractions?""Whatdo (don't)
you like about them?"
The child is then given a sheet of pink paper with five fractionswrittenon it
and is asked a series of questions; as always, spontaneousproblem solving is
allowed before the next question:
1 1 2 3 4
2 3 3 4 6
* "Whatfractionsdo you see here?""Canyou explain ...what one of these frac-
tions means?""Why is it writtenthis way?" "Couldyou show me [using] the
materials?"
* "Which fraction is the (smallest) (largest) fraction in the group?" "Why?"
"Couldyou show me what you mean?""Are there any fractionsin this group
50 ObservingMathematicalProblem Solving
that are the same size?" "(Why?)(Why not?)""Couldyou show me what you
mean?"
Next some pictorialrepresentationson a sheet of yellow paperand new ques-
tions are given (see Figure 4.3):
1 2 3 4 5
11
l I 11
II I I2
I
I I3l I r4'I I 5
1 2 3 4 5
Figure 4.3. PictorialrepresentationspresentedduringTask-Based Interview5.
*"Could you use a fraction to describe any of these pictures?""Whatfraction

or fractionswould you use?" "Why?""Couldyou show me what you mean?"
All spontaneousanswers are accepted, after which the clinician asks aboutthe
picturesthe child may have omitted and whetherthe pictureson the yellow sheet
of papergo with each other or with the fractionson the pink sheet.
The child is next given a sheet of blue paperwith five new fractionswrittenon it.
5 3 5 11 10
5 1 4 8 8
GeraldA. Goldin 51
For the balance of the interview,the child solves up to four problemtasks, one
at a time), each accompaniedby exploratory,nondirective questions. It is not
expected that all problemswill be completed. When 5 minutes remain,the clin-
ician skips to the final retrospective:
* [A circularshapeis presented.]"How could you show one thirdof this shape?"
"Why is that one third?""Is there any other way to show one third?""How
could you show one fourthof this shape?""Whyis that one fourth?""Is there
any other way to show one fourth?"
* [The 1" x 1" x 5" piece of wood is presented.]"Pretendthis is a stick of but-
ter. You need a tablespoon of butterto make a cake. You don't have a mea-
suring spoon, but you know that there are 8 tablespoons in a stick of butter.
Here is the butter.How could you find exactly one tablespoon?"[If the answer
is imprecise, ask once] "Is there any way to find out more exactly?"
* "Imaginea big birthdaycake shapedlike a rectangle.Can you imagine what it
looks like?""Describewhatit looks like." "Now imagine thatthereare 12 peo-
ple coming to the birthdaypartyand they each want a piece of cake. Yourjob
is to cut the cake so that each person gets the same-size piece. How will you
cut the cake?" "Could you show me what you mean?""Are there any other
ways to cut it?" [The clinician continuesto explore cuttingthe cake, including
the situationof icing on the cake.]
* "A toymakerfound some wooden shapes in the corer of her workshop.Some
were squares,and some were triangles.She decidedto put themtogetherto make
little houses [demonstratesusing a squareand a triangle).The squareslooked
like this [gesturesto the pile of squares].The triangleslooked like this [(gestures
to the pile of triangles].The houses looked like this [placesthe triangleon top of
the squareto make a figure that looks like that shown in Figure 4.4]. After a
while, she noticed that she had matchedexactly 3/5 of the squareswith exact-
ly 2/3 of the triangles. How many squares and triangles were there to start
with?""Using these materials,could you show me how she did that?"[If time
permits:]"Couldthere be a differentnumberthat works?"...
Figure 4.4. House composed of a squareand a triangle.
After each of these four problems, the child is asked, "Have you ever done a
problem like this before?" (If yes) "When?What do you rememberabout it?"
and so on. Interview5 ends, like the others, with a retrospectivediscussion.
Selected interviews with the children form the basis of a numberof studies.
The thesis of Zang (1994) examines the developmentof strategicthinkingin four
of the children, comparingInterview 1 and Interview 4; the thesis of DeBellis
(1996) studies affect in four of the children,using Interviews 1, 3, and 5; and the
thesis of Passantino(1997) looks at the developmentof fractionrepresentations
for all of the children,comparingInterviews2 and 5 (see also DeBellis & Goldin,
1997; Goldin & Passantino, 1996; Zang, 1995). With these scriptsas examples,
we now consider some generalperspectiveson structured,task-basedinterviews
of this sort as a researchtechniquein mathematicseducation.
ON THE SCIENTIFICNATURE OF TASK-BASED INTERVIEWS
The longitudinalstudy, like many thatuse task-basedinterviews,is explorato-

ry. Consisting as it does of a collection of individual case studies, its outcomes
are not in a strict sense scientifically reproducible,and it might seem at first that
this is all thatcan be said. Nevertheless,thereare certainrespectsin which meth-
ods of scientific inquiryhave been carefullyregardedin the creationand admin-
istrationof the interview scripts.I believe aspects such as these to be essential if
we are to make real progressin understandingthe natureof mathematicallearn-
ing and problem solving throughempirical observation.Thus, it is possible to
envision the researchbeing extended in a directionthat permitsreplicability.
First, it is crucial to maintaincarefully the scientific distinctionbetween that
which is observedand inferencesthat are drawnfrom observations.In this study
we (at best) are able to observe children'sverbaland nonverbalbehavior,as cap-
turedon videotapeduringthe sessions. From these observations,we (and others
who use similar methods) seek to infer something about the children's internal
representations,thought processes, problem-solving methods, or mathematical
understandings.We cannot "observe"any of the latterconstructs.
Second, our inferences are going to depend on (often tacit) models and pre-
conceptions about the natureof what we are trying to infer and its relation to
observable behavior. A scientific goal of the theory of mathematicseducation
must be to make such models as explicit as possible. As we do this, we move
away from dependingon the ad hoc design of task-basedinterviewstowardcon-
structingthem more consciously on the basis of explicit theoreticalconsidera-
tions. The task-basedinterview is like an instrumentof scientific experimenta-
tion, and it is theory that describes how such an instrumentis expected to inter-
act with the system observed (in this case, the child as problem solver) so as to
permitthe drawingof valid inferences from the observationsand measurements
made. This point is discussed furtherin the next section.
Third,inferencesfrom task-basedinterviewsare likely to be unreliable,in that
differentobservers may disagree about what inferences they would make after
observing the same videotape-even when they agree on the theoretical con-
structsfor which they are looking. The process of drawinginferencesaboutchil-
dren's thinking is fraughtwith uncertainty.At least at the outset, then, another
GeraldA. Goldin 53
scientific goal must be to describethe criteria thatareto be used when inferences

are drawn,so that the inferencingprocess itself becomes open to discussion.
For these issues to be addressedmeaningfully,there must be a sense in which
the task-basedinterviews are themselves explicitly characterizableas research
instruments, subject to reuse, refinement, and improvement by different
researchers.Thus in the study describedhere, we devoted greateffort to structur-
ing the interview scripts-ahead of their actual administration-to achieve two
features:(a) flexibility and (b) reproducibility.Let us considerthese twin goals.
Flexibility by the clinician in a task-basedinterview means being able to pur-
sue a variety of avenues of inquirywith the learneror problem solver, depend-
ing on what takes place duringthe interview. Such flexibility is essential for our
investigationsto allow for the enormousdifferencesthat we know occur in indi-
vidual problem-solvingbehaviorsand that we infer exist in individualchildren's
meaning-makingactivity. Because a majorgoal is to elicit and identify process-
es the childrenuse spontaneously(i.e., without direct hints or coaching), flexi-
bility is necessaryto avoid "leading"the child in a predetermineddirectionin the
problem solving.
Reproducibilityin contrast means that the clinician is not merely inventing
questionsextemporaneouslyas the child responds.It permits,to a certainimper-
fect but improvabledegree, "thesame interview"to be administeredby different
clinicians to differentchildrenin differentcontexts. The degree to which this is
possible increases as the researchexperiencebase with each particularinterview
accumulates.In assertingreproducibilityas a fundamentalgoal, I am fully aware
that I take a position at variance with the version of radical constructivismthat
asserts on a priori grounds its impossibility. The argumentis sometimes made
that since no two individualsever solve "the same"problem,reproducibilityis a
fiction. The mistake of those asserting this position is confusing the problem
instrument(the task, or task-based interview instrument,as structuredby the
researcherapartfrom the child) with the interactionbeing observedor measured
(the problemsolving that occurs when the child participateswith the clinician in
the actual interview). Of course, no two sequences of interactionsare identical.
From a scientific perspective,however, the wide differencesthat are observedto
occur from interview to interview can be betterunderstoodand attributedwhen
variables that are in principle subject to control (i.e., the task variables) are, in
fact, controlled.Thus, the creationof reproducibletask-basedclinical interviews
is an essential scientific step.
To accomplish this step, sufficiently many problem-solving contingencies
must be anticipated.The criteriafor the clinician's choices of questions or sug-
gestions must be made as explicit as possible in advance for each contingency,
with the balancecovered by generalinstructions.This is what we have sought to
do in the process of interview design.
For example, in Interview 1, three cards are presented.After a brief pause (to
allow any spontaneousresponsesto the presentedcards),the child is asked,"What
do you thinkwould be on the next card?"Contingenciesthen include "response"
54 Mathematical
Observing ProblemSolving
and "don'tknow."If the child responds,the next contingenciesinclude "offersa

complete, coherentreason"or "has not yet given a complete, coherentreason,"
with or withouthaving constructeda "coherentexternalrepresentation." The def-
initions (from the directionsin the Interview1 script)are as follows:
A completeandcoherentverbalreasonmeansonebasedon a describedpattern.A
coherentexternalrepresentation meansa drawing,picture,or chipmodel.It is not
requiredthatthe"canonical"fourthcard(with7 dots)be drawn,orthecanonicalpat-
terndescribed,fora responseto be considereda completeandcoherentreasonanda
coherentexternalrepresentation.An answersuchas "7, becauseit's 2 more"is a
coherentverbalreason,but it is not consideredcompletebecauseit refersonly to
findingthe next cardandnot to the basis for the pattern.An answersuch as "7,
becausethiscardhas2 morethanthatone, so thenextone has 2 morealso"would
be consideredcoherentandcomplete.If thereis a discrepancy betweenthe number
of dotsstatedandthe numberin an externalrepresentation,the verbalreasonis not
considered"coherent." This [describes]the "boundary"betweenresponsesthatare
andarenotacceptedas completeandcoherent....
The clinician's next question or suggestion (e.g., "Why do you think so?" or
"Canyou show me what you mean?"that leads, if necessary,to "Canyou show
me using some of these materials?")depends on the contingency that best
describesthe child's response.This is the level of detail at which many (although
not all) contingencies are consideredin the scriptdesign. We thus seek to make
explicit the usually tacit conditions that ordinarilyinfluence a skilled clinician.
But this level of detail demandsmuchpreparationandrehearsalby the clinicians.
In principle, such detailed structuredinterview descriptions lead to several
desirable features: (a) increased replicability of the interview itself, although
contextualand otherfactorswill still vary widely from occasion to occasion, and,
of course, the knowledge structuresof individualchildrenarehighly variable;(b)
a degree of comparabilityof interview outcomes between different children,
across different populations of children, across different conditions of school
learning, and so forth; (c) subsequentexperimentsto investigate the generaliz-
ability of observationsmade in individual case studies; (d) explicit discussion
and critique of the contingenciesbuilt into the interview,which permitsthe cri-
teriafor the clinician's responsesto be analyzedand improved;and (e) an explic-
it basis for discussing the analysis of outcomes, that is, the process of drawing
inferences from observations.For earlierperspectiveson these ideas, see Cobb
(1986), Goldin and McClintock(1980), Hart(1986), and Steffe (1991a).
THE ROLE OF THEORY

One purpose of clinical task-basedinterviews in mathematicseducationis to
permitus to characterizechildren's strategies,knowledge structures,or compe-
tencies-perhaps to be able to look at the effectiveness of instruction,to under-
stand developmentalprocesses better, or to explore problem-solvingbehavior.
However we choose to define our inferentialgoals, a theoreticalframeworkfor
describing or characterizingwhat we seek to infer is necessary. But the role of
GeraldA. Goldin 55
theory is not limited to this. Theory must also tell us something about how the
characteristicsof the task in the task-basedinterview(e.g., its language,its math-
ematical content and structure, its appropriatenessfor particular cognitive
processes, the interview context) are expected to interactwith the cognitions we
are tryingto infer, so thatthe interviewcan be designed to elicit processes of the
desired nature.To say that the problems in the task-basedinterviews described
here are of a level of complexity thoughtto permit a variety of strategiesto be
employed, or internal representationsto be constructed, already presupposes
majortheoreticalassumptions.
The questions asked and the observationsmade duringany scientific investi-
gation, including investigations using task-based, clinical interviews, depend
heavily on the theory we bring to it. Thus, in my view, the main question is not
whether theory should influence us in this enterprise.I maintain,in agreement
with R. B. Davis (1984), that it always, inevitably does:
Perhapstheattemptsto usethemethodsof science[ineducation] havefailedbecause
sciencehasbeenmisunderstood.
In these attemptsit had been assumedthatsciencewas primarilyfactual,that
indeedit dealtalmostsolely in facts,thattheoryhad no role in science.Careful
observationof sciencerevealsthisto be false.It mightbe closerto the truthto say
that"facts"-atleastinterestingfacts-are almostunableto existexceptin thepres-
enceof an appropriate theory[emphasisin original].Withoutanappropriate theory,
one cannotevenstatewhatthe"facts"are.(p. 22)
The questionpertainingto clinicalinterviewsis the extentto which the influence
of theory remains tacit, taking place throughunconsciousassumptionsof clini-
cians, researchers,and/orteachers,or becomes explicit andthusopen to discussion
and challenge.Ourgoal in the presentstudyis to be as explicit as possible.
The theoreticalunderpinningof this series of interviews includes the concept
of (internal)competenciesand structuresof such competencies. These are envi-
sioned as developing over time in the child and as being capable of being
inferred from observable behavior-when the appropriateconditions exist for
the individualto take certaincognitive steps and some correspondingbehaviors
are seen. Anotherfundamentaltheoreticalassumptionis the idea that competen-
cies are encoded in several different kinds of internalrepresentationsand that
these interact with one another and with observable, external representations
duringproblemsolving. A thirdassumptionis that representationalacts occur in
which representationalconfigurations(internalor external)are taken to symbol-
ize or standfor other representationalconfigurations.
The model that most stronglyinfluenced the developmentof the scriptsis one
thatI have been developing for some time as a way of characterizingmathemat-
ical problem-solvingcompetency. It includes five kinds of mutuallyinteracting
systems of internal, cognitive representation(Goldin, 1987, 1992b): (a) a ver-
bal/syntactic system (use of language); (b) imagistic systems (visual/spatial,
auditory,kinestheticencoding); (c) formalnotationalsystems (use of mathemat-
ical notation);(d) planning, monitoring,and executive control (use of heuristic
strategies);and (e) affective representation(changingmoods and emotions dur-

ing problem solving). The interplaybetween the children's internalrepresenta-
tions and externalrepresentationsthatthey use or constructduringthe interviews
provides one of the most importantmeans of drawinginferences.
For example, from children's descriptive statements about what a birthday
cake looks like (Interview 5, Problem 3) we infer internalvisual/spatialrepre-
sentations.From their gestures as they describe how they would cut a birthday
cake into 2 or 3 pieces (Interview3) or 12 pieces (Interview5), with accompa-
nying drawings, we infer simultaneous internal, kinesthetic representations.
Children'sexplanationsof the fractionswrittensymbolically in Interview5 per-
mit inferencesconcerningtheirinternalrepresentationsof this formalmathemat-
ical notation. Steps they take relating one sequence of cards to another in
Interview4 permitinferencesconcerninginternalexecutive control (heuristicor
strategicrepresentations).Affective representationis inferrednot only from the
child's statementsin response to questions,but also from facial expressions and
spontaneouscommentsand gestures.I would stress again thatthe whole process
of inferencing is, at this stage in the research, of limited reliability, but that
strengtheningthe degree of reliabilityis an importantgoal.
Since the study is longitudinal,a majorfocus is how systems of representation
develop in the child over a period of time. In this respect, the theoreticalmodel
incorporatesthreemain stages: (a) an inventive/semioticstage, in which internal
configurationsare first assigned "meaning,"(b) a period of structuraldevelop-
ment, driven by the meanings first assigned, and (c) an autonomous stage, in
which the representationalsystem functions flexibly and in new contexts. We
hope to be able to infer representationalacts associatedwith each of these stages.
The distinction between external and internal representationmeans that we
must attendcarefully to both. We regardthe tasks posed as externalto individ-
ual children,embodyingsyntax, content,context, and structurevariablesthat we
select when we design the interviews. In particular,the mathematicalstructures
of the tasks (semanticstructuresand formal structures-additive, multiplicative,
and so forth) are consciously chosen. The behaviorsobserved result from inter-
actions between the task environmentand the child's internalrepresentations.
To posit interactionsbetween internal and external representationalsystems
thus requiresa great deal of analysis of mathematicalstructuresassociatedwith
the tasks. Parallelbut not identical structures-in some instances,homomorphic
structures,in other instances, structuresless directlyrelated-were intentionally
included in the differentinterviews.For example, a certainadditive structureis
embodied in the (canonical)sequence in Interview 1. Otheradditive and multi-
plicative structuresrelate to the sequences in Interview4, which are also struc-
turallyrelatedto each other.The cardsequencesare all presentedin parallelways
to the children.A certainmultiplicativestructureunderliesthe cube-cuttingtask
in Interview2. Reflection symmetriesare embodiedin the cards in Interviews 1
and4, in the cutoutand cube-cuttingtasksin Interview2, and in the birthdaycake
task in Interview 3. More subtle, hidden symmetry is present in the jellybean
GeraldA. Goldin 57
problemin Interview3. Rationalnumberstructuresoccur in Interviews2 and 5.

The analysis of all these relationshipsis theorybased, and many assumptionsare
being madejust in assertingthat a structuralrelationshipbetween tasks exists.
Anotherkey theoreticaldistinctionis between the child's spontaneousbring-
ing to bear of any particularcompetency, or the child's doing so only when
prompted. This is a subtle but crucial distinction, which involves the child's
exercise of planningcompetencies to call on other competencies (verbal, imag-
istic, formalnotational).For example, from a child's spontaneousresponseto the
task in Interview 1 that each card is two more than the previous card can be
inferredthe implementationof at least partof a problem-solvingplan. Should the
child make the same observationonly after being asked by the clinician, infer-
ence of such a planningrepresentationwould be unwarranted.These ideas have
influenced the task-basedinterview developmentas follows: We pose tasks that
permitthe childrento performat each step spontaneously.We explore not only
the child's overt behavior, but the reason the child states for taking each step.
Recognizing that competency structuresmay be partiallydeveloped, we do pro-
vide hints or heuristicsuggestions when blockage occurs. This often permitsthe
child to demonstratecompetenciesthatotherwisehe or she would never "get to"
during the problem solving, which adds to the information gained. There is
always a trade-offentailed here, in that the more specific the hint or suggestion
provided by the clinician, the less extensive the informationgained about the
child's representationof planningand executive control in problemsolving.
We seek informationabouteach kind of internalrepresentationalsystem; thus,
not satisfied with a coherentverbal explanationonly, we nearly always encour-
age the child to constructa concrete,externalrepresentation.We include a cross
section of questionsexploringvisualization,affect, and strategicthinking.In par-
ticular,Interview2 is designed especially to detect and explore in greaterdepth
imagistic systems (visual/spatial and tactile/kinesthetic) in problem solving
while attendingto affect and to otherkinds of internalrepresentation;Interview
3 focuses on affect in greaterdepth (see also McLeod and Adams, 1989), where-
as Interview4 returnsto tasks selected for the possibility of eliciting certainplan-
ning or strategicrepresentationalcapabilities.
It is my view that the characteristicsof the task-basedinterviewsare variables
that are inevitably built into clinical interview designs. Considerationsof task
structurein these interviews are sufficiently complex to form the basis them-
selves of several articles, yet task structureis an essential componentto under-
standing and making inferences from observed problem-solving behavior. It
needs to be examined independentlyof the individual children as a part of the
process of drawing inferences from children's interactionswith the tasks. My
main point is that there is no way to avoid this interplaybetween theory and
observation.It is not a sufficient answer to respond,as some do, that task struc-
tures do not "exist" apartfrom individualproblem solvers. We simply have the
choice of proceeding unscientifically,choosing tasks that seem interestingand
just "seeing what happens," or trying to proceed systematically with tasks
explicitly described and designed to elicit behaviors that are to some degree
anticipated.
Although the analysis of outcomes in these interviews is theoreticallybased,
we seek not only to observe and draw inferences from expected processes but
also to search for unanticipatedoccurrences.The hoped-for results include the
furtherrefinementand developmentof the theoreticalmodel for problem solv-
ing, including the identificationof inadequaciesand progresstowardan assess-
ment framework,as well as conjecturesfor furtherinvestigationthroughfuture
experimentalstudies.
THE ROLE OF CONTEXT
Task-basedinterviews do not take place outside of a social and psychological

context. That context influences and places constraintson the interactionsthat
occur during an interview and puts limitations on the inferences that can be
drawn.It is one of the componentsthat theory must address,if we are to validly
interpretinterview outcomes.
The view taken here is that "social and psychological context" affects the
interview interactionsthrough internal representationsthat the child has con-
structed, which are in principle subject to description. These are considered
"contextual"because the semantic content of the representationalsystems
involved is not, at least initially, mainly derived from, or related to, intended
mathematicalrepresentationsassociatedwith the tasks posed in the interviews.
We observe, for example, thatthe child's expectationsin an interviewmay be
influencedby the fact thatit is conductedby a relative stranger,the clinician.The
interviewtakesplace in school andthus mightbe assumedby the child to involve
some kind of test that "counts"toward an evaluation. Childrenoften seem to
think, especially at the outset, that the tasks are likely to have "right"and
"wrong" answers and that certain methods will meet with the clinician's
approval,whereas others will not. The interview itself may be taking place at a
momentwhen the child is alert, tired, hungry,distracted,or excited. On the one
hand,the child might preferto be back in his or her regularclass with friendsor
might, on the other hand, be looking forwardto an interestingbreak from the
classroomroutine.The fact of being videotapedwas for the childrenin our study
a familiarexperience (owing to the projectin which their teacherswere partici-
pating);the context of their experience would be differentwere the video cam-
eras a complete novelty.
It seems to be an almost inevitable feature of task-based,clinical interview
methodology that the tasks are unrelated,at least initially, to a goal or purpose
generatedby the child. For example, the butterproblemand the toymakerprob-
lem (Problems2 and 4 in Interview 5) are both posed in a stated context. The
butterproblem(or one like it) is a problemthatcould conceivably arise as a prac-
tical need in a varietyof real-life situationsnot too differentfrom the statedcon-
text. It would very likely be experienceddifferentlyif the child were actually in
GeraldA. Goldin 59
one of those situationsand had generatedthe problemgoal (as opposed to solv-

ing the problem as part of a clinical interview). The toymakerproblem, in con-
trast, is a rewrite of a ratherwell known mathematicalproblem involving mar-
ried couples in a village. We rewrotethe problemto presenta concrete,external
representationwith which the child could experimentif desired. Although the
context of making toys is one the child can easily imagine, the problem goal is
not one thatoccurs "authentically"in thatcontext. It is posed as an almost whim-
sical question, arisingperhapsas a curiosity (curiosity-basedproblemsolving is,
of course, an essential aspect of mathematicalinquiry) but not as a practical
question thatneeds to be answeredfor the toymakingto proceed. Thus, the con-
texts of these two problemsare differentin an importantrespect. Such contextu-
al factors could influence, for example, the importancethat the child ascribes to
the problemgoal and, in turn,the child's persistence,enthusiasm,choice of strat-
egy, and so forth.
Anothermeaningof context, one thatmight be called "mathematicalcontext,"
refersto unstatedaspects of the tasks themselves as they arepresentedduringthe
interview-aspects that although seemingly small may have importanteffects.
For example, in presenting the three cards in Interview 1 and again (several
times) in Interview4, we permit the child to see the cards being drawn from a
stack of cards in a manila envelope. From this minor contextualfeature (which
was intentionallyincluded),the child may infer thatthereis a deck of cardslarg-
er than the three that are shown and, possibly, that there is a patternin the cards.
Three cards presented wholly out of context might not so readily elicit this
expectation.Evidently, certaincontextualinfluences are undesirable(e.g., those
thatmight mask our ability to observe competenciesthatarepresentin the child),
whereas others are helpful (e.g., those that would facilitate the child's "thinking
mathematically").
Since so much thatmay occur duringa task-basedinterviewis context depen-
dent, how can we considerwhat we observe to be more thanaccidental,one-time
events? One importantcondition is to requirethat the constructs we inferfrom
our observationsbe reasonablystable against contextualvariations.For exam-
ple, suppose we infer, in Interview 2, a child's ability to representimagistically
(visually, kinesthetically,or both) the cuttingof a cube across two perpendicular
directions.The inference may be drawnfrom the child's coherentdescriptionof
the componentpieces of the cube, with appropriategestures indicatinghow the
cube was imagined to be cut. Although it is indeed the case that this child's
behaviormay vary considerablyfrom one context to another,when we infer such
particularcompetenciesor structuresof competenciesfrom thatbehavior,we are
inferringaspects of the child's cognition that we expect to be fairly stable. If the
inferredcompetency were to disappearin shortorder,it would not be useful in a
theory of mathematicallearning.
Understandingthe contextual dependence of the interviews also means rec-
ognizing how very difficult it is to establish advance criteria for all the infer-
ences about each child's cognition and affect that we want to draw from our
observations. When observations are interpretedin context, new likelihoods

occur. The plan we have been following is to make the best conjecturespossi-
ble and to try to be explicit aboutthe reasons for conjectures,including relevant
contextual factors, as these occur (Zang, 1994).
Such discussions of contextual issues barely scratch the surface. For task-
based interviewmethodologyto be pursuedseriously, a deeperunderstanding-
indeed, a theory of how social, psychological, and mathematicalcontextualfac-
tors may influence mathematicalproblem solving during a task-based inter-
view-is essential to the interview design process.
PRINCIPLESOF INTERVIEWDESIGN
I conclude this chapterby summarizingwhat, in my opinion, are some of the
most importantunderlyingcharacteristicsof the five interviews describedhere
and try to abstractfrom these the most salient general principles behind their
design. Althougheach interviewhas its own particularfocus, certainbroadchar-
acteristicsare maintainedin all of them:
1. Each interviewis based on particularmathematicalideas appropriatefor the
age group of the children(grades 3-6) and on mathematicaltopics with associ-
ated meaningful,semantic structures,as well as formal, symbolic structures,for
example, additive or multiplicativestructures,sequences, schemataunderlying
rationalnumberconcepts, and so forth.We want the mathematicalcontentto be
based on topics thatcan be studiedin depthand are flexible enough to allow evi-
dence of widely differing capabilitieson the partof the students.
2. Each interview consists of a series of questions posed in one or more task
contexts. These begin at a level that all the childrenare expected to understand
(of course, in differing ways). They become increasinglydifficult, culminating
in questionsthatcan still be attemptedby all the childrenbut thatwill pose major
challenges even to the most mathematicallyastute students.
3. The childrenengage in free problem solving to the maximumextent possi-
ble. This prioritizesexploring the strategiesthat the children use spontaneous-
ly-whatever method or methods seem most appropriateto them as they work
on the task. They are remindedoccasionally to talk aloud about what they are
doing and to describe what they are thinking.Hints and prompts,or new ques-
tions, are offered only after the opportunityfor free problemsolving and are then
followed by a furtherperiod of observinghow the child respondswithout direc-
tive intervention.This rule is (in view of time constraints)occasionally broken
because of our desire to ensurereachinga subsequentsection of the interviewin
the allotted time, but it is broken with the recognition that possibly important
informationis necessarilybeing lost.
4. All studentproductionsare "accepted"during the interview;the clinician
does not impose preconceivednotions aboutappropriateways to solve the prob-
lem but does treat"wrong"answerssimilarlyto "correct"answers(withoccasional,
GeraldA. Goldin 61
specified exceptions). Responses elicit follow-up questionswithoutan indication

of correctness.The rare exceptions, involving guiding the studentstoward par-
ticularunderstandings,are decided in advance and occur only where the under-
standingsare essential for subsequentinterview questions to be meaningful.
5. Materialsfor constructinga varietyof externalrepresentationsare available
for studentuse and vary from task to task:paperandpencil, markers,cards,chips
and other manipulatives,paper cutouts, hand calculators.A major task goal is
always the constructionof representationsby the children-ideally, a multiplic-
ity of them.
6. Each interview includes reflective questions, typically posed retrospective-
ly, that addressthe child's problem-solvingprocesses and the child's affect.
7. Because the interviews are designed for use in a longitudinal study, there
is a conscious effort to incorporate into later interviews some tasks that are
similar in context, mathematicalcontent, structure,or all three, to those posed
earlier.
Building on these specific characteristicsand the issues discussed in this chap-

ter, I propose to formulatethe following tentativeand partialprinciplesof inter-
view design and constructionwith the goal of trying to establish the strongest
possible scientific foundationand maximizing the informationgatheredthrough
a task-basedinterview.
1. Accessibility. Interviewtasks should embody mathematicalideas and struc-
turesappropriatefor the subjectsbeing interviewed.Subjectsmust be are able to
representtask configurations,conditions, and goals internallyand, where appro-
priate,externally.
2. Rich representationalstructure.Mathematicaltasks should embody mean-
ingful semantic structurescapable of being representedimagistically, formal
symbolic structurescapable of notational representation,and opportunitiesto
connect these. Tasks should also suggest or entail strategiesof some complexity
and involve planning and executive-control-level representation.Opportunities
should be included for self-reflection and retrospection.
3. Free problemsolving. Subjectsshouldengage in free problemsolving wher-
ever possible to allow an observationof spontaneousbehaviorsand reasons for
spontaneouschoices. Providingprematureguidanceresults in a loss of informa-
tion. This principlemay mean some sacrifice of the speed with which the subject
understandsthe problemor progressesthroughit.
4. Explicit criteria. Majorcontingencies should be addressedin the interview
design as explicitly and clearly as possible. These contingencies should distin-
guish "correct"and "incorrect"responses (but rarely) with structuredquestions
designed to give subjects opportunitiesto self-correctin any contingency. This
is an importantkey to the replicabilityand generalizabilityof task-basedinter-
view methodology.
5. Interaction with the learning environment. Various external representation-

al capabilitiesshould be provided,which permitsinteractionwith a rich, observ-
able learningor problem-solvingenvironmentand allows inferencesaboutprob-
lem solvers' internalrepresentations.
It is hoped that the discussion in this chapterfurthersthe goal of understand-
ing mathematicallearningand problemsolving scientifically throughthe use of
task-basedinterviews as researchor assessmentinstruments.
63
Chapter 5
Phenomenography: Exploring the

Roots of Numeracy
DagmarNeuman
This chapteruses the findings of a study of beginning school children's expe-

riences of subtraction to illustrate various aspects of phenomenographic
research.The focus of the chapteris placed on the model, developed from the
data, thatpicturesthe phenomenonbeing studied.A detailed analysis is present-
ed of the phenomenonas it is experiencedby the researcher,together with the
researchproblemthat the phenomenongave rise to. Later,in the descriptionof
the categoriesrelatedto the model, an accountis made of the ways the phenom-
enon seemed to appearto the subjectsin the study This informationprovidesthe
foundationfor a discussion of the criteriaof "interpretativeawareness,"proposed
as a means for establishing reliability in phenomenographicresearch,and also
for a discussion of two aspects of validity-the coherence and intelligibility of
the model and the possibility of demonstratingits pragmaticvalidity.
The study reported here had its origin in two conflicting observations
(Neuman, 1987). On the one hand, pupils with difficulties in mathematics,who
were interviewedat the end of their school career,were found who did not know
all the subtractionfacts within the numberrange 1-10. On the otherhand, infor-
mal observationsof 25 first-gradeclasses found that in all these classes, one or
more children already seemed to know these facts on their first day of school.
Both observationswere contradictoryto the commonly held view that all chil-
dren, when they do sums and drill tables duringtheir early school years, sooner
or later learn the basic facts by heart.
Findings from Swedish research(e.g., Kilborn, 1979) and from my early pilot
studies indicatedthat subtractionwithin the range 1-10 was an interestingarea
for investigation. These findings revealed that children who displayed difficul-
ties in mathematicsknew the additionfacts. They also revealedthatchildrenwho
I wish to thank Ference Marton,who was my advisor in this first attemptto carry out
phenomenographicresearch;Shirley Booth and Jorgen Sandberg,who patiently read ver-
sions of this paper and provided constructivecriticism that has helped to shape the final
product;and Anne Teppo, who made the most competenteffort to bring the chapterinto a
reasonablycomprehensibleform. I also wish to thankAnne Teppo for her thoroughwork
on the language used in the article. The studies that form the basis for the chapterwere
funded by the Swedish Board of Education,the Swedish Ministry of Education,and the
Solna Local EducationAuthority,who are acknowledgedwith gratitude.
64 Phenomenography:Exploringthe Roots of Numeracy
could subtractwithin this range in due time learnedto add and subtractwithin
higher numberranges, whereas those who could not never seemed to develop
mental calculationskills.
Four pilot studies that I carried out revealed that pupils with difficulties in
mathematicsused a counting approach,counting forwardor backwardin ones,
in their attemptsto calculate. Other children used what I called a structuring
approachthat helped them avoid counting. Insteadof counting four steps back-
ward to solve the task 9 - 4, for instance, they answered 5 instantly, with the
younger children explaining, if asked, that they knew because 4 + 4 = 8 or
because 5 + 5 = 10. Those explanations illustrate that their basic facts were
anchoredin a sense of numberand in a conceptualunderstandingof the inverse
relationbetween additionand subtraction.
I searchedfor a suitablemethodfor studyingboth the cause of the difficulties
some pupilsexperiencedandthe ways otherchildren,priorto enteringschool, had
begun to createa more viable sense of number.At GoteborgUniversity,I became
acquainted with the INOM group (INlarning och OMvarldsuppfattning,or
"Learningand ways of experiencingour world"),whose researchand education-
al aims fit my intentions.The professorin the center of the group was Ference
Marton,who latercoined the wordphenomenography(Marton,1981; Marton&
Booth, 1997) for the kind of researchthe groupcarriedout. I use examples from
the study I conductedto describecertainaspects of phenomenographicresearch.
PHENOMENOGRAPHY
To fully understanda qualitativeresearchmethodology,we must place it in the

context of an ontology, concerning the natureof the world we live in, and an
epistemology, concerning how we acquire knowledge about this world.
Therefore,in this part of the chapter,I discuss some of the fundamentalphilo-
sophical assumptionsunderlyingphenomenographicresearch.
Phenomenographyis a theoreticaland methodologicalresearchspecialization,
anchored in a nondualistic ontology, akin to a phenomenologicalphilosophy.
According to this ontological position, our world is a real world that is experi-
enced by all our senses but interpretedand understoodin differentways by dif-
ferent humanbeings, dependingon our earlierexperiences.This world, howev-
er, is seen as one world only, not as one subjectiveworld representedin the mind
of the individualand one objective externalworld. It is one world that is
objectiveandsubjectiveat the sametime.An experienceis a relationship between
objectandsubject,encompassing both;the experienceis as muchan aspectof the
objectas it is of the subject....The expression"howthe subjectexperiencesthe
object"is synonymouswiththe expression"howthe objectappearsto the subject"
(Marton& Neuman,1996,p. 317).
Yet phenomenographyis not phenomenology. Even if it definitely shares
some of the phenomenologicalphilosophy, it must be seen as an approachwith
its own fundamentalassumptions,methods, and goals.
Dagmar Neuman 65
The wordphenomenonis a key concept in phenomenography.It is used in the

sense it has in phenomenology-"the thing as it appearsto us"-opposed to the
Kantiannuomenon-"the thing as such." A phenomenon is not identical with
one of the ways in which it can be experiencedbut with the whole complex of
intertwinedways it has appearedto all people, both historically and in the pre-
sent. Whereas the goal of phenomenology is to describe the essence of all the
ways in which a phenomenoncan be experienced,the goal of phenomenography
is to explore and identify the variationin the ways in which it can appear.
A phenomenonappears,as a rule, in a limited numberof ways. This is a cen-
tral assumptionin phenomenography,where findings are mainly grounded in
empiricalanalyses. It has also been empiricallyshown thatif 20 to 30 individuals
are interviewed,and otherpeople from the same populationare interviewedlater,
thererarelyappearsany new way of experiencingthe phenomenonthatis studied
(e. g., Giorgi, 1986; Marton,Beaty, & Dal'Alba, 1993; Saljo, 1979, 1982).
If there are only a limited number of ways in which a phenomenoncan be
experienced,it must be of educationalvalue to try to reveal these ways. Marton
and Booth (1997) suggest that the models depicting the outcomes of phenom-
enographicresearchhave educationalvalue in helping teachers and curriculum
designersidentify "a notionalpathof developmentalfoci for instruction"(p. 81).
PHENOMENOGRAPHICRESEARCH
To think of researchwith a fallibilist epistemology-and with results that are
interpretationsof other peoples' ways of experiencing something-in terms of
reliability and validity may be seen as a contradiction.The crucial thing for
establishingreliability in phenomenographicresearchis the use of phenomeno-
logical reduction (Sandberg, 1996) or, in Ihde's (1997) words "to circumvent
certain kinds of predefinition"(p. 31). Sandbergsees the researcher'sinterpre-
tive awarness as one possible criterionof reliability. Following Ihde, he gives
five guidelines for how to maintainsuch an awarenessthroughoutthe research
process. The first criterionis thatthe researchermust be continuouslyorientedto
the phenomenonbeing studied throughoutthe researchprocess. To be oriented
to the phenomenonalso means to be orientedto the formulationof the research
question. Referring to Kvale (1994), Sandberg points out that a weakness in
many qualitativestudies "is the lack of a clear definition of the researchques-
tion" (p. 157). This, ratherthan the variationin possible interpretationsof the
data,often makes the presentedresultsdifficult to understand.Second, the analy-
sis and presentationof the outcomes should consist of a descriptionof the ways
of experiencingthe phenomenon,not of explanationsof why these experiences
appearthe way they do. (Researchersare often tempted to use their arsenal of
theories and models to explain things outside the experiences reportedby the
interviewees.) Third, all aspects of the experiences that are observed should, at
the beginning of the analysis, be seen as equally importantin orderto faithfully
interpret the essential aspects of the interviewees' ways of experiencing the
phenomenon. Fourth, the researchermust continually adapt the different pos-

sible interpretationsthat appearwhen he or she reads throughthe data until the
basic meaning structurehas been stabilized. Finally, the researchershould not
only identify what the interviewees experience but also how they experience
this "what."The concluding model of the descriptions should relate the inter-
viewees' expressions of what they seem to experience to how they seem to
experience it. This chapter focuses on the first, second, and fifth criteria by
using the phenomenographicmodel developed in the School StarterStudy to
illustrate the implementationof interpretativeawareness.
Study Design
The assumptionof educationalbenefits relatedto phenomenographicresearch
influenced the design of my study and made it ratherembracing.I decided to
interview all children in two new beginner classes during their first weeks at
school before they received any formalteachingin mathematics.I then followed
these children for 2 years in a teaching experiment based on the knowledge
obtainedin the interviewstudy.After these 2 years, I carriedout a new interview
study with those childrenwho had not been able to solve any of the subtraction
problemsin the interviewsconductedwhen they began school. I also interviewed
childrenin a control groupboth when they began school and 2 years later. As in
the researchclasses, only those control childrenwho had not been able to solve
any of the subtractionproblemsin the interviewsat the startof school were inter-
viewed later. I also regularlyvisited the two classes using the teaching experi-
ment and met the teachersat least once every fortnightto listen to their experi-
ences and to plan the new work.
One importantgoal for me in the interviewstudywas to reveal maximumvari-
ation in the ways the phenomenonI studiedcould appearto all pupils in a typi-
cal Swedish school beginnerclass before formal instructionin mathematicshad
begun. A goal of similarimportancewas to reveal the maximumnumberof ways
thatpupils in "therisk zone" might experiencethe phenomenon.To have at least
20 pupils in this zone, I also carried out one extra interview study with two
school beginner classes 1 year after the first interview. This latter study also
enabled me to determinethat no new categories could be identified.
To be able to study a dynamic learningprocess duringan extended period of
time, I also decided to follow in more depthtwo childrenin the researchclasses.
I used the results of the interviewscarriedout at the startof school to select two
children who had displayed the earliest, that is, the least developed, ways of
experiencingthe studiedphenomenon.I then met with them twice a week for 2
years in clinical interview lessons that were tape recorded and transcribed
(Neuman, 1994).
Thus, three kinds of methods were used: clinical interviews, a quasi-teaching
experiment, and a longitudinal case study using recorded clinical interview
lessons. The teaching experimentand the case study were mainly undertakento
Dagmar Neuman 67
deepen the understandingacquiredthroughthe interview study. They also pro-

vided validity throughtechnical triangulation(Larson, 1993), in which different
forms of data are assembled concerning the same phenomenon,and pragmatic
validity (Kvale, 1989) through putting the results of the interview study into
practice.This chapterfocuses on informationdeveloped from the initial clinical
interviews.
The Definition of the Phenomenon

