Paul Christian Dawkins, Amy J. Hackenberg, Anderson Norton - Piaget's Genetic Epistemology For Mathematics Education Research-Springer (2024)

Research in Mathematics Education
Series Editors: Jinfa Cai · James A. Middleton
Paul Christian Dawkins

Amy J. Hackenberg
Anderson Norton Editors
Piaget’s Genetic
Epistemology
for Mathematics
Education
Research
Series Editors
Jinfa Cai, Newark, DE, USA
James A. Middleton, Tempe, AZ, USA
This series is designed to produce thematic volumes, allowing researchers to access
numerous studies on a theme in a single, peer-reviewed source. Our intent for this
series is to publish the latest research in the field in a timely fashion. This design is
particularly geared toward highlighting the work of promising graduate students
and junior faculty working in conjunction with senior scholars. The audience for
this monograph series consists of those in the intersection between researchers and
mathematics education leaders—people who need the highest quality research,
methodological rigor, and potentially transformative implications ready at hand to
help them make decisions regarding the improvement of teaching, learning, policy,
and practice. With this vision, our mission of this book series is: (1) To support the
sharing of critical research findings among members of the mathematics education
community; (2) To support graduate students and junior faculty and induct them
into the research community by pairing them with senior faculty in the production
of the highest quality peer-reviewed research papers; and (3) To support the
usefulness and widespread adoption of research-based innovation.
Amy J. Hackenberg • Anderson Norton
Editors
Piaget’s Genetic Epistemology

for Mathematics Education
Research
Editors
Paul Christian Dawkins Amy J. Hackenberg
Department of Mathematics Department of Curriculum and Instruction
Texas State University Indiana University
San Marcos, TX, USA Bloomington, IN, USA
Anderson Norton
Department of Mathematics
Virginia Tech
Blacksburg, VA, USA
ISSN 2570-4729 ISSN 2570-4737 (electronic)

ISBN 978-3-031-47385-2 ISBN 978-3-031-47386-9 (eBook)
https://doi.org/10.1007/978-3-031-47386-9
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Switzerland AG 2024
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the
editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Paper in this product is recyclable.

To our teachers, most prominently the
children who teach us their mathematics as
well as Les Steffe and Pat Thompson who
paved the way for this flourishing work.
Contents
Part I Introduction to Piaget’s Genetic Epistemology and the

Tradition of Use Featured in This Book
1
Introduction to Piaget’s Genetic Epistemology�� 3
Paul Christian Dawkins, Amy J. Hackenberg, and Andy Norton
2 Historical Reflection on Adapting Piaget’s Work for Ongoing
An
Mathematics Education Research�� 11
Leslie P. Steffe
Part II Key Constructs from Genetic Epistemology Being Used

in Ongoing Mathematics Education Research
3 Schemes and Scheme Theory: Core Explanatory Constructs
for Studying Mathematical Learning�� 47
Erik S. Tillema and Andrew M. Gatza
4
Operationalizing Figurative and Operative Framings of Thought �� 89
Kevin C. Moore, Irma E. Stevens, Halil I. Tasova, and Biyao Liang
5 Figurative and Operative Imagery: Essential Aspects of Reflection
in the Development of Schemes and Meanings �� 129
Patrick W. Thompson, Cameron Byerley, and Alan O’Bryan
6 Empirical and Reflective Abstraction�� 169
Amy Ellis, Teo Paoletti, and Elise Lockwood
7
Groups and Group-Like Structures �� 209
Anderson Norton
8 Reflected Abstraction�� 239
Michael A. Tallman and Alan E. O’Bryan
vii
viii Contents
9 The Construct of Decentering in Research on

Mathematics Learning and Teaching�� 289
Marilyn P. Carlson, Sinem Bas-Ader, Alan E. O’Bryan,
and Abby Rocha
10
Logic in Genetic Epistemology �� 339
11 Students’ Units Coordinations�� 371
Amy J. Hackenberg and Serife Sevinc
12
Modeling Quantitative and Covariational Reasoning�� 413
Steven Boyce
Part III Commentaries on Genetic Epistemology and Its Use

in Ongoing Research
13 Genetic Epistemology as a Complex and Unified Theory
of Knowing�� 447
Anderson Norton
14
Second-Order Models as Acts of Equity�� 475
Amy J. Hackenberg, Erik S. Tillema, and Andrew M. Gatza
15
Reflections on the Power of Genetic Epistemology by the Modern
Cognitive Psychologist �� 511
Percival Matthews and Alexandria Viegut
16 Skepticism and Constructivism�� 541
Part IV Using Constructs from Genetic Epistemology to Develop

Agendas of Research
17
Researching Special Education: Using and Expanding Upon Genetic
Epistemology Constructs�� 565
Jessica H. Hunt
18
Research in Subitizing to Examine Early Number Construction�� 573
Beth L. MacDonald
19
Researching Coordinate Systems Using Genetic Epistemology
Constructs�� 585
Hwa Young Lee
20 Researching Quantifications of Angularity Using
Genetic Epistemology Constructs�� 595
Hamilton L. Hardison
Contents ix
21 Using Constructivism to Develop an Agenda of Research

in Stochastics Education Research �� 605
Neil Hatfield
Index�� 613
About the Authors
Sinem Bas-Ader is an assistant professor in the Department of Mathematics and

Science Education at Istanbul Aydin University. She received her PhD from Middle
East Technical University. She is studying teacher noticing of students’ mathemati-
cal thinking and she is working with both pre-service and in-service mathematics
teachers in various professional development contexts. In her postdoctoral study as
a visiting scholar at Arizona State University, she joined an NSF-funded Pathways
Project team directed by Professor Marilyn P. Carlson. She developed an interest for
constructivism in mathematics teaching and particularly focused on Piaget’s con-
struct of decentering as a key competence for responsive teaching.
Steven Boyce is an associate professor of Mathematics Education at Portland

State University in Portland, Oregon. His passion is modeling students’ mathemati-
cal thinking, in particular connections between adolescents’ and adults’ coordina-
tions of numerical and quantitative units across contexts. Dr. Boyce is the author of
dozens of articles and conference proceedings pertaining to units coordination, frac-
tions, pre-calculus, calculus, and teacher education. He is currently serving as guest
editor of a special issue on Units Construction and Coordination across the
Curriculum, PK-20, to appear in the journal Investigations in Mathematics Learning
in Fall 2024.
Cameron Byerley is an assistant professor in the Department of Mathematics,

Statistics, and Social Studies Education at the University of Georgia. She is inter-
ested in using models of mathematical thinking to improve communication of criti-
cal real-world quantitative information to the public.
Marilyn P. Carlson is a professor of Mathematics Education studying the teach-

ing and learning of ideas of precalculus and beginning calculus. Her current research
is studying mechanisms for supporting precalculus level instructors in transitioning
their instruction to be more coherent, engaging, and meaningful for students. She is
exploring the interaction between a teacher’s mathematical meanings for teaching
an idea, their decentering actions, and their effectiveness in modeling students’
xi
xii About the Authors
thinking. She and her colleagues are working to design interventions to support
precalculus instructors in modeling students’ thinking and using these models when
planning their lessons, interacting with students, and assessing student learning.
Paul Christian Dawkins is a professor of Mathematics Education in the

Department of Mathematics at Texas State University. He studies the teaching and
learning of proof-based mathematics at the undergraduate level. He focuses particu-
larly on real analysis, geometry, and logic. His teaching experiments focus on mod-
eling student thinking about mathematical proofs and developing teaching sequences
and learning trajectories by which students can learn the epistemology of mathe-
matical proving. He received the Selden Award for research in undergraduate math-
ematics education from the Mathematics Association of America.
Amy Ellis is a professor of Mathematics Education at the University of Georgia.

She received her Ph.D. in Mathematics Education from the University of California,
San Diego and San Diego State University. Amy’s research agenda addresses stu-
dents’ algebraic reasoning, teachers’ pedagogical practices for fostering meaningful
student engagement, and playful math, which is the investigation of students’ play-
ful engagement with mathematical ideas. Amy has received 15 grants from national
and state organizations, including the National Science Foundation and the Institute
of Education Sciences, and has published three books for the National Council of
Teachers of Mathematics Essential Understanding Series.
Andrew M. Gatza is a former middle school mathematics teacher and completed

doctoral programs in Mathematics Education (Indiana University, Bloomington)
and Urban Education Studies (Indiana University, Indianapolis). He is currently an
assistant professor of Mathematics Education in the Department of Mathematical
Sciences at Ball State University. His work focuses on bringing together equity and
justice issues with rich mathematical problem sequences to investigate the kinds of
reasoning in which students engage. Additionally, he is focused on cultivating
mutually supportive school partnerships and developing socially conscious mathe-
matics educators who foster positive mathematics identities and understand and
support students’ ways of reasoning.
Amy J. Hackenberg taught middle and high school students for 9 years in Los
Angeles and the Chicago area. She earned her MAT degree in Mathematics
Education from the University of Chicago and her PhD from the University of
Georgia. Amy is a professor of Mathematics Education at Indiana University,
Bloomington. She studies how middle and high school students construct fractions,
rational number knowledge, and algebraic reasoning; how to differentiate instruc-
tion for students with diverse ways of thinking; and how to support teachers to sup-
port their students to feel a stronger sense of belonging in math class.
About the Authors xiii
Hamilton L. Hardison is an assistant professor of Mathematics at Texas State

University. He earned his PhD in Mathematics Education at the University of
Georgia, and his primary research interests lie in investigating students’ mathemati-
cal thinking. His current research focuses on modeling how students quantify angu-
larity, how these quantifications change over time, and how they vary across
contexts. His additional interests include studying mathematical classroom dis-
course and discussing radical constructivism.
Neil Hatfield is an assistant research professor in the Department of Statistics at

the Pennsylvania State University (USA). His research foci include cognition
around the concept of distribution and related stochastic concepts, the teaching of
stochastics (statistics, probability, and data science), and on diversity, equity, and
inclusion issues in STEM.
Jessica H. Hunt began her career in education as a middle school mathematics

teacher in a technology demonstration school in Florida. From that work, she grew
to love teaching students at risk for mathematics difficulties or disabilities. Hunt
argues that mathematics instruction for these students should (a) uncover strengths,
(b) give access to their mathematical reasoning, and (c) support the advance of that
reasoning.
Hwa Young Lee received her PhD in Mathematics Education from the University
of Georgia and is an associate professor in the Department of Mathematics at Texas
State University. Her main research interest is in investigating students’ mathemati-
cal thinking—specifically, students’ constructions of frames of reference, coordi-
nate systems, and graphs—and in learning how teachers can facilitate and support
their students’ mathematical thinking.
Biyao Liang obtained her PhD degree in Mathematics Education at the University
of Georgia. She is currently an assistant professor at The University of Hong Kong.
Her research program covers the areas of mathematical cognition, social interac-
tions, and teacher education. She is interested in students’ and teachers’ mathemati-
cal thinking in a variety of contexts. Her current projects investigate the affordances
of computer programming in engendering and supporting uncanonical and produc-
tive ways of reasoning with mathematics. She is also interested in teachers’ mathe-
matical learning through social interactions with students and designing educational
opportunities, tools, and materials that can support teachers’ learning about students.
Elise Lockwood is a professor in the Department of Mathematics at Oregon State

University. She received her PhD in Mathematics Education from Portland State
University and was a postdoctoral scholar at the University of Wisconsin–Madison.
Her primary research interests focus on undergraduate students’ reasoning about
combinatorics, and she is passionate about improving the teaching and learning of
discrete mathematics. Her work has been funded by the National Science Foundation
and Google, and her current work explores ways in which computation and
xiv About the Authors
programming can be leveraged to support students’ combinatorial thinking and

activity. Elise currently serves as co-editor in chief of the International Journal of
Research in Undergraduate Mathematics Education. In 2019, she was a Fulbright
Scholar to Norway, working with researchers at the University of Oslo.
Beth L. MacDonald is an associate professor of Early Childhood Mathematics

Education in the School of Teaching and Learning at Illinois State University. Beth
taught elementary school for 15 years and served as a PreK-5 Instructional Specialist
for 2 years in Virginia. While teaching, Beth completed her Master’s degree and her
Doctor of Philosophy degree, each with a Curriculum and Instruction concentration
and focus on Mathematics Education, both from Virginia Tech. Her research broadly
focuses on young students’ development of number and subitizing activity. Beth
collaborates with colleagues examining teachers’ specialized content knowledge
development and marginalized students’ number understanding development.
Percival Matthews is an associate professor of Educational Psychology at the

University of Wisconsin–Madison. He earned is PhD in Psychology from Vanderbilt
University and completed his postdoctoral research at the University of Notre
Dame. The bulk of his research is organized around two primary goals: (1) under-
standing the basic underpinnings of human mathematical cognition and (2) finding
ways to leverage this understanding to create effective pedagogical techniques that
can be used to impact the life chances of everyday students.
Kevin C. Moore Born and raised in Ohio, I attended The University of Akron
from 2001 to 2006. I worked as a Graduate Assistant in the Department of
Mathematics and grew curious about my students’ mathematical thinking when
teaching as part of the assistantship duties. This curiosity landed me at Arizona
State University under the guidance of Professor Marilyn P. Carlson. I immediately
grew interested in the constructivist movement in mathematics education, and spe-
cifically the ability to take a scientific-inquiry approach to modeling students’ math-
ematical thinking. Since this initial interest, I have rooted myself with other
researchers who participate in this progressive research program in the hopes of
better understanding students’ mathematical thinking, improving the teaching and
learning of mathematics, and opposing outcome-based forces in education.
Anderson Norton is a professor of Mathematics Education in the Department of

Mathematics at Virginia Tech. His research focuses on building psychological mod-
els of students’ mathematical development, particularly in the domain of fractions
knowledge, and the epistemology of mathematics. He has served as chair for the
editorial panel of the Journal for Research in Mathematics Education, chair of the
steering committee for the North American Chapter of the International Group for
the Psychology of Mathematics Education, and lead editor for the Springer book,
Constructing Number. In 2013, in recognition of his outreach efforts, he received
the Early Career Award from the Association of Mathematics Teacher Educators.
About the Authors xv
Alan O’Bryan has worked for almost 20 years as a mathematics education

researcher, curriculum developer, professional development leader, and teacher
trainer. His research focuses on examining student learning of mathematical ideas
and teacher change through the lens of cognitive psychology. He is the co-author of
six secondary and post-secondary mathematics textbooks and has delivered nearly
100 professional development workshops across the United States for teachers of all
levels from elementary to post-secondary instructors. His presentations and work-
shops are characterized by the way they make mathematics education research find-
ings accessible and practical for classroom teachers.
Teo Paoletti is an associate professor specializing in Mathematics Education in

the School of Education at the University of Delaware. The primary goal of his
research agenda is to explore student understanding of mathematical ideas at vari-
ous levels (e.g., middle school through post-secondary). He leverages design-based
methods to explore ways in which students can leverage reasoning about relation-
ships between quantities to construct and reason about critical mathematics con-
cepts. His recent work entails designing task sequences that leverage various
dynamic mathematical software (e.g., GeoGebra, Desmos, GSP) to support middle-
school students developing meanings for various function classes, graphs, and
inequalities.
Abby Rocha completed her PhD in Mathematics Education at Arizona State

University, where she was recognized with multiple awards for her research and
teaching excellence. Abby’s research investigates relationships between teachers’
mathematical meanings for teaching and their instructional practices, including
their actions to decenter. Currently, Abby is a postdoctoral fellow in the STEM
Learning Center at the University of Arizona. During her fellowship, Abby’s
research will focus on supporting STEM faculty in engaging and integrating cultur-
ally responsive pedagogical and curricular practices through professional develop-
ment for faculty teaching gateway lecture and lab courses.
Serife Sevinc (Şerife Sevinç; in native language) is an associate professor at the

Department of Mathematics and Science Education at Middle East Technical
University, Turkiye. She received her doctoral degree in Curriculum and Instruction
with Mathematics Education specialization at Indiana University and a minor PhD
degree in Inquiry Methodology. Her main research interests involve investigating
students’ and preservice teachers’ mathematical thinking in problem-solving and
problem-posing processes, mathematical modeling process that enhances students’
and teachers’ conceptual understanding in mathematics, and the nature of pre-ser-
vice teachers’ knowledge in designing realistic mathematics problems.
Leslie P. Steffe is a research professor emeritus of Mathematics Education,

University of Georgia. He received his Ph.D. from the University of Wisconsin in
1966 and joined the Department of Mathematics Education at the University of
Georgia in 1967 from which he retired in 2017. He developed the research program
xvi About the Authors
Interdisciplinary Research on Number (IRON) with the epistemologist, Ernst von

Glasersfeld, and the philosopher, John Richards, starting circa 1977 and lasting
until circa 2000. From that point on, he formed the research program Ontogenetic
Analysis of Algebraic Knowing (OAK) which lasted until he retired in 2017.
Irma E. Stevens is an assistant professor in the Department of Mathematics and

Applied Mathematical Sciences at the University of Rhode Island. She earned her
PhD in Mathematics Education and MA in Mathematics at the University of Georgia
and her BS in Mathematics at the University of North Carolina at Charlotte. Her
research interests involve teacher decision making and student learning at the sec-
ondary and postsecondary level. Her current projects include understanding under-
graduate students’ reasoning with dynamic quantities and formulas. Her work uses
technology that has resulted in the construction of materials used for research, pro-
fessional development, and classroom instruction.
Michael A. Tallman received his BSc and MA in Mathematics from the

University of Northern Colorado and his PhD in Mathematics Education from
Arizona State University. Dr. Tallman has taught mathematics at the secondary and
post-secondary levels. His primary research is in the area of mathematical knowl-
edge for teaching secondary and post-secondary mathematics. Dr. Tallman’s work
informs the design of teacher preparation programs and professional development
initiatives through an investigation of the factors that affect the nature and quality of
the mathematical knowledge teachers leverage during instruction. In particular, his
research examines how various factors like curricula, emotional regulation, identity,
and teachers’ construction and appraisal of instructional constraints mediate the
enactment of their subject matter knowledge.
Halil I. Tasova is an assistant professor of Mathematics Education at California

State University, San Bernardino. He earned his PhD in Mathematics Education
from the University of Georgia and holds an MS and BS in Secondary Mathematics
Education from Marmara University in Türkiye. With prior experience as a high
school math teacher, Dr. Tasova brings valuable classroom insights to his current
role. His research focuses on students’ mathematical thinking and learning, particu-
larly their reasoning about quantities. He explores the construction and interpreta-
tion of graphs from the perspective of quantitative and covariational reasoning.
Patrick W. Thompson is professor emeritus of Mathematics Education in the

School of Mathematical and Statistical Sciences at Arizona State University. He
spent his career researching the learning and teaching of mathematics at elementary
school, high school, and university levels through a lens of quantitative reasoning.
He has contributed to the understandings of learning and teaching of arithmetic,
algebra, statistics, and calculus, and to areas of research methodology in mathemat-
ics education.
About the Authors xvii
Erik S. Tillema is a former middle and high school mathematics teacher and cur-
rent associate professor of Mathematics Education at Indiana University. His work
focuses on making second-order models of middle grades and secondary students’
reasoning with the aim of having such models be the basis for curricular design and
instruction. He is particularly interested in second-order models of students’ combi-
natorial and quantitative reasoning and how generalizations of this reasoning can
support students’ algebraic reasoning. His recent work involves collaborating with
secondary teachers in the endeavor of using combinatorics problems to support their
students to generalize algebraic structure.
Alexandria Viegut is an assistant professor of Psychology at University of

Wisconsin-Eau Claire. She earned a PhD in Educational Psychology from UW–
Madison. Alex’s current research investigates students’ understanding of fractions,
the connections between fractions knowledge and algebra knowledge, and the role
of visual representations in mathematical cognition.
Part I
Introduction to Piaget’s Genetic
Epistemology and the Tradition of Use
Featured in This Book
Chapter 1
Introduction to Piaget’s Genetic
Epistemology
Paul Christian Dawkins, Amy J. Hackenberg, and Andy Norton
Introduction
Piaget is known for his work in developmental psychology, but he began his career
as a biologist whose primary interests evolved into epistemology; that is, theories of
knowledge and knowing. While studying snails, he was introduced to Bergson’s
(1998) idea of creative evolution, in response to which he later said,
The problem of knowing (properly called the epistemological problem) suddenly appeared
to me in an entirely new perspective and as an absorbing topic of study. It made me decide
to consecrate my life to the biological explanation of knowledge. (Vidal, 1994, p. 52)
Piaget began to study children’s psychological development as a means of investi-

gating the biological origins of logic and mathematics. In this pursuit, he followed
the biogenetic law that “ontology recapitulates phylogeny”: the development of the
individual follows a similar trajectory as the development of humankind.
P. C. Dawkins (*)
Department of Mathematics, Texas State University, San Marcos, TX, USA
e-mail: pcd27@txstate.edu
A. J. Hackenberg
Department of Curriculum & Instruction, Indiana University, Bloomington, IN, USA
e-mail: ahackenb@indiana.edu
A. Norton
Department of Mathematics, Virginia Tech, Blacksburg, VA, USA
e-mail: norton3@vt.edu
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 3

P. C. Dawkins et al. (eds.), Piaget’s Genetic Epistemology for Mathematics
Education Research, Research in Mathematics Education,
https://doi.org/10.1007/978-3-031-47386-9_1
4 P. C. Dawkins et al.
We can sum up the constructivist epistemology1 with the mantra, “All knowledge
is constructed” (Glasersfeld, 1995, p. 160). Here, we focus on Piaget’s genetic epis-
temology, which implies more than its constructivist tenets. We frame Piaget’s
genetic epistemology in terms of his lifelong pursuit to understand the power and
origins of logico-mathematical operations. It directs us back to the biological ori-
gins of our logico-mathematical operations and to the structures that organize them
so that they, in turn, might organize the world.
To say “all knowledge is constructed” is to challenge the notion that knowledge
comes in from outside a person in some simple or reliable way. Certainly, humans
build knowledge based on their experiences in their environments, but those experi-
ences are unique to each individual. In the same way, the physical form that an
organism takes is largely driven by its internal structure (e.g., DNA), humans’ con-
struction of knowledge operates through internal structures in a complex interplay
with (their sensorimotor experience of) their environment. This is primarily an epis-
temological claim about the construction of knowledge, not an ontological claim
about the extent to which the knowledge so constructed is, in some sense, an “accu-
rate” depiction of the world. A key feature of this epistemology is that it can explain
learning without a strong notion of correspondence to reality.
Modern biology shows that different kinds of animals experience the world very
differently from humans. By framing human psychology as fundamentally biologi-
cal, we gain at least two important perspectives. First, we learn to respect how much
our experience of the world is structured by our bodies and by our cognitive pro-
cesses. Second, we may consider how much those experiences have changed over
the course of our lifetimes. Children at different stages of development may experi-
ence different worlds and reason in markedly different ways from us as adults. We
may question how much of our current experience depends intrinsically on earlier
constructions that we cannot recall ever having done without.
In his genetic epistemology, Piaget adopted a Kantian perspective, but empirical
results from his research with children challenged core assumptions in Kant’s phi-
losophy. Kant’s (1781) Critique of Pure Reason blended empiricism with rational-
ism by accepting a few principal cognitive structures as innate. These innate
structures included space, time, and number, which enable us to organize our expe-
riences in the world. They also explain how we might take experiences as shared. If
we all construct reality within the same God-given space–time framework, we can
expect some commonalities among those realities. However, Piaget demonstrated
that children construct these foundational structures too, during their first few years
of life. In other words, Piaget’s research countermanded Kant’s assumptions. Few,
if any of us, can recall the years we spent playfully organizing our worlds, so we
take those constructions (e.g., space, time) for granted.
1
We have found Noddings (1990) explanation of constructivist epistemology, especially
illuminating.
1 Introduction to Piaget’s Genetic Epistemology 5
Consider Kant’s (1781) famous statement: “The concept of Euclidean space is by

no means of empirical origin, but is the inevitable necessity of thought.” From our
perspective now, this statement is an error. The mere possibility of non-Euclidean
geometries, such as the one Gauss invented during the nineteenth century, refutes
Kant’s claim. Another century later, Piaget and Inhelder (1967) demonstrated that
children construct space during their first years of life, on the basis of their own
sensorimotor activity in the world. Thereafter, the objects children experience have
a home to persist in even when they are out of sight (i.e., object permanence).
Piaget demonstrated that children construct number, too. Steffe has elaborated
on this construction through learning levels that he referred to as children’s number
sequences. At every stage, development depends upon the coordination of actions—
first sensorimotor, then internalized as mental actions, and finally organized within
structures that render them logico-mathematical operations. These structures, both
spatial and numerical, serve the role Kant envisioned for them, but they are the
result of years of labor. Once we have constructed them, it becomes difficult to
imagine a world without them.
What then allows us to build up concepts like space, time, and number if we do
not, as Kant claimed, begin with certain structures already in place? The heart of
Piaget’s answer to this is the organization of our own activity. Children act in their
experiential worlds, and their organization of those actions provides the basis for the
organization of their experiential worlds.
Piaget’s genetic epistemology emphasizes the unique status of logico-
mathematical operations within human knowledge. It also affords a different
account of mathematical objects themselves. If knowledge corresponds to reality in
a strong sense, then it raises questions about the nature and source of abstract con-
cepts such as number, line, function, and set. The philosophical stance known as
Platonism classically solves this problem by asserting the real existence of abstract
entities (an ontological claim). This allows us to somehow learn abstract concepts
under the assumption that they come in from the outside world (an epistemological
claim). Since constructivism provides an alternative account of how concepts form,
it provides a resolution of this epistemological issue that can remain ontologically
neutral. It thus provides an alternative to Platonism in explaining the power of logic
and mathematics. This power owes to the structures that we construct through the
coordination of our own mental actions rather than structures imposed upon us by
the worlds we ourselves organize through those very same structures.
In all, the chief apparent advantage of Platonism, which is to account for the objective
robustness of logico-mathematical entities and structures, is guaranteed in the same way by
the concept of the general and internal co-ordinations of actions and operations. That
hypothesis that ideal entities are external is thus unnecessary to guarantee the independence
of structures in view of the free will of individual subjects. (Beth & Piaget, 1966, p. 294)
Why Is Piaget’s Genetic Epistemology Useful?
For Glasersfeld (1995), a way of knowing is valuable to the extent it is useful.

Another way to say this is that people construct ways of knowing to serve purposes.
We apply this orientation repeatedly in this book to articulate why Piaget’s genetic
epistemology—and the research tradition in Mathematics Education that has been
built from it—is useful. One reason it is useful is that it acknowledges that people
build up knowledge to organize their experiential worlds and pursue goals within
those worlds, not to describe an observer-independent world. As Glasersfeld (1995)
pointed out, Piaget was not the first to take this position on knowledge, but he was
the first to take a developmental approach (p. 13). As we have introduced above,
Piaget viewed the construction of knowing in an individual (1) to be a process of
construction over a lifetime and (2) to reflect the construction of knowing in humans
as a species. The first point means that no person’s ways of knowing are ever com-
plete—they are always evolving. The second point means that understanding the
nature of knowing requires studying its ontogenesis—its development in humans
across their lives.
In our experience, as researchers and teachers in mathematics education, Piaget’s
views on knowing provide the basis for generating rich tools for describing and
accounting for students’ mathematics. Piaget’s views also enlarge what is consid-
ered mathematical—and, therefore, who is considered a mathematical thinker.
Piaget’s views on knowing imply that a great variety of ways of knowing and think-
ing can be admitted into mathematical knowledge, including children’s mathemat-
ics (Steffe & Olive, 2010). So, researchers can co-construct with participants,
including children, ways of thinking that can be understood as mathematical beyond
what has traditionally been viewed as mathematical. These ways of thinking are
models of participants’ mathematical knowledge that researchers can use to support
future interactions with other participants (see Chaps. 9 and 14). As a consequence,
students whose ways of thinking differ from what has traditionally been considered
standard mathematical ideas can be legitimated, and these students can be seen
more fully as mathematical thinkers (e.g., Hackenberg, 2013; Hackenberg & Sevinc,
2021; Norton & Boyce, 2015).
Piaget is known for developmental stages (e.g., concrete operations, formal oper-
ations) that have been critiqued as being rigid and nonrepresentative of all people.
We would like to address that directly using children’s number sequences (Steffe
et al., 1983; Steffe & Cobb, 1988) as an example. In this research, Steffe and col-
leagues studied how young children construct whole numbers by studying how they
count and how the nature of counting changes with successive constructions. They
found that children constructed approximately four number sequences, and these
occur in order because later number sequences involve more complex organizations
of units. Such descriptions can help teachers and researchers organize instructional
interactions with a range of elementary school students. Yet, the descriptions of
number sequences of children do not follow a lockstep set of stages at the same
ages—that is not what the developmental aspect of genetic epistemology means. So,
the developmental aspect of genetic epistemology means that children tend to con-
struct number sequences in a certain order, but there is great variation in how stu-
dents at a particular age conceive of number.
And yet, the usefulness of genetic epistemology does not stop with description.
The use and development of other Piagetian tools, such as accommodation and
reflective abstraction (discussed in Chaps. 3 and 6), provide the means for explain-
ing students’ mathematical thinking and learning.2
Together, descriptions of mathematical thinking with explanations of how learn-
ing proceeded—or did not proceed—are key components of making models of stu-
dents’ mathematical thinking and learning (Steffe & Thompson, 2000; Ulrich et al.,
2014). Consistent with Piaget’s overall insight that knowledge need not be explained
in terms of correspondence with reality, researchers using genetic epistemology
recognize research as their process of knowledge construction. As a result, we must
be careful to distinguish what we try to learn about student knowledge (our models
of their knowing) from student knowledge itself. We cannot know whether these
models, co-constructed with students, are what we would experience if we some-
how became these students or otherwise fully adopted their ways of knowing.
Students’ ways of knowing are not directly accessible to us. Instead, the models are
our ways of knowing that fit with our interactions with the students—they are what
Steffe refers to as second-order knowledge (2010), or the mathematics of students.3
Such models take extensive work for researchers to construct and refine (see Chap.
14). Robustly developed models can be regarded as legitimate mathematical ways
of knowing—and thus, what gets considered to be mathematics gets expanded.
For example, when working to build fractions knowledge, students who are try-
ing to draw 3/5 of a bar can learn to partition the bar into five equal parts, take out
one part, and repeat the part to make three parts (Fig. 1.1). In other words, they can
create 3/5 of a bar as 1/5 of the bar, another 1/5, and another 1/5. To observers, it
might look like the student thinks of 3/5 as 3 times 1/5. However, that may not be
the case. Third- through fifth-grade students taught Steffe and Olive (2010) that they
may not think of 3/5 in this way. Rather, they may rely on part–whole meanings for
the result, thinking of 3/5 as three parts out of five, despite the actions they took to
make the 3/5. Because this way of thinking was a regularity in how students oper-
ated in a longitudinal teaching experiment (Steffe & Olive, 2010), Steffe and Olive
formulated a scheme (see Chaps. 2 and 3 of this volume) to describe these students’
way of thinking about fractions, the partitive fraction scheme (Steffe & Olive,
2010). This way of thinking about fractions is challenging to understand for those
who conceive of fractions as multiples of unit fractions; it is hard to see that the
students’ meaning could be non-multiplicative when the actions look like what a
person would do who thinks of 3/5 as 3 times 1/5. Indeed, a person with multiplica-
tive meanings could engage in the same physical behavior for 3/5. However,
2
We will use the term “students” rather than “children,” since not all research has been with young
children.
3
First-order mathematical knowledge is the mathematical ways of knowing we have built to orga-
nize our own experiential worlds.
Fig. 1.1 3/5 as three of 1/5
differences quickly arise for students when fractions exceed the whole (e.g.,
Hackenberg, 2007; Norton & Wilkins, 2012; Steffe & Olive, 2010). Students who
have constructed a partitive fraction scheme find drawing, for example, 7/5 of a bar,
very mysterious. How can a person draw 7 parts out of 5?
This example shows both aspects of the usefulness of Piaget’s views: The parti-
tive fraction is an example of the use of scheme as a powerful tool for modeling
student thinking, and this scheme is an expansion of ways of thinking with fractions
that can be considered legitimately mathematical.
Second-order models of particular students can be very satisfying to make: When
they are developed, they represent to the researcher an understanding of the ways of
thinking of the students, and they can show why it makes sense that a student solved
a problem or thought about a topic in a particular way. Thus, models can provide a
researcher with a great sense of fit. And yet, the models are actually instruments of
interaction (Steffe & Olive, 2010): They allow researchers to better interact with
these particular students because the researchers can base problems and questions
on the ways of thinking in the model. Doing so can facilitate communication about
mathematical ideas with particular students. This aspect of models can also feel
satisfying because it can engender a sense of connection between the researcher and
students (see Chaps. 9 and 14).
And yet, if the models were only useful for the particular students with whom
researchers were working in particular studies, that would be quite limiting as
research. Fortunately, experience shows that is not at all the case. Another reason
genetic epistemology is useful is that the models developed with a few students usu-
ally allow researchers to interact more broadly with other students who have simi-
larities to the students from whom the models were made (see Chap. 14). So, as
researchers build models for particular students, they are usually building models
that are useful with a wide range of students.
Organization of the Book
We mention these models of students’ mathematics because they portray how

Piaget’s genetic epistemology has contributed to mathematics education research.
More importantly, Piaget provided a rich set of theoretical tools for pursuing this
kind of research. The goal of this book is not to describe particular models of stu-
dents’ mathematics that have been developed but rather to describe the tools that
mathematics educators use to construct such models. We have thus organized the
main body of the book—Part 2, Chaps. 3, 4, 5, 6, 7, 8, 9, 10, and 11—around clus-
ters of related constructs. To be precise, the first seven of those chapters describe
constructs directly descended from Piaget’s research, and the last two describe con-
structs developed later on but whose importance to mathematics education research
warranted their inclusion in this volume.
The rest of the book—Parts 3 and 4, Chaps. 12, 13, 14, 15, 16, 17, 18, 19, 20, and
21—contains two parts corresponding to two different ways of building on the con-
struct chapters in Part 2. The chapters in Part 3 each contain commentaries on the
first part and on genetic epistemology more broadly. The chapters in Part 4 each
summarize the research agenda of a younger mathematics education scholar who
draws upon genetic epistemology in their work. These final chapters provide further
examples of the utility and fecundity of this body of theory.
We could have adopted other organizational approaches such as a historical
account of how ideas developed, by focusing on the various scholars who drew upon
Piaget in their research, or by surveying key findings and contributions developed in
this tradition. We adopted the current organization because we anticipated it would
be most useful to scholars who want to learn about these tools to engage in mathe-
matics education research. In other words, we organized the book looking forward
to future research rather than trying to survey or summarize previous research. As a
result, the contributions of many important mathematics educators who draw heav-
ily upon Piaget’s work may be underrepresented or omitted in these pages.
We have included Chap. 2 as an acknowledgment of the history of research and
the intellectual heritage by which this body of theory has come to us. Dr. Les Steffe
is one of the central scholars who draws upon Piaget’s work to study children’s
mathematics and who trained many of the other authors to do the same. Chapter 2
presents Steffe’s historical reflection on Piaget’s influence on mathematics
education.
This book was formulated to serve as a graduate textbook for those studying to
become researchers in mathematics education. We sense that our field needs more
such texts, especially regarding rich and complex bodies of theory such as genetic
epistemology. We sincerely hope that this book provides a helpful starting point for
those newer to these ideas and a productive resource for those more experienced.
We have learned much from the chapters our excellent coauthors contributed, which
makes us confident that what follows will be of value to the field. It is a joy to be
continually engaged as learners: learners of mathematics, especially students’ math-
ematics, and learners among the community of researchers trying to support quality
mathematics education.
References
Bergson, H. (1998). Creative evolution (A. Mitchell, Trans.). Dover. (Original work published
in 1911).
Beth, E. W., & Piaget, J. (1966). Mathematical epistemology and psychology. D. Reidel Publishing
Company.
Hackenberg, A. J. (2007). Units coordination and the construction of improper fractions: A revi-
sion of the splitting hypothesis. Journal of Mathematical Behavior, 26(1), 27–47. https://doi.
org/10.1016/j.jmathb.2007.03.002
Hackenberg, A. J. (2013). The fractional knowledge and algebraic reasoning of students with the
first multiplicative concept. Journal of Mathematical Behavior, 32(3), 538–563.
Hackenberg, A. J., & Sevinc, S. (2021). A boundary of the second multiplicative concept: The
case of Milo. Educational Studies in Mathematics., 109, 177. https://doi.org/10.1007/
s10649-021-10083-8
Kant, I. (1781). Critique of pure reason (N.K. Smith, Trans.). Macmillan.
Noddings, N. (1990). Constructivism in mathematics education. Journal for Research in
Mathematics Education (Monograph): Constructivist Views on the Teaching and Learning of
Mathematics, 4, 7–18.
Norton, A., & Boyce, S. (2015). Provoking the construction of a structure for coordinating n+ 1
levels of units. The Journal of Mathematical Behavior, 40, 211–232.
Norton, A., & Wilkins, J. L. M. (2012). The splitting group. Journal for Research in Mathematics
Education, 43(5), 557–583. https://doi.org/10.5951/jresematheduc.43.5.0557
Piaget, J., & Inhelder, B. (1967). The child’s conception of space (F.J. Langdon & J.L. Lunzer,
Trans.). Norton.. (Original work published in 1948).
Steffe, L. P., & Cobb, P. (1988). Construction of arithmetical meanings and strategies.
Springer-Verlag.
Steffe, L. P., & Olive, J. (2010). Children’s fractional knowledge. Springer Science &
Business Media.
Steffe, L. P., & Thompson, P. W. (2000). Teaching experiment methodology: Underlying principles
and essential elements. In A. E. Kelly & R. Lesh (Eds.), Handbook of research design in math-
ematics and science education (pp. 267–306). Erlbaum.
Steffe, L. P., von Glasersfeld, E., Richards, J., & Cobb, P. (1983). Children’s counting types:
Philosophy, theory, and application. Praeger Scientific.
Ulrich, C., Tillema, E. S., Hackenberg, A. J., & Norton, A. (2014). Constructivist model building:
Empirical examples from mathematics education. Constructivist Foundations, 9(3), 328–359.
Vidal, F. (1994). Piaget before Piaget. Harvard University Press.
von Glasersfeld, E. (1995). Radical constructivism: A way of knowing and learning. Falmer.
Chapter 2
An Historical Reflection on Adapting
Piaget’s Work for Ongoing Mathematics
Education Research
Leslie P. Steffe
Piaget and Modern Mathematics
Piaget was “rediscovered” (Ripple & Rockcastle, 1964) during the 1960s by math-
ematicians and mathematics educators whose goal was to reform mathematics cur-
ricula based on modern mathematics (e.g., Allendoerfer & Oakley, 1959; School
Mathematics Study Group, 1965). Logical–mathematical structure served as the
basic rationale for the new math programs that, in many cases, resembled collegiate
mathematics. Although classical idealism, the doctrine that reality, or reality as we
know it, is fundamentally mental, served operationally as the epistemological posi-
tion of the reformers, empiricism and realism were still the more general positions
in the United States as indicated by a return to behaviorism in the decade following
the modernist programs. Problem solving, along with learning by discovery, was the
major psychological emphases among the reformers (Pólya, 1945, 1981) for which
Wertheimer’s1 (1945) work on productive thinking served as a basic rationale.
Piaget’s genetic epistemology (Piaget, 1970) did not serve as an epistemological
basis for the modern programs, nor was it explicitly emphasized at a conference
held at Cornell University and the University of California to investigate implica-
tions of Piaget’s work for mathematics education (Ripple & Rockcastle, 1964). The
interest of the conference organizers was in exploring the implications of Piaget’s
Wertheimer was one of the three founders of Gestalt psychology along with Kurt Koffka and
1
Wolfgang Köhler.
L. P. Steffe (*)
Department of Mathematics, Statistics, and Social Studies Education, University of Georgia,
Athens, GA, USA
e-mail: lsteffe@uga.edu

https://doi.org/10.1007/978-3-031-47386-9_2
12 L. P. Steffe
stages of cognitive development as a rationale for the elementary programs because

those programs were left without a psychological rationale (Ripple & Rockcastle,
1964). Piaget (1964) was invited to present four papers at the conference that he
titled, “Development and learning,” “The development of mental imagery,” “Mother
structures and the notion of number,” and “Relations between the notions of time
and speed in children.” Although he made no reference to genetic epistemology in
these papers, by presenting them, he did implicitly explain the concept of genetic
epistemology that he presented at the Woodbridge Lectures at Columbia University
in 1968 (Piaget, 1970).
Genetic epistemology attempts to explain knowledge, and in particular scientific knowl-
edge, on the basis of its history, its sociogenesis, and especially the psychological origins of
the notions and operations upon which it is based. (p. 1)
Development vs. Learning
Although he could have oriented his papers as elaborations of genetic epistemology,

his emphasis at the conferences was on explaining the cognitive development of
number, space, and time as opposed to teaching such concepts and expecting them
to be learned. He made a sharp distinction between development and learning in that
development is a spontaneous process tied to the whole process of embryogenesis.
Embryogenesis concerns the development of the body but it concerns as well the develop-
ment of the mental functions. In the case of the development of knowledge in children,
embryogenesis ends only in adulthood. … In other words, development is a process—
which concerns the totality of the structures of knowledge. (Piaget, 1964, p. 8)
In contrast, he explained learning as presenting the opposite case.

In general, learning is provoked by situations. … It is provoked in general as opposed to
spontaneous. In addition, it is a limited process—limited to a single problem or to a single
structure. (Piaget, 1964, p. 8)
Although I do not regard learning as such a limited process, the developers of the
modern programs firmly believed that their programs could be learned and would
either accelerate Piaget’s account of the cognitive development of basic mathemati-
cal notions or essentially replace developmental processes.
Most, if not all, of the major curricular reform projects have taken the logic (or structures)
of the subject matter as a point of departure rather than psychological learning theory or
studies of cognitive development. This point is made abundantly clear, for example, in
Jeremy Kilpatrick’s paper on the SMSG Program included in this report. (Ripple &
Rockcastle, 1964, p. iii)
The curriculum developers considered Piaget to be an observer rather than a teacher,

and the elasticity of the limits of children’s minds was not considered as having
been established.
2 An Historical Reflection on Adapting Piaget’s Work for Ongoing Mathematics… 13
These reformers (and I speak now not only of SMSG) have been so successful in teaching
relatively complex ideas to young children, and thus doing considerable violence to some
old notions of readiness, that they have become highly optimistic about what mathematics
can and should be taught in the early grades. (Kilpatrick, 1964, p. 129)
The lack of appreciation for genetic epistemology was addressed by Eleanor

Duckworth, a former student of Piaget, who served as an intermediary between
Piaget and the conference attendees. She addressed the teaching the “structure” of a
subject matter.
The pedagogical idea is that children should be taught the unifying themes of a subject mat-
ter area, after which they will be able to relate individual items to this general structure.
(This seems to be what Bruner often means by ‘teaching the structure’ in the Process of
Education). (Duckworth, 1964, p. 3)
Developmental vs. Mathematical Structure
The structural emphasis of the modern programs was not compatible with Piaget’s
emphasis on the structure of operational thought.
An operation is an interiorized action. … Above all, an operation is never isolated. It is
always linked to other operations and as a result it is always a part of a total structure.
(Piaget, 1964. p. 7)
A major difficulty was that “structure” had very different meanings for Piaget and
for the curriculum developers. Piaget’s structures were second-order models, “the
hypothetical models observers may construct of the subject’s knowledge in order to
explain their observations (i.e., their experience) of the subject’s states and activi-
ties” (Steffe et al., 1983, p. xvi). The mathematical structures of the modern pro-
grams were first-order models, or renditions of the mathematical knowledge of the
curricular developers (cf. Steffe et al., 1983). This distinction between the mathe-
matical thought of the child from the point of view of the adult and the adult’s own
mathematical knowledge that he or she would not attribute to the child has been and
remains a major issue in the mathematics education of children. Even though it is
assumed in genetic epistemology that the mathematical thinking of children as it
evolves over time has something to do with mature mathematical thinking, it takes
major decentering for an adult mathematical thinker to think as if he or she is a child
(Thompson & Thompson, 1994). Further, in those cases where the adult does learn
to think as if he or she is a child, developing models of how such an evolution might
occur is quite exacting. Hermine Sinclair succinctly pointed out such difficulties at
the level of child thought in attempts by mathematics educators to use Piaget’s
genetic epistemology at a conference held at Columbia University in 1970.
At first sight it would seem that a psychological theory that is regarded by its author as a
“by-product” of his epistemological research and is therefore principally directed toward
the investigation of knowledge and its changes in the history of mankind, as well as in the
growing child, is ideally suited to educational applications. (Sinclair, 1971, p. 1)
14 L. P. Steffe
Sinclair used a metaphor to explain what she regarded as difficulties in trying to

apply Piaget’s stage theory of cognitive development in an attempt to provoke non-
operational children to become operational.
Piaget’s tasks are like the core samples a geologist takes from a fertile area and from which
he can infer the general structure of a fertile soil; but it is absurd to hope that transplanting
these samples to a field of nonfertile soil will make the whole area fertile. (Sinclair,
1971, p. 1)
Preludes to IRON (Interdisciplinary Research on Number)
Piagetian Research
Professor Henry Van Engen introduced me to Piaget’s work while I was a doctoral
student at the University of Wisconsin, working as a research associate in the
Research and Development Center for Cognitive Learning. Following Bridgeman
(1927), mental operations were at the heart of Van Engen’s meaning theory of arith-
metic, where he defined the meaning of a symbol as an intention to act (Van Engen,
1949a, b). Piaget’s emphasis on mental operations and operational thought was a
major point of convergence and served as the basis of Van Engen’s interest in Piaget
(Van Engen, 1971). As a research associate, it was my wont to apply Piaget’s theory
of the development of number in the mathematics education of young children
using scientific methods, which in part translated to investigating the importance of
conservation of numerosity of first-grade children on arithmetical tasks using
research design (Stanley & Campbell, 1963) and statistical methods (Steffe, 1966).
I continued on with this program of research, which became known as “Piagetian
Research” (Steffe & Kieren, 1994), for 7 years after joining the faculty of mathe-
matics education at the University of Georgia, a time during which I directed ten
doctoral students in applying Piaget’s research in the mathematics education of chil-
dren. Professor Charles Smock of the Psychology Department, who was a Piagetian,
served on the committees of most of my doctoral students during that time, which
was quite important because he was the one who eventually introduced me to Ernst
von Glasersfeld.
A Change in Research Program
I was working as a realist and an empiricist in my attempts to apply Piaget’s devel-

opmental theory in the mathematics education of children, and I was making only
accretional rather than recursive progress. As a consequence of making only mini-
mal progress, I abandoned my attempts to apply Piaget’s research on number and
quantity as well as my statistical method of application and taught a group of first
graders with the help of two of my advanced doctoral students for an academic year
in an attempt to let children teach me what was important in their numerical ways
of operating (Steffe et al., 1976a, b). The importance of this work was two-fold.
First, it was the impetus for the development and use of the constructivist teaching
experiment as a legitimate scientific method of doing research (Steffe, 1983, 1991b;
Steffe & Cobb, 1983; Steffe & Thompson, 2000; Steffe & Ulrich, 2013). Second,
the finding that counting was the children’s basic method of solving arithmetical
tasks led to abandoning attempts to apply Piaget’s idea that number is constructed
as a synthesis of classification and seriation operations concurrently with the intro-
duction of the arithmetical unit and to investigating children’s construction of num-
ber sequences in the context of their spontaneous use of counting in solution of
arithmetical tasks in teaching experiments (Steffe, 1994, 1996; Steffe et al., 1983;
Steffe & Cobb, 1988).
A Fortunate Introduction
It was extremely fortunate, not only to me at the time but also to my doctoral stu-
dents and to the field of mathematics education at large, that Professor Smock
invited me to a seminar given by Ernst von Glasersfeld (Steffe, 2013). The seminar
event arranged by Smock occurred around 1974, shortly after the demise of the
modern mathematics movement and during the move back to behaviorism that
occurred in the 1970s. At the time, the question concerning whether mathematics
was invented or discovered held little sway with me even though I had read Piagetian
basic books such as The Child’s Conception of Number (Piaget & Szeminska,
1952), The Child’s Conception of Geometry (Piaget et al., 1960), The Child’s
Conception of Space (Piaget & Inhelder, 1967), and The Growth of Logical Thinking
from Childhood to Adolescence (Inhelder & Piaget, 1958). I understood that chil-
dren developed mathematical knowledge, but what developed I regarded as a pre-
lude to what was “out there” in some mathematical reality. My conception of
mathematics was, and still is, widely shared by mathematics educators as well as
mathematicians. According to Stolzenberg (1984), it is indisputable that the con-
temporary mathematician operates within a belief system whose core belief is that
mathematics is discovered rather than created or invented by human beings.
My belief in the objective existence of mathematics was seriously questioned by
a story Glasersfeld recounted in the seminar. The story, taken from Letvin et al.
(1959), clarified that the only contact we have with what is “out there” is through
our senses. When talking about a frog as a fly-catcher, he commented that:
The system [the frog’s visual system] as a whole makes the frog an efficient fly-catcher,
because it is tuned for small dark “objects” that move about in an abrupt fly-like way. In the
frog’s natural habitat, as we, who observe the frog see it, every item that possesses the
characteristics necessary to trigger the frog’s detectors in the proper sequence is a fly or bug
or other morsel of food for the frog. But if the frog is presented with a black bead, an air-gun
pellet, or any other small dark moving item, it will snap it up as though it were a fly. In fact,
16 L. P. Steffe
to the normal frog’s visual apparatus, anything that triggers the detectors in the right way,
is a “fly.” (Glasersfeld, 1974, p. 15; reprinted in Glasersfeld, 1987, pp. 106–07)
The point of the story, of course, is that whatever is perceived is basically a compo-
sition of sensory signals generated in our various sensory channels.
We are, of course, free to consider these original signals the effect of some outside causes.
But since there is no way of approaching or “observing” these hypothetical causes, except
through their effects, we are in the same relation to that “outside” in which the first cyber-
neticists found themselves with regard to living organisms—that is to say, we are facing a
“black box.” (Glasersfeld, 1974, p. 16; reprinted in Glasersfeld, 1987, p. 107)
Soon after the colloquium, Ernst presented his seminal paper on radical constructiv-
ism at a conference sponsored by Charles Smock’s Mathemagenic Activities
Program (von Glasersfeld, 1974). In the paper, Glasersfeld argued that radical con-
structivism constituted a legitimate interpretation of Piaget’s genetic epistemology.
The radical constructivist’s interpretation of Piaget’s Genetic Epistemology then, consists
in this: The organism’s representation of his environment, his knowledge of the world, is
under all circumstances the result of his own cognitive activity. The raw material of his
construction is “sense data,” but by this the constructivist intends “particles of experience”:
that is to say, items which do not entail any specific “interaction” or causation on the part of
an already constructed reality that lies beyond the organism’s experiential interface. (p. 22)
Ernst once told me that he had discussed his interpretation of Piaget’s genetic epis-
temology with Barbel Inhelder, Piaget’s close collaborator, and she agreed with it.
Interdisciplinary Research on Number (IRON)
Original Members of IRON
After becoming acquainted with Glasersfeld, along with the philosopher John
Richards and Patrick Thompson (who was then a doctoral student), we organized
the radical constructivist research program, Interdisciplinary Research on Number
(IRON). Although our general goal was to start a constructivist revolution in math-
ematics education of no less magnitude than the modern mathematics movement,
we each had our own goals. Glasersfeld (1995, p. 15) later commented that his
motivation for joining IRON was to develop the empirical data to countermand
empiricism and, further, to engage in work in a field where he would be taken seri-
ously as an epistemologist and philosopher. His particular goal was to develop his
attentional moments model (cf. Glasersfeld, 1981) and use it in an analysis of how
children construct numerical units in counting. Richard’s goal was to reconstitute
the foundations of mathematics based on how mathematics is constructed by human
beings. I had a videotape library of teaching episodes with first-grade children in a
year-long teaching experiment that I had conducted after my first encounter teach-
ing first-grade children (Steffe et al., 1976a, b). My initial purpose for engaging in
IRON was to construct explanations of the unit items children established and
counted in the previous teaching experiment.2
Ernst as Scientist
During the time we engaged in IRON, I knew Glasersfeld best as a coinvestigator,

as a scientist conducing conceptual analysis of children shown on the video-tapes
when they engaged in solving numerical situations (Steffe, 2011). Of these analy-
ses, he commented that,
He [Les Steffe], a graduate student of his [Pat Thompson], the philosopher John Richards,
and myself would spend countless hours viewing these tapes and trying to agree on what we
gathered from them. We had heated arguments and for all of us it was a powerful lesson,
hammering in the fundamental fact that what one observer sees is not what another may see
and that a common view can be achieved only by a strenuous effort of mutual adaptation.
(Glasersfeld, 2005, p. 10)
Ernst’s comment is testimony to how we, as an interdisciplinary team, proceeded in

interdisciplinary research in the context of analyzing children’s counting. It was
through these intensive social interactions that we learned the language and thinking
of the other members of the team, and I carried this lesson on throughout the IRON
research program.
Analyzing children’s counting may have seemed inconsistent with our identifica-
tion as “Piagetians” because Piaget and Szeminska (1952) excluded children’s
counting in their account of children’s construction of number. At a meeting of the
International Group for the Psychology of Mathematics Education at Grenoble,
France, in 1981, that question was put to Ernst in no uncertain terms. His answer
was incisive, and it at once established the originality of the research we were doing
in IRON.
Piaget’s account of children’s construction of number should not be regarded as an objec-
tive mental reality independent of the observer, a point on which he later elaborated. (Steffe,
2011, p. 173)
By making this comment, Glasersfeld made clear that Piaget’s models of the epis-
temic child were Piaget’s explanations of his observations. Our models differed
from Piaget’s because we used teaching experiments rather than clinical interviews
as our method of observation as well as different analytical constructs.
2
For example, I thought that when children put up a finger synchronous with uttering a number
word in an act of counting, putting up a finger was necessarily an arithmetical unit item.
18 L. P. Steffe
Piaget’s Epistemic Child
There was also widespread interest in Piaget’s account of the spontaneous develop-
ment of proportional reasoning during the era of modern mathematics and on into
the next decade (e.g., Lovell, 1971; Steffe & Parr, 1968). He used the operations that
he called inversion, negation, reciprocal, and correlative, which, when working in
consort, formed a model of proportional reasoning that he called the INRC group.
Piaget thought that proportional reasoning emerged at approximately 12 years of
age. Similar to Piaget’s models for children’s construction of number, the INRC
group was not of interest in later research of IRON that was done with fractions and
algebraic reasoning primarily because we constructed our own models of children’s
reasoning in these areas based on teaching experiments.
A reader might infer that we abandoned Piaget’s analysis of situations that he
used to analyze preoperational, concrete operational, and formal operational thought
as contained in the basic books cited above. That is, a reader might infer that we
abandoned Piaget’s analysis of his observations. But nothing could be further from
the case. Piaget’s grouping structures that he used to model concrete operational
thought and his INRC group that he used to model formal operational thought were
models of the epistemic child, which Piaget (1966) explained as “that which is com-
mon to all subjects at the same level of development, whose cognitive structures
derive from the most general mechanisms of the co-ordination of actions” (p. 308).
Although we had abandoned Piaget’s accounts of his epistemic child, we had not
abandoned his analyses of his copious observations and, at times, used them in our
analyses of our observations (cf. Steffe & Olive, 2010). We had mounted what
Lakatos (1970) referred to as a progressive research program in which we were
interested in exploring children’s construction of mathematics in teaching experi-
ments throughout the elementary, middle school, and early secondary school years.
At the outset of IRON, our immediate concern was with young children’s construc-
tion of units and number. The explanatory models that were later constructed were
simply not available at the outset of our work in IRON.
Conceptual Analysis
The First Type
Before Glasersfeld came to the United States, he had worked with Silvio Ceccato
(1974) in the Italian Operational School (Glasersfeld, 2005) and brought with him
the idea of conceptual analysis (Glasersfeld, 1981). Much of our initial effort that
led to Glasersfeld’s above comment was devoted to engaging in conceptual analysis
using the concept of attention in explaining the units children create and count. In
explaining word meaning, the goal of Silvio Ceccato’s Italian Operational School
was to “reduce all linguistic meaning, not to other words, but to ‘mental
operations’” (Glasersfeld, 1995, p. 78). This goal is defined by a question from