A clear definition of the phenomenon to be studied, as experienced by the
researcher,is of great importancein phenomenographicresearch. If the phe-
nomenon is not well defined, it is impossible to formulateappropriateinterview
questions or to present the outcomes of the study as representingthe variations
in the ways in which this phenomenonhas appearedto the interviewees.The def-
inition of the phenomenonis closely related to the formulationof the research
problemthat the researcherhas set out to study.
My research problem concerned subtractionwithin the range 1-10. I had
alreadyobservedthatfor pupils displayingmathematicaldifficulties, subtraction
problems could be hard or easy depending on the numbersused. For instance,
ProblemA "Andyhad two pencils, but nine childrenwanted to make drawings;
how many more did he need?" could be hard, whereas the same problem with
seven pencils (ProblemB) seemed easy. In the same way, ProblemC, "Andyhad
nine pencils and lost seven of them; how many are left?" could be hard,where-
as the same problemwith two pencils lost ProblemD could be easy.
An analysis of the strategies used by the pupils in the pilot study who dis-
played mathematicsdifficulties revealed why ProblemsA and C could be con-
sidereddifficult. For these pupils, exactly as for very young children,the seman-
tic structureof the problems, not numericalfactors, seemed to be the focus of
attention when they solved word problems (Carpenter& Moser, 1984). They
seemed to experience missing addends (A and B) as addition,thinkingof addi-
tion as counting forward,and "take away" problems (C and D) as subtraction,
with subtractionas counting backward.Thus, the pupils solving ProblemA had
to count seven steps forwardand in ProblemC, seven steps backward,whereas
they had to count only two steps forwardin ProblemB and two steps backward
in problemD.
In his research,Fischer (1992) has illustratedthat there is an enormousdiffer-
ence between how childrenexperience threeand four visually presentedobjects.
Whereas nearly all 3- to 4-year-olds immediately denoted a collection of three
objects as "three"without counting or grouping, hardly any of these children
could denote a collection of four objects correctly without counting the objects
or groupingthem into two groups of two.
Figure 5.1 is a picturesimilarto those I used to communicatethe phenomenon
I wanted to study. Children in my pilot studies sometimes solved problems
through these kinds of drawings. They depicted, as I saw it, the one-to-one
68 Phenomenography: Exploring the Roots of Numeracy
correspondencebetween numbersymbols and counted objects that, according

to research on early numerical thinking (e.g., Gelman & Gallistel, 1978;
Steffe, von Glasersfeld, Richards, & Cobb, 1983), represents an important
aspect in the child's early conception of number.
Figure 5.1. Child's drawingof "nine stones" (Skolverket, 1995).
In Figure 5.2, I have transformedFigure 1 to fit Problems A and C as they

seemed to appearto those children who focused their attentionon the situated
aspects more than on the numericalaspects of the problems. (In the following
discussion, "number"is considereda part-part-wholepattern.)In ProblemA, the
part of the numberexperienced as "the missing part"appearsas the last part,
since it is thoughtof as addedto the known part"two."Similarlyin ProblemC,
the partof the numberthat is experiencedas lost appearsas the last part,since it
is thoughtof backwardfrom 9.
> Problem A: 2 + =9
123456789
000000000
< Problem C: 9 - 7 =
Figure 5.2. Depiction of situatedaspects of ProblemsA and C.
To perceive the numerosityof the last partof a numbercan be extremelydiffi-

cult, since it can never be experiencedin an ordinal way, throughwhat Fuson
(1992) has called "count-to-cardinal" (p. 134). The counting word relatedto the
last objectof this partnevertells you anythingaboutits cardinality.Neitherin the
missing addend,2 + _ = 9, nor in the take-awaysubtraction,9 - 7 = _, does the
word nine tell you aboutthe numerosityof the last part,seven, in Figure 5.2. In
ProblemC, nine is actuallythe first word the pupils experienceif the subtracted
partis countedbackward,andthe secondwordin the problem(seven)does not tell
themanythingaboutwhereto stopthe backwardnumerationto find the wordat the
limit of this part.If the numerosityof the last partis outsidethe subitizingrange,it
also cannotbe subitizedas a cardinalityor experiencedintuitivelyin any way at
all. We have to invent some idea or some methodto be able to know aboutit.
DagmarNeuman 69
In the pilot study, the pupils who displayed mathematics difficulties had
invented a method of "doublecounting"to deal with the numerosityof the last
part. They put up one finger for each enumeratedword when they counted this
part.Finally they "readoff' the finger configuration,for instancein ProblemsA
and C, as "one hand + 2 = 7." This double countinghelped childrenperceive the
numerosityof the last partin Problem A and told them where to stop the back-
ward enumerationin Problem B. Yet it was a solution strategythat seemed to
stay concrete and procedural,never being transformedinto objects that could be
operatedon in thought.(See Grey and Tall [1994] for laterreportson this never-
ending proceduralbehavior.)
The preceding analysis enabled me to define the phenomenon I wanted to
study. It concerned the variation of ways in which to experience a numerosity
larger than three when the numerosityis presented in a subtractionproblem as
the part of the whole numberthat constitutesthe missing or lost part.
I will subsequentlydenote this missing part"thepuzzling part,"since to me it
was very confusing that preschoolerscould become aware of the numerosityof
this part without any need of the laborious double counting that seemed to be
necessary for the pupils with mathematicsdifficulties. How children became
aware of the numerousityof the puzzling part was the researchquestion that I
wantedto answer throughmy study.
To present this phenomenonto children,I needed to develop interview ques-
tions thatinvolved subtractionsof the missing addendkind and of the take-away
kind. The puzzling partin these questionsshouldbe largerthan3 and thusbe dif-
ficult to perceive intuitively. The whole numbershould not be largerthan 10 or
smallerthan7, preferablyclose to 10. (If the whole is too small, it might even be
possible to subitize the numerosityof the puzzling part.)
Four subtractionproblemsof this kind were formulated:
1. Your teacherhas 3 pencils, but there are 7 childrenwho all want to write.
How many more pencils does she have to fetch?
2. If you have 10 pencils in your rucksackand lose 7 of them, how many do
you have left?
Two additionproblemsfor which the addedpartwas the largerone were also
given to compare ways of experiencing addition and subtractionamong the
school beginners.
I also decided to let the children take part in a guessing game in which they
could make five guesses of how nine buttons were hidden in two boxes. They
counted the buttons themselves and gave them to me before I hid them. The
guessing game was given first in the interview, and it was primarilyintendedto
catch the children's interest.It did not always confrontthem with the phenome-
non, but it was still a good problembecause five answers were given to the one
question.
The Contextual Analysis
The phenomenographicanalysis is seen as a process of exploration.It is "con-

textual"(Svensson, 1989), going back and forth in a sort of hermeneuticspiral,
where the partsare interpretedfrom an understandingof the whole andthe whole
from a closer analysis of the parts.The analysis in my study began in the inter-
view situationand then continuedas one after another,the tape-recordedinter-
views were listened to, transcribed,and readseveraltimes. A diffuse, global idea
about the phenomenonbegan to take form duringthis process. That global idea
was subsequentlydifferentiatedinto its constituentparts, and a more detailed
analysis was begun when enough transcribedinterviews were at hand. This
analysis was carriedout throughtwo coinciding interpretativeprocedures.In one
of them, all the answers to one question at a time were analyzed, and the indi-
viduals to whom they were relatedwere disregarded.The intentionwas to iden-
tify different ways of experiencing the phenomenon.In mutual interplaywith
this procedure,a second procedurewas carriedout wherebythe ways of experi-
encing were relatedto each individualin an attemptto observe whetheranswers
to several questions expressed one and the same way of experiencingthe phe-
nomenonwithin a single interview.Tentativecategories were formed and given
tentative names. The characteristicsof each category were describedin a sum-
maryin which excerpts from the transcriptswere included.The intertwinedpro-
cedureswere carriedout severaltimes. At the beginning,the interpretationsoften
were overlappingor were suddenlyseen as wrong.Thus, one interpretationgrad-
ually succeeded another,with the summarieschangingaccordingly.An interpre-
tative analysis of this kind can never be said to be definitely finished. Still, at
some point, the categories are seen as satisfactorilystable, and the researcher
decides that they can be presented.
THE MODEL REPRESENTINGTHE OUTCOMES

After the analysis is completed, an attemptis made to let the categories form
a model that, accordingto the researcher'sinterpretations,depicts the phenome-
non as experiencedin the collective awarenessof the individualswho took part
in the study. These categoriesreflect what the phenomenonseems to be to those
interviewedas well as how this "what"seems to be experiencedby them. It is
often possible to relate the categories hierarchically,orderedaccordingto their
graduallygreater inclusiveness. Categories describing more inclusive ways of
experiencingthe phenomenoninclude well-integrated,still functionalaspects of
less inclusive ways. The model, thus, often picturesa kind of evolution. Yet this
does not mean thatit envisions the orderin which children'smentaldevelopment
DagmarNeuman 71
of some cognitive competenceoccurs. It is thephenomenonappearingin its vari-

ation of ways that is picturedin the model.
The model as pictured in Figure 5.3 presents my interpretationof the ways
in which the interview children experienced the phenomenonunder study. The
prenumericalcategories in the original model have been omitted from Figure
5.3 because the purpose of this chapteris to illustratephenomenographicmeth-
ods rather than to present complete research findings. In Figure 5.3 the four
categories (written in lower-case bold letters), extents,finger numbers, word
numbers,and abstract numbers,depict what the whole numbers,including the
puzzling part, seemed to be to these children. The two superordinatecategories
(written in upper-case bold letters), numerically unstructuredand numerically
structured, depict how the whole numbers in the interview problems were
experienced.
Math difficulties Additive structures Multiplicativestructures
~I tSTRUCTURED
Abstract numbers < > Abstract numbers
Transformed Transformed
Structuredby biggestfirst Structuredby
|I <,___ Multiplesof 2 and 3
Word numbers
Structuredby biggestfirst
Finger numbers
Transformed
Structuredby biggestfirst
(Withthe help of the
Undividedfive)
UNSTRUCTURED
Word numbers > Finger numbers < Extents
Counted Counted
Estimated Estimated Estimated
Limited Limited Limited
Figure 5.3. Ways of experiencingthe phenomenonamong the 7-year-oldSwedish school beginners.
The categories written in lower-case italics depict more specifically how the
numerositywithin the puzzling part was experienced.When the whole number
was numericallyunstructured,the children seemed to experience this numerosi-
ty as limited, estimated, or counted. The answers to the interview questions in
these situationswere hardlyever correct.However, when the whole numberwas
perceived as numericallystructured,the answers were correct,and the numeros-
ity of the puzzling part of the numberappearedto be perceived intuitively. In
these situationsthe part-part-wholepatternof the numberappearednumerically
restructuredby a biggestfirst structure,transformations,or multiplesof 2 and 3.
It is not the purposehere to describe the model in detail but ratherto present
informationto illustratethe natureof a phenomenographicmodel and the use of
the criteriafor intuitive awarenessmentionedby Sandberg(1996) in reporting
this model. The following discussion provides more informationof the "what"
and "how"aspects of the categoriesgiven in the unstructuredand structuredsec-
tions of the model. The informationdescribes the variationin the way the phe-
nomenon is experiencedratherthan presentsan explanationof why these expe-
riences appearthe way they do.
The Four "Whats"

An analogueexperienceof numberakinto the one picturedin Figure5.1, or its
concretecounterpartin the form of finger configurationsthat I have called "fin-
ger numbers,"always seemed to be a part of the interpretationframe through
which the childrensaw the problems.When numbersappearedas extents,the chil-
drendid not seem to pay attentionto the single units in this "picture."The size of
the whole extent could only be experiencedordinallywith the help of the count-
ing word relatedto its last object. Wordnumberswere also experiencedthrough
the position of the last countingword. Yet, here the childrenillustratedthat they
paid attentionto each single unit-to each countingword they enumerated.
Childrenpaid attention to each single unit also when finger numbers were
used. A finger configurationcategorizedas a finger numberbegan with one hand
and continued with the thumb, forefinger, middle finger, and so forth, on the
other hand. Thus, a finger numbercould be simultaneouslyexperienced in an
ordinalway by the position and name of the finger where it ended and in a car-
dinal way by the "5 + some number"configuration.Mostly, but not always, fin-
ger numberswere put up by the childrenwithout any need of counting. Several
children did not use their fingers in a concrete way but, after an immediately
given correctanswer,illustratedhow they had "thoughtwith their hands."
For children experiencing numbersas abstract numbers,the numerositiesof
both partswithin the numberwere immediatelyexperiencedin an abstractway,
each partand also the whole relatedto one single symbol. Yet some kind of ana-
logue experienceof numberakin to the one picturedin Figure5.1 still seemed to
constitutethe backgroundof the words they used to communicatetheirthoughts.
The Three "Hows" When the Numbers Appeared as Numerically Unstructured
When, accordingto my interpretation,the situationdescribedin the problem

was of focal attentionin the children'sawareness,the part-part-wholepatternof
the numberappearednumericallyunstructured.The puzzling part was the last
one (as in Figure 5.2)-the partthat never can be experiencedordinallythrough
the counting word relatedto its last object.
The earliestidentified way of experiencingand communicatingthe numerosi-
ty of the puzzling partI have called limited. Here, the childrenused the number
word at the end of this partwhen they communicatedits numerosity-probably
Dagmar Neuman 73
to tell me where the partended. They did not yet seem to be aware of any con-
tradictionin the use of the same counting word for whole and part.They could,
for instance in the guessing game, repeatedlyguess that there were 9 buttonsin
one of the boxes, but still some buttonsin the otherbox, or say that therewere 7
pencils missing in Problem 1 (3 pencils and 7 children)and 6 left in Problem2
(10 pencils, 7 of them lost). In Problem2, they seemed to think of the lost pen-
cils as the ones relatedto the words ten, nine, eight, and seven. Thoughtof back-
ward, the word seven denoted the limit of the part that should be taken away;
thoughtof forward,the word six denotedthe limit of the partleft. Maryand Joan
exemplify this early ordinalway of experiencingProblem2:
Mary: Seven ... six, five, four, three, two, one ... six left.
Joan: ThenI've got one,two,three,four,five, six left.
The puzzlingpartheremainlyseemedto be experiencedin an earlyordinalway;
the focus was put on the point in the counting-wordsequencewhere it ended.
In the category I have called estimated, on the contrary,the puzzling part
seemed to be experiencedin an early cardinal way. It was still an experience of
an extent or of a manifold, but now it was the extension in between the limit
words that seemed to be of focal attention,not the limit words per se. Emma,for
instance, solves Problem2 (10 pencils, 7 of them lost) in the following way:
E: Then I've got four left ... or two ... four or two ... you can't be sure....
I: No ... can'tyou workoutin somewayif you'vegot fouror two left...?
E: Maybe three ... I think it's two.
I: Butisn'tthereanywayof workingouthow...
E: No.
I: Couldn'ttherebe eightleft?
E: Eightleft!!??
I: Whyisn'tthatpossible?
E: Well,becauseI've dropped... well if you lose thatmuch,it can'tbe thatmuch!
Emma seems to be well awareof the relationbetween the two parts,but as she
explains, there is no way for her to work out the exact numberof pencils left-
somethingthat does not seem to botherher much.
One type of answer was categorizedas a counted finger number.This way of
experiencingnumberwas displayed by only one child in one of his answers (to
Problem3). After the finger numberwas displayed,this boy countedthe last part
of it with the word one related to the third finger, the word two related to the
fourth, and so on. Thus, the fingers were what constitutedthe analogue experi-
ence of the whole number,and the words were used to count the last partof the
finger number.
Anothertype of answer was categorizedas a counted word number.Here, the
words constitutedthe analogue experience, and the fingers were used to count
the last partof the word number.This way of experiencingnumberswas the one
observed in practicallyall the answers given by the pupils displayingmathemat-
ics difficulties.
Numericallyunstructuredword and finger numbersappearedrarely,and when

they did, their puzzling partwas only rarelycounted. I interpretedthese "count-
ed" ways of experiencing numbers as being transitionalbetween numerically
unstructuredand numericallystructuredways of experiencing numbers.These
ways seemedto be inventedwhen childrenbeganto understandthatadultsexpect-
ed exact answers to mathematicalproblems, but when they still did not know
aboutany, or only aboutsome, of the ways in which numberscan be structured.
The Three "Hows" Whenthe NumbersAppearedas NumericallyStructured

The three ways of experiencingnumbersas numericallystructuredhelped me
understandhow children could avoid double counting, even when they under-
stood thatthe problemscould, and should, have "exact answers."These ways of
experiencingnumberswere qualitativelydifferentfrom the unstructuredways of
experiencingthem.
Two of the numericallystructuredways of experiencingnumbers-the ones cat-
egorized as biggestfirst and transformations-could be observedwhen the solu-
tion was concretelyformedby finger numbers.The "biggestfirst"idea was then
formedwhen childrenfocused theirattentionon the first handas constitutingone
single object-an "undividedfive"-instead of seeing it as five single unit items.
When the first handwas kept undivided,the biggest parthad to be relatedto this
hand.Thus, as soon as the biggest partwas five or larger,it became the first part.
To solve Problem2, for instance,the childrenput up all fingers and then fold-
ed seven before answeringthree. To solve Problem 3, nine fingers were put up
and the ninth and the eighth fingers were folded. Susie, solving Problem3 (two
pencils and nine children),exemplifies how the "undividedfive" could appearin
the children'sresponses while they formed a finger number:
Susie: First I sort of counted five ... then I put up two ... then I put up two more ...
thenI putthemdown[thelasttwo fingers].
We can see how the two parts in the analogue experience picturedin Figure
5.2 become swappedwhen the numbersget the biggest first structure.Then there
is no "puzzlingpart"any more.
Peter, Caroline, and Charles illustratethe biggest first structureas related to
word numbersand abstractnumberswhen solving Problem 2 (10 pencils, 7 of
them lost). All three children had answered"three,"just before they explained
their thinking:
Peter: [Quietlyandrapidly]One,two, three,four,five (inaudible)... andtheneight,
nine,ten.
Caroline:Seven ... eight, nine, ten.
Charles: 'Cause if you have ten and drop seven ... then ... you have ten ... dropseven ...
well ... you can count seven plus three ... then it's ten.
Instead of counting seven steps backward, or subtracting7 from 10, Peter
counts all and Carolinecounts on while they easily subitizethe words eight, nine,
ten as a "threeness."Charlesthinks of abstractnumbersandjust adds 3 to 7.
DagmarNeuman 75
Whereasthese three childrenillustratehow problemsexperiencedas subtrac-

tive sometimes have to be thought of forward, when the problem becomes
numericallystructuredby the biggest first experience, Ann and Andy illustrate
how additively experiencedproblems, such as Problem 3 (two pencils and nine
children), sometimes have to be thought of backward. Both children had
answered"seven"before they explained their thinking:
Ann: One, two, three,four, five, six, seven, eight, nine, ... andthen I thought... count
to nine ... seven ... and then I said it!
Andy: I thoughtlike this ... nine ... and then I took away two from the nine children
whohadgot thosepencils... andthenthereweresevenleft.
Andy seems to use some kind of egocentric speech in his explanation,but it is
evident that he has subtracted2 from 9. No double counting of the last part is
necessary when the numberhas this biggest first structure.
Transformationsof one numbercombinationto anotherresultedin this same
intuitive and immediate experience of parts and whole of the number.
Transformationswere also identified when they were concretely formed by fin-
ger numbers.In Problem3 (threepencils and seven children),for instance, none
of the partswas as large as five; the first hand, thus, could not be undivided.For
this problem, the children often moved the thumb from the first to the second
hand-or sometimes the thumbplus the forefinger-to transform5 + 2 into 4 +
(1 + 2) or into 3 + (2 + 2). Transformationsof finger numberswere also common
in the guessing game and in Problem4 (fourpencils and ten children).Often the
answers to these problems were given directly. The children did not illustrate,
before they were asked to explain their immediateanswer, that it was relatedto
their imagining a transformedfinger number.
Transformednumbercombinationswere also frequentlydisplayed when the
numbersappearedas abstract.For instance,the most frequentlyused correctway
to solve Problem 4 (four children and ten pencils) was to perceive the number
combination6 + 4 as transformedfrom 5 + 5.
Numbers structuredby multiples of 2 and 3 were expressed in guesses like
"Fourin this box and five in that,"with the explanation"'cause four and four's
eight." After the immediately given answer "four"to Problem 1 (three pencils
and seven children), the children gave explanations of this kind: "I know...
'cause three and three's six."
Validityof the Model

In phenomenographicresearch,validity is related to the intelligibilityof the
outcomes. These outcomes are judged partly with respect to the collected data
and partlywith respect to the model depictingthe outcomes. On the one hand, if
the categories describing different ways of experiencing a phenomenoncannot
be understoodas parts of a structuredwhole constitutingthe phenomenon,the
model is not intelligible. On the otherhand, if the seed of latercategoriescan be
found in earlierones, and if it is possible to see transformationalforms between

the categories,then the structureis consideredto be intelligible.
If the model can be used to interpretdatapresentedby otherresearchers'find-
ings or to interpretphenomena identified historically or in other cultures, not
only the phenomenographicmodel but also the other observationsappearmore
intelligible. For example, some of the categories I found have also been identi-
fied by other researchers.The idea to choose the most convenient way-count-
ing forwardor backwardin subtractionof all kinds-has been observed being
used by most primaryschool children(Resnick, 1983). Resnick namedthis way
of handling subtraction "the choice model" (p. 119). Interestingly enough,
Resnick (personalconversation)had mainly observed this model in a Swedish
study carriedout by Svensson, Hedenborg,and Lingman (1976). Unstructured
word numbershad also been described previously by Steffe, von Glasersfeld,
Richards,and Cobb (1983), who referredto them as "verbalunit items."Further,
Russell and Ginsburgh(1984) had shown thatthe primedistinctionbetween chil-
dren who find mathematicsdifficult and those who do not is that the former
groupof childrenseem to know fewer basic numberfacts.
The importanceof the "undividedfive" had been reportedlong before I found
it in my study, but related not to finger numbers but to the tiles that young
Japanesechildrenuse in the first grades(Hatano,1982). Brissiaud(1992), study-
ing his son, has pointed out the importanceof finger numbersstructuredby the
hand (five) in children's early acquisitionof numbersense. These "fingersym-
bol sets" (p. 41), he underlines,functionas an analoguenumericalreferencesys-
tem of an ordinal-cardinalkind that can be used without any need of counting.
Brissiaudsees an analoguesystem of some kind as being "a necessaryunderpin-
ning for the constructionof a logico-mathematicalsystem, such as number"(p.
65). Werner(1973) put forwarda similaridea more than 50 years earlier,stating
thatthe hand seems to representa naturalnumberarea.His researchindicated"a
definite relationshipbetween the ability to articulatethe fingers and the early
developmentof numberconcept"(p. 296).
A phenomenographicmodel can also be validatedthroughhistoricalcompar-
isons. The early Chinesepeople, for instance,placed a single horizontalrod, pic-
turing the undivided hand, above vertically placed rods, picturingthe fingers.
Similarlythe RomannumeralV picturedthe hand.If the left line in V is thought
of as representingthe fourfingers,andthe rightline, the thumb,all part-part-whole
relationswithin the numberrangeVI-X (VIIIII)(with the exceptionof 31316)can
be experiencedintuitivelyby using two of the ways of structuringdisplayedby the
school beginners-the biggest first structure,createdby the undividedfive, and
transformations carriedout by moving one thumbbetweenthe hands(Figure5.4).
Pragmatic Validity
Pragmaticvalidity, as described by Kvale (1989), provides another way to

defend the outcomes of qualitative research. This type of validity can be
Dagmar Neuman 77
VI VI VIII VIIII Vlll I

VI V II VI II VIIII VIIIII
3+3 V II V III VI III VIIIII
VIII V ll VI II11
V 11111
Figure 5.4. Roman numeralsVI-X structuredby the undividedfive and transformations.
obtained,for instance,throughputtingthe resultsof an interviewstudy into prac-

tice. For example, see the teaching experiments of the phenomenographic
researchersAhlberg (1992), Lybeck (1981), Neuman (1987, 1993, 1994, 1997),
and Pramling(1991, 1994). I finish my discussion of phenomenographicmeth-
ods by briefly describinghow I used pragmaticvalidity to test the validity of the
outcomes of my study.
The interviewstudy gave me severaluseful ideas for my teachingexperiment.
Since experiencingnumbersas extents was the most frequentlyreportedway of
experiencingnumbersin my study,one idea was to introducemathematicsthrough
measuringinsteadof throughcounting.The intenthere was to help all childrensee
the extents as being divided into units. A second idea was to avoid the use of the
countingwords to begin with, since these words seemed to be used in ways that
would not be takenas sharedby all pupils. A thirdidea was to introducesubtrac-
tion before addition,since in subtraction,childrenknow aboutthe whole number
and have, especially if they form finger numbers,the possibilityto become aware
of the two partswithinthis whole withoutany need of doublecounting.
To create situationsin which the childrencould get these kinds of experiences,
the mathematicsclassroom became a fantasy land-the Long Ago Land. In this
land there were no counting words, no measures, no coins, no mathematicsat
all-but problems continually appeared for which mathematics was needed.
Thus, the children were required,bit by bit, to create together the mathematics
they needed for the moment. As they invented measures of differentkinds, for
instance, to comparetwo quantitiesof liquid, pouring one after the other of the
units into a bucket,they quickly discoveredthatthey needed tallies of some kind
to rememberhow many units had been poured. The fingers were proposed as
useful tools by most children,but some thought that strokes drawnin the sand
(thatis, on the chalkboard)could also be suitable.When discretequantitieslater
were recorded,the strokessoon became too many to subitize or representby fin-
gers. At this, we told the childrenthat the people in the Long Ago Land record-
ed handsand fingers as well as just fingers, and the symbol V was introducedfor
the hand.Thus, analoguenumberrepresentationbecame structuredby the "undi-
vided 5" before the childrenhad a reason to invent the counting words.
This is not the place for describingthe teaching experimentin detail. It might
be said, however, thatin the interview study carriedout afterthe second year, all
childrenin the two researchclasses could add and subtractwithin the range 1-10
over the 10 limits without any need to count or keep track.This was very differ-
ent from the ways the childrenin the control group solved these kinds of prob-
lems. The outcomes of the interview study could, as suggested by Martonand
Booth (1997, p. 81), be used as "a notational path of development foci for
instruction."
CONCLUSION
The methods presentedin this monographare different, as are the goals and
underlyingperspectivesof each author'sresearch.This variationmakes the pic-
ture of what mathematicsteaching and learning mean and of how they can be
researchedrich and detailed.However, it also createsdifficulties in our attempts
to formulate criteria for what counts as acceptable qualitative research. For
instance,at the ICME-8conferencein 1996, suggestionswere made to formulate
a set of researchguidelines. Many participants,however, saw this as placing a
straightjacket upon the researcher,preventingcreativity and excluding certain
kinds of research.
To describe fundamentalassumptions,goals, methods, and knowledge (valid-
ity) claims for a given researchstudy, as was done in this chapter,would help us
judge individual studies from within their respective methodological frame-
works. I hope that such descriptionscan serve as a step towardcreating a con-
sensus for judging the quality and acceptabilityof qualitativeresearchin mathe-
matics education.
79
Chapter 6
Working Toward a Design for

Qualitative Research
Susan Pirie
This chapter can be considered a response to the call by Merton (1968)

for the practice of incorporatingin publicationsa detailed account of the ways in
which qualitative analyses actually developed. Only when a considerablebody of
such reportsare available will it be possible to codify methods of qualitativeanaly-
sis with something of the clarity with which quantitativemethods have been articu-
lated. (p. 444)
If this is true for the sociologist, how much more important it is within math-
ematics education, where many of the practitioners of research originate in the
enumerative discipline of mathematics.
Quantitative and qualitative methods are not alternative paradigms for the
same research activities. Each has much to offer, but what is offered and what
constitutes the goals of any project must together guide the choice of methodol-
ogy. Is the intention to build or to test theory? To survey an issue or look at it in
depth? To look at large quantities of data for similarities and to abstract these
because of their general applicability and hence assumed essentiality, or to look
at individual cases and abstract essential features, generalizing them because of
their perceived vital nature? It is not my intention to argue here for one paradigm
over the other in a variety of given situations. Rather, I set out my reasons for the
methodologies I have espoused in my own research and the processes by which
the decisions on method and methodology were made.
I am offering not a textbook on a particular mode of conducting research but
an account of the reality of selecting methodology and method for one specific
project. The reader will be taken through the decision making that must accom-
pany any research and be shown the need for a robust research design that will
accommodate the slings and arrows that outrageous fortune throws at intrepid
researchers, be they novice or not. Experience offers insight into potential prob-
lems, but care and a secure theoretical base for the design of the research offer
the surest way to bring the undertaking to a satisfactory close. The structure of
This account of the explorationof discussion between pupils in the mathematicsclass-

room would not be complete withoutmy expressionof loss and gratitudefor the work done
by Rolph Schwarzenberger,whose illness and deathtragicallycurtailedthe invaluablecon-
tributionhe made to the project.
80 WorkingTowarda Design for QualitativeResearch
this chapteris based on the questionsthat must be asked of any researchand the
illuminationof these questionsthroughdiscussion of how they were tackled and
answeredwithin one specific researchproject.
The researchtopic comes first, and the methodologyand methodsmust be not
just "taken,readymade, off the shelf' but be chosen and adaptedcarefullyto best
access the answers sought (Burgess, 1984, p. 4). My researchexemplifies the
philosophy of theory as process:theory changing,evolving and being construct-
ed to fit or explain new cases. One focus of my interest-and it is this that I
examine in detail in this chapter-is the phenomenon of discussion between
pupils in the mathematicsclassroom.My intentionis to discuss the processes of
selecting the appropriateresearch methodology and methods. I then illustrate
how, throughthe examinationof purposefullygathereddata,we sought features,
properties,and categories within the phenomenonthat would enable or enhance
the positing of theoriesrelatingto the effects of discussion on pupils' mathemat-
ical understanding.
RESEARCHQUESTIONS
What Provoked the Original Research Question?

The research project titled "Discussion: Is it an aid to mathematicalunder-
standing?"was initiatedby Rolph Schwarzenbergerandme as a resultof a report
by a British governmentcommittee of inquiryinto the teaching of mathematics
in schools (Cockcroft, 1982) and the general reactionof teachersto that report.
The reportclaimed in paragraph243 that "discussionbetween teacherand pupils
themselves"was one of the six elements that should be includedin mathematics
teaching at all levels. As a result of this highly readablereport,teachersin great
numberstook this to mean thatdiscussion was per se a good thing, withouteither
questioningthe meaning of discussion or querying what it could achieve in the
mathematicsclassroom. The report gave the following reason for the recom-
mendation:
Theabilityto "saywhatyoumeanandmeanwhatyousay"shouldbe oneof theout-
comesof goodmathematics teaching.(p. 72)
This implies that mathematicstrainsone to talk clearly (which may be a wor-
thy general educationalgoal) but not that discussion aids the understandingof
mathematics.
In fact, the Committee itself had been unable to find much evidence of the
existence of discussion. There existed at thattime considerableresearchinto the
effects of discussion on learning (Barnes & Todd, 1977; Edwards & Furlong,
1978; Mehan, 1979), and more particularly,discussion between teachers and
pupils (Barnes, Britton, & Torbe, 1986; Mehan, 1985), but nowhere could we
find any specific research to support the value of discussion between pupils
themselves (Pirie & Schwarzenberger,1988a). It was, therefore, on this latter
areathat the researchprojectfocused. Schwarzenbergerand I wanted to explore
Susan Pirie 81
the notion and natureof discussion between pupils in mathematicsclassrooms

without the influence of the teacher's presence. We intended to look for links
between such discussion and the growth of pupils' mathematicalunderstanding.
In other words, we hoped to answer this question: "How does pupil-pupildis-
cussion influence understandingof mathematics?"
The recommendationsof the Cockcroft Committee were based on the notion
thatin general talkingto othershelps to clarify one's ideas and focus one's think-
ing. We believed, however, that there was reason to perhapsquestion the gener-
alizability of this notion to mathematics. Our concern was that a mismatch,
unique to mathematics, exists between the written and the spoken languages
used. One of the strengthsof mathematicsis thatthe symbolic languagein which
it is written is brief, precise, and unambiguous-for the initiated!For learners,
one problemis thatthe languageused to verbalizethese symbols can be lengthy,
multiple,and ambiguous(Pirie, 1991a, 1996a, forthcoming).Ourcontentionwas
that the imprecisionof talk in a subjectdependentfor its power on the precision
of its symbolic language might radically alter the effect of discussion on learn-
ing in mathematicsfrom the effect expected in other disciplines.
WhatSecondaryQuestionsAre Raised?
Initial areas of research interest are generally broad and in need of sharper
focusing, both in terms of questions asked and of intendedoutcomes.
It became abundantlyclear from our first exploratoryclassroom visits that
there would be no simple answer to the research question we had posed, nor
indeed had we expected one. We observed incidents where discussion appeared
to advance the understandingof the pupils involved and where it quite definite-
ly confused and misinformedthe participantsand inhibitedprogress towardthe
solution of their problems. Our tasks, therefore, became first, attempting to
understandthe natureof pupils' mathematicaldiscussion, and second, finding a
way to assess its effect on their mathematical understanding (Pirie &
Schwarzenberger,1988a). It is with the first of these, the wish to understandbet-
ter the phenomenonof pupil-pupildiscussion before commenting on its value,
thatthis chapteris concerned.To this end, we were interestedin the whole range
of forms that such discussion could take, ratherthan, at this stage, in the typical-
ity or frequencyof aspects of the phenomenon.
For reasons of time and scale the projectfocused mainly on secondarypupils.
Two questions faced us:
1. Given our purpose,what methodologicalstrategieswould enable us to fulfill
our aims?
2. How could we obtain relevant data-data that would aid us in our specific
task?
We were concerned about producingadequatedescriptionsof the ways that
pupilstalkto one another.We neededto exploreways of analyzingsuch discourse
82 WorkingTowarda Designfor QualitativeResearch
and its contexts-ways that would offer insights into those processes inherentin
pupil-pupiltalking that might influence the growth of the mathematicalunder-
standing.We wished to generalize,thatis, to producea generalpictureof the phe-
nomenonof pupil-pupildiscussion,by abstractingfrom specificallygathereddata
those featuresthat recurand appearto characterizethis particularform of class-
room interaction.More precisely, we were interestedin the generalfeaturesthat
seemed to affect the growthof mathematicalunderstanding.
APPROPRIATEMETHODOLOGIESAND METHODS
Ethnography
Several general research paradigms offered us potential methodological
stances, which I lay out briefly and examine for theirpossible applicabilityto our
project. The first is the ethnographicparadigm,the interpretationof which has
taken various forms over the last few decades (Burgess, 1984; Wolcott 1975).
Ethnography,in its broadestsense, is concernedwith the socioculturalfeatures
of an environment;with how people interactwith each other;and with the rules,
the structures,and the processes of these interactions.In general, it borrows
methods of operation-data collection and analysis-from the discipline of
anthropologyand applies them to specific subgroupingsof people within a larg-
er defined group. We can talk of the "cultureof the classroom"and investigate
how, within this particularclosed environment,a culturedevelops that differs in
manyrespectsfrom the normallived experiencesof its participants.Ethnography
as defined by Wolcott (1975, 1982) involves the suspensionof one's own judge-
ment that is based in one's own culturalassumptionsand demandsthat one look
throughthe eyes of those who are themselves the membersof the cultureunder
scrutiny.The intentionis to illuminatean understandingof the culture,not to pre-
dict futurebehaviors.
The notion of ethnographywas seductive.At first sight it seemed thatit would
yield the in-depthexplorationthatwe sought.On closer inspection,however,two
crucialaspects made the adoptionof this paradigm,as our unadulteratedmethod-
ological choice, unsuitable.
The first concernedthe method of data collection. Participantobservation,so
often associatedwith ethnography,would in all probabilitypreventthe phenom-
enon that we wished to observe from coming into existence! It was talk away
from the presence of the teacher,but still within the cultureof the mathematics
classroom,thatwe wished to focus on, and it seemed highly likely that a knowl-
edgeable adult observerwould by her or his very presence alter the pupils' ver-
bal interactionswith one another.Interviewingthe pupils after a lesson on how
they had talked among themselves would also not be appropriatebecause it was
theirinitial discussion, not theirreflective reportingon it, colored by theirexpec-
tations of what an adult would deem important,that we wished to capture.The
second problemrelatedto the suspension of the researchers'culturalview. The
Susan Pirie 83
assessment of the effect of the discussion on the growth of pupil understanding

had to be from an external,mathematicallyinfluenced viewpoint. It was depen-
dent on detailed knowledge of the natureof the mathematicaltopic under con-
sideration. We did not wish to investigate what the pupils thought they had
learned,but what, from our observationsof their currentand subsequentactions
and talk, we judged them to have learned.
Notwithstandingthese criticisms, the general idea of the long-termgathering
of in-depthdata was an appropriateone. We were going to need to listen to spe-
cific groups of pupils discussing their mathematics,over the length of time they
devoted to learningseveral particularmathematicaltopics, to come to know the
ways they interactedverballyandthe effects these interactionshad on theirlearn-
ing. We were also going to need to catch this talk in a way that would not be
influenced by the listener. After much deliberation,we decided against video-
taping as the method of data gatheringbecause at that time we consideredit too
intrusive.We settled for audiotaping,and we explainedto those pupils whom we
wished to recordthat they were not being tested in any way, that their teachers
would not hear the tapes, and that we were just interestedin how pupils talked
about mathematics.We focused on three or four small groups in each class, and
we recordedevery lesson on each particulartopic, which meant that we would
usually be with the class for two or threeweeks at a time. An audio recorderwas
placed on the table for each groupof pupils, and the researcher,sitting across the
room so that the pupils would not feel that they were being watched, took field
notes on the lesson in general. These notes-on what was writtenon the chalk-
board,for instance, or on the relevantactivities of otherpupils or the ostensibly
importantactions of the pupils being audiorecorded-were made on time sheets
that could be later coordinatedwith the tape recordings.
Ethnomethodology
The focus of our interest was thus talk-a specific kind of talk it is true, but
talk nonetheless-and the ways in which such talk was constructed,the language
that was used, and the types of verbal interactionsthat took place. In fact, we
were looking for some methodof categorizingthe specific pupil-pupildiscussion
that we hoped to trace. Otherresearchfocused on the analysis of talk was, there-
fore, a likely place to look for methodological strategies.Ethnomethodologists
and those workingthroughconversationalanalysis concernthemselves with talk,
treatingit not as a resourcefor informationon some other topic but considering
it the object of study itself.
This emphasis on the talk itself appearedto come close to some of our think-
ing. To understandteachingand learning,of which talk is a natural,integralpart,
we must understand the talk, its coherence, its structure, and its context.
"Conversationalanalysis has developed a conceptual machineryfor unraveling
the organization of conversation so that it may be described and analyzed"
(Hitchcock & Hughes, 1992, p. 162). Partof our aim was to describe the nature
of pupil-pupildiscussion in the mathematicslesson, but we wanted to do more

than this. We were, ultimately,still concernedwith the results of particularver-
bal interactionsin termsof the understandinggained. Jacob (1987) describesthe
more specific "ethnology of communication,"but although talk among pupils
can be the focus within such a paradigm,the aim of the researchis againto exam-
ine speakingas an activity in its own right, not as a process of conveying mean-
ing. There is no significance placed on the, say, mathematical,meaning of talk.
One is looking at pupils' speakingas a cultural,social construct.
Sociolinguistics
Having, with some profit, surveyedthe field of anthropology-relatedmethod-

ologies, we turned to that of linguistics. The narrow dissection of words and
meanings was not of relevance to us, but a sociolinguistic approachmight be
appropriateto our quest within the field of educationalresearch.Sociolinguists
concern themselves with the wide diversity of linguistic forms and speech pat-
terns that occur within communicationthroughlanguage.Their aim is to unpick
the featuresof these patterns,to discover how social factors affect the develop-
ment of such patterns,and to explore the effects of specific speakersand hearers
and contexts on such patterns.Their interest lies in how these social environ-
ments affect verbal interactions (for examples of this approach, see Barnes,
1982; Sinclair & Coulthard,1975). Here, too, lay our interests. What was the
effect of the classroomcontext on the languagethe pupils used when talkingwith
their peers? But more than that, what was the effect of the interactionon their
understandingof mathematics?
Sociolinguistic and ethnographicapproachescan both be used to investigatea
concern with what people are doing when they are making sense of verbalinter-
actions. Both approachescan be concernedwith looking at patternsand regular-
ities in the interactionsor, indeed, with searchingfor the irregular.By examin-
ing the possibilities offeredby researchmethodsdrawnfrom both paradigms,we
hoped to be able to constructmethodsof datacollection and analysisthatallowed
us to look not so much for patternsin the speech itself as for categories of ver-
bal interactionsrevealedby the speech.
DATA COLLECTIONAND ANALYSIS

What Will Constitute the Data?
Before proceeding further,we needed to be more precise about what would
form our primarydata. Given the way we were intendingto work in the class-
rooms, we would be dealing both with a complete audio recordof each lesson as
it applied to each group we were consideringand with field notes that could be
collated with the recordings.What we chose to do with all this raw information
would directlyaffect the methodsof analysis availableto us. Two considerations
stood out as important.First,it was clear thatwe were not interestedin analyzing
Susan Pirie 85
in detail every utteranceof every child throughoutthe entire period of data col-
lection! Ourfocus was discussion-a task thatwe had not previouslyundertaken.
We needed to producea definitionthatfit with the ideas impliedin the Cockcroft
report,but was preciseenoughto use for identificationpurposeswhen we listened
to verbal interactionsbetween the pupils. We producedthe following working
definition,which in fact remainedunchangedthroughoutthe rest of the research.
Discussion is-
*purposefultalk.Therearewell definedgoalsevenif noteveryparticipant
is aware
of them.Thesegoalsmayhavebeenset up by the groupor by the teacherbutthey
are,implicitlyor explicitly,acceptedby thegroupas a whole.
* on a mathematical subject.Eitherthe goalsthemselves,or subsidiary[goalsthat]
emergeduringthecourseof thetalking,areexpressedin termsof mathematical con-
tentorprocess.
* in which there are genuinepupil contributions.[Thereis] inputfrom at least some
of thepupilsthatassiststhetalkingor thinkingto moveforward.We areattempting
hereto distinguishbetweenthe introduction of new elementsto the discussionand
merepassiveresponses,suchas factualanswersto a teacher'squestions.
* and interaction.[Thereis indication]thatthe movementwithinthe talkhas been
pickedupby otherparticipants. Thismaybe evidencedby changesof attitudewithin
thegroup,by linguisticcluesof mentalacknowledgement, orbyphysicalreactions that
showthatcriticallisteninghastakenplace(Pirie& Schwarzenberger, 1988a,p. 460).
The second considerationwe took into accountwas that much of the meaning
of a live interactioncan be lost when the talk is reducedto a transcript.Take, for
example, the utterance"It's four over three."As written,it appearsto be simply
a statementof the improperfraction four-thirdsas the solution to some given
problem.With an emphasison the word over, however, the implicationmight be
a contradictionof some previous action or expression;in other words, that divi-
sion by three is the correct process, not, say, multiplication as previously
thought.Alternatively,spoken with a rising inflection and a suitably astonished
facial expression, it could be expressing total incredulitythat someone would
think that "fourover three"could possibly be the answer!
At the end of each lesson, therefore,the observerlistened to the tape record-
ings and amalgamatedthe field notes with a detailed summaryof the pupil-pupil
interactions.All occasions of pupil-pupiltalk thatfit our definitionof discussion
were noted, and for these and othercentralincidents,both the time of occurrence
and the tape-counternumberwere recorded.We did not at this point transcribe
the tapes because intonation,pauses, and otheraudibleactivities were considered
importantto the categorizationof the talk. In fact, we decided that these notes
and recordingswould form the data from which we would performall our sub-
sequent analyses. We transcribedportions of the tapes only for the purpose of
writingaboutthe researchfor an outside audience.We did, however, make a sec-
ond copy of every tape in case of accident! It is importantto make overt this
explicit decision about the nature of the data used because it is from this per-
spective that results must be judged.
What Are Appropriate Methods of Analysis?