Ceccato’s group: “What mental operations must be carried out to see the presented
situation in the particular way one is seeing it?” (Glasersfeld, 1995, p. 78). We car-
ried out two major types of conceptual analysis (Thompson, 2000). The first one is
first-order conceptual analysis, or an analysis of one’s own conceptions, of the kind
that Glasersfeld carried out using the basic idea of attentional moments. The first-
order conceptual analysis extends, of course, to analysis of one’s own mathematical
conceptions. For example, Thompson (2008, p. 43) used conceptual analysis to
“devise ways of understanding an idea that, if students had them, might be propi-
tious for building more powerful ways to deal mathematically with their environ-
ments than they would build otherwise.”3 An interesting twist in Thompson’s
conceptual analysis is that he often referenced students’ thinking.
Attentional Moments and the Unitizing Operation
Glasersfeld explained attention as a mechanism that compounds sensory material

from various sensory channels into a “whole” or “thing.”
Attention is not to be understood as a state that can be extended over longish periods.
Instead, I intend a pulselike succession of moments of attention, each one of which may or
may not be “focused” on some neural event in the organism. (Glasersfeld, 1983, p. 126)
Focused moments of attention in between unfocused moments constitute the opera-

tion that is used to establish a unitary item—the unitizing operation.
According to our analysis, the attentional operations that create unitary items in all areas of
experience are essentially the same as those that generate units of units that have been
called “number” since Euclid. If that is so, it follows that these “unitizing” operations play
a crucial role in all activities that involve numbers. (Glasersfeld & Richards, 1983, p. 4)
Glasersfeld (1981, 1983) used his attentional model to develop a hypothetical, first-
order model of the operations that produce number starting with the infant’s earliest
sensory items. At the end of his presentation, he commented,
To conclude, I should like to stress two points that, to me, are the central ideas explicated
by this model. First, when we speak of “things,” “wholes,” “units,” and “singulars,” on the
one hand and “plurals,” “pluralities,” “units,” “collections,” and “lots,” on the other, we
refer to conceptual structures that are dependent on material supplied by sensory experi-
ence. Insofar as these concepts involve sensory-motor signals, they do not belong to the
realm of number. They enter that rarified realm through the process of reflective abstraction,
which extricates attentional patterns from instantiations in sensory-motor experience.
(Glasersfeld, 1983, p. 134)
I recall Glasersfeld commenting that extracting attentional patterns from sensory-

motor items and recording them is the simplest form of reflective abstraction.
Glasersfeld’s (1981) attentional model of the conceptual construction of units and
3
See Thompson and Saldanha (2003) for an example.
20 L. P. Steffe
number has not itself been widely used in mathematics education research. But
what follows from the model, units and their systems, have been widely used. The
model doesn’t specify the material of construction, but it does provide a way of
thinking about the unitizing operation and its products.
The Second Type
The second type of conceptual analysis is found in the first publication of IRON,
“Children’s Counting Types” (Steffe et al., 1983), which included our analysis of
the video tapes that I had stored in my tape library at the outset of IRON using
Glasersfeld’s attentional model (Glasersfeld, 1981). This type of conceptual analy-
sis is a second-order conceptual analysis in that it is an analysis of the conceptions
of the other. The goal of this analysis is to produce second-order models of the
“black boxes” constituted by children’s minds. To account for our observations, we
found it necessary to distinguish counters of perceptual unit items, figurative unit
items, and abstract unit items. The importance of making these distinctions among
children’s counting types was that they were later interpreted in terms of distinct
stages in children’s construction of counting schemes, multiplicative schemes, and
fraction schemes (cf. Steffe, 1992; Steffe & Cobb, 1988; Steffe & Olive, 2010).
Schemes
After the original publication of IRON (Steffe et al., 1983), I mounted a new 2-year
teaching experiment starting in 1980 along with a doctoral student, Paul Cobb, with
six first-grade children in order to replicate the results of my original experiment
and to generate new empirical content not previously available in the original exper-
iment. Glasersfeld’s (1980) paper, “The Concept of Equilibration in a Constructivist
Theory of Knowledge,” was his most important paper for our work in this teaching
experiment as well as those that followed. In the paper, Glasersfeld interpreted
Piaget’s (1980) concept of scheme, which Piaget defined as “action that is repeat-
able or generalized through application to new objects” (p. 24).4 In harmony with
Piaget, Glasersfeld (1980) explained the concept as an instrument of action or inter-
action and elaborated it in a way that opened the possibility of focusing on assimila-
tion in the construction or use of a scheme as well as what might go on subsequent
to action or interaction.
Schemes are basic sequences of events that consist of three parts. An initial part that serves
as trigger or occasion.… The second part, that follows upon it, is an action (“response”) or
4
Cf. Thompson and Saldanha (2003) for an example
an operation (conceptual or internalized activity).… The third part is… what I call the
results or sequel of the activity.
…
This feature of the Piagetian model, as I see it, constitutes its main basis as a constructivist
theory of cognition in which “knowledge” is no longer a true or false representation of real-
ity but simply the schemes of action and the schemes of operating that are functioning
reliably and effectively. (pp. 80, 83)
He illustrated his concept of scheme using the sucking scheme, a scheme that Piaget
considered a fundamental scheme babies use in their construction of object con-
cepts. I have no access to the schemes that Glasersfeld might have constructed
regarding the second teaching experiment because he did not participate in the con-
ceptual analyses of the video tapes. The schemes that did emerge from the chil-
dren’s behavior were novel and not known to me prior to the teaching experiment,
although I had experienced children’s counting in my first teaching experiment that
served in the original publication of IRON. To make a claim that a child has con-
structed a scheme follows from constructing the scheme for oneself as a result of
intensive as well as extensive interaction with one or more children and conceptual
analysis of their mathematical behavior. It is a major claim based on original contri-
butions children make to task solutions as well as on repeated observations over a
period of time long enough so that one is convinced their behavior is characteristic
and stable (Steffe, 1994, 1996; Steffe & Cobb, 1988; Steffe & Olive, 2010).
Systems of Schemes
Thompson (2013) critiqued Glasersfeld’s concept of scheme in his work on quanti-

tative reasoning.
My understanding of what Piaget meant by “scheme” differs from that proposed by Cobb
and Glasersfeld (1983) and by Glasersfeld. (1995, 1998)
…
I believe what Cobb and Glasersfeld described fits better with what Piaget called a schema
of action (Piaget, 1968, p. 11; Piaget & Inhelder, 1969, p. 4). Piaget spoke of a child’s suck-
ing schema, for example. I believe Piaget had larger organizations in mind when he spoke
of schemes—organizations of operations, images, schemata, and schemes—that did not
have easily identified entry points that might trigger action. (p. 60)
Glasersfeld did illustrate the concept of scheme as well as assimilation and accom-
modation using the sucking scheme, as I noted above. Because of his interest in
using his attentional model for the conceptual construction of units and number,
however, I never felt constrained to using his interpretation of scheme in the case of
only his most elementary composite wholes that he called “collection” or “lot” as
the first part of a scheme (Glasersfeld, 1981, p. 90). When teaching the children in
the teaching experiment, it became evident that the nature of counting varied within
22 L. P. Steffe
a particular child over time as well as across the children. As a consequence, five
different counting schemes using Glasersfeld’s characterization were isolated and
explained: the perceptual counting scheme, the figurative counting scheme, the ini-
tial number sequence, the tacitly nested number sequence, and the explicitly nested
number sequence (Tzur, 1999). The latter three schemes were operative schemes5
whose assimilative structures were numerical structures. In two later teaching
experiments (Steffe, 1992; Steffe & Olive, 2010), the explicitly nested number
sequence was reorganized in such a way that the reorganizing children constructed,
or were capable of constructing, systems of schemes that I find compatible with
Thompson’s conception of scheme. In fact, beyond the most elementary mathemati-
cal schemes, it has always been necessary to construct systems of schemes to
explain children’s mathematical behavior. But it is not identical because of the kinds
of conceptual analyses in which we engaged.
The Initial Number Sequence as a Scheme
To illustrate the concept of scheme, I use the initial number sequence for which
counting-on is the behavioral indicator (Steffe, 1988a; Steffe & Olive, 2010,
pp. 35ff). The first part of the scheme, the structure that is used in assimilation,
consists of a pattern of numerical unit items that contain records of counting.6
Consider this question: “If four jacks are placed in a bag holding six jacks, how
many jacks are in the bag?” Upon assimilation of the question, as heard by the stu-
dent, it may constitute a perturbation that I understand as “not just inputs but only
such inputs as upsets the organism’s equilibrium” (Glasersfeld, 1980, p. 78). Since
the goal is to restore equilibrium, it can be thought of cognitively7 as the student
generating a goal to find how many jacks are in the bag in order to restore equilib-
rium. The goal activates the activity of the scheme, which in this case is to count
four more times beyond six. As I have already pointed out, the activity of the scheme
is “contained in” the first part of the scheme, as the scheme is an operative numeri-
cal scheme.8 The counting activity feeds back into the goal of the scheme as the
activity is in progress, and vice versa. The third part of the scheme consists of a
numerical composite of definite numerosity produced by counting that completes
the feedback into the goal and closes the scheme.9 The feedback system operates as
a monitor of the activity and is essential in knowing when to stop counting (Steffe
& Olive, 2010, p. 23) (Fig. 2.1).
5
A counting scheme is judged as operative if a child can at least count-on.
6
When the unitizing operation operates on counted items, records of the acts of counting as well as
of the involved sensory items are part of the abstracted unit items.
7
The disequilibria can also have emotional or sensory overtones.
8
“A stimulus is really a stimulus only when it is assimilated into a structure and it is this structure
which sets off the response” (Piaget, 1964, p. 15).
9
The perturbation has been eliminated and the scheme is in a state of equilibrium.
Generated Goal
• Find how many jacks are

in the bag
Situation
Result
• Heard question
• Activated numerical • A numerical composite,
composite ten
• Disequilibrium • Equilibrium
Generated Activity
• Counting four more times

beyond six
Fig. 2.1 The concept of scheme
Operationalizing Reflective Abstraction
While Paul Cobb and I were involved in the teaching experiment with the six chil-
dren, Glasersfeld continued on with his analysis of Piaget’s two volumes on reflec-
tive abstraction, which were available only in French (Piaget & collaborators, 1977).
Glasersfeld’s (1991) analysis was particularly important for our work in the teach-
ing experiments.10 Of reflective abstraction, Glasersfeld (1991) quoted Piaget as
follows11:
Reflective abstraction always involves two inseparable features: a “réfléchissement” in the
sense of the projection of something borrowed from a preceding level onto a higher one,
and a “réflexion” in the sense of a (more or less conscious) cognitive reconstruction or
reorganization of what has been transferred. (Piaget, 1975, p. 41) (p. 58)
Reflective abstraction provided a general explanatory construct that Piaget used to

explain the transition from the preoperational stage of representation to the stage of
concrete operations. In the previous stage, children do not reason logically whereas
in the latter stage, children can reason logically. For example, given a collection of
sticks all of different length, children in the stage of concrete operations order the
sticks according to their length, whereas preoperational children would most likely
order them pairwise if they ordered them at all. Although Piaget thought that four
factors contribute to development—social interaction, experience, maturation, and
equilibration (Duckworth, 1964), to explain the general transition, he used reflective
10
I have a draft of Glasersfeld’s paper written in 1982 which he first published in 1991.
11
Glasersfeld was fluent in four languages of which French was one. So, this was an original trans-
lation that he made.
24 L. P. Steffe
abstraction. One of the major goals of the teaching experiment was to operationalize
reflective abstraction, so it could be useful to mathematics teachers. Operationalizing
reflective abstraction was similar to operationalizing Glasersfeld’s scheme theory in
constructing second-order models of children’s mathematics.
Constructing the Initial Number Sequence
Tyrone
I found occasions that could be interpreted as Piaget’s reflective abstraction if one
does not expect both aspects to occur in close proximity. One occasion occurred in
my work in the teaching experiment when I was working with a child, Tyrone, in his
construction of the initial number sequence (Steffe, 1988b, pp. 284–322). Tyrone
had constructed the figurative counting scheme, which observationally means that
he could count items that were not in his immediate visual field, but he was yet to
count on. He was presented with a collection of items, seven of which were covered
by one cloth and five by another cloth, and was asked to count the items.12 He
pointed to places on the first cloth he took as covering an item as he subvocally
uttered number words “1, 2, 3, 4, 5, 6, 7.” He then proceeded to touch the second
cloth six times in a row, whispering, “8, 9, 10, 11, 12, 13,” which indicates that he
did not have a five pattern that he could use to keep track of counting five more
times. Failing to recognize a pattern that would close his scheme, he started over
twice. On his last attempt, he again started counting from “one” and stopped at
“twelve” with conviction. When counting five more times past “seven,” he stared
into space while touching the cloth synchronously with uttering number words.
I interpreted this final count over the second cloth as Tyrone creating a linear
pattern for “five” by intentionally monitoring his counting acts in the following way.
After saying “8, 9,” say, and making two pointing acts, he must have re-presented13
the results of these counting acts and “held the results at a distance,” reflecting on
them as indicated by staring into space. Distancing himself in this way requires an
operation not provided by re-presentation. I inferred that he took each visualized
counting act as a unit using the unitizing operation, creating an arithmetical unit
item containing records of the counting act from which it was abstracted. This is the
process that creates interiorized counting acts14 as distinct from the internalized
counting acts from which they were abstracted. The process continued on until
Tyrone established a pattern of five interiorized arithmetical unit items or a numeri-
cal composite (cf. Steffe & Cobb, 1988).
12
This task was designed so that Tyrone would at least need to visualize the items he counted as
well as eliminate visual cues for when to stop counting.
13
A re-presentation is a re-generation of a past experience (Glasersfeld, 1991, p. 49).
14
An interiorized counting act is purged of its sensory-motor material. I think of it as a slot that
contains records of the sensory-motor material.
The establishment of a numerical composite is a projection of the five counting

acts, “8, 9, 10, 11, 12,” to the interiorized level, which is one level beyond his figura-
tive counting acts, which were at the level of his figurative counting scheme. I called
the accommodation he made in his figurative counting scheme an engendering
accommodation because he came back from Christmas vacation able to count-on.
Counting-on indicated that a metamorphosis had occurred in his counting scheme
over Christmas vacation.
To account for the observed metamorphosis, the numerical pattern “8, 9, 10, 11,
12” that Tyrone established was an interiorized pattern of arithmetical unit items,
whereas his figurative counting scheme consisted of internalized counting acts.
That is, the numerical pattern “8, 9, 10, 11, 12” had been disembedded from his
internalized counting scheme as an abstracted numerical concept and was estab-
lished at a level “above” his internalized counting scheme of which the figurative
counting acts were still a part. However, I wouldn’t say that the sequence of interior-
ized counting acts is applied to the higher level. Rather, monitoring counting is that
process that creates the level of interiorization. The presence of the numerical pat-
tern would have the effect of setting up a systemic perturbation that would drive the
interiorization of “close by” counting acts that would continue until the perturbation
dissipated. I don’t know how far his interiorized number sequence extended forward
from “one,” but it was far enough for him to engage numerically in solving the
arithmetical tasks that were presented to him after Christmas. I referred to this sec-
ond accommodation as a metamorphic accommodation that followed on from the
accommodation that engendered it and interpreted it as the “réflexion” part of
Piaget’s reflective abstraction.
Jason
The primary reason that three children were selected for the teaching experiment
whose counting schemes were judged as figurative schemes at the outset of the
experiment was to explore if the accommodations the three children made in the
schemes in the solution of presented tasks could legitimately be judged as equiva-
lent. So, the integration operation emerged for Tyrone’s two cohorts in the teaching
experiment in the context of patterns as well confirmed the crucial role of patterns
in the construction of numerical composites. In what follows, I focus on the task
where Jason was observed establishing a numerical pattern. The teacher/researcher
(Paul Cobb) pretended to place eight cookies under a cloth and then some more and
told Jason that there were ten cookies under the cloth (Steffe & Cobb, 1988). Jason
interpreted the task as the teacher/researcher placing eight cookies under the cloth
and then placing ten more with them.
Like Tyrone, Jason concentrated intensely on keeping track of extending his first
eight counting acts ten more times when the “breakthrough” occurred; that is, he
kept track of counting ten more times and stopped after his counting act, “eighteen.”
This counting episode is taken as indication of Jason projecting the involved inter-
nalized counting acts to the interiorized level. After this observation on March 18,
1981, Jason’s counting scheme remained figurative (he always started counting
from “one”) until May 21, 1981 of his first grade when he applied the integration
26 L. P. Steffe
operation sequentially and counted on. So, like Tyrone, the projective aspect of
reflective abstraction occurred in the context of solving tasks on March 18, but it
wasn’t until approximately 2 months later that his figurative counting scheme
emerged as an interiorized number sequence, which was indeed a metamorphosis of
the figurative counting scheme.
So, similar to Tyrone, the “réflexion” aspect of Piaget’s reflective abstraction,
which he characterized as “a (more or less conscious) cognitive reconstruction or
reorganization of what has been transferred,” did not occur in close temporal prox-
imity to the observation of the “projective” aspect. It is important to emphasize that
operationalizing reflective abstraction was accomplished in the context of intensive
communicative interaction with the two children in a teaching experiment. It is also
important to emphasize that it shouldn’t be inferred that the initial reflective abstrac-
tion that I have explained, whose result was the construction of an interiorized
counting scheme, was sufficient for the children to construct, for example, subtrac-
tion as the inversion of addition nor to strategically find the pairs of numbers whose
sum is ten. In general, children’s schemes and their affordances and constraints are,
in part, constructed by a teacher/researcher in the context of conceptual analysis of
recorded teaching episodes. More importantly, they are also constructed by means
of actually teaching children in teaching episodes. I have been emphasizing the
former without emphasizing the creativity of the researcher as a teacher in con-
structing what I call experiential models of the participating children’s mathemat-
ics. It is these experiential models that ground the conceptual analyses of the corpus
of video material. In short, the basic premise of teaching experiments is that the
teacher/researcher must construct the mathematics of children, which are second-
order models of children’s mathematics.
Further Operationalizing Reflective Abstraction
Three of the six children in the teaching experiment, one of whom was a child
named Brenda, entered the experiment as counters of perceptual unit items, mean-
ing they could not count unless the countable items were in their perceptual field. To
illustrate, an interviewer covered six of nine marbles with his hand and asked Brenda
to count all the marbles. Brenda first counted the interviewer’s five fingers and then
counted the three visible marbles pointing to each in turn. The interviewer pointed
out that he had six marbles beneath his hand, and Brenda replied, “I don’t see no
six!” (Steffe & Cobb, 1988, p. 23). I inferred that Brenda was aware that marbles
were hidden by the interviewer’s hand as indicated by her initializing counting the
interviewer’s fingers as substitute15 countable items. Had she pointed to the inter-
viewer’s hand at places she thought that marbles were hidden rather than point to the
15
The interviewer’s fingers were not representatives of the hidden marbles. Rather, they were per-
ceptual replacements.
interviewer’s fingers, I would have inferred that she re-presented more than one, or
a plurality, of marbles and took these images as countable items in a way similar to
Tyrone when he counted to “seven.”
I taught these three children approximately 60 times over the course of the exper-
iment, and they also participated in their regular mathematics classrooms.
Unfortunately, counting-on did not emerge in these three children over the 2 years
that they spent in the teaching experiment in spite of my best efforts to provoke it
and, presumably, the best efforts of their teachers. Rather than re-present collections
of perceptual unit items and count figurative unit items as Tyrone and Jason did, the
three children re-presented finger patterns during their first grade. The re-presented
finger patterns served as substitutes for, but not representatives of, spatial patterns.
These children’s re-presentation of finger patterns is an elementary example of pro-
jection in Piaget’s account of reflective abstraction in its functional forms. Although
the three children were not without a semblance of the projective aspect of reflective
abstraction, their counting scheme remained as a perceptual scheme throughout
their first grade.
Learning Stages
A major consequence of Piaget’s reflective abstraction that is still not taken seri-
ously in mathematics education is that it produces developmental stages in chil-
dren’s spontaneous constructive activity. Although I had abandoned Piaget’s
epistemic child and his cognitive models of the preoperational and concrete opera-
tional child, as I have already commented, I had not abandoned the essence of his
observations. However, rather than speak of what preoperational children could not
do, which was to conserve numerosity or to solve the class inclusion problem, I now
spoke more positively of what children taught me that they could do, which was, in
the case of the three children, to engage in the goal-directed activity of counting
perceptual unit items to specify the plurality of perceptual collections. My hypoth-
esis at the outset of the teaching experiment was that the three children’s counting
schemes would remain essentially equivalent throughout the time that I taught them,
but it was my goal to provoke the maximal progress that I could in their constructive
activity.
The Perceptual Stage
The Period Criterion

I didn’t start the teaching experiment with the assumption that the children’s con-
structive activity would occur in learning stages (cf. Ulrich, 2015 for an elaboration
28 L. P. Steffe
of stages).16 The children imposed that observation upon me. I was compelled to
infer that being a counter of perceptual unit items constituted a stage according to
Glasersfeld and Kelly’s (1983) criteria for stages. The first of these is that an item
must remain constant throughout a period of time, which was counting perceptual
unit items. I don’t know when these three children learned how to count perceptual
unit items, but it preceded their entry into their first grade. So, there must have been
a period of time when they were 5 years of age or younger when they first learned
to count perceptual unit items. They remained as counters of perceptual unit items
throughout their first grade in school, so I regarded the period criterion as satisfied
for these three children.
The Incorporation Criterion
Piaget required that an earlier developmental stage be “incorporated” into a suc-
ceeding stage, a criterion that was adopted by Glasersfeld and Kelly. To discuss this
criterion for these three children, I appeal to their construction of collections of
perceptual unit items for which our use of Glasersfeld’s (1981) attentional moments
model provided a second-order model. But rather than say that collections of per-
ceptual unit items were incorporated in the perceptual counting scheme, I would
rather say that the perceptual counting scheme is based on children having already
constructed collections of perceptual unit items. Further, it is based on children’s
awareness of a plurality, or more than one, of such items, an awareness that serves
to establish a goal to make this quantitative property of the collection definite. That
is, being a counter of perceptual items means a lot more than knowing how to count.
As I have already commented, the goal of children counting collections of percep-
tual items is to find how many items are in the collections.
The Invariant Sequence Criterion
Glasersfeld and Kelly (1983) also required that stages appear in an invariant
sequence. It was very alarming that Brenda, Tarus, and James did not construct figu-
rative collections and count their elements to specify their plurality during their first
grade, as did Tyrone and Jason. So, I had to wait until the children’s second grade to
evaluate this criterion. As I explained above, the three children did re-present finger
patterns, but not being able to produce a figurative collection and count the elements
of the collection was a major constraint for the children throughout their first grade.
They could recognize perceptual patterns—finger patterns, spatial patterns, spatio-
motor patterns, and auditory patterns—as meaning of number words, but they did
not re-present such patterns and count the re-presented elements of the patterns
during their first grade, nor did they engage in monitoring counting to produce a
numerical pattern, as did Tyrone or his two counterparts in spite of my best attempts.
16
A major distinction between developmental stages and learning stages is that the former are the
result of spontaneous development whereas the latter are the result of interacting socially or other-
wise in teaching episodes as well.
The Figurative Stage
It wasn’t until the second grade that the three children made progress and con-
structed figurative collections. To illustrate this point, consider Tarus’ attempt to
solve the following situation at the end of January of his second grade. I presented
him with a cylindrical tube open at one end where a marble would just fit into the
tube. After Tarus put 11 marbles into the tube, I poured three marbles out of the tube
into his hand and asked him to find how many marbles were left in the tube. Tarus
buried his head in his arms and played with a marble and then said, “ten.” In expla-
nation, he said, “I count.” Based on Tarus burying his head in his arms and playing
with a marble, I inferred that he was aware of marbles in the tube (an awareness of
a figurative collection) and “ten” referred to marbles. But I also inferred that his
figurative counting scheme did not contain the numerical operations necessary to
find how many marbles were left. The internal constraints that Tarus experienced
because his counting scheme was not an operative scheme were striking throughout
the 2 years that I taught him (cf. Steffe & Cobb, 1988). Still, Tarus made progress
over the 2-year period in that his counting scheme became a figurative scheme dur-
ing that period. Tarus could count-on while he was in his third grade, but I have no
information on the two other children. The best scenario is that all three children
constructed their initial number sequences by that time. In any event, I inferred that
the invariant sequence criterion for stages was satisfied in the case of the perceptual
and figurative stages in the construction of the number sequence for these three
children.
Percent of First-Grade Children in the Perceptual Stage
Children who are counters of perceptual unit items at the beginning of their first
grade in school are among those children who will essentially fail mathematics all
the way through school because their schemes and operations are out of syn-
chrony with school mathematics curricula. It is not just a few children who enter
their first grade in school as counters of perceptual unit items. Based on my expe-
rience teaching the three children, on my experience in teacher education at UGA,
and on data that was supplied to me by Professor Bob Wright of Southern Cross
University, Australia, who started the Mathematical Recovery Program (Wright
et al., 1998; Wright et al., 2000) and who was associated with IRON as a doctoral
student, I estimated that 40% of entering first graders in the United States are
counters of perceptual unit items. Of this estimate, Professor Wright commented
that “I think that is a good estimate for the number in the perceptual stage or
lower, that is, the children who can’t yet count perceptual items. I think the per-
centage would be lower in Australia and New Zealand, say about 30%” (Personal
Communication, 2010).
30 L. P. Steffe
Further Criteria of Learning Stages
I conducted a 2-year teaching experiment starting in September of 1985 with six

third-grade children to investigate their construction of multiplicative schemes. My
immediate goal here is to recount the progress of two children who had constructed
only the initial number sequence. Zachary was the more advanced of the two chil-
dren, so I concentrated on him in my conceptual analysis (Steffe, 1992).17 In what
follows, I recount my efforts to explore criteria for learning stages that are not rel-
evant to Piaget’s developmental stages in the context of Zachary’s productive
attempts to solve tasks. If a productive attempt constituted a functional accommoda-
tion18 of Zachary’s initial number sequence, the results of the revised scheme (a
revision of the initial number sequence) would need to close the scheme, and the
change would need to be permanent. If a functional accommodation led to a meta-
morphic accommodation like Tyrone’s and Jason’s, it would be necessary to be able
to legitimately infer that a change in the assimilating structure (or operations) of the
initial number sequence occurred. So, of central concern was whether it could be
judged that a functional accommodation left the operations of the initial number
sequence invariant, which is a criterion for learning stages.
Modifications in the Initial Number Sequence
In an exploration of Zachary’s possible functional accommodations, I presented

him with a situation where four rows of blocks with three per row were hidden,
some more rows with three blocks per row were hidden with the four and told
Zachary that seven rows were now hidden. He was to find how many more rows
were hidden. This task was designed to explore if Zachary could take the three
blocks in a row as a composite unit and operate as if the rows were units of one. That
is, would Zachary simply count from “four” up to “seven” to find how many more
rows were hidden? Zachary initially formed the goal of finding how many blocks
rather than rows were hidden. After counting the blocks in the first two rows, “3, 6,”
he proceeded as follows. He sequentially put down his thumb, index finger, and
middle finger repeatedly as he whispered number words, “7, 8, 9; 10, 11, 12; 13, 14,
15; 16, 17, 18; 19, 20, 21; 22, 23, 24.”19 Apparently realizing that he didn’t know
when to stop counting, he started over, this time sweeping a finger over each of the
17
I had no information on Zachary while he was in his second grade. But based on how he used his
initial number sequence at the outset, I would say that he constructed it at some point in his sec-
ond grade.
18
A functional accommodation of a scheme occurs in the use of the scheme.
19
The semicolons represent breaks between completing three counting acts and starting three more
counting acts.
hidden rows and then sweeping it once more as if counting one more row. He then
stopped and said, “I can’t do that one.”
When counting, Zachary used his finger pattern for “three” by sequentially put-
ting down three fingers repeatedly. Apparently, his intention was to use his finger
pattern seven times. But he apparently didn’t know when to stop counting, which
indicates that he didn’t keep track of how many times he used it. Had he kept track,
I would have inferred that he took each finger pattern as a composite unit—as one
thing—and counted those composite units. That is, he would have monitored repeat-
edly using his finger pattern. His comment, “I can’t do that one,” confirms a lack of
monitoring activity and a lack of making composite units of three as he proceeded.
His behavior is consistent with the interpretation that the assimilating structure of
his initial number sequence was numerical composites rather than composite units
containing them. It is also consistent with the interpretation that he was aware of a
plurality20 of rows with a trio of blocks in each row prior to activity.
Because Zachary’s modification in how he used his initial number sequence
didn’t close the scheme, the modification in how he counted couldn’t be said to
constitute a functional accommodation. I interpret Zachary’s modification in how he
counted as producing a pseudo-empirical abstraction, which is any abstraction con-
cerning an activity that a child has produced but whose results cannot yet be obtained
without actually carrying out the activity (Steffe & Cobb, 1988). Corroboration that
he made a pseudo-empirical abstraction is based on how he later tried to find how
many piles of three blocks he could make from a pile of 12 blocks (Steffe, 1992).
He squeezed his right thumb, index finger, and middle finger together and placed
them on the table in front of him and said, “three.” He continued on in this way four
times until he reached “twelve.” He then looked at the nine fingers that he used and
said, “Nine times.” Upon saying “Nine times,” he apparently experienced a conflict
because he started over, but he was again unsuccessful.
The above modification that Zachary made in his initial number sequence did
lead to a functional accommodation in that I was able to infer that he constructed
experiential composite units in activity in the context of engaging in enactive units-
coordination. After Zachary made four rows of blocks with three per row, I covered
them and asked him how many blocks were in the rows. He looked straight ahead as
he sequentially put up his right thumb, index finger, and middle finger four times
and then said, “Twelve.” He only used one finger pattern for three and clearly kept
track of how many times he used it as he was counting to “Twelve” in modules of
three, which indicates that he took his finger pattern as one thing. Taking his finger
pattern for three as one thing in activity constitutes making experiential composite
units, and keeping track of how many finger patterns he made constitutes enactive
units-coordination of the composites four and three, or a units-coordination in activ-
ity. Zachary made still further progress, but there was never an occasion where I
could infer that he made a functional accommodation in his number sequence that I
could judge as metamorphic. So, he remained in the stage of the initial number
20
The term “plurality” means more than one unitary item.
32 L. P. Steffe
sequence for his entire third grade in school with the proviso that he progressed to
making experiential composite units in activity during the latter part of the
school year.
Modifications in the Tacitly Nested Number Sequence
The child Maya had constructed the tacitly nested number sequence at the outset of
the multiplication teaching experiment, which is characterized in part by the con-
struction of composite units (Ulrich, 2016a, p. 34). To illustrate Maya’s number
sequence, I analyzed her solution to a task where I asked her to place eight marbles
in a cup and seven in another. To find how many were in both cups, I inferred that
she double counted seven onto eight, which means that she counted “Eight. Nine is
one; ten is two, …, fifteen is seven. Fifteen.” She then placed three more marbles
with the seven upon my request and I asked her how many marbles were in the cups
now. She immediately said, “Eighteen” and explained, “Because there were eight
and seven, and you know that eight and seven is fifteen, and then you put the other
three, and there’s fifteen, 16, 17, 18.” Her solution to this task was quite beyond any
solution that I observed Zachary produce. To explain how she operated as she did, I
appealed to the recursive property of her number sequence, which means that she
could take the number sequence as its own input to create countable items, as illus-
trated by double counting (Steffe, 1992).
My Construction of Units-Coordinating Operations
As I was analyzing the teaching episodes after the teaching experiment with the
children was completed, the way in which Maya solved the following task had me
“staring at the wall” for the better part of one summer in an attempt to viably inter-
pret her solution. I recount it because it was the occasion for constructing units-
coordinating operations as constitutive multiplicative operations. I also recount it to
indicate the challenges of engaging in a conceptual analysis of the video records of
a teaching experiment that I have referred to as a retrospective conceptual analysis.
In the presentation of the task, I placed a red piece of construction paper in front of
Maya and several congruent rectangular blue pieces cut so that six blue pieces fit on
the red piece. After she said, “Six,” I removed three of them and covered one of
them with two orange pieces that fit exactly and asked her how many orange pieces
go on the blue piece. After looking straight ahead, she subvocally uttered number
words and said, “Twelve.” In explanation, she tapped the table twice with each of
six fingers, “1, 2; 3, 4; 5, 6; 7, 8; 9, 10; 11, 12.” Tapping each of her fingers twice
indicated that she was counting the orange pieces on each red piece. My issue,
which led me to “staring at the wall,” was whether the solution was based on simply
re-presenting placing an orange piece on each blue piece or whether she inserted a
unit of two into each unit of six, which would be an operative rather than simply a
figurative solution. How she organized her explanation into modules of two count-
ing acts eventually was the determining element that eliminated the purely figura-
tive interpretation because she was definitely counting six twos. So, the interpretation
that she coordinated inserting a unit of two into each unit of a unit of six either
before or in operating was the more viable interpretation. As indicated by other
tasks that she solved, Maya was explicitly aware of units of two or three as well as
the items they contained.
I interpreted Maya inserting a unit of two into each unit of her composite unit of
six units as an accommodation. The results of the scheme closed the scheme, and
she was observed using the scheme in other cases that, from my perspective, were
multiplicative situations. This scheme was quite distinct from regarding multiplica-
tion in elementary school as repeated addition, which is essentially what Zachary
had constructed when he found how many blocks were in four covered rows with
three blocks in each row that I explained above. In that case, he made an enactive
units-coordination (or a units-coordination in activity) of the four rows and the three
blocks in each row. In Maya’s case, she made a units-coordination prior to initiating
units-coordinating activity. As the issue was whether Maya’s units-coordination
constituted a metamorphic accommodation on a par with Tyrone’s and Jason’s met-
amorphic accommodation or whether it was a functional accommodation that pro-
duced a scheme, her units-coordinating scheme that was at the same level of
reflective abstraction as her tacitly nested number sequence. To make that decision,
I looked to other tasks that involved units-coordination to solve.
Constraints in Provoking Reflective Abstraction
I presented a task to Maya that children who have constructed the explicitly nested
number sequence could solve in an attempt to analyze if Maya’s units-coordinating
scheme was a stage beyond her tacitly nested number sequence. Had Maya sponta-
neously solved the tasks, I would have inferred that her construction of her units-
coordinating scheme involved a reflective abstraction. Although Maya solved the
tasks upon my urging, her solution attempts were not spontaneously initiated, which
prohibited inferring that her accommodation was produced by a reflective abstrac-
tion (Steffe, 1992). In spite of my best attempts, I should be clear that I did not
experience Zachary nor Maya solve a task I presented to them that could be ana-
lyzed as entailing either a reflective abstraction or the results of a reflective abstrac-
tion throughout their third grade in school. They remained in their respective stages
throughout that school year. I did document solutions of tasks that involved impor-
tant functional accommodations that I have referred to as “lateral learning” rather
than “vertical learning” (Steffe, 1991a). The question concerning whether reflective
abstraction or vertical learning can be provoked in children like Zachary or Maya in
their third grade remains a crucial question because both children were out of syn-
chrony with what is expected of a third-grade child in the mathematics curricula.
Modifications in the Explicitly Nested Number Sequence
Numerical part-to-whole reasoning is the identifying characteristic of the explicitly

nested number sequence (Steffe, 1992, p. 290). That is, children who have con-
structed the explicitly nested number sequence can disembed a composite unit from
34 L. P. Steffe
a given composite unit and use the disembedded unit in further reasoning.21 To illus-
trate, in a “missing rows” situation, Johanna, a child in the multiplication teaching
experiment, made four rows of blocks with three per row. I then covered some more
rows of three and told her that there were now seven rows of three and asked her to
find how many rows were hidden. She said, “three” almost immediately, treating the
rows as if they were singleton units in contrast to Maya’s attempted solution of a
similar task (Steffe, 1992).
To again compare Johanna’s and Maya’s task solutions, I presented the same task
to Johanna that I spent most of one summer “staring at the wall” when analyzing it
in Maya’s case. In solution, Johanna said “twelve” after about 15 seconds and
explained, “Well, six plus six is twelve, and each two blocks fit on one big block,
and that makes twelve” (Steffe, 1992, p. 292). I inferred that Johanna was reflective
aware of six units of two and decomposed them into two units of six, whereas Maya
took six units of one as a given and then substituted a unit of two for each unit of
one. Based on Johanna’s spontaneous reasoning, I inferred that her units-coordinating
scheme constituted a functional accommodation in her explicitly nested number
sequence, an inference that was corroborated throughout the teaching experiment.
Johanna continued making progress and used her units-coordinating scheme in
the construction of three levels of units, or a unit of units of units. For example, after
hiding five rows of blocks with four per row that she had made, I asked her how
many blocks were in the first three rows. After about 25 seconds, when she was in
deep concentration, she said, “Twelve.” I then asked her how many blocks were in
all five rows after she found that eight blocks were in the remaining two rows. After
sitting silently for about 15 seconds, she replied “twenty” and explained, “Because
I added up. Twelve plus four is 16, and 16 plus four is 20!” (Steffe, 1992, p. 292). I
inferred that Johanna united the first three units of four into another composite unit
containing them whose numerosity she found as “twelve.” My inference that she
operated in this way is also based on that it took her 25 seconds to answer “twelve”
(Steffe, 1992). My inference that making units of units of units was an accommoda-
tion in her units-coordinating scheme would not have been viable unless there were
other composite unit tasks where she exhibited similar productive mathematical
activity in solution. Corroboration is an essential element of the process of con-
structing second-order models, particularly when inferring the occurrence of
accommodation.
Corroboration is but one aspect of the teacher/researcher constructing interior-
ized experiential models of the children’s mathematics that the teacher/researcher
uses to predict how a particular child will operate in other tasks. That is, the teacher/
researcher becomes the child and attempts to think as if he or she is the child (Steffe
& Thompson, 2000). I have already commented that such experiential models are
critical in the retrospective second-order analysis of the video-recorded material at
the end of the teaching episodes in constructing explanations of children’s mathe-
matics. In constructing second-order models like the number sequences, it is
21
An analogy is conceiving of a subset of a set.
important to stress that these models are explanations the researcher makes that are
grounded in the activity of children that the teacher/researchers are involved in pro-
voking through experimental teaching. So, the researcher is inextricably involved in
creating the observations for which the researcher creates explanations, which is
compatible with thinking in second-order cybernetics.22
The Fractions Teaching Experiment
After the completion of the multiplication teaching experiment, my interest turned

to exploring children’s construction of fractions. My hypothesis was that fraction
knowing can emerge as accommodations in number sequences (Steffe & Olive,
2010). At the time, there was a general belief that whole number knowing interferes
with the learning of fractions, a belief that is still pervasive. It translates in school
mathematics curricula as fractions being presented independently of work with
whole numbers. To test the reorganization hypothesis, John Olive and I, along with
Ron Tzur and Barry Biddlecomb, who were then graduate students, started a 3-year
teaching experiment that we referred to as the fractions teaching experiment in
199123 with third-grade children. This project continued on until 2002, and the book
Children’s Fractional Knowledge was printed in 2010 (Steffe & Olive, 2010).24
The Equipartitioning Scheme: A Functional Accommodation
The reorganization hypothesis was unexpectedly partially confirmed by two chil-

dren, Jason and Patricia, both of whom had constructed the explicitly nested number
sequence. The children partitioned a computer-generated drawing of a segment that
we referred to as a “stick” in the experiment (Steffe & Olive, 2010). To make a share
of a stick for one of four people, Jason used his concept, four, as a partitioning tem-
plate to mark off one of four equal parts. He then reassembled the parts into a unit
of four equal parts of the stick and compared the four parts with the original stick to
find if four of them together were the same length as the original. In doing so, he
constructed a partitioning scheme that I called the equipartitioning scheme that sat-
isfied six of the seven criteria for Piaget’s operational subdivision25 (Piaget et al.,
1960). The criterion it didn’t satisfy was the criterion that the units of the subdivi-
sion can be taken as units to be subdivided further. Patricia, too, demonstrated that
22
See Steier (1995) for an introduction to second-order cybernetics.
23
Dr. Heide Wiegel also worked on the teaching experiment organizing, cataloging, and analyz-
ing data.
24
See Hackenberg et al. (2016) for an interpretation for teachers.
25
Partitioning is fundamental in constructing fraction schemes.
36 L. P. Steffe
she could use her number concepts as partitioning templates in the establishment of
connected numbers.
Pre-fractional Children
Partitioning continuous units using numerical concepts like Johanna’s was found to
be one of the key elements in corroboration of the reorganization hypothesis.
However, for children like Zachary, who are in the stage of the initial number
sequence, fragmenting a continuous unit proved to be problematic due to a general
lack of coordinating the number and size of the parts with exhausting the whole,
which meant that they are yet pre-fractional (Biddlecomb, 2002; Steffe & Olive,
2010). These limitations in fragmenting were in part removed by children like
Maya, who had constructed the tacitly nested number sequence because number
words now symbolized composite units prior to action, which enables them to parti-
tion a stick into four equal parts should that be their goal. What would be still lack-
ing is an understanding that any of the unit parts could be used to reconstitute a
continuous but segmented unit equivalent to the original unit—that is, the unit parts
are not iterable units like they were for Jason because the units of the composite unit
used in fragmenting are not iterable (Steffe & Olive, 2010; Ulrich, 2016b). In fact,
after working with a child who had constructed the tacitly nested number sequence
throughout the child’s fourth grade, the child did not construct any fraction schemes
(Steffe & Olive, 2010).
The Partitive Fraction Scheme
A burning question of the fraction teaching experiment was whether children who
had constructed the explicitly nested number sequence could also construct the fun-
damental elements of fraction knowing. Jason and Patricia did use their operative
equipartitioning scheme to construct what Tzur (1999) called partitive fractions, or
“proper” fractions of the fractional whole (cf. Steffe & Olive, 2010, pp. 323) by
means of a functional accommodation. At the time it was startling that the partitive
fraction scheme was not sufficient to construct fractions beyond the fractional
whole. Other than being restricted to making proper fractions, children who had
constructed the partitive fraction scheme were not able to, say, given a segment that
is said to be three-fifths of some other segment, construct the other segment by par-
titioning the given segment into three parts and use those parts in making the other
segment (Steffe & Olive, 2010).
The Iterative Fraction Scheme
It was a major finding that three levels of units were necessary to construct fractions
greater than the fractional whole. For example, after Patricia constructed three lev-
els of units, she explained why a stick was eight-sevenths in length was because the
eight parts of the stick were the same as the little pieces in the seventh stick (the
fractional whole), so its eight-sevenths (Steffe & Olive, 2010). I interpreted her
comments as indicating that one-seventh was freed from the fractional whole of
which it was a part and was constituted as a unit fraction without reference to the
fractional whole. That is, “one-seventh” referred to any one of the seven parts and,
in that sense, it was a hypothetical part of seven-sevenths that could be iterated eight
times to produce eight-sevenths. So, the eight units of the eight-sevenths were con-
ceived of as one-seventh eight times.
To illustrate, using the available computer operations, Patricia used the calcula-
tor to multiply a 14/99 stick by ten and explained why the stick she made would be
a 140/99 stick was because “you always have the same little stick you started off
with” (Steffe & Olive, 2010, p. 334). That is, 140/99 was constituted as a fractional
number because it was 140 times the fractional unit that was used in producing it,
analogous to one iterated 140 times to produce the number 140. This was a critical
conception of fractions in that the entirety of the composite unit comprised by
140/99 doesn’t need to be produced to give meaning to the fraction, so the numeral
can be a symbol for those operations. In fact, Patricia and her cohort produced a
fraction number sequence <1/12, 2/12, …, 12/12, 13/12, 14/12, 15/12, 16/12,
17/12, …> that Patricia thought would continue on to infinity quite similar to how
she could use her unit of one to produce indefinitely large whole numbers.
We referred to Patricia’s fraction scheme as the iterative fraction scheme. It was
one stage beyond her partitive fraction scheme, a claim that was based not only on
the fractional numbers that she produced but also on the level of units on which it
was based. We never observed what could be considered a metamorphic accommo-
dation that produced the three levels of units. But I believe that the third level was
constructed by means of a metamorphic accommodation in the context of the teach-
ing experiment. In Jason’s case, he returned from summer vacation able to operate
in ways that were beyond him when he left for vacation after his fourth grade. When
he returned from summer vacation, he reasoned in such a way that implied that he
had constructed three levels of units (Steffe & Olive, 2010). There were no observa-
tions that could be used to operationalize reflective abstraction during his fourth
grade, so if he made a reflective abstraction, it occurred over the summer months.
38 L. P. Steffe
Final Comments
To provide perspective on the learning stages in the school population, I present my

estimates of the percentage of children in the first, second, third, fifth, and sixth
grades who have constructed the various counting schemes. For the 40% of the
school population who enter the first grade in the perceptual stage, I have argued
that it is reasonable to expect that the majority of them will construct the initial
number sequence during their third grade. Wright estimated that from 5 to 8% might
not be counting on by the third grade (Personal Communication). From that point
on, Zachary, notwithstanding, my expectation is that the majority of them will
remain in that stage throughout the fifth grade. The relative percentages are not
certain, but my best estimate is that approximately 30% of the children entering the
sixth grade will be in the stage of the initial number sequence. Wright’s estimate
was “that about thirty percent of kids entering the sixth grade in the US will only be
able to count on” (Personal Communication). And those who are in that stage will
remain there until their eighth grade.26
I also estimate that at most one child in ten in the first grade will have constructed
the explicitly nested number sequence. Based on this estimate, no more than
approximately 50% of children entering the first grade will have constructed the
initial number sequence sometime during their first grade. The 10% estimate is
based on Piaget et al.’s (1960) finding that only one child in ten of those from 6 to
7 years of age (first graders) attain the iterable length unit, which I take as analogous
to the explicitly nested number sequence. Because of the homogeneity of the Swiss
population, I consider the 10% as an upper bound on the percent of first-grade chil-
dren who have constructed or will construct the explicitly nested number sequence.
In the second grade, my estimate is that approximately 40% of the children are
in the figurative stage because these are the children who are in the perceptual stage
in the first grade. Piaget et al. (1960) estimate that one-half of Swiss children from
7 to 7 years 6 months (second graders) attain the iterable length unit. Because of the
homogeneity of the Swiss population, I consider it implausible that 50% of second
graders have constructed the explicitly nested number sequence and estimate that no
more than 30% do so. On that basis, 30% of second graders remain in the stage of
the initial number sequence. Whatever these two percentages may be, approxi-
mately 70% of the second-grade population is yet to construct the explicitly nested
number sequence, which is the number sequence I consider is assumed in normal
second-grade mathematics classrooms given the curriculum. And 40% are yet pre-
numerical (figurative), which means that they are seriously at risk in school mathe-
matics not only in their first and second grades but also throughout their mathematics
education (Table 2.1).
26
In a 3-year teaching experiment with sixth, seventh, and eighth graders, three children who had
constructed the initial number sequence remained in that stage for the first 2 years of the experi-
ment (Hackenberg, 2005; Tillema, 2007).
Table 2.1 Estimated percent of children at each level of units by grade

Initial/
Grade/ Perceptual/ tacitly Explicitly nested Three levels of Iterative
stage figurative nested (%) (%) units (%) fraction
First 40% P 50 10 –
Second 40% F 30 30 –
Third – 60 40 –
Fifth – 35 40 25 No estimate
Sixth – 30 30 40 No estimate
Throughout the third grade, I assume that a majority of the children in the figura-
tive stage will construct the initial number sequence (approximately 35%). Of the
remaining 65%, my best estimate is that 40% of the total population will construct
the explicitly nested number sequence, and the remaining 25% will remain in the
stages of the initial or the tacitly nested number sequence.27 Based on the fraction
teaching experiment, children who have constructed the explicitly nested number
sequence in the third grade can construct three levels of units by the time they enter
their fifth grade under the influence of intensive teaching. So, optimistically, I expect
that 25% of the population will have constructed three levels of units by the time
they enter the fifth grade, and 40% will have constructed the explicitly nested num-
ber sequence.
I estimated that 25 and 40% of the fifth- and sixth-grade populations have con-
structed three levels of units. Data presented by Norton and Wilkins (2009) in
Table 2.2 confirm my estimates in broad outline. These are the children who should
have constructed the iterative fraction scheme and fractional numbers. However,
Norton and Wilkins found that only 14% of the fifth graders and 18% of the sixth
graders have constructed the iterative fraction scheme and that these percentages do
not improve for the seventh and eighth grades. This is quite concerning because the
children in our fractions teaching experiment who had constructed three levels of
units did construct that scheme and its use in producing fractional numbers, which
confirms that researchers should be involved in teaching the children in their teach-
ing experiments. Even more surprising is the finding by these two authors that fewer
than 20% of students throughout the middle school grades can produce a fractional
whole given a proper fraction of the whole (Wilkins & Norton, 2018). Clearly, the
presence of learning stages in children’s constructive activity presents major chal-
lenges for constructivist researchers in mathematics education in a way similar to
the challenges that spontaneous development presented for curriculum developers
during the era of modern mathematics.
27
Zachary and Maya were among this 25%.
40 L. P. Steffe
Table 2.2 Norton and Grade/level Three levels of units (%) Iterative fraction (%)
Wilkins data
Fifth 34 14
Sixth 35 18
Seventh 13
Eighth 18
References
Allendoerfer, C. B., & Oakley, C. O. (1959). Fundamentals of freshman mathematics. McGraw

Hill Book Company, Inc.
Biddlecomb, B. (2002). Numerical knowledge as enabling and constraining fraction knowledge:
An example of the reorganization hypothesis. Journal of Mathematical Behavior, 21, 167–190.
Bridgman, P. W. (1927). The logic of modern physics. Macmillan.
Duckworth, E. (1964). Piaget rediscovered. In R. E. Ripple & V. N. Rockcastle (Eds.), Piaget
rediscovered: A report of the conference on cognitive studies and curriculum development
(pp. 1–5). School of Education. Cornell University.
Hackenberg, A. J. (2005). Construction of algebraic reasoning and mathematical caring relations.
Unpublished doctoral dissertation, University of Georgia.
Hackenberg, A. J., Norton, A., & Wright, R. J. (2016). Developing fractions knowledge. Sage.
Inhelder, B., & Piaget, J. (1958). The growth of logical thinking from childhood to adolescence.
Routledge & Kegan Paul.
Kilpatrick, J. (1964). Cognitive theory and the SMSG program. In R. E. Ripple & V. N. Rockcastle
(Eds.), Piaget rediscovered: A report of the conference on cognitive studies and curriculum
development (pp. 128–133). School of Education. Cornell University.
Lakatos, I. (1970). Falsification and the methodology of scientific research programs. In I. Lakatos
& A. Musgrave (Eds.), Criticism and the growth of knowledge (pp. 91–195). Cambridge
University Press.
Letvin, J. Y., Maturana, H. R., McCulloch, W. S., & Pitts, W. H. (1959). What the frog’s eye tells
the frog’s brain. Proceedings of the Institute of Radio Engineers, 47, 1949–1951. Reprinted in
W. S. McCulloch (Ed.) (1965). In Embodiments of mind. MIT Press.
Lovell, K. (1971). Proportionality and probability. In M. Rosskopf, L. P. Steffe, & S. Taback
(Eds.), Piagetian cognitive development and mathematical education (pp. 136–148). National
Council of Teachers of Mathematics.
Norton, A., & Wilkins, J. L. M. (2009). A quantitative analysis of children’s splitting operations
and fraction schemes. Journal of Mathematical Behavior, 28(2/3), 150–161.
Piaget, J. (1964). The Piaget papers. In R. E. Ripple & V. N. Rockcastle (Eds.), Piaget redis-
covered: Report of a conference on cognitive studies and curriculum development. Cornell
University Press.
Piaget, J. (1966). General conclusions. In E. W. Beth & J. Piaget (Eds.), Mathematical epistemol-
ogy and psychology (pp. 305–312). Dordrecht.
Piaget, J. (1968). Six psychological studies. Vintage Books.
Piaget, J. (1970). Genetic epistemology. Columbia University Press.
Piaget, J. (1975). L’equilibration des structures cognitives. Presses Universitaires de France.
Piaget, J. (1980). The psychogenesis of knowledge and its epistemological significance. In
M. Piattelli-Palmarini (Ed.), Language and learning: The debate between Jean Piaget and
Noam Chomsky (pp. 23–34). Harvard University Press.
Piaget, J., & collaborators. (1977). Recherches Sur l’abstraction reflechissante (Vol. I & II).
Presses Universitaires de France.
Piaget, J., & Inhelder, B. (1967). The child’s conception of space. W. W. Norton & Company, Inc
(First published 1956).
Piaget, J., & Inhelder, B. (1969). The psychology of the child. Basic Books.
Piaget, J., & Szeminska, A. (1952). The child’s conceptions of number. Routledge & Kegan Paul.
Piaget, J., Inhelder, B., & Szeminska, A. (1960). The child’s conception of geometry. Basic Books.
Pólya, G. (1945). How to solve it (2nd ed.). Princeton University Press (1957).
Pólya, G. (1981). Mathematical discovery. (Volumes 1 and 2, Combined paperback edition). Wiley.
Ripple, R. E., & Rockcastle, V. N. (Eds.). (1964). Piaget rediscovered: A report of the conference on
cognitive studies and curriculum development. School of education. Cornell University Press.
School Mathematics Study Group. (1965). Mathematics for the elementary school: Grade 6, Parts
I and II. A. C. Vroman, Inc.
Sinclair, H. (1971). Piaget’s theory of development: The main stages. In M. Rosskopf, L. P. Steffe,
& S. Taback (Eds.), Piagetian cognitive development and mathematical education (pp. 1–11).
National Council of Teachers of Mathematics.
Stanley, J. C., & Campbell, D. T. (1963). Experimental and quasi-experimental designs for
research on teaching. In N. L. Gage (Ed.), Handbook for research on teaching (pp. 171–246).
Rand McNally & Company.
Steffe, L. P. (1966). The performance of first grade children in four levels of conservation of
numerousness and three IQ groups when solving arithmetic addition problems. Technical
Report No. 14, Research and Development Center for Cognitive Learning and Re-Education.
University of Wisconsin.
Steffe, L. P. (1983). The teaching experiment in a constructivist research program. In M. Zweng,
T. Green, J. Kilpatrick, H. Pollack, & M. Suydam (Eds.), Proceedings of the fourth interna-
tional congress on mathematical education (pp. 469–471). Birkhauser.
Steffe, L. P. (1988a). Children's construction of number sequences and multiplying schemes.
In J. Hiebert & M. Behr (Eds.), Number concepts and operations in the middle grades
(pp. 119–140). Lawrence Erlbaum Associates.
Steffe, L. P. (1988b). Modifications of the counting scheme. In I. L. P. Steffe & P. Cobb (Eds.),
Construction of arithmetical meanings and strategies (pp. 284–322). Springer Verlag.
Steffe, L. P. (Ed.). (1991a). Epistemological foundations of mathematical experience (pp. 45–67).
Springer. http://www.vonglasersfeld.com/130
Steffe, L. P. (1991b). The constructivist teaching experiment: Illustrations and implications. In
E. von Glasersfeld (Ed.), Radical constructivism in mathematics education (pp. 177–194).
D. Reidel Company.
Steffe, L. P. (1992). Schemes of action and operation involving composite units. Learning and
Individual Differences, 4(3), 259–309.
Steffe, L. P. (1994). Children’s construction of meanings for arithmetical words. In D. Tirosh (Ed.),
Implicit and explicit knowledge: An educational approach (pp. 131–169). Ablex.
Steffe, L. P. (1996). Stages in the construction of the number sequence. In J. Bideaud, C. Meljac,
& J.-P. Fischer (Eds.), Pathways to number (pp. 83–98). Routledge.
Steffe, L. P. (2011). The honor of working with Ernst von Glasersfeld: Partial recollections.
Constructivist Foundations, 6(2), 172–176.
Steffe, L. P. (2013). Establishing mathematics education as an academic field. Journal for Research
in Mathematics Education, 44(2), 353–370.
Steffe, L. P., & Cobb, P. (1983). The constructivist researcher as teacher and model builder. Journal
for Research in Mathematics Education, 14(2), 83–94.
Springer Verlag.
Steffe, L. P., & Kieren, T. (1994). Radical constructivism and mathematics education. Journal for
Research in Mathematics Education, 25(6), 711–733.
Steffe, L. P., & Olive, J. (2010). Children’s fractional knowledge. Springer.
Steffe, L. P., & Parr, R. B. (1968). The development of the concepts of ratio and fraction in the
fourth, fifth, and sixth years of the elementary school. Technical Report No. 49, Research and
Development Center for Cognitive Learning and Re-Education, C-03, OE 5-10-154. University
of Wisconsin.
42 L. P. Steffe
Steffe, L. P., & Thompson, P. (2000). Teaching experiment methodology: Underlying principles
and essential elements. In A. Kelly & R. Lesh (Eds.), Research design in mathematics and sci-
ence education (pp. 267–307). Lawrence Erlbaum.
Steffe, L. P., & Ulrich, C. (2013). The constructivist teaching experiment. In S. Lerman (Ed.),
Encyclopedia of mathematics education (pp. 102–109). Springer.
Steffe, L. P., Hirstein, J., & Spikes, C. (1976a). Quantitative comparison and class inclusion as
readiness variables for learning first grade arithmetic content. Technical Report No. 9. ERIC
Document Reproduction Service No. ED144808. Project for Mathematical Development of
Children.
Steffe, L. P., Hirstein, J., & Spikes, C. (1976b). Quantitative comparison and class inclusion as
readiness variables for learning first grade arithmetic content. Technical Report No. 9, Project
for Mathematical Development of Children, (ERIC Document Reproduction Service No.
ED144808).
Philosophy, theory, and application. Praeger.
Steier, F. (1995). From universing to conversing: An ecological constructionist approach to learn-
ing and multiple description. In L. P. Steffe & J. Gale (Eds.), Constructivism in education
Stolzenberg, G. (1984). Can an inquiry into the foundations of mathematics tell us anything inter-
esting about mind? In P. Watzlawick (Ed.), The invented reality (pp. 257–308). W.W. Norton
& Company.
Thompson, P. W. (2000). Radical constructivism: Reflections and directions. In L. P. Steffe &
P. W. Thompson (Eds.), Radical constructivism in action: Building on the pioneering work of
Ernst von Glasersfeld (pp. 412–448). Falmer Press.
Thompson, P. W. (2008). Conceptual analysis of mathematical ideas: Some spadework at the
foundation of mathematics education. In O. Figueras, J. L. Cortina, S. Alatorre, T. Rojano,
& A. Sèpulveda (Eds.), Proceedings of the annual meeting of the International Group for the
Psychology of mathematics education (Vol. 1, pp. 45–64). PME.
Thompson, P. W. (2013). In the absence of meaning. In K. Leatham (Ed.), Vital directions for
research in mathematics education (pp. 57–93). Springer.
Thompson, P. W., & Saldanha, L. (2003). Fractions and multiplicative reasoning. In J. Kilpatrick,
G. Martin, & D. Schifter (Eds.), Research companion to the principles and standards for school
mathematics (pp. 95–114). National Council of Teachers of Mathematics.
Thompson, P. W., & Thompson, A. G. (1994). Talking about rates conceptually, Part I: A teacher’s
struggle. Journal for Research in Mathematics Education, 25(3), 279–303.
Tillema, E. (2007). Students’ construction of algebraic symbol systems. Unpublished doctoral dis-
sertation, The University of Georgia.
Tzur, R. (1999). An integrated study of children’s construction of improper fractions and the
teacher’s role in promoting that learning. Journal for Research in Mathematics Education, 30,
390–416.
Ulrich, C. (2015). Stages in constructing and coordinating units additively and multiplicatively
(part 1). For the Learning of Mathematics, 35(3), 2–7.
Ulrich, C. (2016a). Stages in constructing and coordinating units additively and multiplicatively
Ulrich, C. (2016b). The tacitly nested number sequence in sixth grade: The case of Adam. The
Journal of Mathematical Behavior, 43, 1–19.
Van Engen, H. (1949a). An analysis of meaning in arithmetic. I. The Elementary School Journal,
49(6), 321–329.
Van Engen, H. (1949b). An analysis of meaning in arithmetic. II. The Elementary School Journal,
49(7), 395–400.
Van Engen, H. (1971). Epistemology, research, and instruction. In M. F. Rosskopf, L. P. Steffe,
& S. Taback (Eds.), Piagetian cognitive-development research and mathematical education
(pp. 34–52). National Council of Teachers of Mathematics.
von Glasersfeld, E. (1974). Piaget and the radical constructivist epistemology. In C. Smock &
E. von Glasersfeld (Eds.), Epistemology and education: The implications of radical construc-
tivism for knowledge acquisition (pp. 1–26). University of Georgia.
von Glasersfeld, E. (1980). The concept of equilibration in a constructivist theory of knowledge.
In F. Benseler, P. M. Hejl, & W. K. Kock (Eds.), Autopoisis, communication, and society
(pp. 75–85). Campus Verlag.
von Glasersfeld, E. (1981). An attentional model for the conceptual construction of units and num-
ber. Journal for Research in Mathematics Education, 12, 83–94.
ber. In L. P. Steffe, E. von Glasersfeld, J. Richards, & P. Cobb (Eds.), Children’s counting
types: Philosophy, theory, and application (pp. 124–136). Praeger.
von Glasersfeld, E. (1987). The construction of knowledge: Contributions to conceptual seman-
tics. Intersystems Publications.
von Glasersfeld, E. (1991). Abstraction, re-presentation, and reflection: An interpretation of expe-
rience and of Piaget’s approach. In L. P. Steffe (Ed.), Epistemological foundations of math-
ematical experience (pp. 45–67). Springer. http://www.vonglasersfeld.com/130
von Glasersfeld, E. (1995). Radical constructivism: A way of knowing and learning. Falmer Press.
von Glasersfeld, E. (1998). Scheme theory as a key to the learning paradox. Paper presented at the
15th Advanced Course, Archives Jean Piaget.
von Glasersfeld, E. (2005). Thirty years constructivism. Constructivist Foundations, 1(1), 9–12.
http://constructivist.info/1/1/009
von Glasersfeld, E., & Kelly, M. (1983). On the concepts of period, phase, stage, and level. Human
Development, 25, 152–160.
von Glasersfeld, E., & Richards, J. (1983). The creation of units as a prerequisite for number:
A philosophical review. In L. P. Steffe, E. von Glasersfeld, J. Richards, & P. Cobb (Eds.),
Children’s counting types: Philosophy, theory, and application (pp. 1–20). Praeger.
Wertheimer, M. (1945). Productive thinking. Harper.
Wilkins, J. L. M., & Norton, A. (2018). Learning progression toward a measurement concept of
fractions. Journal of Stem Education, 5, 27.
Wright, R. J., Stewart, R., Stafford, A., & Cain, R. (1998). In K. Norwood & L. Stiff (Eds.),
Assessing and documenting student knowledge and progress in early mathematics
(pp. 211–216). Proceedings of the Twentieth Annual Meeting of the North American Chapter
of the International Group for the Psychology of Mathematics Education.
Wright, R. J., Martland, J., & Stafford, A. K. (2000). Early numeracy: Assessment for teaching and
intervention. Paul Chapman Publishing.
Part II
Key Constructs from Genetic
Epistemology Being Used in Ongoing
Mathematics Education Research
Chapter 3
Schemes and Scheme Theory: Core
Explanatory Constructs for Studying
Mathematical Learning
Erik S. Tillema and Andrew M. Gatza
The notion of scheme has given and is still giving rise to different interpretations. (Inhelder
& de Caprona, 1992, p. 41)
As part of writing this chapter, we have had the enjoyable experience of re-
reading many of Glasersfeld’s articles related to schemes. Re-reading Glasersfeld’s
articles gave us an appreciation anew for the contributions he made in interpreting
Piaget’s construct of scheme. In his later writings, Glasersfeld (e.g., 2001) came to
call the network of ideas—schemes, adaptation, viability, assimilation, accommo-
dation, perturbation, and equilibration—“scheme theory.” He points out that this
network of ideas was not laid out for readers of Piaget in a single place as a theory.
Rather Piaget created and refined these ideas over a significant period of time
(Glasersfeld, 1997). One contribution, then, of Glasersfeld’s work is a coherent
interpretation of Piaget’s writing, including the ideas that make up scheme theory.
A second contribution of his work is an explicit articulation of an epistemological
position, radical constructivism, and its role in scheme theory.
As we highlight in this chapter, Glasersfeld’s (1995) epistemological position
has a significant impact on the way that he interprets Piaget’s constructs. Therefore,
we begin this chapter with some epistemological considerations related to interpret-
ing scheme theory. We see these epistemological considerations as particularly
helpful to interpreting how Piaget developed and used the construct of scheme in his
own work and how these constructs have been subsequently taken up in mathemat-
ics education research. To this end, we provide a brief history of the development of
schemes in Piaget’s work with discussion of how mathematics educators have
E. S. Tillema (*)
e-mail: etillema@iu.edu
A. M. Gatza
Department of Mathematical Sciences, Ball State University, Muncie, IN, USA

https://doi.org/10.1007/978-3-031-47386-9_3
48 E. S. Tillema and A. M. Gatza
interpreted anew a few key ideas from Piaget for their work in mathematics educa-
tion (e.g., Steffe et al., 1983; Thompson, 1985). Once these initial pieces are in
place, throughout the rest of the chapter, we focus on illustrating constructs related
to scheme theory relative to what we call a multiplicative pairing scheme (MPS), a
scheme students can construct in the context of solving combinatorics problems.
Our discussion of an MPS begins with illustrations of two definitions of the term
scheme, Glasersfeld’s (1995) and Steffe’s (2010a) definitions. These initial exam-
ples are intended to show “schemes in action” and to highlight important features
related to using the respective definitions of scheme. We then situate our discussion
of schemes within a framework for studying learning, introducing and illustrating
the constructs of assimilation, perturbation, and functional accommodation. We
conclude the chapter with a discussion of an MPS relative to stages of units coordi-
nation and levels within a stage of units coordination where we see stages and levels
as a way to situate schemes within a broader framework. We have chosen to discuss
the same scheme throughout the chapter to help the reader focus on the different
constructs related to scheme theory that we develop in the chapter. That is, we
intend for the discussion to highlight issues that are of importance in general to
using schemes and scheme theory in mathematics education research.
Schemes are living constructs, and as such, researchers continue to make inter-
pretations of key epistemological issues related to scheme theory (e.g., Steffe &
Thompson, 2000a; Thompson, 2008), including interpretations of the term scheme
itself (e.g., Steffe, 2010a; Thompson, 2013; Thompson et al., 2014). We make this
point up front to highlight that Inhelder and de Caprona’s (1992, p. 41) quote is
relevant to this chapter in two ways. The first is that what we are presenting is an
interpretation of schemes and scheme theory situated within a radical constructivist
epistemology. The second is that there is continued space for discussion and devel-
opment of scheme theory in mathematics education research. Thus, we hope this
chapter both informs readers about how schemes and scheme theory have evolved
and inspires them to contribute to the continuing discussion and development of
these constructs—constructs that we see as still possessing significant power for
studying mathematical learning.
rief Overview of Glasersfeld’s Radical