A method of analysis was needed that would look at all the incidents of dis-
cussion that occurredwithin the data. We were thus not attractedto any system
of categorizationthat dependedon a time-controlledsampling of the data. The
methodof "systematicobservation,"or "interactionalanalysis,"seemed to offer
one obvious way of dealing with the data. Basically, in these approaches,
responses or episodes are coded with a set of preselected categories (Flanders,
1970; Galton, Simon, & Croll, 1980). The choice of categoriesis inevitablysub-
jective, althoughit can have its basis in theory or previous research.The effect
of this methodis normallythatthe observerreturnsfrom the classroomwith only
a numericalrecordof the frequencyof the occurrenceof incidents in each cate-
gory. We had a far richer primarydata source-the entire recordings-but we
could have applied some form of interactionalanalysis to it. Our reason for
rejectingthis methodof workingwas, however, not due to potentialloss of data.
We were undertakingthe researchpreciselybecause we did not know what those
categories might be. We wished to extract the categories from the data, not to
impose them on the data (Glaser & Strauss, 1967).
A proven, powerful method whereby every episode of classroom interaction
can be categorized is that of discourse analysis (Sinclair & Coultard, 1975;
Stubbs, 1981). Under such a scheme, the whole lesson is broken down into a
hierarchyof episodes, and the individualepisodes can be furthercoded as to the
type of interactionthatthey typify. In its originalinception,the methodwas used
to deal with the form of the interactions,not the content,but it was a methodthat
held appealfor us and seemed to be adaptableto our needs.
We were able to allocate all the episodes of talk in our entireprimarydatacol-
lection to threebroadcategories:
(i) Talk clearly relatedto mathematics
(ii) Verbal exchanges that were incomprehensible
(iii) Social chat
Category (i) was then subdividedinto episodes that fit our definition of dis-
cussion and those that did not, and category (ii) was subsequentlyreanalyzedto
draw out those episodes that, although at first pass seemed incoherent to us,
clearly held meaning for the pupils (Pirie, 1991b). We were still faced with the
problem of how to categorize the episodes of discussion, and some element of
subjectivity seemed inevitable. Our goal was to be as open as possible to all
interpretationsof the episodes.
Barnes(1982) has done much to open up discoursein the secondaryclassroom
to the scrutinyof others.His snapshotobservationsof individuallessons, and his
analysisby meansof personalinsightfulcomments,have laid his workopen to crit-
icism from puristswho deplorehis overt subjectivity.What such detractorsmiss,
however,is the very realvalue of his reflectiveinterpretations andthe necessityof
just such an exploratoryapproachto pointthe way to fruitfulareasfor further,more
rigorousinvestigation.His focus is largelyon the role and the effect of the teacher
Susan Pirie 87
and on teacher-pupilinteractions,however, and offers little insight into the possi-

ble categorizationsof the phenomenonin which we were interested.Personal
insight and reflectionwere, indeed, the only ways in which we were going to be
able to startto come to grips with our data. Our aim was to be as systematicand
theoreticallyguidedas was possible in our choices of categories.
We looked to sociology for guidance.Ourcategorieswould be groundedin our
data. Ratherthan spring from our own preconceptions,they would evolve from
the data we gatheredby the process known as "systematictheoreticalsampling"
and be stabilized througha procedureof constantcomparison.Theoreticalsam-
pling is the process whereby data are repeatedly gathered and analyzed, with
each subsequentdata-collectiondecision dependenton the analysis of the previ-
ous data collected. The "process of data collection is controlled [sic] by the
emerging theory"and the "initialdecisions are not based on a preconceivedthe-
oretical framework"(Glaser & Strauss, 1967, p. 45). Througha constant com-
parisonof episodes, initial data are examinedfor features,trends,and properties
and are explicitly coded according to these categories; new data are gathered
specifically to illuminate or contradictthese categories and are systematically
comparedagainst all previous data;furtherdata are collected and analyzeduntil
the emergingtheoryor categorizationappearsto be stable.This methodof analy-
sis is inductive-it moves from data to tentativetheory, to new data, to refined
theory.As a methodof datacollection and analysis within the field of sociology,
it is well documentedin Glaser and Strauss(1967) and seemed ideally suited to
our intentions.Its strengthsare revealed by the ways we were able to deal with
practicalissues as they arose.
How Are Data to Be Gathered?
Turningto questions of practicality,how then should we set about gathering

the necessary data?We needed a system that was both flexible and systematic.
(See Burgess [1982] for a discussion of some of the issues related to sampling
thatarise in qualitativeresearch.)Randomsamplingof schools and classes made
no sense in our situation;we needed to observe classes where we had the maxi-
mum chance of encounteringpupil-pupil discussion. We were looking to talk
about neither "frequencyof occurrence"nor "typicalcategories"but about the
range of interactionsthat can be present. We had to collect examples of pupil-
pupil discussion to be able to categorize them; we alreadyknew that these were
hardto find.
Various methods were available whereby we could observe the phenomenon
of mathematicaldiscussion between pupils and its effects on understanding.One
techniqueopen to us was to take any groupof pupils, deliberatelyprovokemath-
ematical discussion among them, and record the event for subsequentanalysis.
We had no doubt that this would be both possible and interestingand would fil-
ter out some of the complexity and wealth of uncontrollablefactors inherentin a
classroom.We should, however, then be wary of the gap between how we would
idealistically like childrento learn and how they did learn in the realities of the
classroom. Ourconcernwas pupil discussion within a normalclassroom setting,
in particular,discussion unaffectedby the participationof a teacher.Hence, this
method was discarded.
We focused the initial observationphase on the classrooms of four secondary
teachers who consciously and deliberately used discussion as a part of their
teaching style (Pirie & Schwarzenberger,1988a). These teachers were, in fact,
hard to find, and their classes must be considered atypical. Were our observa-
tions, therefore,in dangerof being biased by the sample we were taking?On the
contrary,we intendedto base our theorizingon the essential featuresof the phe-
nomenonas they emergedfrom the data.Edwardsand Mercer(1987, p. 26) refer
to this biased samplingas "anintentionalconsequenceof [the] researchdesign."
We were interestednot in counting or comparingcases, but in examining and
categorizingincidents of discussion. The selection of our sample was based on
theoretical,not statistical,grounds;the validity of our findings flowed from the
evolution of the categories ratherthan from the representativenessof the study
population.
The selection of the pupils whom we would observe was also done to maxi-
mize our ability to observe pupil-pupildiscussion. The teachersidentified chil-
drenwho were in the habitof discussing theirwork with theirneighbors,and we
explained to them that we wished to audio record their interactionsduring the
next few lessons. The teachersidentified a variety of mathematicalexperiences
and topics from the normalteaching schedules that were likely to provoke dis-
cussion among the pupils, and these we observed.
One of the crucial featuresof theoreticalsampling is that furtherdata collec-
tion is guided repeatedly by the analysis of existing data until saturationis
achieved, that is, until the emergentcategories remainunchanged.For this rea-
son, althoughit is likely that any researchprojectwill have one main method of
data collection, there are "no limits to the techniquesof data collection, the way
they are used, or the types of data acquired"for the end purposeof illuminating
the phenomenon under examination (Glaser & Strauss, 1967, p. 65). Several
weeks afterthe initial classroomvisits, with a view to gaining deeperinsight into
the effect of discussion, we interviewed the pupils we had observed in their
workinggroupsto elicit theirunderstandingof the topics they had been working
on. We used a loosely structuredclinical interviewing technique (Ginsburg,
1981, 1983) whereby pupils talked their way througha task and the interviewer
followed theirpaths of thinking,sometimesreturninglaterto probe more deeply
into relevantkey ideas (Pirie, 1988).
How Is the Analysis to Be Performed?

A precise, systematic method of analysis was followed throughoutthis first
data-gatheringphase. The two researchersinitially observed the same classes.
Then, after the visit, they individually correlatedtheir own tapes and personal
Susan Pirie 89
field notes and categorized the data under the three headings previously dis-
cussed (mathematical,incomprehensible,and chat). Still withoutconsultingeach
other, we next identified episodes of discussion accordingto our agreed defini-
tion. Finally, we came together to comparethe kinds of things we had deemed
importantto record in our field notes, our individual categorizations,and our
interpretationsof the definition when we applied it to specific classroom inci-
dents. The purpose of this duplication of observation was to consolidate the
methods of recordingand the focus of the field notes to ensure as far as possible
that there was a common understandingof the method of observing and analyz-
ing. Subsequently,we visited separateclassrooms, following one teaching topic
through from initiation to close. At regular meetings, each observer presented
identified incidents of discussion for joint analysis, and each incident was then
coded both by the mathematicaltopic it concernedand by any notablefeaturesit
presented.Theoreticalsamplinglends itself well to collaborativeworking, since
it depends on drawingfeatures and potentiallymeaningfulcategories out of the
data gathered.What can be seen as a problemwithin a researchdesign, namely,
differentinterpretationsof the data, can be turnedinto a strength,since interpre-
tation by more than one person can lead to a richerfirst analysis.
The category labels for featuresthat seemed pertinentduringthat first analy-
sis included"usingmathematicallanguage,""abouttheirlack of understanding,"
"abouthow to do it," "reflectionon their mathematics,""proof,"and "focus on
the meaningof the mathematicalproblem."None of these categorieswas intend-
ed to be exclusive, and indeed we found episodes in which we were using mul-
tiple categorization,such as "aboutthe task,""recording,"and "usingnonmathe-
matical language"as the following example illustrates.
Four 12-year-oldgirls were tackling an investigation(Frogs) in which black-
and-whitecubes are moved by slides andjumps. Their task was to exchange the
places occupied by black cubes for those occupied by the white ones and to
recordthis in some way. They decided to count and recordthe numberof moves
made.
Susanne: I knowhow it works-you havethewhiteson one side.
Tracy: OK.Youhaveto havethewhitesa certainside,don'tyou?
Ann-Marie: I know!
Joanne: Howdo youreckonwe aregoingto recordthis?
Susanne: So thatwe canremember
it.
Tracy: I'll counthowmanymovesyou make.
Joanne: Come on-watch her.
Ann-Marie: Can't I jump that one? Can't I do that?
Joanne: Do it again and I'll count how many moves you make.
What we were beginning to see duringour discussions was the emergence of
threegroups of categories.The first of these was what it was thatgave the speak-
ers somethingto talk about.Within this groupthe episodes could be classified as
to whether(a) they had a task or concreteobject as the focus of theirtalk;(b) they
did not have an understandingof something,but knew this and thus had some-
thing to talk about; (c) they did have some understanding,which gave them
somethingto talk about.
The second group was concerned with the kind of language used-the focus
being on the languagein which the discussionwas conductedand not on the con-
tent of the statementsmade.
Again three classificationssuggested themselves. Those were (f) the speakers
lackedappropriatelanguage-they did not have the corrector useful words,(g) the
speakersused ordinarylanguage,(h) the speakersused mathematicallanguage.
It could be conjecturedthat the categorizationof language as "ordinary"or
"mathematical"would be somewhat arbitrary,since "mathematical"language
for young childrenmight be "ordinary"language for them a few years later. In
practice,however, viewing the discussion in the context in which the pupils were
workingenabledus to make decisions with little difficulty or disagreement.
The thirdgroup thatemerged was the kind of statementsthe pupils were mak-
ing. A varietyof statementscould exist within any one episode. These were clas-
sified as (p) incoherent-that is to say, interactionsthatfit all our criteriafor dis-
cussion but contained statementsthat were incoherentto us, the observers;(q)
operational,or in otherwords, aboutspecific (frequentlynumerical)examples of
mathematics;(r) reflective, which we subsequentlyrenamed"abstractive"as its
naturebecame evident more in terms of statementsof generalizationsof mathe-
matics than in terms of statementsreflecting on mathematics.
We reviewed all the datacollected and categorizedeach episode on each of the
threegroups.The example given above was categorized(a,g,q). The pupils were
talking aboutthe task of recording,using ordinarylanguage, and making opera-
tional statements.Furtherexamples of the use of this categorizationcan be seen
in Pirie and Schwarzenberger(1988a) and Pirie and Schwarzenberger(1988b).
The essence of the method of constant comparisonis the repeatedreanalysis
of existing data.Even before the decision was made to codify all the episodes in
termsof the threegroupsabove, when any new, interestingfeaturewas observed,
all the tapes previously discussed were rescrutinizedfor signs of that new fea-
ture.Althoughwe were happyto agree with Glaserand Strauss(1967, p. 30) that
"a single case can indicate a general conceptualcategory or property,"we also
wantedto capturethe richnessof each of the categoriesthatwe were identifying,
and so we looked for furtherexamples to roundout our descriptions.
An emerging problem was how to identify the growth of understandingand
relateit to the examples of discussionwe were observing.Althoughthe abilityto
talk purposefullyabout mathematicsis-as are also the abilities to write mathe-
matics and to solve problems-prima faci evidence of mathematicalunderstand-
ing, it is not necessarilythe cause of, or even an aid to, such understanding.At this
stagewe based ourjudgementaboutthe growthof mathematicalunderstandingon
the interviewdata, noted it, but did not do much in the way of theorizingon the
effects of the discussion for the following importantreason. It rapidly became
apparentthat the paucity of examples of discussion seen by the Cockcroft
SusanPirie 91
Committee was no accident-the incidence of discussion was rare even in the

classrooms of teachersworking to promoteit. Although pupils talked about the
mathematicsthey were doing, this was more in the natureof workingaloud than
with the intentionof initiatingan interactionwith the hearer.Less commonly, the
goals were not well defined and the talk had a scattershotnature,going in no par-
ticular direction.It took the form of disconnectedbrainstorming.Although this
may be a very valuableactivity, it is not discussion within our definition.
To give us a wider, and at the same time deeper, base for categorization,our
second round of data was gatheredin the same way as the first, with the same
teachers and pupils we already knew. At this stage we also defined more pre-
cisely our definition of "pupil-pupil"as opposed to "pupil-teacher"discussion,
and we agreed that we would consider discussion that was provokedby a com-
ment by the teacher but not discussions in which the teacher took any further
part.Given the natureof our data and our methodof analysis, this change in def-
inition caused no problems for the research.We were simply able to review all
existing data from this new perspective.
How Is the Design Adaptedin the Face of Emerging Questionsand Unforeseen

Events?
In our aim to preserve an open mind to the possible emerging categories of
interaction,we had deliberatelynot intendedto startfrom any explicit hypothe-
ses. We could not avoid, however, having some initial, generativequestionsthat
we believed could be fundamental,althoughwe realized that as featuresof the
phenomenon were revealed, we might revise our emphases. Hammersleyand
Atkinson (1990) referto this as a "shoppinglist of issues" that "clearlydrawson
the authors'priorknowledge"but does "not constitutea researchhypothesis or
set of hypotheses, nor ... provide a researchdesign as such" (p. 35). Our origi-
nal list included questions like the following:
* Are there categoriesof pupil-pupildiscussion that appearto be of more value
for the growth of mathematicalunderstandingthan are others?
* Are there specific topic areas for which pupil-pupildiscussion is appropriate
or valuable?
* Are there special benefits to be derived from such discussion?
* What are the effects of leaving pupils with the misconceptionsthey take from
their privatediscussions?
* Do pupils need to be immersed in this style of working for some length of
time before any benefits become evident?
* Does making the aims of the discussion explicit to pupils enhance their
learning?
This original "shoppinglist" was concernedwith mathematicaltopics and the
influences of discussion on understanding.Ourinclinationhad been to watch for
the effects of spoken language,but we had a growing feeling thatwe were ignor-
ing a whole area of possible interest, the behaviors of the pupils-both their
mathematicalbehavior and the roles they verbally adopted within their small
groups.This generationof new questionsin the course of the examinationof the
datais one of the hallmarksof theoreticalsampling,and we approachedthe sec-
ond analysis, therefore,from a differentframeof reference.This time, using the
second round of data and using the same systematic individual and then joint
methodof interpretation,we deliberatelycategorizedthe incidents of discussion
from the point of view of verbalbehavior.Among others,the following headings
crystallized as relevant: "defining," "into algebra," "verbalizingfor approval
(frequentlytheir own)," "confusingeach other,"and "collaborativechecking."
This last category is worth a comment here because it illustrateshow cate-
gories evolved throughan examinationof data from a new perspective,both at
this point and at a later stage in the analysis. Up until now, we had not seen
episodes in which pupils checked their work as fitting our definition of discus-
sion. It was quite common for one pupil to check the working of the group, but
we had not seen it as more than an aside to the generaldiscussion abouta task in
hand. The following episode, however, suggested a new category of discussion
was called for.
Janie and Meg were workingwith the image of balancing-scalepans as a rep-
resentationof linear equations.For example, they had worked with the picture
shown in Figure 6.1.
Figure 6.1. Balancing scale-panpicturefor the equation3t + 3 = t + 9.
Janie: So, you've got to take the same off both sides. The same numberof tins, or
weights off this side and off there so you can get rid of them.
They later progressedto writing the picture versions with "t" for tins. (The
foregoing example would have been written3t + 3 = t + 9.) Laterstill they were
faced with 8t- 9 = t + 12.
Janie: Eight tins take away nine ...
Meg: Takenineoff bothsides [shecrossesout the 9 andcrossesout the 12 and
replaces it by 3].
3
8t- 9 =/ + y
Janie: And take off a "t,"so 7t = 3 and t equals [uses a calculator;pause] ... That
can't be right. Let me do it again. Write it out again.
Susan Pirie 93
Meg: [Writes 8t - 9 = t + 12, does the canceling again, writes 8t - 0 = t + 3] But

takeawaynothingis just nothing.Takethe "t"off eachside [writes7t = 3,
takesthecalculatorawayfromJanie,anddoes3 - 7]. It doesn'twork.A tin
0.42853.It mustbe a wholenumber.Tryfour.
can'tbe [readingcalculator]
Janie: [Usingthecalculator]8 times4 equals32, takeaway9 equals23. 4 add12
equals16.No, thatdoesn'twork.[Theycall theteacherover.]
We believed that this piece of discussion meriteda category of its own; it was
more thanjust an adjunctto the precedingconversation.We called it "collabo-
rative checking."
The third analysis consisted of looking at the first batch of data gathered
throughthe lens of the second categorizationand looking at the second batch of
datathroughthe lens of the first analysis. When reviewing the first batchof data,
we noted two other instances of "collaborativechecking"that we had not previ-
ously extractedas separatelycategorizableepisodes. Thus the intentionsof sys-
tematicityand constantcomparisonbetween and among items of data can clear-
ly be seen to be intentionsadheredto.
Theoretical Sampling
The guiding questions at each stage in theoreticalsampling are What groups

do I turnto for data this next time? and For what theoreticalreasons?The crite-
ria must be goal directedand those of theoreticalrelevance. They can arise from
new questionsbeing generatedby the dataalreadycollected. One option open to
those workingwithin this methodologyis, accordingto Strauss(1987), to widen
the field within which dataare sought.An obvious choice for us would have been
to look at discussion in other fields-not with the purpose of creating a more
encompassing theory but to "stimulatetheoretical sensitivity in the service of
generatingtheory"(Strauss, 1987, p. 17) in the areaof mathematicaldiscussion.
For the reasons outlined at the beginning of this chapter-the possible non-
generalizabilityof work on discussion in other disciplines to the field of mathe-
matics-we did not choose to do this.
An alternativeapproachwe considered, before collecting furtherdata at this
stage, was to look at the types of mathematicalenvironmentswe were observing
to see whetherthese could be significant.One particularareain which discussion
was evident (although we had only two examples) was when the pupils were
working in small groups at a computer.We made the decision that this situation
was very different from the ordinaryclassroom because the pupils moved en
masse to anotherroom for these lessons, which is typical of many schools' use
of computers.Also, since we knew thata researchprojectat the LondonInstitute
was looking exclusively at this environment (Hoyles, Healey, & Sutherland,
1990), we decided to exclude these episodes from futureobservation.
One of our teachershad the practiceof teaching most of the mathematicssyl-
labus through investigations with backup from time to time delivered through
routine practice. This seemed to be a good environment to explore further,
because he made it plain to his classes thathe expected them to work togetherin
threes or fours on all the investigations and totally alone on the practice days.
Classes familiarwith his methods automaticallyworkedin this way. The advan-
tage of observingovert problemsolving of this naturewas thatthe solutionof the
problemcould be seen as a growth in understandingwithin the group, although
not necessarilyfor individuals.We became interestedin the questionDoes prob-
lem solving benefit from pupil-pupildiscussion? To get a wider perspectiveon
the effects on problem solving, we identified anotherteacherwho, by contrast,
used investigative group work as non-topic-specific relaxation between more
orthodoxspells of teaching throughexposition.
Most of the discussion that we observedtook place in small groups,but one of
our original teachers espoused a philosophy of learning based on whole-class
discussion, which frequently splintered to heated small-group discussion in
which all the pupils had to be able to justify the argumentsthey put forward.At
the beginningof the year, it was common to hearhim say, in responseto a ques-
tion from a pupil, "Don't ask me, ask her, she put the idea forward."But by a few
weeks into the term,many of the pupils totally disregardedhis presence afterhis
stimulatingand often provocativeinput at the startof the lesson. A furthersub-
sidiary, generative question became Are there significant differences between
whole-class and small-groupdiscussions in terms of pupils' learning?
The life of a researcheris never plain sailing! At this momentthe "whole-class
discussion"teacherleft the school in which he was working to head the depart-
ment of a school with a contrastingview of how mathematicsshould be taught,
namely, quietly, with individual learning materials.Ratherthan enter as a new
broom sweeping all along with him, he determinedto change the departmental
teaching approachby his own graduallychanging example. Initially the classes
he taught would still use the same individualizedmaterials,but studentswould
work at them in pairs. Although we lost the opportunityto gathermore data on
whole-class discussion, this change gave us the opportunityto observe discus-
sion in a quite different,structuredenvironment.The natureof our methodology
meant that this for us was nothing but an advantage.A differentcontext for the
discussion might enable us to spot importantfeatures of pupil-pupildiscussion
that were not evident to us previously, which did indeed prove to be the case.
An interestingvariationof the category of "collaborativechecking" came to
light, as the following episode demonstrates.Jonaand Bette were workingon the
same material,sometimes together and sometimes doing the tasks individually
and then looking at each other's work to check what the other was doing.
Presumablybecause talking was an encouragedbut unfamiliaractivity for these
pupils in a mathematicsclassroom, they frequentlytalked aloud, but ostensibly
to themselves, as they worked.
Jona: [They are expanding(x - 2)(x + 3) and Bette has writtenx2 + 3x + 2x - 6 and
x2 + 5x - 6] Tell me howyou didthatagain.[Shehasobviouslyoverheard
Bettetalkingherwaythroughherwork.]
Bette: Right, so you times the x's, x squared,OK? Then x times 3 ...
Jona: ... plus 3x [writingx2 + 3x].
Susan Pirie 95
Bette: Thenyou do theminus2. Timesx.

Jona: [Writing]minus2x.
Bette: ... and the numbersmake minus 6.
Jona: [She has writtenx2 + 3x - 2x - 6 . Below she writes x2 + x - 6.] Last time
you saidfive. Five x, but it's 3 subtract[emphasis]2. You put addthere.
That'swhereI didn'tget it, whatyou'ddone.
Here the notion of working individuallywhile being encouragedto talk about
their work had offered Bette an opportunityto correcta mistake in her working.
We created a subcategoryof "collaborativechecking" and called it "revealing
errorsto themselves." In terms of our category groupings outlined previously,
this episode was coded (c,h,q); they were reviewing theirunderstandingof a task
because Bette's original result did not fit with Jona's understanding-c, mathe-
matical language was present-h, and the statementswere operational-q.
When Can I Stop?
The thirdphase of data collection and analysis was conductedwith the same
methods as before but in the new, changedand focused environments.Each new
set of dataanalyzedwas used to confirmexisting categoriesor suggest new ones.
In the latter situation, all previous data were examined to see whetherthe new
classification was an anomaly of the particulardata set or a universallyimpor-
tant grouping.This may sound very tedious and time consuming, but of course
with each new analysis, the researcherbecomes very familiarwith the previous
data. In addition,the process is not of itself unending.The aim is to attaintheo-
retical saturationin each category. This is judged to have been achieved when
new examples of the category add nothing to the developmentof its properties.
When similarinstances are encounteredrepeatedly,the researchercan be empir-
ically confidentthatthe categoryis saturatedand can then cease to examine such
data in the future. Data collection is then concentratedon filling gaps in other
areas pointed up by emerging theories and questions based on existing data.
When furtherdata sets suggest no new categories, then the researchcan be con-
sidered theoreticallystable.
SUMMARY
By offering a detailed account of the design of a specific researchproject, I

have illustratedthree key considerationsthat need to be addressedif we are to
successfully adaptthe researchparadigmsof otherdisciplines to appropriateuse
within mathematicseducation.The first of these concerns the researchquestion.
Tempting as it is to wish to demonstrateour independence from the enumera-
tive methods of mathematics, the research question must always drive the
choice of methodology. Too often we hear such statementsas "I'd like to do an
ethnographicstudy"before the focus of the study has been selected. Only when
first the broad interest and then the refined questions have been teased out is it
appropriateto turnto the selection of a methodology. This second key decision,

the choice of methodology, should not be undertakenhastily. We must review
imaginativelythe range of possible approachesto answeringour researchques-
tions. One approachmay at first sight appearseductive, but it is in the details
that the connections between questions and successful explorations lie. The
final considerationis one concerningreporting.It is necessary for us to be very
explicit, when writing about our research, about what exactly are our data, so
others may, for themselves, judge the validity of the conclusions and recom-
mendationsthatwe make. Are we basing our analysis on the tape itself (whether
video or audio), on transcriptsof the tape, on field notes made at the time of the
incident, on notes writtenup afterwards,or on some combination?Any of these
and more can be appropriatedata sources, but the kinds of analyses that we can
make will differ (Pirie, 1996b). That said, we must also be explicit about how
we performedthe analysis of these data. If we are to hasten the day when it will
be "possible to codify methods of qualitative analysis with ... clarity," then I
suggest that we need to pay close attentionto the three considerationsoutlined,
but most especially to the last.
The purpose of this chapterwas to demonstratea qualitativeapproachto the
researchdesign for a specific projectand to illustratesystematictheoreticalsam-
pling with constantcomparisonas a possible methodologyfor mathematicsedu-
cation researchthat intends to build ratherthan test theory. It would be cruelly
tantalizing,however, to leave the readerwith no insight into the theory emerg-
ing for the specific project described, and so I conclude with a postscriptthat
traces the emergence of one particularcategory and present the inferences we
drew from it. I then state very briefly some of the tentative answers that are
emerging to our generativequestions.
Postscript
I referredabove to the emergence of new categoriesfrom the data we collect-
ed in the "thwartingindividualizedlearning"environment.There were several
such categories,the most strikingof which we called "pupilas teacher."This cat-
egory was characterizedby one of the pupils clearly enactingthe role of a teacher
in the interaction.We took as evidence for this role play the pupil's adoptionof
the functions of teacher talk and language offered by Sinclair and Coulthard
(1975). Having reanalyzedall previousepisodes of discussion for the possibility
of inclusion within this new category and, interestingly,having found very few
examples in any other collections of data, we looked in detail at each included
episode. Some of the more extreme episodes we separatedinto furthernew cat-
egories labeled "usingpupil answerbook" and "pupilas lecturer,"and the main
category was subdividedaccordingto whetherthe pupil spontaneouslyadopted
the role of teacheror was cast as teacherby anotherpupil. We also looked at how
faithfully each pupil played the assigned role. This subcategorization is
explainedin detail in Newman and Pirie (1990). In the cases in which the pupils
Susan Pirie 97
were working on the individualizedmaterials,which by their very design had to

have progressionbuilt in, the criterionwe used for judging the effect on under-
standingin the shortterm was whetherthe pupils were able to successfully com-
plete the task or move to the next activity.
It is importantto rememberthatwe are not in a position to make sweeping gen-
eralizationsfrom our data. Instead, we wish to draw out common features that
form significantfactors in the creationof a theory describingpupil-pupildiscus-
sion. One such featureis thatwhen pupils elected to spontaneouslyadoptthe role
of teacher,they tended to stay faithful to the role of mathematicsteacher,using
mathematicallanguage and making abstractivestatements(groupr). Those cast
as pupils, too, tended to play their role in an almost caricature,typical manner
(listening carefully, trying to do what was asked of them, respondingpolitely),
and the outcome was successful in the shorttermat least. We could theorizethat
pupils who choose the teacher's role, who offer to help their peers, see the situ-
ation as one in which mathematicallanguage must be used and the discussion
cannot be at the operationallevel. As a consequence of the adoptionof the role,
they believe it necessary to continue the interactionuntil there is evidence that
the "pupil"has gained in understanding.In contrast,when forced into the role of
teacher,the majorityof the "teacher"pupils do not use mathematicallanguage,
nor do the discussions contain abstractstatements.In addition,there is no guar-
antee of a successful outcome, which might lead one to suggest that simply ask-
ing one pupil who "knows"to explain or discuss some piece of mathematicswith
anotheris not necessarily an activity of value. This latter statementis born out
by evidence we have, in the category"confusingeach other,"of occasions where
confusion, sometimes of both pupils, results from the event (Pirie, 199la).
The category "pupilas lecturer"illustrateshow discussion does seem to be of
value. Therewere also othercategoriesfor which this was the case, such as "into
algebra."Here pupils were seen to create their own personal "algebra,"usually
in an attemptto generalize or representa problemsymbolically. This frequently
involved an extended discussion, which led to a much clearerunderstandingof
the mathematicalproblem.This contrastsstrongly with "using algebra,"a cate-
gory in which pupils discussed the use of standardalgebraic techniques, often
with disastrousresults (Pirie 1991a)!
As previously indicated,much data still remain to be analyzed and reported.
Indeed the richness of the data collected ensures that much still lies to be dis-
covered by those who care to revisit the data.Perhapsthe markof a fruitfulpiece
of researchis that it fulfills some or all of its originalintentionsbut uncoversthe
way to furtheravenues of exploration.To be of value, theory must remaina liv-
ing comment, adaptingto new environmentsas they arise.
98
Chapter 7
Studying the Classroom Negotiation

of Meaning: Complementary
Accounts Methodology
David J. Clarke
The researchproceduresthat form the focus of this chapterwere developed in

an attemptto study learning in legitimate classroom settings while minimizing
the need for researcherinference regardingparticipantthought processes and
maximizing the richness of the research data base. The focus of the research
stems from a perceived need to model empirically the process of learning in
classroomsand, in particular,the so-called "negotiationof meaning."The use of
the metaphorof negotiationgroundslearningin social activity, but this research
adopts the epistemological stance that the meaning that is constructedas a con-
sequence of this social activity is experiencedas personalmeaning, and any the-
ory of learningmust accommodatethis constructedself. Classroomsarecomplex
social settings, and researchthat seeks to understandthe learningthat occurs in
such settings must reflect and accommodateto that complexity. This accommo-
dationcan occur througha data-collectionprocess that generatesan appropriate-
ly rich data set. Such a complex data set can be adequatelyexploited only to the
extent that the research design employs analytical techniques sensitive to the
multifacetedand multiply connected natureof the data.
The focus of this chapteris a qualitativeresearchapproachI have called "com-
plementaryaccountsmethodology."Available technology is utilized to combine
videotape data with participants'reconstructionsof classroom events. This inte-
grateddata set then provides the basis for complementaryaccounts constructed
by the research team. Complementaryaccounts methodology is distinguished
from other approachesto classroom researchby-
* the nature of the data collection procedures,leading to the constructionof
"integrateddata sets" combining videotape and interview data;
* the inclusion of the reflective voice of participantstudentsin the data set;
* an analyticalapproachthat utilizes a researchteam with complementarybut
diverse areas of expertise to carry out a multifacetedanalysis of a common
body of classroom data.
I would like to express my gratitudeto Sue Helme, whose insightful interpretativesug-

gestions contributedgreatlyto the researchreportedin this chapter.
David J. Clarke 99
Withinthe scope of this monograph,it is possibleto set out only certainkey fea-
tures of the research activity: the principal means of data collection, and an
overview of the multipleformsof analysisthatare demandedby the complexityof
the settingand made possible by the complexityof the data.In the discussionthat
follows, specificresearchtechniquesof datacollectionandanalysisareoutlined,by
which the natureof classroomlearningmightbe put on a moreempiricalfooting.
PURPOSEIN CLASSROOMRESEARCH
The particularaim of the research discussed here is to contributeto a con-

structivistmodel of learning that adequatelyaccommodatesthe social activity
typical of classroom settings with specific regard to negotiation and the con-
structionof meaningsby learnersin mathematicsclassrooms.
Attemptsto model the learningprocess in classroom settings have employed
the metaphorof the "negotiationof meaning,"and we need explicit, viable, and
falsifiable models for the process representedby this phrase.Such models would
then serve to directour attentionto those constructsand associatedbehaviorsthat
we might study productively and, subsequently, to the methodological tools
demandedby such research.
A research procedure is requiredthat is designed to reveal the process by
which meanings are negotiated and constructed by students in mathematics
classrooms. At the heart of such researchis the question of whose accounts of
classroom activity are privileged for the purpose of understandinglearning
processes in such settings. The key to the approachdescribedhere is to ground
student accounts of classroom activities (including thoughts, motivations, and
construedmeanings) in a videotape record of specific sharedclassroom events
and to supplementeach student's account with an associated data base of other
students'accounts,researcherfield notes, and transcribedvideotaperecords.By
this approach,the researchtechniquesof classroom videotape analysis and stu-
dent clinical interview are combined to best effect in a mannerdesigned to be
reciprocallyvalidatingand illuminating.
The theoreticalconstructsof meaning (Bakhtin, 1979), sources of conviction
(Frid, 1992), and classroom consensus processes (Clarke, 1986) informed the
interpretativeframeworkfor the study and, therefore,the method of data collec-
tion. Sources of conviction refer to how one determinesfacts, legitimacy, logi-
cality, consistency, and accordance with accepted mathematicalor scientific
principlesand standards(i.e., academiccontent meanings) and to the authorities
cited by individuals to justify their statements, actions, or interpretations.
Consensusprocesses are typified by groupcompromise,refinement,and accom-
modation-taken to be those interactionswhereby conjectures and arguments
arising in classroom discourse are comparedand assessed (including the devel-
opment of social-contextmeanings).
With regardto meaning:The presumptionsof meaning are based on commu-
nity, purpose,and situation.It is futile to discuss the meaning of a word or term
100 Studyingthe ClassroomNegotion of Meaning
in isolation from the discoursecommunityof which the speakerclaims member-

ship, from the purposeof the speaker,or from the specific situationin which the
word was spoken. Indeed, it is not the word that has meaning,but the utterance.
The emphasis on utterance,derived from Bakhtin(1979), is evident later in this
chapterin a discussion and an illustrativeanalysis focusing on intersubjectivity.
The resultantframeworkbased on the characterizationof these termsservedto
identify key data types (and the associated researchtechniques)essential to the
researchdesign and also provided the theoreticalframeworkfor the interpreta-
tion of the resultantdata. Data related to negotiative situations and associated
social activity were considered essential to any analysis of such consensus
processes. The structuralanalysis of classroom text is informedsignificantlyby
the theoreticalemphasison utterance.It mustbe stressedthata theoreticalframe-
work used in such a way confers coherence on the researchdesign and tells the
researcherwhere to look withoutpredeterminingwhat will be found.
The challenge for this type of classroom researchwas to portraythe learning
process of an individual embedded in a highly complex social context. This
learning process was taken to be an integrationof not just the obvious social
events that might be recordedon a videotape,but also the individual'sconstrual
of those events, the memories invoked, and the constructionsthat arise as a con-
sequence. The research procedure recounted here was designed explicitly to
achieve this integration.
Centralto this researchprocedurewas the use of videotapedclassroomlessons
and video-stimulatedrecall techniqueswithin an interview protocol that sought
to obtain the following:
1. Students'perceptionsof their own constructedmeanings in the course of a
lesson and the associatedmemories and existing meaningsemployed in the con-
structiveprocess
2. Students' sources of conviction for the constructionof their mathematical
meanings
3. The individuals, experiences, arguments, or actions in which students
believed mathematical(academiccontent) authorityto reside
Since the concern was with the context of learningand descriptionsof teach-
ers' andlearners'mathematical(and social) interpretationsin classroomsettings,
the researchmethods were qualitativein character.
THE PRACTICALITIESOF DATA COLLECTION
Procedure:In the Classroom

Two video cameraswere used:One focused consistentlyon the teacher,andthe
other, on a selected group of about four students.This approachdid not assume
that the studentswould be working in collaborativegroups, only that they were
seated sufficientlyclose to one anotherthat the entire groupcould be simultane-
David J. Clarke 101
ously in view and a single microphonecould be used to recordthe conversations

of all four students. The teacher's conversationwas recorded through a radio
microphone.Once in place, the video camerafocused on the studentsrequiredno
attentionunless the studentschanged location within the classroom;however, a
researchassistantwas requiredto keep the teacherin view on the othercamera.
The dual video images of teacher and pupils were combined into a single
image through a compact audiovisual mixing board situated at the rear of the
classroom. This image was viewed on a small portablemonitor.The composite
image was structuredso that the studentsoccupied most of the viewing screen
with the teacherincluded as a small insertin one corer of the screen. The same
mixing device combined the audio input from both microphones.The relative
volume of teacherand studentconversationscould be adjusted.
The combined image was recorded onto video-8 tape using a very compact
video recorder. This video recorder was linked to a laptop computer. The
researcherwas seated at the rearof the classroom with easy access to the mixing
board and the ability to view the image recordedon the video monitor. Using
headphones, the researcherwas able to listen simultaneouslyboth to the stu-
dents' conversationsand to the teacher's utterances.Using the CVideo software
(Roschelle, 1992), the researcherwas able to recordfield notes onto a word pro-
cessing document on the computerand also "time-tag"the notes to the corre-
spondingevents in the video record.
The following example illustratesa typical set of field notes (numbersrefer to
hours:minutes:secondselapsed on the video tape):
00:15:04to 00:18:35 Studentsdoingnumber2
K sayssomething; A: "Idon'tget it"
AftersometimeA hesitantlyputsherhandup andwithdraws it.
A putsherhandup againwithoutgettinga response;USETHIS
00:18:36to 00:22:19 Studentsstillworkingon number2
T demonstrates a particular
method
Studentscall out"122"[Kmoveshermouth-thinkingwhat?PURSUETHIS]
A hesitantbutattentive.K staringat herpencil-discouraged?
Such field notes serve two purposes: first, as a record of the researcher's
immediateimpressionsof significantsocial interactionsand learningevents; sec-
ond, as referencemarkersfor the subsequentinterviewingof the studentsubjects.
Relating the Interviewto the Video Record

In the video-stimulatedinterview, prioritywas given to those events to which
the studentattachedsignificance. The researchemphasison negotiation,consen-
sus, and conviction was used to attach significance to other events (not neces-
sarily regardedas significant by the student)involving disagreementor uncer-
tainty that then became the subject of student reconstruction.This process, as
with all researchtechniques,has its limitations.It is possible that later analysis
of the videotapemight identify an event, whose significance was not recognized
at the time by eitherthe studentor the researcher.As with any clinical interview,
the researchermay miss one insight throughthe pursuitof another.Nonetheless,
the videotape record was substantially enriched by the student's immediate
reconstructionof those events perceivedto be significantat the time. Subsequent
analysis of interviewand videotapetranscriptscan be used to reveal relationships
of meaning that were not apparentat the time of the interview.
An importantdesign considerationis that without the videotape as stimulus,
the student'saccountis likely to be superficialand groundednot in actualclass-
room events but ratherin the student's uncertainreconstructionof the lesson.
Inconsistenciesamong studentaccountsof classroomlessons provide a fascinat-
ing study in themselves, but it is the learningstimulatedby the student'spartic-
ipation in particularclassroom events that is the focus of this research,specifi-
cally, the relationshipbetween these events and the student'sconsequentknow-
ings. Where the purposeof analysis is an understandingof the learningprocess,
the researcher'sinterpretationof the videotapedatais likely to be inadequateand
possibly misguided without the student's account. The researcherwould lack
insight into the associations,memories, and meaningsthateach classroomevent
evoked for the student.
Procedure:The Interview
At the end of the lesson, using the video record as stimulus, the researcher
interviewedthe targetstudentsindividually.This techniqueis in widespreaduse
(see, e.g., Anthony, 1994). For the interview stage, the researcherused the video
recorder,the video monitor,the laptop computer,and a compact audio recorder
to recordthe interviewon audiotape.The video recordof the lesson was sampled
as requiredin response to the identificationof a particularepisode by either the
student or the researcher.The following example is typical of the commence-
ment of such interviews.
I: Whatdo youthinkthatlessonwasabout?
S: Oh,linearfunctionsandhowyou graphthem.
I: Wasthatsomethingthatyou understood beforethelessonstarted?
S: Not really.I meanI knewaboutlinearfunctions,andwe haddonea bit of stuffon
graphing,butI couldn'tsayI reallyunderstood it.
I: Wouldyou say thatyouunderstand it now,afterthelesson?
S: Yes, I thinkso.
I: At whatpointin thelessonwouldyousaythatyoucameto "understand" aboutgraph-
ing linearfunctions?
S: I'm not sure.ProbablyafterI hadtrieda few of the problemsin the book,andthey
seemedto be comingoutOK.
I: Wastheresomethingthathappened in thelessonthatreallyhelpedyouto understand?
S: Maybe.I'm notsure.
I. It seemedto me thatyou spenta lot of timetalkingto Simoneat one point,whatwas
all thatabout?
David J. Clarke 103
S: Really. I don't remember.I suppose it was somethingto do with the problems.Yeah.

She asked me about the first one, and then we startedtalking about how to do them.
I: Let's look at that bit now, and you can tell me what you were thinking.
The use of the CVideo software enabled the researcher to locate within the
field notes reference to actions of the student that seemed to be of significance
either to the researcher or to the student. Having found this point in the word doc-
ument, the software was used to find the corresponding moment on the video
record, which was then played back and discussed. The contrast between the
superficiality of the student's recollections without the aid of the videotape and
the comparative richness of the student's subsequent comments regarding a spe-
cific videotaped incident provide a recurrent endorsement of this technique.
The audio record of the interview was transcribed onto the relevant section of
the word document and time-tagged to the corresponding video incident.
Together, the video record and the word document incorporated the student and
teacher actions and utterances throughout the lesson, the researcher's field notes,
and the student's interpretations and explanations of significant events. This inte-
grated data source was then available for analysis. One example of such an inte-
grated word document is given below (bold = field notes; plain text = transcript
from videotape; italics = transcript from interview). The following sample text
includes two students' reconstructed accounts of the same event, after viewing
the video record of the interaction transcribed below.
00:39:27 to 00:41:30 T(teacher) asks students K and L what they've done. K
explains, T is dubious, then says I think you're right, L explains, says we're right.
T: Where'd you get 180 from?
K: Width. Equalspoint-
T: Why did you multiply them together?
K: To get the area. Forty-five thousand,thereforeyou'd need.
T: Forty-five thousand?
K: Forty-five thousand.That's what we got.
T: Forty-five thousand?Can you place that-can you place that-can you do that again?
Two hundredand fifty times one hundredand eighty, oh, hang on, hang on, I think
you're right. I think they're wrong.
K: Yup, they're wrong. We're right.
[L holds up calculator].
I (Interviewerwith student L): Uh huh. So why were you so sure your answer was right?
Or were you sure your answer was right?
L: Um, because when she asked us what it was, she thoughtit was right too.
I: I'd like you to tell me this last bit. So say that again for me.
L: She came and asked us to do the answer that we found and we had a differentanswer
to the one that another group had given her and when she heard our answer it must
have clicked that, um, it sounded more right than the other one did, so she went to tell
them that they were wrong.
I: It musthave clicked with her, so that's why she thoughtit was right. Whydid you think
it was right?
L: 'Cause if Mrs. Burtonthinksit's right, it probably is.

I (Interviewerwith student K): Thereshe's going over to your group.
[videotape continues] Yeah,so you seem prettysure, you got the, you had answersfor
everythingshe said.
K: Yeah,because, um don't say anythingto her, but the girls said that she'd pick on me
a bit because I'm new, so if I show her that I know what I'm talkingabout then she'll
lay off.
I: Oh, well that makes sense.
K: Because I don't want to seem like I'm going, "Well,you're right."
I: But you did know, didn't you?
K: Yeah,but I just thoughtit, 'cause yeah.
I: Yeah.
K: So I just wanted to show her that I know what I was talkingabout, because otherwise
she'd keep on at me.
I: Yeah.
K: Admit that actually sort of thinkthat she made a little bit of a mistake. Yeah.I knew
she'd lay off if I sort of had an answerfor everything,so that's whyIjust said, straight-
away she 'd ask me a question,I'd have an answer, and she 'd go, um, thinkabout itfor
a while and then straightawaysay, "You'reright" or "You'rewrong." Yeah,I just
wanted to say everythingquicklyso she didn't have time to thinkof another question.
In combination, the researcher's field notes, the transcribed classroom dia-
logue, and the student's explanations augment the video record to provide a rich
data base for subsequent analysis. Other examples are given in the following dis-
cussion of data analysis.
ANALYSIS: FINDING STRUCTURE IN DIVERSITY
There are many comments that might be made about the preceding data.
Comparisons might be made of the two students' interpretations of the exchange
with the teacher. In a study whose major focus is classroom negotiation, student
K's account of her motives and her perception of the interaction offer some
insight into what it is that is actually being negotiated in an exchange whose sur-
face content is mathematical. The example above strongly suggests that without
the student's reconstruction, the researcher's account of the interaction would be
unlikely to capture the student's motivations and construal of the particular
social situation. Yet, lacking this detail, any inferences the researcher might
make regarding the student's participation in the classroom and her associated
learnings would be extremely restricted.
Inference is an obligation the researcher cannot escape; the minimization of
the inferential gap remains a methodological imperative if our research is to
claim either authority or application different from that of the novelist or the
poet. To infer student thought processes and the significance of classroom
events on the basis of only videotape data seemed an unnecessary and unjusti-
fied extrapolation. An important, possibly essential, perspective on the class-
room was obtained from the students themselves in interview situations with the
David J. Clarke 105
assisting promptof the classroom video record. Students reconstructed,in full

knowledge of the ultimate outcomes of those actions, the significance of the
events and their associated motivationsand meanings. Such accounts informthe
researcher's interpretation of events, either by their similarity with the
researcher's interpretationof the interaction or by their difference from that
interpretation.
This chapterfocuses on two fundamentallyrelatedresearchactivities:the col-
lection of a rich set of dataand the use of analyticprocesses essential to the real-
ization of the potential of these data. This section describes a set of different
interpretativetechniques that can be used to provide multiple and complemen-
tary interpretationsto adequatelyportraythe many facets of classroom learning
containedin the integrateddatasets. Examplesfrom the videotapetranscriptsare
used to illustrateanalysis by exemplificationand a structuralanalysis of text. A
brief discussion is then presentedof several other analytic methods.
Analysis by Exemplification
The key constructs from which our theories of classroom learning are con-
structedmust be empiricallywell founded. The operationalizationof these con-
structs occurs through the accumulation of research examples, in which the
obligations of the researcher are to provide theoretical justification of the
claimed exemplificationsand to subjectthe proposedoperationalizedconstructs
to the scrutinyof the informedcommunityof researchersand learningtheorists.
The following discussion illustratesthe use of "analysisby exemplification"to
identify empiricalexamples of the key constructsof "negotiation"and "intersub-
jectivity." The sample analysis from the integrateddata set that follows starts
from evidence of students'intersubjectivityand documentsalternativeforms of
uncertainty and the negotiative process whereby resolution is achieved-a
process in which intersubjectivityhas a centralrole. In this accountof classroom
learning,intersubjectivityenters as a mediatingagency, essential to the negotia-
tive process, wherebyuncertaintyis resolved and new knowings are constructed.
Both negotiationand intersubjectivity,and the relationshipbetween them,require
definitionto providea theoreticalframeworkfor the analysis thatfollows.
Negotiation has been characterizedin some detail elsewhere as a cyclic
process of refraction (construal), reflection, and representation-the goal of
which is consensus (Clarke, 1996). Lave and Wenger associate learning with
participationin practiceand assert that "participationis always based on situat-
ed negotiation and renegotiationof meaning in the world" (Lave & Wenger,
1991, p. 52). Cobb and Bauersfelddefine the negotiationof meaning succinctly
as "the interactive accomplishmentof intersubjectivity"(Cobb & Bauersfeld,
1995, p. 295).
Negotiation depends on language (or at least on some form of communica-
tive process), and language is constitutively intersubjective(Todarov, 1984, p.
30). Thus, a level of student-studentand student-teacherintersubjectivity is
prerequisite to the negotiative processes by which the resolution of uncertain-

ty is attempted. A relationship between the constructs negotiation and inter-
subjectivity can be summarized in the argument that one pathway to knowing
is by means of the resolution of uncertainty, that the process of resolution is
fundamentally negotiative, that negotiation is mediated by language, that lan-
guage presumes intersubjectivity, and that the matter of intersubjectivity is
meaning (Clarke & Helme, 1997, p. 117).
The theoretical constructs of negotiation and intersubjectivity both require
empirical substantiation. Similarly, general statements of principle, which define
one construct in terms of another, require empirical demonstration of the postu-
lated relationship. This is particularly obligatory when the proposed relationship
is one of process and product. Since classrooms represent legitimate sites of sit-
uated mathematical practice, this perspective supports the need for the empirical
documentation of negotiative processes in the classroom. For example, an under-
standing of the means by which the resolution of uncertainty might be achieved
requires an understanding of intersubjectivity as a phenomenon of social inter-
action. To establish this point, consider the following videotape transcript. (All
utterances are by students. K and L were subsequent interviewees, S19 and S20
were not.)
Episode 1
1. S19: It says how many sheets of graphpaperwould you need to show one million 1-
millimetersquares.
2. L: To show one million, you know you don't divide it by 100, because there's more
thana hundred1-millimetersquares.I mean you're going to find the areaof this.
3. K: What?
4. L: You've got to find the areaof this, there's more thanone hundred1 millimeters.
5. K: That's right. I was doing length by-oh screw that.
6. L: One hundred1-millimetersquares.Take length ...
7. K: Um, there's how many down here?
8. L: And along that side there is ...
9. K: 10, 20, 30, 40, 50. How many are there down there?
10. L: There's a hundred1 millimetersthere.
11. L: No, there wouldn't be.
12. K: There wouldn't be, that's not right.
13. L: There'd be 250.
14. K: Yeah.
15. L: Yeah, there'dbe 250.
16. K: And we just totally screwed it all ...
17. L: Length of graph.
18. K: OK, so it would be length times width [inaudible]
19. L: And uh, 250 millimeters.Width ...
20. K: What's width?
21. L: That's...
22. K: That's 10, 20, 30, 40, 50, et cetera.
David J. Clarke 107
23. L: 18, 180.

24. K: Times 180. OK herewe go. 250 times 180 equals45 thousand.OK,that's45
thousand. Weneeda million.What'sa milliondividedby 45 thousandandtimes
it by that?
25. L: Hangon, hangon, hangon, hangon. Don'tgo too fast.OK.Thereforethereare
45 thousandmillionmillimetersquares.
26. S20:45 thousandmillion?
27. L: Yeah.
28. S20:45 thousand.
29. K: 22.2.
30. L: Ononepiece.Of graphpaper.
The goal of analysis by exemplification,in this instance,is the operationaliza-
tion of intersubjectivitythroughthe accumulationof examples. In the preceding
transcriptare several indicatorsof intersubjectivity.First, much of the recorded
dialogue is incoherentas written text: That is, sentences are ungrammaticalor
incomplete;pronounsare used withouttextualclues as to theirreferrents;single-
word utterancesare frequent.Communicationin this form is interpretedas being
sustainableonly because the participantsshare understandingsof the referrents
of the pronounsor key words and of the processes, actions, or relationshipssug-
gested (but not stated)by the sentence fragmentsand the participants'gestures.
Second, evidence of intersubjectivitycan be interpretedfrom the manner in
which one speakercompletes the sentence of the previous speaker,as occurs in
lines 8 and 9, and in 28 and 30 (I am using line as a shorthandfor utterance).The
overt text in the form of the literal transcriptionis here being distinguishedfrom
the implicit text being coconstructedby the participants.The existence of an
implicit text is inferredon the basis thatthe interactionappearsto have been both
purposefuland successful.
The findings of this type of analysis reside in the claim that the preceding
episode constitutes an example of intersubjectivity in a classroom setting.
Evidence for the enactmentof the constructof intersubjectivityshould assist us
in theoreticallylocating intersubjectivitywithin the learningprocess as agency
or as outcome or as both.
StructuralAnalysis of Text
Anothertype of analysis of the integrateddatasets used to achieve the research
goals involves the classification of the text into three levels: the episode, the
negotiative event, and the utterance.Episodes comprise the dialogue and activi-
ties that studentsengage in as they approach,work on, and complete a particular
classroom activity, such as a problem-solving task. Thus, each episode is a
coherent unit of activity unified by a single purpose. Each such episode may
involve several negotiative events. A negotiative event is defined by an identifi-
able intermediatepurpose-a purpose whose realizationis an intermediategoal
within the encompassing episode. Each negotiative event may be composed of
several utterances, each with its own immediate purpose. The interpretation of
the significance of a given episode requires an interpretation of each constituent
level: the negotiative event and the utterance.
The following episode illustrates the partitioning of text according to the
occurrence of "negotiative events" within an "episode." (K and L are student
interviewees, S20 and S22 student noninterviewees, and T is the teacher.)
Episode 2
EVENT 1
1. L: [Writing]500 sheets. Height equals.
2. K: OK, question 2 [Find the height of a stack of one million sheets of paper].
3. L: Does everyone understandwhat we did with numberone?
4. K: No, but. Anyway. 500 sheets.
5. L: And how many sheets do we need?
6. K: 500 sheets of what? 500 sheets.
7. L: Their height equals five point eight.
8. S20: We've done that.
9. K: I know. But we've got to do it all togetherso.
10. L: One point oh times ten to the power of six divided by five hundred.
11. K: Oh veah sure everyday what are you talking about?What are you talking
about?
12. L: Finding out how many five hundredsthere are in a million.
13. K: How many five hundredsthere are in a million. That would make it one
thousand.How many thousandsare therein a million? Thatwould make a
thousand,two thousand.What? [to anotherS]. I have a lot to say.
14. L: [Uses calculator]Two thousand.Well done!
15. K: This is called skill. This is what you do. Five hundredinto a hundredwhich
is two. Then you do a hundred[correctingherself] which is a thousand.
16. L: Times five point eight. Shush.
17. K: Which is 2. Then you do a thousandinto a million which is one thousand,
so one thousandtimes two is two thousand.
18. L: Eleven thousandsix hundred.
EVENT 2
19. T: With your working out folks I want you to tell me what you are mult-
Matthew-what you are multiplyingby, and you simply put a little arrow
telling me what and why.
20. K: What are we doing? Is it a million sheets of paperthough?
21. L: Yeah. One point ...
22. K: We're doing a million sheets of paper.
23. L: Yeah, you need ...
24. K: Yes we do. We do, shut up.
25. L: Therefore,I did that wrong.
26. K: Two thousandtimes five point eight is eleven, six, zero, zero. [i.e.,11,600]
27. L: [Soundingout letters] M-ms?
28. K: Centimeters-which would make it eleven point six meters,right?
David J. Clarke 109
EVENT 3
29. L: Or eleven-yeah. It'd be eleven point six meters, wouldn't it, 'cause you
take off one to get the centimeters,and anotherone, yeah. [pause]
30. K: [Lookingup] That's quite high, isn't it?
31. L: All right. And you've got to point out what the units [?] are, right?
32. K: You've got to point out what the what is?
33. L: We have to show what we're multiplyingby.
[S22 says somethingto K, K laughs]
34. S20: That's not how you know, you look like you know what you're doing and
you just do it.
35. K: Exactly, you go into a state of total concentration,it lasts about2 seconds,
that's when you get the answer, and then you don't know what you're
doing, so it doesn't matter.Five hundredsheets equals, height equals five
point eight centimeters.I don't even understandwhat I wrote. [pause as L,
K write]
EVENT 4
36. K: But why do we divide a million by five hundredto get that answer?
37. L: Because you know, if you know what the height is ...
38. K: So what am I doing? Tell me what I'm doing here, tell me what I've done.
39. L: All right. You know that five hundredsheets equals fifty-eight ...
40. L and K: Five point eight centimeters.
41. K: There is a point there, it's up there.
42. L: Oh, I can't see it.
43. K: Get some other glasses.
44. L: Now, we need to know-we need to know the height of a million sheets of
paper.Thereforeyou must divide a million by five hundredand times that
numberby five point eight.
45. K: [Writing]Equals two thousand.Sheets of paper.
46. L: OK.
47. K: Two thousandtimes five point eight centimetersequals eleven thousand
six hundred centimeters, equals eleven point six meters of paper. [bell
rings]
The preceding transcription may be considered one episode in a lesson con-
sisting of many episodes. Using structural analysis, I partitioned this episode into
four events that can be characterized as follows:
* Event 1 combines the refinement of intersubjectivity within the group with
L's first solution attempt (1 to 18).
* Event 3 involves the negotiation of appropriate units of measurement (29 to
35).
* Event 2 revisits the procedure employed in Event 1 (19 to 28).
* Event 4 reviews the procedure again and links it to the task (36 to 47).
These structural elements identified within the text may reflect parallel struc-
tures within the process of learning.
Analysis of Dialogue as Text

The use of indexing softwarepackages for textual analysis of the dialogue in
the integrateddata sets can identify points at which the student's state of know-
ing demonstrablychanges from uncertaintyto comparativecertaintyor convic-
tion. It may be that a studentrecounts,in an interview,a situationin which he or
she came to "know"or to "understand"somethingrelatedto the topic dealt with
in the observed lesson. In particular,any studentuse of the verbs know, under-
stand, or learn can be analyzedin detail with regardto the subjectand object of
each verb's use. That is, what sort of things can be known, understood, or
learned;what sort of experiences,events, images, or people appearto be associ-
ated with learning,understanding,or coming to know; and who is it that knows,
understands,or learnsthings?
In this form of analysis, such software packages as NUDoIST(Qualitative
Solutions and Research, 1994) can be employed to undertaketextual analyses of
the frequencyof associationof terms, such as those mentionedabove, in student
classroomdiscourseand in studentdiscussion of videorecordedclassroom situa-
tions (Clarke& Kessel, 1995). Such software is in widespreaduse in studies in
which text analysis is a key feature.The use of an indexing tool such as NUD?IST
offers a form of replicabilityfor texturalsearchesnot previously associatedwith
conventionalcontent analysis.
ComplementaryAccounts
Different perspectives,reflectingdifferentobjects of study within the encom-
pasing goal of modelingclassroomlearning,can be used to analyzethe integrated
datasets. Forexample,Clarkeand Helme (1997) used the datain Episode2 to dis-
cuss the role of negotiationin the resolutionof uncertaintyin mathematicsclass-
rooms.Episode2 is also amenableto analysiswith respectto the severalinstances
of studentmetacognitioncontainedwithin the text (Lines 13, 20, 25, 35, 38) and
the functionof these self-evaluativereflectionswithinthe learningprocess.
These two accounts of Episode 2-one from the perspectiveof the resolution
of uncertaintyand one from the perspectiveof studentuse of metacognition-are
not in competition; they representcomplementaryinterpretationsof the same
integrateddata set. Such complementaryaccountshave the potentialto be mutu-
ally informingand to constitutein combinationa richer portrayalof classroom
learningthan would be possible by the considerationof either account separate-
ly. The acknowledgementof the relativistcharacterof any particularaccount of
social activity commits us to defining explicitly the nature of our focus of
inquiry,and those terms we employ in the analysis and comparisonof the vari-
ous accounts, since we are no longer able to presume any specific (absolute)
meaningby implicit appealto the authorityof cultureor the conventionsof use.
A key design element in the complementary accounts methodology is the
bringing together of a researchteam with sufficiently diverse expertise to ade-
David J. Clarke 1ll
quately implement this approach.The research project from which the exam-
ples in this chapterwere taken used an internationalteam of more than a dozen
university academics with expertise in mathematicseducation, developmental
psychology, sociology, epistemology, values analysis, motivation,mathematics,
science education, children's conceptual frameworks, metacognition, gender,
and a range of qualitativeand quantitativeresearchmethodologies. Ratherthan
seek a consensus interpretationof an event, an episode, or an interaction,indi-
vidual members of the research team were encouraged to interpretthe docu-
mented interactionfrom their own distinct, carefully articulatedtheoreticalper-
spective and use their particularselection for focus of study. The goal of such a
process is complementarityratherthan consensus, and each researcher'sinter-
pretationis accordedparity of status, subject to the same criteriaof coherence,
consistency with the videotape data, and plausibility.
CONCLUSION
In the researchdiscussed in this chapter,an attemptwas made to optimize the

use of currentlyavailable technology throughthe synthesis of classroom video-
tape and interview data in an integratedvideo and text document.An important,
possibly essential, perspectiveon the classroom was obtainedfrom the students
themselves in interview situations,in which the significanceof classroomevents
and their associated thought processes were reconstructedby the studentswith
the assisting promptof the classroom video record.The analysis of the resultant
data was enhancedby a frameworkfor text analysis that distinguishedepisode,
negotiative event, and utteranceand the use of an indexing tool with the capa-
bility to undertakecomplex analyses of textual data.
Recent developments in educational research (and in learning theory) have
led to the acceptanceof the idiosyncraticand legitimate subjectivityof both the
research subjects and the researcherand to the consideration of what can be
learnedfrom the comparisonof the multiple stories compiled from the accounts
of the various participants in the social setting and from the reconstructed
accounts of a research team. Judgments regarding the relative merit of one
account over anotherrelate to the purpose for which the comparison is being
made and do not call into question the value of either accountwith regardto any
otherpurpose.I would like to suggest thatit is throughthe accumulationof such
complementaryaccounts in relation to a common integrateddata set that our
portrayal of classroom learning will approach the complexity of process we
seek to model.
112
Chapter 8
The Centrality of the Researcher:

Rigor in a Constructivist Inquiry into
Mathematics Teaching
Barbara Jaworski
This chapter is about the research methods employed in a study that explored
the nature of the teaching in a setting that might be called an "investigative
approach" to the learning of mathematics (Jaworski, 1994). There were strong
parallels between the teaching method and the investigative nature of the
research itself. Both were embedded in a theoretical base of radical and, subse-
quently, social constructivism. The research methodology was broadly ethno-
graphic, using data-collection techniques of participant observation and inter-
viewing and verification techniques of triangulation and respondent validation.
It was conducted from a researcher-as-instrument position; in other words, the
main instrument in both data collection and analysis was the researcher. This was
both inevitable and a source of serious issues, particularly where validation of
interpretations and emergent theory were concerned. The meaning of research
rigor in this context is central to the discussion that follows.
THE STUDY
I begin by briefly describing the theoretical background for the research. This
description introduces needed terminology and raises some questions and issues
that will be addressed in the subsequent discussion.
The Theoretical Background to the Research

The term investigative approach is one that gained some popularity with
respect to mathematics learning and teaching in the United Kingdom during the
1970s and 1980s. It was not well defined, but was related to the use of mathe-
matical investigations and inquiry methods in the classroom. The research study
was undertaken to explore more precisely what such an approach involved, in
terms of classroom practices and teachers' thinking, and what issues it raised for
I should like to thankDoug Grouws and Anne Watson for their very helpful comments on
an earlierversion of this chapter.
BarbaraJaworski 113
the mathematicsteacher.A purposeof the researchwas to try to characterizean

investigative approach.
It seemed obvious from the inception of the project that the subjects of the
researchwould be teachersof mathematicsand theirpupils and thatthe research
would take place in mathematicsclassroomsand theirrelevantsurroundings.An
early issue arose in choosing teachers to participate.Should their teaching be
requiredto be investigativein orderto study its characteristics?If so, what char-
acteristics would one look for in selecting such teachers?The circularityhere
was a problem. This problem was resolved initially by choosing, for the first
phase of the research, secondary mathematicsteachers who were interestedin
exploring what an investigative approachmight mean in their classrooms.
Workingwith these teachersled to a clarificationof the theoreticalbasis of the
research, which resulted in the embedding of the theory of an investigative
approachinto a constructivistview of mathematicalknowledge and learning.
Initially, this was a radical constructivist perspective (e.g., von Glasersfeld,
1987). During the later phases of the research,it developed into a social con-
structivistperspective (Jaworski, 1994). In the radical view, learnerswere con-
ceived of as constructingtheir own knowledge, not as taking knowledge from
outside of themselves. The process of constructionwas seen to be an adaptation
of their existing knowledge (resultingfrom previous experience) to accommo-
date new experience.This constructivistposition implied thata teachercould not
give mathematicalknowledge to students:teaching had to be seen as more than
just a transferof knowledge.
An investigativeapproach,based in this constructivistframework,was seen as
providingexperiences for studentsthroughwhich they could build a broadrela-
tional understandingof mathematics(Skemp, 1976). Implicit in references to
mathematicalunderstandinghere is that students would be (a) encouraged to
addressmathematicalconcepts recognizably standardto most secondaryschool
curriculaand (b) encouragedto develop a conceptual understandinginvolving
wide contextualembeddingand linking of concepts.
The need to move, theoretically, from radical to social constructivismwas
apparentto the researcheras soon as the influence of classroomethos and social
interactionon the development of mathematicalunderstandingswas addressed.
In the classrooms observed, social processes-including school and classroom
norms and interactionsbetween students and between teacher and students-
formeda majorpartof the dataand hence were significantto the analysis. In any
classroom, shared norms and expectations instigate and guide the learning
process. The relationshipbetween social or common knowledge and individual
knowledge constructionwas a source of issues for the teachersconcernedas well
as for the researcher.Two chief issues were involved:
1. The reconciling of knowledge constructedby students in exploratorysitua-
tions in the classroom with knowledge as defined by the curriculum,that is,
knowledge from the thinking and negotiation of mathematiciansthroughout
114 Centralityof the Researcher:Rigor in a ConstructivistInquiry
history,which has resultedin a socially derivedbody of knowledge:Although

socially derived, this knowledge takes on perceived absolutist properties.
Those interactingwith it, teachers and students, may come to regard it as
objective and external to human endeavor. Noddings (1990) addresses this
issue from an epistemologicalviewpoint.
2. The interactiveconstructionof knowledge in the classroom: It seemed that
throughdiscussion and negotiationbetween participants,knowledge grew in
the social domain.However, the statusof this knowledge is problematicfrom
a constructivistperspective,particularlyfrom a radicalview. This phenome-
non is importantto a social constructivistperspectiveand needs to be taken
into account (Ernest, 1991; Jaworski, 1994).
These two issues resulted from observations of teaching and learning in
classroom events and rationalizationwith a constructivist theoretical perspec-
tive. They figured strongly in the analysis of data. Practical manifestationsof
the theoretical issues were significant in illuminating and challenging estab-
lished theory.
Since the researchprocess itself is designed to generateknowledge, it too can
be regardedfrom a constructivistperspective. What is learned can be seen to
derive from the constructiveprocess of the researcher.In qualitativeresearchof
an interpretativenature, this knowledge depends heavily on a reconciling of
interpretations,or intersubjectivity,between participantsin the research. The
research might therefore be considered social constructivistin nature. Such a
rationalecan lead to very dangerousgroundwhere researchrigor is concerned.
For example, how is it possible to validate conclusions that are no more than a
synthesis of interpretations?Hammersley(1993) has pointed out the potential-
ly slippery slope leading from researchinto fiction or ideology. However, given
the epistemological base outlined previously for the scrutiny of mathematics
teaching, it seemed unavoidable that the research would be conducted from a
constructivist perspective. It is perhaps unsurprising,therefore, that issues of
validity dominatedmuch of the analysis and, ultimately, the synthesis that led
to conclusions.
The Practical Basis of the Research

In the second and thirdphases of the research,teacherswere selected to fit the
outlined theoreticalframework.They were chosen because they seemed to be
developing an investigative approachin their teaching. In some cases this was
their declaredaim, and in others it was a judgmentof the researcherafter initial
observations.
The study of the practiceof selected teachersled to the identificationof prac-
tical manifestationsof the outlinedtheory and subsequentlyto the theory's clar-
ification, modification,and enhancement.As indicated,throughoutthe research
there was a symbiotic relationshipbetween theory and practicethat is centralto
the methodologicalissues addressedin this paper.
BarbaraJaworski 115
In the research,I wished to characterizeinvestigative teaching on the basis of

its practical manifestationsin classrooms. I therefore decided to look toward
ethnographyas a methodologicalapproach.An ethnographicapproachinvolved
a classroom observerstudyingand trying to make sense of the whole activity of
a classroom-perhaps beginning with a clean slate and writing on it some
descriptionof what was seen to occur that would then form the basis of future
analysis. I learnedthatthis was a somewhatsimplistic view of ethnography,par-
ticularlywith respect to the relationshipbetween interpretationsof observations
and the theoreticalperspectivesof the research.No researchergoes into a class-
room as a blank slate. There are always theories and preconceptions, often
implicit, that guide observationand that need to be made explicit for the rigor of
the research.
In my study, developmentof methodologyand awarenessof theoreticalimpli-
cations went hand in hand. I had to start from the simplistic view, previously
expressed, and learn from the questions that arose as I proceeded.This necessi-
ty is supportedby Ball (1990), who writes the following:
Theprimeethnographic skillscannotbe communicated or learnedin a seminarroom
or outof thetextbook.Studentscanbe prepared, forewarned,oreducatedin ethnogra-
phy,buttheonlywayto learnit is to doit.Theonlywayto getbetteratit is to domore
of it. Mypointis thatethnographicfieldworkreliesprimarilyontheengagement of the
self [italicsadded],andthatengagement canonlybe learnedenactively.(p. 157)
Ethnographyseemed to allow the constructionof knowledge regardinginves-

tigative teaching from observationsof the practice of teaching. This led to the
interrogationof my own previousknowledge and experiencethat formedbroad-
ly my theoreticalbase. The "engagementof self' was from the beginning a cen-
tral featureof the research.
The field work took place over 4 years, in threephases. I studiedsix secondary
school mathematicsteachers,two in each of three schools, duringa period of 6
to 9 months for each school. Typically, I visited a school for 1 day each week.
Each teacherselected one class of students,in the 11 to 15 age range, for obser-
vation. The first phase acted as a pilot study in which researchquestions were
refined and research methods evolved. I engaged in participantobservationof
lessons and informalinterviewingof teachersand studentsas my chief methods
of data collection. My levels of participationvaried from being the teacher to
being as unobtrusive as possible in observing the lesson of another teacher.
Formsof interviewingvaried,too. With students,I had to ask very specific ques-
tions. With teachers, interviews were often long monologues from the teacher
with minimal interruptionfrom me. Othersinvolved unstructuredconversations,
with participantsoften seeking common understandingsof issues that arose.
Significance
An importantissue for the researchwas thatof "significance."All observation
is selective, if only because when one looks to the left, one misses somethingto
the right. Neithermy field notes nor audio or video recordingscould capturethe
totalityof the classroom.WhateverI recordedin field notes musthave caughtmy
attentionand so had some level of significance. In the case of electronicrecord-
ing, choices had to be made about where to point a camera or place an audio
recorder.When I listened to a recordingof a lesson or interview, some aspects
stood out more than others;thus, they carriedsignificance. A majorpart of my
analysis was to recognize and accountfor items of significance.
The following text is structuredto provide details of issues of significance
throughan example of an episode that was accorded significance. In the main
text, I discuss the methodological concerns, whereas the indented text offers
descriptionsand analyses from the research.
A SIGNIFICANTEPISODEAND ITS ANALYSIS
In this episode, a teacher,Mike, in Phase 2 of the research,stoppedhimself in

the middle of an instructionto his class. My descriptionfollows:
Hebeganwiththewords,"Ingroups,decideon differentthingsto try,andaskewhat
happens?'Whileyou'redoingit...." At this pointhe stoppedandpaused,thenhe
said,"WhatamI goingto askyou to do?"He startedgivingan instruction, seemed
to thinkbetterof it, andinsteadaskedthe classwhatinstruction
he hadbeenabout
to give.Oneresponsefromtheclasswas,"Keepquiet,"whichhe acknowledged with
a nod,butotherhandswereupandhe tookanotherresponsewhichwas,"Askques-
tions."His replywas, "YES!"Otherhandswentdown.It seemedto me thatothers
hadbeenaboutto offerthisresponsetoo. (Jaworski,1994,p. 113)
This episode struckme as being an importantindicatorof Mike's approachto
teaching-thus it had significancefor the research.This significanceneeded clar-
ification andjustification.First,it was importantto be clear aboutwhat occurred
in terms of classroom actions and spoken words. I can supportmy accountwith
referenceto my audio recordingof the event. I also fed back to the teachermy
writingaboutthe event. His agreementwhen he readwhat I had writtenprovided
furtherjustification-respondent validation. Second, I needed to explain the
event's significance for the study. My analysis of this episode follows:
The teacher'swordsseemedto say blatantly,"Guesswhat'sin my mind,"but it
appearedthat most of the class knew the answerto his question-"(You're going to
ask us to) Ask questions!"As I hadn'tknown what he wantedthem to do, I was very
struckby this. A part of his classroom rubricwas that the studentsshould ask their
own questions. He acknowledged in interview after the lesson that he was always
asking them to ask questions,hence they knew that this is what he expected of them
and knew what he wanted without his having to spell it out. When I subsequently
offered him my text to read, for respondentvalidation,he furthersaid, "I believe I
did this deliberatelyto stress the 'you can get into my head, and do.' I had not had
them long, remember."(Jaworski,1994, p. 114)
In the preceding words, I have clarified the event's significance for me in
terms of my own experience of it, and I have explained and justified my inter-
pretation by pointing to sources of evidence. It was necessary then to fit the event
BarbaraJaworski 117
into my theoreticalperspective and explain its significance for my study. I shall

say more of this shortly.Although I could be accused of indulging in subjectiv-
ity-perhaps of creating a fiction-if I were to offer my interpretationalone, it
is the fitting of the event into its situationalcontext as fully as possible thatpro-
vides its credibility.Is the accountas writtenvalid in the eyes of the reader-ini-
tially the teacher, but eventually someone external to the research?A reader
needs to know on what basis interpretationsare made in orderultimatelyto judge
the validity of what is presented.
Interpretation,Reflexivity,and Rigor
As I have indicated, an importantpart of my analysis was the attributionof
meaning in classroom situations. In this context (of qualitative, interpretative,
and possibly constructivist research), Burgess (1985) claims that the main
researchinstrumentis the researcher-the researcheris central to the research.
The researcher'stask is to seek out the meaningsin a situationwith referenceto
declared interests or goals. Cohen and Manion (1989) state, "One can only
impute meaning to [experiences]retrospectively,by the process of turningback
on oneself and looking at what had been going on" (p. 32). This reflectionof sig-
nificance back to the goals of the researchis known as reflexivity.Thatis, reflex-
ivity is a to and fro movement between the researcherand the research-a con-
stantquestioningand critiquingof observationsand analyses relative to the total
situationand context in which the researchtakes place. The personaltheories of
the researcherare one aspect of this totality.
I identified a reflexive cycle in analysis and synthesis throughoutmy research,
whereby interpretationswere centralobjects. Where possible, I sought to verify
accounts by triangulatingdata-seeking the interpretationsof the other partici-
pants in an event, for example, teacher and student, or of another observer.
However, Cicourel (1973) points out limitationsin this statement:
It is difficultfortheobserver"toverifyhis interpretation
of theothers'experiences by
checkingthemagainsttheothers'ownsubjective Theobserver
interpretations."... is like-
ly todrawonhisownpastexperiences asa common-sense actorandscientific
researcher
to decidethecharacter of theobservedactionscene.(p.36,citingSchutz,1964)
Cicourelcites Schtitz(1964), who goes furtherin recognizingthe special posi-
tion of the researcher:
Theobserver'sschemeof interpretation
cannotbe identical,of course,withtheinter-
pretativescheme of either partnerin the social relationobserved.The modifications
of attention which characterizethe attitude of the observer cannot coincide with
those of a participantin an ongoing social relation.For one thing, what he finds rel-
evant is not identical with what they find relevantin the situation.Furthermore,the
observer stands in a privileged position in one respect: he has the ongoing experi-
ences of both partnersunder observation. On the other hand the observer cannot
legitimately interpretthe "in-order-to"motives of one participantas the "because"
motives of the other, as do the partnersthemselves, unless the interlockingmotives
become explicitly manifestedin the observablesituation.(p. 36)
Cicourel furtherclaimed that the researchercan "only objectify his observa-

tions by making explicit the properties of interpretativeprocedures and his
reliance on them for carryingout his researchactivities"(p. 36).
Such interpretations,and the issues involved in making them, were the sub-
stance of this research.It was my task in presentingthem to the readerto make
their basis explicit. Cicourel's use of objectify was interesting to me because
this seemed to mirrorthe sense in which I sought to avoid subjectivity. The
word objectivityis redolent of positivist research,and in constructivistterms it
is not definable because true objectivity in terms of knowledge external to an
individual or social group can never be known (von Glasersfeld, 1987). The
term intersubjectivityhas been used by Ball (1982) and othersto capturea sense
of common knowledge arising from group negotiation in sharing and compar-
ing interpretations.Shared meanings within a mathematics lesson might be
regarded as mathematicalintersubjectivity,central to an epistemology of the
social constructionof mathematicalknowledge. In a researchcontext, intersub-
jectivity involves the shared meanings that can be seen to develop through
processes of interviewing, triangulation,and respondentvalidation. It derives
from conversationand negotiation between participantsin the research,which
includes the researcher.
Such a reflexive accountingprocess in which intersubjectivityplays a partis
the "rigor"that Ball (1990) speaks of when she talks of a researchbiography:
Theproblemsof conceptualizing qualitativeresearchincreasewhendata,andthe
analysisand interpretation
of data,are separatedfromthe social processwhich
generatedthem.In one respect,the solutionis a simpleone. It is the requirement
for methodologicalrigorthatevery ethnography be accompaniedby a research
biography,thatis a reflexiveaccountof the conductof the researchwhich,by
drawingon fieldnotesandreflections,recountsthe processes,problems,choices,
and errorswhich describethe fieldworkupon which the substantiveaccountis
based.(p. 170)
So, in reportingmy analysis of the episode of Mike, a research biography
requiresdetails of the incidentitself, the classroomcontext in which it occurred,
the environmentin which this classroom was situated,the teacher's interpreta-
tion of the event, the teacher's comments on my analysis of the event, the rea-
sons for this event's significancein termsof my theoreticalbase, and the relation
of my analysis to my own experience as a practitionerand researcher.
I found this complexity of detail challenging and fascinating,but a majordis-
advantagewas its lengthy natureboth in the time requiredfor analysis and in the
space taken to present an account. This latterconsequence meant that very few
episodes could be reportedin detail, and many that could have contributedvalu-
ably to conclusions were left out. For example, from two school terms of obser-
vation of lessons of a teacherand class, only three or four of those lessons were
discussed in any detail in the reportof the research.This made critical demands
in ensuringthatthose episodes selected were sufficiently generic to representthe
validity of the theory they supported.
BarbaraJaworski 119
Generalizibilityand TheoryGeneration
In additionto fieldnotes, transcriptsfrom audio- and videotapes,and one set of
questionnairedata of student's views of mathematicslessons, I had as data my
own reflective notes writtenthroughoutthe study. These consisted of day-to-day
jottings regardingincidentsI had experiencedand my own ideas and perceptions.
Sometimesthey were elaborationsof anecdotesthathad significance. Sometimes
they involved incipienttheorizing-expressing patternsI observedor attempting
explanations.Eisenhart(1988) refers to this type of data collection as researcher
introspectionin which "theethnographertries to accountfor sources of emergent
interpretations,insights, feeling, and the reactive effects that occur as the work
proceeds"(p. 106).
The incipient theorizing attemptedto express levels of generality within the
research.A criticism of qualitativeresearchmethods is that it is very difficult to
make and justify generalizationsthat apply to other settings. In-depthresearch
necessarily results in small samples from which it can be hard to extrapolate.
Delamont and Hamilton (1986) addressthis issue by recognizing the difficulty,
yet claiming that some degree of generalizationmakes sense:
Despitetheirdiversity,individualclassroomssharemanycharacteristics. Through
thedetailedstudyof oneparticularcontext,it is stillpossibleto clarifyrelationships,
pinpointcriticalprocesses,andidentifycommonphenomena. Laterabstracted sum-
mariesandgeneralconceptscanbe formulated, whichmay,uponfurtherinvestiga-
tionbe foundto be germaneto a widervarietyof settings.(p. 36)
In my own study, it was importantto consider how far the classroom charac-
teristics I found significant were indicative of investigative approachesmore
generally or were of relevance to other teacherswishing to interpreta construc-
tivist philosophy in mathematicsteaching. Furlong and Edwards (1986) make
this comment:
Althoughthe ethnographer is committedto havingas opena mindas possibledur-
ing his periodof observation, it is inevitablethathe will beginhis workwithsome
preconceptions andsomeforeshadowed problemswhichwill leadhimto payatten-
tionto certainincidentsandignoreothers.If he presentshis observations as "objec-
tive description,"
he is probablynaivelyunawareof his ownselectivity.Ontheother
hand,if he followsa theorytoo closely,he will be accusedof selectingobservations
to supporthis ownpointof view. (p. 54)
There seems to be some skill required in weaving a path between the two
polarizationsexpressed here, and I was very much aware of the implicationsof
this for my own work. I recognizedpotentialfor what Glaser and Strauss(1967)
refer to as "an opportunisticuse of theory,"which they call "exampling":
A researchercan easily find examplesfor dreamed-up, speculative,or logically
deducedtheoryafterthe ideahas occurred.But sincethe ideahasnotbeenderived
fromthe example,seldomcan the examplecorrector even changeit (evenif the
authoris willing),sincetheexamplewasselectivelychosenforits confirmingpower.
Thereforeone receivesthe imageof a proofwherethereis none,and the theory
obtainsa richnessof detailthatit didnotearn.(p. 5)
For example, in the early stages of my researchI was exhilaratedby the way
that radical constructivism (then, a theory that I had only just encountered)
seemed to underpinmy perceptionsof an investigativeapproachto mathematics
teaching. It was possible to look at examples of students'mathematicalthinking
as they arose in the lessons I observed and cast these in radical constructivist
terms. However, it then became necessary to look critically at the relationships
involved to see how a theory such as radicalconstructivismwould fit the com-
plexity of my researchas a whole.
Consideragain the episode of Mike's requiringhis class to ask questions.This
episode emerged, in my analysis, as significantfrom the lesson of which it was
part;therefore,some selection had takenplace from the dataat this stage. I need-
ed to accountfor this significance theoretically.The theoreticalsignificancehad
two levels: (a) it fit with my own constructivistperspectiveof knowledge growth
and hence learning, and (b) it contributedto emergent theory of the practical
implications(for a mathematicsteacher,for example) of a constructivistview of
knowledge and learning.
I thus had to justify my analysis at these levels. This justification is summa-
rized in the indentedtext that follows:
The process of asking their own questions encourages students to become
immersedin the ideas thatthe teacherwantsto be the focus of the lesson. From
asking questionsand the resultinginvestigation,studentsgain ownershipof the
mathematicsthey generate,whichprovidesan experientialgroundingfor synthe-
sis of particularmathematicalideas. The teacher'sapproachfostersquestioning
and investigatingand,moreover,an independenceof thinkinganddecisionmak-
ing thatcan lead to studentstakingmoreresponsibilityfor theirown learning.
This analysis rests on the constructivistview that (a) students' own con-
structionsare central to their developing mathematicalconcepts and (b) the
practicalteachingacts (such as requiringstudentsto ask questions)put empha-
sis on students' own constructionsto make these more evident in students'
knowledge growth.
Clearly, some statementshere need furtherexplanationandjustification (e.g.,
the importanceof experientialgroundingand its practicalmanifestations).It was
centralto my researchto providethis kind of critique,thatis, the requisiteexpla-
nations andjustification.There were various levels of complexity. One of these
involved the characterizationof an investigative approachin terms of "asking
questions,""ownershipof ideas," and "takingresponsibilityfor own learning."
Were these constructs emergent from the data, or did they accord with the
researcher'spreconceptions?Was the researcherengaged in the productionof
grounded theory or in a process of exampling? Another level of complexity
involved reconciling observationswith theory such as radicalconstructivism-
for example, in seeing the teacher'semphasison students'askingtheirown ques-
tions and makingtheirown decisions as contributingto students'constructionof
mathematicalconcepts.
BarbaraJaworski 121
Addressing these questions and issues requireda very detailed study of the
research data. A disadvantage of trying to provide a flavor of the research
process througha particularepisode is that, by its very nature,the episode can-
not carry a sense of the interweavingof observations,perceived attributes,and
analyticalcategories.It cannot show, for example, how "askingquestions"relat-
ed to "patternspotting" and "making conjectures"in other lessons and other
classrooms. It cannot show how this teacher's approachin this lesson compared
or contrastedwith approachesin his other lessons or in other teachers' lessons.
One episode cannot show that one teacherworks accordingto an investigative
approach,let alone have consequences for describing an investigative style of
teaching more generally. The overall research synthesis demands a rational
weaving of such researchoutcomes and a clarityof presentationthatallows other
researchersto judge its validity. Where an episode is concerned,I can say little
beyond what this teacheraimed to achieve and what seemed to occur in his class-
room. However, by taking many such episodes from differentlessons of differ-
ent teachersit is often possible to see a patternof interactionsin which students
question and investigate mathematicalsituations and in which the groundwork
for synthesis of mathematicalconcepts is prepared.Subsequently,it is possible
to take these practicalmanifestationsof aspects of theory and flesh out the theo-
ry. This is the symbiotic process that I describedearlier.
In summary,initial theory gives startingpoints for observationand selection.
Episodes selected are rich in details thatthe theoryis too narrowto predict.From
this richness, patternsemerge that not only substantiatethe theory, but make
clearer what such theory means for the practice of teaching and learning. This
enhancedtheory can then be reappliedto furtherpracticalsituationsfor substan-
tiation and enrichment.
THE THEORY-PRACTICEINTERFACE
In my research,an example of theory arising from data was a theoreticalcon-
structI called the Teaching Triad (Jaworski,1992). This constructarose from a
close scrutiny of all the data from one teacher, which involved categorizing
attributesand classifying emerging patterns.It was possible to characterizeher
teaching under three headings: managementof learning, sensitivity to students,
and mathematicalchallenge. These categorieshad distinctas well as interrelated
properties.This was theorygeneration.The teachingtriademergeddirectlyfrom
the data.I conjecturedthatthis teacher'steachingcould be characterizedthrough
the teaching triad. I tested the triad on furtherlessons that had not been part of
the original analysis to see whether these also fit with the triad or whether the
triadcould offer a characterization.Considerableevidence supportedthe triad's
potential to characterizethis teaching. It was then importantto test the triad
againstotherteaching to see whetherit had potentialbeyond one teacher.Again,
evidence suggested it had. The next stage was to rationalizethe teaching triad,
an emergentconstruct,with the theoreticalbasis of the research,a constructivist
view of knowledge and learning. Details of this rationalization can be found in