B
Constructivist Epistemology
Throughout Piaget’s work, he insisted that knowledge always requires an

active agent:
To know an object, to know an event, is not simply to look at it and make a mental copy or
image of it. To know an object is to act on it. To know is to modify, to transform the object,
to understand the process of this transformation, and as a consequence, to understand the
way the object is constructed. (Piaget, 1964, p. 176)
This orientation to knowledge broke with some of the foundational relationships

between knowledge and reality as expressed in Western philosophy (Glasersfeld,
3 Schemes and Scheme Theory: Core Explanatory Constructs for Studying… 49
1974). That is, Western philosophers often position knowledge as true when it rep-
resents “a state or feature of an experiencer-independent world” (Glasersfeld, 1990,
p. 127) where the “experiencer-independent world” is considered to be ontological
reality. Thus, in Western philosophy, truth is a central way of relating knowledge to
reality—something is true if it corresponds to a reality that is independent of a
knower (Glasersfeld, 1983).
Piaget’s (1970) genetic epistemology breaks with this relationship precisely
because an active agent is required in the production of all knowledge. Put another
way, a person only has access to their own ways of perceiving and conceiving and
thus has no way to step outside of themselves to determine whether their knowledge
matches an ontological reality (Piaget, 1970, p. 15; Glasersfeld, 1983). For this
reason, Glasersfeld (1990) proposed speaking of a person’s experiential reality
rather than speaking about a person coming to know an ontological reality, where a
person’s experiential reality is the reality they construct in the context of interacting
in their physical and social world.
These considerations led Glasersfeld (1980a, 1990) to replace the notion of truth
as the fundamental relation between knowledge and reality with the notion of via-
bility—a person’s knowledge is viable within their experiential reality. Within the
physical world, viability means that a person is able to survive environmental con-
straints, a link to biology. Within the conceptual world, viability means that a per-
son’s knowledge is noncontradictory, where noncontradictory means that a person
can attain a state of equilibrium, meaning a person is not in a constant state of per-
turbation (Glasersfeld, 1981a). We point out that this definition of noncontradictory
is a statement from an actor’s perspective, not an observer’s perspective; it is a state-
ment of what noncontradictory means for an actor, not a statement about whether or
not an observer deems another person’s knowledge to be noncontradictory.
Philosophically, we see two important questions that arise from the shift to an
experiential reality: (1) Is this shift a denial of the existence of any form of what
people call ontological reality? and (2) Why is it common for people to act as if
there is an ontological reality? To respond to the first question, Glasersfeld (1984,
1995) is clear that shifting from an ontological to an experiential reality is not a
denial of an ontological reality, simply a statement that a person cannot come to
know ontological reality. Moreover, this position is not a statement that a person’s
constructions can be made up freely. Instead, it is a statement that a person experi-
ences constraints within their experiential reality of both the physical and concep-
tual variety; it is not possible for me to walk through my desk (physical constraint),
and I cannot call my concept of one-fourth both one-fourth and one-fifth and main-
tain a state of equilibrium (conceptual constraint). In this sense, our experiential
reality limits what is possible, but it does not determine or cause how a person
responds to these constraints (Glasersfeld, 1983, p. 4).1
1
We note that conceptual constraints are always relative to a person’s current schemes. Therefore,
we are not suggesting here that, for example, elementary grade students would necessarily experi-
ence a perturbation (i.e., a loss of equilibrium) when talking about fractions like thirds, fourths, or
fifths using the language “halving it up.”
A response to the second question can be found in Elkind’s introduction to Six

Psychological Studies when he says:
Once a concept is constructed, it is immediately externalized so that it appears to the subject
as a perceptually given property of the object and independent of the subject’s own mental
activity. (Elkind in Piaget, 1967, p. xii, italicized in the original)
That is, for Piaget, knowing entails an active agent introducing actions or operations
into their experiential world and then subsequently, once a concept is constructed,
treating these actions or operations as if they are external to and independent of the
acting agent. Once the process of externalization takes place, a person’s sense of the
“existence of their concepts in an independent world” is strengthened in two ways.
First, when a person’s constructions prove to be viable in future experience, a per-
son’s sense of “their existence” is strengthened. As Thompson (2013, p. 61) puts it,
“we construct stable understandings by repeatedly constructing them anew.”
Second, when a person attributes to other people compatible constructions, “their
existence” is also strengthened. The process of two or more people attributing to
each other compatible constructions is what Glasersfeld (1995) calls establishing
intersubjectivity, which he identifies as the highest form of reality.
We start with these epistemological considerations because schemes are one of
the primary tools that Piaget used to account for the construction of what Glasersfeld
calls a person’s experiential reality. This constructive process has a definite begin-
ning—at birth. Therefore, it is not surprising that Piaget developed schemes from
his adjustment of reflexes, which evolutionary biologists consider to be present at
birth. We discuss Piaget’s adjustment of reflexes to create schemes next.
iaget’s Development and Glasersfeld’s Three-Part Definition

P
of Schemes
Schemes from Reflexes
Glasersfeld (1982, 1993) identifies that evolutionary biologists typically conceive

of reflexes as involving a “stimulus-response” mechanism where the response is a
“fixed action” (Fig. 3.1, Glasersfeld, 1995, p. 64). An example of a reflex in humans
is the rooting reflex, where the “stimulus” is a baby’s experience of a brushing sen-
sation on his or her cheek, and the baby’s response is to turn its head and root.
Evolutionary biologists consider reflexes to be due to random genetic variation,
where a reflex provides a critical advantage to members of a species who possess
Fig. 3.1 Glasersfeld’s depiction of a reflex

them because they increase the likelihood of survival. The advantage that the root-
ing reflex offers is a higher likelihood of providing nutrition to a baby.
Piaget took the idea of a reflex as a starting point for creating schemes. He con-
ceptualized schemes as having three parts rather than two (Glasersfeld, 1993): a
perceived or conceived situation, an activity, and a result (Fig. 3.2, Glasersfeld,
1995, p. 65). To make sense of Piaget’s replacement of a stimulus with a perceived
or conceived situation involves understanding his intent. He intended to make
explicit that stimuli neither exist in an observer-independent environment nor do
they carry information (Glasersfeld, 1980b). Rather an active agent has to select and
coordinate sensory signals or reproduce them in their absence to establish a per-
ceived or conceived situation. Piaget also adjusted the notion of a “fixed action” in
a reflex, removing the notion that a response is fixed (Glasersfeld, 1993). For exam-
ple, as methods of feeding a baby change so too does the baby’s response to having
its cheek brushed. Removing the fixedness between stimulus and response removes
the idea that a stimulus and response are in one-to-one correspondence with one
another. That is, multiple perceived or conceived situations may activate the same
scheme (Thompson, 2013).
The final adjustment Piaget made explicit in creating schemes was the inclusion
of a result, which is a component that is not typically explicitly identified as part of
a reflex. In the case of the rooting reflex, the result would be that the baby’s hunger
is sated. Adding the notion of a result to reflexes is important to scheme theory
because it introduces the idea that a person can develop an expected result. An
expected result is an anticipation of what the result of a scheme will be based on
prior experience. Thus, once a baby has gained sufficient experience, the rooting
reflex can be considered a sucking scheme (sufficient experience for Piaget included
the first 6 weeks of life); a baby may perceive a situation (e.g., brushing against its
cheek) as an occasion for the activity of rooting and come to expect that this activity
will satisfy its sensation of hunger. To sum up, Glasersfeld proposes that Piaget
made these adjustments in creating schemes from reflexes and proposes a three-part
definition of Piaget’s term scheme, a scheme entails a perceived or conceived situa-
tion, an activity, and a result (shown in Fig. 3.2).
Sensory Motor and Conceptual Schemes
For researchers to effectively use Glasersfeld’s three-part definition of scheme

involves understanding more about two issues in Piaget’s framing of the construct:
(1) the genesis of mental operations and their role in schemes, and (2) the distinction
Fig. 3.2 Von Glasersfeld’s

depiction of a scheme
between figurative and operative. To discuss the first issue, we note that Piaget
(1971) used schemes as an explanatory construct for knowing and learning in both
the sensory motor and conceptual domains (i.e., both the physical world and world
of ideas). As he states:
All knowledge is tied to action and knowing an object or an event is to use it by assimilating
it to an action scheme … This is true on the most elementary sensory-motor level and all the
way up to the highest logical-mathematical operations. (Piaget, 1971, pp. 14–15, 17)
For Piaget, schemes in the sensory motor domain involve the coordination of sen-
sory motor actions—e.g., walking and opening a jar—and within the conceptual
domain the coordination of mental operations—e.g., partitioning a linear whole
into five equal parts (Glasersfeld, 1995). Piaget (1964) considered these two
domains to be linked in the following way:
An operation is an interiorized action. But, in addition, it is a reversible action; that is, it can
take place in both directions, for instance, adding or subtracting, joining or separating.
(p. 176)
For Piaget, the genesis of mental operations was sensory motor actions (i.e.,
kinesthetic activity, composition of sensory motor data in perception). So, for exam-
ple, a partitioning operation has its origins in sensory motor situations that involve
sharing a continuous whole, like breaking a cake into two parts or marking a line
segment into four parts (Piaget et al., 1960; Steffe, 2010b).2 These sensory-motor
actions get internalized, which means a child can carry them out in visualized imag-
ination without carrying out any kinesthetic actions on perceptually available mate-
rial like a cake or a line segment (Norton et al., 2018). However, internalized actions
are not yet reversible, in that a child cannot move backward and forward through an
experience that they carry out in visualized imagination (Steffe, personal communi-
cation, 12/08/21). So, for example, a child cannot move from a partitioned cake
back to a whole cake. Once an internalized action becomes reversible, it is an inte-
riorized operation, which means, for example, that a child can in visualized imagi-
nation move back and forth from whole cake to partitioned cake, and back to whole
cake again.
This account of the genesis of operations from sensory motor actions that get
internalized and then interiorized is a reasonable starting point for understanding
Piaget’s perspective on mental operations. However, there is a missing subtlety that
is not addressed. To address this subtlety, we consider Steffe et al.’s (1983) work on
counting that is rooted in Glasersfeld’s (1981b) account of the genesis of the unit-
izing operation. Glasersfeld proposes that the origins of the unitizing operation are
in a person’s focused and unfocused moments of attention. This account differs
subtly from an account of operations as internalized and interiorized sensory motor
actions in that sensory motor actions are often also characterized as visible to an
2
Steffe (2010a, b, Chap. 4) offers a much more nuanced account of the origins of partitioning that
include five levels of fragmenting where at the fourth level of fragmenting there is the potential for
a student to engage in what he calls equipartitioning.
observer. With such a characterization of sensory motor actions, the distinction

between making an account of visible behavior and making inferences about opera-
tions can easily be conflated. Glasersfeld’s account of the genesis of the unitizing
operation based on focused and unfocused moments of attention highlights that
visible behaviors are only a marker for making inferences about even the most basic
operations; that is, moments of attention are not visible to an observer, and must be
introduced by an active agent. These observations are critical to making sense of
Piaget’s notion of an operation, and they are not captured solely from the first
account we gave of operations as interiorized sensory motor actions.
Another subtle difference that Steffe (personal communication, 12/08/21) made
in transitioning these constructs to mathematics education research is that he relaxed
the criteria that operations necessarily had to be reversible. For our work, we have
followed Steffe’s definition of operations and we have construed Piaget’s (1964)
criteria that—“Above all, an operation is never isolated. It is always linked to other
operations, and as a result, it is always a part of a total structure (p. 176)”—as a way
to relate operations to conceptual schemes. We see operations as the building blocks
for conceptual schemes where operations make up the activity of a conceptual
scheme.3 A significant part of the work, then, of using Glasersfeld’s three-part defi-
nition of scheme is to specify the operations that constitute it.
The second issue, the distinction between figurative and operative, is also an
important distinction for characterizing students’ conceptual schemes. We begin our
discussion of this distinction with Glasersfeld’s (1995) quote about Piaget’s use of
these terms:
Throughout Piaget’s work, the distinction he makes between “figurative” and “operative”,
and the concomitant distinction between (physical) “acting” and mental “operating” are
indispensable for an understanding of his theoretical position. “Figurative” refers to the
domain of sensation and includes sensations generated by motion (kinaesthesia), by metab-
olism of the organism, and the composition of specific sensory data in perception. “Acting”
refers to actions on that sensorimotor level, and it is observable because it involves sensory
objects and physical motion. Any abstraction of patterns composed of specific sensory and/
or motor signals is what Piaget calls “empirical.”…In contrast, any result of conceptual
construction that does not depend on specific sensory material but is determined by what
the subject does, is “operative” in Piaget’s terminology. “Operations,” therefore, are always
operations of the mind and, as such, not observable. Whatever results reflection upon these
mental processes produces are then called “reflective abstraction.” (p. 69)
As Glasersfeld frames it here, Piaget’s distinction between figurative and opera-

tive focuses on the difference between physical acting and mental operating, and in
this framing the figurative domain is tied to actions in the sensory motor domain.
So, if we return to our example of partitioning a cake, when a young child, 3 years
of age, say, cuts a perceptually present cake into parts, that would fall into Piaget’s
figurative domain if the child is constrained to carrying this activity out on a percep-
tually present cake. However, this seemingly sharp distinction can easily get
3
Piaget does not discuss what he means by “total structure” in the immediate context of this quota-
tion. He could mean either a scheme or a grouping structure. We have chosen to interpret his mean-
ing to be a scheme.
muddied: how should a researcher characterize a sixth-grade student who seem-

ingly needs to carry out partitioning activity on a fraction bar to determine the result
of multiplying a fraction by a fraction? Here a sixth-grade student is not, generally,
constrained to actions in the sensory motor domain. To sharpen our distinction, we
return to our more basic example of a 3-year-old; in this example, it is possible to
differentiate between the cake and the act of cutting the cake. In Piagetian terms, the
cake is an object concept, and the act of partitioning the cake is either a sensory
motor or internalized action. This distinction between the material that gets oper-
ated on (i.e., the cake or an object concept) and the actions or operations that get
performed on it (i.e., the partitioning of the cake) helps to sharpen the distinction
between figurative and operative. The material that gets operated on is figurative
and the actions or operations performed on it can be considered operative
(Glasersfeld, 1991).
This distinction is the basis for the way Thompson (1985) generalized Piaget’s
distinction between figurative and operative. Thompson proposed that the differen-
tiation between figurative and operative could be seen as a distinction between a
controlling scheme and a subordinate scheme where both schemes had the potential
to be operative in the sense that both could involve mental operations. With
Thompson’s generalization, the distinction, then, between figurative and operative
is about a relationship between schemes; a current operative scheme needs material
to operate on, and this material is a previously constructed scheme, which serves as
figurative material for the operative scheme. This observation means that figurative
material is always relative to the current cycle of operations or what is currently
operative for a person. This generalization makes the distinction between figurative
and operative useful even for studying students’ schemes related to advanced math-
ematical concepts (Chap. 4; Thompson, 1985).
mpirical Example of Glasersfeld’s Three-Part Definition

E
of Scheme: Michael Solves the Outfits Problem
With the above distinctions in mind, we now use an empirical example to illustrate
Glasersfeld’s three-part definition of scheme. To do so, we use Michael’s solution of
the Outfits Problem.
Outfits Problem. You have 3 shirts and 4 pants. An outfit is 1 shirt and 1 pants.
How many outfits can you make?
Michael was an eighth grader who participated in a 3-year teaching experiment.

He solved the Outfits Problem at the beginning of his third year in the project. We
use this example for two reasons. First, it is an example of a student’s solution to a
problem that is not particularly complex, especially for an eighth-grade student,

which allows us to make a clear exposition of the mental operations that we have
proposed were involved in Michael’s scheme. Second, Michael had developed com-
plex multiplicative reasoning with whole numbers and fractions during the first
2 years of the teaching experiment (Hackenberg, 2010). However, he did not ini-
tially solve the Outfits Problem in the way that we anticipated (i.e., the situation did
not initially activate his powerful multiplicative reasoning with whole numbers).
Therefore, his solution to the problem was an occasion for us to consider that
he used novel mental operations. With this context, we now present Michael’s solu-
tion to the problem.
Data Excerpt 1: Illustrating a Conceptual Scheme
M: Okay, you said four pairs of pants. [Begins to draw a picture of one pair of pants and one
shirt and writes “x 4” next to the pants and “x 3” next to the shirts.] You could take one
(shirt and one pants) and make one outfit. Take another one (shirt and pants to make a
second outfit) and take another one (shirt and pants to make a third outfit) and then you
have one pants (leftover). Wait do we do every possibility?
T: Every possibility you could possibly get.
M: [draws three tally marks followed by four tally marks, making two rows of tally marks
(Fig. 3.3).] You could do that, that, that [connects each tally mark in the first row with
the first tally mark in the second row.] You could do a lot of possibilities of this one. [M
draws similar connecting lines between the rows of tally marks that end up in a jumble.]
There is three (outfits) for every pants so basically twelve.
We interpret Michael’s initial response of three outfits with one pair of pants
leftover (Lines 1–5) as part of his establishing a situation. We note that from the
researcher’s (and likely the reader’s) perspective, the statement of the problem is
sufficient to establish a well-defined situation. However, that is often not the case for
the learner—the learner may make initial interpretations of a situation that get
revised as part of establishing, for themselves, a definite problem situation. Michael’s
question at the end of Line 5 indicates that he established the Outfits Problem as a
situation about creating possibilities, which contrasted with his initial interpreta-
tion, where he considered himself constrained to using a shirt or pants only once.
The activity of Michael’s scheme involved two mental operations, ordering and
pairing, that he carried out on two composite units,4 four and three, where the com-
posite units were the figurative material for his scheme (Lines 7–11). Elsewhere we
Fig. 3.3 Michael’s tally

marks
have defined an ordering operation as creating a first, second, third, and so on unit
in a composite unit, and pairing as uniting a unit from each of two composite units
to create a single unit, a pair, that contains two units (Tillema, 2013, 2014). To trans-
late the abstract language of operations back into context, he created a first, a sec-
ond, and a third shirt; created a first, a second, a third, and fourth pants; and paired
a shirt with a pants to create an outfit (i.e., a pair) 12 times.5 He organized creating
pairs in a lexicographic order, fixing the first pants and pairing it with the first, sec-
ond, and third shirt before creating any outfits with the second pants where he again
repeated pairing it with the first, second, and third shirt (see English, 1991, odom-
eter method).
Michael did not anticipate a definite number of possibilities before he solved the
problem. Partway through his solution, he had an expectation that there would be “a
lot of possibilities (Line 9),” but even after he stated this expectation, it did not seem
to be accompanied by him knowing a definite number of possibilities. When he
finished operating on the tally marks, we interpret his statement, “There is three for
every pants so basically twelve (Line 11),” as him interpreting the result of his
scheme in terms of his composite units being iterable units (i.e., I can make three
with the first pants and so could do that with every pants), and his scheme closed.
However, we claim that the four threes he produced in this situation were a different
kind of composite unit; he produced a composite unit of pairs, three pairs, as
opposed to a composite unit of ones, three ones. We have used this distinction as a
way to make claims about the multiplicative relationships students establish between
one- and two-dimensional discrete units where a pair is a discrete two-dimensional
unit. Later in this chapter, we discuss these differences in more detail to differentiate
among students at different stages of units coordination.
Here, we use this data excerpt to focus on moving from a data excerpt to making
an account of the data excerpt using the three-part definition of schemes. We call
Michael’s scheme a multiplicative pairing scheme (MPS), where he established a
situation, engaged in mental operations on figurative material, and produced a
result. He established the situation using two composite units where the units of
these composite units were iterable units of one. He operated on these two compos-
ite units, using ordering and pairing operations, and interpreted the result he pro-
duced in terms of his iterable composite units (i.e., each three I made is the same).
We have claimed there were differences between his iterable composite units and
the result he produced in this situation, a composite unit of pairs. However, Michael
4
Michael’s units of one and composite units were both iterable. Steffe (2010b) considers a unit of
one to be iterable when a student can take a number word like three as implying three iterations of
a unit of one without having to actually make these iterations. He considers composite units to be
iterable when a student can take one composite unit to stand in for any number of iterations of the
composite unit.
5
From our perspective, the contextualized language helps make some sense of the language of
operations, but it does not help identify key differences among students’ reasoning because the
same contextualized language can be used to describe students who establish qualitatively differ-
ent multiplicative relationships in these situations.
did not make any differentiation in this situation, and so his scheme closed. We note
that Thompson’s (1985) reformulation of figurative material is essential in making
sense of our interpretation of Michael, in that Michael’s composite units, which we
have called figurative material for Michael, were the result of a prior conceptual
scheme (i.e., a scheme based on mental operations) that made the composite units
available to him as material to operate on in this situation.
Michael’s data brings up one last interesting question: When should a researcher
claim that a student has constructed a scheme? Piaget (1936/1977), as quoted in
Thompson et al. (2014, p. 10), characterizes a scheme as “organized totalities whose
internal elements are mutually implied.” Based on this characterization, we would
claim that Michael was in the process of constructing a new scheme rather than that
he had already constructed one (Thompson et al.’s, 2014, understanding in the
moment versus stable understanding, pp. 12–14). This characterization is aligned
with our claim that Michael produced a new type of unit in this situation, a compos-
ite unit of pairs, but that he himself did not differentiate a composite unit of pairs
from a composite unit of ones. Therefore, this data excerpt represents a student
whose scheme was under construction. In such cases, there are two important ques-
tions for a researcher to consider: (1) is it necessary to define a new scheme? and (2)
if so, what relationship is there between a new and prior scheme? One potential
relationship is that a new scheme supersedes a prior scheme, in which case a person
can use the new scheme to solve all the situations that they used the prior scheme to
solve, and they can also use it to solve additional situations. In these cases, the new
scheme, in essence, replaces the old scheme. In our example, we did not see the new
scheme as superseding Michael’s old scheme for whole number multiplication
because this new scheme involved him in establishing a novel relationship between
one- and two-dimensional discrete units. Therefore, we chose to define a new
scheme, an MPS, that was related to his scheme for whole number multiplicative
reasoning but distinct from it.
uilding on Glasersfeld’s Definition of Schemes: Steffe’s

B
Tetrahedral Model
Nuances in Steffe’s Definition of a Scheme
Steffe (2010a) presents a tetrahedral model for schemes that builds on Glasersfeld’s
three-part definition in order to make several additional features of schemes explicit.
One aspect the tetrahedral model makes explicit is that schemes function in relation
to a goal or goals. He places goals at the apex of a tetrahedron because a person’s
goals guide all three parts of a scheme (Fig. 3.4a); this relationship between a per-
son’s goals and their scheme is portrayed by the bi-directional arrows between the
goals and the three parts of a scheme, perceived or conceived situation, activity, and
result, where the three parts of a scheme are represented on the vertices of a tetrahe-
dron (Fig. 3.4b). Throughout his work, Steffe has emphasized the importance of
Fig. 3.4 Steffe’s tetrahedral model of a scheme (a, top left; b, top right; c, bottom)
internal goals rather than external stimuli in guiding schemes. He uses the sucking
scheme to illustrate this issue where he emphasizes that upon experiencing a sensa-
tion of hunger a baby may establish a goal to satiate hunger (Steffe, 2010a,
pp. 21–22). His discussion highlights that an internal goal, the goal to satiate hun-
ger, not external stimuli, brushing the cheek, is what guides even very basic schemes
like the sucking scheme. This example indicates the depth of re-formulation that
took place in developing schemes from reflexes.6
Steffe (2010a) also includes the bi-directional arrows from the goal to the parts
of a scheme (Fig. 3.4b) to make explicit that any one of the three parts of a scheme
coupled with a goal can activate the other parts of a scheme. The potential for any
one of the parts of a scheme to activate the other parts is represented by the bi-
directional arrows between the three parts of the scheme itself (Fig. 3.4c). This
aspect of Steffe’s model highlights that schemes are not linear, i.e., they do not
necessarily flow from perceived situation to activity to expected result (compare
Fig. 3.4c with Fig. 3.2). Rather there is a dynamic relationship among the parts of a
scheme where any one of the parts can activate the others.
We give a hypothetical example of the nonlinearity of a scheme based on extend-
ing the example we started with Michael. It is possible to imagine Michael getting
dressed in the morning and creating an outfit by pairing a shirt with pants that he has
in his dresser. This activity might support him in establishing a goal of determining
the total number of possible outfits he could create from the shirts and pants in his
6
We warn the reader against thinking of Fig. 3.4a or b as indicating that a person’s goals are a part
of the scheme itself. Rather once a person establishes a goal or goals, the goal or goals mediate all
three parts of the scheme.
dresser. This hypothetical example, then, illustrates how the activity of pairing a
shirt with a pants and his goal to determine the total number of possibilities might
activate his MPS. Once activated, it could be an occasion for him to establish a situ-
ation, similar to the Outfits Problem, based on the number of shirts and pants he has
in his dresser and to determine the total number of possible outfits he could make,
the result of his scheme. Thus, Michael could use an activity and goal to establish a
situation and a result. The main point is that any of the three parts of a scheme
coupled with a goal can activate the other parts; a scheme does not necessarily pro-
ceed in a linear fashion.A final aspect of Steffe’s tetrahedral model is that it explic-
itly portrays a person’s expectation, which Steffe shows with bi-directional arrows
between a conceived situation and result (see Fig. 3.4c and compare to Fig. 3.2).
The arrow in Steffe’s model that points from establishing a situation to a result is
intended to convey that as part of establishing a situation a person may also estab-
lish an expectation of what the result of their scheme will be. The arrow from the
result back to establishing a situation is intended to convey that once a person pro-
duces an actual result, they may compare the actual result to their expected result
(cf. Glasersfeld, 1980b, p. 75). We would add that a person can develop expecta-
tions for any of the three parts of a scheme, as well as how these parts are related to
each other.
mpirical Example of Nuances in Steffe’s Definition of Scheme:

E
Nico’s Reversible Scheme
This observation of the dynamic relatedness of the parts of a scheme is of particular

interest to mathematics education researchers when a person can start from the
result of a scheme and return to the situation that produced it. In these instances, we
consider a scheme to be reversible (Hackenberg, 2010; Inhelder & Piaget, 1958).
The reason this kind of reasoning is of particular interest to mathematics educators
is because reversibility is the basis for establishing important relationships, like
subtraction as the inverse of addition, division as the inverse of multiplication, or
square rooting as the inverse of squaring. We illustrate the reversibility of a scheme,
using Steffe’s tetrahedral model, where Nico, a seventh-grade student, starts from a
result and a goal to establish a situation and activity that produced that result. Nico
was solving the following version of the Flag Problem, where he agreed to respond
to this single question outside of the context of working on any other related prob-
lems. The interviewer had the following interaction with Nico.
Flag Problem. You create a two-striped flag using one color in the top stripe
and one color in the bottom stripe. The color in the top and bottom stripe can
be the same (e.g., red-red flag is allowed), and you count as different, for
example, a red-orange and an orange-red flag. How many colors would you
need to make 225 two-striped flags?
Fig. 3.5 A two striped flag and Nico’s notation for the Flag problem (a) Left and (b) right
Data Excerpt 2: Illustrating Reversibility in Steffe’s Tetrahedral Model

of Schemes
T [shows Nico a flag that has two horizontal stripes (Fig. 3.5a).]: If I told you that there
were enough colors to make 225 (two-striped) flags, can you tell me how many colors
there were to start out with? Where a flag would be like red and blue or yellow and green
or red and green.
N: Let’s see…ah, no [indicating initial uncertainty]. [He continues thinking] Um, 15.
T: Okay. Do you want to say why?
[Nico says he took the square root of 225. He has some difficulty articulating why he took
the square root of 225; he says that he does not want to write down all 225 possible flags.
To avoid doing so, he initiates showing the interviewer a smaller case of the problem.
He proposes a case where there are 9 two-striped flags and makes a list to show that 3
colors would produce 9 flags (Fig. 3.5b). He then explains why 15 colors would produce
225 flags.]
N: So you have 15 colors, and then what you could do is like match, if you had like, red,
and then you had just a bunch of other colors you could match red with all of the other
colors. [He momentarily wanders in his explanation and then continues]
….
So you match red with all the colors, then you could move on to blue and you would match
that with all the colors. And then move onto the next one and match that with all the
colors. And eventually you’d get 225.
We interpret this data excerpt as Nico starting from what he considered to be a
result of his MPS (225 possible flags) coupled with a goal of determining how many
colors he would need to produce that number of flags. Doing so activated his expec-
tation that taking the square root would produce the correct number of possible
colors (Lines 6–7); he expected that 225 was produced from a situation where he
would square some number of possible colors.7 He experienced difficulty explain-
ing what activity he would engage in to connect a situation involving 15 colors to
the 225 possible flags, where his difficulty largely seemed to be about his not want-
ing to write down all 225 possible flags (Lines 7–8). To address his difficulty, he
created an analogous situation that involved nine flags where he could more easily
instantiate the activity of his scheme. He ordered the three colors and paired the first
7
We largely use contextualized language for this example because the main purpose is to show the
dynamic relatedness of the parts of a scheme. In the language of unit structures, we consider Nico’s
expectation to have arisen from his establishing a multiplicative relationship among two composite
units and a unit of units of pairs. We discuss this multiplicative relationship later in this chapter. For
this particular situation, the numerosity of the two composite units was unknown, and their numer-
osity had to be equal.
color with each of the other colors (red with red; red with orange; red with yellow),
continuing in this way until he had produced nine ordered pairs (Lines 8–11). Once
he had done so, he was able to explain what activity he would use to move from 15
colors (an initial situation) to 225 flags (a result) without producing all 225 flags
(Lines 12–18).
This example, like the hypothetical example with Michael, shows how Steffe’s
formulation of schemes posits a dynamic relatedness among the parts of a scheme;
in this case, Nico used a result and goal which activated the other parts of his
scheme, a situation and activity. A careful reader might object to this characteriza-
tion in that Nico clearly must have established a situation involving 225 flags (i.e.,
he started from a situation he established). The point, though, is that he considered
that situation to be the result of his MPS. It is in this way that he used a result to
regenerate a situation and activity. This observation indicates the importance of see-
ing schemes as relative to one’s current model of a learner and situated within a
broader goal of making meaningful distinctions in qualities of student reasoning. In
this case, one quality of student reasoning that is important is that Nico’s scheme
was reversible. When a scheme is reversible, the result and the initial situation imply
one another (two composite units of 15 implies 225 ordered pairs, and 225 ordered
pairs implies two composite units of 15). For Nico, this implication enabled him to
link taking a square root and squaring a number.
Using Schemes to Investigate Learning
Assimilation, Perturbation, and Accommodation
One difference between Nico’s data excerpt (Data Excerpt 2) and Michael’s data
excerpt (Data Excerpt 1) is that we do not think Nico’s data excerpt was a situation
of learning, whereas we do think Michael’s data excerpt had the possibility to be a
situation of learning. We consider the potential for learning to occur when a person
experiences a perturbation. Understanding perturbations requires an understanding
of Piaget’s constructs of assimilation and accommodation. Piaget (1971, pp. 4–6)
considered assimilation to be the process that an organism uses to integrate some
material (whether that be physical or conceptual) into its structures. He gives many
biological examples at the beginning of Biology and Cognition of assimilation
where, for example, a plant assimilates light energy into its structures, and in the
process of integration, transforms the light energy into chemical energy. Piaget
emphasized that assimilation always involves a transformation.
We find Glasersfeld’s (1995) characterization that assimilation involves a person
treating a current experience (current problem situation) as an instance of a prior
experience (prior problem situation) to capture the transformative component of
assimilation, but perhaps in an unexpected way. That is, a person transforms a cur-
rent experience vis-à-vis assimilation into one that is already known. An observer
may see differences between a current and prior experience that the assimilating
person (i.e., the actor) either ignores or simply does not consider. It is in this sense
that Steffe and Thompson (2000a) have described assimilation as constitutive for an
actor: an observer may be able to point to differences between current and past
experiences that have activated a person’s scheme, but the person themselves may
not see or does not act on these differences. Glasersfeld (1995) maintains that
assimilation is active both when a person establishes a problem situation (first part
of a scheme) and in comparing an expected result to an actual result (last part of a
scheme). When a person assimilates their actual result to their expectation, their
scheme closes—the scheme has functioned as intended, producing the
expected result.
On the other hand, a person may experience a perturbation when they establish a
difference between what they expect—where their expectations have been devel-
oped in prior uses of their scheme—and a current instantiation of a scheme
(Glasersfeld, 1980b, p. 75). Two common examples of when perturbations occur
are: (1) when a current situation activates a scheme but does not meet the same
conditions as prior situations that have activated it, or (2) when a result of a scheme
differs from the result the person expects (Glasersfeld, 1995). We note that perturba-
tions often occur outside of a person’s conscious awareness, and in such cases,
“establishing a difference” may be entirely implicit for the person.
In response to a perturbation, a person may review any of the three parts of their
scheme and make changes to it. Following Steffe (1991), we call changes a person
makes in their scheme, while the scheme is in use, functional accommodations, and
we consider functional accommodations to be one kind of learning. We note that
making an inference that a person has experienced a perturbation is not the same as
making an inference that a person has made a functional accommodation. To infer
that a person has made a functional accommodation requires evidence that the per-
turbation fed back into the scheme in a way that led to changes in one or more parts
of the scheme. Moreover, we consider an inference of a functional accommodation
to involve establishing that the changes were more or less permanent, similar to
Thompson et al.’s (2014, p. 13) distinction between in the moment and stable. Here
we use the terminology more or less permanent to mean that future situations acti-
vate a scheme in a way that indicates the person integrates the novel ways of operat-
ing into their future uses of the scheme. To illustrate assimilation, perturbation, and
accommodation, we now present an extended empirical example.
mpirical Example of Assimilation: Carlos’s Solution

E
of the Flag Problem
To illustrate assimilation, we use a data excerpt of Carlos, an eighth-grade student,

who worked with Michael as a partner. Carlos also participated in the first 2 years
of a teaching experiment, and his eighth-grade year was his third year in the project.
Carlos spent the first two teaching episodes of his eighth grade solving problems
Fig. 3.6 Carlos’s list for

the Flag Problem
like the Outfits Problem. Unlike Michael, we could attribute to Carlos his construc-
tion of an MPS from the start of the teaching episodes. He could represent these
problems using lists, tree diagrams, and arrays and was able to anticipate correct
multiplication problems for these situations. At the beginning of the third teaching
episode, he solved a version of the Flag Problem that did not involve reversibility.
Flag Problem. You have 15 colors with which to make a two-stripe flag. One
color in the top stripe and one color in the bottom stripe. How many two stripe
flags could you make?8
To set up the Flag Problem, the teacher-researcher (first author) asked Carlos and
his partner to make some example flags. Together they made some flags with red in
the top stripe and some flags with red in the bottom stripe, but neither made a flag
with the same color in both stripes nor did they make two flags that had the colors
switched (e.g., the red-blue and blue-red flag). Once the boys had made some exam-
ple flags, the teacher-researcher asked them to make a list of all the possible flags,
8
When I say problems “like the Outfits Problem” or problems “like the Flag Problem,” I mean that
the statement of the Outfits Problem refers to two composite units (4 pants and 3 shirts) whereas
the statement of the Flag Problem refers to only a single composite unit (15 colors). This differen-
tiation is ours; a student may assimilate problems like the Flag Problem using two composite units,
depending on the state of their 2-slot MPS.
leaving open whether they should count flags that had the same color in each stripe,
and whether they should count flags that had the colors switched (i.e., count as dif-
ferent the red-blue and blue-red flag). The following exchange between Carlos,
Michael, and the teacher-researcher took place over an 11-minute period. Carlos’s
response during this time illustrates well features of how assimilation functions.
Data Excerpt 3: Assimilation as Constitutive
C [in response to the Flag Problem writes “15 × 15” on his paper and uses his multiplication
algorithm to find the result should be 275 (an error in carrying out the algorithm). He
begins to write possible colors, associating a number with each. He writes out the first
column of notation by writing “1,1”, “1,2”, etc. (Fig. 3.6). He writes out the second
column the same as the first. When he gets to the third column, he writes, “3”, thirteen
times, and then fills in the appropriate number next to each “3”. He writes the fourth and
fifth column of notation in a similar manner to the third column. He is writing down one
less flag in each column of Fig. 3.6 so the total number of flags in the first five columns
is 15 + 14 + 13 + 12 + 11]
[Six minutes have passed. Carlos is finishing his ninth column of notation]
W: You guys are going to need more paper.
C: I already have the answer. That is [points to where he has written “15 × 15”] how many
pairs you could make. [Carlos has incorrectly identified that his list will contain
“15 × 15” flags when he is done. Michael tells Carlos not to give away the answer. Both
continue to make their list where Michael’s list is similar to Carlos’s list.]
M [indicating that the number of flags in each column of his list is one less than the number
in the previous column]: It just decreases by one every time.
T [To Michael]: Yeah, it just decreases by one every time.
C: [Carlos starts the tenth column writing “10” four times. He counts the number of flags
he has symbolized in the ninth column, finds he has symbolized seven flags, and writes
“10” twice more for a total of six times. He then writes “10”; “11”, etc. next to each “10”
he has written to complete the tenth column. He repeats this process for the remaining
columns.]
M [finishes writing a similar list of notation]: See there is like fifteen and it decreases by one
every time so it’s fifteen plus fourteen plus thirteen plus twelve until you get to one.
T: Go ahead and write that down and see if you can figure out how many you got that way.
[Carlos finishes his list of notation. 10 minutes have passed since he began. Carlos sits
as if satisfied that he has solved the problem.] Carlos, I wonder if you could figure out
how many pairs you got here [pointing to his list in Fig. 3.6]?
C: Two hundred and seventy-five [points to where he has computed 15 × 15.]
T: Why don’t you check that by figuring out what this [points to Fig. 3.6] is?
C: Okay [writes out the sum 15 + 14 + 13 + … + 1 vertically.]
T [To Michael & Carlos]: I wonder if you could figure out a quick way to add those up.
C: Fifteen times fifteen.
We give an interpretation of this data excerpt and then relate it back to the idea
that assimilation is constitutive. From this data excerpt, we infer that for Carlos, the
Flag Problem activated his MPS just as the Outfits Problem had. We note that his
response of 15 × 15 yields a correct total number of outcomes for the Flag Problem
had he counted ordered pairs (e.g., counted the red-blue and the blue-red flag).
However, as his solution went on, it became apparent that he was not intending to
count ordered pairs—he did not include them in his list. Therefore, we do not inter-
pret his multiplication statement as indicating his intent to count ordered pairs.
Rather, we see his response as him treating the Flag Problem as if it were the same
as the Outfits Problem, where, for him, the Outfits Problem did not involve ordered
pairs and did involve multiplication. We infer he assumed the same to be the case for
the Flag Problem. We think it is noteworthy that the way Carlos created his list
indicated that he anticipated the number of flags he would write in each column
before writing them; that is, he made the third column of his list by writing 13 “3s,”
and then filling in the numbers 3 through 15 afterward, indicating that he anticipated
he would make 13 flags before he actually made them. This way of creating his list
is noteworthy because it shows that he anticipated he would write down one less flag
each time he started a new column. It is not surprising, then, that he wrote a sum,
15 + 14 + 13 + … + 1, for the total number of flags in Fig. 3.6. However, he thought
the sum would produce the same total number of flags as his multiplication prob-
lem, 15 × 15. We take that as indicating he did not have a way to compare the result
he produced, 15 + 14 + 13 + … + 1, to his expected result, 15 × 15, except through
a computation.
Here, we do not make a full account of why that occurred or why we think it is
important that it did occur (see Tillema, 2014). Instead, we use this data excerpt to
illustrate that assimilation involves treating a new situation (i.e., the Flag Problem)
as an instance of an already known situation (i.e., problems like the Outfits Problem).
Moreover, doing so was constitutive for Carlos in that he gave no indication that the
Flag Problem was in some way different from problems like the Outfits Problem. In
fact, his response illustrates how robust a previously constructed scheme can be; he
had multiple opportunities during the time he was making his list (an 11-minute
period) to identify differences in his solution of the Flag Problem relative to his
solution of problems like the Outfits Problem. Nonetheless, at the end of that time,
he still asserted that his sum and the multiplication problem would produce the
same number of flags. We are confident that Carlos knew that the sum was not the
multiplication problem, so we infer his sense of their sameness came from assuming
each would produce the same total number of flags. Doing so is another instance of
assimilation; Carlos treated the actual result (the sum) as an instance of his expected
result (the multiplication problem) without differentiating between them. We infer
he made no differentiation between them because he did not, at that moment, have
a way to make a comparison between the two, and without a way to compare them
he simply considered them to be the same.
mpirical Example of Perturbation: Carlos’s Solution

E
of the Flag Problem
We continue with Carlos’s solution of the Flag Problem to illustrate a perturbation

that arose from comparing his actual result (the sum) to his expected result (the
multiplication problem). We note that when a person assimilates their actual result
to their expected result, we would assert that their scheme closes; the scheme has
functioned in the way that the person has expected and no differentiation is made
between the current and prior uses of the scheme. In Carlos’s case, the
teacher-researcher wanted to provide a further opportunity for Carlos to differenti-

ate his actual and expected result. So, he requested that Carlos continue his solution
of the Flag Problem by adding up the numbers in the sum. The following interaction
took place.
Data Excerpt 4: Comparing an Actual and Expected Result
C [evaluates his sum by adding 15 and 5, then adding 9 and 1, 8 and 2, 7 and 3, and 6 and
4. He then adds 14, 13, 12, 11, and 10 in his head. The teacher asks both boys what
they got.]
M: I got one hundred and ten possibilities.
C: I got one hundred and thirty.
T [Neither of the boys’ responses is correct. The total number of flags should be 120]:
Uh-oh. We got to figure out some way to add those numbers up and actually check.
[two and a half minutes have passed]
W: What’s Carlos doing over here? Are you doing the same way (evaluating the sum in the
same way as Michael)?
C [with excitement]: No, I’m not subtracting. I just figured out something! That once you
got that one [C points to the “15” in his sum], then you got that one and that one [Carlos
points to “14” and “1”], then you make another fifteen, that one and that one [Carlos
points to “13” and “2”] and make another fifteen. You keep on going all the way down
until you get to seven and eight which (means) you have eight fifteens so you should
have just timesed fifteen times eight and you would have got the answer [Carlos crosses
out where he has written “15 × 15”.]
Carlos initially evaluated the sum and found that there were 130 flags. Neither
275 from the multiplication problem nor 130 flags from evaluating the sum was a
correct calculation. However, we infer it allowed him to numerically compare his
actual and expected result, and that in doing so he experienced a perturbation.
Carlos never stated that he made a numerical comparison—we could only infer this
comparison and his perturbation from what he did next. He related his actual result,
the sum, to his expected result, the multiplication problem, by transforming the sum
into 15s. We note that perturbations are sometimes construed as consciously con-
flictive, but in this case, the evidence that Carlos experienced a perturbation did not
come from an explicit statement that he was experiencing a conflict. Rather the
inference that he experienced a perturbation came from his expression that he “just
figured out something,” indicating that he was searching for a way to make a com-
parison between his actual and expected result. Thus, in this case, it was the fact that
Fig. 3.7 Carlos’s list for

the Handshake Problem
he was searching for a way to compare the two that led to making the inference that
he had experienced a perturbation.
We highlight that an important component of a researcher making an inference
of a person experiencing a perturbation is the researcher’s current model of a stu-
dent’s schemes. We make this statement because the less consciously conflictive or
obviously different an actual result is from an expected result, the less likely a
researcher is to identify opportunities to occasion perturbations for students. To
highlight this point, we return to the example of Michael where we claimed that he
produced a composite unit of pairs in solving the Outfits Problem, but that he assim-
ilated this result using his composite units where his composite units were compos-
ite units of one. Doing so meant that he did not experience a perturbation in the
Outfits Problem; he treated the actual result as if it were the same as his expectation
of what the result would be. It is possible for a researcher himself or herself not to
make this differentiation, in which case they might reduce the likelihood of respond-
ing in ways that could occasion perturbations for students. We consider the strength
of the models a researcher makes, then, to be connected, though not causally (Steffe,
1996), to the learning opportunities a student might have.
mpirical Example of Functional Accommodation: Carlos’s

E
Solution of the Handshake Problem
Carlos’s solution of the Flag Problem does not provide evidence that his perturba-
tion fed back into his MPS—it simply provides evidence that he found a way to
compare his actual and expected result. Therefore, we look at the next teaching
episode to illustrate data that we would take as sufficient evidence for inferring he
had at least made temporary changes to his MPS, which we frame as necessary, but
not sufficient evidence of a functional accommodation.
Carlos began the next teaching episode solving the Handshake Problem. His
solution illustrates that his perturbation in the Flag Problem did indeed feed back
into his scheme, which is shown in the following data excerpt.
Handshake Problem. There are 10 people in a room. Each person wants to

shake every other person’s hand. How many handshakes will there be?
Data Excerpt 5: Initial Signs of a Functional Accommodation

C [begins by recording the handshakes shown in Fig. 3.7. Later he writes A = 1, B = 2, etc.
to show that A and 1 represent the same person. In his list in Fig. 3.7, he is eliminating
self-handshakes, for example, he has not recorded 1A, and also duplicate handshakes,
for example, he has not recorded 2A because he had already recorded that handshake as
1B. Michael is also making a list for the handshakes.]
M: Same problem as last time, minus one and keep going and going and going.
C: Minus two actually.
M: Uhnun (no).
C [Emphatically]: Unhun (yes), cause they can’t shake their own hand. [Michael shakes his
head, no. Carlos says to the teacher-researcher] Right? So, one couldn’t shake A and two
couldn’t shake…
We take this excerpt as sufficient evidence to claim that Carlos’s perturbation
from the Flag Problem fed back into his scheme. That is, we interpret Carlos’s com-
ment, “minus two actually,” as an indication that he was comparing the total number
of possible handshakes that the second person could have had (10 possible hand-
shakes) to the total number of handshakes that he counted the second person actu-
ally making (8 handshakes). This statement indicates that he was comparing how
the activity of his scheme in the Handshake Problem differed from the activity he
produced in solving problems like the Outfits Problem. In addition, he insisted that
the Handshake and Flag Problems were different—in one, he counted flags with the
same color stripes, and in the other, he did not count self-handshakes. We take this
statement as further evidence that he was comparing his activity in this problem to
his activity in the Flag Problem. These statements are sufficient evidence that the
perturbation he experienced in the Flag Problem fed back into his MPS and that he
had at least temporarily made a change in the activity of his scheme.9
We would require further evidence that this change was more or less permanent
to make a claim that Carlos made a functional accommodation in his MPS. By per-
manent, we mean that should Carlos’s scheme get activated in future situations, he
would, with relative ease, establish similarity between the future situations and the
distinctions he made here. We also suggest that making a claim that a student has
made a functional accommodation is different from characterizing what the func-
tional accommodation was. To make an account of a functional accommodation
requires a characterization of changes the person made to their scheme, which
involves detailing how the operations in their scheme or coordination among
schemes changed. As we discussed with Michael, it also involves determining
whether the functional accommodation produces a new scheme; unlike in Michael’s
case, here we do not assert that Carlos’s functional accommodation entailed the
construction of a new scheme. Instead, we frame the initial signs of a functional
accommodation as changes he made to his MPS that allowed him to compare differ-
ences among the situations that activated his scheme. We do not provide details here
about what changes we propose occurred in the parts of his scheme; we simply state
that it is important to detail such changes in making a claim of a functional accom-
modation (see Tillema, 2014, for details about the functional accommodation).
9
Michael’s statement, “same problem as last time, minus one and keep going and going and
going,” indicates he was comparing his anticipation of what the result of his scheme would be in
this situation to the result he had produced in his solution of the Flag Problem, the sum.
ituating Investigations of Learning Within a Broader

S
Framework: Stages
Glasersfeld’s Definition of Stage
We find investigations of learning using scheme theory to be most powerful when

they are situated within a broader framework for interpreting how students might
progress within a given mathematical domain and what possibilities may be open to
them across distinct mathematical domains. For us, the broader framework that has
informed our work on combinatorial schemes builds from Steffe’s (1992, 1994,
2010b) work on number sequences that Hackenberg et al. (2016) have subsequently
used to identify three distinct stages of units coordination: stage 1, stage 2, and
stage 3 (Chap. 11; Ulrich, 2015, 2016).10 Glasersfeld and Kelley (1982) define a
stage as:
Designat[ing] a stretch of time that is characterized by a qualitative change that differenti-
ates it from adjacent periods and constitutes one step in a progression…the concept of
qualitative change and, consequently, the concept of stage, cannot be derived from a quan-
titative measurement but requires a binary judgment of presence versus absence. (Glasersfeld
& Kelley, 1982, p. 155)
We interpret Glasersfeld’s definition of stage as indicating that students at different

stages construct qualitatively distinct schemes. Moreover, we see a claim that two
students are at different stages as a claim that the difference in stage is unlikely to
change over the course of multiple interactions with a student.11 As Steffe (personal
communication, 4/20/22) puts it, the role that experience plays in stage change is
ambiguous where we interpret Steffe as meaning that the relationship between
experience and stage change is unclear or inexact not that experience is unimportant
in promoting such changes. Thus, a claim that two students are at different stages
helps to provide a boundary for the schemes a researcher might expect a student to
construct in a particular domain of reasoning. With this purpose in mind, we illus-
trate the relationship between stages and schemes by: (1) outlining qualitative dif-
ferences in students’ MPS at each of Hackenberg et al.’s (2016) stages of units
coordination, and then (2) illustrating a qualitative difference in stage 2 and stage 3
students’ schemes for solving 3-D combinatorics problems.12
10
Hackenberg (2007, 2010) and colleagues (e.g., Hackenberg & Sevinc, 2022) have referred to
these as multiplicative concepts rather than as stages in some research publications. However, she
and Norton have both used stages as well (e.g., Hackenberg et al., 2016; Norton et al., 2015)
because they fit Glasersfeld’s definition of stage.
11
This assertion is supported in the research literature in that there are only a few examples of
researchers working in longitudinal studies reporting that a student has made a stage change (e.g.,
Steffe, 1991, 1994; Hackenberg et al., 2021).
12
Characterizing combinatorics problems as 3-D is our characterization of the problems.
ackenberg’s and Norton’s Stages of Multiplicative Reasoning

H
and a 2-slot MPS
We return to the Outfits Problem to characterize qualitative differences in students’

MPS across the three stages of multiplicative reasoning. As our earlier analysis of
Michael’s solution to the Outfits Problem indicated, ordering and pairing operations
are central to students’ construction of an MPS. The way students use these two
operations, along with their disembedding operation to produce distinct multiplica-
tive relationships, are the primary qualitative differences in students’ MPS across
students at different stages of units coordination. A disembedding operation allows
a student to treat a part of a composite unit as independent from, but related to, the
composite unit without destroying the composite unit (e.g., 2 can be treated as inde-
pendent of 6 in the process of, for example, adding 6 and 8 by adding 2 to 8 to make
10 and then adding the remainder of 4).
Stage 1 students’ MPS involves them in operating on two composite units13 by
ordering the units of each composite unit, disembedding a single unit from each
composite unit, and pairing the unit of each composite unit to create a unit that con-
tains two units, but is counted as a single unit, a pair. Stage 1 students’ use of an
ordering operation can support them in producing the pairs in a lexicographic order.
However, stage 1 students have yet to construct a disembedding operation, so they
need the support of perceptual material (e.g., a list or array) to treat the pairs they
produce as independent from the units of one that they use to create pairs. We have
identified that stage 1 students (Tillema, 2018) produce a multiplicative relationship
among a unit of one, a unit of one, and a pair (Fig. 3.8) as the result of their scheme;
it is not a relationship that they can take as a given prior to their activity. Two com-
mon consequences of establishing this relationship as the result of their scheme are
that: (1) they have to create pairs before they can take them as countable and (2) they
may not see points in an array as pairs unless they have created these pairs in imme-
diate past experience. Establishing the multiplicative relationship shown in Fig. 3.8
involves establishing the basic multiplicative relationship between two discrete one-
dimensional units and one discrete two-dimensional unit (one unit times one unit
produces one pair).
Stage 2 students differ from stage 1 students in that they can interiorize the mul-
tiplicative relationship between a unit of one, a unit of one, and a pair (Fig. 3.8). The
interiorization of this multiplicative relationship means that stage 2 students no
longer need to create a pair in immediate past experience to (1) take them as count-
able and (2) treat the points in an array as pairs. Similar to stage 1 students, stage 2
students use ordering, disembedding, and pairing operations in their solution of
problems like the Outfits Problem and can learn to produce pairs in lexicographic
13
We note that stage 1 students have not yet constructed iterable units of one so their composite
units are qualitatively distinct from stage 2 students. Stage 2 students have constructed iterable
units of one, but have not constructed iterable composite units so their composite units are qualita-
tively distinct from stage 3 students whose composite units are iterable (Chap. 11; Steffe, 2010b).
Fig. 3.8 A multiplicative

relationship created by
stage 1 students using
their MPS
Fig. 3.9 A multiplicative

relationship created by
stage 2 students using their
2-slot MPS
order. However, they have constructed a disembedding operation, and so can disem-
bed a unit of one from one composite unit and the entire composite unit from the
other composite unit. In doing so, these students can produce a multiplicative rela-
tionship that stage 1 students do not—a multiplicative relationship between a unit of
one, a unit of units, and a unit of pairs (Fig. 3.9). They produce this multiplicative
relationship as the result of their scheme; it is not a relationship that they can take as
a given prior to their activity. We have considered producing this multiplicative
relationship as akin to establishing a row or column of a 2-D array multiplicatively
where a student differentiates the one-dimensional units from the two-dimensional
units that make up the row or column (e.g., one unit times a unit of four units pro-
duces a unit of four pairs) (Tillema, 2018).
Stage 3 students differ from stage 2 students in that they can interiorize the mul-
tiplicative relationship in Fig. 3.9. The interiorization of this multiplicative relation-
ship means that stage 3 students can take a row or column of an array as
multiplicatively created in the absence of creating it in immediate past experience.
Similar to stage 1 and stage 2 students, stage 3 students use ordering, disembedding,
and pairing operations in their solution of problems like the Outfits Problem, and
they can learn to produce pairs in lexicographic order. However, they use their dis-
embedding operation differently than stage 2 students in that they can disembed the
entire composite unit from each of the composite units. Doing so, coupled with the
unit structure they establish on the pairs, means that they can also interiorize a more
Fig. 3.10 Multiplicative relationship interiorized by stage 3 students using their 2-slot MPS
advanced multiplicative relationship between a unit of units, a units of units, and a

unit of units of pairs (Fig. 3.10). We have considered establishing the multiplicative
relationship shown in Fig. 3.10 as akin to establishing a 2-D array multiplicatively,
where a student differentiates between the one- and two-dimensional units in a 2-D
array (e.g., a unit of three units times a unit of four units produces a unit of four units
of three pairs or a unit of three units of four pairs) (Tillema, 2013). We also refer to
stage 3 students’ scheme as a 2-slot MPS (as opposed to just an MPS) because the
interiorized multiplicative relationship involves two composite units.
The qualitative differences in students’ MPSs at each stage involve differences in
the way they use operations and in the multiplicative relationships that they have
interiorized between one- and two-dimensional units. It is within the framework of
these differences that we have investigated subsequent functional accommodations
that students make to their MPSs—for example, the functional accommodations
Michael and Carlos made, as well as those that lead to the construction of ordered
pairs rather than just pairs (e.g., Tillema, 2021). Now that we have introduced
stages, we discuss an important feature of schemes that Thompson (2013; Thompson
et al., 2014) captures in his formulation that helps us to identify a qualitative differ-
ence in the functional accommodations stage 2 and stage 3 students establish in the
context of solving 3-D combinatorics problems.
Recursion in Thompson’s Definition of Scheme
Thompson et al. (2014) elaborate on Glasersfeld’s (1995) three-part definition and

Steffe’s (2010a) tetrahedral model of schemes to “make [their] explanatory power
more evident (p. 10).” He defines a scheme as “an organization of actions,
Fig. 3.11 Mikayla’s representation for the Sandwich Problem (a, upper left; b, upper right; c, bottom)
operations, images, or schemes—which can have many entry points that trigger
action—and anticipations of outcomes of the organization’s activity (p. 11).” One
feature of Thompson’s definition of scheme that helps to make the explanatory
power more evident is that his definition highlights the recursive nature of schemes:
a current scheme can, and usually does, contain prior schemes and/or operations.14
Given the recursive nature of schemes, it is important to determine what a
researcher considers to be a single scheme versus what is considered an instantia-
tion of multiple schemes. To capture this difference, we have contrasted a sequential
use of two or more schemes with the insertion of one scheme inside another scheme
where this insertion creates a new single scheme from what was previously two dis-
tinct schemes (Tillema & Burch, 2022). By a sequential use of schemes, we mean
that one scheme closes (i.e., the goal of the scheme is satisfied) before the second
scheme opens and the two (or more) schemes remain distinct from one another. In
contrast, an insertion of a scheme or operations into another scheme entails creating
a new singular scheme composed of multiple prior schemes or a new singular
scheme that contains novel operations where we frame this insertion as a recursive
insertion. We use these ideas to illustrate a qualitative difference that can occur in
the schemes of stage 2 and 3 students in their solution of 3-D combinatorics
14
There are differences between Thompson’s, Steffe’s, and Von Glasersfeld’s interpretation of
scheme. Thompson’s definition of scheme, however, explicitly captures the recursive nature of
schemes, which we see as a powerful addition in the context of using either Von Glasersfeld’s
three-part definition or Steffe’s tetrahedral model.
problems: in our examples, the stage 2 student sequentially uses her MPS, while the
stage 3 student recursively inserts the operations of his 2-slot MPS inside of the
scheme.
mpirical Example of a Stage 2 Student Sequentially Using Her

E
MPS: Mikayla Solves the Sandwich Problem
We illustrate a sequential use of schemes with data from Mikayla’s solution to the
Sandwich Problem. Mikayla, a stage 2 student, was a preservice secondary teacher
in her senior year of college. She and a partner participated as students in 12
60–90-minute teaching episodes whose goal was to use combinatorics problems as
a launch point for producing algebraic identities (Tillema & Burch, 2020). During
the first teaching episode, she solved the Outfits and Flag Problems (2-D combina-
torics problems), and then was given the Sandwich Problem (3-D combinatorics
problem). Her solution to the Sandwich Problem is as follows.
Sandwich Problem. Subway has 4 kinds of bread, 3 kinds of cheeses, and 5

kinds of meats. A sandwich is 1 bread, 1 cheese, and 1 meat. How many pos-
sible sandwiches can Subway make?
Data Excerpt 6: Mikayla’s Sequential Use of Her Scheme

M [writes Fig. 3.11a, then writes Fig. 3.11b, and next writes Fig. 3.11c.]
T: Why don’t you go ahead and just say what you did?
M: So, I started with one combination between the two, first. I started with the bread and the
cheese [referring to Fig. 3.11b]. So, I did what I did with the colors [reference to the
Flag Problem]. I did bread in the x-column (first position) or the first part of the stem
and leaf plot, and then the cheese in the other one (in the second position). And got three
combinations for all of those (each bread), so a total of twelve combinations just for
bread and cheese. And then I numbered each one of those combinations, b1c1 was one,
b1c2 was two, and so on all the way up to b4c3 is twelve [referring to the numbers in the
bread and cheese column in Fig. 3.11c]. And then I made another one the same way
[referring to the fact that she considers Fig. 3.11c to be similar to Fig. 3.11b] using the
numbers one through twelve to not write those all over again. (The numbers) repre-
sented the bread and cheese combinations, and then the possibility that they have for the
meat combinations. So like b1c1, or choice one, could have then five meats next to it.
And then I did that all the way down. So, there were twelve possible bread and cheese
combinations times the, five, for each one of those for the meat combinations to get
sixty total sandwiches.
We take Mikayla’s solution of the Sandwich Problem as indication that she cre-
ated the ordered triples through a sequential use of her MPS. That is, she ordered
the units of all three composite units (Fig. 3.11a), but then paired only the units of
two of the three composite units to make 12 ordered pairs. We infer from her cre-
ation of Fig. 3.11b that her goal was to produce all of the ordered pairs (i.e.,
bread–cheese combinations) and that her scheme closed once that goal was satis-
fied. We make the inference that her scheme closed because she made no attempt to
incorporate the units of the third composite unit into her creation of ordered pairs
until after she had completed making all of the ordered pairs.
Stage 2 students have interiorized a relationship between a unit of one, a unit of
one, and an ordered pair, which is what we infer enabled her to take the result of her
MPS, the 12 ordered pairs, as input for a second use of this same scheme. We note
that her language (“So like b1c1, or choice one”) indicated these retained their status
as ordered pairs. She explicitly ordered these pairs with the numbers 1 through 12
(Fig. 3.11c) and paired each one with the units of the third composite unit to make
60 ordered triples. Using language back in the context of the problem, she ordered
the bread–cheese combinations and paired each bread–cheese combination with the
five meats to create a sandwich. We interpret Mikayla’s solution as involving a
sequential use of her MPS. She used this scheme first to make ordered pairs and
then a second time to make ordered triples, where we consider her to have interior-
ized the relationship between a unit of one, a unit of one, a unit of one, and an
ordered triple. Moreover, in her sequential use of her MPS, we consider her to have
established a multiplicative relationship between a unit of one, a unit of one, a unit
of five units, and a unit of five ordered triples in activity (akin to multiplicatively
producing a row or column in a 3-D array).
We illustrate one consequence of Mikayla’s sequential use of her MPS from the
ninth teaching episode when we presented her with the Passwords Problem.
Passwords Problem. You have the letters A, B, C, and D. How many four-
character passwords can you make if the letters in a password cannot be
repeated (i.e., do not count AABD) and order matters (i.e., ABCD and
ABDC are different passwords)?
From our perspective, the Passwords Problem involves determining the number
of permutations of four letters. At this point in the teaching episodes, Mikayla had
solved a variety of different combinatorics problems, including other problems that
we would consider permutation problems.
Data Excerpt 7: Mikayla’s Initial Solution of the Passwords Problem
TR: Do you know how many total, of these passwords, there would be?
M: Sixteen?
TR: Where are we getting sixteen?
M: Because there is four (letters) and we have four choices.
We interpret Mikayla’s response as indication that the Password Problem acti-
vated her MPS, and that she interpreted the solution of the Password Problem as
involving two composite units, four and four, the number of letters, and the length
of the passwords, respectively. Because she was constrained to sequentially using
her MPS, it precluded her from anticipating, for example, a four-factor product like
4 × 3 × 2 × 1. She could determine such a product through the sequential use of her
scheme, but she did not anticipate a multifactor product. Her lack of anticipation of
multifactor products was one constraint that we experienced and attributed to her
sequential use of her MPS.
mpirical Example of a Stage 3 Student Recursively Inserting

E
Operations into a Scheme: Tyrone Solves the Card Problem
We contrast Mikayla’s sequential use of her MPS with Tyrone’s recursive insertion
of operations inside his 2-slot MPS. Tyrone, a stage 3 student, was a tenth grader
who participated in an interview study that consisted of 2-hour-long interviews.
During the first interview, he solved the Sandwich Problem like Mikayla, and then
he solved the Card Problem. We focus on his solution to the Card Problem because,
during his solution to this problem, he inserted his pairing and ordering operations
into his 2-slot MPS, a recursive insertion of operations, which entailed an extension
of his 2-slot MPS.
Card Problem. You have the 2, 3, and King of Spades, a friend has the 2, 3,
and King of Hearts, and another friend has the 2, 3, and King of Diamonds. A
three-card hand consists of one card from each person’s hand (order does not
matter). How many different three-card hands are possible to make?
Tyrone began his solution to the Card Problem by saying he thought there would
be “eighteen hands” because he thought his “three spades” could be paired with “six
combos.” We interpret this statement as evidence that his 2-slot MPS was activated,
and that he combined the heart and diamond cards to get six non-spade cards that he
anticipated he could pair with the three spade cards. At the interviewer’s request, he
attempted to show the 18 three-card hands using actual cards. He did so by fixing
the king of spades and simultaneously moving a heart and diamond card next to the
king of spades, and then simultaneously moving the heart and diamond card away
from the king of spades. He was unable to keep track of which three-card hands he
had made until he fixed a card in both the first and second position, which was a
novelty he introduced into his solution. This novelty is illustrated in the following
excerpt.
Data Excerpt 8: Tyrone Recursively Inserts Operations into His Scheme
T [puts the king of diamonds next to the king of spades. He leaves those two cards next to
each other, and places the two of hearts next to them to create one three-card hand.] One.
[He moves only the two of hearts away from the king of spades and king of diamonds and
places the three of hearts next to the king of spades and king of diamonds to create a
second three-card hand.] Two.
[He moves the two of hearts away from the king of spades and king of diamonds and moves
the king of hearts next to the king of spades and king of diamonds to create a third three-
card hand.] Three [He has made the first three three-card hands that appear in Fig. 3.12a.
He has made them with actual cards and not written them down like in Fig. 3.12a].
[He leaves the king of spades out and removes the king of diamonds and king of hearts. He
places the three of diamonds next to the king of spades and one at a time moves each of
the heart cards next to the king of spades and three of diamonds to create the next three
three-card hands shown in Fig. 3.12a, saying]. Four, five, six.
[He leaves the king of spades out and removes the three of diamonds and three of hearts
cards from next to it. He puts the two of diamonds next to the king of spades and one at
a time moves each of the heart cards next to the king of spades and two of diamonds, the
last three three-card hands shown in Fig. 3.12a. He says,] Seven, eight, nine.
[He leaves the king of spades out and moves the two of diamonds and the king of hearts card
away. He moves the three of hearts next to the king of spades and puts the two of spades
next to those two cards.] Ten. [His tenth three-card hand includes two spade cards,
which is not a three-card hand that should be counted.] Hold on, hold on. See I’m get-
ting lost. Alright I gotta re-start. I got lost. I got ahead of myself. [Over the course of
eight minutes he makes four distinct attempts to produce all three-card hands using
actual cards. On his final attempt he is successful. The interviewer then asks him to cre-
ate a list for the three-card hands. He produces Fig. 3.12b, which includes all of the
three-card hands produced in a lexicographic order]
We contrast our interpretation of this excerpt to the one of Mikayla. Across all of
Tyrone’s attempts, he worked with the units of all three composite units15 at the
same time as he produced triples. That is, we infer that he paired the first unit of one
composite unit with the first unit of the other composite unit, and then recursively
inserted his pairing operation into his 2-slot MPS, pairing the first, second, and third
units of the last composite unit with the pair he had created. Using contextualized
language, he paired the king of spades with the king of diamonds to create a two-
card hand and then paired this two-card hand with the two, three, and king of hearts
(see the first three three-card hands in Fig. 3.12a). We make the inference that he
used his pairing operation recursively based on his fixing the cards in the first and
second position, and then cycling through all possible cards in the third position.
One reason it took Tyrone multiple attempts before he finally produced all three-
card hands was that he did not initially organize his production of triples using his
ordering operation. This facet of his reasoning can be seen in Fig. 3.12a, where he
did not use a consistent order for the heart cards; for the first three three-card hands,
he ordered the hearts as 2, 3, K and for the second three three-card hands he ordered
the hearts as K, 2, 3 (Fig. 3.12a, yellow circles). He also ordered the diamonds as K,
3, 2 (Fig. 3.12a, red circle), which was different from any of the orders he used for
the heart cards (Fig. 3.12a, yellow circles). On his final attempt with the cards, and
in his list (Fig. 3.12b), he ordered the cards in each suit as 2, 3, K (Fig. 3.12b, yellow
circle, red circle, and blue circles). His list indicated that he eventually recursively
used his ordering operation; he ordered the units of all three composite units before
he began creating triples.16 This recursive insertion of his ordering and pairing oper-
ations into his 2-slot MPS is evidence of extending his 2-slot MPS through a recur-
sive insertion of his ordering and pairing operations into his 2-slot MPS.
15
Tyrone’s units of one and composite units were iterable units as described earlier.
Fig. 3.12 Three-card hands Tyrone made on his first attempt (a) and in his final list (b)
Revisiting Theoretical Constructs Relative to the Data Examples
The difference between Mikayla (sequential use of MPS) and Tyrone (recursive
insertion of operations in the extension of his 2-slot MPS) is an instance of a qualita-
tive difference in schemes that students at different stages of units coordination
construct. Our claim, then, about Mikayla was that she was constrained to (Steffe &
Thompson, 2000b) this sequential use of her scheme over the course of the 12
teaching episodes. The claim of being constrained to sequentially using her MPS
means that the functional accommodations she made during the teaching episodes
were relative to her sequential use of her MPS. In contrast, the functional accom-
modations Tyrone he during the interviews were relative to the recursive insertion
of his ordering and pairing operations in his 2-slot MPS.
This comparison fits the two criteria that Glasersfeld provides for differences in
stages: inferring a difference in stage is a judgment of presence versus absence, and
there is a qualitative difference in the schemes of students at different stages. For us,
this qualitative difference in schemes helps us to situate future observations of the
functional accommodations Mikayla and Tyrone made during the teaching episodes
16
We do not claim that Tyrone produced ordered triples just that he produced triples. Producing
ordered triples entails ordering the composite units themselves, which in contextualized language
would involve Tyrone treating the spades as the first suit, the diamonds as the second suit, and the
hearts as the third suit. The fact that Tyrone did not establish ordered triples contrasts with our
inference about Mikalya who did establish ordered triples in the context of the Sandwich Problem.
We attribute this particular difference to Tyrone’s lack of experience with combinatorics problems
(i.e., constructing ordered triples was a functional accommodation he could make, but one he had
not yet made). Generally speaking, stage 2 and stage 3 students conceive of ordered triples in
qualitatively distinct ways.
within a broader framework. This broader framework contributes to the creation of

a constructive itinerary (i.e., what is possible) for stage 2 and stage 3 students in
terms of their combinatorial schemes. It is in this way that qualitative differences in
schemes can serve as a boundary for the claims a researcher makes about student
learning. With that said, we do not interpret stages to be deterministic for two rea-
sons: (1) within a given stage there is a broad range of schemes students construct,
and they make functional accommodations relative to these schemes and (2) a
researcher always remains open to the possibility that a student makes a stage
change in the context of interactions with a student.
Investigating Learning of Stage 3 Students: Levels of Schemes
Researchers have not proposed stages of units coordination beyond stage 3.

However, they have begun to identify significant qualitative differences in the rea-
soning of stage 3 students in empirical studies (e.g., Hackenberg & Sevinc, 2022;
Shin et al., 2020; Tillema & Burch, 2022) or proposed what would be significant
differences in students’ schemes through conceptual analyses (e.g., Thompson
et al., 2014). We propose that it may be fruitful to use Glasersfeld’s definition of
level as a way of interpreting at least some of these differences. Glasersfeld and
Kelley (1982) contrast stages with levels where they define a level as:
not refer[ring] to a stretch of time. Level is a relative concept interpretable only with regard
to the kind of scale employed….[Levels refer to] hierarchical divisions within a system of
classification or measurement and refers to that system and not to any structural or qualita-
tive feature per se. (Glasersfeld & Kelley, 1982, p. 158)
Glasersfeld does not tie his definition of level to a stretch of time, or to making a
qualitative judgment of presence or absence. In this way, we see Glasersfeld’s defi-
nition of level as a tool to identify differences and similarities among students who
are operating within the same stage, and specifically stage 3 students because no
further stages have, as yet, been proposed. Therefore, we consider levels to be a use-
ful tool for creating a more nuanced landscape of schemes within a given stage. We
illustrate one instance of using levels with empirical data for a stage 3 student where
we use the recursive aspect of Thompson et al.’s (2014) definition of scheme to dif-
ferentiate among different levels of a scheme.
mpirical Example of Different Levels of a Scheme for Stage 3

E
Students: Armando Solves the Colored Digits Problem
Armando, a stage 3 eighth-grade student, participated in 2-hour-long interviews.

Overall, the goal of the interviews was to investigate students’ development of the
quadratic identity that (nx)2 = n2x2 using cases of the Colored Digits Problem (two
cases are given). Figure 3.13 shows Armando’s 2-D array for the 4-case of the
Colored Digits Problem, part b.
The Colored Digits Problem. You have a deck of number cards. You draw a card,
replace it, and draw a second card in order to make coordinate points, such
as (1,1).
(a) Suppose the deck of number cards includes the numbers 1 through 7. The
numbers are colored yellow. What color combination(s) are the coordinate
points (e.g., first digit yellow and second digit yellow gives a yellow–yellow
color combination)? How many coordinate points could you make? (1-Case:
1 color and 7 digits)
(b) Suppose the deck of number cards has the numbers 1 through 28 on them.
Every 7 digits is a different color (digits 1 through 7 are yellow, 8 through 14
are red, 15 through 21 are green, and 22 through 28 are pink). How many
color combinations could you make (e.g., first digit yellow and second digit
red gives a yellow–red color combination)? How many coordinate points are
in each color combination? How many total coordinate points could you
make? (4-Case: 4 colors and 28 digits)
For the 4-case of the problem, the goal was for him to establish the equivalence that
282 = (4 × 7)2 = 42 × 72 where 282 refers to one way to count the total number of
coordinate points; (4 × 7)2 is a way to re-express 282 to show the number of col-
ors and number of digits per color; and 42 × 72 is a way to express the total num-
ber of coordinate points where 42 expresses the number of color combinations
and 72 the number of coordinate points in each color combination. Upon entering
the interviews, Armando had not constructed a 2-slot MPS. However, he con-
structed a 2-slot MPS during the first 30 minutes of the first interview. Once
Fig. 3.13 Armando’s array

Armando had constructed a 2-slot MPS, he solved cases of the Colored Digits
Problem where part a was an initial case and part b was a later case. We show
data of Armando explaining part a of the problem to illustrate his construction of
a 2-slot MPS, and then show him explaining part b of the problem in order to
discuss distinct levels of a 2-slot MPS.
Data Excerpt 9: Identifying Levels of a Scheme
A [Provides the following explanation for part a of the problem]: It’s 49…the first 7 (coor-
dinate points) would be like one-one, one-two, one-three, one-four, one-five, one-six,
and one-seven. And it would go on like that. Just switching the one to a two, then a three,
then a four, five, six, and seven. [Later he says that 49 is 7 times 7 or 7 squared.]
….
[A has represented part b of the problem using an array (Fig. 3.13) where he has shown the
different color combinations and drawn all coordinate points in the yellow–yellow color
combination. He has just made a box to show the coordinate points in the other color
combinations. He has recorded the following notation for his picture:
“784 = 42 × 72 = 282 = (4 × 7)(4 × 7)”]
TR: Do you want to talk about your notation and just relate it to your picture?
A: Alright, um, in all it would be 28 squared (for the total number of coordinate points),
since there’s 28 (digits), and so it would be 28 times 28 or 28 squared to put it simple.
And digit and, what’s the word for it? Digit and plotted (color) wise, it would be 7 times 4,
I forgot to add something in this [adds the dot multiplication sign in his notation “784
= 42 × 72 = 282 = (4 × 7) ∙ (4 × 7)”], 7 times 4 times 7 times 4.
TR: Do you just want to show on your picture where you see 7 times 4 times 7 times 4?
A: Well since there are all these 7s, like 7 here [spanning the yellow digits on the horizontal
axis with his thumb and index finger], 7 there [spanning the red digits on the horizontal
axis with his thumb and index finger], there [spanning the green digits on the horizontal
axis with his thumb and index finger], and up to there [spanning the pink digits on the
horizontal axis with his thumb and index finger]. Those all work out to 7 times 4.
And then the same thing up here 7, 7, 7, 7 [each time he says seven he spans his fingers on
seven digits on the vertical axis]. Those two (the 4 sevens times the 4 sevens) multiplied
together would give me the same answer as 28 squared would give me.
And then over here would be 7 squared [points to the coordinate points in the yellow–yel-
low color combination] meaning 7 (squared) in (each of) the colors (combinations), I
want to say.
And 4 squared meaning all the colors, like 4 there [uses his thumb and middle finger of his
right hand to span the colors on the horizontal axis] and 4 here [uses his thumb and
middle finger of his left hand to span the colors on the vertical axis], multiplied together
[moves his right and left hand into the interior of the array touching his right and left
thumb and his right and left middle finger].
As mentioned, his explanation of part a of the Colored Digits Problem is indica-
tion that the problem activated his 2-slot MPS, and that he considered there to be a
multiplicative relationship among two composite units, a unit of seven units, a unit
of seven units, and a unit of seven units of seven pairs (Lines 1–5). We consider his
solution of part b as indication that he inserted his 2-slot MPS inside itself twice.
That is, we infer that part b of the problem activated his 2-slot MPS, and that his
multiplication statement 282 was a result of reasoning that was similar to his reason-
ing in part a (Lines 13–14), where he established a multiplicative relationship
among a unit of 28 units, a unit of 28 units, and a unit of 28 units of 28 pairs.
He also structured the 28 digits along each axis as a unit of 4 units of 7 units
(Lines 20–26, shown with the four colors along each axis in Fig. 3.13). He then
envisioned the total number of coordinate points as the number of coordinate points,
7 squared, within each color combination (Lines 28–30), where there were a total of
4 squared color combinations (Lines 30–35). The 4 squared color combinations are
shown in his array with the colored pairings inside the boxes. The 7 squared coordi-
nate points are shown in the bottom left of his array, after which Armando said, “It’d
take a really long time to make all of those points so I boxed off where they are for
the others.”
We see his array and explanation of notation as evidence of him inserting his
2-slot MPS inside itself twice; that is, contained in 28 squared (i.e., all the coordi-
nate points, which he produced using his 2-slot MPS) was 4 squared (i.e., the color
combinations, which he produced using his 2-slot MPS) where each color combina-
tion contained 7 squared (i.e., the coordinate points within each color combination,
which he produced using his 2-slot MPS). Thus, we see his solution of part b of the
problem as a recursive insertion of his scheme inside itself twice; he inserted his
2-slot MPS inside his 2-slot MPS inside his 2-slot MPS.
We use this data excerpt to propose three levels of a 2-slot MPS in order to dif-
ferentiate among stage 3 student reasoning. The first level is simply the one that
Armando used to solve part a of the problem. It involves establishing a multiplica-
tive relationship among a unit of units, a unit of units, and a unit of units of pairs, as
Armando did when he determined the total number of coordinate points for part a
of the problem.
The second level of a 2-slot MPS entails creating an equivalence like 282 = 42 × 72,
where the equivalence is generated because 282 and 42 × 72 are two distinct ways of
counting the total number of coordinate points (Gatza, 2021). A student might then
use a single instantiation of their 2-slot MPS to determine that the coordinate points
can be counted as 282. Then their scheme closes (i.e., they satisfy their goal of
counting the total number of coordinate points) before they find a second way of
counting the total number of coordinate points. The second way of counting the
coordinate points involves recursively inserting their 2-slot MPS inside itself; the
student recursively inserts the total number of coordinate points per color combina-
tion (one instantiation of their 2-slot MPS) inside each color combination (a second
instantiation of their 2-slot MPS).
The third level is what we proposed Armando did, in which he inserted the inser-
tion of his 2-slot MPS inside itself (three levels of recursive insertion). That is, he
inserted the coordinate points per color combination into the color combinations
(symbolized as 42 × 72), all of which were contained in the total number of coordi-
nate points (symbolized as 282). We consider his doing so as a crucial part of pro-
ducing the (4 × 7)2 in the equivalence that 282 = (4 × 7)2 = 42 × 72, because it made
available to him each of the 28s that he could restructure into a unit of 4 units of
7 units.
evels, Functional Metamorphic Accommodation,

L
and Reflecting Abstraction
This example fits Glasersfeld’s definition of level in that we do not attribute a period
of time to the distinct levels of the scheme we have outlined. Moreover, the levels
are a way to capture what we see as discernably distinct ways of reasoning within a
particular stage that are hierarchical in nature. The differences in this example that
we outlined are hierarchical in nature precisely because they entailed recursive
insertion of a scheme inside itself; thus, the third level could only be attained after
the second, and the second only after the first.
We consider there to be several important reasons to define distinct levels of
reasoning, especially for stage 3 students. The first is that it provides the opportunity
to differentiate between two kinds of learning, a functional accommodation and a
functional metamorphic accommodation (cf. Steffe, 1991). Our earlier example of
Carlos’s learning in his solution of the Flag Problem we considered to be a func-
tional accommodation, where we have documented the changes he made to his
2-slot MPS. His functional accommodation, however, did not involve us in propos-
ing distinct levels of learning. As we documented elsewhere (Tillema, 2014), Carlos
changed the criteria he had for situations that activated his 2-slot MPS, used opera-
tions in a novel way that restructured the activity of his 2-slot MPS, and made changes
to the expected result of his 2-slot MPS. None of these changes to his scheme
involved defining distinct levels.
On the other hand, we consider Armando’s learning over the course of the two
interviews to be in the province of a functional metamorphic accommodation. We
consider his accommodation to be metamorphic precisely because we could define
distinct levels of his scheme based on the recursive insertion of his 2-slot MPS
inside of itself. Further, we consider it to be a way to operationalize the projective
aspect of Piaget’s (1977/2001) reflecting abstraction. That is, two key components
of Piaget’s reflecting abstraction are that a coordination of schemes gets projected
from a lower to higher plane of learning and that this coordination gets reorganized
at the higher plane of learning (cf. Chap. 6). We have considered the insertion of
operations or schemes as a way to account for the projective aspect of Piaget’s
(1977/2001) reflecting abstraction (Tillema & Burch, 2022). When a person recur-
sively inserts operations or schemes into a scheme to create a new single scheme, we
have considered the new single scheme to be at a higher plane of learning precisely
because the new single scheme incorporates multiple prior schemes or operations
into a new singular scheme. Thus, we consider levels, and identifying functional
metamorphic accommodations, that produce changes between levels to be one way
to operationalize the projective aspect of reflecting abstraction. More broadly, we
see these constructs as an important way to make distinctions about different quali-
ties in the kind of learning that a researcher is documenting.
Conclusion
Schemes and the corresponding constructs that we have outlined in this chapter
provide researchers with a set of tools to make accounts of another person’s math-
ematical reasoning. Steffe et al. (1983) have called these accounts second-order
models. We often begin the work of creating a second-order model with questions
like the following in mind:
• Can I clearly characterize the three parts and goal of a person’s scheme?
(Michael’s example in Data Excerpt 1)
• What is figurative and what is operative relative to a person’s current scheme?
(Michael’s example in Data Excerpt 1)
• How are the parts of a scheme dynamically related to each other? (Steffe’s model
of scheme, Michael’s hypothetical example)
• Is a person’s scheme reversible? What affordances does reversibility have in the
students’ reasoning? (Nico’s example in Data Excerpt 2)
These questions serve as a basis for examining data where our initial goal is to
create a scheme or schemes that are a local account of student reasoning, including
qualities of a scheme, like whether it is reversible or not. We often find responding
to the second question to be the most difficult; that is, since figurative material is
always relative to a person’s current operative schemes (Thompson, 1985;
Glasersfeld, 1991), identifying the relation between the operations and figurative
material of a scheme requires situating the person within a broader framework of
mathematical reasoning. We have used the stages of units coordination (Hackenberg
et al., 2016) to accomplish this goal in our work where the stages of units coordina-
tion offer a way to understand key differences in the operations available to a stu-
dent at a certain stage, and differences in the nature of, for example, students’
composite units (i.e., the figurative material for the MPS) at each stage. Based on
this observation, we consider a broader framework to be helpful in scheme creation
because it supports a researcher to consider the nature of the operations available to
a student and the nature of the figurative material they are operating on, both of
which are critical to scheme creation.
We consider a researcher’s creation of a scheme to be a baseline from which to
make further observations about student reasoning. These further observations
include efforts to investigate learning—where one kind of learning is what we have
called functional accommodations, which are accommodations that occur while a
scheme is in use and do not involve what a researcher deems to be distinct levels of
learning (cf. Steffe, 1991). As we engage in this work, we are often considering
questions like the following:
• What situations activate a scheme? How do the activating situations reveal infor-
mation about a person’s expectations? (Carlos’s assimilation in Data Excerpt 3)
• What markers are there that a person is experiencing a perturbation? (Carlos’s
perturbation in Data Excerpt 4)
• What evidence is there that a perturbation feeds back into a scheme, leading to a
functional accommodation? (Carlos’s initial functional accommodation in Data
Excerpt 5)
• What evidence is there that the changes to a scheme are more or less permanent?
(Criteria for claiming a functional accommodation)
We use these questions as a way to support making accounts of learning. As part
of doing so, we find it helpful to think of assimilation as the mechanism by which a
person transforms a given situation into an already known situation (e.g., for Carlos,
the Flag Problem became like the Outfits Problem). From an explanatory stand-
point, considering how assimilation is functioning in a situation can support making
an account of a person’s expectations, their expectations both in terms of situations
that activate a scheme and expectations they have for the result of their scheme.
Making an account of a person’s expectations can then support seeing when and
why a person experiences a perturbation. As we suggested, evidence for a perturba-
tion may be subtle in that it may not be consciously conflictive for a person. Once a
researcher has identified a perturbation, it is important to determine whether and if
the perturbation has fed back into a person’s scheme. Doing so returns a researcher
to the baseline characterization of a scheme with the aim of making an account of
what changes a person made to their scheme and whether there is evidence that the
changes are more or less permanent. These criteria support making an inference of
a functional accommodation.
The tools we have described so far help a researcher in the initial creation and
subsequent use of schemes to model learning. We consider stages and levels to add
two additional layers to a researcher’s creation of schemes. Stages can help a
researcher make an account of how students’ schemes at different stages are quali-
tatively distinct. Levels can support a researcher in making hierarchical distinctions
among students’ schemes within the same stage. When considering stages and lev-
els in scheme creation, we often have the following kinds of questions in mind:
• What schemes do students at different stages of units coordination construct?
How are these schemes qualitatively distinct from one another? (Mikayla and
Tyrone in Data Excerpts 6 and 7)
• How can differentiating levels help to situate observations about students’
schemes who are at the same stage? (Armando in Data Excerpt 8)
• How do levels support differentiation in the kind of learning that a researcher is
observing? (Discussion of functional versus functional metamorphic
accommodation)
Consideration of the first question helps us to capture important differences in
students’ schemes that are at different stages. These differences go beyond ones that
we would expect could be explained only through functional accommodations. That
is, we view functional accommodations relative to differences in the schemes stu-
dents at different stages construct where our expectation is that, without a change in
stage, the schemes of students at different stages would remain qualitatively dis-
tinct. We do not think of stages as deterministic but rather orienting when we are
making second-order models; we remain open to the possibility that a student makes
a stage change while at the same time valuing the learning of a learner at a
given stage.
On the other hand, levels support us to differentiate what we see as discernably
distinct ways of reasoning within a particular stage that are hierarchical in nature.
One of the primary reasons for introducing levels is to signal potential differences
in qualities of learning where we have considered changes in levels to entail a func-
tional metamorphic accommodation as opposed to just a functional accommoda-
tion. In the above example, we used the recursion of a scheme to capture the
hierarchical nature of levels. However, characterizations of different levels of a
scheme are one area for research that is relatively underdeveloped, and we specifi-
cally consider there to be a need to develop such accounts for stage 3 students.
These accounts, accompanied with the use of functional metamorphic accommoda-
tion, would contribute to mathematics educators’ work to operationalize Piaget’s
construct of reflective abstraction (see Chaps. 6 and 8 for other ways to do the same).
These tools, as well as the accompanying questions, have helped us to see nuance
in student reasoning and to organize this nuance into connected accounts of learn-
ing. This work, in turn, has supported us in having productive interactions with
students and served as a basis for our interactions with pre- and in-service teachers.
It is in this way that we consider scheme theory to be a living tool that dynamically
supports us in our work in research and teaching.
Acknowledgments This research was supported by the National Science Foundation under
Grants No. DRL-1920538 and No. DRL-1419973. The views expressed do not necessarily reflect
official positions of the foundation.
The authors would like to thank the following people for their helpful comments on this chap-
ter: Les Steffe, Pat Thompson, Paul Dawkins, Derek Eckman, and Amy J. Hackenberg.
References
English, L. (1991). Young children’s combinatoric strategies. Educational Studies in Mathematics,

22(5), 451–474.
Gatza, A. M. (2021). Not just mathematics, “just” mathematics: Investigating mathematical learn-
ing and critical race consciousness [Doctoral dissertation, Indiana University, Bloomington].
ProQuest.
Glasersfeld, E. V. (1974). Piaget and the radical constructivist epistemology. In C. D. Smock
& E. von Glasersfeld (Eds.), Epistemology and education (pp. 1–24). Follow Through
Publications.
Glasersfeld, E. V. (1980a). Viability and the concept of selection. American Psychologist, 35(11),
970–974.
Glasersfeld, E. V. (1980b). The concept of equilibration in a constructivist theory of knowledge.
In F. Benseler, P. M. Hejl, & W. K. Koeck (Eds.), Autopoiesis, communication, and society
(pp. 75–85). Campus.
Glasersfeld, E. V. (1981a). The concepts of adaptation and viability in a constructivist theory of
knowledge. In I. E. Sigel, D. M. Brodzinsky, & R. M. Golinkoff (Eds.), Piagetian theory and
research (pp. 87–95). Hillsdale.
Glasersfeld, E. V. (1981b). An attentional model for the conceptual construction of units and num-
ber. Journal for Research in Mathematics Education, 12(2), 83–94.
Glasersfeld, E. V. (1982). An interpretation of Piaget’s constructivism. Revue Internationale de
Philosophie, 36(4), 612–635.
Glasersfeld, E. V. (1983). Learning as constructive activity. In J. C. Bergeron & N. Herscovics
(Eds.), Proceedings of the 5th annual meeting of the North American Group of Psychology in
Mathematics Education (Vol. 1, pp. 41–101). PME-NA.
Glasersfeld, E. V. (1984). An introduction to radical constructivism. In P. Watzlawick (Ed.), The
invented reality (pp. 17–40). Norton.
Glasersfeld, E. V. (1990). An exposition of constructivism: Why some like it radical. In R. B. Davis,
C. A. Maher, & N. Noddings (Eds.), Constructivist views on the teaching and learning of math-
ematics (pp. 19–29). National Council of Teachers of Mathematics.
Glasersfeld, E. V. (1991). Abstraction, re-presentation, and reflection. In L. P. Steffe (Ed.),
Epistemological foundations of mathematical experience (pp. 45–67). Springer.
Glasersfeld, E. V. (1993). Learning and adaptation in the theory of constructivism. Communication
and Cognition, 26(3/4), 393–402.
Glasersfeld, E. V. (1995). Radical constructivism: A way of knowing and learning. Falmer Press.
Glasersfeld, E. V. (1997). Homage to Jean Piaget. The Irish Journal of Psychology, 18(3), 293–306.
Glasersfeld, E. V. (2001). Scheme theory as a key to the learning paradox. In A. Philipp &
J. Vonèche (Eds.), Working with Piaget: Essays in honour of Bärbel Inhelder (pp. 139–146).
Psychology Press.
Glasersfeld, E., & Kelley, M. F. (1982). On the concepts of period, phase, stage, and level. Human
Development, 25(2), 152–160.
sion of the splitting hypothesis. The Journal of Mathematical Behavior, 26(1), 27–47.
Hackenberg, A. J. (2010). Students’ reasoning with reversible multiplicative relationships.
Cognition and Instruction, 28(4), 383–432.
Hackenberg, A. J., & Sevinc, S. (2022). A boundary of the second multiplicative concept: The case
of Milo. Educational Studies in Mathematics, 109(1), 177–193.
Hackenberg, A. J., Norton, A., & Wright, R. J. (2016). Developing fractions knowledge. Sage.
Hackenberg, A. J., Walsh, P. A., & Valero, J. R. (2021). A case of units coordination stage change
in middle school. In Brief research report at the forty third annual meeting of the International
Group for Psychology of Mathematics Education in North America (pp. 1281–1286). Towson
University.
Inhelder, B., & de Caprona, D. (1992). Vers le constructivisme psychologique: Structures?
Procédures? Les deux indissociables. In B. Inhelder & G. Cellérier (Eds.), Le cheminement des
déscouvertes de l’enfant (pp. 19–50). Delachaux et Niestlé.
Inhelder, B., & Piaget, J. (1958). The growth of logical thinking from childhood to adolescence: An
essay on the construction of formal operations (Vol. 22). Psychology Press.
Norton, A., Boyce, S., Phillips, N., Anwyll, T., Ulrich, C., & Wilkins, J. (2015). A written instru-
ment for assessing students’ units coordination structures. International Electronic Journal of
Mathematics Education, 10(2), 111–136. https://doi.org/10.12973/mathedu.2015.108a
Norton, A., Ulrich, C., Bell, M. A., & Cate, A. (2018). Mathematics at hand. The mathematics
educator, 27(1), 33–59.
Piaget, J. (1964). Part I: Cognitive development in children: Piaget development and learning.
Journal of Research in Science Teaching, 2(3), 176–186.
Piaget, J. (1967). Six psychological studies (D. Elkind, Trans.). Random House.
Piaget, J. (1971). Biology and knowledge: An essay on the relations between organic regulations
and cognitive processes. University of Chicago Press.
Piaget, J. (1936/1977). The Origin of Intelligence in the Child. United Kingdom: Penguin.
Piaget, J. (1977/2001). Studies in reflecting abstraction (R. Campbell, Trans.). Psychology Press.
Piaget, J., Inhelder, B., & Szeminska, A. (1960). The child’s conception of geometry. Basic Books.
Shin, J., Lee, S. J., & Steffe, L. P. (2020). Problem solving activities of two middle school students
with distinct levels of units coordination. The Journal of Mathematical Behavior, 59, 1–19.
Steffe, L. P. (1991). The learning paradox: A plausible counterexample. In Epistemological foun-
dations of mathematical experience (pp. 26–44). Springer.
Steffe, L. P. (1994). Children’s multiplying schemes. In G. Harel & J. Confrey (Eds.), The devel-
opment of multiplicative reasoning in the learning of mathematics (pp. 1–41). SUNY Press.
Steffe, L. P. (1996). Social-cultural approaches in early childhood mathematics education: A dis-
cussion. In Mathematics for tomorrow’s young children (pp. 79–99). Springer.
Steffe, L. P. (2010a). Perspectives on children’s fraction knowledge. In L. P. Steffe & J. Olive
(Eds.), Children’s fractional knowledge (pp. 13–25). Springer.
Steffe, L. P. (2010b). Articulation of the reorganization hypothesis. In L. P. Steffe & J. Olive (Eds.),
Children’s fractional knowledge (pp. 49–74). Springer.
Steffe, L. P., & Thompson, P. W. (2000a). Interaction or intersubjectivity? A reply to Lerman.
Journal for Research in Mathematics Education, 31(2), 191–209.
Steffe, L. P., & Thompson, P. W. (2000b). Teaching experiment methodology: Underlying prin-
ciples and essential elements. In R. Lesh & A. E. Kelly (Eds.), Research design in mathematics
and science education. Kluwer.
Philosophy, theory, and application. Praeger Publishers.
Thompson, P. W. (1985). Experience, problem solving, and learning mathematics: Considerations
in developing mathematics curricula. In E. Silver (Ed.), Teaching and learning mathematical
problem solving: Multiple research perspectives (pp. 189–243). Erlbaum.
Thompson, P. W. (2008). Conceptual analysis of mathematical ideas: Some spadework at the foun-
dation of mathematics education. Plenary paper delivered at the 32nd Annual Meeting of the
International Group for the Psychology of Mathematics Education. In O. Figueras, J. L. Cortina,
S. Alatorre, T. Rojano, & A. SÈpulveda (Eds.), Proceedings of the annual meeting of the
International Group for the Psychology of Mathematics Education (Vol. 1, pp. 45–64). PME.
Thompson, P. W., Carlson, M. P., Byerley, C., & Hatfield, N. (2014). Schemes for thinking with
magnitudes: A hypothesis about foundational reasoning abilities in algebra. In K. C. Moore,
L. P. Steffe, & L. L. Hatfield (Eds.), Epistemic algebra students: Emerging models of students’
algebraic knowing (WISDOMe Monographs) (Vol. 4, pp. 1–24). University of Wyoming.
Tillema, E. S. (2013). A power meaning of multiplication: Three eighth graders’ solutions of
Cartesian product problems. The Journal of Mathematical Behavior, 32(3), 331–352.
Tillema, E. S. (2014). Students’ coordination of lower and higher dimensional units in the con-
text of constructing and evaluating sums of consecutive whole numbers. The Journal of
Mathematical Behavior, 36, 51–72.
Tillema, E. S. (2018). An investigation of 6th graders’ solutions of combinatorics problems and
representation of these problems using arrays. The Journal of Mathematical Behavior, 52, 1–20.
Tillema, E. S. (2021). Students’ solution of arrangement problems and their connection to Cartesian
product problems. Mathematical Thinking and Learning, 22, 23–55.
Tillema, E. S., & Burch L. J. (2020). Leveraging combinatorial and quantitative reasoning to
support the generalization of advanced algebraic identities. Invited paper presentation at the
International Congress on Mathematical Education to the Topic Study Group on the Teaching
and Learning of Discrete Mathematics in Shanghai, China.
Tillema, E. S., & Burch, L. J. (2022). Using combinatorics problems to support secondary teach-
ers understanding of algebraic structure. Zentralblatt für Didaktik der Mathematik, 54, 777.
https://doi.org/10.1007/s11858-022-01359-1
(Part 1). For the Learning of Mathematics, 35(7), 2–7.
(Part 2). For the Learning of Mathematics, 36(1), 34–39.
Chapter 4
Operationalizing Figurative and Operative
Framings of Thought
Kevin C. Moore, Irma E. Stevens, Halil I. Tasova, and Biyao Liang
Introduction
Theory is in the end, as has been well said, the most practical of all things, because this
widening of the range of attention beyond nearby purpose and desire eventually results in
the creation of wider and farther-reaching purposes and enables us to use a much wider and
deeper range of conditions and means than were expressed in the observation of primitive
practical purposes. For the time being, however, the formation of theories demands a reso-
lute turning aside from the needs of practical operations previously performed.
(Dewey, 1929)
Theory is the stuff by which we act with anticipation of our actions’ outcomes and it is the
stuff by which we formulate problems and plan solutions to them. (Thompson, 1994b,
p. 229, in reference to Dewey’s quote)
K. C. Moore (*)
Department of Mathematics, Science, and Social Studies Education, University of Georgia,
Athens, GA, USA
e-mail: kvcmoore@uga.edu
I. E. Stevens
Department of Mathematics and Applied Mathematical Sciences, University of Rhode Island,
Kingston, RI, USA
e-mail: irma.stevens@uri.edu
H. I. Tasova
Department of Teacher Education and Foundations, California State University, San
Bernardino, San Bernardino, CA, USA
e-mail: halil.tasova@csusb.edu
B. Liang
Academic Unit of Teacher Education and Learning Leadership, The University of Hong
Kong, Hong Kong, SAR, China
e-mail: biyao@hku.hk

https://doi.org/10.1007/978-3-031-47386-9_4
90 K. C. Moore et al.
One framing of theory is that it provides ready-made tools to be applied as ways

to understand, organize, and explain phenomena. In such a framing, theoretical con-
structs are mostly static in their definitions, with researchers using them as defined.
For instance, Moore (2014a) used Carlson et al.’ (2002) covariation framework to
model a student’s construction of the sine relationship in graphical and unit circle
contexts. In adopting the mental actions introduced by Carlson et al. (2002), Moore
did not make modifications to their definitions, instead applying them as defined in
order to provide a hypothetical account of the cognitive actions driving the student’s
activity.
Another framing of theory is that it provides malleable tools for inquiry and
research pursuits. Rather than providing ready-made tools to be applied, theory and
theoretical constructs provide a general roadmap or orientation. Theoretical con-
structs are dynamic in their definitions, with researchers adapting constructs as they
carry out empirical work so that the researcher can operationalize theoretical con-
structs in the context of novel settings and data. The adaptations result in an evolu-
tion of the theoretical constructs that account for novel settings while maintaining
some central tenets of those constructs. For example, in adopting Ellis’s generaliza-
tion framework (Ellis, 2007b), Ellis and colleagues (2022) found it necessary to
adapt that framework to model the reasoning of students across different grade
bands and content areas. We write this chapter in the spirit of this latter framing.
The notions of figurative and operative thought were introduced over a half-
century ago by Piaget (1969; Piaget & Inhelder, 1971, 1973), and since then, math-
ematics education researchers have used, extended, and shaped them in several
ways. Their uses have been varied, and no singular definition of either sufficiently
captures its development and use. Thus, in this chapter, we present various uses of
the two notions of thought with the intention of highlighting their value in modeling
student thinking. In doing so, we situate evolutions in the two constructs, while
underscoring common tenets that span across these evolutions. We discuss method-
ological concerns regarding notions of figurative and operative thought, including
how the constructs can be used in task design of empirical studies. We also discuss
the constructs’ implications for researchers’ claims regarding students’ meanings.
Because theoretical constructs should be continuously tested in new areas to assess
their viability and expand their generalizability, we end the chapter with various
ideas for future research that can contribute to continued evolutions in the notions
of figurative and operative thought.
Some “Definitions”
As mentioned in the introduction, we do not intend to provide definitive definitions

of figurative and operative thought. This choice is not to meant to imply that defini-
tions are not important. Definitions help paint a picture of the constructs’ central
tenets when combined with researchers’ uses of the constructs, but what ultimately
4 Operationalizing Figurative and Operative Framings of Thought 91
matters is the use of the constructs. As a resource to the reader, we provide a non-
exhaustive collection of researcher quotes related to defining figurative and opera-
tive thought (Table 4.1).
Uses and Evolution of Figurative and Operative Thought
As can be inferred from Table 4.1, one of the more difficult aspects of reading
Piaget is that he did not always provide concise definitions, and when he did, his
definitions evolved to capture changes in his conception of knowledge and know-
ing. Piaget’s approach to his writing and work mirrors his genetic epistemology and
his conception of objects. To know an object is to act on it or engage in its use, as an
object is comprised of operations of thought. Or, “To my way of thinking, knowing
an object does not mean copying it—it means acting upon it” (Piaget, 1970, p. 15).
It is through acting on an object (e.g., operationalizing a construct) that an individ-
ual furthers their understanding of the object via the mental operations imbued to
the object.
Fortunate to our role as authors of this chapter, the figurative and operative dis-
tinction did not see significant evolutions in Piaget’s writing and use. We hypothe-
size that this is because the distinction formed a general categorization of different
knowledge forms, as opposed to constructs that emerged through nuanced charac-
terizations of knowledge development (e.g., his various forms of abstraction). In
fact, references to figurative or operative forms of thought occur in the context of
Piaget describing forms of abstraction, in which he situates operative forms of
thought with more sophisticated forms of abstraction.
As Montangero and Maurice-Naville (1997) reported, the figurative and opera-
tive distinction emerged during writings by Piaget and his colleagues on perception,
representation, mental imagery, and memory (Piaget, 1969; Piaget & Inhelder,
1971, 1973). Defined generally, figurative aspects of thought include those things
that involve perception, sensation, sensory objects, and physical motion. Figurative
aspects of thought deal with configurations and states, of which an individual looks
to mimic in production. Operative aspects of thought include the coordination of
mental actions and their transformations, and are those mental operations that define
objects and allow individuals “to free themselves from the illusions and deforma-
tions aroused by perception” (Montangero & Maurice-Naville, 1997, p. 144).
Operative forms of thought enable a student to isolate numerosity or make compen-
sations of dimensions when engaged in Piaget’s famous pebble counting or glass
and volume tasks. Figurative forms of thought lead to children relying on perceptual
properties (including spatial positioning of endpoints) to draw conclusions about
numerosity or volume.
Before transitioning to the ways in which the figurative and operative distinction
has been used in depth by mathematics educators, we perceive Piaget’s distinction
to be related to (and have likely informed) his work on children’s conception of
Table 4.1 Figurative and operative quotes

Figurative and operative thought quotes
Figurative thought quotes Source
Piaget did admit an inferior species of knowledge that merely pictured static Piaget (2001,
configurations; he called the lesser kind figurative knowledge. For Piaget p. 2)
perception and—usually—language are figurative.
Figurative knowledge is knowledge of observables… Piaget (2001,
p. 298)
‘Figurative’ refers to the domain of sensation and includes sensations generated von
by motion (kinaesthesia), by the metabolism of the organism (proprioception), Glasersfeld
and the composition of specific sensory data in perception. (1995, p. 69)
[the figurative aspect of thought] tends to include the figural character of reality, Piaget and
i.e., configurations as such. With it can be grouped: (a) perception, which Inhelder
functions only when an object is present and through the intermediary of a (1971,
sensorial field; (b) imitation in the broad sense of the term (gestural, phonic, pp. 11–12)
graphic, etc.) which functions either with or without the presence of the object,
but in any case through overt or covert motor reproduction; (c) the mental
image, which functions only when there is no object present.
The figurative functions have no tendency to transform objects, but tend to Piaget and
supply imitations of them…static configurations, which are relatively easy to Inhelder
translate into images; even when they concern movements or transformations, (1973, p. 9)
they do so in order to produce the appropriate configurations, not the changes of
state.
When a person’s actions of thought remain predominantly within schemata Thompson
associated with a given level (of control), his or her thinking can be said to be (1985, p. 195)
figurative in relation to that level.
Adopting Thompson’s framing, a researcher drawing distinctions between Moore et al.
figurative and operative thought is thus an issue of characterizing whether an (2019b, p. 2)
individual’s meanings are tied to carrying out particular actions and their
results…
Operative thought quotes
For Piaget, knowledge is not a matter of pictures or sentences or symbolic data Piaget (2001,
structures… knowledge is fundamentally operative; it is knowledge of what to pp. 2–3)
do with something under certain possible conditions. Or it is knowledge of what
that thing will do under different conditions… operative knowledge consists of
cognitive structures.
…whereas operative knowledge, which pertains to actions and operations, is Piaget (2001,
knowledge of transformations. Now, logico-arithmetic operations cannot be pp. 298–299)
rendered figuratively, unless this is done symbolically.
The operative aspect of thought deals not with states but with transformations Piaget (1970,
from one state to another. p. 14)
In contrast, any result of conceptual construction that does not depend on von
specific sensory material but is determined by what the subject does, is Glasersfeld
‘operative’ in Piaget’s terminology. (1995, p. 69)
(continued)
Table 4.1 (continued)