Jaworski (1994).
Here, I am concerned with the methodological processes involved and shall
explicate these further by showing how an analysis of a classroom episode (quot-
ed from Jaworski, 1994) provided support for the teaching triad as a construct to
characterize the teaching observed.
The Case of Packaging
The class was beginning a project on packaging. In the first lesson of the pro-
ject, the teacher, Clare, and the students had brought into the classroom packages
and bottles of various kinds from domestic products. Clare had organized a
brainstorming session with the class, seeking questions that might be explored
with regard to packaging of various kinds. A set of twenty questions had result-
ed, and Clare had photocopied a sheet of these for each member of the class.
Students were encouraged to start by exploring a question of their own choice.
The videotaped episode consists of about 5 minutes of class time from the sec-
ond lesson of the project. A number of girls sitting together at a table had chosen
to work on the questions "Which shapes are scaled down versions of other
shapes? How can you tell? How can you check?" They had identified three vari-
ables, volume, surface area, and shape, that they were trying to relate to one
another. They had decided that they needed to fix one of these variables in order
to explore the other two. The one they decided to fix was shape, and they decid-
ed to make it a cuboid. A transcription of the episode appears in Figure 8.1. An
analysis of the episode follows.
It's a Cuboid
This involves a group of girls, including Rebecca and Diana, who were working on ques-
tions relating to volume and surface area using a large collection of packets from com-
mercially producedproducts.The teacher, Clare, listened to their conversationfor some
moments, and then interjected:
(1) C1 We're saying, volume, surface area and shape, three, sort of variables, variables.
And you're saying, you've fixed the shape-it's a cuboid. And I'm going to say to you
[pause]hm
[She pauses and looks around.]
Cl I'll be back in a minute
[but she continues talking.]
C1 Thatis a cuboid.
[She picks up a tea packet.]
C1 Thatis a cuboid.
[She picks up an electric light bulb packet.]
(5) C1 and...
[She goes away-then returnswith a meter rule.]
(table continues)
Barbara Jaworski 123
Cl This is a cuboid.
[She looks aroundat their faces. Some are grinning.]
Cl And you're telling me that those are all the same shape? [Everyonegrins.]
(8) Reb Well, no-o. They've all got six separatesides though.
Cl They've all got six sides. But I wouldn't say that that is the same shape as that.
(10) Reb No-o
Cl Why not?
Di Yes you would ...
[Thereis an inaudibleexchange between the girls D and R.]
Cl What's different?
[Thereare some very hardto hearresponseshere. They include the words size and longer.]
Cl Different in size, yes.
[Clarereachedout for yet anotherbox, a large cereal packet,which she holds alongside the
small cereal packet.]
(15) Cl Would you say that those two are differentshapes?
Reb They're similar.
Cl What does similar mean?
Reb Same shape, differentsizes. [They all laugh.]
[Duringthe last four exchanges therewas hesitancy,a lot of eye contact,giggles, each per-
son looking at othersin the group,the teacherseeming to monitorthe energy in the group.]
C1 Same shape but differentsizes. That's going roundin circles isn't it? [R nods exag-
geratedly. Others laugh. Teacher laughs.] We still don't know what you mean by shape.
What d'you mean by shape?
[She gathers three objects, the two cereal packets and the metre rule. She places the rule
alongside the small cereal packet.]
(20) Cl This and this are differentshapes, but they're both cuboids.
[She now puts the cereal packets side-by-side.]
C1 This and this are the same shape and different sizes. What makes them the same
shape?
[One girl refersto a scaled-downversion. Anotherto measuringthe sides-to see if they're
in the same ratio. Clarepicks up their words and emphasises them.]
(22) Cl Right. So it's about ratio and about scale.
Figure 8.1. Cuboid transcript(Jaworski,1994).
The episode seems to split into three parts, which I characterize as follows:
statements 1-7, the teachers' initial challenge; statements 8-14, students'
engagement; and statements 15-22, further challenge. In terms of the teaching
triad, parts 1 and 3 show strong elements of mathematical challenge, part 2 pro-
vides evidence of sensitivity to students, and the ethos within which the episode
occurs is indicative of the teacher's management of the learning environment.
At the beginning of the episode, the teacher has listened to the students' dis-
cussion and made a decision to intervene. From her point of view, the problem
seemed to be thata cuboid was not mathematicallyappropriateto theirneeds and

she wanted somehow to draw their attentionto this. She set up a dramaticlittle
scene in which she asked them to comparethreedifferentcuboids. Her departure
to get the meter rule addedto the dramabecause there was a pause between her
pointingto the two boxes andthen returningwith the rule. Hertone was provoca-
tive. The girls' attentionwas captured.There were half-smiles, almost as if they
were asking, "Whatis she up to?" When she producedthe rule they grinned.It
seemed obvious that the rule was not the same shape as either of the boxes.
I saw this situationto be similar to Inhelderand Piaget's (1958) rods experi-
ment, which was designed to get studentsto considerrelationshipsbetween vari-
ables. In theirexperiment,researchersasked only neutralquestions,not tryingto
teach but to elicit evidence of students' thinking and understanding.However,
here there is a teaching act to be considered, and this might be seen ratherin
terms of Vygotsky's (1978) Zone of Proximal Development (ZPD)-judging
students'potentialfor makingprogressand providingthe necessary scaffolding.
The teacher, it seemed, believed that students would not make the required
progress without some interventionfrom her. Thus, she influenced their devel-
opment, allowing them to progressin a way that might not otherwise have been
possible. Of course we can never know what would have been possible without
the intervention,which is one of the difficulties of using the ZPD constructin a
practicalarena.
One interpretationis thatthe teacherwas aggressive. She was denying students
the opportunityto formulatefor themselves that this notion of "shape"is too
imprecisea variable.She was forcing onto them her perspective,forcing the pace
of theirthinking,directingthem mathematically.Asking more neutralquestions,
correspondingto the Piagetian position, would have left the girls to move for-
ward only as their own thinkingallowed.
An alternativeinterpretationis thatthe teachersaw the girls' thinkingas fuzzy
and not seeming to make much progresstowardrelatingthe variables.She could
see an opportunityto focus theirthinkingin a way thatwould lead into some use-
ful mathematics. She had to decide whether to push them in this direction.
Having made the decision (for whateverreason),ratherthanoffer an explanation
of why cuboids would be too imprecise, she set up a provocative situationand
challenged them with an apparentcontradiction,capturing their interest and
attention.This might be seen as promotingconflict discussion (Bell & Bassford,
1989), or, in radicalconstructivistterms, as creatingconstraintsto challenge the
viability of the girls' currentknowledge (von Glasersfeld, 1987). In social con-
structivist terms, the situation could be seen as stimulating intersubjectivity
throughwhich meaningscould develop; in Vygotskianterms, as providingscaf-
folding to enable progress (Bruner, 1985). There was a relaxed and friendly
atmosphere,but at the same time, there was a build up of tension as the contra-
diction became apparent.
Clear evidence of a high degree of mathematicalchallenge can be seen in the
first part of the episode. There was also considerable risk involved on the
BarbaraJaworski 125
teacher's part.The girls might not take up the challenge. It might be inappropri-
ate. They might not be able to cope with it. They might lose their own, perhaps
precarious,thinkingand possibly their confidence. The teacher,in having some-
how to salvage the situation,might increase students'dependencyon her. These
were some of the dilemmas facing Clare as she chose her course of action with
the students.
For Clare,mathematicalchallenge seemed always to be allied to sensitivity to
students.She knew her studentswell, for which I had much evidence. The three
girls had previously demonstratedtheirability to thinkwell mathematically.She
believed that they had high mathematicalpotential. She also had a very good
relationshipwith them. The risks she took were allied to this knowledge. These
factors are all facets of the culturalethos of the classroom and the teacher'srole
in encouragingstudents' mathematicalconstructions.Situationswere createdin
which mathematicscould be discussed-in this case, the packaging questions.
Discussion and negotiationwere actively encouraged,with the teacherproviding
stimulationor provocationwhere she believed it was needed. Such knowledge of
students,creationof tasks, and acts of encouragementand stimulationwere part
of the teacher's managementof the learningenvironment.The classroom ethos
was a productof this managementof learning.
The next partof the episode, in statements8-14, shows a lessening of the ten-
sion as the girls began to think throughwhat had been offered. It was almost as
if they were thinking aloud, ratherthan participatingin discussion. The shapes
all have six sides. However, the meter rule and the bulb box are not the same
shape. How are they different?Well, they are the same in some respects. In this
section the teacherwas less intrusive,perhapsprovidingspace for students'con-
structivethinkingto internalizethe problem,but her remarkswere still focusing.
"Whatis different?"
Success in a situationlike this dependsvery greatly on the teacher's sensitivi-
ty to students' perceptions, both mathematicaland social. In analyzing why I
believed that this episode was successful with regardto the teacher's objectives
and the students'gain, I attributedit to the decision makingby the teacherat var-
ious strategicpoints. Clearlythe teacherhad to make the initial decision to inter-
vene and to do so as provocativelyas she did. However, there was anothercru-
cial decision hovering in the middle stage of the episode. Were the studentsable
to take up the challenge?Could they make progress?Whatelse should she offer?
It was in makingan appropriatedecision here that sensitivity to the studentswas
most crucial. The success, or otherwise, of such episodes is very rarely due to
just chance, but usually involves a high degree of vital decision makingbased on
teaching knowledge and experience (Calderhead,1987; Cooney, 1988).
Finally, in statements15-22, the teacherseemed to decide thatshe shouldpush
further.She chose two cereal packets of different sizes but the same shape and
asked if they were the same. She was rewardedinstantlyas one girl offered the
crucial word, similar. So she pushed harder:"Whatdoes similar mean?"The
reply was not helpful. They were going roundin circles. She diffused the tension
by acknowledgingthis andlaughing,andthey all laughedwith her. However, she

persevered,and in the interchangethat followed, she was given furtherappro-
priatelanguageby the students-ratio and scale.
She could, of course, have gone on to ask, "Whatdoes ratiomean?"However,
she chose to leave it there.Her emphasison ratio and scale, picking up the girls'
own words, was probably sufficient to provide a new startingpoint for their
thinking and progress. It seemed that the girls had entered into her thinking, as
she had initially entered into theirs. She seemed to be convinced that the girls
were involved sufficiently to be able to make progress.
There was evidence, subsequently,that the students' thinking became more
focused as a result of the exchange with the teacher. They moved from vague
articulationsof shape to much more precise ones involving similarity,ratio, and
scale. With a correspondinglymore precise conceptualfoundation,they could be
more likely to make progressin relatingtheiroriginalvariables.Ultimatelysome
assessmentcould be made of the episode in terms of what the girls did next and
where their thinking eventually led. However, judging only this episode, there
seemed to be an effective balance between challenge and sensitivity. The teach-
ing situationseemed to be effective in terms of what the girls gained from it and
what the teachermight have hoped to achieve.
Decision makingin the episode was well founded on knowledge and prepara-
tion by the teacher. She had createda situationin which studentscould engage
in meaningful and potentiallyproductivework. Studentshad a set of questions
on which to start.They had been instrumentalin devising these questions, so the
questions were meaningful to them. They could choose whichever questions
interested them most, which increased motivation. They were encouraged to
work in groups-to articulateand share ideas, to develop intersubjectivity,and
to challenge and supportindividual conceptions. Thus, what the studentswere
engaged in here owed much to the teacher's overall managementof the learning
environment.I suggest that the three elements (managementof learning,mathe-
matical challenge, and sensitivity to students)can be seen in evident interaction
in this episode, characterizingits naturein contributingto students' mathemati-
cal learning.
CONCLUSIONS
The precedingexample illustratesmy analysis of one episode from classroom

practice against a theoretical construct that arose from earlier analysis. As a
result, it is likely that the theoretical construct-the teaching triad-became
clearer throughits practical manifestations.It now becomes available to other
researchersor practitionersto test againstothertheory or research.The detail of
such manifestationsand the bases of interpretationsandjudgmentsare crucialto
this availability.
The relation between ethnographyand theory raises thorny issues for ethno-
graphicresearch.Hammersley(1990, p. 102) arguesthat too little theory results
BarbaraJaworski 127
from such research,perhapsas "an over-reactionto positivism."He claims that

ethnographicresearchin schools puts much emphasison qualitativedescriptions
of behavior, supportedby extracts from field notes or transcripts,but little on
explanationsfor patternsdiscovered. He suggests that for the developmentand
testing of theory to be pursuedeffectively, the researchfocus has to be narrower
than is often the case in ethnographicresearch.
There is a serious tension here for the researcher.In my case I wished to study
the natureof an investigativeapproachin mathematicsteaching.It was necessary
to be sufficiently aware of what this term meant in theory in orderto select sub-
jects for research. This was a narrowing;I did not go out to observe just any
mathematics teaching. One possibility for following up my initial selection
would have been to devise a coding schedule whereby I could identify, in class-
rooms, aspects of an investigativeapproachthatmy theorypredicted.This would
have been a furthernarrowingof focus and could have resultedin my finding lit-
tle more to characterizean investigative approachin practicethan was predicted
by my theory. Thus, it seemed importantto study the classrooms in as open a
manner as possible, allowing for the diversity and richness, but also for the
researcherbeing overwhelmedby too much complexity. It was the recognition
of significance within this complexity that allowed furthernarrowingto occur
and subsequentlyled to theory generationand enrichment.
Despite talk of narrowing,it is importantto recognize, when talking of valid-
ity, thatthere is no way in which any of the interpretationsor conclusions in this
research could be regardedas right or true. Validity, here, does not have the
objective meaning it takes from positivistic research.One might arguethat such
meaning often derives from unrealisticnarrowingand definition.Researchrigor
in this study lay in embeddingresults in their fully situatednatureand context,
and in making the details of this embedding overt. The centrality of the
researcherand the resultingimplicationsmust be judged throughthe validity of
what is offered in this and other writings, since validity resides ultimatelyin the
degree to which an informedreaderis convinced by what is written.
128
Chapter 9
Using a Computer in Synthesis of

Qualitative Data
Judith Mousley, Peter Sullivan, and Andrew Waywood
Forms of educational inquiry are shaped not only by the traditions of scientif-
ic and naturalistic research but also by researchers' epistemological assump-
tions-that is, by their beliefs about the nature and scope of knowledge itself. If
it is believed that knowing is a matter of saying what is objectively real, then
researchers will aim to describe and measure observable entities, to explain
these, and then to suggest ways in which the resulting understandings might best
be used. Such a belief about knowledge leads to the selection of relatively objec-
tive research methods and data, with researchers attempting to remove them-
selves from the subject matter of the study. However, if it is believed that know-
ing is a matter of having a stance within a world, then researchers will aim for
new ways of participating in that world through the construction of tentative the-
ories and through demonstrating an openness to changing understandings and
alternative viewpoints. Such a belief about knowledge is likely to lead to
research methods and data that are more varied and more qualitative in nature.
Contrasting beliefs about what it is to know are linked with different under-
standings of terms like understanding. Such serious words are difficult to define
in that they come to have meaning only as they are situated (uttered) in particu-
lar discourses, such as in the alternative discourses currently vying for
researchers' allegiances. What was initially a contrast of style between research
in the physical sciences and that in the human sciences has become a method-
ological divide, and the notion of what it is to understand has become a measure
of the very depth of this divide. In the physical sciences, the term clusters with
words like explain, cause, grasp, and discover. However, in the human sciences,
it clusters with interpret, reason, express, and experience. This contrast relates
closely to discourses about knowledge in that the gaining of understandings in
the physical sciences is about researchers grasping reality (e.g., by measuring or
predicting an event), whereas in the human sciences it is about researchers being
grasped by reality (e.g., by experiencing and interpreting an event). Although the
former act necessitates a stance outside the event, the latter creates changes in the
lifeworlds of researchers-altering lived experience in fundamental ways.
On the assumption of a dialectical relationship between understanding and
interpretation, what it is to interpret is to be similarly defined within these con-
JudithMousley,Peter Sullivan,and AndrewWaywood 129
trastingdiscourses. Gadamer(1975) explains this relationshipwhen he claims

the following:
is not an occasionaladditionalact subsequentto understanding,
Interpretation but
ratherunderstanding is alwaysan interpretation,and hence interpretationis the
explicitformof understanding.Inaccordancewiththisinsight,interpretivelanguage
andconceptsarealsoan innerstructural elementof understanding. (p. 274)
Gadamer'sposition moves the whole problem of language "from its periph-
eral and incidental position into the center of philosophy" (p. 274) and leads
him furtherto question whether a researchercan escape the constraintsof lin-
guistic use. He questions whether language might performthought and claims
that this raises
fundamental doubtsaboutthe possibilityof ourescapingfromthe sphereof influ-
ence of oureducationwhichis linguistic,of ourcivilizationwhichis linguisticand
of ourthoughtwhichis transmittedthroughlanguage,as well as thedoubtaboutour
capacityforopennessto realitywhichdoesnotcorrespond to ouropinions,ourfab-
rications,ourpreviousexpectations.
(p. 491)
It might be assumed that using technology in qualitativedata analysis in the
human sciences would encouragemovement of researchersaway from the sub-
jects of their studies as well as add an element of objectivity to processes of
analysis and synthesis in that it would control the effects of human experience
and operate outside human theoretical, methodological, and political frame-
works. However, this can never be the case. The act of programminga comput-
er, as well as such researchprocesses as deciding on what will be considereduse-
ful data, resides within those frameworks.The beliefs that a programdesigner
holds about knowledge, understanding, and interpretation, aside from those
about effective research processes, shape the software produced and, conse-
quently, shape the products(includingdata,interpretations,and explanations)of
the projectsfor which it is employed.
Some developersof computerprogramsfor dataanalysis recognize that value-
laden and subjectivechoices are made at every instanceof a researchprojectand
so have createdenvironmentsthatfacilitatethe recordingof these choices as well
as the keeping of notes as new understandingsdevelop (or new theories are
built). These recordscan make the role of the researchermore open to examina-
tion and, hence, renderthe role of human agency more recognizable.One such
computerpackage, an applicationdesigned to be used as a tool in the analysis
and examinationof qualitativedata, is NUD*IST.(Richards& Richards,1990).
OUR USE OF NUDoIST
NUD.ISTstands for Non-numericalUnstructuredData Index Searching and

Theorizing.The version used was the most recent available at the time (version
2.3). Later versions have improved graphic displays, menu bars, and dialog
boxes to make the whole process more user friendly.
130 Using a Computerin Synthesisof QualitativeData
This chapterfocuses on how using NUD*ISTmade us more aware of the theo-

ries and expectationsthat we broughtto the analysis task. Ourworkingas a team
to use this softwarefor handlingresponses to a survey raised our consciousness
of roles that researchers,and tools they employ, play in shapingthe productsof
educationalinquiry.In this chapter,our focus on a growthof understandingand
on processes of interpretationsignifies our attemptto stay close to the text of
subjects,and to participatein its meaning,ratherthan to have our discoursepro-
duce an object thatis distancedfrom ourselves. We recognize thatthe productof
our researchprojectis not one of the "things"of the physical sciences-it is an
expressionof a viewpoint and is open to critiqueas well as reworkingby others
as well as to furtherdevelopmentby ourselves.
It is appropriateto describebriefly the processes in using this particularpack-
age. We stress that this descriptionis not meant as an instructionguide and that
we are not arguingthat the programis betterthan othercomputer-basedqualita-
tive analysis tools.
Basically, NUD*ISTis software that allows the managementand organization
of data through indexing. It can handle any digitized text-from one word-
processed sentence to a whole book scanned into the computer's memory.
Although off-line data (e.g., videotapedor audiotapeddata) cannot be handled,
they can be linked (using other software) to digitized data and thus be manipu-
lated by text.
The first key decision is aboutthe way the dataare categorized.Ultimatelythe
data will be entered into a treelike hierarchy,with the data points called nodes
and subnodes.The nodes at the top of the tree may be determinedby the research
design or can be derived from the data themselves. In both cases, the subnodes
will usually be determinedby inspecting the data. One node, and its subnodes,
can be reservedfor factualinformation,such as gender,datasource,and so forth,
so that cross-categorizationcan draw on this at a later date. Note that all of this
can be changed as the analysis unfolds, so researchersare not bound by initial
categorizations.
The data are typed into text files in a particularstyle. One key decision is
whetherdata are to be managedin words, phrases,lines, or paragraphs.The text
units are separatedby a <return>.The files are then importedinto the program.
It can be useful to get a printoutat this stage, since the text units have associat-
ed line numbers.
The categorization structureis then applied to the data. Each text unit is
assigned to one or more of the subnodes througha system of numericalcodes.
This can be done by researchers,or NUD?IST has the capabilityto conductsearch-
es for key words or phrasesand respondsto Boolean commands.This facilitates
the examinationof currentor potentialsubcategories.
NUD*ISToffers options to support analyses from this stage onward. These
include recoding and recategorizingentries and subnodes, adding and deleting
subnodes,analyzingdataby backgroundcategories,addingnew materialinto the
project,and recordingnotes at each stage of the analysis process.
JudithMousley, Peter Sullivan,and AndrewWaywood 131
The next section of this chapter elaborates our use of NUD*IST.It uses one
applicationof the tool in a study by the first two authors.It outlines the stages of
the researchand the steps in the use of NUD*IST. Concurrently,some issues aris-
ing from the use of such tools are discussed as are some of the additionaloptions
available in the program.
THE EVOLUTIONOF THE PROJECT
In the example of the application of NUD?ISTreported in this chapter, we

describe the evolution of the project, the instrumentused, and the development
of a conceptualframework.We attemptto develop a useful frameworkfrom the
raw data and acknowledge the difficulty of accomplishingthis goal.
In 1990, Clements and Mousley reportedthe results of their researchinto stu-
dent teachers' perceptions of the teaching observed during practicumsessions
(Mousley & Clements, 1990). Studentteachershad reported,on questionnaires,
that much of what they were observing in classrooms was the antithesisof both
currenttheory and the behaviors espoused in their preserviceteacher-education
course. It seemed that despite its considerablecost to universitiesas well as the
considerableenergy expendedby teachersand traineeteachers,the work experi-
ence was not enabling studentteachersto observe the regularuse of calculators
and othertechnology, cooperativegroup work, mathematicaldiscussions among
students, problem-based teaching of mathematics, the asking of open-ended
questions, and other nontraditionaltechniques. Further research by Sullivan
involving teacher-educationstudents from other institutions produced similar
results (Mousley, Sullivan, & Clements, 1991).
It seemed likely either that (a) desirablepracticeswere not being implement-
ed in schools, and thus the observationsmade by the teacher-educationstudents
were an accurateportrayalof currentteaching,or that(b) these teacher-education
studentsmay have witnessed some high-qualityteachingbut may not have iden-
tified the featuresof such teachingbecause of the subtle, sophisticated,and com-
plex nature of classroom interaction.Probably,both hypotheses were correct,
with the weighting of factorsvarying among differentteachers,studentteachers,
and individualschools.
This raisedthe questionof how teachereducatorscould preparestudentteach-
ers better for close observationof exemplary practices. The key seemed to be
better preparationfor school-experiencerounds with the honing of observation
skills by using a resourcebase that included examples of exemplaryteaching.
This decision, in turn,led to discussion of which featuresexperiencedteachers
and mathematicseducatorsbelieve are desirablecomponentsof qualitypractice.
It was realized that there would not be complete agreementaboutone set of fea-
tures of quality teaching.The effectiveness of pedagogicaldecisions and actions
depends on factors like their appropriatenessfor the subject matterat hand, the
currentincident,and the particularpupils' needs at any given moment.However,
it was thoughtthatit would be useful to gatherdataon some generallyagreed-on
featuresof high-qualityteachingas identifiedby experiencedteachersand math-

ematics educatorsand then to use these in creatinga resourcebase that included
videotapedlessons.
The "Quality Teaching" Questionnaire

The first step in this process was the developmentof a survey of mathematics
educatorsto seek consensus on some characteristicsof qualityteachingof math-
ematics. Such a questionnairewas preparedand administered.Ourresearchdrew
on both "quantitative"and "qualitative"paradigmsto allow differentbut com-
plementarysets of survey items to provide broaderdata with epistemological
touchstones.The diverse methods of data collection were intendedto make pro-
vision for later assimilation of evidence and for close examinationof patterns
and interestinginconsistencies.
The first section of the survey consisted of one open-responseitem:
We wantyouto imaginea mathematics lesson,at anyyearlevel,wherethestudents
arelearning,forexample,to estimatethemassof variousobjects,orto addfractions,
or to recordgiveninformation as a graph....Pleasewritedownthe mostimportant
whicha qualitymathematics
characteristics lessonon any of theseconcepts/skills
wouldusuallyhave.
Participantswere asked to complete this item first, before they read any cate-
gories presentedin the latterpartof the questionnaire,in orderto capturerespon-
dents' own initial reactionsto the request.
The second, more structured,part of the instrumentused fixed format items.
Seventy-eightpairs of descriptorswere compiled from researchsummaries(e.g.,
Bell, Costello, & Kuchemann,1983), from teaching frameworks(e.g., Good,
Grouws, & Ebmeier, 1983), from recent recommendations(e.g., Australian
EducationCouncil, 1990), and from the earlierquestionnaireused with the student
teachers(Mousley & Clements, 1990). The descriptorswere clusteredunderthe
headingsteachingenvironment,lessonaims,lessoncontent,presentation, class activ-
ities,questions,aids,assessment,andclosure.Descriptorswerepresentedas bipolars.
Forinstance,one set of descriptorslistedin the lesson aimsclusterreadas follows:
there was no clear purpose <-- there was a clear purpose
pupils were aware of the aims <--- pupils were not aware of the aims
the goals were not achieved <- the goals were achieved
the aims were negotiated <- > the aims were imposed
Respondents were asked to (a) write in any importantcharacteristicsthey
thoughtwere missing, (b) delete any pairs that seemed not necessary for quality
mathematicslessons, then (c) mark on the continua whether quality teaching
would be closer to one of the pair of featuresor the other, and (d) rankthe fea-
turesin orderof importance.The analysis of the second structuredcomponentof
the survey is not discussed here other than to note that the data can be linked to
responses to the open-responseitems by using NUD*IST.
Judith Mousley, Peter Sullivan, and Andrew Waywood 133
The survey was first piloted with twelve teachers and teacher educators.To
test the layout, length, and complexity of the form; to determineparticipants'
thinkingduringthe completion process; and to discuss with them possible solu-
tions for any difficulties experiencedin completing the form, these respondents
were observed and interviewedas they completed the questionnaire.
After revision in accordancewith the feedbackgained from the pilot study, the
questionnairewas mailed to threegroupsof people involved in mathematicsedu-
cation. These were mainly colleagues and fellow members of two professional
associations.Forty survey responses came from experiencedteacherswho were
graduatestudents in mathematicseducation (100% return),56 from Victorian
teachereducators(80% return),and 29 from Americanteachereducators(40%
return).These groupswere selectedbecausethey representedan informedview of
currentissues in teaching and learningyet would presentopportunitiesfor com-
parisonsto be made between sets of datafrom differentgroupsof respondents.
DATA ANALYSIS
All 125 responses to the open-response item were word processed. The
NUDoISTprogramproduced a printoutthat identified the respondent'soccupa-
tion, country,and so forth, and the raw data (a list of descriptors,a sentence, or
a paragraph).For example, the printout reproduced as Figure 9.1 was the
response from Number 70, a mathematicseducation colleague from Victoria,
Australia.
*2
70 MESH 2
* QUESTION 1 3
- well-prepared,i.e., lesson structuredappropriately 4
- if relevant,a relatedconcrete/investigative
project/activity 5
- guided interactivediscussion leadingto conceptualunderpinnings 6
- should be related,where possible, to previouslearningand relatedto 7
children'sconceptualexperiences 8
Figure 9.1. The formatof the text.
The first step in the analysis process was to separatethe phrasesinto units that
seemed to convey a particularthought. The phrases were categorized by hand
with a numericalclassification system involving a taxonomy of classroom prac-
tices-a taxonomy not developed in advance of the categorizationprocess, but
resultingfrom discussions as each phrasewas examined.
Initially, some key ideas were identified by inspection of the data. These
formedthe top-level nodes. One node was used for backgrounddata, such as the
country of the respondent,years of experience, and so forth. Other top-level

nodes included pupil activity and communication.Ideas that emerged within
communication,such as clear explanation,became subnodes.
At times, furtherlevels of division were created.New categories also needed
to be formed for text that did not fit into existing categories. Similarly, if we
found thatdatawe had been splittingbetween two categorieswere so similarthat
the divide was somewhat artificial and arbitrary,the overlappingsubcategories
were merged. This process of placing data accordingto emerging ideas and cat-
egories continueduntil each phrasein the responses had been coded.
Key words and common phrases were useful in deciding on how to classify
phrases. For instance, the phrases "sharingideas," "time for sharing percep-
tions," "sharing alternative solutions," and "listening to alternative solution
strategies"seemed to be linked by common words as well as conceptually,so all
were classified as "sharingstrategies."
Figure 9.2 shows a schematicrepresentativeof one section of the framework
tree duringone stage in this process. Pupil's activity representsa top-level node.
Thereare subnodesof this node, which in this case are contentand activities and
interaction,and these in turnhave their own subnodes.
1 2
pupil'sactivity
11 12 13
I
contentand activities
I
interaction
I 17 I 1
I I I I
discussion
investigations/
problemsolving between pupils
Figure 9.2. Schematicrepresentationof one section of the frameworktree.
Althoughwe sketchedout such possible ways of organizingthe dataon butch-

er paper,NUD*ISTallows users to merge, cut and paste, compare,reorganize,and
rename sets of data onscreen-with a log of such theory-buildingstrategies
being maintainedby the program.
Our aim was to create a structurethat could allow categorizationand organi-
zation of the data rather than to devise a picture of some ultimate reality.
Successful developmentof the taxonomy at this stage dependednot on its cor-
rect interpretation(for many equally viable presentationscould have resulted
from the originaldata)but on its usefulness in buildingplausibleunderstandings
and richerinsights aboutelements of qualityteaching.The taxonomycould then
be put to the mathematicseducationcommunityfor discussion and development

of increasedunderstandingin this area.
Putting the data into categories and subcategories was time-consuming but
enlightening. The responses, and the need to decide where they fit, stimulated
probingdiscussions aboutteachingand learning.Therewere some lively debates
about possible interpretationsof responses and the meanings of jargon, such as
nonthreatening.Many problematic aspects of interpretationarose. Are there
essential differences in what respondentscall problem solving andproblempos-
ing, and if so, what might they be? Is the word curriculumbeing used to describe
what is intendedto be learned,what is learned,or the learningprocess? If peda-
gogy expresses a dialecticalrelationship,does it belong more with the teacher's
realm or student's actions? Many such points raised our awarenessof the diffi-
culties thatresearchersface when dealing with the lack of specificity both of our
professionallanguage and of emerging categories.
We also discussed whether the structure arising from the data-reduction
process was largely predetermined-either by our own pedagogical notions or
by the wording of the original question. Although we never suffered from the
delusion that categories for analysis emerge from data untouched by human
hand (Glaser & Strauss, 1967), we became more and more aware of the impact
that our experiences as teachers (including our teachertraining),as teacheredu-
cators, and as readers of currentmathematicseducation literaturewere having
on the shaping of these categories. Our discussions led to a growing awareness
thatour theorizingshapedreality, ratherthanvice versa, and thatour varied (but
similar) prior experiences and preunderstandingslimited possible subjective
and intersubjectivemeaning as well as the ways in which we communicated
about the data.
We were awarethatas soon as researchersreador hear subjects' interpretation
of particularphenomena,credence is given to those data. The new knowledge
then influences the ways in which furtherdata are interpreted.Thus developing
understandingschange researchers'stances (perhapsonly by reaffirmation)so
thatthey can never see raw datarelatingto the same phenomenain the same way.
Thus datachange researchers-they change the way they see as well as what can
be seen. But, as (dman (1988) points out, the synthesizing of qualitativedata is
not merely a process of attempting to mirror other people's understandings;
interpretationand understanding interact closely, since making something
explicit is likely to imply a change in the understandingof it:
Interpretation, andpre-understanding
understanding, aredialecticallycoupledwith
ourexistenceas humanbeings.People'sself-definitions anddefinitionsof life area
resultof theirunderstanding,
andtheirunderstanding developsthroughtheirexperi-
ences.Theunderstanding of theirlives is theultimatehermeneutical
circle.(p. 64)
As the data analysis progressed,there was also concern aboutpracticalissues,
such as the effect of includinga phrasein two differentclassificationsor whether
a particulartermand an example of what was meantshouldbe includedas one or
two entries.We decided that some of the responsesfit neatly into two categories,
and because we were less interestedin the numberof phrasesin each category
than in what was actually said (and what we could use in planning exemplary
lessons), we allowed the methodology to respond to our needs as researchers
ratherthanto be constrainedby traditionsof any researchapproaches.
It was to our benefit that we were working as a team at this stage; otherwise,
individual understandingswould not have been challenged so frequently.We
found that the NUD-ISTprogramis, as its manualclaims (Richards& Richards,
1990), more than a "code-and-retrieve"system:
The indexingdatabase can be of any level of complexity-fromthe flat lists of
codesnecessaryfor mostcode-and-retrieveprograms to highlyorganizedandcom-
indexesof categoriesandsub-categories.
plextree-structured Thisstructuring allows
indexingconceptsto be organizedandmanagedas theoreticalsystems,notjust as
labels.(p. 7)
Being forced to consider alternativeinterpretations,structures,and ways of

workingwas an educativeprocess in itself. We startedto discuss the notion that
the act of classificationpromotestheorybuilding, and being able to use NUD?IST
quickly and efficiently (to call up sets of phrases for comparison,combination,
and redistribution)assisted this process. It is importantto note that theory was
developing as sharedunderstandingswithin the group-not by the computer.As
Richardsand Richards(1990) note,
whattheuserdoeswiththeretrievals,usingsuchsoftware,is essentiallyoffline:all
the theory-building, the shapingof understandingandthe reshapingof the datain
accordance withthechangingunderstanding, is doneoutsidethecomputer.(p. 6)
Emerging Categories
At this stage, numbercodes were assigned to each of the text units and entered
into the computer.For the backgrounddata-for example, country of origin-
the whole of the survey was coded at the particularsubnode. This was done
because NUD?IST checks aspects of one category againstthose of anotherrather
than match informationwithin any node. For example, a researchermay later
wish to ask such questions as, "Whatdid males say about problem solving and
how does this comparewith what females said?"or "How did comments about
use of manipulativesvary between Australianand USA teachers?"Searches of
this naturerequirethe personalinformationto be storedin one node and the spe-
cific topics, such as "problemsolving" or "manipulatives,"to be in different
nodes (or subnodes)so that sortingand matchingprocesses can take place.
Initially, the major categories into which all data were placed were teacher
action (1) and pupil action (2). Figure 9.3 shows a set of example phrasesused
by a respondent.The first number2 given in the tripleto the rightof each phrase
relatesto the classificationpupil action. The middle numbersindicateminorcat-
egories from the next level of the classification tree: content and activities (12)
and interaction(13). The right-handnumber in the triple designates a further
JudithMousley, Peter Sullivan,and Andrew Waywood 137
subclassification-discussion between pupils (1) and investigations/problem-

solving (17). It would have been possible to create many levels of classification,
but we stoppedat the point at which phrasesaboutclassroomactionwere grouped
in relationto commonly used termsaboutteachingand learningstrategies.
Physicalinvolvement (2 12 3) 4
Cooperativelearning (2 12 3) 5
Discussion (2 13 1) 6
Problemsolving (2 12 17) 7
Risktaking (2 12 8) 8
Figure 9.3. Examples of coding a response.
The numbersin the right-handcolumn of Figure 9.3 are not classifications-

they are "addresses"for the original data. This program feature is important
because it enables the machine, on request,to undertakesearches for particular
strings (words, phrases,and so forth) or to reorganizedata while retaininglinks
with particularrespondentsor identifiedgroups of respondents.More important,
it enables researchersto skip quickly to the original paragraphs-the context of
the subject and his or her utterances-to check, for instance, that a phrase has
been categorizedwith consistency of meaning despite shuffling of the data dur-
ing the process of building and rebuildingcategories. If we aim to stay close to
the text of subjectsand to participatein its meaning, such facilities for recall and
review are vital.
In the process of refining categories, the collections placed into particular
nodes and subnodes were printed and examined to determine whether they
formed a coherentset. Any phrasesnot congruentwith the othersin the set were
moved. To determinenames for the categories that had been created,groups of
descriptors were printed and academics were asked to suggest appropriate
nomenclature.We found that most of the suggestionsmatchedthe names we had
alreadybeen using in discussing the nodes.
It was realizedafterwardthatit would have been useful to use a coding process
to enable later ready identificationof phrases that raised interestingpoints for
furtherdiscussion-such as two obviously conflicting beliefs or a particularly
thought-provokingcomment. However, such data were not lost, for one of the
benefits of using a computerfor qualitativedata analysis is that original words,
phrases, sentences, or whole paragraphs(and indeed whole chapters,papers, or
journals in other studies) can be called up later for furtherreview. Data remain,
relational and interactive,and can be traced back to their origins if necessary.
The retentionof originaldataalso means that analyses can be replicated,for pur-
poses of adding robustness,for checking the logic of the currentcategories, for
testing the consistency and coherence of the conclusions, or for examining the
impact of a differentteam of researchers.
Threefacilities of the programenabledthe keeping of an "audittrail"(Lincoln
& Guba, 1985, p. 319) throughoutthe analysisstage. The firstis the abilityto save
(as documents)and recall datafrom any stage of the analysisprocess.The second
is thatNUD*IST keeps a log of changesautomatically.A historyof what was done
thereforeforms the basis for keeping a record of, and retracingif desired, the
researchprocess.The thirdusefulfacilityis the capacityfor the researcherto notate
any entryfromwithinthe program.These facilitiesallow researchersto enterques-
tions arising, points of interest,links to publishedtheory (includingappropriate
quotations),commentsaboutthe researchprocess,and othernotes. They therefore
facilitate(a) the process of developing understandings,(b) the recordingof con-
ceptualdevelopmentsthathave takenplace, and (c) the writingup at a laterstage.
They also open up the researchprocessto examinationby others.
The Categories Arising
The printoutof respondentphrasescategorizedinto two differentsubnodesis