Figurative and operative thought quotes
Figurative thought quotes Source
The operative aspect on the other hand takes in those forms of cognitive Piaget and
experience or of deduction whose function consists of modifying the object in Inhelder
such a way as to apprehend transformations as such. This includes: (a) (1971,
sensorimotor actions (with the exception of imitation), the sole instruments of pp. 11–12)
sensorimotor intelligence to be organized before language; (b) internalized
actions that prolong previous ones right from a preoperational level; (c) the
operations proper of the representational intelligence, or reversible internalized
actions which organize themselves as a set of structures or as transformation
systems.
When the actions of thought move to the level of controlling schemata, then the Thompson
thinking can be said to be operative in relation to the level of the figurative (1985, p. 195)
schemata.
…[operative thought] foregrounds the coordination of internalized mental Moore et al.
actions so that figurative aspects of thought are subordinate to this (2019b, p. 2)
coordination… the individual can call forth and control a scheme and its results.
geometry (Piaget et al., 1981). Recall that to Piaget, objects are defined by opera-
tions of thought and thus the understanding of geometric objects involves moving
beyond the observable. Despite the ease at which the study of geometric objects can
be framed as the study of the observable, Piaget’s work underscores that the con-
struction of geometric objects involves dissociating operative forms of thought from
their figurative content so that the operative takes precedence over its figurative
content, forms, and states (Norton, 2022; Piaget et al., 1981). Take, for instance, a
circle. While the reader can likely evoke an image of a circle with ease, the mental
coordination of rotating a fixed length about a fixed point and anticipating the trace
of the rotating end-point defines the circle. We pull this tenet—one that involves
identifying whether it is figurative or operative aspects of thought that take prece-
dence in one’s meanings—through the rest of the chapter.
Transitioning the Constructs to Mathematics Education
Piaget and his colleagues’ work has influenced many mathematics educators’
research programs. Whether it be aspects of his genetic epistemology, his discussion
of schemes and operations, his identified stages of development, or his framing of
abstraction, Piaget’s body of work has proved fruitful in providing mathematics
educators with tools to research, model, and promote individuals’ mathematical
development. With respect to the figurative and operative distinctions, von
Glasersfeld (1987, 1995) and Steffe (1991a, b; Steffe & Olive, 2010), as close col-
laborators, were early adopters who did not make significant adaptations to the
distinctions.
von Glasersfeld (1995) incorporated Piaget’s distinction in his exposition on the

nature of knowing and learning in order to distinguish between two forms of thought
with categorical differences. Consistent with Piaget, he explained that figurative
thought involves sensation, perception, sensory objects, and physical motion, and is
thus directly related to any abstractions based in specific sensory material, including
motor signals. He contrasted this by explaining that operative thought involves con-
ceptual operations and their coordination. He explained that the enactment of such
operations occurs in the context of specific figurative material, their abstraction is
such that they are not dependent on specific sensory material, and these abstractions
can provide the figurative material for subsequent abstractions. At a more specified
level, one of von Glasersfeld’s (1987, 1995) primary interests was with representa-
tion, whether in terms of language, icons, or symbols. In reference to the figurative
and operative distinction, von Glasersfeld (1987) explained,
With regard to icons, Piaget’s distinction between the “figurative” and the “operative”
would seem to be of some importance. Number is not a perceptual but a conceptual con-
struct; thus, it is operative and not figurative. Yet, perceptual arrangements can be used to
“represent” a number figuratively. Three scratches on a prehistoric figurine, for instance,
can be interpreted as a record of three events. In that sense, they may be said to be “iconic”—
but their iconicity is indirect. They do not depict “threeness”, they merely provide the
beholder with an occasion to carry out the conceptual operations that constitute threeness
(Glasersfeld, 1981, 1982). Carrying out these operations does not involve reference to some
prior sensorimotor item or elements of such items—it is the operating itself that each time
constitutes the abstract conception of threeness…
An analogous distinction must be made in the case of symbols. On the one hand, there are
symbols that refer to figurative items or sensorimotor situations, such as the King of France
or the act of smoking; on the other, there are symbols that do not refer to sensorimotor
experience at all but are merely indicators that a certain conceptual operation is to be per-
formed. I would call this second category operative symbols and would list among them not
only number words, numerals, and mathematical signs, such as “+”, “—”, and “=”, but also
prepositions, conjunctions, and certain other words whose interpretation does not depend
on the recall of sensorimotor experiences but requires the construction of some operative
conceptual relation. (Glasersfeld, 1987, pp. 7–8)
von Glasersfeld (1987) described that an icon refers to another item or experience
via sensorimotor similarity, whereas a symbol refers to other items or experiences
more arbitrarily. With respect to both icons and symbols, von Glasersfeld drew on
Piaget’s figurative and operative distinction to categorize those icons and symbols
that form or refer to figurative items and sensorimotor situations, and those icons
and symbols that (possibly in addition to forming figurative items) occasion or point
to the coordination of mental operations.1 In von Glasersfeld (1995), he echoed this
distinction in reference to words. He explained that an individual’s meaning for a
word can be abstracted from sensorimotor experience and thus be figurative, or it
can point to a coordination of mental operations and thus be operative. In the con-
text of words, he added that a symbol or icon might merely cause a response in the
1
It is important to note that von Glasersfeld used symbol differently than Piaget as he described in
his writings on radical constructivism (von Glasersfeld, 1995).
form of a carried-out action, a case in which the symbol or icon acts as a signal as
opposed to an abstracted coordination of mental operations (Moore, 2014b;
Glasersfeld, 1995). von Glasersfeld’s application of Piaget’s distinction to icons,
symbols, and words is reflected in graphical associations related to static and emer-
gent shape thinking discussed in section “Informing generalized models of student
thinking”.
With respect to Steffe and his colleagues, they adopted the figurative and opera-
tive distinctions to characterize children’s development of schemes in the context of
counting and quantities, including length. With respect to the example of length,
Steffe and Olive (2010) identified that figurative length entails a visualized path,
motion producing the path, and some sense of the duration of the motion. Steffe and
Olive (2010) further identify that figurative length provides the foundation for an
individual to construct quantitative properties through the individual’s actions and
reflection upon those actions in order to construct a system of coordinated opera-
tions (i.e., an operative meaning). Steffe and Olive’s characterization of length
underscores that characterizing a meaning as figurative need not carry negative con-
notations. Instead, a figurative meaning may be a natural part of conceptual devel-
opment as an individual transitions from meanings based in perception, sensation,
and motor actions to meanings based in mental actions introduced and eventually
coordinated by the individual.
With respect to counting, and situated within children’s construction of number,
Steffe and colleagues (1991b; Steffe & Olive, 2010) distinguished between figura-
tive and operative counting schemes. Whereas figurative length primarily fore-
grounded notions of motion and duration, a figurative counting scheme requires
re-presenting perceptual material or engaging in some form of sensorimotor action
when counting.2 In order to support their counting of objects, especially if hidden, a
child with a figurative counting scheme will produce or imagine material that is
visual, such as an array of objects, or that is sensory, such as pointing in space or
tapping their finger (Steffe, 1991b). Steffe and Olive (2010) introduced the term
figural unit items for these re-presentational objects due to their basis in figurative
material and sensorimotor action. An operative counting scheme is one in which the
individual has constructed “a sequence of abstract unit items that contain the records
of the sensory-motor material” (Steffe, 1991a, p. 35). Such a construction occurs
through the individual coordinating the mental operations involved in counting so
that they no longer need to produce or imagine figurative material or sensorimotor
actions. The individual thus constructs unitized records of counting that they can
then operate on to construct more sophisticated notions of quantity. Important to the
figurative and operative distinction, including Thompson’s extension described
below, these unitized records or abstract unit items are not tied to specific
2
We use “re-present” (or “re-presentation”) and “representation” in distinct ways. Drawing on
Piaget and von Glasersfeld (von Glasersfeld, 1995), as well as Liang and Moore (2021), we use
“re-presentation” to refer to the enactment and regeneration of schemes and operations. We use
“representation” in the canonical sense to refer to the modes of display and symbolization associ-
ated with the field of mathematics (e.g., graphs, inscriptions, and verbal statements).
re-presentational actions and associated figurative material. The regeneration of

figurative material is necessary to re-present the records of counting, but it is the
abstract unit items and unitized records of counting that drive the regeneration of
that material.
Transitioning the Constructs to Higher Level Mathematics
For Steffe and colleagues, the figurative and operative distinctions proved fruitful in
describing number meanings and the extent figurative material or sensorimotor
activity were a necessary aspect of counting actions. Their distinction provided a
productive avenue to expanding Piaget’s discussion of concrete-operational intelli-
gence (e.g., thought involving logical operations but limited to concrete or physical
situations), which was sensible given that the development of number and counting
schemes was their content area of study. This distinction is not as useful when dis-
cussing students’ meanings for more advanced mathematics topics built on com-
plex, intertwined systems of meanings, representational activity, and symbolization.
Researchers have thus reframed the figurative and operative thought distinction as
illustrated by the work of Thompson, Moore, and colleagues (Liang & Moore,
2021; Moore, 2021; Moore et al., 2019a, b; Thompson, 1985).
Thompson (1985) described Piaget’s distinction as one of “the most significant
that I know of for mathematics education” (p. 194). He proposed generalizing
notions of figurative and operative thought to any level of meanings in order to
account for the developmental nature of mathematical meanings, and the fact that a
system of meanings at one developmental level can become the source material
(i.e., the figurative ground) for a system of meanings at a subsequent developmental
level. Thompson explained,
When a person’s actions of thought remain predominantly within schemata associated with
a given level (of control), his or her thinking can be said to be figurative in relation to that
level. When the actions of thought move to the level of controlling schemata, then the think-
ing can be said to be operative in relation to the level of the figurative schemata. That is to
say, the relationship between figurative and operative thought is one of figure to ground.
Any set of schemata can be characterized as figurative or operative, depending upon
whether one is portraying it as background for its controlling schemata or as foreground for
the schemata that it controls. For instance, the thinking of a college mathematics major in
an advanced calculus course, which certainly would be classified as being formal opera-
tional in Piaget’s fixed sequence of cognitive development, could nevertheless be classified
as figurative regarding the kind of thinking required in a graduate course in real analysis. Of
course, we would have to make apparent to ourselves the possibility that the “objects” of
such a student’s thinking are things like functions, classes of functions, and associated
operations. (Thompson, 1985, p. 195)3
3
Here, we believe Thompson’s use of schemata is consistent with scheme or meanings than his use
of schema (see Thompson, 2013b). For a more extensive discussion of schemes, see Tillema and
Gatza (Chap. 3).
Thompson’s discussion of operative and figurative thought reflects his interest in

characterizing complex systems of knowing that entail a web of meanings that build
on themselves during a student’s development. Importantly, this development can
occur both in the short-term and over a long period. For example, in the short-term,
a student could construct a directional relationship (see Carlson et al., 2002) between
two covarying quantities, which then can become the material on which they oper-
ate to construct an amounts of change relationship. Alternatively, over a longer
period, a student could construct particular additive partitioning schemes that then
become the source material for making multiplicative comparisons and constructing
rate of change meanings (Byerley, 2019; Thompson, 1994a). Importantly, by iden-
tifying that what is operative at one level can become figurative at the next level as
the individual attempts to operate on prior abstracted mental actions, Thompson
provides researchers a way to characterize students’ actions or meanings as figura-
tive or operative no matter the mathematical content or “object.”
Also expanding the applicability of figurative and operative thought, Thompson’s
reframing breaks free from a distinction resting on the availability or production of
figurative material and sensorimotor actions. Thompson instead foregrounds the
extent an individual’s meanings are tied to carrying out particular actions and
obtaining particular results, as opposed to their being able to transform their actions
to account for novel experiences and associated figurative material. Said another
way, Thompson’s framing differentiates the focus of reasoning being properties of
figurative material itself (including the results of activity) from the focus of reason-
ing being the mental actions and their coordination that generate associated figura-
tive material and results.4
Thompson broadened the figurative and operative distinction to allow for the
presence of figurative material (a point upon which we subsequently expand upon as
it relates to task design), but it is not incompatible with Steffe and Olive’s descrip-
tion of figurative counting schemes. A child with a figurative counting scheme is
constrained to carrying out particular actions, and specifically that of producing
sensorimotor actions or using available perceptual material. On the other hand, a
child having constructed a sequence of abstract unit items is in a position to use
those items in the presence of novel experiences and associated figurative material.
Thompson’s reframing is also faithful to tenets of Piaget’s use of figurative and
operative thought. Piaget included imitation in his characterization of figurative
aspects of thought but with a focus on sensorimotor reproduction. Piaget also noted
that figurative aspects of thought involve imitations of objects and the production of
configurations and observables. Thompson extends Piaget’s notions of imitation
and reproduction to include those meanings or actions that foreground imitating or
habitually reproducing previously enacted actions, whether mental or sensorimotor,
for the purpose of obtaining particular results and states. As Ellis et al. (Chap. 6)
explain, such meanings are often tied to abstractions based on the results or products
4
We note that this framing from Thompson is related to various forms of abstraction. For more
information on those forms, see chapters by Ellis et al. (Chap. 6) and Tallman (Chap. 8).
of mental actions, as opposed to the coordination of the mental actions that pro-
duced particular results and objects (Moore, 2014b; Moore et al., 2019a, b). With
respect to operative thought, Piaget’s descriptions foregrounded internalized actions,
organized and logico-mathematical structures, and an individual’s ability to reason
about transformations. This is consistent with Thompson’s use of operative, with
Thompson adding that what is operative on one level can become figurative in
nature during subsequent experiences and development. Importantly, Thompson’s
addition is consistent with the goal-directed foundation of Piaget’s genetic episte-
mology (Piaget, 1970; Steffe, 1991a; Glasersfeld, 1995).
Thompson’s reframing has enabled researchers to characterize marked differ-
ences in students’ meanings for the representations and symbolizations they con-
struct, including the extent those meanings are tied to carrying out particular actions
and their results. To illustrate the usefulness of Thompson’s framing of figurative
and operative thought in this regard, we first draw on results from investigations of
students’ graphical meanings and activity. This line of research has been prevalent
because graphing necessarily entails aspects of both figurative and operative thought
(Moore et al., 2019b).
Before we report on data and empirical studies that draw on Thompson’s fram-
ing, we provide a few concise definitions and resources (Table 4.2) that will be of
use to the reader. We include a list of relevant citations with each definition for the
reader interested in reading more detailed works that use the associated term or
construct. The connection between notions of figurative thought, operative thought,
quantitative reasoning, and covariational reasoning is not by chance. Quantitative
and covariational reasoning are apropos examples of operative thought, and specifi-
cally the enactment of logico-mathematical operations and associated meanings
(Moore, 2019b; Norton, 2014; Thompson, 1985). We attempt to provide examples
in the following sections so that they do not require a significant background in
quantitative and covariational reasoning, but familiarity with research in those areas
will undoubtedly assist the reader in situating our illustrations of figurative and
operative notions of thought.
Models of Students’ Graphical Thinking
Moore, Stevens, et al. (2019b) adopted Thompson’s (1985) reframing of figurative

and operative thought to characterize prospective secondary teachers’ (PSTs’)
graphical meanings. The authors did so in order to highlight a marked difference in
the PSTs’ actions as they engaged in solving particular problems involving the
interpretation or construction of graphical representations. The authors explained
that some PSTs’ actions suggested their enacting meanings were dominated by
fragments of sensorimotor experience and perceptual properties of shape, while
other PSTs’ actions suggested their enacting meanings were dominated by the coor-
dination of mental operations in the form of quantitative reasoning.
Table 4.2 Useful definitions

Construct Definition Other references
Quantity “A quantity is a quality of something that one has Smith III and
conceived as admitting some measurement process” Thompson (2007),
(Thompson, 1990, p. 5). Steffe, (1991b) and
Thompson (1993)
Magnitude “The idea of magnitude, at all levels, is grounded in the Liang and Moore
idea of a quantity’s size” (Thompson et al., 2014, p. 1). (2021)
“Magnitude refers to the size or amount-ness of a
quantity that remains invariant with respect to changes
in the unit used to measure the quantity” (Moore et al.,
2019b, p. 3).
Value (or “The numerical result of a quantification process applied Moore et al.
measure) to [a quantity]” (Thompson, 1990, p. 6). (2013), Smith III
and Thompson
(2007), and
Thompson (2011)
Number We use “number” to refer to when a student is Smith III and
referencing a numerical signifier that is not the result of Thompson (2007)
a quantification process applied to a quantity to obtain a and Thompson
measure. et al. (1994)
Quantitative We use quantitative operations to refer to both “the Liang and Moore
operations conception of two quantities being taken to produce a (2021), Steffe and
new quantity” (Thompson, 1990, p. 11) as well as the Olive (2010) and
operations that generate quantity or are involved in Thompson (1994a)
measuring a quantity, such as partitioning, unitizing, and
iterating (Steffe, 1991b).
Arithmetic or “[Operations] used to calculate a quantity’s value; there Ellis (2007a),
numerical is no direct correspondence, except in a canonical sense, Smith III and
operations between quantitative operations and the arithmetic Thompson (2007)
operations actually used to calculate a quantity’s value and Thompson
in a given situation” (Thompson, 1990, p. 12). et al. (1994)
Quantitative “A quantitative relationship is the conception of three Moore et al. (2013)
relationship quantities, two of which determine the third by a and Thompson and
quantitative operation” (Thompson, 1990, p. 13). Thompson (1996)
Quantitative “Quantitative reasoning is the analysis of a situation into Ellis (2007a),
reasoning a quantitative structure—a network of quantities and Moore (2014a) and
quantitative relationships” (Thompson, 1990, p. 13). Smith III and
Thompson (2007)
Covariational “…someone holding in mind a sustained image of two Carlson et al.
reasoning quantities values (magnitudes) simultaneously…one (2002), Johnson
tracks either quantity’s value with the immediate, (2012) and Paoletti
explicit, and persistent realization that, at every moment, and Moore (2018)
the other quantity also has a value” (Saldanha &
Thompson, 1998, p. 298).
Directional “Coordinating the direction of change of one variable Moore (2014a), and
change with changes in the other variable” (Carlson et al., 2002, Thompson and
p. 357). Carlson (2017)
(continued)

Construct Definition Other references
Amounts of “Coordinating the amount of change of one variable Ellis et al. (2015),
change with changes in the other variable” (Carlson et al., 2002, Johnson (2015) and
p. 357). Liang and Moore
(2021)
Moore, Stevens, et al. (2019b) provide an apropos example of Thompson’s

(1985) reframing because all PSTs in the study undoubtedly enacted mental opera-
tions when problem solving. Furthermore, they did so in the presence of available
or produced figurative material, thus not having to regenerate material entirely from
scratch. Accordingly, early distinctions of figurative and operative thought would
not have been productive in providing differentiated accounts of their actions. By
adopting Thompson’s framing, Moore, Stevens, et al. (2019b) differentiated whether
the PSTs’ actions foregrounded properties of figurative material itself, including
that which resulted from their activity, or whether their actions foregrounded the
coordination of mental operations not necessarily tied to particular properties of
figurative material. In all, they identified that some PSTs’ actions foregrounded
figurative aspects of thought, including requiring starting a graph on the vertical
axis, drawing or interpreting a graph left-to-right, drawing a graph that passes the
“vertical line test,” and drawing a graph that maintains some template or recalled
shape, to mention a few. On the other hand, they identified some PSTs’ actions that
foreground operative aspects of thought in the form of quantitative and covaria-
tional reasoning so that their construction and interpretation of graphs were domi-
nated by those operations no matter the resulting shape of their graph; the shape was
entirely a consequence of their coordination and tracking of quantities’ variation.
As an illustrative example, Moore, Stevens, et al. (2019b) identified some PSTs’
meanings that involved their perceiving “slope” in terms of left, right, up, and down
physical movements, and then associating paired movements with numerical prop-
erties for slope. For instance, one student had constructed the following (movement,
slope) pairs: (left-up, negative), (right-down, negative), (left-down, positive), and
(right-up, positive). Importantly, these (movement, slope) pairs were not dependent
on axes’ orientations or labeling, but instead properties of a line qua line. Moore,
Stevens, et al. (2019b) illustrated that such a meaning is viable when particular
Cartesian conventions are maintained, but its foregrounding of particular senso-
rimotor movement results in significant perturbation when those conventions are
not maintained.5 For instance, a PST in their study perceived the two displayed
graphs in Fig. 4.1 to have a positive and negative “slope,” respectively. She was
unable to reconcile to her satisfaction the perturbation that stemmed from each dis-
played graph having different “slopes,” yet each being associated with the same
formula that implied a positive “slope” of 3.
5
As Moore, Silverman, et al. (2019a) described, our use of “convention” here is not precise. What
the researchers perceived as a convention was certainly not a convention to the referenced partici-
pants of the study.
Fig. 4.1 Two displayed graphs of y = 3x, each with noncanonical axes orientations
Fig. 4.2 A displayed graph of y = 3x when scales are homogeneous; and y is oriented with positive
values horizontally to the right and x is oriented with positive values vertically downward, or y is
oriented with positive values horizontally to the left and x is oriented with positive values verti-
cally upward
On the other hand, Moore, Stevens, et al. (2019b) illustrated other PSTs’ mean-
ings involved their perceiving properties of “slope” as a consequence of rate of
change and its representation under the constraints of a coordinate system. These
PSTs thus associated slope with the coordination of quantities’ values and a com-
parison of amounts of change. Reflecting the generativity of operative forms of
thought, such a meaning proved viable for the PSTs no matter the coordinate orien-
tation, as their reasoning persistently foregrounded the coordination of quantitative
operations so that perceptual properties of displayed graphs were always a conse-
quence of those coordinations. For instance, a PST in their study conceived numer-
ous axes-orientations and labeling to achieve a positive “slope” with the displayed
graph in Fig. 4.2. The PST’s comment on a student who claims that the line has “a
negative slope” further illustrates the foregrounding of quantitative and covaria-
tional reasoning (Excerpt 4.1).
Excerpt 4.1 Annika Coordinating Two Quantities’ Values to Address a Student
Claim Annika: You’d have to notice that even though it looks like a negative slope
[making a hand motion down and to the right] because we call it slope because it’s
visual and it’s easy to visualize a negative and positive slope [making hand motions
to indicate different slopes]. But that’s only visual on our conventions of how we set
it up. Um, but like [pointing to the graph] if slope is rate of change we can still see
that for like equal increases of x [making hand motions to indicate equal magnitude
increases] we have an equal increase of y [making hand motions to indicate equal
magnitude increases] of three. And so for equal positive increase of one [sweeping
fingers vertically downward to indicate an increase of one], we have an equal posi-
tive increase of three [sweeping fingers horizontally rightward to indicate an
increase of three]. And so it is a positive slope. (Moore et al., 2019b, p. 12)
Informing Generalized Models of Student Thinking
Whereas Moore, Stevens, et al. (2019b) focused on students’ graphical meanings

enacted in-the-moment of problem-solving, the distinctions between figurative and
operative thought are also useful in articulating generalized models of students’
graphical meanings. These generalized models, referred to as epistemic subjects,
are conceptual models that specify categorical differences among students’ mean-
ings and ways of thinking (Steffe & Norton, 2014; Thompson, 2013a). Specifically,
the distinctions between figurative and operative aspects of thought informed Moore
and Thompson’s constructs of static (graphical) shape thinking and emergent
(graphical) shape thinking, respectively. Moore and Thompson (Moore, 2021;
Moore & Thompson, 2015) introduced static shape thinking as a way to refer to
meanings that involve conceiving a graph as if it is essentially a malleable piece of
wire (graph-as-wire). They introduced emergent shape thinking as a way to refer to
meanings that involve conceiving a graph as a trace entailing corresponding values
(or magnitudes) that are produced through the covariation of quantities.
Examples of static shape thinking include iconic translations—associations
between the visual features of a situation and those of a graph (Monk, 1992)—and
thematic associations—associating properties of event phenomena with shape prop-
erties of a graph that are superfluous to the stated quantitative referents (Thompson,
2016). An example of an iconic translation is a student perceiving a graph as con-
taining a circle because the situation entails a circle. An example of a thematic
association is a student perceiving an object traveling at a constant speed as neces-
sarily implying a line for a graph or an object traveling at a varying speed necessar-
ily implying a curve for a graph, regardless of the quantities being graphed. Other
examples of static shape thinking include facts of shape in which figurative
properties of a graph (or perceptual comparisons between graphs) imply particular

equations, analytic rules, names, mathematical properties, and so on. For instance,
an individual might associate a graph that curves up with an exponential relation-
ship, or an individual might associate a line with a linear relationship regardless of
the coordinate system. As another example, a student might associate the shape of a
graph with only one analytic rule, which constrains their ability to accommodate
novel graphing experiences (Moore, 2021). In each case (Fig. 4.3), the associations
foreground perceptual properties of shape and associations based on those properties.
Examples of emergent shape thinking include any instance in which an individ-
ual conceives a displayed graph as the product of coordinating and tracking two or
more varying quantities. This could occur in the context of a student conceiving a
displayed graph as having emerged through tracking quantities’ covariation, or it
could occur in the context of a student producing a displayed graph as a trace that
captures some conceived covariational relationship. The latter case is often for the
purpose of producing a displayed graph that is a quantitative model of some situa-
tion or relationship held in mind. Producing such a displayed graph involves the
student re-constructing the coordination of covarying quantities to produce a trace
they perceive as capturing a covariational relationship equivalent to the relationship
conceived in the situation or relationship held in mind (Fig. 4.4). Relatedly, a stu-
dent can reason emergently to compare and contrast displayed graphs in terms of
the covariational properties involved in producing their trace. For instance, a student
can reason emergently to conceive two displayed graphs as representing an invari-
ant covariational relationship despite perceptual differences between the graphs due
to their being represented under different scales (Fig. 4.5), orientations (Fig. 4.1), or
coordinate systems (Fig. 4.6) (Moore et al., 2013). Similarly, emergent shape think-
ing could involve a student conceiving a displayed graph as being produced by two
different covariational relationships despite the result of each trace being perceptu-
ally equivalent (Fig. 4.7, left and right, each illustrating a progressive trace). In each
case, any association a student makes with a graph is an implication of the covaria-
tional relationship they understand the displayed graph to have emerged as a trace of.
The relationships between Thompson’s framing of figurative and operative
thought and Moore and Thompson’s graphical shape thinking constructs are appar-
ent by the object of reasoning—a displayed graph—and its associations. With
respect to static shape thinking, such a way of thinking foregrounds figurative
aspects of thought; static shape thinking entails actions based in perceptual cues and
properties of shape. A displayed graph is essentially an object in and of itself, and
associations with the displayed graph are learned facts of the shape qua shape. Such
associations are consistent with von Glasersfeld’s (1987, 1995) discussion of figura-
tive icons, symbols, or words. In the case of constructing a displayed graph, a stu-
dent thinking statically carries out activity for the purpose of drawing a shape
consistent with learned or memorized associations. Those learned or memorized
associations are not conceived as organic or inherent to the production of the graph,
but rather they are simply taken as assigned to a produced shape that is perceptually
consistent with the association. The learned or memorized associations are abstrac-
tions based on the products of activity and particular states. Furthermore, because a
Fig. 4.3 Examples of static shape thinking and its implications

Fig. 4.4 Constructing a covariationally equivalent emergent trace
Fig. 4.5 Two displayed graphs of y = 45 sin(2πx)

Fig. 4.6 Two displayed graphs equivalent to the relationship defined by z = 2 k+1
student thinking statically foregrounds perceived properties of a produced or antici-

pated shape, they can reject a displayed graph generated via reasoning emergently
when the resulting properties of the produced shape are incompatible with the stu-
dent’s static shape thinking (see Polly in Moore et al., 2019b).
With respect to emergent shape thinking, such a way of thinking foregrounds
operative aspects of thought; emergent shape thinking entails the coordination of
mental operations in the form of quantitative and covariational reasoning. A dis-
played graph is understood as reproducible through the coordination of the mental
operations that resulted in its trace, and associations with the graph are anticipated
as the mathematical properties organic to the operations that produce it. Any learned
or memorized associations are understood as properties of or symbolizing the coor-
dination of operations represented by the displayed graph. Such associations are
consistent with von Glasersfeld’s (1987, 1995) discussion of operative icons, sym-
bols, or words. Furthermore, because a student’s thinking emergently foregrounds
the coordination of mental operations, their meanings are positioned to accommo-
date novel graphing experiences. The coordination of those mental operations and
their perceived mathematical properties provide actions of thought that can be trans-
formed into novel figurative material, all the while maintaining invariance in the
logico-mathematical form. That is, emergent shape thinking enables the enactment
of operations that can differ in terms of their figurative entailments yet be compared
on the basis of similarities and differences in the mathematical properties of those
operations.
Adapting the Distinctions to Other Representations
Thus far, we have illustrated examples of figurative and operative thought through
graphical representations and phenomena. However, definitions of figurative and
operative thought have been adapted in ways to address other representations. In
Fig. 4.7 One completed displayed graph, two covariational relationships (left, r = sin(2θ); right,
r = |sin(2θ)|)
this section, we consider these adaptations when discussing students’ reasoning

with formulas.
As a reminder, Thompson (1985, p. 195) said, “Any set of schemata can be char-
acterized as figurative or operative, depending upon whether one is portraying it as
background for its controlling schemata or as foreground for the schemata that it
controls.” In considering how to adapt the constructs of figurative and operative
thought in a new representation, one needs to consider what mental operations are
involved in students’ acts of re-presentation. With formulas, a researcher character-
izing an enacted meaning does not rely on perceptual actions taken on material,
such as tracing right to left as Moore, Stevens, et al. (2019b) described, or repeating
partitioning activities across contexts as Liang and Moore (2021) described.
Whereas graphical representations and phenomena provide material to engage oper-
atively or figuratively, formulas merely provide symbols in the form of inscriptions
or glyphs. Thus, a researcher characterizing a student’s formula meanings as figura-
tive or operative relies on the associations evoked by a symbol (or collection of
symbols) in a formula.
In a teaching experiment in which Stevens (2019) attempted to understand and
support students’ meanings for formulas through reasoning with dynamic situa-
tions, she built upon research involving the Piagetian constructs of multiplicative
objects and figurative and operative thought to model students’ ways of thinking
about graphs. Her strategy of the experiment focused on designing tasks that would
enable her to construct models of students’ mathematics by identifying opportuni-
ties for students to reason quantitatively and covariationally with formulas, and her
analysis focused on characterizing students’ mental actions in terms of the nature of
associations evoked by given or constructed formulas. In particular, she adopted the
constructs of figurative and operative thought to differentiate between associations
that stemmed from perceptual similarities or learned facts and those that stemmed
from enacting quantitative operations and abstracting their mathematical properties.
To discuss figurative thought in more detail, consider how students might answer
the following prompt: “Describe a situation in which the formula A=2πrh describes
a relationship between quantities. How does your situation describe that relation-
ship?” The meanings students have for the formula A = 2πrh dictate the associations
they will draw between a situation and the formula. Consistent with static shape
thinking examples above, a meaning foregrounding figurative aspects of thought
would foreground associations between attributes or shapes (e.g., area of circle, or
surface area of a cylinder or spherical cap) and aspects of the symbols constituting
the formula. These associations may or may not have quantitative entailments, but
importantly those entailments are not inherent to the association. For example, a
student may conceive the formula A = 2πrh as the normative formula for the surface
area of an open cylinder, and only an open cylinder (see Stevens (in press) for this
example). As a second example, a student might also identify “2π” as necessarily
pointing to a feature of a circle and, thus, that the relevant situation must include a
circle. As Stevens (in press) illustrated, a student stated that A = 2πrh could not
represent a formula associated with a parallelogram. They explained, “Because I
don’t know what it [2π] would represent in a parallelogram. Like, in a circle, it’s
because you can like divide a circle into two pi radii, but you don’t have anything
even here that you could do that with.” In both examples, the reasoning foregrounds
associations between the arrangement or presence of the glyphs in the formula and
particular shapes. Thus, in the event that the associations do have quantitative entail-
ments, those entailments are constrained to the objects of association.
Alternatively, operative thought would entail meanings that focus on the formula
as representing quantitative relationships that could be re-presented in a variety of
contexts. As an example, A = 2πrh represents a linear relationship between height
and area that could be relevant to a rectangle, cylinder, or spherical cap, depending
on what quantities the symbols represent. A student could thus generalize that
A = 2πrh means that for equal changes in height, the surface area of the relevant
shape, whatever it is, increases by an amount that is always 2πr times as large as that
change in height. Stevens (2019) described a student who reasoned that such a rela-
tionship was implied by the formula, and the student then generated that relation-
ship using cylinder and spherical cap contexts. Figure 4.8 provides an illustration of
this way of reasoning; the corresponding strips of the cylinder and spherical cap
have equal surface area. Consistent with the notion of operative thought, the stu-
dent’s reasoning foregrounds quantitative relationships, and the coordination of
quantitative operations drives their activity and products.
It is important to note that figurative associations, such as associating πr2 with the
area of a circle, can be useful in constructing a quantitative structure in a context.
Similar to the notion of expert shape thinking (Moore & Thompson, 2015), as long
as a student can unpack those associations quantitatively (similar to the reasoning
illustrated in Fig. 4.8), while also anticipating that there can be other potential asso-
ciations. Such associations reduce cognitive effort; it is inefficient to always enact
quantitative operations.
It is also important to note the effects of foregrounding figurative associations in
constructing formulas. A student who constructs figurative associations between
geometric shapes and formulas in a way that dominates any quantitative entailments
may attempt to identify perceptual features in the context and then attempt to incor-
porate formulas associated with those figurative elements in their construction of a
formula, resulting in a nonquantitative formula. For instance, in Stevens (2019), a
student, Kimberley, identified both a dynamic cone and cylinder as including circles
and varying height. She concluded that only one should be able to be associated
with the formula A = 2πrh. She eliminated the cone by foregrounding figurative
aspects of thought and breaking the cone into particular shapes that she could then
associate with area formulas. As a result, Kimberley’s formula for the cone com-
bined her formulas for the area of the circle with the area of a triangle (Fig. 4.9).
In summary, although mathematics education researchers have primarily focused
the figurative and operative distinctions on representations or phenomena in which
students can enact quantitative operations, the distinctions are also applicable to the
Fig. 4.8 An example of operative thought with A = 2πrh by considering A and h as variables to
conclude that for equal changes in height, there are equal changes in surface area for cylinder and
spherical cap (A = surface area, r = radius length of circle/sphere, h = vertical height)
Fig. 4.9 Example of figurative thought used to construct formula for surface area of a cone by
multiplying together shapes in geometric figure
associations students construct for symbols, groups of symbols, or formulas.

Furthermore, such associations are consistent with von Glasersfeld’s (1987, 1995)
broader discussion of icons or symbols with attention to the extent a formula evokes
a learned association or a set of anticipated quantitative operations. Incorporating a
focus on formulas also helps illustrate that although associations that are figurative
in nature (e.g., A = πr2 provides a way to calculate the area of a circle) are beneficial
for reducing cognitive load, it is important that such associations are not constructed
at the loss of formulas representing a quantitative relationship specific to a context
or particular object (e.g., A = πr2 representing that the area of a circle increases by
increasing amounts as the radius increases by a constant, successive amount).
Furthermore, it is important that such operative associations are not constructed at
the loss of a formula potentially representing a variety of quantitative relationships
(e.g., A = πr2 is the area of a rectangle with a side π times as large as the other side).
In summary, operative meanings that foreground symbols and groups of symbols in
formulas as pointers to quantities, quantitative operations, and covariational opera-
tions are generative in that they enable a student to conceive formulas as relevant to
both experienced and unexperienced situations, all the while understanding the for-
mula captures some invariance property across those situations.
ransitioning the Constructs Back to the Study

T
of Meaning Construction
The examples in the prior section used figurative and operative distinctions when
discussing undergraduate students’ reasoning with representations (e.g., graphs and
formulas). The aforementioned work occurred after students had constructed par-
ticular meanings for those representations through schooling; the studies did not
address students’ initial construction of graphical meanings. In this section, we
illustrate the extension of these constructs to the construction of graphs, which was
led by Tasova (2021). Tasova (2021) adopted figurative and operative thought to
characterize middle-school students’ graphing meanings as they engaged in solving
problems involving the construction of graphical representations of varying quanti-
ties for what appeared to be the first time (at least within formal settings). In doing
so, he illustrated the viability of explaining middle-school students’ reasoning with
the figurative and operative distinction, while also adapting those distinctions to
explain his data. Ultimately, he identified meanings compatible with figurative or
operative aspects of thought. Underscoring von Glasersfeld’s (1995) framing that
operations always have to operate on something, Tasova also identified important
meanings that entailed aspects of both figurative and operative thought in the same
activity. We summarize those findings here.
Before supporting middle schoolers in developing meanings for graphs in two-
dimensional space, Tasova (2021) provided them opportunities to engage with
quantities’ magnitudes represented by varying lengths of directed bars placed on
empty number lines (also called magnitude lines, see Fig. 4.10, right). Students had
opportunities to use these dynamic segments to represent individual quantities in
one-dimensional space. As one example, students played the bike animation (see
Fig. 4.10, left) and pulled the blue segment in ways that could represent how the
bike’s distance from Arch (DfA) varied in the situation.
Tasova (2021) identified two meanings for the dynamic segments that align with
figurative and operative aspects of thought. An individual who conceives the bike’s
DfA in the situation as decreasing and increasing might vary the length of the seg-
ment on the magnitude line accordingly. This is an example of an operative aspect
of thought as the individual disembeds the conceived quantity’s magnitude from the
situation, re-presents it on the magnitude line, and then coordinates the length of the
Fig. 4.10 Downtown Athens Bike Task and magnitude line

blue bar, ensuring to preserve its length as the animation played. As an alternative
conception, an individual can simulate the physical bike on the magnitude line. For
example, they might coordinate the direction they “pull” one end of the blue seg-
ment with the bike’s right and left movements, thus associating the movement of the
bike with the movement of the segment endpoint. Tasova identified this as an exam-
ple of a figurative aspect of thought as the student’s thought is dominated by imitat-
ing the moving object in the situation.
As a second example of using the figurative and operative distinction to charac-
terize a nuanced difference in middle-school students’ actions, we use another case
that involves re-presenting a quantity on the magnitude line. In this case, the bike in
the animation and the dynamic segment on the magnitude line are synchronized—
the length of the blue bar on the magnitude line varies according to the bike’s move-
ment on the map by design (see the animation at the following link: https://youtu.
be/6kdbDeVEF9w). For example, while moving the bike to the right from its posi-
tion (Fig. 4.10, left), the right end side of the blue bar on the magnitude line moves
to the left (indicating the bike’s DfA is decreasing). This is different from the previ-
ous task in that the endpoint moves in the opposite direction of the bike.
Explaining how this could happen, a seventh-grade student referred to the map
and claimed, “The bike is getting closer to Arch.” Then, by pointing to the blue bar
on the magnitude line and tracing the pen over the line from right to left, the student
said it “is gonna get closer to right here [pointing to the zero point on the magnitude
line], which is Arch.” Tasova argued that the student was not conceiving the bar as
a varying length, but instead that the student conceived of the Arch and bike as
physically placed on the left and right endpoints of the blue bar, respectively. He
thus perceived the blue bar’s change to be a product of the bike moving closer in
proximity to the Arch along the blue bar. Tasova contrasted this with a response
from another student who first determined that the bike’s DfA is decreasing, while
moving the bike to the right on the map. The student then conceived the length of
the blue bar on the magnitude line as a re-presentation of the bike’s DfA, thus
requiring that the length decrease to remain equivalent to the length in the situation.
She explained, “it [pointing to the blue bar] is gonna get smaller because distance is
smaller on the number line too.” Moreover, she labeled the starting point as “zero,”
whereas the previous student conceived the same point on the magnitude line
as “Arch.”
Reflecting on the two students’ actions, they are compatible in that they each had
the same result from an observer’s perspective: the students moved the blue bar
endpoint so that it decreased in length. But their meanings were notably different.
In the first case, and consistent with reasoning that foregrounds figurative notions of
thought, the meaning foregrounded the spatial proximity of two objects and moving
them accordingly. Any change in the blue bar came as a consequence of changing
their spatial proximity. In the second case, and consistent with reasoning that fore-
grounds operative notions of thought, the meaning foregrounded the distance
between two objects. The length of the blue bar was persistently in mind, and the
student’s goal was to maintain a length equivalent to that in the situation. Although
these differences are subtle, Tasova (2021) illustrated they have significant
Fig. 4.11 (a) DABT, (b) Zane’s graph, and (c) Melvin’s graph
consequences for students’ graphing actions. To illustrate, we use two middle-

school students’ activity during the Downtown Athens Bike Task (DABT).
DABT prompts students to use a Cartesian coordinate system to draw a sketch of
the relationship between the bike’s DfA and the bike’s distance from Cannon (DfC)
as the bike moves on Clayton St. in Downtown Athens (see Fig. 4.11a). The bike
starts from the West side of the street and moves at a constant speed. To illustrate an
example of figurative thought, we draw on Zane’s (a seventh grader) activity in
which his graphical meanings representing literal pictures of a particular situation.
Zane conceived the Arch and Cannon placed at physical locations on the vertical
and horizontal axes, respectively, as implied by the labels (see blue and purple dots
on each axis for Arch and Cannon, respectively, in Fig. 4.11b). Because Arch is at
the top and Cannon is at the bottom of the plane, Zane then imagined rotating the
plane in a way that Arch is at the bottom and Cannon is at the top, just as with the
map. He then drew a segment on the plane, identifying it as the path of the bike. He
also added dots on the plane (i.e., one that he labeled “Bike” and the other at the
origin), which he placed to match the dots at each end side of the bike’s path on the
map (see Fig. 4.11a). Zane went to the extent of using the same coloring for the seg-
ment on the plane as that on the map (i.e., light blue). Drawing on the notion of
iconic translation (Clement, 1989; Monk, 1992), Tasova (2021) characterized
Zane’s meaning as transformed iconic translation. Zane translated a transformed

version of perceptual features of the situation to the plane as he rotated the plane and
overlaid it into the map in order to perceptually match the graph that he drew on the
plane with the bike’s path on the map.
Whereas Zane’s example is a rather typical example of a graphical meaning that
foregrounds figurative aspects of thought, Melvin’s activity in DABT provides a
novel illustration of how an individual’s meaning might involve both figurative and
operative aspects of thought. Melvin conceived the entire vertical and horizontal
axes of the plane as the physical Cannon and Arch, respectively. The axes did not
represent magnitude or number lines, but instead were referents to the objects them-
selves. Melvin then drew a line upward from left to right to represent “where the
bike travels.” He added tick marks and dots on his line graph “to represent like
where the bike could be.” For example, when the bike is at location 2 in the map
(i.e., DfA and DfC are each at their minimum), he said, “It [the bike] would go right
here [pointing to the tick mark on his graph near 2 in Fig. 4.11c].” Similar to Zane,
and consistent with reasoning that foregrounds figurative aspects of thought, Melvin
conceived each point on his graph as the physical bike moving on its path on
the plane.
Differing from Zane, Melvin also imagined moving the bike on the plane accord-
ing to the variation of its DfA and DfC. He conceived DfC and DfA as being re-
presented by vertical and horizontal segments drawn on the plane, respectively. The
vertical segments on the plane represented the bike’s DfC because Melvin assimi-
lated the horizontal axis as a reference ray that he measured the bike’s DfC from.
Similarly, the horizontal segments on the plane represented the bike’s DfA because
he assimilated the vertical axis as a refence ray he measured the bike’s DfA from.
Increasing length of the horizontal and vertical segments on the plane indicated an
increase in the bike’s DfA and DfC. Moreover, Melvin understood that the length of
the horizontal segment is shorter than the length of the vertical segment for each
point on the graph as the bike’s DfA is always less than the bike’s DfC on the map.
Thus, although Melvin assimilated his graph as “where the bike travels,” Melvin’s
graph was not the bike’s path as it is seen on the map; it was not an iconic transla-
tion. Melvin’s meaning of the points included determining quantitative features of
the bike in the situation (i.e., its DfA and its DfC), and he ensured they preserved
those quantitative properties on the plane.
We consider Melvin’s graphing activity as a different way of graphing relation-
ships because his initial activity foregrounded figurative aspects of thought, but he
then engaged in quantitative operations to make sense of the results of that initial
activity and produce a graph to his satisfaction. Specifically, he represented the
quantities’ magnitudes in the space by committing to two frames of reference (i.e.,
the axes), and then placing a point where those magnitudes meet on the path he
produced. Tasova and Moore (2020) termed this meaning of points, which has both
operative and figurative entailments, as a spatial-quantitative multiplicative object,
which involves an individual envisioning a point on the plane as a location/object
that also entails quantitative properties. This example underscores that the figurative
and operative distinction is not merely binary with respect to students’ meanings.
As we discuss below, we believe this meaning, along with other meanings Tasova
identified, illustrates the need to better understand how figurative and operative
notions of thought play a role in students’ construction of meanings that serve as
both immediately and longitudinally productive.
Implications for Methodology and Task-Design
In addition to providing tools to construct second-order models of students’ math-

ematics, as illustrated above, we have, in turn, found the figurative and operative
distinctions useful in designing empirical studies. Specifically, we have found the
distinctions useful in designing tasks and framing evidence for students’ quantita-
tive and covariational reasoning. Their usefulness in this regard can be organized by
Moore et al. (2022) notion of an abstracted quantitative structure (AQS), which the
authors introduced as criteria for concept construction from a quantitative reasoning
perspective. They described an AQS as a system of quantitative operations that an
individual has interiorized so that it:
C1. is recurrently usable beyond its initial experiential construction;
C2. can be re-presented in the absence of available figurative material including that in
which it was initially constructed;
C3. can be transformed to accommodate to novel contexts permitting the associated quan-
titative operations;
C4. is anticipated as re-presentable in any figurative material that permits the associated
quantitative operations. (Moore et al., 2022).
The criteria were directly informed by the figurative and operative distinctions
(Moore et al., 2022). C2 reflects the distinction between figurative and operative
counting schemes as discussed by Steffe and colleagues (Steffe & Olive, 2010), as
well as Glasersfeld’s (1995) distinction between recognition and re-presentation.
C3 captures the distinction introduced by Thompson (1985), allowing for the pres-
ence of figurative material and resting on the extent an individual is able to trans-
form their actions to account for novel experiences and associated figurative
material. C4 further extends C2 and C3 in a way that combines aspects of both
Steffe’s and Thompson’s framings. Reflecting Steffe’s framing, C4 involves an indi-
vidual anticipating particular quantitative operations and their mathematical proper-
ties in the absence of figurative material. Reflecting Thompson’s framing, C4 also
involves the individual understanding that those operations and their mathematical
properties are generalizable to any instance in which those operations are viable,
thus allowing the individual to anticipate their enactment in the presence of novel
figurative material. Such anticipation enables us to attribute particular quantitative
and covariational relationships to quantities and figurative material too complex for
us to enact those operations within. For instance, temperature is far too quantita-
tively complex for most individuals to enact particular quantitative operations, but
we can anticipate variations in temperature and use proxies (e.g., values, number
lines, and coordinate systems) to represent those variations in a way that enable us
to enact quantitative operations as if we are reasoning about temperature.6
As mentioned above, Moore et al.’s (2022) main intention of introducing the
construct of an AQS was to provide guiding criteria for concept construction, but it
also provides a framing for designing empirical studies and situating researcher
claims regarding students’ quantitative and covariational reasoning. Said succinctly,
a researcher’s claims regarding a student’s quantitative and covariational reasoning
are necessarily constrained by the nature of the tasks they use with a student.7 To
illustrate this, we discuss each framing criteria above with respect to task design and
potential claims regarding students’ quantitative and covariational reasoning.
C1 refers to a quantitative structure that is recurrently usable beyond its initial
experiential construction. From a researcher’s perspective, and when considered
independently of C2–C4, C1 is the most trivial of the criteria because it refers to an
individual re-enacting a previously constructed quantitative structure in the pres-
ence of previously experienced context or figurative material. Thus, from a task-
design perspective, this involves engaging the student in a task the student has
already experienced and enacted the relevant quantitative operations. Here, “the
same task” need not mean a carbon-copy of the previous task. Rather, “the same
task” encompasses the figurative material provided and the quantitative operations
needed to solve the task in the way the researcher hypothesizes. The researcher
might change aspects of the tasks they consider to be surface-level features like
values (e.g., the numbers being summed in a counting task) based on their model of
the student’s mathematics and what might be perceived as the same from the stu-
dent’s perspective.
A contraindication of C1—the student not being able to re-enact particular oper-
ations in the presence of an identical task—is evidence of the absence of quantita-
tive or covariational reasoning. We do not consider indications of students having
constructed an AQS consistent with C1 to enable strong claims regarding a student
reasoning quantitatively or covariationally. Indications of C1 can be produced by a
student mimicking or reproducing memorized actions from previous experiences.
In such cases, a student is able to rely on the available figurative material and recall
traces of activity from previous experience to reproduce previous results. However,
to the researcher, the observable actions of such a student might be indistinguish-
able from those of a student enacting quantitative operations. It is thus necessary to
design tasks that enable differentiating from a student mimicking or reproducing
6
This underscores one of the more critical roles graphing plays in mathematics. Coordinate sys-
tems and their graphs provide quantities (e.g., oriented lengths or angle-openness) of which indi-
viduals can enact quantitative operations, thus serving as useful proxies for those quantities on
which we cannot enact quantitative operations.
7
It is more accurate and compatible with our theoretical perspective to say that a researcher’s
claims regarding a student’s quantitative and covariational reasoning is necessarily constrained by
the observable actions of the student’s engagement. For the purposes of this chapter, it is more
straightforward to discuss tasks as if they have some inherent or objective design feature, but we
remind the reader that no such features exist and a task is always defined by the student’s
engagement.
memorized actions from a student whose reasoning foregrounds quantitative opera-

tions, which leads us to C2.
Much like C1, C2 is based on a student being able to re-enact particular quantita-
tive operations. Unlike C1, the re-enactment comes in the absence of all or some
subset of figurative material. With respect to a counting situation that involves an
individual summing two collections of stones, the researcher might adjust the task
so that one collection or both collections of stones may be hidden from the student
with them only being told the numerosity of the hidden collections. As another
example drawn from our own work, after a student constructs a particular relation-
ship (e.g., the sine relationship) in the context of a Ferris wheel rider (Fig. 4.12), the
researcher might prompt the student to re-construct that relationship with no figura-
tive material provided (Fig. 4.13). Or, the researcher might provide a subset of figu-
rative material as we did with the Which One? task in which a participant is asked
to choose which red segments, if any, appropriately represent the height above the
center (Liang & Moore, 2021, p. 300) (Fig. 4.14, also see the following link: https://
youtu.be/2pVVGl8eEr0). Note that these collections of tasks vary with respect to
what is perceptually available to the student and what the student might be asked to
mentally (or physically) generate, thus supporting a researcher in characterizing a
student’s re-presentation capacity.
In the Which One? task, we removed the Ferris wheel spokes because those
spokes provide figurative material that can assist with partitioning the rider’s trip
Fig. 4.12 Amounts of change and the sine relationship for π/2 radians of rotated arc
Fig. 4.13 A task with no provided figurative material
Fig. 4.14 The Which One? task environment as presented to students (Liang & Moore, 2021,
p. 300), in which they are prompted to determine which red segments, if any, represent the point’s
height above the center of the Ferris wheel in relation to the arc length traversed by the point
into equal increases in arc length and corresponding variations in height. We also
designed the task to include a digital environment in which they did have the avail-
ability of red segments, and these red segments could be re-oriented and embedded
in the circle if the participant chose to do so. However, we designed the task so that
the participant could not physically draw within the environment. Thus, as they
moved the point around the circle, the chosen red segment varied accordingly, either
in its given position or in a re-oriented position if they chose to do so. This required
that the participant be able to hold in mind different states of the red segment and
mentally produce amounts of change in that segment without the assistance of con-
stantly available figurative material (Fig. 4.15).8
A contraindication of C2—the student not being able to re-enact particular oper-
ations in the absence of figurative material—is evidence that their meanings are tied
to particular figurative material and carrying out activity tied to perceptual proper-
ties of that material. On the other hand, indications of students having constructed
an AQS consistent with C2 enables a researcher to make relatively stronger claims
8
The reader might note that the correct red segment can be determined solely through identifying
that it is the correct length at each state. If a student bases their choice by checking states, we ask
“Show us how to see the amounts of change in the red segment?” in order to prompt them to
describe or generate partitions illustrating appropriate amounts of change.
Fig. 4.15 The Which One? task and re-positioning a red segment as the object moves around
the circle
regarding a student reasoning quantitatively or covariationally. An indication of C2

provides some evidence that the individual is able to re-enact quantitative opera-
tions to generate or anticipate figurative material that is no longer available.
However, like C1, in our experience, students are able to provide evidence of C2 by
what amounts to mimicking previous activity in order to produce absent figurative
material. The production of that material often provides them source material for
reflection and comparison to recalled previous activity and its results. Although the
individual might enact quantitative operations in reflection and with the assistance
of the material they have produced, this is not consistent with C2 as the operations
enacted in reflection were not driving the production of the figurative material.
Whereas C2 involves an individual re-enacting quantitative operations in con-
texts previously experienced, C3 captures instances of an individual accommodat-
ing previously enacted quantitative operations to account for novel contexts. Much
like a researcher determining what counts as “the same task” is dependent on their
model of a student’s mathematics, what a researcher determines to be “a novel con-
text” is also dependent on their model of a student’s mathematics. A novel context
to a student could be a new phenomenon, a graphical representation, a graphical
representation under a different coordinate orientation, or a graphical representation
under a different coordinate system, to name a few (Fig. 4.16) (see the following
studies for a collection of tasks designed with C3 in mind: Liang & Moore, 2021;
Moore et al., 2013, 2019a, b; Paoletti et al., 2018). In some instances, such as chang-
ing a phenomenon or graphical coordinate system, what is novel may be the phe-
nomenon and quantities under consideration. In other instances, such as considering
different coordinate orientations, what is novel may be the figurative orientation of
the quantities due to a change of referent. Regardless, a task designed with C3 in
Fig. 4.16 Various task-design options in order to transition from a previously experienced context
to a potentially novel context
mind affords the student the opportunity to re-enact previously enacted quantitative
operations in the presence of different quantities or figurative orientations with the
goal of perceiving an invariance across the differing contexts.9
9
We note that C2 can be incorporated when designing tasks with C3 in mind. As an example, a
researcher might intend to gain insights into whether a student can conceive a graphical representa-
tion as quantitatively equivalent to some relationship they conceive in a phenomenon. Taking C2
into account, the researcher might ask the student to construct that graphical representation, or they
might provide figurative material in the form of displayed graph or graphs and prompt the student
to determine which graph(s) accurately represent the relationship.
Fig. 4.17 A displayed graph emerging in a way that is invariant with that of the phenomenon
A hallmark of operative thought, per Thompson’s (1985) framing, is that the

coordination of mental operations and their transformation dominate an individual’s
activity. For that reason, indications of C3 provide a researcher with a strong evi-
dentiary basis for claiming a student’s reasoning foregrounds quantitative and
covariational reasoning. Due to figurative differences across previously experienced
and novel contexts, a researcher implementing tasks with C3 in mind is positioned
to gain insights into whether a student can conceive quantitative invariance across
those contexts despite figurative differences in their actions (Fig. 4.17). A researcher
is simultaneously positioned to gain insights into when a student is mimicking pre-
vious activity. In instances in which a student provides a contraindication of C3, it
suggests that there is some aspect of their meaning tied to figurative aspects of the
contexts in which they previously enacted that meaning. This is not to say that the
student’s reasoning in those prior contexts was not quantitative, but rather that some
aspect of their reasoning was reliant on particular figurative features of that context
and they thus looked to maintain those figurative features. For instance, in transi-
tioning from a phenomenon like a Ferris wheel ride, the student might produce a
graph from memory and then try to maintain partitions along a curve because there
are partitions along a curve in the phenomenon (Fig. 4.18). Here, the graph does not
emerge as a product of quantitative operations, but rather is produced to provide
material to mimic activity or enact operations. As we discuss in the following sec-
tion, students’ attempts to maintain both quantitative and figurative aspects of their
actions often lead to experienced perturbations due to incompatibilities between
those maintained aspects.
Because C4 depends on an individual’s capacity to anticipate some structure of
quantitative operations, designing tasks incorporating this principle is complex.
Furthermore, our research team has not concentrated on designing such tasks to the
extent we have with that of C1–C3. In working with PSTs, one fruitful approach is
designing tasks to include hypothetical student work that is nonnormative. Doing so
provides a researcher insight into whether or not the PST anticipates that the hypo-
thetical student might have produced the work through the viable enactment of the
relevant operations. In the case that the PST does anticipate such an event, they may
then attempt to determine that viable enactment (i.e., C3). In the case that the PST
does not anticipate such an event, the PST is likely to enact meanings that identify
some incorrect element of the student’s work. Annika’s response to Fig. 4.2 and
determining numerous viable ways the figure is a displayed graph of y = 3x is an
example of the former. An example of the latter would be a PST immediately reject-
ing Fig. 4.1b as a potential displayed graph of y = 3x because of it sloping down-
ward left-to-right, or their rejecting Fig. 4.1a because the axes are incorrect (see
Moore et al., 2019a for other examples).
Because C4 itself does not entail the enactment of quantitative operations, we do
not consider indications of it to be evidence for quantitative or covariational reason-
ing. Rather, indications of C4 suggest that the individual has abstracted meanings
that foreground quantitative and covariational reasoning to the extent that the rele-
vant system of operations has become a way of thinking (Harel, 2008a, b; Thompson
Fig. 4.18 Student work in which they drew a graph from memory and then attempted to re-
produce partitions that maintained orientations and placements along segments or curves (Liang &
Moore, 2021, p. 306)
et al., 2014). On the other hand, contraindications of C4 imply the individual has
abstracted meanings that could entail quantitative and covariational reasoning, but
they are such that they foreground some other aspect or form of thought. As the
individual then enacts that meaning to make sense of the task, the researcher gains
insights into the operations and associations constituting that meaning.
Moving Forward
The progress of mathematics education research is contingent on researchers push-