shown in Figure 9.4.
THINKING
(21218)
Aimsto develop understandingopportunityfor studentthinking
Lesson requiresthinkingratherthan repetitionand mimicking
Promotesthinking
Lotsof thinkingby students abouttask
Speculatingon how best to solve them and solving
Thoughtaboutthe problem
CHALLENGING
(212 9)
Challenging
Challengingtask withinreach
Challengingactivities
Challenge
Shouldhave abilityforextension
differences
Challengingbut caters for individual
Figure 9.4. Responses categorizedwithin two particularsubnodes.
These groups of phrases are ready to be inspected for coherence, sense, and
relevance.Phrasesthatdo not fit can be easily moved elsewhere. Fromsuch data,
we can infer that some teacher educators see student thinking and provoking
JudithMousley, Peter Sullivan,and AndrewWaywood 139
challenge as characteristicsof quality teaching, and so these constructsformed

partof the model that we developed.
Data gatheredunderthe headings used in the survey (teacheractions, content,
activities, and so forth) had now been organized,throughthe process of alloca-
tion to emerging categories as described above, into six quite-new categories:
building understanding,communication,engagement,problemsolving, task ori-
entation,and teacherconcern.
Efforts to place particularphrasesand the examinationof the content of each
node led to the realizationthat building understandingrelied largely on factors
and activities listed in other nodes, and it seemed that each of the other nodes
could be considered a vehicle for building mathematicalunderstanding.For
instance,the phraseslisted in such nodes as task orientationand communication,
and in such subnodes as use of materials, although significant in themselves,
seemed directedat teacher and studentactions that in turn would lead to build-
ing understanding.In other words, the main categories and the subcategories
arose from the process of grouping like comments together and establishing
coherence and sense from the groups. The placement of the building under-
standingnode at a higher level that the others was a decision of the researchers.
Figure 9.5 shows both the nodes and subnodes that illustratethe components
of quality teaching that emerged from the analysis and also a taxonomy or hier-
archy that we imposed on that data.
Building
Understanding
Materials
Prior
knowledge
Mathematical
thinking
Connections
Conceptual
development
Reflection
Sequence
Review
I I I
Organizing Nurturing Engaging Communicating Problem
forLearning Ability Active Pupil
topupil
discussion Solving
Clear purpose Levels Personal Sharing
strategies Investigation
Clear instructionNon-threatening
Enjoyment Cooperative Open-ended
Classorganization Rapport Realworld Recording Challenging
Questions Relationships Motivation Posing
Assessment Goalsetting Variety
Enthusiasm
Figure 9.5. Hierarchyof categories of quality mathematicsteaching arising from the data.
Other Possibilities for Analysis
Having the data formattedappropriatelyin the NUD-ISTprogramwould enable

the comparisonof commentsby experiencedteachers(e.g., 5+ years)with those of
less-experiencedteachersor with teachereducators.Similarly,words and phrases
used by teachereducatorsfrom differentcountries,or of differentgenderor from
differentage groups,could be compared.Althoughsuch lists were not called up, it
would only be a matterof a minute'sworkto requestandprintany lists of interest.
Furthercross-referencing,cross-indexing,and reindexingbecame possible as
the data were refined, and the potentialfor using NUDoIST for manipulatingdata
to answernew researchquestions became clearer.For instance, the results from
othersections of the questionnairecould be fitted with the new categoriesand the
reasons for inconsistencies explored with the synthesis of disparateinformation
serving as clues for more complex and richerreadings.Similarly,the additionof
new data-perhaps from repeatingthe study with other samples from the math-
ematics-educationcommunity-would allow interfaces and contrasts between
different sets of data to be examined. To carry out such comparisonsby hand,
and at several levels of the inquirytree, would be a huge and dauntingtask. It is
at this stage that NUD?IST makes such analyses practicable.
Although we were not interestedin using the qualitativedata for statistical
analysis at this stage, it is now in a form where this would not be difficult. For
instance, it is easy to examine the data to find out that the node building under-
standingwas outstandingin termsof the numberof times respondentsreferredto
its features (112 times out of the 125 responses), so this proportioncould be
comparedwith responsesto such featuresin the structuredsection of the survey.
An analysis such as this would have the potentialto be informativebut would be
in dangerof removingfrom the reportingprocess much of what is useful for our
specific purposes.
COMPUTERANALYSIS AND QUALITATIVEDATA
It is important to note that although reduction to general categories has

involved some loss of depth of meaning as well as a coalescence of discrete
ideas, the conclusions remainopen to verification.All the richness of the origi-
nal data is retainedand can be recalled at any stage of the project-or even in
furtherprocesses of inquiry. The original and sorted data can all be used and
reportedmeaningfully.Handlingdata in this way is one solution to the common
qualitative data-analysis problem that Sowden and Keeves (1988) identified
when they claimed that
whileanincreasingamountof empiricalresearchusingqualitative datais beingcar-
riedout, ... it is beingreportedin sucha waythattheconclusionscannotbe verified
.... Thisgivesriseto the anomaloussituationthatwhilein researchthe evidenceis
richanddetailed,theveryrichnessanddetailof thedatacollectedpreventpresenta-
tionin a coherentformthatwouldleadto acceptanceof the findingsas a contribu-
tionto scholarlyinquiry.(p. 525)
In comparison with manual indexing and matrixing systems, it is clear that

using a programlike NUDoISTenables the efficient handling of large quantities
of textual materials and offers several advantages for educational inquiry.
Readersof researchreportsgenerally rely on accurateclassification of raw data
by authorsand usually have no way of checking the interpretationsmade or the
appropriatenessof orientingconstructs.However, when data are stored as elec-
tronicdocumentsandreadilyaccessible, looking anew at materialgatheredor the
replication of the analysis processes by other researchersdoes not require the
repetitionof a lengthy processes. Whereas the reductionof data to generalities
frequently closes off options for cross-indexing and reindexing, or at least
requiresthe use of a tedious index-cardsystem, NUD*ISTenables-and in fact
encourages-such processes. Similarly, numericaldata (such as frequenciesfor
sets and subsetsof data)can easily be reworkedin the light of new researchques-
tions or developing understandings.In addition to their use for indexing, pro-
grams such as NUDoIST become tools for the building of theory by emphasizing
the emerging natureof understandingswhile maintaininglinks between original
utterancesand their discursive contexts.
Notwithstandingthe contributionthat computer-basedtools can make to qual-
itative analyses, there are limitations.Using a computerfor categorizingqualita-
tive data does not overcome some of the problems inherent in the research
design. It does assist with the focusing of data,enablingit to be reducedto a state
where it can be analyzed in terms of quantity,discussed underbroad headings,
or comparedwith other data. Data can then be displayed in a compressed,orga-
nized form thatis accessible to teachers,teachereducators,and otherresearchers.
Conclusionscan be drawn.As theory aboutmathematicspedagogy is developed
further, irregularities,patterns, and plausible explanations together with the
notions of robustness, sturdiness, and validity can be explored (Miles and
Huberman,1990, p. 349).
Althoughthe originaldataare still accessible, the act of classificationinvolves
a reorientation,rebuilding, and modification of the original data as well as
removal of individual words and phrases both from their original contextual
paragraphsand from their speakers.Whereasthe aim of the researchermight be
to keep as close as possible to the intentionof the contributorof a given phrase,
two factors impinge on this process. First, the act of categorizationis by its very
naturesubjective. Selection of orientingcategories depends on researchers'per-
sonal constructsof the task at hand.These, in turn,are shapedby perceivedpos-
sibilities within the fields of researchand educationaltheory and then by devel-
oping notions of the overall "findings." Second, phrases and terms used to
describe the teaching and learning of mathematicsare themselves attemptsto
capturenebulous qualities and diverse ways of behaving. These are used in var-
ious ways according to participants'prior experiences and understandingsas
well as their perceptions of audience. To place phrases within categories is to
assume a meaning and a set of associative propertiesand perhapsto "bend"the
data to fit with an emerging structure.As with most analysis of most qualitative
data,factors (such as inflection and body language)that would have assisted the
accuracyof placementor summationare lost in the transcriptionprocess.
It can be noted that these problemsare relatedto the subjectivenatureof deci-
sion making and the interpretationof humanbeliefs and actions and are not in
any way specific to the use of NUD*IST. As Giddens (1984) notes, knowledge is
framedas individualsview the real world in terms of their personalunderstand-
ings, and these interpretationscan be made only in the light of their current
understandingsof the theories, ideas, and concepts. Furtherinterpretationtakes
place as the ideas are published and take new form in the praxis of everyday
social use. Although "moments"of decision making are not so apparentin
empirical research, they are still present, and attemptsto control such factors
bring their own set of limitationsto researchprojects.
It mustbe recognizedthatusing a computerdoes not change the fact thatinter-
pretive research is based on the gathering of qualitative data and that such
researchaims to understandratherthanto explain. The intentionis to captureand
interpretratherthanto generalizeor predict.The complexities and uniquenessof
an event (such as the beliefs of professionalsat a particulartime) are recognized
as objects worthstudyingand using. These objects can bring aboutchange in the
stance of individualresearchersbut should not be consideredthe replicablephe-
nomena of empiricalresearch.
ISSUES FOR CONSIDERATION
Using a computerfor handlingqualitativedatahas certainlyfacilitatedourexplo-

rationof social actionsandunderstandings andassistedin the constructionof tenta-
tive understandingsaboutthese. One groupof questionsraisedin our discussions
aboutthe use of tools suchas NUD*IST relatedto how we, as educationalresearchers,
makejudgmentsregardingwhatinformationto seek, how to seek it, how to analyze
it, and how to use it. We discussedhow the resolutionof these questionsis influ-
enced by, and also limitedby, our own socioculturalhistoryas well as by our indi-
vidualprofessionalhistories,personalunderstandings, andcurrentneeds.
Anothergroupof issues raisedin our discussionswas aboutthe sharingof find-
ings and what is consideredby a researcherto be worthsharing.Whatis withheld
from the reader?How do we shareour findings with particularaudiences,and is
this done selectively? How accessible are the ideas presentedto the consumers
and producersof educationaltheoryin use, thatis, teachersin classrooms?
There was also discussion about how we raise the issues of the problematic
natureof judgements that have been made duringthe researchprocess. At each
stage of the process, decisions were being made abouthow to handlethe data in
its currentform, but in most researchreports,the natureof this decision making
is not presentedas problematic.The collaborativenatureof the coding, and the
verification process using other researchers,provided checks and balances and
was an integral part of the research method. We also sought to address such
issues explicitly in reports.
JudithMousley, Peter Sullivan,and Andrew Waywood 143
The experience of thinkingaboutthe researchprocesses undertakenhas sensi-

tized us to the workings of interpretationwithin educationalresearch. Starting
points and further developments, analyses and theory, conceiving and recon-
ceiving, part and whole, what-is-seen and what-is-to-be-seen-the interplays
among these are ongoing throughoutthe research.All reflect a process of gain-
ing understanding.The researcheris involved not only in processes of represen-
tation and the creationof boundariesbut also in processes of expandinghorizons
of understanding.However, readersof publishedarticlesrarelysee them as artis-
tic creations-pictures of the educational researchers'experiences, intentions,
and growing conceptions.And the act of readinga researchreportin many ways
parallels the creative acts of the researchersthemselves in that the theoretical
stances of the readersinform their interpretationsof, as well as their abilities to
use, the work at hand.
The data-analysisprocess using NUDoIST taughtus as much aboutour own (as
well as each others') concepts of quality lessons and the act of writing a ques-
tionnaireitem as it did aboutthe beliefs our colleagues hold aboutquality math-
ematics teaching. We also learneda lot about the way differentphrasesare used
in describingteaching and were remindedcontinuallyof the vagueness of most
of the terms used. It is now hoped that we can use our collected data to engage
people in furtherdiscussion about the appropriatenessof the categories formed
and about the possibilities of developing a more common language for describ-
ing teaching. It might be noted that the developmentof a common language is
one of the characteristicsof a discipline of study, and one that educational
researchgenerally could well pursuemore actively.
144
Chapter 10
Using Research as a Stimulus for

Learning
Beatriz S. D'Ambrosio
Teacherresearchprovides a tool for developing, encouraging,and sustaining

teachers' reflective practice. The process of carrying out such research leads
teachersto analyze their classrooms, their practice, and their students' learning
to depths that are difficult to reach with other types of professional activities.
Furthermore,the results of such researchprovide the largermathematics-educa-
tion community with greater insight into aspects of successful teaching.
Qualitativemethods furnishessential tools for teachersto use in this endeavor,
as they study the reality of their classrooms and their students'learning.
In this chapter,I describe a preserviceand an in-service experiencein which
teacherresearchwas used to engageteachersin exploringhow studentslearnmath-
ematics. First, I discuss the importanceof teacherresearchin bridgingthe gap
between researchand practice.Then I examine the parallelsbetween the act of
learningand the act of researchingand the use of practitionerresearchin develop-
ing K-12 mathematicsinstructionthatis moreconstructivistin nature.The chapter
focus on qualitativeresearchmethodsis then presentedthroughan example of a
teacher-research experiencedevelopedfor, and implementedwith, preservicesec-
ondarymathematicsteachers,followed by the descriptionof an in-serviceexperi-
ence centeredon the use of teacherresearchas the primaryteachingmethodology.
ISSUES IN TEACHERRESEARCH
The questionof how to bridge the gap between researchand practiceis a con-
cern of the internationalacademic researchcommunity in mathematicseduca-
tion. Several participantsat the Eighth InternationalCongress on Mathematics
Education,held in Seville, Spain, in July 1996 addressedthis issue in a working
group that explored the theme "connectionsbetween research and practice in
mathematicseducation."Among the many dimensions that exacerbatethe gap
are the lack of communicationbetween academic researchersand practitioners,
the silenced voices of practitionersin much academic research,the demeaning
approachto practitionerknowledge in much academicliterature,and the view of
practitionersas users of knowledge about teaching and learning ratherthan as
active participantsin the generationof thatknowledge.
Beatriz S. D'Ambrosio 145
It has been useful to me to invite teachersto engage in researchactivity and to

learn from this experience what constitutes some of the issues that result in the
lack of communicationbetween researchersand practitioners.Two issues, in my
mind, need resolutionif we hope to involve teachersin the dialogue aboutteach-
ing and learning:the acceptanceof teacherresearchas scholarshipand the accep-
tance of teacherresearchas partof the process of generatingthe knowledge base
about teaching and learning.
In recent encounterswith colleagues, I have been asked why I insist on using
the phraseteacher research in my work with teachers.My colleagues have sug-
gested that I insteaduse the phrasepractical wisdom to describe what we might
learn from teachers'researchactivity. I am very concernedthat using the phrase
practical wisdom instead of the word research underminesthe legitimacy and
authorityof the knowledge base of teachers-a knowledge base that is ground-
ed in the articulationof theoreticalframeworks,personalhistories, and practice.
Teachers'researchon teachingand learningprovidesan insider's perspectiveon
the complexities of the lives of teachersand studentsas they interactto construct
understandingand meaning aboutthe world aroundthem.
The acceptance of teacherresearchas scholarshiprequiresthat the academic
communitylook to thatbody of work as a new genre of researchwith its own cri-
teriaof rigor,uniquemethodology, form, and style. It also requiresa reconsider-
ation of the purposes of research.Teacherresearchersdo not look to academic
research for the accumulated"knowledge base" about teaching and learning.
Instead, academic researchserves as a source for them of intriguinginterpreta-
tions, conflicting information,multiple conceptual frameworks,confirming or
discrepantevidence from other settings, and new questions and problems. As
teachers problematizetheir teaching and seek to understandtheir classrooms,
their students,and their practices, they build theoreticalframeworks,raise new
questions, and contribute to the knowledge base on teaching, learning, and
research.As a community,we face the challenge of including teachers' ways of
knowing and understandingin our dialogue about teaching, learning, and
research.
Also, to more fully embracethe researchof practitioners,our traditionalviews
of theory must be redefined.This new genre of research,still in its early stages
of development, requires an understandingof theory as a combination of per-
spectives, that is, as a set of interrelatedconceptual frameworksgrounded in
practice(Cochran-Smith& Lytle, 1993).
LEARNINGTHROUGHTEACHERRESEARCH
Currentreform initiatives in the teaching of mathematicshave placed new
demands on teachers. In particular,the curricularperspectives adopted by the
NationalCouncilof Teachersof Mathematicsrequireteachersto takeon roles in the
classroomthat differ from those that have typicallyengaged teachers.These new
roles include guiding and encouragingchildren's investigationsof mathematics,
146 Using Researchas a StimulusforLearning
basing instructionon children'spriorknowledge,tying mathematicsto the world

of applications,and being a learnerand researcherthemselves. For preservice
teachersand practicingteachers,these new expectationsoften differ greatlyfrom
theirbeliefs aboutthe activitiesof a teacher.
An effective source of change of teachers' conceptions of teaching and learn-
ing is that of the teacheras researcher(e.g., Darling-Hammond,1996; Hubbard
& Power, 1994; Meyer, 1995; Nias & Groundwater-Smith,1988; Short& Burke,
1996; Wells, 1994). This approachis believed to develop an investigativedispo-
sition in teachers,helping them to become reflective, questioning,and flexible.
The area of language arts has the longest history of teacher as researcher
(Goswami & Stillman, 1987; Phinney & Ketterling, 1997; Sega, 1997; Short,
Harste, & Burke, 1996). The work done by teachers in this field, as they study
their students'learningof readingand writing,reveals how practitionerresearch
can serve as an agent of change. The impact on teachers of conducting class-
room-basedresearchincludes the following: (a) teachers come to better under-
stand their classrooms, (b) they are in a betterposition to make decisions about
theirclassroompracticeand the classroomenvironment,(c) they take ownership
of change initiatives, (d) they come to betterunderstandtheir studentsby listen-
ing to them in ways they have never listened before, (e) they model and foster an
inquisitivedispositionin theirclassroomsby redefininglearningas research,and
(f) they engage in classroom-basedresearch that renews their enthusiasm for
learning.Finally, classroom-basedresearchwill bring the voice of the teacherto
the researchcommunity and the voice of the researcherto the classroom, thus
bridgingthe everlastinggap between researchand practice(Altrichter,Posch, &
Somekh, 1993; Boero, Dapueto,& Parenti,1996; Bullock, 1987; Cochran-Smith
& Lytle, 1993; Crawford& Adler, 1996; Goswami & Stillman, 1987).
The parallelsbetween the act of learningand the act of researchingareremark-
able when learning is understoodas a constructiveprocess as is advocated in
many currentreforminitiatives.In fact, one might ventureto claim that learning
is a researchprocess. Teachers' pursuitof research studies in their classrooms
provides opportunitiesfor teachersto learnmore about studentunderstandingof
mathematics, mathematics itself, and themselves as teachers. These insights
enable teachersto tackle the new roles envisioned by reforminitiatives.
Creatingan environmentthat encourages studentsto constructmathematical
ideas requiresteachersto understandwhat studentsbring to the learningexperi-
ence. Effective classroom activities capitalize on students'priorunderstandings
of mathematics,which allows studentsto constructnew meaningsto solve prob-
lems. Teachersnot only must assess studentunderstandingof mathematicalideas
priorto planningexperiences,but they must also devise means of assessing stu-
dent growth in knowledge duringthe activities.
Teachers are encouragedto assess much more than students' knowledge of
mathematics. The Curriculum and Evaluation Standards for School Mathematics
(NationalCouncil of Teachersof Mathematics,1989) recommendsthat teachers
assess students'willingnessto pursuean investigation,perseverance,flexibility of
use of heuristics,decision-makingprocesses,and otherprocess skills thatenhance

theirparticipationin the communityof learnersof mathematics-in this case, the
mathematicsclassroom.Creatinga researchdispositionin teachersis essential if
they are to successfully assess the many dimensionsof studentthinking.
To successfully plan investigative activities for students,teachersmust under-
standmathematicsas an inquiry-baseddiscipline (Borasi, 1992; Siegel & Borasi,
1994). This requiresteachersto actively investigatemathematicsthemselves.For
many teachers,this is a totally new perceptionof the natureof mathematicsthat
requires them to reconstructtheir understandingof the act of learning mathe-
matics. Many teachersperceive mathematicalresearchas an activity restrictedto
researchmathematicians.It is crucial,if changes in the natureof classroomactiv-
ities are to occur, that teachersbelieve thatthey too can investigatemathematics
and pursue researchin that field of inquiry.The disposition towardmathemati-
cal inquiryfostered in their studentsmust be one that they themselves practice.
Professional-development experiencescan be used to createan investigativedis-
positionin teachers.Researchactivitiesin mathematicscan be used thatare similar
to the researchactivitiesthatteacherswould engage in with their studentsas they
participatetogetherin the communityof mathematicslearners.Activitiesdesigned
to encourageteachersto reflecton theirpracticewould stimulateteachersto recon-
sidertheirperceptionof what constitutessuccessfulteaching.In addition,teachers
shouldbe encouragedto systematicallyplan innovationsand analyzethe successes
and difficultiesencounteredduringthe implementationof these innovations.
In conclusion, classroom-basedresearch, carried out by teachers, provokes
teachers to analyze their classrooms, their practice, and their students' learning
to depthsthat are difficult to reach with othertypes of professional-development
activities. Furthermore,such researchprovidesthe largermathematics-education
communitywith greaterinsight on aspects of successful teaching.In the follow-
ing pages, I describe two programsthat have been designed to engage teachers
and futureteachersin researchactivities. A goal of these programswas to engage
participantsin reflective practice.I give examples of the ways in which partici-
pantswere able to reflect on their students'learning,on mathematicsitself, or on
their teaching practicethroughtheir researchstudies.
A PRESERVICEEXPERIENCE
A teacher-educationseminarheld at the Catholic University of S,,o Paulo in
Brazil is an example of how a teacher-researchexperiencewas designed for pre-
service secondarymathematicsteachers.The goal of the seminarwas to initiate
the participantsin researchexperiences that would shape their beliefs about the
teaching and learning of mathematics and about the nature of mathematics.
Together, the seminar participants (future teachers, university faculty, and
guests) constituteda researchteam. The chargeto the researchteam was to iden-
tify a researchquestion,propose a study, collect data,analyze the data,and write
up the findings in the form of a report(Researchreport,no date).
148 Using Researchas a Stimulusfor Learning
The group decided to work on studyingchildren'sunderstandingof fractions.

They proposedto design a diagnostictest, apply the test to a large numberof stu-
dents, and analyze the data. Unlike most researchseminars,the group was pur-
posely not taughtany researchmethodology.The philosophyadoptedby the fac-
ulty was thatknowledge is constructedby the learner;thus the approachtakento
learningabout researchshould be constructivistas well. Hence, studentswould
experience research activities and construct their own understandingof what
constitutesresearchon the basis of reflections on the experiencesthemselves.
This researchgroup was not "hindered"by criteriaand standardsof what the
mathematics education research community would consider "good" research.
Instead,the drivingforce for theirresearchstudy was the pressingquestionsthat
had emerged duringtheir field experiencesin preparingto become teachers.
The researchgroupmet twice a week until data collection began. During data
collection, meetings occurred once a week. During data analysis, meetings
occurredtwice a week. The role of the faculty membersduringthe meetings was
to serve as participantobservers. They took field notes of the discussions and
asked questions that focused the groups' attentionon the researchactivity. For
example, as items were producedfor the diagnostictests, the faculty would often
ask the group for a rationalefor including specific test items. Also, the students
often turnedto the faculty to identify readingsthat were relatedto the questions
they were asking.
The group's initial research design reflected their beliefs about research-
beliefs gained from much of the readingthey had done on educationalresearch
in general and on mathematicseducationresearchin particular.They were con-
vinced that they would not be able to say much about children's understanding
of fractions unless they tested a very large numberof students. Thus, the first
step takenby the groupwas to test 182 studentsbetween fifth and eighth grades.
This was followed by an analysis of the test results. This analysis was, in a first
instance,focused on the frequencyof correctandincorrectanswers.The research
team was amazed at the results and distressed about how little they could say
aboutthe results.Only vague commentscould be made, such as "88%of the chil-
drensolved question 15 incorrectly"(Researchreport,p. 6). As the researchteam
was pressed by the faculty to say more about the findings, they began to look
beyond the numericalresults.The researchteam began to raise many conjectures
as indicatedby the following comments:"It seems that the 7th and 8th graders
were so worried about using the correct rules for operatingwith fractions that
they were unable to finish the test.... We could not understandhow 8th graders
had to find a common denominatorin orderto solve 1/2+ 1/4+1/4. It seems they
couldn't transfertheirinformallearningof part/wholeto the formalizationof the
ideas" (Researchreport,p. 6).
Many more questionswere raisedfrom this first step than were answered.The
students refined the study by restructuringthe diagnostic instrument. They
revised the instrumentso that it no longer focused on both proceduraland con-
ceptual knowledge of fractions,but instead focused on items that might reveal
students' conceptual understandingof fractions. The second test was adminis-

tered exclusively to 10- to 12-year-oldchildren and was given to only 76 chil-
dren.As the studentssearchedfor meaning and for explanationsof how children
were making sense of fractions,tests were analyzed much more qualitatively.
Althoughthis secondtest was a lot morerevealingthanthe first,it was clearthat
anotherform of data collection was needed before the questionsraised could be
adequatelyaddressed. It is worthmentioningthattherewas evidence of growthin
these preservice teachers' thinking about children's understandingof fractions
even as they developed the second instrument.Every question had a rationale
behindit, and many children'sanswerswere anticipated.For every potentialchil-
dren's answer,the studentshad a conjecturefor a possible explanation.
It was during the analysis of the results of the second test that the students
became aware that the interview process might be a more effective form of data
collection for the types of questions they wished to address. Discussions of the
various readingson researchabout the learningof fractionsand the ineffective-
ness of the initial datain giving insight into children'sunderstandingof fractions
pointed the group to the use of interviews.
Reflections on the Preservice Experience

It is not the purpose of this chapterto reporton the findings about children's
learningof fractions.As stated earlier,my goal is to reporton the study of pre-
service teachers' growth in understandingof the researchprocess and how an
investigative disposition becomes an importantpartof the teaching process.
The decision to use interviews with the children was an importantbreak-
throughin the students' thinking about research. It was also interestingto see
how creative they were in thinking of ways to elicit children's thinking. The
decisions they made about the interview process were the following: (a) they
would interview some children individually and others in pairs, with the hope
that the childrenin the pairs would communicatewith each other and that their
thinkingwould be verbalized;(b) they would have two membersof the research
team conductingan interview (one interviewing,the othertakingnotes); (c) they
would audiotapethe interviews; and (d) they would pilot the interviews before
finalizing the interview structureto improve the questions posed and to learn
how to conduct interviews.
Decisions were also made about how to pick the children to be interviewed.
Childrenwere classified into three groups on the basis of the results of the sec-
ond test: childrenwho got everythingcorrect,those who made consistent errors,
and those whose errors were inconsistent. The pairs were chosen in various
ways: (a) one child who made no errorswas paired with one who made errors,
(b) one child who made consistent errorswas pairedwith one who made incon-
sistent errors,and (c) one child who made errorswas paired with a child who
made a very differenttype of errors.At this point it was clear that this research
groupwas strugglingwith a very importantquestionthat faces researchersusing
interview procedures.The most effective means of conductinginterviews with

children is not clearly established in the research community. This research
group was exploring an open-ended question about pairing students for inter-
viewing purposesin an effortto elicit as much informationas possible abouttheir
understandingof fractions.
Some of the indicators of growth in reflective thinking among the students
were the following: (a) the research group's ability to anticipate children's
responsesto the questionsposed, (b) the rationalegiven for each questionposed,
(c) the emerging need for the use of differentresearchmethodsto furtherinves-
tigate a question, (d) the emerging new questions arising with the analysis of
each new data set, (e) the emerging feelings of anxiety about teaching as they
analyzedthe consequences of traditionallearningexperiences, and (f) a realiza-
tion by the studentsof how childrenconstructknowledge and make sense of their
experiences. There was also evidence that teaching experience alone would not
generate the level of reflection necessary to analyze the data about children's
work with fractionsthatwas gatheredin this project.A groupof practicingteach-
ers was asked to use in their classes the second diagnostic test producedby the
researchgroup and to analyze the results.The teachershad very little ownership
in the diagnostic test, they had little sense of why the questions were composed
as they were, and they themselves had not raised the researchquestions driving
the analysis of the test results. This resultedin a much less sophisticatedanaly-
sis of the test results as evidenced by the teachers' informalpaperspreparedfor
an in-service course they were attendingat the time.
The following comments from three student members of the researchgroup
may help the readersense how reflective the studentswere towardthe end of the
group researchproject,and especially how open they were to continuouslearn-
ing and self-assessment.
I feel thatI am still not secureaboutwhatto do, butI realizethatI havechanged
tremendously in whatI believeaboutteachingandaboutevaluation....I feel anxious
everytimeI correcta test,andI knowI am expectedto quantifythe resultsrather
thandescribethe students'understanding. (Maria,interview,March1989)
I amnowconcernedwiththedirectionstakenin education,withmy students....We
havetakenthefirststep,to detectandworryabouttheproblemsin learning(evenif
we arestillunableto resolvethemfully).... Thisis my expectation, to continuemy
questforchangesin my practice.(Ana,interview,March1989)
... reflectaboutstudents'perspectives.Thisis importantin all of pedagogicalprac-
tice, in theobjectives,lessonplans,contentanalysis,andmethodology... theclass-
roomhasbecomeonebig laboratory, whichin factmotivatesone'swork.(Roberto,
interview,March1989)
AN IN-SERVICEEXPERIENCE
The setting for the in-service experience describednext was a graduateclass

called Teachersas Researchers,held at the Universityof Delaware.The class was
BeatrizS. D'Ambrosio 151
a requirementfor the master'sdegree in educationand usually had an enrollment

of about 30 students.The intentof this class was to engage practicingteachersin
classroom-basedresearchactivities.Most of the studentsin the class were practic-
ing teachers,and theirresearchstudieswere carriedout in theirown classrooms.
The main goal of the class was to encourage teachers to identify a research
questionrelatedto theirclassroompracticeand to conducta study addressingthe
question they had identified to gain insight into their teaching practiceand how
it relates to their beliefs about teaching and learning.Questions, rangingfrom a
focus on a single child to a focus on school policies andprocedures,variedgreat-
ly among teachers.
Built into the process was a supportnetworkfor the teachersas they pursued
their studies. The class met formally on a weekly basis for 3 hours. Part of the
class meeting consisted of teachers working in small groups to share their
progress,their strugglesand difficulties, and their insights. Duringthis time, the
instructorcirculatedthroughoutthe class and noted the similarities and differ-
ences in the issues addressedby the various groups. These issues became the
focus of the large-groupdiscussion duringthe final hour of the class meeting.
Inquirybegan as the teacherswere asked to keep personaljournalsfrom which
a researchquestion would emerge. The journals were describedto the teachers
as a personalrecordof theirday. Teacherswere asked to write in thejournalreg-
ularly, daily if feasible, and reflect on the accumulatedjournal entries weekly.
Their goal throughoutthe first month of the experiencewas to use the journalas
a tool to help them reflect on their practice. They were to find something that
intriguedthem about their practice or about their students or, more generally,
abouttheir lives as teachers.These journals took on many differentroles for the
differentteachers.Whereassome kept a journalthat resembleda log or diary of
classroom activities, others used the journal to raise issues about teaching and
learningand to reflect on the issues. The importanceof the teacher'sjournal in
identifying a researchquestion is reflected in Strieb's (1993) comment:
I'd foundmy idealmodeof datacollection.In theintroduction
of thepublishedver-
sionof myjournal,I wrote,"themoreI wrote,themoreI observedin my classroom,
andthemoreI wantedto write."As I re-readmyjournalI got moreideasforteach-
ing. (p. 122)
Initially the readings used for the course included primarilywork done by
teachers (see examples in Bissex & Bullock, 1987; Goswami & Stillman, 1987;
Rowland, 1984). Many teachersentered the course with great skepticism about
researchand the researchprocess. Even those who viewed researchas an activi-
ty limited to members of the university community were interested in, and
intriguedby, the pieces writtenby teachersthemselves. Accordingto Kincheloe
(1991), "Traditionalresearchershad weeded out the self, denied their intuitions
and inner voices, in the process producingrestrictedand object-like interpreta-
tions of socio-educationalevents. Using the traditionaldefinitions, these object-
like interpretationswere certainand scientific"(p. 30). Whereasuniversity-based
authorstypically strive for "objectivity,"purposefullystrippingtheir writing of

any emotion or passion, practicingteacherswrite with all the emotion and pas-
sion that characterizetheir lives. The experiences of teachersin this class were
clearly resonating with the experiences described by the writings of teacher
researchers,that is, classroom teachers who had systematically studied a ques-
tion of concern to them in their professionalpractice.
As questions began to emerge for each of the teachers, I noticed certainpat-
ternsin teachers'choices of researchquestions.Severalteacherschose to look at
how to bettermanagetheirclassrooms.Theirfocus was on discipline and behav-
ior. Theirrationalewas typically thatthe class as a whole would benefit if it were
more effectively managed.Otherschose to study a studentor a small group of
studentsand their learningdifficulties. All of these latterstudies clearly focused
on children.
Several teachersused the researchliteratureand the findings in that literature
as the springboardfor their inquiries.One teachercame to questionthe assump-
tions within the misconceptions in literature in science education. Another
teacherused her learningsaboutproblemsolving to analyze what implicationsit
had for her and her teaching. Several teachers studiedemergentliteracy in their
students.One teacherpiloted the use of a curricularprogram,MathematicsTheir
Way(Barrata-Lorton,1976), in her first-gradeclassroom.
Two teachersfocused on policy issues that affected classroom practices.One
teacherrealized from reflecting on her journal entries how often she was inter-
ruptedthroughouther teachingand chose to study how interruptionsaffected her
teachingand her students'learning.Her systematicinquiryprovidedevidence to
arguethatschool policies concerningclassroominterruptionsfrom the office and
administrationbe revised. Yet anotherteacher set out to prove how resistant
teacherswould be towardincludingresearchas a partof theirdaily activities. She
was convincedthat this was an impossibletask to include in theirdaily lives and
set out to convince the class what a crazy idea this was. Interestingly,in spite of
all her "evidence,"she could not argue the case strongly, and in fact, her col-
leagues' experiencesseemed to disproveher conjecturesratherthansupportthem.
To enable teachersto develop a study that could be managedwithin a semes-
ter, my role in the initial phase of the study was to help them focus their ques-
tions. Similarto the process in the typical masters'study or doctoralstudy,much
effort was put into clearly defining the researchquestion and focusing on that
question. Once each question was defined, my task became one of advising the
teacheron the researchmethodologythatwas most reasonablefor addressingthe
questionand most feasible to fit as unobtrusivelyas possible within the teacher's
classroom practice.
The methodologyused by the teacherresearchersvariedaccordingto the types
of questions they addressedand included primarilyqualitativesources of data.
Throughoutthe experience, many teachers were reluctantto accept qualitative
data sources as legitimate researchtools. However, as they addressedthe feasi-
bility of using quantitativeexperimentaldesigns in the classroom and the nature
Beatriz S. D 'Ambrosio 153
of the types of questions they were asking, it soon became clear that the most
informativesources of data would be mostly qualitative.The methodology used
turnedout to be quite dynamic, and new sources of data and forms of data col-
lection were tried throughoutdifferentstages of many of the studies.
The analysis of the datawas done throughoutthe study. Every week the teach-
ers presentedtheir findings to their small working group. The group discussion
often led to insights and suggestions. My role at this stage was to point out some
of the research literaturethat could offer insight into what the teachers were
observing and finding in their own studies. I realized that the process of analyz-
ing and interpretingthe data providedintrinsicmotivationfor readingthe exist-
ing, related researchliterature,a level of motivation that I had little success in
generatingwith other forms of in-service experiences.
Susan's Inquiryand Her Growthas a Teacher