ing into new areas. In some cases, such pursuits require adopting or developing new
theoretical perspectives or constructs. In other cases, such pursuits involve testing
and adapting available theoretical perspectives or constructs. The work described in
the previous sections is an example of the latter. Such an orientation toward theory
stresses that theoretical perspectives and constructs should be pushed into new areas
of study in an attempt to put that theory to the test. Doing so enables testing the
viability of the theory and expanding its generalizability through adaptations that
respond to constraints experienced in applying or developing the theory. Thus, our
general suggestion for future work is that mathematics education researchers look
to test the viability of the figurative and operative distinction within whatever area
is of interest to them. Relatedly, we underscore that in pursuing such work, research-
ers keep in mind that they should not merely look to apply the distinctions as other
researchers have, but rather view the notions of figurative and operative thought as
malleable to their needs and experiences carrying out empirical work. Synthesizing
the opening quotes by Dewey and Thompson, the weight of theory is in its ability to
aid researchers’ pursuits of problems and explanations of phenomena.
With respect to specific suggestions for research, we provide two potentially
productive avenues. Our first suggestion is to continue the extensions taken by
Tasova (2021) and Stevens (2019) to investigate the ways in which the figurative
and operative distinction is relevant to students’ construction of mathematical ideas
and other representational systems. We envision at least two pivotal insights such
research might provide. Research along the suggested lines will provide insights
into how and the extent to which students construct meanings that transition from
foregrounding figurative aspects of thought to those meanings in which operative
aspects of thought dominate figurative entailments. Relatedly, such research will
provide insights into those experiences that support or inhibit students’ construction
of meanings that foreground either figurative or operative notions of thought.
Research along the suggested lines might also provide insights into individuals’
construction of incompatible or competing meanings. Tasova (2021) introduced the
term competing meanings when expanding on Moore, Stevens, et al.’s (2019b)
observation of a PST (Polly) having enacted two graphical meanings, one that fore-
grounded operative notions of thought and one that foregrounded figurative notions
of thought. In the case of Moore, Stevens, et al. (2019b), the PST perceived her
enacted meanings to be incompatible with each other, which generated a perturba-
tion and led to the PST rejecting the displayed graph she had constructed via reason-
ing emergently due to her having drawn the graph from left-to-right; she claimed
that drawing a graph that way was “backwards.” Similarly, in his work with middle-
school students, Tasova (2021) noted that students enact multiple graphical mean-
ings. In some cases, the enacted meanings were incompatible with each other,
leading to a student experiencing a perturbation and ultimately having to make an
accommodation to their meanings or abandon one of their meanings. In other cases,
a student’s enacted meanings were not necessarily incompatible with each other, but
instead, each provided different avenues to solving the perceived problem.
Regardless, Tasova’s (2021) notion of competing meanings generates a few ques-
tions for research that could benefit from incorporating the figurative and operative
distinction: What are the different meanings students construct for a particular topic,
and to what extent are those meanings compatible or incompatible? To what extent
are particular competing meanings epistemological obstacles (Harel & Sowder,
2005) or natural to cognitive development, as opposed to artifacts of canonical edu-
cational approaches? How might competing meanings be productively addressed
and leveraged by educators and researchers? With respect to this last question, there
is the potential to generate productive learning experiences via students reflecting
on competing meanings, particularly when such reflection is engendered by a
moment of perturbation.
Our second suggestion for a research line moving forward is also an extension,
and one in the direction of upper-level mathematics (e.g., undergraduate mathemat-
ics and above). This suggestion is motivated by personal conversations with
Anderson Norton, Shiv Smith Karunakaran, David Plaxco, and Paul Dawkins.
These conversations revolved around the potential role of the figurative and opera-
tive distinction in the areas of proof and proving, logic, and the study of advanced
mathematical objects. In short, those conversations have revolved around the ten-
dencies of mathematicians and mathematics educators to reduce those areas to for-
malisms, including syntactical rules, proof structures or templates, logic models,
algebraic systems, and other objects that they take to encapsulate operations. As
Norton explained, such a framing emphasizes “the conceptualization of formal
mathematics, as opposed to formalism deriving from the conceptualization of oper-
ations” (Personal communication). A productive line of research, which is sug-
gested by Norton’s (2022) recently published book, may instead be to consider how
the coordination of mental operations and meanings rooted in such provide a spring-
board for the formalizations required in upper-level mathematics.
Acknowledgments This material is based upon work supported by the National Science
Foundation under Grant Nos. DRL-1350342, DRL-1419973, and DUE-1920538. Any opinions,
findings, and conclusions or recommendations expressed in this material are those of the authors
and do not necessarily reflect the views of the National Science Foundation or our respective uni-
versities. Thank you to PME-NA and SIGMAA on RUME for the opportunity to present a previous
version of this manuscript. Thank you to Anderson Norton for several conversations that shaped
and contributed to this chapter.
References
Byerley, C. (2019). Calculus students’ fraction and measure schemes and implications for teach-
ing rate of change functions conceptually. The Journal of Mathematical Behavior, 55, 100694.
https://doi.org/10.1016/j.jmathb.2019.03.001
Carlson, M. P., Jacobs, S., Coe, E., Larsen, S., & Hsu, E. (2002). Applying covariational rea-
soning while modeling dynamic events: A framework and a study. Journal for Research in
Mathematics Education, 33(5), 352–378. https://doi.org/10.2307/4149958
Clement, J. (1989). The concept of variation and misconceptions in Cartesian graphing. Focus on
Learning Problems in Mathematics, 1(1–2), 77–87.
Dewey, J. (1929). The sources of a science of education. Liveright Publishing.
Ellis, A. B. (2007a). The influence of reasoning with emergent quantities on students’ generaliza-
tions. Cognition and Instruction, 25(4), 439–478.
Ellis, A. B. (2007b). A taxonomy for categorizing generalizations: Generalizing actions and
reflection generalizations. Journal of the Learning Sciences, 16(2), 221–262. https://doi.
org/10.1080/10508400701193705
Ellis, A. B., Özgür, Z., Kulow, T., Williams, C. C., & Amidon, J. (2015). Quantifying exponen-
tial growth: Three conceptual shifts in coordinating multiplicative and additive growth. The
Journal of Mathematical Behavior, 39, 135–155. https://doi.org/10.1016/j.jmathb.2015.06.004
Ellis, A. B., Lockwood, E., Tillema, E., & Moore, K. C. (2022). Generalization across multiple
mathematical domains: Relating, forming, and extending. Cognition and Instruction, 40(3),
351–384. https://doi.org/10.1080/07370008.2021.2000989
Harel, G. (2008a). DNR perspective on mathematics curriculum and instruction, part I: Focus on
proving. ZDM: The International Journal on Mathematics Education, 40, 487–500.
Harel, G. (2008b). DNR perspective on mathematics curriculum and instruction, part II: With
reference to teacher’s knowledge base. ZDM: The International Journal on Mathematics
Education, 40, 893–907.
Harel, G., & Sowder, L. (2005). Advanced mathematical-thinking at any age: Its nature and its
development. Mathematical Thinking and Learning, 7(1), 27–50.
Johnson, H. L. (2012). Reasoning about variation in the intensity of change in covarying quantities
involved in rate of change. The Journal of Mathematical Behavior, 31(3), 313–330. https://doi.
org/10.1016/j.jmathb.2012.01.001
Johnson, H. L. (2015). Together yet separate: Students’ associating amounts of change in quantities
involved in rate of change. Educational Studies in Mathematics, 1-22. https://doi.org/10.1007/
s10649-014-9590-y
Liang, B., & Moore, K. C. (2021). Figurative and operative partitioning activity: A student’s mean-
ings for amounts of change in covarying quantities. Mathematical Thinking & Learning, 23(4),
291–317.
Monk, S. (1992). Students’ understanding of a function given by a physical model. In G. Harel
& E. Dubinsky (Eds.), The concept of function: Aspects of epistemology and pedagogy
(pp. 175–193). Mathematical Association of America.
Montangero, J., & Maurice-Naville, D. (1997). Piaget, or, the advance of knowledge. In In.
L. Erlbaum Associates.
Moore, K. C. (2013). Making sense by measuring arcs: A teaching experiment in angle mea-
sure. Educational Studies in Mathematics, 83(2), 225–245. https://doi.org/10.1007/
s10649-012-9450-6
Moore, K. C. (2014a). Quantitative reasoning and the sine function: The case of Zac. Journal for
Moore, K. C. (2014b). Signals, symbols, and representational activity. In L. P. Steffe, K. C. Moore,
L. L. Hatfield, & S. Belbase (Eds.), Epistemic algebraic students: Emerging models of students’
algebraic knowing (pp. 211–235). University of Wyoming.
Moore, K. C. (2021). Graphical shape thinking and transfer. In C. Hohensee & J. Lobato (Eds.),
Transfer of learning: Progressive perspectives for mathematics education and related fields
(pp. 145–171). Springer.
Moore, K. C., & Thompson, P. W. (2015). In T. Fukawa-Connelly, N. Infante, K. Keene, &
M. Zandieh (Eds.), Shape thinking and students’ graphing activity (pp. 782–789). Proceedings
of the Eighteenth Annual Conference on Research in Undergraduate Mathematics Education.
Moore, K. C., Paoletti, T., & Musgrave, S. (2013). Covariational reasoning and invariance among
coordinate systems. The Journal of Mathematical Behavior, 32(3), 461–473. https://doi.
org/10.1016/j.jmathb.2013.05.002
Moore, K. C., Silverman, J., Paoletti, T., Liss, D., & Musgrave, S. (2019a). Conventions, habits,
and U.S. teachers’ meanings for graphs. The Journal of Mathematical Behavior, 53, 179–195.
Moore, K. C., Stevens, I. E., Paoletti, T., Hobson, N. L. F., & Liang, B. (2019b). Pre-service
teachers’ figurative and operative graphing actions. The Journal of Mathematical Behavior, 56.
Moore, K. C., Liang, B., Stevens, I. E., Tasova, H. I., & Paoletti, T. (2022). Abstracted quanti-
tative structures: Using quantitative reasoning to define concept construction. In G. Karagöz
Akar, İ. Ö. Zembat, S. Arslan, & P. W. Thompson (Eds.), Quantitative Reasoning in
Mathematics and Science Education (pp. 35–69). Springer International Publishing. https://
doi.org/10.1007/978-3-031-14553-7_3
Norton, A. (2014). In T. Fukawa-Connelly, G. Karakok, K. Keene, & M. Zandieh (Eds.), The con-
struction of cohomology as objectified action (pp. 957–969). Proceedings of the Seventeenth
Annual Conference on Research in Undergraduate Mathematics Education.
Norton, A. (2022). The psychology of mathematics: A journey of personal mathematical empower-
ment for educators and curious minds. Routledge.
Paoletti, T., & Moore, K. C. (2018). A covariational understanding of function: Putting a horse
before the cart. For the Learning of Mathematics, 38(3), 37–43.
Paoletti, T., Stevens, I. E., Hobson, N. L. F., Moore, K. C., & LaForest, K. R. (2018). Inverse
function: Pre-service teachers’ techniques and meanings. Educational Studies in Mathematics,
97(1), 93–109. https://doi.org/10.1007/s10649-017-9787-y
Piaget, J. (1969). The mechanisms of perception. Routledge & Kegan Paul.
Piaget, J. (1970). Genetic epistemology. W. W. Norton & Company, Inc.
Piaget, J. (2001). Studies in reflecting abstraction. Psychology Press Ltd..
Piaget, J., & Inhelder, B. (1971). Mental imagery in the child: A study of the development of ima-
ginal representation. Routledge & Kegan Paul.
Piaget, J., & Inhelder, B. (1973). Memory and intelligence. Basic Books.
Piaget, J., Inhelder, B., & Szeminska, A. (1981). The child’s conception of geometry (E. A. Lunzer,
Trans.). W. W. Norton & Company. (1960).
Saldanha, L. A., & Thompson, P. W. (1998). Re-thinking co-variation from a quantitative per-
spective: Simultaneous continuous variation. In S. B. Berensen, K. R. Dawkings, M. Blanton,
W. N. Coulombe, J. Kolb, K. Norwood, & L. Stiff (Eds.), Proceedings of the 20th Annual
Meeting of the North American Chapter of the International Group for the Psychology of
Mathematics Education (Vol. 1, pp. 298–303). ERIC Clearinghouse for Science, Mathematics,
and Environmental Education.
Smith, J. P., III, & Thompson, P. W. (2007). Quantitative reasoning and the development of alge-
braic reasoning. In J. J. Kaput, D. W. Carraher, & M. L. Blanton (Eds.), Algebra in the early
grades (pp. 95–132). Lawrence Erlbaum Associates.
Steffe, L. P. (1991a). The learning paradox: A plausible counterexample. In L. P. Steffe (Ed.),
Epistemological foundations of mathematical experience (pp. 26–44). Springer-Verlag. https://
doi.org/10.1007/978-1-4612-3178-3_3
Steffe, L. P. (1991b). Operations that generate quantity. Journal of Learning and Individual
Differences, 3(1), 61–82.
Steffe, L. P., & Norton, A. (2014). Perspectives on epistemic algebraic students. In L. P. Steffe,
K. C. Moore, L. L. Hatfield, & S. Belbase (Eds.), Epistemic algebraic students: Emerging
models of students’ algebraic knowing (pp. 317–323). University of Wyoming.
Stevens, I. E. (2019). Pre-service teachers’ constructions of formulas through covariational rea-
soning with dynamic objects [Ph.D. Dissertation]. University of Georgia.
Stevens, I. E. (in press). “A=2πrh is the surface area for a cylinder”: Figurative and operative
thought with formulas. Proceedings of the Twenty-Fourth Annual Conference on Research in
Undergraduate Mathematics Education.
Tasova, H. I. (2021). Developing middle school students’ meanings for constructing graphs
through reasoning quantitatively [Ph.D. Dissertation]. University of Georgia.
Tasova, H. I., & Moore, K. C. (2020). Framework for representing a multiplicative object in
the context of graphing. In A. I. Sacristán, J. C. Cortés-Zavala, & P. M. Ruiz-Arias (Eds.),
Mathematics education across cultures: Proceedings of the 42nd Meeting of the North
American Chapter of the International Group for the Psychology of Mathematics Education,
Mexico (pp. 210–219). Cinvestav/PME-NA.
in developing mathematics curricula. In E. A. Silver (Ed.), Teaching and learning mathemati-
cal problem solving: Multiple research perspectives (pp. 189–243). Erlbaum.
Thompson, P. W. (1990). A cognitive model of quantity-based algebraic reasoning. Annual
Meeting of the American Educational Research Association. 1990, March 27–31.
Thompson, P. W. (1993). Quantitative reasoning, complexity, and additive structures. Educational
Studies in Mathematics, 25(3), 165–208.
Thompson, P. W. (1994a). The development of the concept of speed and its relationship to concepts
of rate. In G. Harel & J. Confrey (Eds.), The development of multiplicative reasoning in the
learning of mathematics. SUNY Press.
Thompson, P. W. (1994b). Images of rate and operational understanding of the fundamental
theorem of calculus. Educational Studies in Mathematics, 26(2–3), 229–274. https://doi.
org/10.1007/BF01273664
Thompson, P. W. (2011). In S. Chamberlin, L. L. Hatfield, & S. Belbase (Eds.), Quantitative rea-
soning and mathematical modeling (pp. 33–57). New perspectives and directions for collabora-
tive research in mathematics education: Papers from a Planning Conference for WISDOMê.
Thompson, P. W. (2013a). Constructivism in mathematics education. In S. Lerman
(Ed.), Encyclopedia of mathematics education (pp. 96–102). Springer. https://doi.
org/10.1007/978-94-007-4978-8_31
Thompson, P. W. (2013b). In the absence of meaning. In K. Leatham (Ed.), Vital direc-
tions for research in mathematics education (pp. 57–93). Springer. https://doi.
org/10.1007/978-1-4614-6977-3_4
Thompson, P. W. (2016). Researching mathematical meanings for teaching. In L. English
& D. Kirshner (Eds.), Third handbook of international research in mathematics education
(pp. 435–461). Taylor and Francis. https://doi.org/10.4324/9780203448946-28
Thompson, P. W., & Carlson, M. P. (2017). Variation, covariation, and functions: Foundational
ways of thinking mathematically. In J. Cai (Ed.), Compendium for research in mathematics
education (pp. 421–456). National Council of Teachers of Mathematics.
Thompson, A. G., & Thompson, P. W. (1996). Talking about rates conceptually, part II:
Mathematical knowledge for teaching. Journal for Research in Mathematics Education,
27(1), 2–24.
Thompson, A. G., Philipp, R. A., Thompson, P. W., & Boyd, B. A. (1994). Calculational and
conceptual orientations in teaching mathematics. In A. Coxford (Ed.), 1994 yearbook of the
NCTM (pp. 79–92). NCTM.
magnitudes: A hypothesis about foundational reasoning abilities in algebra. In L. P. Steffe,
K. C. Moore, L. L. Hatfield, & S. Belbase (Eds.), Epistemic algebraic students: Emerging
models of students’ algebraic knowing (Vol. 4, pp. 1–24). University of Wyoming.
von Glasersfeld, E. (1982). Subitizing: The role of figural patterns in the development of numerical
concepts. Archives de Psychologie, 50, 191–218.
von Glasersfeld, E. (1987). Preliminaries to any theory of representation. In C. Janvier (Ed.),
Problems of representation in the teaching and learning of mathematics (pp. 215–225).
Lawrence Erlbaum.
https://doi.org/10.4324/9780203454220
Chapter 5
Figurative and Operative Imagery:
Essential Aspects of Reflection
in the Development of Schemes
and Meanings
Patrick W. Thompson, Cameron Byerley, and Alan O’Bryan
Our purpose in this chapter is to clarify the role of imagery in students’ mathemati-
cal learning and reasoning. In pursuit of this goal, we address interdependencies
among imagery, reflection, scheme, and meaning as theoretical constructs and illus-
trate their interdependence by examples. Our motive for this expanded charge is that
while the notion of schemes and scheme development is sometimes discussed in
studies of students’ mathematical learning, the role of imagery in that process is
often neglected, yet it is central to the development of productive mathematical
meanings. Often people doing mathematics education research pay insufficient
attention to the very nuanced ways in which people understand a situation—what
their image of the situation is. As a counterpoint, Thompson (1996) spoke of stu-
dents’ images of a solution to an algebraic equation. “Their image of solving equa-
tions often is of activity that ends with something like ‘x = 2.’ So, when they end
with something like ‘2 = 2’ or ‘x = x,’ they conclude without hesitation that they
must have done something wrong” (Thompson, 1996, p. 274).
P. W. Thompson (*)
Arizona State University, Tempe, AZ, USA
e-mail: pat@pat-thompson.net
C. Byerley
University of Georgia, Athens, GA, USA
A. O’Bryan
Rational Reasoning LLC, Gilbert, AZ, USA
https://doi.org/10.1007/978-3-031-47386-9_5
130 P. W. Thompson et al.
Imagery
We state at the outset that by “imagery” we mean far more than “visualization.”
Imagining something one has seen, or producing something as if seen, indeed, falls
within the category of imagery. But recalling something one has said, thought, done,
or felt also falls within the category of imagery. In an act of re-presentation,1 a per-
son recalls having an experience. The experience might have been having a thought
or feeling, having done something in some context, having interpreted something in
a particular way, recalling having recalled something, or other acts of recollection.
In general, by “image,” we mean the re-presentation of experience. We use “imag-
ery” to refer to images collectively.
While we say imagery exceeds visualization, we acknowledge that “visualiza-
tion” has been used sometimes with broad meaning, including what we call “cre-
ative imagery.” Creative imagery is the construction of an image one has not
experienced but is grounded in experience. One example is Galileo’s thought exper-
iment of repeatedly dropping two iron balls of different sizes joined by ever-thinner
filaments to infer all masses in a vacuum fall identically (Clement, 2018; Miller,
1996). Another is Einstein’s thought experiment of a person falling in an elevator to
infer gravity is akin to acceleration within an inertial frame of reference (Miller,
1996). Neither Galileo nor Einstein had experiences that were recalled as such.
Rather, they assembled images from experience in novel ways.
Experiences are never recalled veridically. Our distinction between creative
imagery and recalled experience is therefore muddied by the fact that initial experi-
ences and recalled experiences are constructed using schemes one currently has,
which can change in the interim between initial experience and recollection. For
example, Piaget (1968a) presented experimental evidence of children’s memories of
a visual image improving over time. The improvement, Piaget claimed, was due to
students having developed more coherent schemes by which they remembered (re-
constructed) their original perception. The experiment had first-grade children look
for a short period of time at a display of 10 bars aligned vertically in ascending
length (Fig. 5.1a). Piaget and colleagues asked children to draw what they remem-
bered seeing 1 week later and 6 months later. Children’s drawings 1 week later
tended to show local groups of ascension, but not uniform increase in length
(Fig. 5.1b). Their drawings 6 months later, with no mention of the initial episode
having been made, were considerably improved—many more children’s drawings
resembled the figure presented 6 months earlier. Piaget explained that, in the inter-
vening 6 months, children had developed schemes for order that included
transitivity.2 In other words, their recollections improved because the schemes by
1
Glasersfeld (1991) distinguished between “to represent” and “to re-present” this way. To repre-
sent an experience is to take one thing as standing for another. To re-present an experience is to
bring to mind a record of the experience.
2
One way Piaget inferred that children’s ordering schemes included transitivity was to have them
insert a stick into a series they had constructed, ordered left-to-right ascending by height. If the
5 Figurative and Operative Imagery: Essential Aspects of Reflection… 131
Fig. 5.1 (a) The figure as presented to children. (b) Type of drawing common among children
1 week later
which they re-presented their initial experience of the bars had become more
advanced.
Another source for our position regarding imagery is our appreciation of action
as the foundation of Piaget’s genetic epistemology. To Piaget, actions were cogni-
tive activity that might (or not) be expressed in behavior. “One can say that all
action—that is to say all movement, all thought, or all emotion—responds to a
need” (Piaget, 1968b, p. 6). Actions are the foundation of experience, so the phrase
“recall an experience,” to be in line with Piaget’s genetic epistemology, forces us to
include recalled movement, thought, or emotion as imagery.
We also avoid tacitly equating “action” with observable behavior, which happens
often when people use the phrase “reflection on activity.” Powers (1973a) addressed
this when he explained that living organisms cannot organize themselves around
how their behavior is perceived by others, but instead according to the effects of the
organism’s actions as discerned by the organism (Powers, 1973b, p. 418).
Powers’ message for us is that we cannot take students’ behavior at face value—
it is but an expression of their actions—where we use “action” as Piaget intended.
This is at the root of Steffe and Thompson’s (2000) distinction between students’
mathematics (their mathematical realities) and mathematics of students—an observ-
er’s understanding of how students might be thinking to behave as they did or might
do. Students’ mathematics is the dark matter and dark energy of mathematics edu-
cation. Devising a viable mathematics of students is therefore a core mission of
mathematics education research. Understanding students’ imagery while engaged
in instruction and while recalling their experiences outside of instruction is central
to that mission.
On a related note, it is common for instructors and instructional designers to
include visual presentations to supplement prose or have students engage in some
form of activity. The thought is that activities or visual presentations help students
child found the first stick taller than the one they were to insert and inserted the new stick to the left
of the first-taller stick, the child exhibited transitivity. They knew all sticks to the left of that posi-
tion were necessarily shorter than the one they inserted and all sticks to the right were necessarily
taller than the one they inserted (Piaget, 1965).
understand the goal of instruction. Our emphasis on imagery as recalled experience

is important for putting these efforts in proper light. By taking seriously the stance
that imagery is rooted in recalled experience, we are forced to be mindful that stu-
dents recall their experience of activities or visual presentations; they cannot recall
what we understand as having been presented to them. Students’ experience of an
activity or presentation is conditioned by the schemes through which they under-
stood it and recall it. Since learners are, by definition, new to the ideas being taught,
their experience of activities or presentations will be substantially different from the
originator’s intentions.
Imagery, Schemes, and Meanings
Having spoken of schemes in relation to images repeatedly, we feel obliged to say

what we mean by a scheme and speak to the role of imagery in scheme development.
Piaget’s use of “scheme” was quite utilitarian. It allowed him to speak of mental
organizations that supported flexible reasoning across seemingly disparate situa-
tions. Montangero and Maurice-Naville (1997, p. 155) presented a compendium of
six ways Piaget used “scheme.” They ranged from “[Schemes are] organized totali-
ties whose internal elements are mutually implied” (Piaget, 1936, p. 445) to “A
scheme is the structure or the organization of actions which is transferred or gener-
alized when this action is repeated in similar or analogous circumstances” (Piaget &
Inhelder, 1966, p. 11, footnote not translated in the English version).
Piaget’s statement, “organized totalities whose internal elements are mutually
implied,” derives from his stance that actions are implicative. When someone
engages in an action, it creates a new experiential context that could be the trigger
for other actions. An action in a context implies other actions. Thus, “… elements
are mutually implied” means that a scheme constitutes a locally closed system in
which any of its assumed conditions could activate the scheme in its totality. An
example is when someone has a mature constant speed scheme. They are aware that
a time and a speed are involved when they know an object traveled some distance,
that a distance and a speed are involved when they know it traveled some time, and
that a time and a distance are involved when they know it traveled at some speed.
“Mutual implications” of time, distance, and speed in a person’s constant speed
scheme is evidenced when they understand that there are implied distances in “A car
drove 60 mi/hr for 3 hours and then 40 mi/hr for 5 hours. What was the car’s aver-
age speed?”
Likewise, the statement, “… organization of actions which is transferred when
this action is repeated …” was Piaget’s way to account for how a scheme (as an
organization of actions) can be activated in seemingly different circumstances. An
example of this is when someone uses their constant speed scheme (constant rate of
change of distance with respect to time) to understand constant rate of change of
volume of a fluid in a container with respect to its height in the container.
Fig. 5.2 Steffe’s (2010, p. 23) characterization of schemes
Cobb and Glasersfeld (1984) and Glasersfeld (1995, 2001) proposed that, to
Piaget, a scheme was a three-part mental structure: an internal condition that would
trigger a scheme, an action or system of actions, and an anticipation of what the
action would produce. Steffe (2010) expanded Cobb and Glasersfeld’s definition to
include a scheme’s goal (Fig. 5.2). Steffe’s motivation for including a generated
goal in his definition of scheme was
The Generated Goal can be regarded as the apex of a tetrahedron. The vertices of the base
of the tetrahedron constitute the three components of a scheme. The double arrows linking
the three components are to be interpreted as meaning that it is possible for any one of them
to be in some way compared or related to either of the two others. The dashed arrow is to
be interpreted as an expectation of the scheme’s result. (Steffe, 2010, p. 23)
A main feature Piaget communicated in his characterizations of scheme is they are

cognitive structures that can express themselves in action or behavior. Cobb’s,
Glasersfeld’s, and Steffe’s definitions address Piaget’s intention well. However,
another important aspect of schemes’ functioning to Piaget, not captured by those
definitions, is that people employ schemes to comprehend situations, and to give
meaning to situations (Piaget & Garcia, 1991). To us, a definition of scheme must
support interpretations of a person’s attempt to understand situations in terms of
meanings they have for constituent elements and relationships among them.
Thompson and Saldanha (2003), for example, spoke of a mature fraction scheme as
a network of relationships among schemes for measure, multiplication, division,
and relative size, each of which entails aspects of proportionality, as a means to
understand the broad array of situations one can understand as involving fractions.
To this end, Thompson et al. (2014) expanded Steffe’s, Cobb’s, and Glasersfeld’s
definitions of scheme.
A scheme as an organization of actions, operations, images, or schemes—which can have
many entry points that trigger action—and anticipations of outcomes of the organization’s
activity.3 (Thompson et al., 2014, p. 11)
We point out that, in this definition, “entry points” often are images of contexts, and
anticipated outcomes are most definitely images. Unlike Steffe’s definition, and like
Glasersfeld’s and Cobb’s, Thompson et al.’s (2014) definition of scheme does not
3
This definition is recursive, not circular. A scheme might recruit other schemes when activated.
include goals. In a larger context, though, it is like Steffe’s definition in that a person
can generate a goal in the activity of implementing a scheme or might activate a
scheme because a generated goal matches a scheme’s anticipated outcome.
Thompson et al.’s definition of scheme is at the root of Thompson and Harel’s
(Harel, 2021; Thompson et al., 2014) attempt to give coherence among meanings
for understandings, meanings, and ways of thinking (Fig. 5.3).
This system for the use of “understanding,” “meaning,” and “way of thinking”
aligns with Harel’s and Thompson’s quest to decouple “understand” and “under-
stand correctly.” They do this by resting their system on the idea of assimilation.
They rely on Piaget’s characterization of assimilation as, in effect, giving meaning.
Assimilating an object to a scheme involves giving one or several meanings to this object,
and it is this conferring of meanings that implies a more or less complete system of infer-
ences, even when it is simply a question of verifying a fact. In short, we could say that an
assimilation is an association accompanied by inference. (Johnckheere et al., 1958, p. 59,
as translated by Montangero & Maurice-Naville, 1997, p. 72)
The first entry (Understanding in the moment) in Fig. 5.3 describes a person who
has an understanding of something said, written, or done in the moment of under-
standing it. Technically, all understandings are understandings-in-the-moment.
Some understandings might be a state that the person has struggled to attain at that
moment through functional accommodations to existing schemes (Steffe, 1991) and
is easily lost once the person’s attention moves on. This type of understanding is
typical when a person is making sense of an idea for the first time.
The meaning of an understanding is the space of implications that the current
understanding mobilizes—actions or schemes that the current understanding
implies, that the current understanding brings to mind with little effort. An under-
standing is stable if it is the result of an assimilation to a scheme. A scheme, being
stable, then constitutes the space of implications resulting from the person’s assimi-
lation of anything to it. The scheme is the meaning of the understanding that the
person constructs in the moment. As an aside, schemes provide the “way” in Harel’s
“way of understanding.” Finally, Harel and Thompson characterize “way of
Fig. 5.3 Meanings of “understanding,” “meaning,” and “way of thinking” (Thompson et al.,
2014, p. 13)
thinking” as when a person has developed a pattern for utilizing specific meanings
or ways of thinking in reasoning about particular ideas or situations.
Images and Schemes
Images enter our definition of scheme, and therefore meaning, in three ways: Images
can be contexts that activate a scheme, they can be waypoints in a scheme’s activity,
and they can be anticipations of a scheme’s result within the context of the scheme’s
activation.
Thompson (1994a, 1996) explained the ways in which the notion of image is
intertwined with Piaget’s concept of scheme. He pointed out three levels of imagery
in Piaget’s work. The first level of imagery is when a child engages in deferred imi-
tation. Deferred imitation is when a child enacts the imitated behavior to assimilate
(understand) it. The second level of imagery (figurative) is an image of an initial
state and actions that are associated with it, but the actions and image are tied
tightly—such as a student accustomed to drawing altitudes in a triangle with all
angles less than 90 degrees being confused when asked to draw all altitudes in a
triangle with one angle greater than 90 degrees.
The third level of imagery is what Piaget called “operative.”
[This is an image] that is dynamic and mobile in character … entirely concerned with the
transformations of the object. … [The image] is no longer a necessary aid to thought, for
the actions which it represents are henceforth independent of their physical realization and
consist only of transformations grouped in free, transitive and reversible combination … In
short, the image is now no more than a symbol of an operation, an imitative symbol like its
precursors, but one which is constantly outpaced by the dynamics of the transformations.
Its sole function is now to express certain momentary states occurring in the course of such
transformations by way of references or symbolic allusions. (Piaget, 1967, p. 296)
The three levels of imagery do not differ in type. They are all re-presented experi-
ences. Instead, the levels are differentiated by the ways images are integrated into
individuals’ reasoning and the types of reasoning into which they are integrated. We
unpack the three levels in the following paragraphs.
First-Level Imagery (Deferred Imitation)
Piaget’s examples of deferred imitation are often about infants or toddlers mimick-
ing their experience of others (e.g., opening their mouth to mimic their mother) or
engaging in play to mimic social interactions. But deferred imitation is a broader
phenomenon. In psychotherapy, it is called re-experiencing (Joseph & Williams,
2005)—the replaying of a traumatic event to assimilate (understand) it by either
adjusting one’s understanding of a world in which such a thing could occur or
adjusting one’s understanding of one’s place in the world that makes the event
sensible. The role of deferred imitation in mathematics learning is unclear to us.

This is not to claim it is unimportant. We just say we are unclear as to its role.
Second-Level (Figurative) Imagery
Second-level (figurative) imagery aligns with what Davis et al. (1978) called
“visually-moderated sequences”—activity sequences triggered by a current visual
or cognitive state (e.g., seeing an equation and thinking to add something to both
sides) that end in a new visual or cognitive state (e.g., an expression with no con-
stant terms) that triggers another activity sequence (e.g., dividing both sides by the
same number). Each activity sequence results in a new state, but upon arriving at a
new state it is simply the end of the activity sequence. It is not a goal toward which
the student strove, and thus the end state is not an anticipated result of the activity,
and the actual result does not, to the student, entail an image of the activity lead-
ing to it.
Frank (2017) provided an excellent example of a student whose activity ends
with a result that, for her, did not entail an image of the activity that led to it. The
student (Ali) constructed a graph to represent two quantities’ values as they varied
simultaneously in an animation of the quantities and their magnitudes. Ali ended
with what Frank considered an appropriate graph. But the graph, to Ali, did not
entail the covariational reasoning in which she engaged while making it. Ali spoke
of the graph which she had just made as if it was a static shape as if a piece of wire.
At the outset of this study, I thought that if Ali made a graph by simultaneously tracking two
magnitudes, then she engaged in emergent shape thinking. I had not considered that Ali’s
meaning for her sketched graph might not reflect the thinking she engaged in to make the
graph. (Frank, 2017, p. 193)
Research by Lobato, Stump, and Moore provide additional examples of students

operating with figurative imagery. Lobato and Thanheiser (2002) reported children
thinking the slope of a ramp leading to a platform is affected by the width of the
platform. They included the platform as part of the “over” image in their “up and
over” slope scheme. Stump (2001) reported a student who thought a slope of −5/6
is different from a slope of 5/ − 6 because they entail different images of “up and
over.” Moore and colleagues (2014, 2019) reported students became confused about
the slope of a line when x- and y-axes were switched. They wanted the line to have
the same slope because, to them, the line still pointed in the same direction. Their
slope scheme depended on an image of a line’s direction, and a slope value, to them,
was an index of directionality.
Anyone operating with figurative imagery can lose track of their reasoning eas-
ily. Byerley and Thompson (2017) report several instances of teachers moving from
one meaning of a situation to a contrary meaning within seconds as they employed
schemes that used images figuratively. Figure 5.4 presents an item from the
Mathematical Meanings for Teaching secondary mathematics (MMTsm) inventory
(Thompson, 2016). Its design was motivated by the ambiguity with which teachers
Fig. 5.4 Meaning of “over” (Byerley & Thompson, 2017). (© 2014 Arizona Board of Regents.
Used with permission)
in their samples used the word “over” when speaking of rates of change. Thompson
and his team could not tell whether teachers used “over” to convey an interval dur-
ing which an event unfolded or to convey a spatial arrangement of numerator and
denominator with respect to a vinculum.
For this item, the team took “during” to be a high-level answer to Part A and
f(x0 + Δx) − f(x0) = 4 grams, f(x + Δx) − f(x) = 4 grams, or even Δm = 4 to be a high-
level answer to Part B. Byerley and Thompson reported 113 (45%) of 251 US teach-
ers said “over” meant during or something equivalent in response to Part A, and 71
(28%) of 251 teachers said “over” meant the same as divide in response to Part A. In
response to Part B, only 18 (16%) of the 113 teachers who said “over” meant “dur-
ing” for Part A represented the statement as a difference or change in mass. Forty-
one percent (41%) of the 113 teachers who said “over” meant during for Part A
responded to Part B by representing the statement as a quotient involving mass and
time, writing something like (change in mass)/(change in time) = 4 grams. In other
words, when reading the statement as a plain-language description of a situation, the
word “over” for these 113 teachers suggested an image of something happening in
time. But upon reading the same statement as something to be represented symboli-
cally, the word “over” suggested an image of numerator and denominator separated
by a vinculum.
An interview with James, who had a B.Sc. in mathematics education and was an
experienced teacher of algebra, geometry, and precalculus, illustrates how figurative
imagery leads to conflicts between schemes.
James: [Over means] during or duration. You could also think of it as a ratio, so
change in mass over, yeah so during or duration, so in your math class
when they say, “something over something”, they always mean a divide
sign so a ratio.
Int: Do you think they are both saying the same thing?
James: Well, yeah, I think that. Well yeah, they are saying. I think the during or
duration is more saying conceptually what is going on, and the divided by
or over I see the reason behind that, I think I’m more pointing out mathe-
matically what we mean when we say over with no explanations as to why,
it is just the way it is.
Int: So is the mass, the change in mass divided by the change in time, is that
how you write the idea of duration?
James: Can you repeat the question?

Int: Is the “delta mass divided by delta x” a mathematical way of saying
duration?
James: I want to say the change in x is the way of saying duration. I want to say
the change in x is representing duration. But maybe we could include the
division sign. So no, I would not say that “delta mass over delta x” is a way
of saying duration. So this is funny. (Byerley & Thompson, 2017,
pp. 188–189)
James never reconciled his conflict between “over” suggesting Δx as a represen-
tation of elapsed time and “over” suggesting the quotient Δm/Δx. We explain his
conflict by appealing to the imagery he apparently evoked in relation to his different
purposes for reading the statement. In Part A, his purpose was to read the statement
as a plain-language description of a phenomenon, for which “over” suggested an
image of something happening as time elapsed. In Part B, James’ purpose was to
represent mathematically a situation described in plain language, which suggested
an image of two numbers or expressions separated by a vinculum. What was new
for James is that the interviewer asked him to compare competing implications of
his two assimilations of the same word. The images James evoked were figurative—
they were tied tightly to the schemes he evoked in the contexts of his different
purposes.
We present a second example from Byerley and Thompson (2017) of figurative
imagery leading to competing assimilations of “the same” context. Figure 5.5 shows
another item from the MMTsm. Its purpose was to tease out whether teachers inter-
preted graphs as showing amounts of a quantity despite it being stated explicitly in
two ways that the graph showed rates of change of one quantity with respect to
another. Parts 2 and 3 of this item (not shown here) allowed teachers to reveal their
level of commitment to their initial interpretations of the graph. Thirty-six percent
(36%) of 239 high school mathematics teachers appropriately chose (c) for Part 1 of
this item; 49% chose (a).
The following excerpt shows a teacher who slipped from one interpretation of
the item to a contrary interpretation while explaining her answer to the question.
Fig. 5.5 Part 1 of the three-part item “Increasing or decreasing from rate” (Byerley & Thompson,
2017). (© 2014 Arizona Board of Regents. Used with permission)
Annie: [Reads problem aloud, emphasizes “grams per hour”.] We interpret

increasing. umm...let’s see the function gives the rate of change in grams per hour...
and so umm...what we are going to look at I would look at the rate of change being
positive or negative, if we have a positive rate of change the grams per hour the mass
is increasing per hour, is getting larger
Annie: So I look at where I have a positive rate of change, and I try to identify
where I have no rate of change [highlights maximum at (1.25, 5) where the rate of
change is approximately positive 5, but the acceleration is zero], this is telling me
where the mass is staying the same, and then I have a negative slope so mass is get-
ting small down to a zero rate of change so I’m not getting any smaller or larger...
[Annie chooses (a)]
We do not have direct evidence of where Annie looked as she spoke, but it seems
plausible she gazed at the text during her first utterances and gazed at the graph dur-
ing her second utterances. It appears that in the first utterances, Annie had in mind
values of the function f as values of the rate of change of the culture’s mass with
respect to elapsed time, whereas in the second utterances, Annie had in mind slopes
of tangents to the graph as values of the rate of change of the culture’s mass with
respect to elapsed time. In other words, Annie slipped from one scheme (values of
the function give the rate of change) to a different scheme (slope of tangents to the
graph gave the rate of change), and the “slip” was prompted by her different imag-
istic contexts (text vs. graph). For Annie, with rate of change functions, values of the
function give rates of change; with graphs, slopes of tangents give rates of change.
We note in passing that Annie felt no conflict between her two schemes because she
did not think of the function having positive values in places where its graph had
negative slopes.
Third-Level (Operative) Imagery
At the third level of imagery, students’ schemes are not dependent upon specific
images. Instead, images serve as arbitrary “momentary states” in a scheme’s imple-
mentation. Thompson and Dreyfus (1988) reported two sixth graders’ (Kim’s and
Lucy’s) advancing from second-level imagery, thinking of an integer such as −5 as
a location on a number line and later thinking of −5 as a displacement from any start-
ing point. Their imagery moved from Piaget’s second level to his third level. The
children’s schemes no longer needed a definite starting point. They knew they could
start anywhere to enact −5. They could then think of +3 + −5 as a composition of two
displacements that produced a net displacement of −2, where the second displace-
ment started wherever the first ended. Though they enacted the displacement of +3
from a specific place on a number line, they did not feel required to use a specific
place from which they must enact it. Their image of an actual starting place was not
more general. Rather, it was their scheme that became more general. It did not
require a specific starting place, thus, specific locations on the number line served
as “momentary states” in the activity of their integer composition schemes as they
conceptualized the net displacement (sum) of several displacements.
While Kim and Lucy developed operative imagery regarding composition of

d isplacements, only Kim thought of number in −number as itself being a net dis-

placement. Understanding expressions like 90 30 was unproblematic for Kim

because the expression 90 30 fit within her image of things constituting num-
ber—it was a net displacement. For Lucy, only whole numbers fit her image of
things constituting number. Evaluating the negation of complex expressions was
often effortful for Kim, but she nevertheless knew what she was supposed to end
with—the negation of a net displacement. In sum, Kim and Lucy had operative
imagery with respect to composition of displacements, while only Kim had devel-
oped operative imagery for negation. Our main point here is that you categorize the
level of students’ images as second-level (figurative) or third-level (operative)
according to your judgment of how necessary those specific images are in students’
schemes as they employ them.
Summary
The examples above show our use of “image-level” relatively. Deferred imitation
can happen when anyone re-plays an event or collection of actions that they did not
fully assimilate—they did not fully understand. Young children go home from
school and play school as an attempt to assimilate their new experience of a teacher
who attempts to control their thinking. Graduate students in mathematics replay
specific aspects of a lecture in their attempt to assimilate them—to develop an
understanding. Movement of an hour hand on a circular clock is often offered as a
foundational image for understanding cyclical groups. For persons who must think
of a clock to do arithmetic in a cyclical group, their clock image is figurative because
their scheme for addition in a cyclical group requires it. A person who uses a clock
whose hour hand varies from 0 to 2π hours to think of equivalent angle measures on
a number line, but can also see the repetition as hops on the number line or as the
number line collapsing into equivalence classes by the mapping ℝ → ℝ/2π,4 is
employing images operatively. Any image they employ in their reasoning about
arguments to a trigonometric function is a matter of convenience because it fits their
purpose in the moment.
Piaget’s notion of image is useful because, in developing a scheme, a student
must reflect on her reasoning. To reflect on her reasoning, she must create, as best
she can, images of having reasoned in the way she did. This means she must develop
recollections of “momentary states” in having reasoned. To construct a scheme,
students must repetitively engage in variations of the reasoning that will become
solidified in that scheme and re-present it as best they can to reflect upon it.
Sometimes reflection occurs during moments of confusion, sometimes after having
4
The mapping maps every real number x to the non-negative remainder of x divided by 2π. This
has the same effect as all numbers on the number line falling simultaneously, like molecules of
water vapor, onto their equivalent positions in the interval [0,2π).
engaged in a chain of interpretations, inferences, and decisions. Nevertheless, in the

process of constructing a scheme, images of having reasoned become students’
objects of reflection (Cooper, 1991; Harel, 2008a, b, 2013).
It is worth noting that the number of schemes students develop is immense. Any
word that is meaningful to them is meaningful because hearing or seeing it activates
a scheme. Any symbol or symbolic expression that is meaningful to them is mean-
ingful because seeing it or thinking of it activates a scheme. Any diagram or picture
that is meaningful to them is meaningful because they assimilate it into one or more
schemes. Any time you assess students’ thinking or interview students, they are
interpreting both the setting and your actions through the activation of schemes.
Moreover, any combination of the above that proves meaningful to a student is
meaningful because of minor or major accommodations in their schemes in the
moments of activating them. Any time someone puzzles about the meaning of a
word or phrase and resolves their puzzlement has engaged in some form of reflec-
tion that engendered an accommodation to their schemes. When you create a scheme
as a model of student thinking and impute that scheme to students, you must be
cognizant that you have most certainly omitted a vast number of schemes that were
at play in students’ thinking that might turn out to be important for understanding
their thinking. The art of using scheme and image as explanatory constructs is to
find the appropriate level of analysis that produces tractable explanations of stu-
dents’ successes and difficulties.
The prior paragraph points to a methodological aspect of scheme as a theoretical
construct. On one hand, we say schemes are organizations of a person’s mental
activity that express themselves in what an observer sees as behavior. From this
perspective, schemes reside in individuals. On the other hand, we say scheme is a
theoretical construct that researchers impute to individuals to explain their behavior.
They are a researcher’s construct. This is much like stances taken by natural scien-
tists. They realize anything they say is based on models built from theory-laden
observations, but in doing their science they act as if their models describe reality—
until observations force them to step back and question their assumptions and their
models. Likewise, we infer schemes from students’ behavior in response to care-
fully defined probes. We impute schemes to students to form explanations of their
behavior and to design supports we think will advance their thinking. We step back
and question ourselves when our explanations become inconsistent or inadequate,
or our designed supports do not have their intended effects.
Imagery, Schemes, and Reflective Abstraction
In this section we expand our earlier discussion of imagery, schemes, and meanings
to address what we mean by reflection and the role it plays in a person’s construc-
tion of schemes.
John Dewey placed reflection at the center of his understanding of thinking and
placed thinking at the center of the development of a critically informed democracy.
Dewey defined reflective thought as, “Active, persistent, and careful consideration
of any belief or supposed form of knowledge in the light of grounds that support it,
and the further conclusions to which it tends” (Dewey, 1910, p. 6). Dewey also
anticipated coherence as a characteristic of reflective thought that Piaget came to
see as central to his genetic epistemology. Reflection on one’s thinking leads to
the organization of facts and conditions which, just as they stand, are isolated, fragmentary,
and discrepant, the organization being effected through the introduction of connecting
links, or middle terms. (Dewey, 1910, p. 79)
Dewey was also in line with Piaget as to one’s motive for reflection.
Demand for the solution of a perplexity is the steadying and guiding factor in the entire
process of reflection—i.e., reflection serves a regulatory function. (Dewey, 1910, p. 11)
The key aspect of Dewey’s account of reflection is that “to reflect” means to think
about thinking. This is in line with our prior discussions of imagery with respect to
schemes when one considers that people construct schemes by creating images of
having reasoned and taking those images as their objects of thought.
A vast difference between Dewey’s and Piaget’s accounts of reflection is that
Dewey thought of reflection as a conscious activity, whereas Piaget thought of con-
scious reflection as the tip of an iceberg. He posited processes of unconscious reflec-
tion that must precede anything resembling Dewey’s characterization. Piaget took
the stance that one can be aware only of images one operates upon. You cannot be
aware of the operations you use to operate on an image—to an extent. You can,
however, project your operations of thought to a level where they become images
upon which you operate. But that is different from being aware of the operations of
thought you employ while using them.
As explained by Ellis et al. (Chap. 6) and Tallman and O’Bryan (Chap. 8), the
idea of reflection, or thinking about one’s thinking, has been on philosophers’ minds
at least since the time of Aristotle. They also explain that Piaget was the first to
break the notion of reflection down into constituent cognitive processes. We will not
add to their extensive discussions. Instead, we will highlight essential aspects of
reflective abstraction to complete our picture of how imagery and schemes (and
therefore meanings) develop and interact in students’ thinking across their mathe-
matical development.
Piaget posited five types of abstractive processes: empirical, pseudo-empirical,
reflecting, metareflection, and thematization.
Empirical abstraction is the process of extracting common properties of sensory
experience. To empirically abstract a property, the person comes to distinguish
between objects having and not having the property. It is important to understand
that “the property” is the person’s construction, not a “real” property. A child
abstracting the property of having four legs to apply the word “dog” might at first
also apply “dog” to what we call cats.
Pseudo-empirical abstraction is the process of taking the results of one’s activity
as objects of empirical abstraction. For example, a child constructed the sequence of
arrays in Fig. 5.6. Her actions were to make a vertical line of dots, one more than
Fig. 5.6 A child’s construction of a sequence of arrays
was already there, and then complete the figure by making the same number of hori-
zontal dots as in the current array.
When the child was asked about the number of dots in each array, she noticed
that each array was a square, so she said, “1, 4, 9, 16, 25.” When asked how many
would be in the next array, she squared six to say “thirty-six.” In other words, when
asked about the number of dots in each array, she took the arrays as if given to her.
Her reasoning did not reflect that, by her construction method, the sixth square
would have 52 + 6 + 5 dots, or that the (n + 1)st array would have n2 + (n + 1) + n
dots, which she could then connect to her observation that each array is a square to
conclude that (n + 1)2 = n2 + (n + 1) + n. This is not to diminish the child’s accom-
plishment. Instead, it is to point out the difference between abstracting one’s reason-
ing from the activity of producing the sequence and abstracting an empirical pattern
from the products of one’s reasoning.
Reflecting abstraction was already exemplified in the example of pseudo-
empirical abstraction. A person engages in reflecting abstraction when she brings to
mind (re-presents), as best she can, the reasoning in which she engaged in a prior
occasion. A successful process of reflecting abstraction produces a new action, but
one that does not need the specific contexts accompanying the original. Successful
reflecting abstractions produce reflected abstractions.
Metareflection and thematization are constructs Piaget introduced to account for
the ever-increasing level of persons’ abstraction of logical, mathematical, and scien-
tific structures. He posited that reflecting abstraction produces reflected abstrac-
tions, which then can themselves become objects of reflection. He used
“metareflection” to capture a person reflecting on reflected abstractions (Piaget,
2001, pp. 82–84).
A person thematizes a scheme via metareflection by developing an image of its
major elements, how they work together, and how they might unfold in the context
of actual situations. This is not unlike the way storytellers thematize a story. They
come to think of the story’s major elements, how they work together, and how each
element might be unpacked into its details. As Harel (2008c, 2013, 2021) explains,
thematization of a scheme evolves over repeated occasions of employing the scheme
and reflecting on the activity of employing it. Depending on the scheme’s complex-
ity, this could happen over many years. It is worth noticing that when a scheme is
well-formed at a reflected level, one can have the experience of having implemented
that scheme. Thus, the processes of metareflection and thematization can operate on
images of schemes as well as on images of schemes’ constituent elements.
Dawkins and Norton (2022) draw upon the construct of metareflection (reflect-
ing on reflected abstractions) to account for students’ development of universally
quantified conditional statements as logical structures. “We propose populating,
inferring, expanding, and negating as four mental actions that, upon becoming
reversible and composable, can give rise to the logic of universally quantified con-
ditional statements. We adopt the view that logic is a metacognitive activity in which
people abstract content-general relationships by reflecting across their content-
specific reasoning activity.” (Dawkins & Norton, 2022, p. 1).
We (the authors) had the experience of a calculus student sharing with us her
thematization of an approximate accumulation function in DIRACC calculus
(Thompson et al., 2019). We were in a staff meeting, the conference room door was
open, and the statements in Fig. 5.7 were written on a whiteboard visible from the
hallway.
The student stepped into the room, pointed at the board, and proudly announced,
I can tell you what every line on that board means and how it all works together! You have
a function that gives the rate of change of accumulation for every value of x. You want to
approximate the accumulation from a to x, so you cut up the x-axis into parts all of length
Δx starting from a and define a function so its value for everything in a Δx interval is the
accumulation’s rate of change at the beginning of that interval. Then you assume the accu-
mulation happens at that constant rate over the whole Δx interval. You do that for complete
intervals from a to x, and then you add the accumulation through the partial interval from
left(x) to x. That’s how a value of A(x) gives an approximate accumulation from a to x.
While the student omitted some details, such as how the definition of left(x) works
the way she said and how the summation in the last line gives the accumulation she
claimed, she accomplished essentially what she set out to do. She explained each
element in her approximate accumulation scheme and how they worked together to
Fig. 5.7 Statements defining an approximate accumulation function in DIRACC calculus

produce an approximate accumulation from a to x for every value of x when all one
knows is the accumulation’s rate of change at every moment.
It is our experience that metareflection and thematization are the least empha-
sized of all the forms of reflection in Piaget’s genetic epistemology—in both instruc-
tion and in research. In instruction, we do not envision a teacher commanding his
students to metareflect or to thematize. Rather, a teacher could promote metareflec-
tion and thematization by emphasizing schemes’ stories. While the teacher’s stu-
dents must thematize their own schemes, the teacher emphasizing the story of a
scheme can open students to the possibility that there is a story to understand. In
research, it is inherently difficult to investigate metareflection and thematization as
explanatory constructs or as phenomena to investigate for two reasons. First, these
are processes that, even if a student engages in them, occur largely outside of
instruction, perhaps even in their sleep, during what Hadamard (1954) called peri-
ods of “incubation” and during periods of what Steffe (1991) called “metamorphic
accommodation”—accommodations to schemes that persist over situations and
time. Second, metareflection and thematization are difficult to investigate method-
ologically. Tallman and O’Bryan (Chap. 8) suggest an approach to engendering
metareflection in which researchers engage students in what appear to the students
as very different situations, but which nevertheless can be understood as similar at a
reflected level. Thompson (1994b) used essentially this approach with some success
to investigate a fifth grader’s construction of a reflected constant rate of change
scheme. Research on metareflection and thematization will be a fruitful area for
future advances in theory and practice.
Case Studies
We provide two case studies to illustrate the interconnections among imagery,

schemes, meanings, and reflective abstraction. The first case is a seventh-grader
learning a mathematical game. It will illustrate a youngster’s construction of
schemes regarding the game and the crucial role his imagery played in developing
them. The second case is of an adult who reasons initially at a figural level about a
mathematical task and who projects the situation to a level of existing reflected
abstractions upon becoming confused by an unexpected outcome.
Imagery in the Construction of a Nim Scheme
According to Jorgensen (2009), Nim is one of the oldest games in the world. It is a
game for two players. In one version, they start with the target number 21 and cur-
rent total of zero. Players take turns adding a number from 1 to 3 to the current total.
The player ending with 21 wins.
Diego was a 12-year-old rising seventh grader. We asked him and his parents to
let us teach him Nim and allow us to record our Zoom sessions. They agreed. Diego
played against a computer program that embedded a general winning strategy that
enabled it to win whenever the opportunity arose. Diego played against two ver-
sions of the program: (1) a fixed target of 21 and selecting numbers from 1 to 3. He
could choose which player goes first; and (2) an arbitrary target and arbitrary range
of numbers from which to choose. Diego could choose the numbers and which
player goes first. Sessions were conducted using Zoom. We met for two sessions—
June 17 and 28, 2020. The long break was when Diego attended surf camp.
We report these sessions to highlight the central role of imagery in Diego’s con-
struction of a general Nim scheme by way of the gamut of reflective processes.
Diego’s early imagery was figurative, based on re-presenting states of his play. He
focused on moves he might make based on the game’s current state. He refined his
strategy eventually by taking his reasoning as his object of thought as opposed to
states of the game as his object of thought. His re-presentations of prior reasoning
became the images upon which he operated.
Session 1: June 17, 2020
21 and 3
Pat explained the game’s rules to Diego and explained that Diego would be playing
against a computer program. Pat shared his screen to show the program, which itself
explained the game and gave an example (Fig. 5.8).
Diego said he would go first. He chose numbers from 1 to 3 at random. When the
computer played to reach a total of 17, Diego paused. After a few seconds, he said,
“I lost. No matter what I choose, the computer will get to 21.”
In the second game, Diego went first, again choosing numbers at random. Diego
paused when the computer played to reach 13. “Darn, I’m going to lose again.” He
remembered that the computer reaching 17 led to it winning, and he couldn’t stop
the computer from reaching 17.
In the third game, Diego developed the strategy of “goal numbers.”
Fig. 5.8 The computer screen in a game of Nim

Excerpt 1
Diego: To get to 21 I have to get to 17. To get to 17 I have to get to 13. To get to
13 I have to get to … nine. To get to nine I have to get to … five. To get to five I have
to get to one. To get to one I have to go first.
Diego’s image of his participation in the first two games was simply to select
numbers from 1 to 3 and add his selection to the current total. We say he operated
with figurative imagery because each of his activities (adding a new number)
resulted in a new state (a new total number), but the new total number was simply
the end of an activity sequence (his turn). He was not striving for a particular goal,
and when he considered a new total number, his thinking did not entail an image of
the reasoning he used to get to it. He reasoned about specific numbers in specific
contexts but not yet re-enacting his reasoning to reflect on it.
Diego’s scheme for Nim-21 developed as he reflected on being “blocked.” In the
first game, he experienced being “blocked” from reaching a desired number (21)
because of the total the computer gave him (17). He concluded by trying all possi-
bilities, he could not reach 21; the computer would win regardless of his choice of
number when he is given 17. In the second game, Diego had a similar experience of
being “blocked” when the computer presented him with 13—he saw that the com-
puter would reach 17 regardless of his choice, and it would therefore win for the
same reason he experienced before. In the third game, Diego employed his image of
being “blocked” to devise a strategy in which he could block the computer from
reaching a desired goal number. Diego employed his “blocking” image to block the
computer from reaching 21 by him reaching 17, then blocking the computer from
reaching 17 by him reaching 13, and so on. Diego’s images of being blocked were
at first dependent on thinking about the computer reaching 17 and the possible
moves he could make from 17 to 21. As Diego repeatedly reasoned about being
blocked and additional blocking numbers his image became less dependent on spe-
cific game states. As Diego’s blocking number scheme became more stable and less
dependent on specific states of the game, his imagery of being blocked moved from
figurative to operative. It became less necessary for Diego to consider a specific
game state when he reflected on being blocked. We say that Diego developed a
blocking number scheme to determine goal numbers and used his goal numbers to
ensure he won. He coordinated his blocking number scheme and list of goal num-
bers in playing the game. Diego had developed a Nim-21 scheme.
38 and 8
Pat then ran a program that played Nim with arbitrary target and range numbers,
suggesting they change target and range. Diego chose 38 as the target and 1–8 as the
range. Diego applied his blocking scheme to determine goal numbers of 38, 29, 20,
11, and 2. He chose to go first, selected 2, and won the game.
Diego’s behavior suggested he used more than a Nim-21 blocking scheme to
determine his goal numbers. He generalized his Nim-21 scheme of “subtract four”
to “subtract one more than the largest range number.” This had the effect of general-
izing his Nim-21 scheme to a Nim scheme.
In using his Nim scheme, Diego made a play to reach a goal number, then awaited
the total given to him by the computer to determine his next choice. He paid no
attention to the number the computer chose. He attended only to the current total
given to him. The computer’s choice did not play into his thinking. His underlying
image was to await the computer’s total, then use it to reach the next goal number.
Pat raised the matter of Diego’s number in relation to the computer’s number in
the context of reaching the next goal number.
Excerpt 2 (After Winning 38 and 8)
Pat: Do you notice a relationship between what the computer chooses and what
you choose to get to the next goal?
Diego: No, not really.
Pat: What happens when you are at a goal number so that you get to the next
goal number?
Diego: I add a number.
Pat: What about the computer?
Diego: It adds a number, too.
Pat: What about those two numbers? What has to be true about them so you get
to the next goal?
Diego: (28 second pause.) They have to add up to 9!
Pat: Why is that?
Diego: Because if I’m at a goal number … if I’m at a goal number the computer
will choose a number and then I’ll choose … I’ll choose … I’ll choose
another number to get to the next goal number … and the next goal number
is 9 away from the [goal] number I have.
Diego’s initial responses to Pat’s question confirmed our hypothesis about his
underlying image of a play. It was not an image that combined his and the comput-
er’s moves into one move that satisfies the requirement of reaching the next goal
number. Instead, his image of a play was to take the number presented to him and to
figure out the number needed to reach the next goal he’d already determined. At the
beginning of Excerpt 2, Diego’s image of goal numbers was operative because it
was not tied to a specific state of the game—he was able to discuss goal numbers
generally. Diego’s scheme for goal numbers made it possible for him to use images
of goal numbers decoupled from specific game states. His image underlying his
decision on the next play was figurative because it was tied to a specific state of the
game. He did not have a combined play scheme that would allow him to decouple
his image of his next move from the current game state.
Pat’s question, “What about those two numbers? What has to be true about them
so you get to the next goal?” was instrumental in providing Diego an occasion to
reflect on the relationships among the current goal, computer’s play, his play, and
next goal. We suspect that he organized, at least temporarily, the states of the current
and next goal as being connected by the combination of the computer’s and his plays.
We interpret Diego’s 28-second pause as him re-playing the computer’s and his
move in succession. This gave him an occasion to consider the two moves together,
as one. We cannot know whether Diego would have thought of this himself, but the
fact that he assimilated the question and resolved it suggests that, in re-playing the
game, he modified his image of a “play,” at least in that moment, to include both
players’ moves that together would move the total from one goal number to the
next. Diego’s modification of his image of a “play” in this way was a step towards
him developing an operative image of a “play.” The development of a stable com-
bined play scheme went along with the development of an operative image of a play
that was not tied to a specific game state.
33 and 7
Pat suggested one last game. Diego chose 33 and 1–7 as target and range, respec-
tively. Diego went through his working-back strategy to determine goal numbers 33,
25, 17, 9, and 1. He concluded that to reach 1, he had to go first. Despite the insight
Diego stated after the previous game, he still focused on the number presented to
him and what he had to add to reach the next goal.
Excerpt 3 (in the Midst of 33 and 7)
Pat: How are you deciding your number?
Diego: I’m looking for the number to add to get the next goal number.
After Diego won, Pat asked again if there was a relationship between the com-
puter’s number and his number when reaching the next goal. Diego quickly noted
that the sum of his and the computer’s plays needed to be 8—for the same reason he
stated earlier. However, it is important to note that, with 33 and 7, Diego did not use
the insight he’d stated earlier (regarding the sum of his and computer’s moves) in
deciding his play. Instead of Number Computer Plays + My Number = 8, Diego’s
image of play continued to be Current Total Given Me + My Number = Next Goal.
Pat ended the session by thanking Diego for participating and suggested he play
the game with his family before the next meeting. Pat texted Diego late the next day
to ask if he’d played the game with his family. Diego replied, “No, I’m still trying
to get my head around it.” We presumed by “it,” Diego meant “strategy.”
Session 2: July 28, 2020
Diego’s attendance at surf camp led to an 11-day break between sessions. He had
not played the game with anyone, but he had thought about the game. Pat asked
Diego what he remembered.
Excerpt 4
Pat: Do you remember the game we played?
Diego: Yeah. Nim. Yeah.
Pat: Do you remember how it is played?