The following example is used to illustratethe reflectivenessof one of the par-
ticipatingteachers,Susan,as she conductedher study.This teacherused herjour-
nal entries and her reflections about children's success in problem solving as a
springboardfor her thinkingabouther practice.
In her early journal entries, Susan begins her reflections by describing her
teaching style and philosophy of educationas expressed below:
My styleandphilosophyof teachinghasbeenverymuchcenteredaroundthemas-
teryof thebasicsandteachingthe methodsthatallowone to arriveat the answers.
My instructionhasoftenstartedwithmanipulatives andhands-onpracticeto define
how a basicoperation,suchas multiplication or division,worked.However,from
thereI wouldrevertto the text andpracticeas manyproblemsas possiblewiththe
students,developinga step-by-stepplanto arriveat a correctanswer.I havefound
thatI amveryeffectiveatteachingthisway,andI haveseenstudentsmasterthefour
basicoperationsthroughthismethodof teaching.
Susan was a teacherin her second year of teaching. The teacheras researcher
class was the very first class in her master's program.She had felt rewardedin
her teaching experience with fourth graders by getting children to master the
basic facts in mathematics.It was the journalentriesrelatedto this researchpro-
ject that led Susan to realize that the students were not performingas well on
problem-solvingtasks as they were on the masteryof the basic facts.
The startof thisresearchbeganwithmy ownrealizationthat,in therealworld,my
studentswouldrarelycomeacrossproblemssuchas 27 x 15.Theywouldrarelyfind
key wordsto hintat a solution.Realmathproblemscomein the formof wordsand
numberswithouta signtellingyou whatto do. Realmathproblemsrequirecritical
thinkingandplanningbeforea solutioncanbe reached.
In analyzingher thinkingaboutproblemsolving and her approachto teaching
problem solving, she began to notice some patternsof behavior.
I didnotdwellon theproblemsformorethanone or two classperiods.I wantedto
getthemcompleted,andI didnotconcentrate on the"howdo youknowwhatto do"
154 Using Researchas a StimulusforLearning
aspectof theproblem....I foundmyselfencouraging

strategiestaughtin theearlier
grades,suchas locatingkey words,thatwerenow ineffectivebecausethe mathe-
maticaloperations
weremoreadvanced.
The focus of her study turned to the children's strategies for solving word
problems. She interviewedchildren, she observed their work in small groups,
and she reflected on her actions and their consequences for the students' learn-
ing of problemsolving.
I was horrifiedwhenI realizedthe patternin someof theseproblem-solving
meth-
ods andthewaysin whichI hadbeenhelpingstudentsto solvetheseproblems....I
knewas I beganto see thesemethodsappliedthatI hadindeedsaidin oneof theear-
liersubtractionlessonsthatto spendmoneyyoumustsubtractsincemoneyis being
takenaway.... I alsoknewthatwhenwe startedtheyearwithadditionandsubtrac-
tionI haddisplayeda chartof keywordstypicallyfoundin wordproblemsrequiring
studentsto addandsubtract.I foundthatsuchan ordered,nonanalytical methodof
teachingwouldcometo hauntme onlyseveralmonthslater.
In the midst of her study, I noticed a shift from Susan's thinkingabout what
she was doing in her classes to concentratingon what she could learn from the
children. The focus of her inquiry changed dramaticallytoward the children's
thinkingstrategies.Fromobservingand interviewingthe successful studentsand
comparing their thinking strategies with those of the less successful students,
Susan triedto extrapolatethe thinkingstrategiesthat she could encourageall stu-
dents to use.
The moresuccessfulproblemsolverswerethe studentswho couldanalyzethe size
of the solution,attributemeaningto the words,andgeneratea diagramor picture
relatingto theproblem.
Susan's efforts to understandstudents'thinkingaboutproblemsolving helped
shape her own views about problem solving. Although she had clearly studied
what problem solving was about in her preservice teacher-educationprogram,
what it really meant to teach childrento solve problems had not been incorpo-
ratedinto her practice.It is likely that Susan had not internalizedthe meaningof
a problem-solvingapproachto teaching mathematicsfrom her teacher-prepara-
tion experiences. During the research experience, Susan found and reread the
articles about problem solving that she had alreadybeen exposed to duringher
preservice training.At this point, she was able to draw conclusions about the
implicationsof thatliteraturefor her practicemuch more effectively thanshe had
been able to do prior to this experience. She was able to tie her findings to the
literatureon students'problem-solvingstrategiesand to plan for changes in her
teaching of problemsolving.
Since uncoveringthe differenttypesof problem-solving strategies(botheffective
andineffective),I havetriedto changemy teachingmethodsas well as my attitude
aboutteachingproblemsolving.
Stimulatedby this experience, Susan came to question her values for mathe-
matics instruction.She had been pleased with her results, yet the reflections on
her practice and the group interactionswith her peers led her to question her
assumptions and beliefs about teaching problem solving. The questioning and
raising problems from within through self-analysis took on a very different
dimension thanit might have if an outsiderwere tryingto point to "faults"in her
teaching. This is but one of the many examples whereby the teachers who
engaged in teacher research found themselves questioning their practice and
wonderingand planningwhat they might do differently.
CONCLUSION
This chapter was intended to point to some of the dimensions of teacher
researchthat can serve to foster a dispositiontowardinquiryin teaching.In both
the preservice and in-service experiences described here, teacher researchwas
used as a tool for developing, encouraging,and sustaining teachers' reflective
practice. In these experiences, qualitativemethods were the main tool used by
teachersto studythe realityof theirclassroomsand the learningof their students.
The preservice experience in particularpoints out how traditionalquantitative
methods are unsatisfying;it is not feasible to use these methods to address the
questions that teachers raise regardingthe teaching and learning occurring in
their classrooms.
The chapteralso points out how the teachers' voices are an importantcompo-
nent of our understandingof the effectiveness of teacher researchexperiences.
Their voices also constitutean importantcomponentof our understandingof the
reality of the classroom and of children'sthinkingand learning.As teacherspar-
ticipate in contributingto the knowledge base about teaching and learning,they
become empowered,autonomousdecision makers.At the same time, the acade-
mic researchcommunitygains insights about teaching and learningthroughthe
teachers'perspectivesas they interpret,analyze, and describethe complexities of
their lives as teachersand their students'lives as learners.
156
Chapter 11
Where Do We Go from Here?
Susan Pirie
"Educationliteraturehas usually treated the alternativesto traditionalposi-

tivistic researchas a single approach-often called the 'qualitative'approach"
(Jacob, 1987, p. 1).
Even a brief study of this book would be sufficient to dispel this notion of a
singularityof definition for qualitativeresearchwithin the field of mathematics
education.Indeed,had the editormanagedto assemble all the authorsof the pre-
ceding chaptersin one room, it is my contentionthat a very lively debate would
have ensued and thatno one approachto qualitativeresearchon the learningand
teaching of mathematicswould have emerged as triumphantand representative.
It is not even clear to me that the authorswould have been in unanimousagree-
ment on their placing of the variety of methods in relationto the boundariesof
legitimate mathematics research mentioned in Chapter 2. What we, as
researchers,have done is takenthe traditionsof other,older disciplines and made
them our own to apply them to specific questions arising within mathematics
education.We have, in addition,attemptedto make overt the variationsand pre-
cautions that we have been obliged to employ to remain faithful to our own
notions of genuine "research."
The range of approaches,methods, attitudes, foci, concerns, and situations
encompassedis dramatic.In Chapter3, Ernestattemptsto constructa definition
of what he calls "the qualitativeresearch paradigm,"but he is quick to draw
attentionto the fact that this term can be misleading because it can enshrouda
confusion between methods and methodology; a qualitativemethodology may
well utilize quantitativemethods (Jacob, 1987). It is this confusion that some-
times gives rise to inappropriatecriticismsof the validity of certainresearch,and
I will returnto this topic later in this chapter.
Teppo opens the book with a focus on the diverse ways of knowing, and sub-
sequent authorscapture some of the myriad ways of coming to know that are
available to us. They have not provided recipes for success but illustrationsof
the reality of qualitative research-its problems and its strengths. Recurrent
themes run throughthe chaptersthat addresssome of the fundamentalquestions
facing qualitativeresearchin mathematicseducation.It is appropriate,therefore,
to close this book, but not the debate, by teasing out some of these themes and
acknowledgingthe tasks that lie ahead.
Susan Pirie 157
THE ROLE OF THEORY
Possibly the most crucialconsideration,if not always the most dominantin the
presentations,is thatof the role of theoryin qualitativeresearch.Thereare in fact
two issues here: first, the intentionof the researchitself in regardto its contribu-
tion to the developmentof generaltheorywithin mathematicseducationand sec-
ond, the theoreticalstandpointof the researcherand the theoreticalbasis for the
methods and instrumentsused. A subsidiaryof the second issue is the ability to
shift theoreticalpositions relativeto methodsand instrumentswithinthe research
process. In otherwords, theoretically,where is the researchgoing, where does it
have it roots, and how flexible should those roots be?
Teppo touches on the first issue in her introductionwhen she considersthe dif-
ferentgoals the authorshave with respectto the outcomes of theirresearch.Each
chapterthereafterseeks to characterize,or accountfor, or classify aspects of, the
teaching and learning of mathematics.There is an intention to work toward a
greaterunderstandingof the whole area of mathematicseducation, whetherthe
currentparticularfocus be on overarchingclassroomphenomenasuch as discus-
sion (Pirie), or on more detailed specificities, such as school beginners' concep-
tions of number(Neuman). The encompassing notion is that theory will (even-
tually) be generatedthat will explain in legitimate and precise ways the events
and occurrencesof classrooms, althoughin some cases, before this can happen
we need simply to betterunderstandthe natureof the phenomenon.
For Clarke,the startingpointis a little different.He intendsnot to buildnew the-
ory but to challenge an existing theoreticalassertion.He does not set out, howev-
er, to scientificallydisprovethe currentmodel of "negotiationof meaning"within
the classroom.Rather,employinga qualitative,deeply investigativemethodbased
on case study,he attemptsto elaboratethe model by providinga bettercharacteri-
zation of the ways in which meaningsare constructedwithin the classroom.Only
thenwill he be in a positionto refutethe value of currentmetaphoricalways of see-
ing the meaning-makingprocess. Attemptsto quantifyor rigorouslytest the inap-
propriatenessof the currentposition would themselves be inappropriate.In this
respect,one of the firmcharacteristicsof qualitativeresearch-that of the focus on
concretepracticesand particulars-is being enacted.
The second issue concerningthe theoreticalaspects of the roots of the report-
ed researchis more contentious.Because of its broadand diverse natureand the
span of historical backgroundsfrom which qualitative research springs, it is
essential for the communityto have some understandingof where in this spec-
trumparticularresearchersare situatingthemselves (see Chapters1 and 2). Since
we do not have a unified methodology built over time and universallyaccepted
by the mathematicseducation community, we need to make explicit where we,
as individuals, stand theoreticallyin relation to our work (Silver & Kilpatrick,
1994). Pirie is an excellent case in point. She does not approachthe researchwith
a preconceived attitudeabout the appropriatetheoretical stand but begins with
her research question and seeks out an appropriatetheoretical starting place
158 WhereDo We Gofrom Here?
(Denzin & Lincoln, 1994; Shulman, 1981). Goldin, indeed, states explicitly, in
the words of Davis (1984, p. 22), "without an appropriatetheory, one cannot
even state what the 'facts' are," although some of the authorsmight disagree
about what constitutedan appropriatetheory.
Relatedto the second issue is the interactivenatureof the relationshipbetween
the theoretical underpinningsand the understandingsthat develop during the
researchprocess. It is clear that there is a need to make overt our own theoreti-
cal underpinningsin the way thatJaworskimakes very explicit in her initial rad-
ical constructivistapproach.What is of especial interesthere is that we then see
one of the inner strengthsof qualitativeresearchat work-namely, the ability of
the researcherto make design decisions throughoutthe researchprocess and to
draw on multiple perspectives to understand the phenomenon in question
(Denzin & Lincoln, 1994; Janesick, 1994). As her gatheringof data and initial
analysisdictatethe makingof certainpragmaticdecisions, she acknowledgesthe
transformationof this theoreticalposition to that of social constructivism.I use
the word strengths here because it is the paradigmaticdenial of an absolutist
epistemology that allows for the human element, allied to the specific goals of
the research,to be influential.Indeed, for D'Ambrosio, the shifting of theoreti-
cal positions within the thinkingof the teachersshe was studyingwas one of the
majorfacets in the achievementof her goals.
The ability of preliminaryresearchfindings to influence underlyingmethod-
ological choices is also illustratedby Pirie (see Chapter6) and by Mousley,
Sullivan, and Waywood (see Chapter9). Their requirementsfor multiple, over-
lappingclassificationsof certainsets of datain no way contradictthe theoretical
backgroundsof theirmethods.In contrastto the scientific tradition,thereis need
here not for a dichotomybetween right and wrong categorizations,but for a sys-
tematic, controlled searching for informative classifications that will lead to
greaterunderstandingof the teaching and learningprocesses.
The variousways thattheoryis treatedby the authors,both as a frameworkfor
researchand as the end result of the inquiryprocess, illustratethe situatedrole
that theory plays in qualitativeresearch.Selection of an appropriateframework
should be relative to the contextual needs of each research question.
Additionally,theory that is elaboratedon, or generatedby, the researchshould
be understoodwithin the context of that research(Cobb, 1994; Janesick, 1994).
This recognition of contextualrelevance makes it crucial that researcherscare-
fully position and describe their use of theory within the contexts of their work.
To formjudgmentsrelatedto research,we must know from wherethe researcher
is coming.
VALIDITYAND RELIABILITY
Not anythinggoes, of course. Researchis not simply looking to see what hap-
pens. "Traditionsare important,even when one takes an open stance, because
they provide a set of orientingassumptions"(Jacob, 1987, p. 40). To undertake
Susan Pirie 159
even theory-buildingresearch, we must have specific, justifiable aims that are

based to some degree within a theoreticalreasoning. One approachthat has a
substantialfollowing among researchersis to seek to define and justify qualita-
tive researchfrom the perspectiveof its emergence as a new paradigm.Many of
the authorsdo this, giving legitimate, powerful, theoreticalfoundationsfor their
approaches.An indicationof the length of the road ahead, however, is given by
the set of theoreticalquestions posed by Goldin, to which he (modestly) claims
to give only preliminaryand partialanswers. He faces squarelyan awarenessof
the need to answerthose who challenge the value of qualitativeresearchon their
own terms.This is evident from the strugglehe reveals (andit is this detailedrev-
elation that is of such value to the emergence of a coherentresearchparadigm)
as he attemptsto align his researchwith the criteriafor acceptabilityimposed by
the traditionalmethodologies.
If our findings are to be considered of value by readersexternal to, or unfa-
miliar with, qualitativeapproaches,we must at the very least offer evidence that
the traditionalnotions of "validity"and "reliability,"if not directly applicable,
have their counterpartswithin the new researchparadigm( Eisenhart& Howe,
1992). We cannot say with impunitythat these notions are of no consequenceto
us; we must seek to put in theirplace equally cogent argumentsand tests for the
value of what we report.
The questionthatmust also be explicitly addressedis, "Of whatare we tryingto
establishthe validity?"Is it the descriptionwe are offering of the phenomenon?
The inferencesdrawnfromthatdescription?The theorybasedon these inferences?
At every stage we mustremaincredible,yet also realizethat"validityis relativeto
[the] purposesand circumstances"of the research(Maxwell, 1992, p. 283).
It is clear that as yet, no neat textbook tests or argumentsexist by which qual-
itative researchcan be judged. Indeedit is unlikely, by the very natureand intent
of the research,that such tests will ever exist (Silver & Kilpatrick,1994). At the
most, only very general guidelines for credibility can be applied across all
research.One such standardthat was touched on in the preceding section refers
to that of the coherence or goodness of fit of researchdesign-the fit among the
theoreticalframework,researchquestions, data-collectionprocedures,and inter-
pretativetechniques (Eisenhart& Howe, 1992). Pirie's hypotheticalframingof
her intendedresearchwithin differentqualitativemethodologies illustrateshow
the requirementsof coherence would have created a range of different studies
within the given context of student classroom discourse-these differences
reflecting the need to fit all elements of the researchdesign into a methodologi-
cally coherentwhole.
It is more appropriateto consider design-specific standardsthat relate judg-
ments of quality to particularmethodologicalperspectives.Thus, Jaworskicon-
cerns herself with "rigor"within her interpretative,ethnographicstance;Goldin,
with reproducibility,comparability,and generalizabilityin his use of scientific
inquiry; and Neuman, with the phenomenographiccriteria of interpretative
awareness.Making these criteriaexplicit helps us differentiatethe perspectives
160 WhereDo WeGofromHere?
needed to judge individual research. In addition, the nature of the knowledge

soughtby each authoris heightenedby understandingthe relevantcriteriawhere-
by the differentresearchis evaluated.
A crucialelementthatresearchersmust addressis thatof openness,both in care-
fully reportingthe researchdesign and in makingexplicit the subjectivenatureof
the researcher'srole in collectingand analyzingqualitativedata.Jaworski'suse of
a researchbiographyillustrateshow detailed descriptionsof context, researcher
reflections,and respondentvalidationsilluminateand make credibleher interpre-
tationsof "significant"events relativeto the goals of her research.
For many of the authors,it is the explicitness of the reporting(Ernst;Goldin;
Jaworski;Mousley et al.) and the awarenessof the need to remain"opento dis-
cussion and challenge" (Goldin) that give strengthto the value of the research.
As Mertonremarkedin 1968, when we have a considerablebody of qualitative
researchin which the researchershave exposed their actions and thinkingin the
kind of detail necessaryfor externalscrutiny,we can attainthe securityof foot-
ing accordedto quantitativeresearch.A quarterof a centurylater, that need has
yet to be met.
WHAT ARE THE DATA?
A common method of data collection is throughaudio and video recording.

The rationalefor this methodis usually thatthese recordingsoffer the best avail-
able way of preservingas much of the phenomenaunder scrutiny as possible.
Leaving aside the decisions along the lines of whetherto use one static camera,
or to use a camerafor each groupof pupils with a fixed microphone,or to use a
radio microphoneattachedto the teacher, the researcheris still faced with the
decision aboutwhat to do with the tapes. This is not a trivialdecision, and prag-
matic influences must be acknowledged (Pirie, 1996b). There are those who
remainfaithfulto the originalintentionof recordingeverything,and for them the
data are the tapes and all analysis is done purely from watching or listening to
the tapes. This leads to a very cumbersomeworkingenvironment,althoughmod-
ern software such as C-video is beginning to make the process of scanning,and
selecting data on which to focus, a more manageabletask. There are others who
realize that something will be lost by the process, but for whom, nevertheless,
transcribingis still considered a more appropriateway to work. For them the
data are the transcripts.
Field notes form anothercommon source of data in qualitativework. Within
the scientific paradigm,these will be in the form of codings on some preselect-
ed grid;once they have left the classroom, they can be operatedon, as symbols,
exactly as they stand.The field notes of the qualitativeresearcher,however, may
consist of sheets of notes, diagrams,prose, and personalcomments.Are these to
be the data as they stand? Does the researcherreduce them to coherent sum-
maries and work with those? Or indeed do they serve as aide-memoireto the
writing of a fuller account of events witnessed? What exactly is used for the
Susan Pirie 161
analysis?The interactionhere is not to set up one method as preferableto anoth-

er, but in the interestsof credibilitywhen reporting,we must, once again, return
to the notion of being as explicit as possible in the details we make available to
those who would judge our research.
LANGUAGE
In theiropening paragraphs,Mousley, Sullivan,and Waywood raise one of the
fundamentalconsiderationswithin qualitative research that is not given suffi-
cient prominence by other authors. This is the issue of language. By its very
nature,qualitativeresearchdemands that we move "the whole problem of lan-
guage from its peripheraland incidental position into the center."This means
more than acknowledgingand defendingloss of detail when we reduce auraland
visual data to writtenverbaldata.We tradein words-the words of the teachers,
the words of the students,the words of our own reporting.We need to concern
ourselves with the reality that we can only define and describe categories and
emergenttheories throughlanguage. We observe and personallyinterpretclass-
room contexts, and, leaving aside the question of whetherwe need language for
thought, we certainly need language to attemptto convey these thoughts to an
audience.
"Texts are not simple mirrorsof reality"(Nielson, 1995, p. 8). When we give
voice to our interpretations,our meaning is mediated through words, both by
ourselves and by our audience.The category labels we assign and the constructs
we define are frequently attempts to capture initially nebulous qualities and
behaviors.We can, as Goldin does in his scriptedinterviews,take greatcare with
our own language, but the responses are not given with such care! Students
respondto tasks spontaneouslyand without attentionto the possibility of misin-
terpretation.In the very structuredsetting in which he works, Goldin is able to
endeavorto confirm his interpretationsby encouraging"the child to constructa
concrete,externalrepresentation."In more open classroom settings, such confir-
matoryevidence is rarelyavailable.We constructour own personalmeaningsof
the discourse that we hear-and indeed we can do no other-but there is a need
to be alive to the notion that the meanings are ours and, further,that others will,
in turn,impose meanings on our words that are dependenton their histories and
culture.Mousley et al. make overt theirawarenessof the eventualitythatas ideas
are publishedthey inevitably take on new form in the everyday praxis of social
use. Language is not merely one of "a new set of researchemphases"(Ernest)
but at the very core of all we do in a qualitativeapproachto the study of mathe-
matical education.
REPORTINGTHE RESEARCH
Whateverform of datais selected for analysis and however carefulwe attempt
to be in the linguistic interpretationsof these data, one furtherproblemremains
162 WhereDo We Gofrom Here?
when it comes to reportingqualitativedata. As Jaworskisays, "a major disad-

vantage"of qualitativedataaretheir"lengthynature... in the space takento pre-
sent an account."Whetherthey take the form of shelves of videotapes or reams
of transcriptsand field notes, it simply is not feasible to offer the readerall the
data on which conclusions have been based. We are obliged to offer our sum-
maries, interpretations,and selections of what we deem importantfor the reader
know. We may claim that a single reportedcase can indicate a general concep-
tual category or property,that it is sufficiently generic to be taken as indicative
of the theory we are suggesting. Necessarily, however, it is ultimately "the
degree to which an informedreaderis convinced" (Jaworski)that proves to be
the test by which the researchand resultanttheoreticalimplicationsare judged.
We are forced to returnagain to the problem and necessity of validating our
work,for others.
As Mousley et al. suggest, developing technology may be coming to our aid.
Software such as NUD*ISTmakes handlingthe vast quantitiesof data somewhat
easier, but more than that, such productsallow others to access the data and see
the analyticaldecisions that were made by the researchers.The actual paths of
theory building become accessible; the actual data become available for repli-
cating the analysis or, indeed, for carryingout alternativeanalyses for the same
or differentpurposes.This lattersuggested use of the data raises new questions
concerningthe ethics implied by the transferenceof ownershipof data.This and
other ethical issues are discussed in the next section. Until technological
advances become even more accessible to the general reader,however, we are
dependenton the extent to which we are seen to be trustworthyin our report-
ing-until perhaps the definition of reader takes on another meaning, and
researchis communicatedthroughalternativetechnologies as yet undreamed.
ETHICALISSUES
One final concern, conspicuous by its absence within this collection of writ-
ings, needs raising:a considerationof some of the fundamentalethicalissues that
qualitativeresearchraises. These, I contend, are more acute than in quantitative
approaches.There is the dangerthat we are less able to accordanonymityto the
people involved. Caughtby the need to give sufficient detail to enable readersto
exercise theirpowers of judgmentaboutthe validity of the researchfindings, we
may find that merely changing the names of the participantsmay not disguise
their identity. The specifics of the detail may be enough to identify the school,
class, or subject.Of even greaterconcernis the widespreaduse of video data.At
oral presentationsof research,such as at conferences and seminars,the tempta-
tion to illustrate,enhance, or at least enliven the delivery with video clips from
the data is very strong. Anonymity here is impossible. Traditionally,subjects
have been assuredof the confidentialityof theircontributionto the advancement
of knowledge, but nowadays an additionalfactor is at play. With the ease of
widespread, internationaldissemination of research, there is the desirability,
Susan Pirie 163
even necessity, to specifically illustrate or define the cultural and educational

backgroundagainst which it takes place, thus making the masking of identities
more difficult.
How far can, or should, a confidentiality agreement hold? Locally?
Nationally?Internationally?For the main participants?For the bystanders?For
those whose voices can be heardoff camera?Do, and should, ethical considera-
tions preventus from gatheringcertainclassroom data?Can we never reveal our
original datato others?At what loss to the growthof understandingof the math-
ematical education process? Do our answers to the previous questions conflict
with the intent to make data more widely available throughmoder technology
as suggestedby Mousley et al.? Wheredoes the balancelie between personaland
societal concerns?Thereis a tension herebetween the need to reportin detail and
the protectionof those involved-a dilemmaof nontrivialproportionsthatneeds
to be grappledwith, since the publicationof researchis no longer confined to the
pages of traditionaljournals.
CLARIFYINGTHE RESEARCHPROCESS
It should be evident from the foregoing chapters that we are not out of the
woods in regardto the generalacceptanceof qualitativeapproachesas legitimate
researchparadigmsfor exploringthe understandingof mathematicseducation.It
is also evident thatthe approachesoffered by the authorsin this book have power
and merit and, in many cases, offer the only way that today's quest for under-
standingcan be furthered.As researchers,we must be criticalparticipantsin the
debatethat will eventuallylead to the clarificationof what is acceptableresearch
in the emerging discipline of mathematicseducation.There is "anintimaterela-
tionship between the research process and the findings it produces," and by
studying these processes, we gain insights not only into them but also into the
problems of qualitative, interpretativeresearch (Altheide & Johnson, 1994, p.
486). As participants,we need to addressthe challenge issued by Jacob (1987, p.
41) that it is a "betterunderstandingof the various qualitativetraditions(that)
offers the hope of a richerand fuller understandingof education."
164
References
Ahlberg,A. (1992). Att mota matematiskaproblem: En belysningav barns lirande. [Meeting math-
ematical problems: An elucidation of children's learning]. Goteborg: Acta Universitatis
Gothoburgensis.
Altheide, D. L., & Johnson,J. M. (1994). Criteriafor assessing interpretativevalidity in qualitative
research. In N. K. Denzin & Y. S. Lincoln (Eds.), Handbook of qualitative research (pp.
485-499). ThousandOaks: Sage Publications.
Altrichter,H., Posch, P., & Somekh, B. (1993). Teachersinvestigatetheir work:An introductionto
the methodsof action research. New York:Routledge.
Anthony,G. (1994). Learningstrategiesin the mathematicsclassroom:Whatcan we learnfrom stim-
ulated recall interviews?New Zealand Journal of EducationalStudies,29(2), 127-140.
Ascher, M. (1991). Ethnomathematics.Pacific Grove, CA: Brooks/Cole.
Aspray,W., & Kitcher,P. (Eds.). (1988). Historyand philosophyof modernmathematics(Minnesota
Studies in the Philosophy of Science, Vol. XI). Minneapolis:University of MinnesotaPress.
Austin, J. L. (1962). Sense and sensibilia. (G. J. Warock, Ed.). Oxford:Oxford UniversityPress.
AustralianEducationCouncil. (1990). A national statementon mathematicsfor Australianschools.
Canberra:AustralianEducationCouncil.
Bakhtin,M. (1979). Theaesthetics of verbal creation. Moscow: Bocharov.Cited in Todarov(1984).
Ball, S. J. (1982). The verificationand applicationof participantobservationcase study.In case study
methods5. Perspectiveson case study:Ethnography(pp. 141-164). Deakin, Australia:University
of Deakin Press.
Ball, S. J. (1990). Self doubt and soft data: Social and technical trajectoriesin ethnographicfield-
work. QualitativeStudies in Education,3(2), 157-171.
Barnes, D. (1982). From communicationto curriculum.London:Penguin.
Barnes, D., Britton, J., & Rosen, H. (1971). Language, the learner and the school (revised).
Harmondsworth:Penguin.
Barnes, D., Britton, J., & Torbe, M. (1986). Language, the learner and the school (3rd edition).
Harmondsworth:Penguin.
Barnes, D., & Todd, M. (1977). Communicationand learning in small groups. London:Routledge
& Keegan Paul.
Barrata-Lorton,M. (1976). Mathematicstheir way. Menlo Park,CA: Addison-Wesley.
Bartolini-Bussi,M. G. (1994). Theoreticaland empiricalapproachesto classroom interaction.In R.
Biehler, R. W. Scholz, R. Straesser,& B. Winkelmann(Eds.), The didactics of mathematicsas a
scientific discipline (pp. 121-132). Dordrecht:Kluwer.
Bassey, M. (1990-1991). On the natureof researchin education(Parts 1-3). ResearchIntelligence,
36, 35-38; 37, 39-44; 38, 16-18.
Bauersfeld, H. (1992). Classroom cultures from a social constructivist'sperspective. Educational
Studies in Mathematics,23, 467-481.
Bauersfeld,H. (1994). Theoreticalperspectiveson interactionin the mathematicsclassroom. In R.
Biehler, R. W. Scholz, R. Straesser,& B. Winkelmann(Eds.), The didactics of mathematicsas a
scientific discipline (pp. 133-146). Dordrecht:Kluwer.
Bell A. W., & Bassford,D. (1989). A conflict andinvestigationteachingmethodandan individualised
learningscheme-a comparativeexperimenton the teachingof fractions.Proceedings of the 13th
InternationalConferencefor the Psychologyof MathematicalEducation,(pp. 125-132). Paris.
Bell, A. W., Costello, J., & Kuchemann,D. (1983). Research on teaching and learning. Windsor:
NFER-Nelson.
Berger, P., & Luckmann,T. (1966). The social constructionof reality. Reprinted.London:Penguin
Books, 1967.
Bishop, A. J. (1985). The social constructionof meaning:A significant developmentfor mathemat-
ics education?For the Learningof Mathematics,5(1), 24-28.
References 165
Bishop, A. J. (1988). Mathematicalenculturation.Dordrecht:Kluwer.

Bishop, A. J. (1992). Internationalperspectives on researchin mathematicseducation. In D. A.
Grouws (Ed.), Handbookof research on mathematicsteaching and learning (pp. 710-723). New
York:Macmillan.
Bissex, G., & Bullock, R. H. (Eds.). (1987). Seeingfor ourselves: Case-studyresearch by teachers
of writing. Portsmouth,NH: Heinemann.
Blaikie, N. (1993). Approachesto social inquiry.Cambridge:Polity Press.
Bloor, D. (1976). Knowledgeand social imagery.London:Routledge & Kegan Paul.
Bodner,B. L., & Goldin, G. A. (1991a). Drawing a diagram:Observinga partiallydevelopedheuris-
tic process in college students.In F. Furinghetti(Ed.), Proceedings of the FifteenthInternational
Conferencefor the Psychology of MathematicsEducation (PME) (Vol. 1) (pp. 160-167). Genoa,
Italy: Dipartimentodi Matematicadell'Universia di Genova.
Bodner,B. L., & Goldin, G. A. (1991b). Cognitive obstacles of developmental-levelcollege students
in drawingdiagrams.In R. G. Underhill(Ed.), Proceedings of the ThirteenthAnnual Meeting of
the North American Chapter of the International Group for the Psychology of Mathematics
Education(Vol. I) (pp. 8-14). Blacksburg,VA: VirginiaPolytechnicInstituteand StateUniversity
Division of Curriculum& Instruction.
Boero, P., Dapueto, C., & Parenti,L. (1996). Didactics of mathematicsand the professionalknowl-
edge of teachers. In A. J. Bishop, K. Clements, C. Keitel, J. Kilpatrick,& C. Laborde (Eds.),
Internationalhandbookof mathematicseducation (pp. 1097-1121). Dordrecht:Kluwer.
Borasi, R. (1992). Learningmathematicsthroughinquiry.Portsmouth,NH: Heinemann.
Bredo, E. (1994). Reconstructingeducationalpsychology: Situatedcognition and Deweyian prag-
matism.EducationalPsychologist, 29(1), 23-35.
Breiter,C. (1997). Mathematicalknowledgeability.Paperpresentedat the meeting of the American
EducationalResearchAssociation, Chicago, IL.
Brissiaud,R. (1992). A tool for numberconstruction:Finger symbol sets. In J. Bideaud, C. Meljac,
& J. P. Fischer (Eds.), Pathways to number (pp. 41-65). Hillsdale, NJ: Lawrence Erlbaum
Associates.
Brown, D. (1997). Teachers as researchers:Some historical issues. In V. Zack, J. Mousley, & C.
Breen (Eds.), Developing practice: Teachers' inquiry and educational change (pp. 193-202).
Geelong, Australia:Centrefor Studies in Mathematics,Science and EnvironmentalEducation.
Bruner,J. (1985). Vygotsky: A historicaland conceptualperspective.In J. V. Wertsch(Ed.), Culture,
communicationand cognition. Cambridge:HarvardUniversity Press.
Bruner,J. S. (1990). Acts of meaning. Harvard,MA: CambridgePress.
Bullock, R. H. (Ed.). (1987). Teachers talk about their research:A round-tablediscussion. In G.
Bissex & R. H. Bullock (Eds.), Seeingfor ourselves: Case-studyresearch by teachers of writing
(pp. 145-163). Portsmouth,NH: Heinemann.
Burbules,N. C., & Rice, S. (1991). Dialogue and differences:Continuingthe conversation.Harvard
EducationalReview,61(4), 393-416.
Burgess, R. G. (1982). Field research: A source book andfield manual. London:Allan & Unwin.
Burgess, R. G. (1984). In thefield. London:Unwin Hyman.
Burgess, R. G. (1985). Strategies of educational research. London:FalmerPress.
Calderhead,J. (1987). Teachers' classroom decision making.London:Holt Rinehart& Winston.
Carpenter,T. P., & Moser, J. M. (1984). The developmentof additionand subtractionconcepts in
grades one to three. Journalfor Research in MathematicsEducation, 15(3), 179-202.
Carr,W., & Kemmis, S. (1986). Becoming critical. London:Falmer.
Carter,K. (1993). The place of story in the study of teaching and teacher education.Educational
Researcher,22(1), pp 5-12, 18.
Charles,R. I., & Lester, F. K. (1984). An evaluation of a process-orientedinstructionalprogramin
mathematicalproblemsolving in grades 5 and 7. Journalfor Research in MathematicsEducation,
15(1), 15-34.
Cicourel A. V. (1973). Cognitivesociology. London:Penguin.
166 References
Cizek, G. J. (1995). Crunchygranola and the hegemony of the narrative.EducationalResearcher,

24(3), 26-28.
Clarke,D. J. (1986). Conceptionsof mathematicalcompetence.Research in MathematicsEducation
in Australia, (No. 2), 17-23 (Australia).
Clarke, D. J. (1996). Refractionand reflection: Modelling the classroom negotiation of meaning.
RefLecT,2(1), 46-51.
Clarke, D. J., & Helme, S. (1997). The resolution of uncertaintyin mathematicsclassrooms. In F.
Biddulph & K. Carr (Eds.), People in mathematics education: Proceedings of the Twentieth
Annual Conferenceof the MathematicsEducationResearch Groupof Australasia (pp. 116-123).
Waikato,New Zealand:MERGA.
Clarke,D. J., & Kessel, C. (1995). To know and to be right: Studying the classroom negotiationof
meaning. In B. Atweh & S. Flavel (Eds.), Galtha: MERGA18: Proceedings of the 18th Annual
Conferenceof the MathematicsEducationResearch Groupof Australasia(pp. 170-177). Darwin,
NT: University of the NorthernTerritory.
Cobb, P. (1986). Clinical interviewingin the context of researchprograms.In G. Lappan& R. Even
(Eds.), Proceedings of the Eighth Annual Meeting of the North American Chapter of the
InternationalGroupfor the Psychology of MathematicsEducation:Plenary speeches and sympo-
sium (pp. 90-110). E. Lansing, MI: University of Michigan Departmentof Mathematicsand
Departmentof TeacherEducation.
Cobb, P. (1989). Experiential,cognitive, and anthropologicalperspectivesin mathematicseducation.
For the Learningof Mathematics,9(2), 32-42.
Cobb, P. (1994). Where is the mind? Constructivistand socioculturalperspectiveson mathematical
development.EducationalResearcher,23(7), 13-20.
Cobb, P. (1995). Continuingthe conversation:A response to Smith. EducationalResearcher,24(6),
25-27.
Cobb, P., & Bauersfeld,H. (Eds.). (1995). The emergence of mathematicalmeaning: Interactionin
classroom cultures. Hillsdale, NJ: LawrenceErlbaumAssociates.
Cobb, P., Jaworski,B., & Presmeg, N. (1996). Emergentand sociolculturalviews of mathematical
activity. In L. P. Steffe, P. Nesher, P. Cobb, G. Goldin, & B. Greer (Eds.), Theories of
MathematicalLearning(pp. 3-19). Mahwah,NJ: LawrenceErlbaumAssociates.
Cobb, P., & Yackel, E. (1996). Constructivist,emergent,and socioculturalperspectivesin the con-
text of developmentalresearch.EducationalPsychologist, 31(3/4), 175-190.
Cochran-Smith,M., & Lytle, S. (1993). Inside/Outside:Teacher research and knowledge. New
York:TeachersCollege Press.
Cockcroft,W. H. (1982). Mathematicscounts: Reportof the committee of inquiryinto the teaching
of mathematicsin schools. London:Her Majesty's StationeryOffice.
Cohen L., & Manion, L. (1989). Research methodsin education (3rd ed.). London:Routledge.
Confrey,J. (1995a). How compatibleare radicalconstructivism,socioculturalapproaches,and social
constructivism?In L. P. Steffe & J. Gale (Eds.), Constructivismin education (pp. 185-225).
Hillsdale, NJ: LawrenceErlbaumAssociates.
Confrey, J. (1995b). Studentvoice in examining "splitting"as an approachto ratio, proportionsand
fractions.In L. Meira & D. Carrher(Eds.), Proceedings of the 19th InternationalConferencefor
the Psychology of Mathematics Education (Vol. 1) (pp. 3-29). Recife, Brazil: Universidade
Federalde Pernambuco.
Cooney, T. J. (1988). Teachers'decision making.In D. Pimm (Ed.), Mathematicsteachers and chil-
dren. London:Hodderand Stoughton.
Cooney, T. J., & Henderson,K. B. (1972). Ways mathematicsteachershelp studentsorganizeknowl-
edge. Journalfor Research in MathematicsEducation,3, 21-31.
Crawford,K. (1996). Culturalprocesses and learning:Expectations,actions, and outcomes. In L. P.
Steffe, P. Nesher, P. Cobb, G. Goldin, & B. Greer (Eds.), Theories of mathematicallearning
(pp.131-147). Mahwah,NJ: LawrenceErlbaumAssociates.
Crawford,K., & Adler, J. (1996). Teachersas researchersin mathematicseducation.In A. J. Bishop,
K. Clements,C. Keitel, J. Kilpatrick,& C. Laborde(Eds.), Internationalhandbookof mathemat-
ics education(pp. 1187-1205). Dordrecht:Kluwer.
References 167
D'Ambrosio, U. (1985). Ethnomathematicsand its place in the history and pedagogy of mathemat-
ics. For the Learningof Mathematics,5(1), 44-48.
Darling-Hammond,L. (1996). The rightto learn and the advancementof teaching:Research,policy,
and practicefor democraticeducation.EducationalResearcher,25(6), 5-17.
Davis, P. J. (1993). Applied mathematicsas social contract.In S. Restivo, J. P. van Bebdegem, & R.
Fisher (Eds.), Math worlds: Philosophical and social studies of mathematicsand mathematics
education (pp. 182-194), Albany: State University of New York Press.
Davis, P. J., & Hersh, R. (1980). The mathematicalexperience.Boston: Birkhauser.
Davis, P. J., & Hersh, R. (1988). Descartes' dream. London:Penguin Books.
Davis, R. B. (1984). Learningmathematics:The cognitive science approach to mathematicseduca-
tion. Norwood, NJ: Ablex Publishing.
Davis, R. B. (1992). Reflections on where mathematicseducationnow stands and on where it may
be going. In D. A. Grouws (Ed.), Handbookof research on mathematicsteaching and learning
(pp. 724-734). New York: Macmillan.
Davis, R. B. (1994). What mathematicsshould students learn?Journal of MathematicalBehavior,
13, 3-33.
Davis, R. B., Maher, C. A., & Noddings, N. (Eds.). (1990). Journalfor Research in Mathematics
Education,MonographNo. 4, Constructivistviews on the teaching and learning of mathematics.
Reston, VA: National Council of Teachersof Mathematics.
DeBellis, V. A. (1996). Interactionsbetweenaffect and cognition duringmathematicalproblemsolv-
ing: A two-yearcase studyoffour elementaryschool children.Unpublisheddoctoraldissertation,
RutgersUniversity GraduateSchool of Education,(UniversityMicrofilms No. 96-30716).
DeBellis, V. A., & Goldin, G. A. (1991). Interactionsbetween cognition and affect in eight high
school students' individual problem solving. In R. G. Underhill (Ed.), Proceedings of the
ThirteenthAnnual Meeting of the North American Chapter of the InternationalGroupfor the
Psychology of MathematicsEducation(Vol. I) (pp. 29-35). Blacksburg,VA: VirginiaPolytechnic
Instituteand State University, Division of Curriculum& Instruction.
DeBellis, V. A., & Goldin, G. A. (1993). Analysis of interactionsbetween affect and cognition in ele-
mentaryschool childrenduringproblemsolving. In J. R. Becker & B. Pence (Eds.), Proceedings
of the FifteenthAnnualMeetingof the NorthAmericanChapterof the InternationalGroupfor the
Psychology of MathematicsEducation (Vol. II) (pp. 56-62). Pacific Grove, CA: San Jose State
University Centerfor Mathematicsand ComputerScience Education.
DeBellis, V. A., & Goldin, G. A. (1997). The affective domainin mathematicalproblem-solving.In
E. Pehkonen(Ed.), Proceedings of the Twenty-firstConferenceof the InternationalGroupfor the
Psychology of Mathematics Education (PME) (Vol. II) (pp. 209-216). Helsinki, Finland:
University of Helsinki Departmentof TeacherEducation.
Delamont, S., & Hamilton, D. (1986). Revisiting classroom research: A cautionary tale. In M.
Hammersley (Ed.), Controversies in classroom research (pp. 25-43). Milton Keynes: Open
University Press.
Denzin, N. K., & Lincoln, Y. S. (1994). Introduction:Enteringthe field of qualitativeresearch.In N.
K. Denzin & Y. S. Lincoln (Eds.), The handbookof qualitative research (pp. 1-17). Thousand
Oaks, CA: Sage Publications.
Descartes,R. (1955). A discourse on method.Translationin R. Descartes,Philosophicalworks, Vol.
1. New York: Dover Press. (Originalwork published 1637)
Devlin, K. (1997). Mathematics:The science of patterns. New York: Scientific AmericanLibrary.
Donmoyer,R. (1996). Educationalresearchin an era of paradigmproliferation:What's a journaledi-
tor to do? EducationalResearcher,25(2), 19-25.
Dossey, J. A. (1992). The natureof mathematics:Its role and its influence. In D. A. Grouws (Ed.),
Handbookof research on mathematicsteaching and learning (pp. 39-48). New York:Macmillan.
Dryfus, T. (1991). Advanced mathematicalthinking processes. In D. Tall (Ed.), Advancedmathe-
matical thinking.Dordrecht:Kluwer.
Dunne, M., & Johnston,J. (1992). An awarenessof epistemological assumptions:The case of gen-
der studies. InternationalJournal of Science Education, 14(5), 515-526.
168 References
Eco, U. (1984). Semioticsand thephilosophyof language. Bloomington,IN: IndianaUniversityPress.

Edwards,A. D., & Furlong,V. J. (1978). The language of teaching. London:Heinemann.
Edwards,D., & Mercer,N. (1987). Commonknowledge-the developmentof understandingin the
classroom. London:Methuen.
Eisenhart,M. A. (1988). The ethnographicresearchtraditionand mathematicseducation research.
Journalfor Research in MathematicsEducation,19(2), 99-114.
Eisenhart,M. A., & Howe, K. R. (1992). Validity in educationalresearch.In M. D. LeCompte,W.
L. Millroy, & J. Preissle (Eds.), Thehandbookof qualitativeresearch in education(pp. 643-680).
San Diego: Academic Press.
Eisner, E. W. (1993). Forms of understandingand the futureof educationalresearch.Educational
Researcher,22(7), 5-11.
Erlwanger,S. H. (1973). Benny's conception of rules and answers in IPI mathematics.Journal of
Children'sMathematicalBehaviour, 1(2), 7-26.
Ernest,P. (1991). Thephilosophy of mathematicseducation.London:FalmerPress.
Ernest, P. (1994a). Mathematics,education and philosophy: An internationalperspective. London:
FalmerPress.
Ernest, P. (Ed.). (1994b). Constructingmathematicalknowledge: Epistemologyand mathematics
education. London:FalmerPress.
Ernest, P. (1994c). An introductionto educational research methodologyand paradigms. Exeter:
University of Exeter.
Ernest,P. (1994d). A perspectiveon researchin mathematicseducation.In J. Kilpatrick,J. Fey, &
A. Sierpinska(Eds.), Whatis research in mathematicseducationand what are its results?(ICMI
Study Pre-ConferenceProceedings).College Park,MD: Universityof Maryland.
Ernest, P. (1997). Social constructivism as a philosophy of mathematics. Albany, NY: State
University of New York Press.
Fenstermacher,G. D., & Richardson,V. (1994). Promoting confusion in educational psychology:
How is it done? EducationalPsychology, 29(1), 49-55.
Feyerabend,P. (1975). Against method.London:New Left Books.
Fischer,J. P. (1992). Subitizing:The discontinityafterthree.In J. Bideaud,C. Meljac, & J. P. Fischer
(Eds.), Pathways to number:Children'sdeveloping numericalabilities (pp. 119-131). Hillsdale,
NJ: LawrenceErlbaumAssociates.
Flanders,N. A. (1970). Analyzingteacher behaviour.Reading,MA: Addison-Wesley.
Foucault,M. (1972). The archaeology of knowledge.London:Tavistock.
Freeman,K. (Ed.). (1956). Ancilla to the pre-socratic philosophers. Oxford:Basil Blackwell.
Freire,P. (1972). Pedagogy of the oppressed.London:Penguin Books.
Freudenthal,H. (1991). Revisitingmathematicseducation: China lectures. Dordrecht:Kluwer.
Frid, S. (1992). Undergraduate calculus students' language use and sources of conviction.
Unpublisheddoctoraldissertation,Departmentof SecondaryEducation,University of Alberta.
Fuller, S. (1988). Social epistemology.Bloomington,IN: IndianaUniversity Press.
Furlong,V. J., & Edwards,A. D. (1986). Languagein classroominteraction:Theoryand data.In M.
Hammersley (Ed.), Controversies in classroom research (pp 51-61). Milton Keynes: Open
University Press.
Fuson, K. (1992). Relationshipsbetween countingand cardinalityfrom age 2 to age 8. In J. Bideaud,
C. Meljac, & J. P. Fischer (Eds.), Pathways to number (pp 127-149). Hillsdale, NJ: Lawrence
ErlbaumAssociates.
Gadamer,H. G. (1975). Truthand method.New York: Seabury.
Gage, N. L. (1989). The paradigmwars and their aftermath:A "historical"sketch of researchon
teaching since 1989. EducationalResearcher, 18(7), 4-10.
Galbraith,P. (1991). Paradigms, problems and assessment: Some ideological implications. Paper
presentedat MathematicsEducationResearchGroupof AustralasiaConference,Perth,Australia.
Galton, M., Simon, B., & Croll, P. (1980). Inside the primary classroom (the ORACLE project).
London:Routledge.
References 169
Gardner,H. (1983). Frames of mind. New York: Basic Books.