Diego: Yeah.
Pat: How is it played?
Diego: So … first person to get to 21 wins. If you choose one you go first, if you
choose two you go second. And the first person to get to 21 wins.
Pat: How do you get to 21?
Diego: So, you gotta go first, and you choose … two … no. You have to choose//
Pat: //What numbers are you choosing from?
Diego: Oh. One, two, and three.
Pat: I think you were trying to remember the strategy you used?
Diego: Yeah.
Pat: What was that strategy?
Diego: (4 second pause.) His number and your number have to add up to four.
Pat: Why is that?
Diego: So then that, so then (yawns) so you keep getting to the places where
you know you’re able to get to 21.
Pat: Okay.
Diego: I know … I know that sounds confusing.
Pat: Well, instead of just telling me about it let’s play a game.
Diego: Okay.
We were struck by Diego’s recollection of his strategy (in bold). What had been
a transitory observation in his 7/17/20 session, an observation he never employed,
had become a defining feature of his strategy on 7/28/20, despite no interaction in
the interim regarding the game with us or between Diego and his family. Moreover,
his strategy of C + D = 4 entailed a reason for it—this strategy blocked the computer
from reaching any goal numbers. This suggests to us that on 7/17 Diego engaged in
what Steffe (1991) called functional accommodation—the modification of a scheme
in the context of using it—and that in the interim, he engaged in what Steffe (1991)
called metamorphic accommodation, which we understand as largely equivalent to
Piaget’s meaning of projecting images and actions to a reflected level. In Excerpt 4,
Diego had an operative image of combined play that was not tied to a specific game
state. The development of this operative image was possible due to his reflection
upon combined plays and the development of a stable combined play scheme.
Diego’s combined play image was part of his combined play scheme—the scheme
also included the entry points that trigger action and anticipations of action. As
Diego’s combined play scheme became more stable, his images of combined play
became more mobile and less dependent on specific aspects of the game. We say
more about this in the discussion.
21 and 3
In playing the first game with 21 and 3, Diego employed his strategy of working
backward to determine the first goal number. However, he miscalculated 21 minus
4, saying “sixteen,” getting goal numbers of 21, 16, 12, 8, 4, and 0. He chose the
computer to go first and selected his number according to his new rule that the sum
of the computer’s and his number must be 4. Diego realized he would lose when the
computer presented him with 17.
Excerpt 5
Diego: (Total is 17; D pauses for 21 seconds) Ooohhh. (Pause)
Pat: Why the long pause?
Diego: Um … cause seventeen … if I put three it’s twenty and he wins, if I put two
its nineteen and he wins, and one he could just put three and he wins.
Pat: What do you suppose the problem is?
Diego: (Pauses for 26 seconds; whistles while thinking.) Oh, I went too far back.
I was supposed to get to 17 first. So I lost. So just put two [just to finish
the game].
Pat: So, you were supposed to get to 17, but you said 16?
Diego: Yeah.
We interpret Diego’s 21-second pause as his imagining all the moves he could
make in combination with the computer’s subsequent move and his 26-second
pause as re-presenting (re-playing) his original reasoning to get his list of goal num-
bers. He recalled thinking “21 − 4 = 16” and realized he should have said “seven-
teen.” We also note that this episode confirms Diego’s confidence that applying the
condition C + D = 4 for each pair of moves was sufficient for him to win.
In the next game (21 and 3, again), Diego insisted on enacting his working-back
strategy “just to make sure.” ending at one and deciding he should go first. He
played appropriately, getting to each of his desired goal numbers. When the com-
puter presented Diego with a total of 10, Diego chose 3 to reach 13. Pat asked about
how he was deciding on his choice of numbers.
Excerpt 6
Diego: (Computer presents a total of 10) Three.
Pat: How are you figuring out what number to pick?
Diego: Whatever number he chooses, I just need to pick the number that adds up
to four.
Diego’s “C + D = 4” scheme was now at a reflected level. He knew applying it
would necessarily land him at the next goal number without having to think of what
the next goal number was.
36 and 5
Diego again employed his working-back strategy, but this time just to determine the
starting number. He counted 36, 30, 24, 18, 12, 6, 0, then said, “But we cannot get
to 0, so the computer needs to go first.” He did not mention a goal number while
playing the game. Instead, as the computer played its number, Diego played a num-
ber, so the sum of the computer’s and his numbers was 6.
Excerpt 7
Comp: (Plays 1)
Diego: Okay, five (total is six; computer chooses 5, total is 11). One
Pat: How are you deciding what number to put in?
Diego: Same thing. Whatever number they choose, I just have to choose another
number that adds up to six.
Pat: Why is six special?
Diego: It’s one more than the number we can pick … so that they can’t go over the
number we have to get to but we also can’t … err … and also we … it’s
too … and also it’s not too much for us to get to.
Diego went on to win the game. He again confirmed that his strategy no longer
relied on the total given him or the next goal number. It relied only on him determin-
ing the first number to play, using this number to decide whether he or the computer
should go first (computer first if first goal is 0; otherwise, him first), and knowing
what the sum of his and the computer’s play must be. Diego felt confident that
attending to these conditions would produce a winning strategy. He had developed
a Nim scheme.
General Nim
Pat suggested one last game—77 as the target and 1–10 as the selection range.
Diego again worked backward from the target by 11’s, getting 0 as the first goal
number. Diego chose the computer to go first and won the game. Afterward, Pat
asked about his general strategy.
Excerpt 8
Pat: Let me ask you a question.
Diego: Yeah.
Pat: It seems … and correct me if I’m wrong. It seems you start with the target
number, and then go back to find the next smallest target number//
Diego: //Yeah. That’s what happened.
Pat: And you keep going back [Diego:
Yeah] until you find the first target number [Diego:Yeah] and that tells you whether
or not you go first? [Diego: Yep] Is that right?
Diego: Yep. If it gets to one, then you have to go first, but if it gets to zero then they
have to go first.
Pat: What if you get to two?
Diego: You can still go first … unless … the boundary … the number … unless
the number limit you have to choose from is less than two … which would
be kinda boring.
Diego’s last statement shows he not only reasoned generally about his strategy,
but also considered the implications of alternative conditions (“unless the number
limit is less than two …”). We take this as evidence that Diego’s goal number images
and combined play images were now operative. Because of the generality of Diego’s
summary, Pat decided to raise the issue of efficiency.
Excerpt 9
Pat: That’s very good! Now … could you think of a way to figure out what the
first number should be without having to go back step by step?
Diego: I guess that (pause) if … oh, hold on, I’m just noticing. When it was 21,
they had to go first. Oh no. We had to go first. So I feel like, if it’s 21 … if
the number limit is divisible by the number limit you’re allowed to choose,
then you have to go first.
Pat: So 21 is divisible by 3//
Diego: //Oh, no. But 36 isn’t divisible by 5. And do you remember … can you go
back and see if they had to go first on 36? Scroll up to see the previ-
ous game.
Pat: (Scrolls back to the game of 36 and 5.) You had the computer go first and
it chose one.
Diego: Yeah, they went first. If the target number is divisible by the highest num-
ber you choose from, then you have to go first.
Pat: But 36 is not divisible by five.
Diego: And the computer went first! (Pause) So if it’s not divisible by the highest
number you’re allowed to choose from, then the other person has to go first!
Pat: Why do you suppose it works that way?
Diego: (Looking into the air) Because that … it’s always gonna leave … … like …
it’s always gonna leave … a higher number than … it’s alw … ahh … it’s
always gonna end up like … it’s never gonna end up … hmm, hold on.
(Pause) Yeah. That makes no sense. (Pause.) Here. Let’s do … so … so …
from what we’ve seen right now it’s um an odd number has to make the
computer go first and an even number has to make … wait no, not even
number. A number divisible by that means I have to go first. So let’s just
see … make … do … do like 30 … do like 32 and then do 4 … 32 is divis-
ible by 4, yeah.
Diego’s response to Pat’s question “Could you think of a way to figure out what
the first number should be without having to go back step by step?” has earmarks of
what Piaget called pseudo-empirical abstraction. By this, we mean that Diego
looked for a pattern that related game conditions (target number and range) and the
decisions he had made in light of them. He reflected on the products of his reason-
ing, not the reasoning in which he engaged to create those products. This is not a
criticism. Rather, it is an observation.
Diego tested his hypothesis on 32 as the target number and 1–4 as the range—
and lost. Pat suggested he try his working back strategy again. Diego did this—32,
27, 22, 17, 12, 7, and 2—noting he had to go first and start with 2. Pat determined it
would be a long struggle for Diego to refine his strategy further, so he used an inter-
vention common in exploratory teaching interviews (Castillo-Garsow, 2010; Moore,
2010; Steffe & Thompson, 2000). This was to offer a suggestion to see how Diego
might understand it and how he might subsequently use it. An intervention move
within an exploratory teaching interview can unveil the nature of and boundaries of
the interviewee’s schemes.
Excerpt 10
Pat: So going back … I’m going to remind you of something. Each time you
went back, you subtracted five, right? [Diego: Yeah] You subtracted one
more than the upper limit. [Diego: Hmm hmm] Repeated subtraction is
like division. [Diego: Hmm hmm] So you went back some number of fives
and you got to two. [Diego: Yeah] So, if you divide 32 by 5, what remain-
der do you get? [Diego: Two] (5 second pause) And what’s special about 2?
Diego: You could go first and get to 2, and then you could get to 7 first, you just …
you kinda win.
Pat: So let’s try that. This time I’m going to try 45 and 6. Now, without using
target numbers, see if you can find the first target number.
Diego: Seven divided by 45 is … (40 second pause; D looks in the air). The
remainder would be three. Because the closest number to that number that
is divisible by seven is 42, and 42 divided by seven is six. And then the
remainder would be three. So the first target number is three.
Pat: So who goes//
Diego: //So I would have to go first.
Pat: And when you say 42 divided by seven is six, what does that six mean?
Diego: (19 second pause) It means I would go back by seven six times.
It appeared Diego easily understood that his going back strategy entailed repeated
subtraction, and he appeared to understand Pat’s statement, “Repeated subtraction
is like division.” He also inferred that the remainder after dividing 32 by 5 would be
the first target number. Diego then applied a “division and remainder” strategy to a
game with target 45 and range 1–6, concluding he had to start with 3 and the sum of
plays had to be 7. Diego also understood the relationship between dividing 45 by 7
and his working-back strategy—dividing 45 by 7 would give the number of times he
would go back by 7 to get the starting number, and the remainder would be the start-
ing number. Afterward, Diego celebrated.
Excerpt 11
Diego: So now I know how to do it without having to go back step by step!
Pat: Isn’t that cool that you can figure out [where to start and] what number to
add without knowing any of the target numbers except the last one?
Diego: Yeah.
Pat: So, just … okay, we can finish this now. But if you could, tell me what your
general strategy would be no matter what numbers I pick.
Diego: Umm. Get to the remainder first.
Pat: Remainder of what?
Diego: Get to the remainder of the number … get to the remainder of the goal
number divided by the highest choice number plus one.
Pat: And that remainder tells you what?
Diego: And then … and then … that remainder tells you if you should go first or
if they should go first and then you always have to get … you have to see
what their number is and add another number to get the number … to
get … to get to the highest number plus one.
Pat: When should you have the computer go first? When should you decide to
have the computer go first? (Zoom connection fails) Diego? Diego?
The broken Zoom connection was because Diego’s phone battery expired.
Excerpt 12 shows their exchange of text messages following the Zoom failure.
Excerpt 12
Excerpts 11 and 12 provide evidence that Diego made metamorphic accommo-

dations to his Nim scheme. His final Nim strategy was much more efficient than his
first winning strategy. His final strategy was more efficient and qualitatively differ-
ent from his first winning strategy because he integrated his division scheme to
avoid needing to repeatedly subtract to find each goal number. Diego reflected on
images of goal numbers and combined play to accommodate his Nim scheme to be
more general.
Discussion
It is important to note that Diego did not write anything down, nor was there a vis-
ible record of his thinking. There was only what the computer presented after each
player played. Diego had no visual record of his reasoning to aid his reflections. He
only had what he could recall of his reasoning and its results as images to
reflect upon.5
5
It is plausible that had Diego created a written record, his writing might have prompted him to
engage more frequently in pseudo-reflective abstraction.
Diego’s imagery for the game rested in his re-presentations of his reasoning, and
sometimes re-presentations of his conclusions. The role of Diego’s images evolved
as his schemes evolved. His initial imagery consisted of his reasoning about specific
numbers in specific contexts. Diego re-enacted his reasoning in order to think about
it. This is an instance of deferred imitation. Early on, Diego’s remembered decisions
did not entail the reasoning leading to them, consistent with his scheme at that time
employing imagery figuratively. His decisions were simply the last step in his chain
of actions. Later, Diego differentiated between his decisions and the reasoning lead-
ing to them, which allowed him to begin projecting his reasoning to a reflected level.
Finally, Diego generalized his images from specific game states to arbitrary game
states, consistent with his imagery operating at a third level because he had created
his Nim scheme at a reflected level. We provide a detailed summary of Diego’s
development of his Nim and General Nim schemes below.
Nim Scheme
• An image of being blocked. Diego experienced “I was blocked” twice. He devel-

oped the image, “The computer gave me a number that kept me from reaching a
goal number. There are numbers that ‘block’ a player from winning.”
• A “Blocking” scheme—find all the blocking numbers. A blocking number keeps
the computer from reaching the next goal number and ensures he can reach the
next goal number. Diego’s blocking scheme relied on his image of a block-
ing number.
• A “First number” scheme—if the first goal number is 0, the computer goes first.
If first goal number is not zero, go first and select that number. This scheme
relied on Diego having an image of having executed his blocking scheme.
• A “Combined play” image—a combined play takes the game from one goal
number to the next. In Excerpts 2 and 3, Diego reasoned that the sum of his and
computer’s play had to be a certain number (9 in one instance, 8 in the other), but
he seemed not to have had a “combined play” image. He did not recall these
conclusions upon playing the next game. It was after the 11-day break that Diego
seemed to have an image of the computer’s play and his play as one play.
• A “Combined play” scheme—decide what to play based solely upon the com-
puter’s play and the selection range.
• A Nim scheme—we saw in Excerpt 7 that Diego coordinated his First Number
scheme with his Combined Play scheme to form a Nim scheme. He was confi-
dent his Nim scheme would produce a winning strategy regardless of the target
number and selection range.
Diego had developed a scheme for Nim. He developed it over time by coordinat-
ing his blocking scheme, first number scheme, and combined play scheme to make
a strategy for winning the game regardless of the target number and selection range.
His coordination of schemes was enabled by having projected each to a reflected
level. We are comfortable saying this because he could articulate each in general
terms, not reliant on any specific game. In each case, the projection happened as a
result of reflecting on images of his reasoning that he gained from playing the game.
Diego’s images became more mobile and flexible. Figurative images became opera-
tive as Diego’s schemes developed through reflection and repeated reasoning.
One of Piaget’s defining characteristics of operating at a reflected level is that the
person is aware of their schemes and uses them to explain their reasoning. That
Diego coordinated his blocking, first number, and combined play schemes, and
explained how they worked together confirms to us he had indeed projected them to
a reflected level.
General Nim Scheme
Diego went beyond his initial Nim scheme. He assimilated Pat’s suggestion to think
of repeated subtraction as division and used that idea to develop more than a win-
ning strategy. He employed already-developed schemes for division (as repeated
subtraction) and remainder to understand Pat’s suggestion and saw how it related to
his blocking scheme (repeated subtraction) in determining the first goal number. He
modified his Nim scheme by incorporating the scheme “First Number = mod (N,
M).”6 Had Diego not had a well-developed division scheme, he would not have seen
the relevance of Pat’s suggestion.
We note in closing that Diego’s speech over time gives an indication of how he
generalized his imagery. His earlier statements were about specific numbers. Later
statements were about “computer’s number,” “my number,” “goal number,” and
“one more than the biggest number we can choose.” We take Diego’s use of literal
names for states as enabling him to differentiate his reasoning from the specific
contexts in which his reasoning occurred and from the specific conclusions he drew.
This also supported Diego’s projection of his reasoning to a reflected level. His use
of literal names for objects upon which he acted supported his focus on the actions
he took to get from one state to another.
We interviewed Diego again after a lapse of 15 months. He did not recall Nim
nor how it is played. We reminded him of the rules and played a game of 21. After
a short pause, he recalled his Nim scheme (finding goal numbers and first number)
and used it to win. We then played 45 and 1–8, and he again employed his Nim
scheme to have the computer go first, and he won. In the third game (67 and 7), he
recalled his generalized Nim scheme, found the remainder of 67 ÷ 8, chose to go
first, gave 3 as his first number, and won. This all happened in less than 20 minutes.
The rapidity with which Diego reconstructed his generalized Nim scheme suggests
to us we are correct to have called it a scheme.
6
This is our description, not Diego’s.
Implications for Math Education
We shared Diego’s case study specifically to highlight that students’ images of hav-
ing reasoned are the primary fodder for productive reflection and hence for produc-
tive scheme formation. Students’ images of their reasoning become transformed
into operations of thought they can apply outside specific contexts in which their
reasoning occurred. This fact has implications for mathematics teaching and math-
ematics education research.
Implications for Mathematics Teaching

To help students form images of having reasoned so that they may reflect upon them
is not the same as asking, “Why did you do that?” or “How do you know that?”
Those questions often sound to students like they are being policed. A teacher high-
lights reasoning instructionally by engendering reflective discourse (Cobb et al.,
1997; Stein et al., 2008) with and among students, and by creating didactic objects
to support reflective discourse (Thompson, 2002). Designing instruction to bring
students’ imagery into the open and to support reflective discourse means to orient
students to discuss ways they are understanding situations (“What do you see going
on in this situation?” “Share with us what you imagined about this situation when
you said that?”), meanings and reasoning they are trying to convey (“Help us under-
stand what you meant by that?”), ways they are understanding diagrams and anima-
tions (“What do you see happening in this animation?” “What do you see this
diagram depicting?”), and their reasoning in the context of solving a problem
(“Please share your strategy, if you can.” “What stood out to you when you decided
to divide?”). Of course, those cannot be idle questions. A teacher conveys genuine
interest by incorporating students’ answers into the classroom conversation.
Fostering reflective discourse also entails having students attempt to understand
others’ reasoning and reflect on meanings they might intend. Reflective discourse
takes students’ imagery, meanings, and reasoning as objects of class discussion.
Instructors fostering reflective discourse continually demonstrate that they care
about and value students’ understandings—instructors convey to students that they
are interested in students’ images and meanings-in-the-moment (Hackenberg,
2010). Cobb et al. (1997) and Clark et al. (2008) reported that teachers’ consistent
support of reflective discourse can positively affect students’ attitudes and class-
room atmospheres. Under a teacher’s guidance, sharing the foundations of one’s
thinking, and expecting the same of others, becomes a classroom mathematical
practice (Yackel & Cobb, 1996).
Although Pat’s interactions with Diego were in an interview setting, Pat’s ques-
tions to Diego did have an instructional effect. For example, Pat’s questions, “How
are you deciding what number to select?” and “Do you see a relationship between
the computer’s number and your number?” appear to have caused Diego to reflect
on his reasoning when he might not have done so otherwise. Also, Pat’s questions
prompted Diego to formulate responses to those questions. As noted by Inhelder
and Piaget (1964), the attempt to express one’s reasoning to oneself, or to
communicate one’s reasoning to another person, is a primary stimulant for reflec-

tion. Questions like these, offered at timely moments, can be incorporated into
instruction.
Implications for Mathematics Education Research

Diego’s case exemplifies a methodological focus on students re-presenting their
reasoning to themselves as necessary for reflection. This focus entails timely probes
about what the situation under discussion is to the student. It also exemplifies a
focus on asking questions that prompt students to explain their decisions. But probes
into students’ decision-making processes must be crafted carefully. They must not
appear to the student as demands for justification. Instead, you want probes to con-
vey to students that you are genuinely interested in how they are thinking. “Help me
understand how you thought about this?” exemplifies a genre of questions that can
be useful in gaining insight into students’ imagery and reasoning.
Diego’s case also highlights our stance that it is students’ images of having rea-
soned that provide the fodder for reflection. To live this stance in your research
requires that you take students’ verbal statements and symbolic work as a clouded
window into their thinking—that you must probe to gain insight into what they
meant when they said or wrote what they did and how they imagined their actions
being relevant to the situation as they conceived it.
Imagery in the Projection from Figurative to Reflected Thought
Piaget spoke of two kinds of reflecting abstraction. The first is to construct schemes
at a reflected level while the second is to reason at first with schemes at a figural
level and then move one’s reasoning to counterpart schemes that have been created
at a reflected level.7 This case is a study of the latter.
Michael and Robert are mathematics educators. Michael shared with Robert a
task for students asking them to predict the graph of y = sin (3x + 1) as a transforma-
tion of the graph of y = sin (x). Robert first thought about the problem in terms of
slots—(_ + b), largely as a figural generalization of (x + b). He envisioned the graph
being transformed according to y = sin (_) and then translated according to the
value of b. Robert knew y = sin (3x) would “compress” the graph of y = sin (x) by
moving each value of sin(x) from above the value of x to above the value of x/3. He
knew “+1” would translate the graph of y = sin (3x) one to the left by putting each
value of sin(3x) above the value of 3x − 1. Robert concluded that the graph of
y = sin (3x + 1) is the graph of y = sin (3x) shifted by −1. Michael said, “That’s not
correct.”
7
We have seen this distinction translated in different ways by different translators and we do not
know which terms to use for them.
Robert was immediately puzzled—he was sure his reasoning was correct. He
first wondered how to test his claim. He knew, according to his theory of the situa-
tion, that all points on the graph of y = sin (3x) would be shifted to the left by 1. He
decided to produce both graphs on the same axes and examine one point on the
graph of y = sin (3x) shifted like he thought it should. Robert’s investigation con-
firmed that the graph of y = sin (3x + 1) is not the graph of y = sin (3x) shifted by
1 (Fig. 5.9).
Robert puzzled about why the graph of y = sin (_ + 1) is not necessarily the
graph of y = sin (_) shifted by 1. He used the mouse to highlight points on the graph
and noticed the graph seemed shifted to the left by 1/3 instead of 1. He wondered,
“How is it possible for three as a coefficient of x to affect the effect of adding 1?”
Robert eventually asked himself, “What am I doing when I shift a graph by
changing its argument?” Focusing on the idea of argument opened him to think of
sin(3x + 1) as a composite function. He then considered sin(3x + 1) as sin(ax + b)
and sin(ax + b) as the composite function h(k(x)). Robert employed similar imagery
as initially to understand how a graph is shifted when a function’s argument is itself
a function: Start with a value x = c (Fig. 5.10a), move the value x = c on the x-axis
by the function k to arrive at x = k(c) (Fig. 5.10b), “pick up” the value of h(k(c))
(Fig. 5.10c), move from x = k(c) back to x = c by k−1(k(c)) (Fig. 5.10d), then plot the
value of h(k(c)) above x = c (Fig. 5.10e). He concluded that the graph of y = h(k(x))
will appear to be the graph of y = h(x) but with each value h(x0) plotted above or
below x = k−1(x0), provided k−1(x0) exists.
Robert then applied this result to the graph of y = sin (k(x)) where k(x) = ax + b
to see that the graph of y = sin (x) is “shifted” by the function k−1 so that each value
of sin(x) appears above or below k−1(x) = (x − b)/a. In this particular case, the graph
of y = sin (3x + 1) is the graph of y = sin (x) with each value of sin(x) plotted above
or below (x − 1)/3. He explained his new approach as,
You want to anticipate the graph of y = h(k(x)) given the graph of y = h(x). Imagine standing
on the graph of y = h(x) at, say, x = x0. Where did this value x0 come from? It came from a
value x = c so that x0 = k(c). Where will the value of h(x0) appear on the graph of y = h(k(x))?
It will appear above or below c = k−1(x0). In the case of y = sin (x) and y = sin (3x + 1), any
value sin(x0) will appear at a value x = c so that x0 = 3c + 1, or c = (x0 − 1)/3. This tells me
that to get the graph of y = sin (3x + 1) start with the graph of y = sin (x), shift it to the left
by 1, then compress that graph by 1/3. But this will be true of any function f. The graph of
y = f(3x + 1) will be the graph of y = f(x) but with each value f(x) appearing above or below
(x − 1)/3.
Robert shared his conclusion with Michael, who agreed that the graph of
y = sin (3x + 1) is the graph of y = sin (x) shifted by 1, then compressed by a factor
of 1/3. Michael, however, had not considered the general case of composite func-
tions that Robert used to arrive at this specific result.
How do we possess a record of Robert’s inner thoughts? Because Robert was Pat
Thompson and Michael was Alan O’Bryan. Pat, realizing this was a potentially
important event, wrote a log of his thoughts as he puzzled through his confusion.
Figure 5.10 is our rendition of his unorganized drawings.
The central point of this example is that Pat first employed imagery at a figural
level—the level of action regarding an existing scheme. He initially assimilated the
Fig. 5.9 Robert’s test of his initial claim
question to a well-developed scheme for transforming graphs, which allowed him to

think primarily in terms of imagining specific actions on a graph in relation to the
specific symbolic form y = sin (_ + 1). He thought of y = sin (_ + 1) as y = sin (x + 1),
concluding that the graph of y = sin (_ + 1) would be the graph of y = sin (_) shifted
by 1 to the left. He moved to a level of reflection only upon being faced with the
invalidity of his reasoning.
Even at a reflected level, Pat employed images like what he conjured at a figural
level. He envisioned how the original and new graphs are related by picking an
arbitrary value on the x-axis where a value of the composite function would be plot-
ted and moving to a location on the x-axis (by evaluating a function’s argument)
where the original function would be evaluated. These images initially were figural
with respect to Pat’s “translate a graph” scheme that employed them. They became
operative when he moved to a reflected level to think of y = sin (3x + 1) as y = h(k(x)).
At a reflected level, he thought of a function and its graph, but not a specific function
or a specific graph. Any graph would support his thinking of movements on the
x-axis by an arbitrary function k and its inverse. He also thought of an arbitrary
argument to the composite function. His images were arbitrary while still providing
a context for the transformations he employed—evaluating the composite function’s
argument to get a location for evaluating the original function, then using the argu-
ment’s inverse to move that value of the function back to where it would be plotted
as a value of the composite function. He then thought of this transformation as being
applied to every value in the domain of the argument. Pat resolved his initial confu-
sion about the graph of y = sin (3x + 1) in relation to the graph of y = sin (x) by
answering a general question about the graph of a composite function y = f(g(x)) in
relation to the graph of y = f(x) for any functions f and g. He also understood his
reasoning applied only where g has an inverse function.
It is important to understand that this is not a story about Pat constructing a
higher level scheme, like Diego, through the abstractive phases of empirical,
pseudo-empirical, and reflecting abstractions. He already possessed the schemes for
functions, function notation, function inverse, and function graph he eventually
Fig. 5.10 Our rendition of Robert’s drawings and his reasoning about them
employed. Instead, this is a story of someone projecting their figurative reasoning to

an already-present reflected level—a level of already-existing operative schemes.
Pat understood all along he was solving the original problem, but with the additional
understanding that he was solving a general version of that specific problem. Pat’s
attention was focused initially on the graph of y = sin (3x + 1) in relation to the
graph of y = sin (x). He re-imagined the problem to be about the graph of y = h(k(x))
in relation to the graph of y = h(x) for arbitrary functions h and k.
It is also important to note that the role of Pat’s images changed from his initial
to final thoughts. Initially, Pat’s imagery provided a template for his assimilation of
the problem, assimilating y = sin (3x + 1) as y = sin (_ + 1), where “_ + 1” was a
figurative generalization of x + 1. Even though his imagery for y = sin (_ + 1)
involved movement, it was figural regarding the scheme to which he assimilated
it—a scheme in which the added constant specified the direction and distance the
function’s graph would be translated. Pat’s images at a reflected level served as

arbitrary states of transformations—moving an arbitrary displacement on a number
line with respect to an arbitrary graph. Pat understood movements along an axis
being “caused” by evaluating functions and understood locations on the number line
as related by being values of an argument or its inverse. This role of imagery is in
line with Piaget’s third form, describing the role of images in relation to operative
thought (see quotation on page 135). They served as “momentary states” in Pat’s
reasoning about transformations from original graph to desired graph. The specific
images Pat employed were not necessary to invoke the transformations that
related them.
Although Pat did not construct a higher level scheme through elaborate abstrac-
tive processes of differentiation, integration, and so on, he did construct a new
scheme—and therefore a new meaning—for “transform a graph.” He connected
(assembled) existing schemes in, for him, a novel way. Initially, his meaning for
“transform a graph” was to think of the form of a function’s argument and what it
implied about how a function’s graph changes. His meaning at the end was crystal-
ized as how the original function’s values are “re-positioned” on the independent
axis by the inverse of the original function’s argument when the argument is viewed
as a function.
Implications for Mathematics Education
We shared Robert’s (Pat’s) case study specifically to highlight again that images of
having reasoned are the primary fodder for productive reflection. This time, unlike
Diego’s case, “productive reflection” meant to project the situation as Pat originally
conceived it to a reflected level of already-existing schemes. That imagery of having
reasoned is important even when moving to reflected mathematical thought has
implications for mathematics teaching and mathematics education research.
Implications for Mathematics Teaching and Mathematics Education Research
Solving a specific problem by solving a generalized version of it is a standard move

in higher mathematics. At the same time, it is a rare move in school mathematics
and a move made without students’ noticing it in undergraduate mathematics. We
are unaware of research into this phenomenon from a genetic epistemology founda-
tion. The closest we know is research on problem posing (see Cai et al., 2015),
especially the early work by Brown and Walter (1983, 1993). Problem posing, as
originally crafted by Brown and Walters, is to provide facts about a situation and ask
students to craft problems from these facts. Cai et al. (2015) surveyed research on
problem posing as an instructional technique, concluding that employing it in
instruction has a positive impact on students’ problem-solving abilities.
We suspect students in problem-posing studies engaged in various forms of
reflection to create their problems. Reports that “more able” students pose more and
more complex problems than “less-able” students (Cai et al., 2015; Marsh & Yeung,
1998) suggests to us that “more able” students were engaging in projection to
reflected thought. However, we are unaware of problem-posing studies that exam-
ined students’ behaviors from a perspective of images they formed and their reflec-
tions on those images. One technique employed by Brown and Walters is ripe for
research on students’ imagery and reflection. It is to have students revisit a problem
repeatedly, to relax constraints in each iteration so that they produce ever more gen-
eral versions of the original problem. Generating generalized problems, and dis-
cussing their generalization processes, could provide opportunities for reflection.
Researching their solving activities for the problems they generate could then pro-
vide occasions to explore the connections they actually make between their underly-
ing imagery and reflective processes.
Discussion
Our thesis throughout this chapter has been that images of having reasoned are the
foundation for reflection and scheme development. We stressed that imagery
includes visualization but includes far more than visualization. We recapped Piaget’s
levels of imagery and expanded their meaning to make them useful for modeling
mathematical scheme formation at any level of sophistication. We included the case
study involving composite functions specifically to show Piaget’s constructs can be
used to model higher level mathematical thinking. We illustrated the interplay
among imagery, reflection, and scheme formation through two case studies and
explained the implications of each for mathematics teaching and research.
In this discussion, we will highlight an aspect of Diego’s case study that was
central to the work with Diego but remained tacit in our accounts. It is our prepara-
tion for the interviews—task selection and design together with a conceptual analy-
sis of the game.
We settled on Nim as the context of our interviews for two reasons. First, we
wanted to avoid as much as possible Diego’s need to create written records of his
work. We are not suggesting that symbolizing is unimportant—far from it. Our past
research and teaching, however, convinced us that one effect of students’ mathemat-
ical schooling is they often engage in premature symbolization. We say “premature”
because students often create inscriptions that then turn into the objects of their
attention. Their focus on past inscriptions then diverts their attention from the rea-
soning that led to them. We also considered that readers might think a case study of
learning the game of Nim is unrelated to learning school mathematics. This would
be true if we have in mind standard school mathematics. But reflective discourse is
not standard in school mathematics, and we wished to highlight that it is students’
reflection on their images of prior reasoning that is central to their advancement. We
argue that the case of a student constructing a fairly complex scheme without
recourse to pseudo-empirical abstractions from written work is highly relevant to
ways students could learn mathematics in school. As the last interchange with Diego
showed, he did symbolize his thinking, but he did so only after his thinking advanced
to a state where the symbols retained their meaning within his reasoning.
Glasersfeld defined conceptual analysis by a question, “What mental operations
must be carried out to see the presented situation in the particular way one is seeing
it?” (Glasersfeld, 1995, p. 78). Thompson expanded Glasersfeld’s meaning of con-
ceptual analysis to include four uses:
1. in building models of what students actually know at some specific time and
what they comprehend in specific situations,
2. in describing ways of knowing that might be propitious for students’ mathemati-
cal learning,
3. in describing ways of knowing that might be deleterious to students’ understand-
ing of important ideas and in describing ways of knowing that might be problem-
atic in specific situations,
4. in analyzing the coherence, or fit, of various ways of understanding a body of
ideas. Each is described in terms of their meanings, and their meanings can then
be inspected in regard to their mutual compatibility and mutual support.
(Thompson, 2008, p. 59)
We employed conceptual analysis according to #1 in our analysis of Diego’s inter-
views. We employed it according to #2 in our preparations for the interview—we
analyzed the game of Nim according to what someone must understand to play it at
the highest level and what schemes might be necessary to get there. We also drew
on our own experience playing the game and from watching others play it.
For any game with target N and range 1–M, we considered My First
Number = mod (N, M) coordinated with Computer′s Play + Human′s Play = M + 1
to be the most sophisticated strategy. We also anticipated that the first scheme would
be Blocking and the second would be Goal Numbers. What we had to consider was
how Diego might fill in the gaps between the first schemes and the final scheme.
We anticipated that the Computer′s Play + Human′s Play = M + 1 scheme was
crucial for Diego to advance to using division, for without that scheme he would not
see that successive goal numbers are generated by repeatedly subtracting his and the
computer’s combined play. This was why Pat repeatedly asked Diego, “How are
you deciding your number?” and later asked, “Do you see a relationship between
the computer’s number and your number?”
We also anticipated it would be crucial that Diego see a connection between
repeatedly subtracting the combined play to determine the first number and dividing
the target by one more than the range to find a remainder. When Pat determined this
insight would be long in coming, he pointed out to Diego that he was using repeated
subtraction and suggested the connection between repeated subtraction and division
to see what Diego would make of it. Finally, we decided not to suggest Diego write
anything down for the reasons we explained earlier.
The case of Robert (Pat) illustrates a second form of reflection that turns figura-
tive imagery into operative imagery—the projection of a context assimilated figu-
rately to schemes that employ images operatively. The trigger for this projection
was Pat thinking of the “input” to the sine function as an argument to the sine
function, which then led him to think of the original context more generally as a
composite function. The case of Robert is different from Diego’s in that Pat already
possessed the schemes to which he projected the context. We argued that Pat did not
construct a scheme that employed images operatively. He already possessed those
schemes. Instead, Pat constructed a new meaning for “transform a graph.”
In conclusion, we say again that our main purpose was to highlight three argu-
ments. The first is that imagery, as re-presentations of experience, includes far more
than visualization. The second is that a main function of imagery in students’ math-
ematical learning is that they form images of having reasoned. This includes the
kind of reasoning in which Diego engaged, it includes reasoning students use to
interpret diagrams or animations, and it includes reasoning they engage in to com-
prehend a problem situation along with reasoning they engage in to solve it. The
third is that imagery, as a construct, does not stand alone. Imagery as a construct is
useful only to the extent that it allows one to focus on the contexts of students’
schemes and meanings and to employ reflective abstraction as a construct for
explaining and investigating students’ mathematical learning.
References
Brown, S. I., & Walter, M. I. (1983). The art of problem posing. Franklin Institute Press.
Brown, S. I., & Walter, M. I. (1993). Problem posing: Reflections and applications. Erlbaum.
Byerley, C., & Thompson, P. W. (2017). Secondary teachers’ meanings for measure, slope, and
rate of change. Journal of Mathematical Behavior, 48(2), 168–193.
Cai, J., Hwang, S., Jiang, C., & Silber, S. (2015). Problem posing research in mathematics:
Some answered and unanswered questions. In F. M. Singer, F. M. Ellerton, & J. Cai (Eds.),
Mathematical problem posing (pp. 3–34). Springer.
Castillo-Garsow, C. C. (2010). Teaching the Verhulst model: A teaching experiment in covaria-
tional reasoning and exponential growth [Dissertation, Arizona State University]. http://goo.
gl/9Jq6RB
Clark, P. G., Moore, K. C., & Carlson, M. P. (2008). Documenting the emergence of “speak-
ing with meaning” as a sociomathematical norm in professional learning community dis-
course. The Journal of Mathematical Behavior, 27(4), 297–310. https://doi.org/10.1016/j.
jmathb.2009.01.001
Clement, J. J. (2018). Reasoning patterns in Galileo’s analysis of machines and in expert pro-
tocols: Roles for analogy, imagery, and mental simulation. Topoi. https://doi.org/10.1007/
s11245-018-9545-5
Cobb, P., & von Glasersfeld, E. (1984). Piaget’s scheme and constructivism. Genetic Epistemology,
13(2), 9–15.
Cobb, P., Boufi, A., McClain, K., & Whitenack, J. (1997). Reflective discourse and collective
reflection. Journal for Research in Mathematics Education, 28(3), 258–277.
Cooper, R. G. (1991). The role of mathematical transformations and practice in mathematical
development. In L. P. Steffe (Ed.), Epistemological foundations of mathematical experience
(pp. 102–123). Springer-Verlag.
Davis, R. B., Jockusch, E., & McKnight, C. (1978). Cognitive processes involved in learning alge-
bra. Journal of Children’s Mathematical Behavior, 2(1), 10–320.
Dawkins, P. C., & Norton, A. (2022). Identifying mental actions for abstracting the logic of condi-
tional statements. The Journal of Mathematical Behavior, 66, 100954. https://doi.org/10.1016/j.
jmathb.2022.100954
Dewey, J. (1910). How we think. DC Heath.

Frank, K. M. (2017). Examining the development of students’ covariational reasoning in the con-
text of graphing [Ph.D. dissertation]. Arizona State University.
Hackenberg, A. J. (2010). Mathematical caring relations in action. Journal for Research in
Mathematics Education, 41(3), 236–273.
Hadamard, J. (1954). The psychology of invention in the mathematical field. Courier Corp.
Harel, G. (2008a). DNR perspective on mathematics curriculum and instruction, part I:
Focus on proving. ZDM – Mathematics Education, 40, 487–500. https://doi.org/10.1007/
s11858-008-0104-1
Harel, G. (2008b). DNR perspective on mathematics curriculum and instruction, part II: With
reference to teacher’s knowledge base. ZDM – Mathematics Education, 40, 893–907. https://
doi.org/10.1007/s11858-008-0146-4
Harel, G. (2008c). What is mathematics? A pedagogical answer to a philosophical question. In
R. B. Gold & R. Simons (Eds.), Current issues in the philosophy of mathematics from the per-
spective of mathematicians (pp. 265–290). Mathematical Association of America.
Harel, G. (2013). Intellectual need. In K. Leatham (Ed.), Vital directions for research in mathemat-
ics education (pp. 119–151). Springer.
Harel, G. (2021). The learning and teaching of multivariable calculus: A DNR perspective. ZDM –
Mathematics Education, 53(3), 709–721. https://doi.org/10.1007/s11858-021-01223-8
Inhelder, B., & Piaget, J. (1964). The early growth of logic in the child: Classification and seria-
tion. Routledge & Kegan Paul.
Johnckheere, A., Mandelbrot, B. B., & Piaget, J. (1958). La lecture de l’expérience [Observation
and decoding of reality]. P. U. F.
Jorgensen, A. H. (2009). Context and driving forces in the development of the early computer
game Nimbi. EEE Annals of the History of Computing, 31(3), 44–53.
Joseph, S., & Williams, R. (2005). Understanding posttraumatic stress: Theory, reflections, context
and future. Behavioural and Cognitive Psychotherapy, 33(4), 423–441. https://doi.org/10.1017/
S1352465805002328
Lobato, J., & Thanheiser, E. (2002). Developing understanding of ratio-as-measure as a founda-
tion for slope. In B. Litwiller (Ed.), Making sense of fractions, ratios, and proportions: 2002
yearbook of the NCTM. National Council of Teachers of Mathematics.
Marsh, H., & Yeung, A. S. (1998). Longitudinal structural equation models of academic self-
concept and achievement: Gender differences in the development of math and English con-
structs. American Educational Research Journal, 35(4), 705–738.
Miller, A. I. (1996). Metaphors in creative scientific thought. Creativity Research Journal, 9(2–3),
113–130.
Montangero, J., & Maurice-Naville, D. (1997). Piaget or the advance of knowledge (A. Curnu-
Wells, Trans.). Lawrence Erlbaum.
Moore, K. C. (2010). The role of quantitative reasoning in precalculus students learning central
concepts of trigonometry [Dissertation]. Arizona State University.
Moore, K. C., Silverman, J., Paoletti, T., & LaForest, K. (2014). Breaking conventions to support
quantitative reasoning. Mathematics Teacher Educator, 2(2), 141–157. https://doi.org/10.5951/
mathteaceduc.2.2.0141
Moore, K. C., Silverman, J., Paoletti, T., Liss, D., & Musgrave, S. (2019). Conventions, habits,
and U.S. teachers’ meanings for graphs. The Journal of Mathematical Behavior, 53, 179–195.
Piaget, J. (1936). The origins of intelligence in children (2637414). W. W. Norton.
Piaget, J. (1965). The child’s conception of number. W. W. Norton.
Piaget, J. (1967). The child’s concept of space. W. W. Norton.
Piaget, J. (1968a). On the development of memory and identity (E. Duckworth, Trans.). Barre
Publishers.
Piaget, J. (1968b). Six psychological studies. Vintage Books.
Piaget, J. (2001). Studies in reflecting abstraction (R. L. Campbell, Trans.). Psychology Press.
Piaget, J., & Garcia, R. (1991). Toward a logic of meanings. Lawrence Erlbaum.
Powers, W. (1973a). Behavior: The control of perception. Aldine.
Powers, W. (1973b). Feedback: Beyond behaviorism. Science, 179, 351–356.
Steffe, L. P. (1991). The learning paradox. In L. P. Steffe (Ed.), Epistemological foundations of
mathematical experience (pp. 26–44). Springer-Verlag.
Steffe, L. P. (2010). Perspectives on children’s fraction knowledge. In Children’s fraction knowl-
edge. Springer.
and essential elements. In R. Lesh & A. E. Kelly (Eds.), Research design in mathematics and
science education (pp. 267–307). Erlbaum.
Stein, M. K., Engle, R. A., Smith, M. S., & Hughes, E. K. (2008). Orchestrating productive mathemat-
ical discussions: Five practices for helping teachers move beyond show and tell. Mathematical
Thinking and Learning, 10(4), 313–340. https://doi.org/10.1080/10986060802229675
Stump, S. L. (2001). Developing preservice teachers’ pedagogical content knowledge of
slope. The Journal of Mathematical Behavior, 20(2), 207–227. https://doi.org/10.1016/
S0732-3123(01)00071-2
Thompson, P. W. (1994a). Images of rate and operational understanding of the fundamental theo-
rem of calculus. Educational Studies in Mathematics, 26(2–3), 229–274.
Thompson, P. W. (1994b). The development of the concept of speed and its relationship to con-
cepts of rate. In G. Harel & J. Confrey (Eds.), The development of multiplicative reasoning in
the learning of mathematics (pp. 179–234). SUNY Press.
Thompson, P. W. (1996). Imagery and the development of mathematical reasoning. In L. P. Steffe,
P. Nesher, P. Cobb, G. A. Goldin, & B. Greer (Eds.), Theories of mathematical learning
(pp. 267–283). Erlbaum.
Thompson, P. W. (2002). Didactic objects and didactic models in radical constructivism. In
K. Gravemeijer, R. Lehrer, B. van Oers, & L. Verschaffel (Eds.), Symbolizing, modeling and
tool use in mathematics education (pp. 197–220). Kluwer.
dations of mathematics education. In Proceedings of the annual meeting of the International
Group for the Psychology of Mathematics Education (Vol. 1, pp. 31–49). PME.
Thompson, P. W. (2016). Researching mathematical meanings for teaching. In L. D. English &
D. Kirshner (Eds.), Handbook of international research in mathematics education (3rd ed.,
pp. 435–461). Taylor & Francis.
Thompson, P. W., & Dreyfus, T. (1988). Integers as transformations. Journal for Research in
Mathematics Education, 19, 115–133.
Thompson, P. W., & Saldanha, L. A. (2003). Fractions and multiplicative reasoning. In J. Kilpatrick,
L. L. Hatfield, & K. C. Moore (Eds.), Epistemic algebra students: Emerging models of stu-
dents’ algebraic knowing (Vol. 4, pp. 1–24). University of Wyoming. http://bit.ly/1aNquwz
Thompson, P. W., Ashbrook, M., & Milner, F. A. (2019). Calculus: Newton, Leibniz, and Robinson
meet technology. Arizona State University. http://patthompson.net/ThompsonCalc
rience and Piaget’s approach. In L. P. Steffe (Ed.), Epistemological foundations of mathemati-
cal experience (pp. 45–65). Springer-Verlag.
von Glasersfeld, E. (2001). Scheme theory as a key to the learning paradox. In A. Tryphon &
J. Vonèche (Eds.), Working with Piaget: Essays in honour of Bärbel Inhelder (pp. 139–146).
Psychology Press.
Yackel, E., & Cobb, P. (1996). Sociomathematical norms, argumentation, and autonomy in math-
ematics. Journal for Research in Mathematics Education, 27(4), 458–476.
Chapter 6
Empirical and Reflective Abstraction
Amy Ellis, Teo Paoletti, and Elise Lockwood
The Enduring Attention to Abstraction
Philosophers in the empiricist tradition have long appealed to the notion of abstrac-
tion as a way to understand how people develop knowledge. Over 2000 years ago,
Aristotle developed his concept of abstraction as a way to distinguish and recognize
different aspects of objects. In his account, one can consider a particular attribute of
an object and isolate it, creating a new idea based on the attribute alone (Bäck,
2006). Much later, abstraction as a concept appeared in John Locke’s An Essay
Concerning Human Understanding (1690/1975), in which he stated that it was the
source of all general ideas: “This is called Abstraction, whereby ideas taken from
particular beings become general representations of all the same kind” (Book II, Ch.
X, §9). Locke emphasized the role of abstraction in producing general ideas, as well
as the importance of abstract general ideas to knowledge. A century later, Kant
wrote in his Lectures on Logic (1800/1992) that by comparing objects and attending
to the feature they have in common, one can abstract from all other things to form a
concept:
To make concepts out of representations one must be able to compare, to reflect and to
abstract, for these three logical operations of the understanding are essential and universal
conditions for generation of every concept whatsoever. I see, e.g., a spruce, a willow and a
A. Ellis (*)
Department of Mathematics, Statistics, and Social Studies Education, University of Georgia,
Athens, GA, USA
e-mail: amyellis@uga.edu
T. Paoletti
College of Education and Human Development, University of Delaware, Newark, DE, USA
E. Lockwood
Department of Mathematics, Oregon State University, Corvallis, OR, USA
https://doi.org/10.1007/978-3-031-47386-9_6
170 A. Ellis et al.
linden. By first comparing these objects with one another I note that they are different from
another in regard to the trunk, the branches and the leaves, etc.; but next I reflect on that
which they have in common among themselves, trunk, branches and leaves themselves, and
I abstract from the quantity, the figure, etc., of these; thus I acquire a concept of a tree.
(9:94-5)
In Critique of Pure Reason (1781/2003), Kant also distinguished between the

capacity to construct representations of specific objects by means of the senses,
which he called intuition, and the capacity to form abstract and general representa-
tions, which he called concepts. Similarly, Wilhelm von Humboldt (1795/1907), in
sharing a number of observations about reflection, noted that the mind must “stand
still for a moment in its progressive activity, must grasp as a unit what was just pre-
sented, and thus posit it as an object against itself” (p. 581). Glasersfeld (1991)
pointed out that von Humboldt was describing a kind of abstraction: “Focused
attention picks a chunk of experience, isolates it from what came before and from
what follows, and treats it as a closed entity” (p. 2).
Although other philosophers, such as Berkeley and Hume, critiqued the notion of
abstraction as a source of concept development, it nevertheless became a lasting
theme in philosophy (Laurence & Margolis, 2012). Despite its importance, how-
ever, within the empiricist tradition, philosophers did not typically articulate pre-
cisely how the process of abstraction is supposed to work. That changed with
Piaget’s body of work. As Glasersfeld (1991) remarked, “Few, if any, thinkers in
this century have used the notion of abstraction as often and insistently as did
Piaget” (p. 8). Early in his career, Piaget distinguished empirical abstraction, which
concerns properties of objects, from another type of abstraction, which he termed
reflective. Over the course of his writings, Piaget continued to expand and refine his
ideas about abstraction, ultimately identifying and characterizing three types of
reflective abstraction (pseudo-empirical, reflecting, and reflected). Below, we
explore these different types of abstraction and examine how Piaget and other schol-
ars made sense of the distinctions between them. In doing so, we will also consider
how researchers and educators can benefit from considering the role of abstraction
in fostering student learning.
onsidering Abstraction When Making Sense

C
of Student Reasoning
Why might it be useful to distinguish different types of abstraction in Piaget’s

genetic epistemology? In order to illustrate this utility, consider an example of two
students’ reasoning on an interview task called the Growing Rectangle Problem
(Singleton & Ellis, 2020). For this task, students were asked to compare a rectan-
gle’s growth in length and corresponding growth in area (Fig. 6.1).
Both students, Angelo and Willow, produced the same general statement relating
the rectangle’s length (L) and its area (A), writing the equation A = 1.5 L. However,
the type of abstraction each student engaged in to develop the general statement
6 Empirical and Reflective Abstraction 171
Fig. 6.1 The growing rectangle problem
differed. Furthermore, we will see that these different forms of abstraction ulti-
mately afforded different opportunities for generalizing and justifying their
conclusions.
Angelo
Angelo, an eighth grade student, wrote “6” for the length and “8” for the area when
identifying values to go in the blanks. In justifying this pair, he explained, “Well,
from what I saw, it looks like this box [pointing to the growth part in dotted lines] is
a little bigger than this one [pointing to the original rectangle with solid lines], so I
guessed that it was 2 centimeters bigger.” In doing so, Angelo compared the relative
sizes of the original rectangle’s length and area compared to the rectangle’s growth
in the dotted portion of the figure. In an attempt to help Angelo produce a correct
length-area pair, the interviewer asked him to determine the rectangle’s area if the
length grew by 4 cm:
Int: So, let’s say that this grew out by 4 centimeters. So, it didn’t quite go a ways as far as
where these dotted lines are. The length just grows by 4 centimeters. How much do you
think the area would grow?
Angelo: (Long pause). I think that it would grow by 12, because if four is doubling, and if
that’s 4 centimeters (pointing to the original rectangle’s length) and that’s 4 centimeters
(gesturing along the growing length), then that (the length) would equal 8 centimeters.
And then another 6 centimeters would equal 12 centimeters.
Angelo realized that if the length were doubled, the area would also be doubled.
Under the interviewer’s direction, Angelo was then able to double again to produce
a length-area pair of 8:12, halve to produce 2:3, and halve again to produce 1:1.5.
The interviewer then asked Angelo to determine the rectangle’s added area for two
more added length values, ¾ cm and 6 cm. Angelo reasoned that for the original
length-area pair, 4:6, 4 × 1.5 = 6, and therefore he could multiply other length values
by 1.5 to find their corresponding areas.
172 A. Ellis et al.
Fig. 6.2 Angelo’s list of

length-area pairs for the
growing rectangle
The interviewer then asked Angelo to make a generalization. If the rectangle were
to grow by any length of x cm, by how much would the area grow? Angelo
looked across the results of his prior calculations (Fig. 6.2), tracing his finger
across the pairs, and wrote “1.5(x).” He explained:
Angelo: Well, I looked at my other answers, and then I looked at one. And 1 multiplied by
1.5 is still 1.5 (gestures at the middle row of his table, the pair 1 and 1.5). And with this one,
1.5 would equal that (gestured at the row of 6 and 9). And then I was thinking of the other
ones as well and then that's what I got for those. So I’m guessing they’re all multipli-
able by 1.5.
When asked whether all of the pairs would necessarily follow the pattern he identi-
fied, Angelo was unsure. He began to check each pair, beginning with the 8:12 pair,
then the 4:6 pair, and then the other pairs. The only way Angelo had to explain his
general statement was via an empirical argument that multiplying the length by 1.5
worked for all of the length-area pairs he had tried so far. Angelo was unable to
provide a justification that was deductive, and the value of 1.5 did not hold a quan-
titative meaning for him in terms of the rectangle situation.
Willow
Willow, a sixth grade student, initially made an additive comparison. She compared
the rectangle’s area of 6 cm2 to its length of 4 cm and concluded that the area
“would always be 2 more” than the length. She then, however, made a sudden
shift in how she was thinking about the rectangle growing, by imagining it as
dynamic and moving, rather than static:
Willow: Unless it will start at zero. Because if you start it at zero, if you start it from zero to
find out the actual growth, then say this is like the first they grew and this, kind of, so, this
grew by four first (gestures along the length) and then this grew by six (gestures to the
whole figure, indicating area). So, this could grow by four again (gestures along the length
to continue the figure past the original drawing), and this could grow by six again (gestures
to the whole figure to indicate the area similarly ‘coming along’ with the length). Even
though they start here they could still be like, this (the area) grew by six, this (the length)
grew by four.
When she said the figure “starts at zero,” Willow appeared to imagine the rectangle
to be moving from left to right. Like Angelo, she then engaged in the mental action
of doubling. Willow imagined the rectangle doubling in length and then understood
that area would also have to double. This understanding appeared to be connected
to a mental image of the rectangle dynamically growing. Willow then said, “It
would always be plus four and plus six, so if you said when the length grows by
eight, then the area grows by 12.”
Building on her original mental action of doubling, Willow reasoned that when
the length grew by 16 cm, the area would grow by 24 cm2, because “16 is four times
four, so it grew four times the original length kind of, then the area grows four
times.” Using similar reasoning, Willow determined that the area would grow by
3 cm2 when the length grew by 2 cm because “two is half of four, three is half of
six.” Willow also decided that if the length grew by 1 cm, the area would grow by
1.5 cm2 because “one goes into four four times” and “1.5 goes into six four times.”
At this point, Willow had not identified a direct multiplicative relationship between
length and area, even though she had written down the length-area pair 1:1.5.
Specifically, she had not yet explicitly determined that the area was 1.5 times the
length. Instead, Willow continued to reason with the original 4 cm:6 cm2 pair, find-
ing equivalent ratios by relating each new length with the original 4 cm length to
determine the multiplicative factor and then using that factor to find the new area
value. In this manner, Willow identified the pair 3 cm:4.5 cm2 because three is ¾ of
4 and 4.5 is ¾ of 6. Similarly, she identified the pair 5 cm:7.5 cm2 because five is 1¼
of 4 and 1¼ of 6 would be five 1.5 s, which she added up to get 7.5 (Fig. 6.3):
At this point, Willow stopped and said, “Oh, I think I see something.” She
explained, “One point five is ¼ of six. So, one would be ¼ (of four), so it (the added
area for an added length of 1 cm) would be 1.5. And then this (the length) grew by
one.” Willow concluded, “It’s like 1.5 bigger each time.” The interviewer asked
Willow to explain further, and she said, “Each time the growth in length goes up by
Fig. 6.3 Willow’s method

of adding five 1.5 s to
calculate 1¼ of 6
174 A. Ellis et al.
one, the growth in area, I think the growth in area equals -” [writes “A = 1.5 x L”].
Willow was then able to think of other growth increments in relation to the unit
ratio, imagining other positive increases in length as a scaled version of some factor
times 1 cm, and correspondingly the area must be that same factor times 1.5 cm2.
The Basis for Angelo and Willow’s Abstractions
Angelo’s first pair was based on his perceptual judgment of the relative sizes of the
original and extended rectangles. However, he was then able to create new length-
area pairs through the action of doubling the original 4:6 length-area pair. This then
supported Angelo in repeating his doubling action, then halving, and ultimately
noticing that because 6 = 4 × 1.5, he could multiply other length values by 1.5 to
find their corresponding area values. It was after noticing a pattern across all of his
correct length-area pairs that Angelo was able to produce the expression 1.5(x); this
expression was a representation of the pattern of outcomes across multiple pairs in
his written record. Because it represented only the pattern of outcomes, Angelo did
not have confidence in the accuracy or generalizability of his expression. He felt
compelled to check other length-area pairs and did not see his expression as neces-
sarily useful for determining new pairs beyond what was already in his table.
For Willow, the expression “A = 1.5 x L” represented something different than a
pattern of outcomes. Even when she had produced the length-area pair 1:1.5, Willow
did not yet see it as a unit relationship between length and area. It was only after
determining 1¼ of 6 as five groups of 1.5 that she suddenly realized that the 1.5
could represent the rectangle’s added area corresponding to each additional unit
increment of length: “It’s like 1.5 bigger each time.” By reflecting on her activity of
coordinating the number of groups of area (1.5 cm2) with the corresponding number
of groups of length (five groups of 1 cm), Willow was able to then re-conceive the
ratio 1:1.5 as an amount of area coupled with a 1-cm increase in length. Her expres-
sion, A = 1.5 x L, was therefore a representation of Willow’s image of the rectangle’s
growth in area for a unit growth in length. Once Willow had this expression, she
knew she could use it to determine a corresponding area for a given length.
Angelo and Willow each engaged in a form of abstraction to produce the gener-
alization A = 1.5 L. However, the type of abstraction they engaged in differed.
Angelo abstracted based on reviewing the outcome of his activity and identifying a
pattern in those outcomes, a form of result-pattern generalization (Harel, 2001).
Willow, in contrast, abstracted based on her coordinating groups of area with cor-
responding groups of length, a form of process-pattern generalization. Understanding
the different types of abstraction in Piaget’s genetic epistemology can help us make
a better sense of students’ reasoning and ultimately design meaningful instruction
that can support students’ thinking. In the sections below, we introduce and describe
the different types of abstraction Piaget discussed and then introduce extended data
episodes to characterize examples of these abstraction types.
Empirical Abstraction
Empirical abstraction concerns the abstraction of properties of objects, with senso-

rimotor experience providing the raw material for such abstractions. Piaget described
this process as “extracting the inherent properties of the object” (1948/1967, p. 25)
and “deriving the common characteristics from a class of objects” (1961/1966,
p. 189). In his early work, Piaget referred to abstractions from objects as “simple
abstractions” (e.g., Piaget, 1961/1966). However, he later began to use the phrase
empirical abstraction, in part due to the fact that engaging in such abstractions is
not, in fact, simple (Piaget, 1980). From the actor’s perspective, knowledge derived
via an empirical abstraction may seem to originate from the external objects them-
selves. However, several researchers (e.g., Dubinsky, 2002; Montangero & Maurice-
Naville, 2013; Steffe, 1991; Thompson, 1985), as well as Piaget himself (2001),
have been careful to note that such knowledge derives from internal constructions
made by the actor regarding a property of an object:
Let us note right away that this type of abstraction, even in its most elementary forms, can-
not be a pure “read-off” of data from the environment. To abstract any property whatsoever
from an object, such as its weight or its color, the knowing subject must already be using
instruments of assimilation (meanings and acts of putting into relation) that depend on
sensorimotor or conceptual schemes. And such schemes are constructed in advance by the
subject, not furnished by the objects (pp. 29–30).
We elaborate on the implications of this interplay at the conclusion of the chapter.