Gardner,H. (1987). The mind's new science (rev. ed.). New York: Basic Books.
Garfinkel,H. (1967). Studies in ethnomethodology.Hemel Hempstead:PrenticeHall.
Geertz, C. (1973). Thick description:Towards an interpretivetheory of culture.In C. Geertz (Ed.),
The interpretationof cultures (pp. 3-32). New York: Basic Books.
Gelman, R., & Gallistel, C. R. (1978). The child's understandingof number. London: Harvard
University Press.
Gerdes,P. (1985). Conditionsand strategiesfor emancipatorymathematicseducationin underdevel-
oped countries.For the Learningof Mathematics,5(1), 15-20.
Gerdes, P. (1996). Ethnomathematicsand mathematicseducation.In A. Bishop (Ed.), International
handbookof research in mathematicseducation (pp. 909-943). Dordrecht:Kluwer.
Gergen, K. (1985). The social constructionist movement in modem psychology. American
Psychologist, 40, 266-275.
Giddens, A. (1984). The constitutionof society. Oxford:Polity Press.
Gillies, D. A. (Ed.). (1992). Revolutionsin mathematics.Oxford:ClarendonPress.
Ginsburg, H. (1981). The clinical interview in psychological research on mathematicalthinking:
Aims, rationales,techniques.For the Learningof Mathematics,3, 4-11.
Ginsburg,H. (1983). The developmentof mathematicalthinking.New York:Academy Press.
Giorgi, A. (1986). A phenomenological analysis of descriptions of concepts of learning obtained
from a phenomenographicperspective.Goteborg:Publikationerfrdn institutionenfor pedagogik,
GoiteborgsUniversitet,18.
Gitlin, A. D. (1990). Educative research,voice, and school change. Harvard Educational Review,
60(4), 443-466.
Glaser, B. G., & Strauss,A. L. (1967). The discovery of groundedtheory: Strategiesfor qualitative
research. London:Weidenfeld & Nicholson.
Goffman, E. (1971). Thepresentationof self in everydaylife. London:Penguin Books.
Goldin, G. A. (1982). The measureof problem-solvingoutcomes. In F. K. Lester, Jr. & J. Garofalo
(Eds.), Mathematical problem solving: Issues in research (pp. 87-101). Philadelphia, PA:
FranklinInstitutePress (acquiredby LawrenceErlbaumAssociates, Hillsdale, NJ).
Goldin, G. A. (1985). Studyingchildren'suse of heuristicprocesses for mathematicalproblemsolv-
ing throughstructuredclinical interviews.In S. Damarin& M. Shelton (Eds.), Proceedings of the
Seventh Annual Meeting of the North American Chapter of the International Group for the
Psychology of Mathematics Education (pp. 94-99). Columbus, OH: Ohio State University
Departmentof EducationalTheory and Practice.
Goldin, G. A. (1986). Commentson structuredindividualinterview methods for the study of prob-
lem solving. In G. Lappan & R. Even (Eds.), Proceedings of the Eighth Annual Meeting of the
North American Chapter of the International Group for the Psychology of Mathematics
Education: Plenary speeches and symposium (pp. 111-119). E. Lansing, MI: University of
Michigan Departmentof Mathematicsand Departmentof TeacherEducation.
Goldin, G. A. (1987). Cognitive representationalsystems for mathematicalproblem solving. In C.
Janvier (Ed.), Problems of representation in the teaching and learning of mathematics (pp.
125-145). Hillsdale, NJ: LawrenceErlbaumAssociates.
Goldin, G. A. (1992a). Towardan assessmentframeworkfor school mathematics.In R. Lesh & S. J.
Lamon (Eds.), Assessment of authentic performance in school mathematics (pp. 63-88).
Washington,DC: AmericanAssociation for the Advancementof Science.
Goldin, G. A. (1992b). On developing a unified model for the psychology of mathematicallearning
and problem solving. In W. Geeslin & K. Graham (Eds.), Proceedings of the Sixteenth
International Conferencefor the Psychology of Mathematics Education (PME) (Vol. III) (pp.
235-261). Durham,NH: University of New HampshireDepartmentof Mathematics.
Goldin, G. A. (1993). Observing mathematicalproblem solving: Perspectives on structured,task-
based interviews.In B. Atweh, C. Kanes, M. Carss, & G. Booker (Eds.), Contextsin mathematics
education: Proceedings of the Sixteenth Annual Conference of the Mathematics Education
Research Group of Australasia (MERGA-16)(pp. 303-310). Queensland,Australia:Queensland
University of Technology-Kelvin Grove Campus (MERGA).
170 References
Goldin, G. A., DeBellis, V. A., DeWindt-King,A. M., Passantino,C. B., & Zang, R. (1993). Task-
based interviews for a longitudinal study of children's mathematical development. In I.
Hirabayashi,N. Nihda, K. Shigematsu, & F-L. Lin (Eds.), Proceedings of the Seventeenth
International Conferencefor the Psychology of Mathematics Education (PME) (Vol. I) (pp.
197-203). Tsukuba,Japan:PME ProgramCommittee.
Goldin, G. A., & Landis, J. H. (1985). A problem-solvinginterview with "Stan"(age 11). In S.
Damarin& M. Shelton (Eds.), Proceedings of the SeventhAnnual Meeting the North American
Chapterof the InternationalGroupfor the Psychology of MathematicsEducation(pp. 100-105).
Columbus,OH: Ohio State University Departmentof EducationalTheory and Practice.
Goldin, G. A., & Landis, J. H. (1986). A study of children's mathematicalproblem-solvingheuris-
tics. In L. Burton& C. Hoyles (Eds.), Proceedings of the TenthInternationalConferencefor the
Psychology of Mathematics Education (PME) (Vol. I) (pp. 427-432). London: University of
London Instituteof Education.
Goldin, G. A., & McClintock,C. E. (Eds.). (1980). Taskvariables in mathematicalproblemsolving.
Columbus, OH: ERIC Clearinghousefor Science, Mathematicsand EnvironmentalEducation.
Reissued by Franklin Institute Press, Philadelphia (1984) (acquired by Lawrence Erlbaum
Associates, Hillsdale, NJ).
Goldin, G. A., & Passantino,C. B. (1996). A longitudinalstudy of children'sfractionrepresentations
and problem-solvingbehavior. In L. Puig & A. GutiErrez(Eds.), Proceedings of the Twentieth
International Conferencefor the Psychology of MathematicsEducation (PME) (Vol. III) (pp.
3-10). Valencia, Spain:Universidadde Valencia Dept. de Didactica de la Matematica.
Good, T. L., Grouws, D. A., & Ebmeier, H. (1983). Active mathematics teaching. New York:
Longman.
Goswami, D., & Stillman,P. (1987). Reclaimingthe classroom: Teacher research as an agencyfor
change. Upper Montclair,NJ: Boynton/Cook.
Gray, E. M., & Tall, D. 0. (1994). Duality, ambiguityand flexibility: A procepturalview of simple
arithmetic.Journalfor Research in MathematicsEducation,25(2), 115-141.
Greer,B. (1996). Theories of mathematicseducation:The role of cognitive analysis. In L. P. Steffe,
P. Nesher, P. Cobb, G. Goldin, & B. Greer (Eds.), Theories of mathematical learning (pp.
179-196). Hillsdale, NJ: LawrenceErlbaumAssociates.
Guba,E. G., & Lincoln, Y. S. (1989). Fourthgeneration evaluation.London:Sage.
Habermas,J. (1971). Knowledgeand humaninterests.London:Heinemann.
Habermas, J. (1981). The theory of communicative action (2 vols.) (T. McCarthy, Trans.).
Cambridge:Polity Press.
Hammersley,M. (1990). Classroomethnography.Milton Keynes: Open UniversityPress.
Hammersley, M. (1993). On constructivism and educational research methodology. PME News
(May). Oxford:OxfordUniversity Departmentof EducationalStudies.
Hammersley,M., & Atkinson,P. (1990). Ethnographyprinciples in practice. London:Routledge.
Harding,S. (Ed.). (1987). Feminismand methodology:Social science issues. Milton Keynes: Open
University Press.
Harding,S. (1991). Whosescience? Whoseknowledge?Milton Keynes: Open University Press.
Harr6,R. (1979). Social being. Oxford:Basil Blackwell.
Harre,R., & Gillett, G. (1994). The discursivemind. London:Sage Publications.
Hart,K. (Ed.). (1981). Children'sunderstandingof mathematics:11-16. London:John Murray.
Hart,L. (1986). Small groupproblemsolving as a data source for the individual.In G. Lappan& R.
Even (Eds.), Proceedings of the Eighth Annual Meeting of the North American Chapter of the
InternationalGroupfor the Psychology of MathematicsEducation:Plenary speeches and sympo-
sium (pp. 120-127). E. Lansing, MI: University of Michigan Departmentof Mathematicsand
Departmentof TeacherEducation.
Hatano,G. (1982). Learningto add and subtract:A Japaneseperspective.In T. P. Carpenter,M. M.
Moser, & T. A. Romberg(Eds.), Additionand subtraction:A cognitiveperspective.Hillsdale, NJ:
LawrenceErlbaumAssociates.
References 171
Henriques,J. H, Urwin, W., Venn, C., & Walkerdine,V. (1984). Changingthe subject:Psychology,
social regulationand subjectivity.London:Methuen.
Hitchcock, G., & Hughes, D. (1992). Research and the teacher. London:Routledge.
Hobbes, T. (1651). Leviathan.(Reprintedin FontanaLibrary.)Glasgow: William Collins, 1962.
Hoyles, C., Healey, L., & Sutherland,R. (1990). The role of peer group discussion in mathematical
environments:Reportto the LeverhulmeTrust.London:University of London.
Hubbard,R., & Power, B. M. (1994). The art of classroom inquiry: A handbookfor teacher
researchers.Portsmouth,NH: HeinemannPress.
Ihde, D. (1997). Experimentalphenomenography.New York:Longman.
Inhelder, B., & Piaget, J. (1958). The growth of logical thinkingfrom childhood to adolescence.
London:Routledge & Kegan Paul.
Jacob, E. (1987). Qualitativeresearchtraditions:A review. Review of EducationalResearch,57(1),
1-50.
Janesick,V. J. (1994). The dance of qualitativeresearchdesign: Metaphor,methodolatry,and mean-
ing. In N. K. Denzin & Y. S. Lincoln (Eds.), The handbookof qualitativeresearch (pp. 209-219).
ThousandOaks, CA: Sage Publications.
Jaworski,B. (1992). Mathematicsteaching,what is it? For the Learningof Mathematics,12(1), 8-14.
Jaworski,B. (1994). Investigatingmathematicsteaching: A constructivistenquiry.London:Falmer
Press.
Joseph, G. G. (1991). The crest of the peacock. London:I. B. Tauris.
Kennedy, M. M. (1997). The connection between researchand practice. Educational Researcher,
26(7), 4-12.
Kieran,C. (1994). Doing and seeing things differently:A 25-year retrospectiveof mathematicsedu-
cation researchon learning.Journalfor Research in MathematicsEducation,25(6), 583-607.
Kieren, T. E., & Pirie, S. E. B. (1991). Recursion and the mathematicalexperience. In L. P. Steffe
(Ed.), Epistemological foundations of mathematical experience (pp. 78-101). New York:
Springer-Verlag.
Kilbom, W. (1979). PUMP-project-backgrund och erfarenheter(The PUMP-project,background
and experiences). FoU nr. 17. Stockholm:Liber.
Kilpatrick,J. (1987). What constructivismmight be in mathematicseducation.In J. C. Bergeron,N.
Herscovics, & C. Kieran(Eds.), Proceedings of PME 11 Conference,(Vol. 1, pp. 3-27). Montreal:
University of Montreal.
Kincheloe, J. (1991). Teachersas researchers: Qualitativeinquiryas a path to empowerment.New
York:FalmerPress.
Kitcher,P. (1984). The nature of mathematicalknowledge.Oxford:Oxford UniversityPress.
Kitcher,P., & Aspray, W. (1988). An opinionatedintroduction.In W. Aspray & P. Kitcher (Eds),
History and philosophy of modernmathematics(MinnesotaStudies in the Philosophy of Science,
Vol. XI, pp. 3-57). Minneapolis:University of MinnesotaPress.
Kline, M. (1980). Mathematics:The loss of certainty.Oxford:OxfordUniversityPress.
Kuhn,T. S. (1970). The structureof scientific revolutions(2nd ed.). Chicago:Universityof Chicago
Press.
Kvale, S. (1989). To validate is to question. In S. Kvale (Ed.), Issues of validity in qualitative
research (pp. 73-91). Lund:Studentlitteratur.
Kvale, S. (1994). Ten standard objections to qualitative research interviews. Journal of
PhenomenologicalPsychology, 25, 147-173.
Lakatos,I. (1976). Proofs and refutations.Cambridge:CambridgeUniversity Press.
Lakatos, I., & Musgrave, A. (Eds.). (1970). Criticism and the growth of knowledge. Cambridge:
CambridgeUniversity Press.
Larson, S. (1993). Om kvalitet I kvalitativa studier. (On quality in qualitative studies). Nordisk
Pedagogik, 4(93), 194-211.
Lave, J., & Wenger, E. (1991). Situated learning: Legitimateperipheralparticipation. Cambridge:
172 References
LeCompte,M. D., Millroy, W. L., & Preissle,J. (Eds.). (1992). The handbookof qualitativeresearch
in education. San Diego: Academic Press.
Lerman,S. (1996). Intersubjectivityin mathematicslearning:A challengeto the radicalconstructivist
paradigm?Journalfor Research in MathematicsEducation,27(2), 133-150.
Lesh, R., & Lamon, S. J. (Eds.). (1992). Assessmentof authenticperformancein school mathemat-
ics. Washington,DC: AmericanAssociation for the Advancementof Science.
Lester,F. K., Jr., Kehle, P. E., & Birgisson, G. (1996). In pursuit of practical wisdom in mathemat-
ics education research. Paperpresentedat the ResearchPresession of the Annual Meeting of the
National Council of Teachersof Mathematics,San Diego, CA.
Lincoln, Y. S., & Guba,E. (1985). Naturalisticinquiry.Beverly Hills, CA: Sage.
Livingston,E. (1986). The ethnomethodologicalfoundationsof mathematics.London:Routledge &
Kegan Paul.
Lybeck, L. (1981). Arkimedesi klassen: En amnespedagogiskberattelse. (Archimedesin the class-
room: A subjectpedagogical story) Goteborg:Acta UniversitatisGothoburgensis.
Lyotard,J. F. (1984). Thepostmoder condition:A report on knowledge.Manchester:Manchester
University Press.
Maher,C. A., & Martino,A. M. (1996). Young childreninvent methodsof proof: The gang of four.
In L. P. Steffe, P. Nesher, P. Cobb, G. Goldin, & B. Greer(Eds.), Theoriesof mathematicallearn-
ing (pp. 431-447). Hillsdale, NJ: LawrenceErlbaumAssociates.
Marton,F. (1981). Phenomenography:Describingconceptionsof the world aroundus. Instructional
Science, 10, 177-200.
Marton, F., Beaty, E., & Dal'Alba, G. (1993). Conceptions of learning. InternationalJournal of
EducationalResearch, 19, 277-300.
Marton, F., & Booth, S. (1997). Learning and awareness. Hillsdale, NJ: Lawrence Erlbaum
Associates.
Marton,F., & Neuman,D. (1996). Phenomenographyand children'sexperienceof division. In L. P.
Steffe, P. Nesher, P. Cobb, G. A. Goldin, & B. Greer (Eds.), Theories of mathematicallearning
(pp. 315-334). Mahwah,NJ: LawrenceErlbaumAssociates.
Maxwell, J. A. (1992). Understandingand validity in qualitative research. Harvard Educational
Review, 62(3), 279-300.
McLeod, D. B., & Adams, V. M. (Eds.). (1989). Affect and mathematicalproblem solving: A new
perspective. New York: Springer-Verlag.
McLeod, D. B., Carpenter,T. P., McCornack,R. L., & Skvarcius, R. (1978). Cognitive style and
mathematicslearning:The interactionof field independenceand instructionaltreatmentin numer-
ation systems. Journalfor Research in MathematicsEducation,9, 163-174.
Mead, G. H. (1934). Mind, self and society. Chicago:University of Chicago Press.
Mehan, H. (1979). Learning lessons: Social organization in the classroom. Cambridge:Harvard
UniversityPress.
Mehan,M. (1985). Structureof classroomdiscourse.In T.A. Van Dijk (Ed.), Handbookof discourse
analysis (Vol. 3) (pp. 119-131). London:Academic Press.
Mellin-Olsen, S. (1987). Thepolitics of mathematicseducation. Dordrecht:Reidel.
Merton,R. K. (1968). Social theoryand social structure.New York:The Free Press.
Meyer, R. (1995). Servicing-in:An approachto teacherand staff development.TeacherResearch,2
(1), 1-16.
Miles, M. B., & Huberman,A. M. (1990). Animadversionsand reflectionson the uses of qualitative
inquiry.In E.G. Eisner& A. Peshkin (Eds.), Qualitativeinquiryin education(pp. 339-357). New
York:TeachersCollege Press.
Minsky, M. (1986). The society of mind. New York: Simon & Schuster.
Mousley, J., & Clements,M. A. (1990). The cultureof mathematicsclassrooms.In M. A. Clements
(Ed.), Whithermathematics?(pp. 397-406). Melbourne:MathematicalAssociation of Victoria.
Mousley, J., Sullivan, P., & Clements, M. A. (1991, July). The perceptions which studentteachers
have of teachingpractices in classroom observed duringfield experience.Paperpresentedat the
annualconference of the MathematicsEducationLecturer'sAssociation, Perth.
References 173
National Council of Teachers of Mathematics. (1989). Curriculumand evaluation standardsfor

school mathematics.Reston, VA: National Council of Teachersof Mathematics.
Neuman, D. (1987). The origin of arithmeticskills: A phenomenographicapproach. Goteborg:Acta
UniversitatisGothoburgensis.
Neuman,D. (1993). Exploringnumbersthroughsubtractionin primarymathematics.In T. Breitig, I.
Huntley,& G. Kaiser-Messmer(Eds.), Teachingand learning mathematicsin context(pp. 37-50).
New York:Ellis Horwood.
Neuman, D. (1994). Five fingers on one hand and ten on the other:A case study in learningthrough
interaction.In J. P. da Ponte & J. F. Matos (Eds.), Proceedings of the EighteenthInternational
Conference for the Psychology of Mathematics Education, (Vol. 2, pp. 352-359). Lisbon:
Departamentode Educaqio,Faculdadede Ciencias da Universidadide Lisboa.
Neuman, D. (1997). Immediate and sequential experiences of numbers. In E. Pehkonen (Ed.),
Proceedings of the Twenty-firstInternational Conferencefor the Psychology of Mathematics
Education, (Vol. 3, pp. 288-295). Lahti:Universityof Helsinki.
Newman, C. F., & Pirie, S. E. B. (1990). Watching the growth of understanding. Midlands
MathematicsEducationSeminar,University of Birmingham.
Nias, J., & Groundwater-Smith, S. (1988). Theenquiringteacher:Supportingand sustainingteacher
research. New York: FalmerPress.
Nielson, H. B. (1995). Seductive texts with serious intention.EducationalResearcher,24(1), 4-12.
Noddings, N. (1990). Constructivismin mathematicseducation.In R. B. Davis, C. A. Maher,& N.
Noddings (Eds.), Constructivistviews on the teaching and learning of mathematics(pp. 7-18).
Reston, VA: National Council of Teachersof Mathematics.
Noddings, N. (1994). William Brownell and the search for meaning. Journal for Research in
MathematicsEducation,25(5), 524-525.
Nunes, T. (1992). Ethnomathematicsand everyday cognition. In D. A. Grouws, (Ed.), Handbookof
research on mathematicsteaching and learning (pp. 557-574). New York:Macmillan.
Odman,P. J. (1988). Heumeneutics.In J. P. Keeves (Ed.), Educationalresearch,methodology and
measurement:An internationalhandbook(pp. 63-69). Oxford:Permagon.
Passantino,C. B. (1997). Making connections: A cognitive analysis of children's problem-solving
behavior involvingfraction representationsand strategies. Unpublished doctoral dissertation,
RutgersUniversity GraduateSchool of Education.(UniversityMicrofilmsNo. 97-17241)
Phinney,M. Y., & Ketterling,T. (1997). Dialoguejournals,literature,and a sixth-gradeurbanIndian
class. TeacherResearch, 4(2), 22-41.
Pirie, S. E. B. (1988). Understanding-instrumental, relational, formal, intuitive..., how can we
know? For The Learningof Mathematics,8, 2-6.
Pirie, S. E. B. (199la). Peer discussion in the context of mathematicalproblemsolving. In K. Durkin
& B. Shire (Eds.), Language in mathematicaleducation: Research and practice (pp. 143-161).
Milton Keynes: Open UniversityPress.
Pirie, S. E. B. (1991b). Mathematicaldiscussion: Incoherentexchanges or shared understandings.
Language and Education, 11(3), 243-257.
Pirie, S. E. B. (1996a). Is there anybody listening? In P. E. Elliot (Ed.), Communicationin mathe-
matics K-12: 1996 Yearbook (pp. 105-115). Reston VA: National Council of Teachers of
Mathematics.
Pirie, S. E. B. (1996b). Classroomvideo recording: When,why and how does it offer a valuable data
sourcefor qualitativeresearch? Paperpresentedthe 18th AnnualMeeting of the North American
Chapterof the InternationalGroupfor the Psychology of MathematicsEducation.PanamaCity,
FL. (ERIC Document ReproductionService No. ED 401 128).
Pirie, S. E. B. (forthcoming).Crossing the gulf between thoughtand symbol-language as slippery
stepping stones. In M. B. Bussi, A. Sierpinska, & H. Steinbring (Eds.), Language and commu-
nication in the mathematics classroom. Reston VA: National Council of Teachers of
Mathematics.
Pirie, S. E. B., & Schwarzenberger,R. L. E. (1988a). Mathematicaldiscussion and mathematical
understanding.EducationalStudies in Mathematics,19, 459-470.
174 References
Pirie, S. E. B., & Schwarzenberger,R. L. E. (1988b). Mathematicaldiscussion-it's what you say

and what you do. MathematicsTeaching, 123, 58-60.
Polanyi, M. (1958). Personal knowledge.London:Routledge & Kegan Paul.
Popper,K. (1959). The logic of scientific discovery.London:Hutchinson.
Pramling, I. (1991). Learning about "the shop": An approach to learning in preschool. Early
ChildhoodResearch Quarterly,6, 151-166.
Pramling, I. (1994). Fenomenografioch pedagogisk praktik.(Phenomenographyand pedagogical
practice).NordiskPedagogik, 4, 227-239.
Putnam, H. (1975). Mathematics,matter and method (Philosophical Papers Vol. 1). Cambridge:
QualitativeSolutions and Research. Non-Numerical UnstructuredData * Indexing Searching and
Theorizing (NUD-IST)[Software package]. (1994). Melbourne: Qualitative Solutions and
Research.
Radford, L. (1997). On psychology, historical epistemology, and the teaching of mathematics:
Towarda socio-culturalhistory of mathematics.For the Learningof Mathematics,17(1), 26-33.
Renz, P. (1997). An excess of jargon and a poor aim? Focus, 17(3), 1-3.
ResearchAdvisory Committee.(1994). Reflecting on practice:New directionsfor NCTM research
sessions. Journalfor Research in MathematicsEducation,25(4), 371-374.
Research Advisory Committee. (1997). Clarifying the contributionsof research within NCTM.
Journalfor Research in MathematicsEducation,28(4), 396-397.
Researchreport.(no date). Uma analise da constru9aodo conceito de fraqao[An analysis of the con-
structionof the concept of fractions].Unpublishedmanuscript.
Resnick, L. B. (1983). A developmentaltheory of numberunderstanding.In H. Ginsburg(Ed.), The
developmentof mathematicalthinking(pp. 109-151). New York:Academic Press.
Restivo, S. (1992). Mathematicsin society and history: Sociological inquiries.Dordrecht:Kluwer.
Richards,J. (1991). Mathematicaldiscussions. In E. von Glasersfeld(Ed.), Radical constructivismin
mathematicseducation (pp. 13-51). Dordrecht:Kluwer.
Richards,J. (1996). Negotiating the negotiationof meaning:Commentson Voigt (1992) and Saxe
and Bernudex(1992). In L. P. Steffe, P. Nesher, P. Cobb, G. Goldin, & B. Greer(Eds.), Theories
of mathematicallearning (pp. 69-75). Mahwah,NJ: LawrenceErlbaumAssociates.
Richards,T. J., & Richards,L. (1990). Manualfor MainframeNUD*IST: A softwaresystemfor qual-
itative data analysis on time-sharingcomputers.Melbourne:Replee.
Richardson,V. (1994). Conductingresearchon practice.EducationalResearcher,23(5), 5-10.
Rorty, R. (1979). Philosophyand the mirrorof nature. Princeton,NJ: PrincetonUniversity Press.
Roschelle, J. (1992). C-Video. [Softwarepackage]. San Francisco:Knowledge Revolution.
Rotman,B. (1993). Ad infinitumthe ghost in Turing'smachine: TakingGod out of mathematicsand
putting the body back in. Stanford,CA: StanfordUniversityPress.
Rowland, S. (1984). The enquiring classroom: An approach to understandingchildren's learning.
New York:FalmerPress.
Russell, R., & Ginsburg,H. P. (1984). Cognitive analysis of children's mathematicaldifficulties.
Cognitionand Instruction,1, 271-244.
Ryle, G. (1949). The concept of mind. London:Hutchinson.
Salj6, R. (1979). Learningin the learner's perspective:Some common-sense conceptions. Reports
from the Departmentof Education,Goteborg University,No. 76.
Salj6, R. (1982). Learningand understanding:A study of differencesin constructingmeaningfrom
a text. Goteborg:Acta UniversitatisGothoburgensis.
Sandberg,J. (1996). Are phenomenographicresultsreliable?In G. Dal'Alba & B. Hasselgren(Eds.),
Reflections on phenomenography: Toward a methodology? (pp. 129-140). G6teborg: Acta
UniversitatisGothoburgensis.
Saxe, G. B. (1991). Culture and cognitive development:Studies in mathematicalunderstanding.
Hillsdale, NJ: LawrenceErlbaumAssociates.
References 175
Saxe, G. B., & Bermudez,T. (1996). Emergentmathematicalenvironmentsin children's games. In

L. P. Steffe, P. Nesher, P. Cobb, G. Goldin, & B. Greer(Eds.), Theoriesof mathematicallearning
(pp. 51-68). Mahwah,NJ: LawrenceErlbaumAssociates.
Schoenfeld, A. (1994). A discourse of methods. Journalfor Research in MathematicsEducation,
25(6), 697-710.
Schubert,W. H. (1986). Curriculum:Perspective,paradigm,and possibility. New York:Macmillan.
Schiitz,A. (1964). Collectedpapers II: Studiesin social theory,A. Broderson(Ed.).The Hague:Nijoff.
Schiitz, A. (1970). Conceptand theoryformationin the social sciences. In D. Emmet& A. Maclntyre
(Eds.), Sociological theory and philosophical analysis (pp. 1-19). London:Macmillan.
Schiitz, A. (1972). Thephenomenologyof the social world. London:Heinemann.
Sega, D. (1997). Reading and writing about our lives: Creatinga collaborativecurriculumin a class
of high school misfits. TeacherResearch, 4(2), 101-111.
Sfard, A. (1991). On the dual nature of mathematicalconceptions: Reflections on processes and
objects as differentsides of the same coin. EducationalStudies in Mathematics,26(1), 1-36.
Short, K., & Burke, C. (1996). Examiningour beliefs and practicesthroughinquiry.LanguageArts,
73, 97-103.
Short, K., Harste,J., & Burke, C. (1996). Creating classroomsfor authors and inquirers(2nd ed.).
Portsmouth,NH: Heinemann.
Shotter,J. (1993). Conversationalrealities: Constructinglife throughlanguage. London:Sage.
Shulman,L. S. (1981). Disciplines of inquiry in education:An overview. EducationalResearcher,
10(6), 5-12, 23.
Siegel, M., & Borasi, R. (1994). Demystifying mathematicseducationthroughinquiry.In P. Ernest
(Ed.), Constructingmathematical knowledge: Epistemology and mathematical education (pp.
201-214). New York: FalmerPress.
Silver, E. A., & Kilpatrick,J. (1994). E PluribusUnum:Challengesof diversityin the futureof math-
ematics educationresearch.Journalfor Research in MathematicsEducation,25(6), 734-754.
Simons, H. (Ed.). (1989). Rhetoric in the humansciences. London:Sage.
Sinclair, J. McH., & Coulthard,M. (1975). Towards an analysis of discourse. London: Oxford
University Press.
Skemp, R. R. (1976). Relational understandingand instrumental understanding.Mathematics
Teaching, 77, 20-26.
Skolverket. (1995). Diagnostiskt material i matematik,skoldr 2. (The Swedish School Council.
Diagnostic materialin mathematics,year 2). Stockholm:Skolverket.
Skovsmose, 0. (1985). Mathematicaleducation versus critical education. Educational Studies in
Mathematics,16, 337-354.
Skovsmose, 0. (1994). Towardsa philosophyof critical mathematicseducation.Dordrecht:Kluwer.
Sowden, S., & Keeves, J. P. (1988). Analysis of evidence in humanisticstudies.In J. P. Keeves (Ed.),
Educationalresearch, methodologyand measurement:An internationalhandbook(pp. 513-526).
Oxford:PergamonPress.
Steen, L. A. (1990). On the shoulders of giants: New approaches to numeracy.Washington,DC:
National Academy Press.
Steffe, L. P. (199 la). The constructivistteachingexperiment:Illustrationsand implications.In E. von
Glasersfeld (Ed.), Radical constructivismin mathematicseducation (pp. 177-194). Dordrecht:
Kluwer.
Steffe, L. P. (Ed.). (1991b). Epistemologicalfoundations of mathematicalexperience. New York:
Springer-Verlag.
Steffe, L. P. (1996). Preface. In L. P. Steffe, P. Nesher, P. Cobb, G. Goldin, & B. Greer (Eds.),
Theoriesof mathematicallearning (pp. ix-xi). Mahwah,NJ: LawrenceErlbaumAssociates.
Steffe, L. P., & Gale, J. (Eds.). (1995). Constructivismin education. Hillsdale, NJ: Lawrence
ErlbaumAssociates.
Steffe, L. P., & Tzur, R. (1994). Interaction and children's mathematics. In P. Ernest (Ed.),
Constructing mathematical knowledge: Epistemology and mathematics education (pp. 8-32).
London:FalmerPress.
176 References
Steffe, L. P., von Glasersfeld,E., Richards,J., & Cobb, P. (1983). Children'scounting types. New
York:PraegerPublishers.
Steffe, L. P., & Wiegel, H. (1996). On the natureof a model of mathematicallearning.In L. P. Steffe,
P. Nesher, P. Cobb, G. Goldin, & B. Greer (Eds.), Theories of mathematical learning (pp.
477-498). Mahwah,NJ: LawrenceErlbaumAssociates.
Stenhouse, L. (1975). An introduction to curriculum research and development. London:
Heinemann.
Strauss,A. (1987). Qualitativeanalysisfor social scientists.Cambridge:CambridgeUniversityPress.
Strieb,L. (1993). Visitingandrevisitingthe trees.In M. Cochran-Smith& S. L. Lytle (Eds.),Inside/out-
side: Teacherresearchand knowledge(pp. 121-130). New York:TeachersCollege Press.
Stubbs, M. (1981). Scratchingthe surface:Linguistic data in educationalresearch.In C. Adelman
(Ed.), Uttering,muttering,collecting, using and reportingtalkfor social and educationalresearch
(pp. 114-133). London:GrantMcIntyre.
Svensson, L. (1989). Fenomenografi och kontextuell analys. Phenomenographyand contextual
analysis). In R. Sali (Ed.), Som vi uppfattardet. Elva bidrag om inlmrningoch omvarldsuppfat-
tning (As we see it. Eleven papers on learning and understandingthe world around us) (pp.
33-52). Lund:Studentlitteratur
Svensson, O., Hedenborg, M. L., & Lingman,L. (1976). On children'sheuristicsfor solving simple
additions.ScandinavianJournal of EducationalResearch, 20, 161-173.
Tall, D. (1991). The psychology of advanced mathematicalthinking. In D. Tall (Ed.), Advanced
mathematicalthinking(pp. 3-21). Dordrecht:Kluwer.
Teppo, A. R., & Esty, W. W. (1994). Problemsolving using arithmeticand algebraicthinking.In D.
Kirshner(Ed.),Proceedingsof the SixteenthAnnualMeetingof the NorthAmericanChapterof the
InternationalGroupfor the Psychology of MathematicsEducation(pp. 24-30). BatonRouge, LA:
LouisianaState University.
Teppo, A. R., & Esty, W. W. (1995). Mathematicalcontexts and the perceptionof meaningin alge-
braic symbols. In D. T. Owens, M. K. Reed, & & G. M. Millsaps (Eds.), Proceedings of the
SeventeenthAnnual Meeting of the North American Chapterof the InternationalGroupfor the
Psychology of MathematicsEducation(pp. 147-151). Columbus,OH: ERIC Cleaminghousefor
Science, Mathematics,and EnvironmentalEducation.
Tiles, M. (1991). Mathematicsand the image of reason. London:Routledge.
Todarov, T. (1984). Mikhail Bakhtin:The dialogical principle. (W. Godzich, Trans.) Minneapolis:
University of MinnesotaPress.
Toulmin, S. (1972). Humanunderstanding,I. Oxford:ClarendonPress.
Treffers, A. (1991). Didactical backgroundof a mathematicsprogramfor primaryeducation.In L.
Streefland(Ed.), Mathematicseducationin primaryschool: On the occasion of the opening of the
FreudenthalInstitute.Utrecht:Centerfor Science and MathematicsEducation.
Tymoczko, T. (Ed.). (1986). New directions in the philosophy of mathematics.Boston: Birkhauser.
van Oers, B. (1996). Learningmathematicsas a meaningfulactivity. In L. P. Steffe, P. Nesher, P.
Cobb, G. Goldin, & B. Greer (Eds.), Theories of mathematicallearning (pp. 91-113). Mahwah,
NJ: LawrenceErlbaumAssociates.
Vico, G. B. (1858). De antiquissima Italorum sapientia. (Pomodoro, Trans.). Naples: Stamperia
de'Classici Latini. (Originalwork published 1719.)
Vico, G. B. (1961). The new science. (T. G. Bergin & M. H. Fisc, Trans.).GardenCity, NY: Anchor
Books. (Originalwork published 1744.)
von Glasersfeld,E. (1987). Learningas a constructiveactivity. In C. Janvier(Ed), Problems of rep-
resentation in the teaching and learning of mathematics (pp. 3-18). Hillsdale, NJ: Lawrence
ErlbaumAssociates.
von Glasersfeld,E. (1989). Constructivismin education.In T. Husen (Ed.), Internationalencyclope-
dia of education (Supplementaryvol., pp. 162-163). Oxford:Pergamon.
von Glaserfeld,E. (1991). Radical constructivismin mathematicseducation. Dordrecht,Kluwer.
von Glasersfeld,E. (1995). Radical constructivism:A way of knowingand learning.London:Falmer
Press.
References 177
Vygotsky, L. S. (1978). Mind in society. The developmentof the higher psychological processes.
Cambridge:HarvardUniversity Press.
Walkerdine,V. (1988). The masteryof reason. London:Routledge.
Walkerdine,V., & The Girls and MathematicsUnit. (1989). Counting girls out. London: Virago
Press.
Wang, H. (1974). From mathematicsto philosophy. London:Routledge.
Wells, G. (1994). Changingschoolsfrom within: Creatingcommunitiesof inquiry.Toronto:Ontario
Institutefor the Study of Education(OISE).
Werner, H. (1973). Comparativepsychology of mental development. New York: International
Universities Press. (Originalwork published 1940.)
Wilder, R. L. (1981). Mathematicsas a cultural system.Oxford:PergamonPress.
Wittgenstein,L. (1953). Philosophical investigations( G. E. M. Anscombe, Trans.). Oxford: Basil
Blackwell.
Wittgenstein,L. (1978). Remarkson thefoundationsof mathematics(rev. ed.). Cambridge,MA: MIT
Press.
Wolcott, H. (1975). Criteria for an ethnographic approach to research in schools. Human
Organisation,34(2), 111-127.
Wolcott, H. (1982). Mirrors,models and monitors:Educatoradaptationsof the ethnographicinno-
vation. In G. Spindler (Ed.), Doing the ethnographyof schooling: Educationalanthropologyin
action (pp. 68-95). New York: Holt Reinhart& Winston.
Woolgar, S. (1988). Science: The very idea. London:Ellis Horwood/Tavistock.
Zang, R. (1994). Inferringinternalstrategicproblem representationand its development:A two-year
case study with four elementary school children. Unpublished doctoral dissertation, Rutgers
University GraduateSchool of Education,(UniversityMicrofilmsNo. 95-14127).
Zang, R. (1995). Inferringinternalstrategicproblemrepresentationand its development:A two-year
case study with Marcia.In D. T. Owens, M. K. Reed, & G. M. Millsaps (Eds.), Proceedings of the
SeventeenthAnnual Meeting of the North American Chapterof the InternationalGroupfor the
Psychology of Mathematics Education (Vol. 1, pp. 199-203). Columbus, OH: ERIC
Clearinghousefor Science, Mathematics,and EnvironmentalEducation.
Zaslavsky, C. (1973). Africa counts. Boston: Prindle,Weber & Schmidt.

Anne R. Teppo. Qualitative Research Methods in Mathematics Education PDF

Uploaded by

Document Information

Original Title

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

Anne R. Teppo. Qualitative Research Methods in Mathematics Education PDF

Uploaded by

Copyright:

Available Formats

ISSN 0883-9530

National Council of Teachers of Mathematics

The JRME Monograph Series is published by the Series Editor

Manuscriptsshouldbe sent to Era Yackel; Department

National Council of Teachers of Mathematics

All rights reserved

Library of Congress Cataloging-in-Publication Data:

Qualitativeresearchmethodsin mathematicseducation/ edited by Anne

The publicationsof the NationalCouncil of Teachersof Mathematicspresenta

Printed in the United States of America

David J. Clarke DagmarNeuman

GeraldA. Goldin Anne R. Teppo

The editorwishes to thankthose who providedsupportand encouragementfor

The chaptersin this monographdescribequalitativeresearchmethods used to

Diverse Ways of Knowing

The chapters in this monograph describe qualitative methods used in mathe-

Qualitative research has been described as a field of inquiry (Denzin and

Qualitative Research in Education

Anthropologistsand sociologists have employed qualitativeresearchmethods

that each paradigmoffers a differentway to focus researchon the complexities

Diversity of Perspectives WithinMathematicsEducationResearch

A Need for Dialogue

AN OVERVIEWOF THE FIELD OF MATHEMATICSEDUCATION

Mathematicseducators face a similar problem in considering their field of

The Nature of Mathematics

Mathematicshas been describedas the science of abstractpatternsand char-

In addition, the use of some form of external representationmakes it possible

The Nature of MathematicsEducation

views about the nature of mathematics,one's underlyingepistemological per-

which the learnerreinvents "mathematisingratherthan mathematics;abstract-

Mathematics Education Research

ization of the natureof knowledge from a single and externalreality to a set of

voice," promotesreconceptualizationsof one's own mathematicalunderstanding

DIVERSE WAYS OF KNOWING

Chapters2 and 3 orient the readertowarda critical examinationof the mono-

development of meaning within a social context, the former within a small

she assigned to observedclassroombehaviorand her reasons for attributingsig-

The reportedresearchalso variesas to the authors'perspectiveson the sourceof

Toward a Definition for Research

Before the readerplunges into the remainingchaptersof this book, it is appo-

More fundamentally,what do we consider to be researchin our discipline? We

* On the basis of existing findings relatingto errorsand difficulties that pupils

Some of the previous scenariosyou may have been able to categorizeinstant-

for the whole community, but mathematicseducators need to participatein a

The Epistemological Basis of

theoreticalperspectiveon knowledge and research,that allows specific methods,

From Modernismto Postmodernism

philosophy is beginning to wane. Instead, contemporaryphilosophy is now

feminist epistemology (Harding,1991): Parallel"bottom-up"developmentsin

The Philosophy of Mathematics

Accompanyingthe emergence of these new of conceptions of knowledge and

ethnomathematics,mathematics-education studies,and feminist and multicultural

Postmodernism and the Qualitative Research Paradigm

Postmodernismis a portmanteauterm used to denote a varietyof perspectives

employed in the researchprocess resonateswith developmentsin philosophy,the

in some meaningfulactivity or context. Erklaren,in contrast,is the method of

One of the centralcomponentsof the qualitativeresearchparadigmis the con-

her model of the subject'sunderstandingconcerningeven the narrowestof math-

Von Glasersfeld's (1989, 1995) radical constructivistformulation is that our

Contributions From Social Perspectives

A number of other traditions and developments have helped introduce the

on the work of Foucault (1972) and applied to mathematics education by

These emphasescombine to indicatethat constructivistresearchin mathemat-