Some of the observables central to empirical abstraction may come from objects
themselves (e.g., color, texture), such as in the example Kant (1800/1992) described
regarding the concept of a tree. Other observables can be derived from the material
of the actor’s actions (e.g., pushing, lifting). As such, the empirical abstraction then
“provides a conceptualization which, in a way, is descriptive of the observable fea-
tures of the action’s material characteristics” (Piaget, 1976, p. 351). Describing an
example of such an empirical abstraction, Piaget (1970) remarked:
A child, for instance, can lift objects in her hands and realize that they have different
weights. Usually, big things weigh more than little ones, but that is not always true. She
finds this experientially, and her knowledge is abstracted from objects themselves. This is
simple (empirical) abstraction (pp. 16–17).
In both of the examples of weight and tree, the empirical abstraction entails extract-
ing some property from a set of objects and then classifying the objects or features
of the objects, on that basis. Thompson (1985) described this process as a “separat-
ing of the object or object’s composition into similarities and differences” (p. 196),
and Simon et al. (2004) described it as a “generalization of properties of objects”
(pp. 312–313).
We draw from the second author’s everyday life to provide an example of an
empirical abstraction. Before the age of 2, his child learned that coins could be used
to unlock bedroom doors in his house. The locks had approximately 1-cm-long slots
that could be turned with any rigid object a few millimeters in width. The child
began to test other objects (e.g., thick cardboard books, toy tongs) to explore which
176 A. Ellis et al.
objects could (and could not) turn these locks. Through this process, the child cre-
ated a concept of “objects that can open locks” via an empirical abstraction that
relied on the perceptual size and rigidity of objects within the set. In doing so, the
child extracted “a character x, from a set of objects and then classif[ied] them
together on this basis alone, a procedure we shall refer to as simple abstraction and
generalization” (Piaget, 1961/1966, p. 317). Whereas empirical abstractions, such
as this one, concern properties of objects, we next turn to reflective abstractions,
which concern the learner’s mental operations.
Reflective Abstraction
Piaget (1976) distinguished reflective abstraction as a type of abstraction that is

qualitatively different from empirical abstraction:
While it is clear that any abstractions from objects is thus ‘empirical’, the action’s pole
gives rise to both types: empirical with regard to observable features of the action, that is,
concerning a material process (a movement, a position of the hand, and so forth), and
‘reflexive’ with regard to inferences drawn from the coordinations themselves. (pp. 345–346)
Note that in the above quote, Piaget used the term “reflexive,” which is typically
translated as “reflective.” Reflective is the term we will use in this chapter. As
Montangero and Maurice-Naville (2013) described, reflective abstraction is not a
property of “reality” such as weight, color, or temperature, but rather is a property
of the learner’s activities. In fact, in defining reflective abstraction, Piaget empha-
sized its origin as being the learner’s actions or coordinations: “A reflexive abstrac-
tion is one that derives its information from the subject’s actions or, more specifically,
from their coordinations” (1976, p. 270). Reflective abstraction, Glasersfeld (1982b)
noted, occurs when the learner abstracts mental operations from the sensorimotor
context that may have given rise to them. Succinctly characterizing this difference,
Piaget (1977) explained, “Empirical abstractions concern observables and reflective
abstractions concern coordinations” (p. 319). Reflective abstraction is, therefore, an
endogenous process (driven by internal factors), in contrast to empirical abstraction,
which is exogenous in origin (driven by external factors) (Dubinsky, 2002;
Montangero & Maurice-Naville, 2013).
As we mentioned above, even empirical abstraction draws on the products of
reflective abstraction. Montangero and Maurice-Naville (2013) noted, “Piaget con-
stantly stressed the greater importance of reflective abstraction. The latter is respon-
sible for the creation of forms of knowledge (categories or classes, comparisons)
that make empirical abstraction possible” (p. 61). Indeed, reflective abstraction is
the process Piaget proposed for the development of new knowledge; it is how more
advanced concepts can be built out of existing concepts (Glasersfeld, 1982a; Simon
et al., 2004). Piaget wrote that reflective abstraction “permits the derivation from
those coordinations of new assimilatory frameworks and new structures by combin-
ing reflection onto new cognitive levels with reorganizing réflexion” (1980, p. 111).
Further, he stressed that empirical abstraction is limited to sensorimotor data,

whereas reflective abstraction is structuring at all levels. Reflective abstraction is,
therefore, a critically important process to understand and elaborate, and yet, simul-
taneously, it has been difficult for the field of mathematics education to define and
characterize this process.
The introduction of reflective abstraction is, to our knowledge, unique to Piaget’s
writings. It appeared beginning in the 1940s, but there were hints of these ideas even
prior to that. In his 1928 work, Judgment and Reasoning in the Child, for instance,
Piaget discussed the idea that knowledge is enriched when transferred to the level of
thought; he would not formally define this process, however, until much later
(Montangero & Maurice-Naville, 2013). In The Child’s Conception of Space (1948),
Piaget distinguished a type of abstraction that is drawn from the coordination of the
learner’s actions rather than from an object, and he again made this distinction in his
1950 Introduction to Genetic Epistemology. Later, Piaget began to provide addi-
tional specifications to the concept of reflective abstraction. In particular, beyond
simply differentiating reflective from empirical abstraction, Piaget (1977) intro-
duced the distinction between subtypes of reflective abstractions – pseudo-empirical,
reflecting, and reflected – which we discuss in more detail below. Before describing
these subtypes, however, we first introduce an important separation of reflective
abstraction into two phases, which are useful for making sense of the different
subtypes.
Two Phases of Reflective Abstraction
Piaget used two terms when discussing reflective abstraction, réfléchissement and
réflexion. Réfléchissement was Piaget’s word for taking an activity or mental opera-
tion developed on one level, abstracting from that level of operating, and raising it
to a higher level (Glasersfeld, 1995). This term has been translated not only as
“reflection” but also as “projection” (e.g., Moessinger & Poulin-Dubois, 1981).
Piaget (1970) noted that the term réfléchissement has at least two meanings:
In its physical sense a reflection refers to such a phenomenon as the reflection of a beam of
light off one surface onto another surface. In a first psychological sense abstraction is the
transposition from one hierarchical level to another level (for instance, from the level of
action to the level of operation). (pp. 17–18)
Glasersfeld (1982b) depicted the process of projection in this manner: “It lifts the
construction pattern out of a sensorimotor configuration, leaving behind the actual
sensory material - in other words, holding on to the connecting, relating, integrating
operations, and disregarding the stuff that was connected by them” (p. 18).
Montangero and Maurice-Naville (2013) referred to this projection phase as “the
abstraction proper” (p. 60), the process of extracting forms of organization from
one’s knowledge and reflecting what was abstracted onto a higher level. Others have
also referred to lifting one’s reasoning onto a higher level or a higher plane of
178 A. Ellis et al.
thought (e.g., Dubinsky, 2002). What Piaget meant by a higher level is a matter for
some debate and elucidation, which we discuss in more detail in the sections below.
The second phase Piaget (1977) discussed is réflexion, in which the abstracted
structure is reorganized at the higher level: “‘Reflection’ involving a mental act of
reconstruction and of reorganization on this higher level of that which has been thus
transferred from the lower one” (p. 303). Both nouns, réfléchissement and réflexion,
were formed from the verb réfléchir, whose present participle is réfléchissante.
Piaget used abstraction réfléchissante as a generic term for both phases, which is
likely why the distinction between réfléchissement and réflexion was lost when the
more generic term “reflective abstraction” took hold (Glasersfeld, 1995). Campbell
(2001) noted that Piaget’s emphasis of the optical meaning of reflection with réflé-
chissement and the cognitive meaning with the term réflexion works better in French
than in English; consequently, many call the first phase projection and the second
phase reflection, reconstruction, or reorganization. We will use the terms projection
and reorganization throughout this chapter.
Both phases, projection and reorganization, must occur for us to consider an
abstraction to be a reflective abstraction. On the higher plane of thought, more pow-
erful mental actions are present, which enables a process of constructing new com-
binations by a conjunction of abstractions (Dubinsky, 2002). Piaget considered this
construction process to be key to mathematical development, as did others (e.g.,
Dubinsky, 1991; Gallagher & Reid, 1981; Steffe, 1991). Reorganization essentially
enables the elaboration of new mental actions: “It leads to a generalization that is a
novel composition, preoperatory or operatory because it involves a new scheme that
has been elaborated by means of elements borrowed from prior schemes by differ-
entiation” (Campbell, 2001, p. 11). For some, this new scheme is simply one that is
more flexible or is composed of higher-level operations (e.g., Campbell, 2001;
Thompson, 1985); for others, this scheme must entail new actions that are coordi-
nated as operations within a reversible and composable system (e.g., Norton, 2018)
(see Chap. 3 for an elaboration of Piaget’s scheme theory). Regardless of what one
requires of the novel scheme, the reorganization phase is cyclical, with the resulting
schemes becoming “purer and purer thanks to its internal mechanism of reflection
on its reflections” (Piaget, 1977, p. 319). This further distinguishes reflective
abstraction from empirical abstraction, as reflective abstraction allows engagement
in thought experiments that do not rely on the presence of sensorimotor material.
Through this two-phase process, reflective abstraction begins with old cognitive
structures and generates new ones (Campbell, 2001). Unlike empirical abstraction,
reflective abstraction elevates all of the learner’s cognitive activities – coordinations
of actions, operations, schemes, and cognitive structures – and separates particular
characteristics of those activities and uses them for new constructions (Piaget,
2001). For this reason, reflective abstraction is an essential focus in mathematics
education (Simon et al., 2004). Many have turned to reflective abstraction as “a
guiding heuristic in a search for insight into mathematical learning” (Steffe, 1991,
p. 43) and, indeed, some even consider it to be the goal of instruction (Simon, 2016).
We have established that forms of abstraction that are not empirical fall under the
umbrella of reflective abstraction, but within this umbrella, there are three subtypes.
Glasersfeld (1995) characterized the subtypes in this manner: “The first of the three
reflective abstractions projects and reorganizes, on another conceptual level, a coor-
dination or pattern of the subject’s own activities or operations” (p. 105). This first
subtype is called “reflecting abstraction.” Glasersfeld went on to explain that:
The next is similar in that it also involves patterns of activities or operations, but it includes
the subject’s awareness of what has been abstracted and is therefore called ‘reflected
abstraction’. The last is called ‘pseudo-empirical’ because, like empirical abstractions, it
can take place only if suitable sensorimotor material is available. (p. 105)
Thus, pseudo-empirical abstraction, reflecting abstraction, and reflected abstraction

are all forms of reflective abstraction, and we discuss each in turn below.
Pseudo-empirical Abstraction
Although it was not until the last period of Piaget’s work that he specified and
enriched the different types of reflective abstraction, as early as 1948, he began to
make implicit distinctions between the types of abstraction. For instance, Piaget
(1948/1967) noted that “the ‘abstraction of shapes’ is not carried out solely on the
basis of objects perceived as such, but is based to a far greater extent on the actions
which enable objects to be built up in terms of their spatial structure” (p. 68). In this
case, geometrical shape is not a property of an object like its weight or color, but
rather a phenomenon that results from the learner’s physical or mental actions in
perceiving an object coordinated from their inception of the shape grasped as a
single whole. Piaget’s remark pointed to a key difference between empirical and
pseudo-empirical abstractions, even though he did not develop the latter term until
much later: pseudo-empirical abstraction draws coordination not from the objects
alone, but rather from the learner’s physical or mental actions exerted on objects
(Montangero & Maurice-Naville, 2013). The objects, such as the shapes Piaget
referred to, are indispensable, but the coordination is from the learner’s actions,
such as the imposition of spatial structuring.
In particular, Piaget clarified this distinction in his 1977 book, Recherches sur
L’Abstraction Réfléchissante:
When the object has been modified by the subject’s actions and enriched by the properties
drawn from their coordinations (e.g., when arranging the elements of a collection in a
sequence) the abstraction bearing upon these properties is called ‘pseudo-empirical’
because, while it concerns the object and its actual observable traits as in empirical abstrac-
tion, the facts it reveals concern, in reality, the products of the coordination of the subject’s
actions: this is, then, a particular case of reflective abstraction and not at all a derivative of
empirical abstraction. (p. 303)
Thus, according to Piaget, abstraction that requires perceptual material but draws on
one’s actions imposed on that material is pseudo-empirical (Glasersfeld, 1995).
Piaget (1980) clarified that pseudo-empirical abstraction entails “a reading of the
objects involved, but reading which is really concerned with properties due to the
180 A. Ellis et al.
action of the subject himself” (p. 92, emphasis ours). He referred to this as an initial
form of reflective abstraction and emphasized that it plays a fundamental psychoge-
netic role in all logico-mathematical learning, in that humans need to manipulate
physical objects in order to build up structures that are otherwise too abstract.
As an example, Montangero and Maurice-Naville (2013) shared a case of a child
who is just learning to count. The child took ten stones, counted them, and placed
them in a row. He then decided to check whether there would still be ten stones
regardless of the order in which he counted or arranged them. In rearranging the
stones in different orders and configurations, he found that the total was always 10,
regardless of whether he placed them in a circle or a row or whether he counted
from left to right or from right to left. The child’s abstraction of the conservation of
number was drawn not from a property of the stones, per se, but instead from the
organization the child introduced. But, in what is critical to making this an example
of pseudo-empirical abstraction, the stones themselves were a necessary support to
the child’s activity. Number and order are abstractions not based solely on percep-
tion, and thus they are pseudo-empirical rather than empirical (Campbell, 2001). In
another example, Thompson (1985) distinguished the abstraction of weight from
the abstraction of conservation of weight. As described above, through empirical
abstraction, one can determine that different objects have different weights by hold-
ing them and can conclude that larger objects typically weigh more than smaller
objects. Weight, being a property of an object, is an empirical abstraction. In con-
trast, it is through pseudo-empirical abstraction that one determines that the weight
of an object must remain the same under transformations of elongation or deforma-
tion, as long as nothing is added or taken away. One’s abstraction of this conserva-
tion depends on their actions on objects rather than on properties of the objects
themselves; it is pseudo-empirical, however, in that it still requires the objects in
order to make the abstraction.
Number and conservation of weight are fairly straightforward examples address-
ing young children’s early abstractions, but is pseudo-empirical abstraction a useful
construct for more advanced mathematical learning? Montangero and Maurice-
Naville (2013) seemed skeptical of this notion, pointing out that pseudo-empirical
abstractions are numerous early in life but then, as people age, gradually decrease
as reflecting and reflected abstractions increase. Some researchers do not bother to
distinguish pseudo-empirical abstraction as a subtype of reflective abstraction (e.g.,
Simon et al., 2004). Others, however, emphasize the subtlety in determining what
constitutes perceptual material or “an observable.” For instance, Campbell (2001)
pointed out that even if one accepts the idea that number is a pseudo-empirical
abstraction, it could then itself become an observable: “Even if number did have to
be inferentially constructed by the child at level N, why should this prevent the child
at level N + 1 from gathering empirical data about different numbers of objects
under different circumstances, and engaging accordingly in empirical abstraction?”
(pp. 9–10). Thus, Campbell argued that what is a coordination at one level can then
function as an observable at the next higher level. Hence, what constitutes a pseudo-
empirical abstraction depends on what we take to be an observable versus a con-
struction at the current level.
Campbell’s point introduces some flexibility in what we mean by “observables

that are external, on the one hand, but that are constructed by reflecting abstraction,
on the other” (Piaget, 2001, p. 31). Some researchers, such as Moore (2014) and
Simon (2006), even go so far as to admit a wholly internal phenomenon as an
observable, as long as it is the product of one’s mental actions, rather than the coor-
dination of those actions. From this perspective, an observable could be, for instance,
the outcome of one’s listing activity when solving a combinatorics problem or a
representation of a recursive pattern that is now taken as source material for a new
abstraction. We follow Moore’s (2014) stance by broadening the category of what
can be admitted as a pseudo-empirical abstraction. Namely, we consider it useful to
broaden what constitutes an “observable” to include the product of mental activity
as source material for a new abstraction. This extension is useful for making sense
of students’ learning in advanced mathematics but also for learning in the ear-
lier grades.
From this perspective, we can consider Angelo’s abstraction, which he expressed
as 1.5(x), to be a pseudo-empirical abstraction for two reasons. Firstly, Angelo
relied on the observable that was the two columns of numbers representing the
ordered pairs he had generated (Fig. 6.2). It was in looking across these pairs of
numbers and noticing a consistent pattern that he was able to extract the relationship
that each area value was 1.5 times its corresponding length value. Secondly, and this
speaks to our deliberate extension of what we consider to be an observable, we
would admit Angelo’s abstraction as pseudo-empirical because the source material
he relied on was the outcome of his action of creating equivalent ratios, as instanti-
ated by the observed pattern in the set of ordered pairs. We would still consider this
to be a pseudo-empirical abstraction even if Angelo had not relied on the available
perceptual material of the two columns of numbers, because the source material for
his abstraction was not a coordination of his activity, but rather the outcome of his
activity. In the extended examples below, we continue to rely on this broader stance
of what constitutes pseudo-empirical abstraction to examine students’ sensemaking
with graphs and with combinatorial tables.
Reflecting Abstraction
Reflecting abstraction is a form of reflective abstraction that, like pseudo-empirical

abstraction, is drawn from the coordinations of actions or operations. The key dis-
tinction between pseudo-empirical abstraction and reflecting abstraction, according
to Piaget, is that reflecting abstraction does not require perceptual material. Like
pseudo-empirical abstraction, the source of reflecting abstraction is endogenous, but
it does not require the presence of an object or other sensorimotor materials: “We
are now dealing with the subject himself and no longer with external objects”
(Piaget, 1980, p. 90). As Campbell (2001) noted:
182 A. Ellis et al.
The abstraction of a mental characteristic that qualifies some action scheme and is destined
to bring this characteristic into a more complex scheme (not just into a simple descriptive
concept of internal experience) is reflecting abstraction. Calling it reflecting indicates that
abstraction transforms the very conduct by differentiating it and consequently adds some-
thing to the quality that has been isolated by abstraction. (pp. 10–11)
Such abstraction from mental activity is necessarily constructive. A prior cognitive

structure is what is projected to the higher developmental level, which is then reor-
ganized into a new structure (Campbell, 2001; Tallman, 2021).
Given our broader interpretation of what can be counted as a pseudo-empirical
abstraction, namely, including as an observable the product of mental activity, this
necessarily restricts somewhat the activity that we include as reflecting abstraction.
For instance, Angelo’s abstraction, expressed as 1.5(x), relied on the product of his
mental activity in creating equivalent ratios, rather than on a coordination of actions.
In contrast, we characterize Willow’s abstraction as a reflecting abstraction. Recall
that Willow realized that the rectangle grew 1.5 cm2 in area for each 1-cm increase
in length, which she expressed as A = 1.5 x L. The source material for Willow’s
abstraction was not a pattern of observable outcomes across her set of equivalent
ratios, as it was for Angelo; rather, it was her activity of coordinating the number of
groups of area with the corresponding number of groups of length. Willow did not
require perceptual material to make her abstraction, and she drew, instead, on a
coordination of actions, rather than on an outcome of actions. As such, Willow pro-
gressed through what Tallman (2021) described as three phases: (1) the differentia-
tion of a sequence of actions from their outcome, (2) the projection of the actions
from the level of activity to the level of representation (the reflected level), and (3)
the reorganization of the projected actions. As Tallman (2021) described:
A subject must differentiate (dissociate) actions from their effects before they can construct
an internalized representation of them, what Piaget called projecting actions to the level of
representation (i.e., the level of cognition). Additionally, the subject must coordinate the
actions that produced the effect before they can project and represent them on a higher
cognitive level. Once a subject differentiates actions from their effect and then coordinates
them, they are prepared to project these coordinated actions to the reflected level, where
they are organized into cognitive structures. (p. 3)
We posit that Willow, in making a group of five 1.5 s, reflected on the coordination
between the number of 1.5 s and the number of 1 s: Each group was 5 because 1 was
¼ of 4 and 1.5 was ¼ of 6, and Willow needed to determine the amount of area cor-
responding to 5 cm in length. Willow engaged in the sequence of actions of combin-
ing five groups of 1.5 cm2 to get 7.5 cm2 for the area. In doing so, she differentiated
the sequence of actions, which was coordinating the number of 1.5 s with the num-
ber of 1 s, from the outcome, which was 7.5. She then projected the actions from the
level of activity, adding up five 1.5 s, to the level of internalized representation. This
enabled Willow to then reorganize her projected actions. Specifically, she realized
that she could now directly couple growth in area with growth in length for a 1-cm
length increment: “Each time the growth in length goes up by one, the growth in
area, I think the growth in area equals [writes ‘A = 1.5 x L’].”
Reflected Abstraction
Reflected abstraction is the third type of reflective abstraction, and it differs from
other types of abstraction in that it entails conscious awareness: “We call ‘reflected’
abstraction the result of a reflective abstraction when it has become conscious”
(Piaget, 1977, p. 303). Reflected abstractions are conscious products of other reflec-
tive abstractions, resulting in a reflection of the thought on itself (Piaget, 1976). In
contrasting reflected abstraction from pseudo-empirical and reflecting abstraction,
Piaget (1976) explained:
By contrast, the reflexive abstraction can become conscious, particularly when the subject
compares two steps that he has carried out and tries to discern common factors…In this
second case, we shall refer to ‘reflected abstraction’, the past participle denoting the result
of the ‘reflexive’ process. (p. 346)
Reflected abstraction thus entails the learner consciously examining their prior
experiences in order to reorganize them along with their current activity (Glasersfeld,
1987). It can enable a formulation or even a formalization of the elements that have
been abstracted (Montangero & Maurice-Naville, 2013). Furthermore, reflected
abstraction “can completely free itself from any relationship with material objects”
(Piaget, 1980, p. 92) and indeed can drive much of the activity that gives rise to
higher mathematics. When engaging in pseudo-empirical or reflecting abstraction,
one uses some structure in their operational compositions, and that use is implicit.
As Glasersfeld (1995) explained, these types of reflective abstraction may or may
not involve the learner’s awareness, and in fact, throughout history, mathematicians
have used thought structures without conscious awareness: “A classic example:
Aristotle used the logic of relations, yet ignored it entirely in the construction of his
own logic” (Piaget & Garcia, 1983, p. 37, as cited in Glasersfeld, 1995, p. 106). A
learner can be aware of what they are cognitively operating on, without also being
aware of the operations being carried out. When engaging in reflected abstraction,
however, the learner gains conscious awareness of their operations.
This is not to say that reflected abstraction only occurs in the context of, say,
formal mathematics. It can be observed even at the level of a child’s verbal expres-
sion of an action, such as “I press on the button, and the bell rings” (Montangero &
Maurice-Naville, 2013, p. 59). One must, however, consciously operate on one’s
actions at the level of representation. At the stage of reflected abstraction, the struc-
ture of operational compositions is “teased out” (Piaget, 1980, p. 99) and can sup-
port the establishment of a formal theory. Operations that were initially instruments
of calculation become, at the stage of reflected abstraction, differentiated objects of
thought in their own right, a process Piaget referred to as thematization. Tallman
(2021) stated that this suggests that the learner has symbolized coordinated actions
at a higher cognitive level:
Reflected abstraction thus relies on what Piaget called the semiotic function, or the subject’s
capacity to construct mental symbols to represent aspects of their experience. Reflected
abstraction entails symbolizing coordinated actions at the level of representation so as to
reify the material actions the symbol represents into a form that can be used as an object of
184 A. Ellis et al.
thought at the level of representation. On this higher cognitive level, the subject can con-
sciously manipulate these symbols independently of re-presenting the coordinated actions
they signify, all the while being capable of doing so. The semiotic function is thus the
essential mechanism by which reflecting abstraction becomes reflected abstraction. (p. 4)
Tallman explained that engaging in conceptual operations on the symbols that rep-
resent coordinated actions at the level of representation ultimately supports “increas-
ingly organized and differentiated cognitive structures” (p. 4), i.e., learning. For an
elaborated account of reflected abstraction, see Tallman and O’Bryan (this issue).
Other researchers may not have the same bar for symbolizing coordinated actions
(e.g., Montangero & Maurice-Naville, 2013). Piaget himself referred to explicit
comparison of one’s actions or operations across cases as evidence for reflected
abstraction (e.g., 1976, 2001); we will similarly consider those cases in the follow-
ing data episodes.
Data Episodes
In order to exemplify the different types and stages of abstraction as well as discuss
the standards of evidence we hold for identifying instances of abstraction, we pres-
ent extended data episodes from each of two tasks. The first is a covariation task
called the Faucet Task, and the second is a series of combinatorial tasks called the
Passwords Activity. For each task, we present two sets of data, and in doing so, we
identify specific instances of abstraction and discuss evidence for categorizing them
according to type.
The Faucet Task
We draw from data from two different teaching experiments in which students
addressed the Faucet Task (Paoletti, 2019; see Fig. 6.4a for a screenshot of the
Faucet Task activity and https://bit.ly/36jy0Dn for the task itself). In the Faucet
Task, students are initially asked to coordinate how turning different faucet knobs
influences two quantities: the amount of water leaving the faucet and the water’s
temperature. Eventually, students are tasked with coordinating how the relationship
between these two quantities could be represented graphically. We use the first
example, with a 4-year-old student, Mario, to exemplify empirical and reflective
abstraction. We then use a second example with two middle-school students, Kendis
and Camila, to articulate more detailed differences between pseudo-empirical,
reflecting, and reflected abstractions.
Fig. 6.4 (a) A screenshot of the Faucet Task and (b) Mario’s hand feeling the water below an
actual faucet
Mario Engages in Empirical and Reflective Abstractions
Prior to engaging with the Faucet Task, the teacher-researcher (TR) first provided
Mario with opportunities to consider several scenarios with a real faucet (see
Fig. 6.4b). Specifically, the TR prompted Mario to predict how the temperature and
amount of water would change for four scenarios, each starting with each knob
halfway on. When he asked Mario to predict how the water temperature and amount
of water would change when he turned the cold knob on, Mario predicted that there
would be “less water and colder.” Upon turning the cold knob on and putting his
hand under the water, Mario noted that although the water was colder, there was
more water leaving the faucet, instead of less water as he predicted. A similar inter-
action occurred when the TR asked Mario how the two quantities would change
when turning the cold knob off. Mario predicted there would be “more water and
hotter.” However, when putting his hand under the water after turning the cold knob
off, Mario quickly observed there would be “less [water]…but it’s definitely hotter.”
In this initial interaction, Mario was developing a conception of how quantities
change together in the faucet scenario via an empirical abstraction; the water com-
ing from the faucet provided sensorimotor data that Mario used to determine
whether there was more or less water leaving the faucet, as well as whether the
temperature increased or decreased. By physically engaging with the knobs, and by
observing and feeling these changes, Mario was empirically abstracting the rela-
tionships between turning knobs, the changing amount of water, and the water’s
changing temperature.
Although Mario’s initial conception of the relationships between quantities in
the faucet scenario was grounded in an empirical abstraction, in his next activity,
there is evidence that he began to project his meanings to a higher conceptual level.
After addressing all four scenarios with the actual faucet (Fig. 6.4b), the TR pre-
sented the faucet applet to Mario, which allowed him to turn both knobs on and off
for a digital faucet (shown in Fig. 6.4a). In the applet, the changing width of the
rectangle below the spigot is intended to represent the amount of water leaving the
faucet, and the changing water color is intended to represent the varying
186 A. Ellis et al.
temperature. After Mario familiarized himself with the applet, the TR asked him to
predict how the water temperature and the amount of water would change in each of
the four scenarios, with each knob starting halfway on. In each case, Mario was able
to accurately predict, without testing, how the water temperature and the amount of
water would change for each scenario. We take this as evidence that Mario projected
the outcome of his activity with the actual faucet to the level of mental representa-
tion (rather than physical experience). Then, at this level, he engaged in a reorgani-
zation by coordinating his current activity with the results of his prior activity in
such a way that he was able to anticipate how turning each knob would simultane-
ously affect the water temperature and the amount of water. Mario could therefore
predict how turning any knob, in any of the scenarios, would change both the
amount of water leaving the faucet and the relative amount of hot and cold water,
thereby affecting temperature. Hence, we infer that he engaged in reflective abstrac-
tion as he differentiated his prior actions of turning knobs and observing water
amounts and water temperature from the outcomes of those actions, projected that
differentiation to the level of representation, and reorganized his conception of the
faucet situation at this level.
Of note, the extent to which Mario engaged in a pseudo-empirical abstraction
versus a reflecting abstraction is an open question. As the Faucet applet was percep-
tually available, Mario may have been leveraging this representation to support his
reasoning that leveraged his prior experiences with the physical faucet. If this per-
ceptual material was necessary for Mario, we would contend that his meanings were
grounded in a pseudo-empirical abstraction. However, if Mario could predict how
each quantity would change without needing to rely on any available perceptual
material, we would infer that his meanings were grounded in a reflecting abstraction.
endis and Camila Engage in Pseudo-empirical, Reflecting,

K
and Reflected Abstractions
We present a second example with the Faucet Task to illustrate nuances in different
forms of reflective abstraction, drawing on two eighth grade students’ activity,
Kendis and Camila (see Paoletti et al., 2021, for more detail). In this version of the
task, one goal included supporting students in reasoning about graphs as represent-
ing relationships between two covarying quantities. As such, students interacted
with a GeoGebra applet showing a faucet (Fig. 6.5, https://www.geogebra.org/m/
rdxkrwek), along with a coordinate system that included dynamic vertical and hori-
zontal segments representing two quantities, water temperature and amount of water
leaving the faucet. Additionally, there was a point corresponding to the endpoints of
the two segments (Fig. 6.5). The task is designed to support students conceiving of
a point as a multiplicative object (Saldanha & Thompson, 1998), simultaneously
representing the magnitudes of the two quantities, and eventually conceiving of
graphs as being produced via an emergent trace of this point (i.e., reasoning emer-
gently; see Moore, 2021).
Fig. 6.5 A screenshot from the GeoGebra faucet applets that included a graph representing water
temperature and amount of water leaving the faucet on the vertical and horizontal axes, respectively
Prior to the interactions below, Kendis and Camila had worked collaboratively to
describe how each segment on the axes would vary based on how the quantities
varied (e.g., when the cold knob is turned off, the vertical red segment would get
longer because the temperature increases, and the horizontal pink segment would
get shorter because there is less water leaving the faucet). Further, Kendis had
described that the point “stays in line with both of them,” with “both of them”
referring to the endpoints of the segments. When addressing how the point would
move when turning the hot knob on, Kendis was aware that the point’s final posi-
tion would be up and to the right of its initial starting location. However, she
initially imagined the point’s movement as dictated by the two segments chang-
ing sequentially rather than simultaneously. Specifically, when describing how
the point would move in this case, Kendis argued:
Kendis: It’s going this way (tracing to the right along the horizontal axis from the endpoint
of the pink segment, (1) in Fig. 6.6) and, look, it’s going to stay in a line with it (pointing to
the top of the red segment), so it’s just going to move over and up (traces sequentially to the
right (2), then up (3) in Fig. 6.6).
In (1), moving from left to right, Kendis described how the segment representing the
amount of water increased. She then attended to sequential changes in both quanti-
ties as she indicated that the horizontal, left-to-right, motion near the point in the
plane was followed by a vertical upward motion created by an increase in tempera-
ture. Paoletti et al. (2021) contended that such sequential reasoning is consistent
with the developmental nature of covariational reasoning in which a student thinks
of one quantity, then the next, then back to the first, and so on (Saldanha &
Thompson, 1998).
After making this prediction, Kendis observed the point’s diagonal up-and-to-the-
right motion in the applet when turning the hot water on. Such sensorimotor
188 A. Ellis et al.
Fig. 6.6 Kendis’s hand

motions describing how
the point would move
when the hot knob is
turned on
experience provided Kendis an opportunity to experience a pseudo-empirical

abstraction. Specifically, she differentiated the results of the point moving (i.e.,
its final position) from the sequence of actions that cause the point to move; she
explicitly coordinated how the simultaneously changing quantities would be rep-
resented via simultaneously changing segment lengths. Kendis then argued that
the point would move diagonally because of the covariation of the two quantities.
For example, discussing the point’s motion as the hot knob is turned off, Kendis
described the point’s movement as representing two simultaneously changing
situational quantities:
Kendis: So this (pointing to the top of the red segment along the vertical axis) is going to go
down (drags finger down). It’s going to go down (repeats downward motion along axis), and
then (points to the horizontal axis) it’s less water also. So it’s going to go diagonal (making
a diagonal cutting motion with her hand).
Kendis then immediately engaged in a series of movements without talking. First,

she motioned horizontally left from the point to indicate a decreasing amount of
water (represented by (1) in Fig. 6.7), and then she motioned down from the end-
point of her first motion to indicate a decreasing water temperature (represented by
(2) in Fig. 6.7). Critically—and differing from her earlier activity—Kendis lastly
motioned diagonally down and to the left (represented by (3) in Fig. 6.7).
Based on this activity, we infer that Kendis’s prior experience observing the
motion of the point supported her in engaging in a pseudo-empirical abstraction.
Specifically, Kendis projected the sensorimotor experience of the point’s diagonal
motion to a higher level, which supported her in reorganizing her conception of the
point’s horizontal and vertical movement from sequential to simultaneous. She now
conceived the point as a multiplicative object that simultaneously represented the
two covarying quantities’ magnitudes. At this point, due to Kendis’s motions focus-
ing first on the horizontal and vertical motions on the graph on the computer screen
prior to the diagonal motion, we infer that the perceptual material was important for
Fig. 6.7 Kendis’s hand motions when describing how the point would move when the hot knob is
turned off
Kendis’s conception of the point’s movement, which is why we conclude that she
engaged in a pseudo-empirical abstraction rather than a reflecting abstraction.
By introducing multiple scenarios (e.g., turning the hot water on, turning the hot
water off, etc.), the TR intended to provide the students with opportunities to con-
ceive of numerous graphs as being produced via the trace of a point constrained by
covarying quantities’ magnitudes. He considered that such opportunities may sup-
port them in reflecting across their reasoning in these different cases, thereby poten-
tially fostering the development of reflecting abstractions. These abstractions, in
turn, could then support the goal of the students becoming explicitly aware of their
conceiving graphs as the product of a trace of a point that represented the magni-
tudes of two covarying quantities. We would take such an explicit awareness as an
indication of a student having engaged in a reflected abstraction.
To explore the extent to which the students may have engaged in reflecting and
reflected abstractions in the Faucet Task, the TR provided them with five completed
graphs, with no applet available. He invited the students to describe situations that
would produce each graph, with several graphs involving more than one turn, which
was novel relative to the previous situations. We contend that a student describing a
series of turns that would produce a given graph provides evidence they have
stripped away the particular faucet situations that were central to their prior reason-
ing because they would now be able to both anticipate traces that would produce a
given graph and describe a situation that would produce such a trace. Providing
evidence of such reasoning, when describing the graph in Fig. 6.8a, Camila imag-
ined the graph tracing from the top-left point (traversing the arc labeled “a” and then
the arc labeled “b” in Fig. 6.8a). Camila then explained that the situation started
with cold half on and hot all the way on. She explained that turning the cold knob
the rest of the way on would produce the first arc, as “it’s going down in temperature
and to the right, so it means you’re increasing water and it’s going down, so it means
you have to be adding cold water.” She then described that turning the hot knob off
would produce the arc labeled “b” in Fig. 6.8a. Hence, Camila provided evidence
190 A. Ellis et al.
Fig. 6.8 Two graphs Camila interpreted
that her repeated prior experiences with constructing graphs that represented differ-
ent faucet situations supported her in reorganizing her meanings for graphs at a
higher level. In particular, as she stripped away the specific faucet situations, she
projected her reasoning about graphs as traces to a higher level, and she then
engaged in a reorganization at this level to anticipate that any graph could be pro-
duced via the trace of a dynamic point representing two covarying quantities.
Further, Camila later provided evidence that she was consciously aware of her
meaning that graphs are produced via emergent traces. In this instance, Camilla
discussed the graph in Fig. 6.8b. This graph can be interpreted as representing a
scenario in which both knobs start in the off position, then the cold knob is turned
on (temperature is constant while the amount of water increases), and then the hot
knob is turned on (temperature and amount of water increase). However, the graph
could also be interpreted another way. Namely, it could represent a scenario in
which both knobs start in the on position, then the hot knob is turned all the way off
(temperature and amount of water decrease), and then the cold knob is turned all the
way off (amount of water decreases, while temperature remains constant).
Camila was able to conceive of the same final graph as being producible via two
different action scenarios. When asked to describe a situation that could produce the
graph, Camila provided one explanation, but then the TR spontaneously asked her
if she could interpret the graph a second way:
Camilla: (With both knobs turned off) First step is to turn the cold on (motions hand from
left to right as if tracing the straight part of the graph in Fig. 6.6d) then turn the hot one
on (motions an arc with hand from left to right).
TR: So they’re both starting completely off, turning cold on then turning hot on… So in
terms of the two quantities, how did you know that was (trails off).
Camilla: Well, [the graph] continued to go to the right (motions with hand from left to right)
so it means [amount of water]’s increasing in quantity (repeats motion) and then after
the second transition [the graph]’s going up in temperature (motions as arc with hand
from left to right) which means you’re going to be adding hot water, so the first one we
started off as cold adding it (traces a horizontal line motion from left to right) and then
we had to add more of hotter temperature (see Fig. 6.9 for a recreation of Camila’s
imagined actions).
TR: Cool, yeah, so this one going straight tells you what was happening with the
temperature?
Camilla: It means it’s not changing.
TR: Not changing, right? So it’s got to just be cold water coming on, but then we get here
and the temperature increases, the amount of water increases, which [inaudible]. Could
there be another way this, this plays out?
Camilla: Hot water off.
TR: Hot water, so you start with both of them on, turn hot water off, get to here [crosstalk].
Camilla: [crosstalk] And then the cold is at halfway and then you could also turn it off
(traces fingers in air straight across from right to left; see Fig. 6.10 for a recreation of
Camila’s imagined actions).
In this interaction, Camila described two different action scenarios that resulted
in the same outcome (i.e., the final graph). Such activity is consistent with Piaget’s
(1976) description of reflected abstraction as comparing two processes one has car-
ried out while discerning common factors across those processes. We take this to be
evidence that Camila engaged in a reflected abstraction with regard to her meanings
for graphs as being producible via traces. That is, Camila’s activity entails a(n at
least implicit) comparison of two different sets of actions she carried out as result-
ing in the same common final graph (i.e., the actions depicted in Figs. 6.9 and 6.10).
We take this implied comparison as evidence of her having engaged in a reflected
Fig. 6.9 The first scenario Camila imagined producing the graph in Fig. 6.8
192 A. Ellis et al.
Fig. 6.10 The second scenario Camila imagined producing the graph in Fig. 6.8
abstraction such that she was explicitly aware of her understanding that a graph
could be produced via the trace of a point representing covarying quantities.
The Passwords Activity
We now present two examples of data from the domain of combinatorics, each of
which relate to the same set of tasks we call the Passwords Activity. We first use an
example from a single interview with an undergraduate vector calculus student,
Tyler, to demonstrate pseudo-empirical and reflecting abstraction. Then, we use a
second example with three undergraduate vector calculus students, Carson, Aaron,
and Anne-Marie, to demonstrate reflecting and reflected abstraction. Broadly, the
set of tasks involves counting passwords of varying length and with certain con-
straints; the overall goal of the tasks is to motivate deep understanding of the bino-
mial theorem (for more information on this activity, see Lockwood & Reed, 2016;
Ellis et al., 2022a).
In the Passwords Activity, students consider the question, “How many three-
character passwords can be made using the letters A and B?”, and we explicitly
Table 6.1 (a, b) The three-character and four-character AB tables
direct students to organize their work by completing tables according to the number
of As in the password (Table 6.1a). We aim to have students fill out each row by
listing passwords. Once students complete the three-character passwords problem,
they then repeat this process for passwords of lengths 4 and 5 (the four-character
generating table is seen in Table 6.1b).
The goal is to have the students build (via partial or complete listing) the tables
to see how they would use them when progressing to the next part of the tasks.
Then, students move on to passwords involving the number 1 and the letters A and
B (which we call AB1 passwords). Students are asked to make tables for three-
character and four-character AB1 passwords, organized according to the number of
1 s in the passwords (the four-character AB1 table is in Table 6.21). The table for the
number of four-character passwords that use 1, A, and B can be completed by first
thinking of counting the number of ways of placing the 1 s and then considering the
number of options for the remaining non-1 positions (the positions that are not 1 s
must be As or Bs, which reduces the problem to a previous one involving AB
passwords).
Tyler Engages in Pseudo-empirical and Reflecting Abstraction
Tyler was a first-year college student enrolled in vector calculus, and he participated
in a single, 60-minute individual interview (we have shared excerpts from Tyler’s
work elsewhere, such as in Ellis et al., 2022b). Tyler’s method of solving the tasks
typically involved organized and systematic listing. That is, Tyler’s listing activity
became a mechanism by which he could progress through and solve the problems;
ultimately, both the outcomes of that listing activity and the listing activity itself
would become sources of abstraction for Tyler.
1
Note that the right column of Table 6.1b represents a way of writing the expressions that high-
lights the relationship with the AB tables; in designing the tasks, we hoped students would eventu-
ally recognize this structure, although they could also simply write the totals in the right column.
194 A. Ellis et al.
Table 6.2 The three-character and four-character AB1 tables
Fig. 6.11 (a) Tyler’s list of the eight three-character AB passwords and (b) his three-character
AB table
Tyler’s work on the AB password tasks. We began by asking Tyler for the total
number of three-character AB passwords. In response, Tyler created a list of out-
comes (Fig. 6.11a). Tyler had a strategic, organized way of listing outcomes, and
he used his list to create the correct table for three-character AB passwords
(Fig. 6.11b).2
2
We do not return to Tyler’s specific systematic listing strategy in this paper, but for the curious
reader, he created the list in Fig. 6.11a by first creating the column of AAA, AAB, and ABB and
then creating the column of BBB, BBA, and BAA. He noted that these had consecutive groups of
letters, and then he added the ABA and BAB as passwords that would “mix them up.” When asked
about his strategy, Tyler noted that he listed all the passwords that started with A and then all the
passwords that started with B. The takeaway is that he was systematic in his listing strategy.
We then asked Tyler to create the four-character AB table, and he again gener-
ated the table (Fig. 6.12) by systematically listing and then counting the outcomes
for each row. We note the 1, 4, 6, 4, 1 in the right column; as we will describe, he
would later refer to the 6 in this table when creating a four-character AB1 table.
An important episode occurred when Tyler tried to fill in the entry of the three-
character AB1 table with zero 1 s:
Int.: How about for zero?
Tyler: Zero, um, it’s not gonna be one this time.
Int.: Okay.
Tyler: (Writes AAA, then pauses) A, A, A, um, eight maybe?
Int.: Okay, and why, why’d you guess that?
Tyler: Um, because for this one with just; what I’m doing now I guess is just two letters, um,
so it’s the exact same one I did here, isn’t it (points to the length 3, AB table in
Fig. 6.11b). Yeah, so I’m going to go with eight just because that’s the exact same thing.
We contend that this is an instance of pseudo-empirical abstraction for two rea-
sons. Firstly, we posit that the initial act of writing AAA served as a reminder of
Tyler’s prior listing activity, and he directed his attention toward the written table on
the page, which provided available sensorimotor material. Tyler first listed out
AAA, and then he paused before he guessed that there were eight AB1 passwords
with zero 1s. This suggests that the physical act of writing down the AAA password
triggered for Tyler a reminder of his previous listing. In this way, Tyler engaged in
a reflective abstraction that was tied to sensorimotor material, making it an instance
of pseudo-empirical abstraction. Furthermore, we also draw on our broadened cat-
egory of what might be considered an “observable” (based on Moore, 2014), where
here the 8 served as the outcome of Tyler’s prior mental activity of listing. We see
this as pseudo-empirical abstraction because Tyler drew on the results of his
Fig. 6.12 Tyler’s

four-character AB table
196 A. Ellis et al.
activity, rather than on the coordination of his actions in producing the original
table. Thus, the outcome of his listing activity served as the observable on which
Tyler relied to abstract the table structure. Specifically, Tyler projected the result of
his initial listing activity (the eight total outcomes he produced in the three-character
AB table), which he then fit into the AB1 situation, reorganizing his current activity
with the AB1 table to incorporate the outcome of his prior activity. Tyler could then
conceive of that 8 not just as the result of his work on the AB table but as the total
number of possibilities in his AB1 table in the case where there was no 1.
In this episode, then, Tyler recognized that when he was counting the AB1 pass-
words with zero 1s, he was in a situation that was identical to the previous situation
involving only As and Bs. Tyler connected the current situation with the previous
situation, and he used his activity and his previous table to make sense of (and find
a solution for) the new situation. In particular, he seemed to understand that in his
prior listing activity, he had generated a total of eight three-character AB passwords,
and so finding himself in a similar situation, he could use the result of that prior
activity.
Next, we asked Tyler to fill out a four-character AB1 table, organized according
to the number of 1s. In working on the row for one 1, Tyler did something unex-
pected – he introduced a way of describing a general outcome involving 1s and xs.
Specifically, he wrote out four general outcomes, 1xxx, x1xx, xx1x, and xxx1
(Fig. 6.13a), and he used those general outcomes to fill out the new table. Tyler
discussed his reasoning in the following exchange, and as he did so, he referred
back to his four-character AB table (Fig. 6.13b).
Tyler: And then the 1, so what I was thinking – what I was saying earlier, how there is only
a certain amount of spots for it. Like it has to be … like I’m just going to use x cause,
um, has to be in one of these spots ... (writes the combinations of 1 and xs in Fig. 6.13).
Int.: Great.
Tyler: So there’s, now there’s just three xs, um, and I know that for ... three spots with two
different letters there’s going to be eight different ways to do it (points back to the
previous 3-character AB table, see Fig. 6.13b)…Um, so I guess eight … there’s eight
different of each of those just using this same table (points to the three-character AB
table). Um, there’s just 32 so I want to say there’s going to be um, 32 for just the one.
Int.: Okay and you got, you’re thinking of that as kind of the 4 times 8?
Tyler: Yeah I, just adding them all up.
Fig. 6.13 (a) Tyler’s list of one 1 and three xs, and (b) Tyler explicitly refers back to the AB table
This was a key moment in Tyler’s work. He articulated a generalized outcome,

expressing four-character AB1 passwords as arrangements of xs and 1s. Tyler con-
tinued to use 1s and xs in filling out the rest of the four-character AB1 table.
Figure 6.14a shows his list of xs and 1s in the four-character AB1 case, with exactly
two 1s. There are exactly six of them, and the following exchange demonstrates
Tyler’s meaning of those six general outcomes as they relate back to his previous
work. Specifically, Tyler seemed to understand why six such outcomes would make
sense, because he could recognize that he was in a situation of arranging two dis-
tinct objects, which is what his previous work involving AB passwords also entailed.
He ultimately arrived at the correct table for four-character AB1 passwords
(Fig. 6.14b).
Tyler: Yeah there you go. Is that all of them? Yeah so six, because that would make sense…
Int.: Does that six make sense?
Tyler: Does it? Uh, well that would – that’s um, two variables like instead of doing three
things there’s two, um, with the four combo, so two, was six over here (points back to
the 6 in the correct entry of the four character AB password table, Fig. 6.12), so that’s
why I thought it made sense.
This was a crucial revelation in terms of Tyler’s work and his ultimate success on
the activity. We note that Tyler could see the arrangements of 1s and xs and the
arrangements of As and Bs as being essentially “the same.” One example of this
phenomenon was when Tyler was able to justify why there were six arrangements
of two 1s and two xs. As seen previously, Tyler said that the 6 made sense by noting,
“Like instead of doing three things there’s two, um, with the four combo, so two was
six over here [points back to the 6 in the correct entry of the four-character AB
table], so that’s why I thought it made sense.” Tyler’s response suggests that he
Fig. 6.14 (a) Tyler’s six combinations of two 1s and two xs and (b) his four-character AB1 table
198 A. Ellis et al.
recognized that he was in a situation of arranging two 1s and two xs, and he explic-
itly related the result of that activity to the number of ways of arranging two As and
two Bs, which he had solved previously. Tyler extrapolated a similarity between
arranging AABB and 11xx, as he recognized that, ultimately, he was just rearrang-
ing two kinds of things. In doing so, Tyler stripped away the specifics (e.g., it did not
matter whether they were As and Bs or 1s and xs) as he focused on the coordination
of his actions to make a generalization across the two scenarios.
In this case, we infer that Tyler engaged in a reflecting abstraction, as he began
to consider a general way to represent ways to count the number of passwords with
one 1. Here, the six ways to list two As and two Bs were not just the result that he
viewed as separate from (the outcome of) a particular process; rather, he reflected
on the process of listing those outcomes themselves. He reflected on the repeated
action of arranging two types of two characters, and the particulars of what those
characters are were stripped away. We infer that Tyler projected his specific listing
activity of arranging two As and two Bs in the AB passwords case to the higher level
of the AB1 case (i.e., a more general level of thinking of listing structurally). He
then engaged in a reorganization, which involved making sense of his current activ-
ity in the AB1 case, listing two 1s and two xs; he thus coordinated his current spe-
cific activity with a generalized structure of arranging two of each of two types of
characters.
Thus, to summarize the episodes we have described with Tyler, we contend that
he engaged in both pseudo-empirical and reflecting abstractions, and the nature of
his initial activity facilitated each abstraction. In the first case, he abstracted 8, the
result of a listing process for three-character AB passwords, and that 8 was tied to
the sensory material of the table he had written and the physical listing process he
had engaged in. In the second case, he looked at the result of that process (the 6), but
he did not abstract that result but rather the process itself. We follow Moore (2014)
in considering the difference between abstracting the result of a process and abstract-
ing a coordinated process itself as being a key distinguishing feature between
pseudo-empirical and reflecting abstractions.
Group of Students Engage in Pseudo-empirical, Reflecting,

A
and Reflected Abstraction
We now present an example in which a small group of four first-year college math-
ematics students (Carson, Anne-Marie, Aaron, and Josh) were working through the
Passwords Activity. We highlight a slightly different aspect of their work, focusing
particularly on their reasoning about symmetry within the tables, but in doing so, we
demonstrate instances of pseudo-empirical, reflecting, and reflected abstraction. We
have reported aspects of these data with a different focus elsewhere, specifically
examining a construct we call empirical re-conceptualization (Ellis et al., 2022a), as
well as exploring students’ notions of equivalence (Lockwood & Reed, 2020) and
transfer (Lockwood & Reed, 2021).
In prior work creating tables for five-character AB passwords, the students had
determined that there were 25 = 32 total five-character AB passwords, which they
could justify using the multiplication principle. The tables for these passwords
yielded a column of 1, 5, 10, 10, 5, 1, as seen in Aaron and Carson’s respective work
(Fig. 6.15). In the following excerpt, we see that both Aaron and Carson used
numerical patterns to complete the table. They reasoned combinatorially for the 1,
5, 5, and 1 entries (note, e.g., Aaron wrote there are five passwords with one A
because “A goes into one of five slots”). They knew the total had to be 32, and they
hypothesized that, because of prior symmetry they had observed, the two middle
numbers would be the same, and the total would need to add to 32.
As Aaron explained, “I just knew that it had to add up to 32, so, you know, 20 added
up to what was already there.” Carson agreed with him and said the following:
Carson: I did the exact same thing. I started with the symmetry again where you know you
have one option for each of the monogamous sets, and then five options for sets where one
of the things is different than all of the others, and then two empty slots. And because it’s
symmetric, then we know that those two slots need to be the same number, and knew that
the total had to go up to 32 just based off of, you know, if you [have] five slots and two
options for each slot, you’re going to get 25. [...] So, there was a remainder of 20, so half of
20 is 10, so 10 for each of those slots.
We claim that Aaron and Carson’s statement that rows for 2 and 3 would be 10 was
a case of pseudo-empirical abstraction. They abstracted the pattern of symmetry
they had observed by looking across previous tables. We argue that their abstraction
was pseudo-empirical because, like Angelo, they reflected on the outcomes of their
initial activity, as represented as a pattern in the prior table, rather than on a coordi-
nation of actions that they engaged in while listing or creating the table. The out-
comes were available to them as sensorimotor material in the prior table, but even if
the prior table had been absent, we would still consider this a pseudo-empirical
abstraction due to the students’ reliance on the results of their activity.
The interviewer asked the students to explain how they became aware of the
symmetry and why it would make sense. Anne-Marie and Carson’s response to this
question demonstrates having made a reflecting abstraction:
Fig. 6.15 (a) Aaron’s and (b) Carson’s five-character tables

200 A. Ellis et al.
Anne-Marie: But, I just see it as, like, it’s where the As and Bs flip. So, like, when I was
writing out the list, I was moving the A through the matrix of Bs. And, like, so that’s
when…the symmetry…I think it was kind of when you were moving the B instead
of the A.
Int.: Okay, nice.
Carson: Yeah, so just, it’s…asking to move one A around in a group of Bs is the same thing
asking to move one B around a group of As. It’s going to return the same number of
results.
Int.: Okay.
Carson: So, it’s like that flip she was talking about, you know, where you were going down
this list of how many As you have. Eventually, it’d be easier to ask how many Bs you
have, because it’ll be the same list, just backwards.
In this excerpt, Anne-Marie described her listing process as writing outcomes
with fewer As and “moving the A” through the Bs and then changing to “moving the
B instead of the A.” Carson expanded on Anne-Marie’s statement, and he described
moving one (general) character through a string of another (general) character.
Carson’s language suggests that he was engaging in reflecting abstraction as he
considered a general listing activity based on the coordination of his actions. In
particular, the source material of the abstraction was the students’ reflection on their
general process of listing rather than the outcome of the listing itself. Here, we
interpret that Carson extracted the regularity he observed in his listing activity –
specifically, the idea of moving one character through a string of other characters –
and he projected that listing activity into a new situation in which he was working
with different characters (now a B instead of an A) and different strings (moving
through a string of As instead of a string of Bs).
In a final example, we describe an episode in which Carson engaged in a reflected
abstraction. Here, the interviewer again asked the students about the symmetry:
“Someone else just tell me a little bit more about why that’s the case. [...] Two As
and three Bs, three As and two Bs, like, yeah, why is it the same number of things?”
Carson responded by bringing up a connection to a problem that the students had
solved in an earlier session, an “arrangement with restricted repetition” problem,
which involves arranging characters, some of which are identical (or repeated). For
example, such problems would be to arrange letters in the word MAMA or
RACCOON. Such problems can be solved by arranging all letters and then dividing
by the number of arrangements of repeated letters. In the following excerpt, Carson
describes a new approach to thinking about the AB passwords problems, which was
different from the work he had done so far on the Passwords Activity. Carson looked
at the third row of the AB table and considered the original solution as a problem
involving arranging three As and two Bs:
Carson: Another way you could look at this [the fourth row in the 5-character AB table,
Fig. 6.15b] is if you have the A, A, A, B, B, how many different ways can you arrange
the letters in that word? So that’s going to 5!, which is the total number of letters over
the repeat letters, so 3!, 2! (writes “AAABB” and “5!/(3!2!)”). Right?
Int.: Mm hmm.
Carson: And that’s the same equation there. That will hold true whether the repea letters are
three A’s and two B’s, or three B’s and two A, or three C’s and two D’s. Right?
Int.: Okay.
Carson: And I mean, that’s a mathematical relationship we’ve talked about before, because
you’re just pulling out the redundant, um, redundant arrangements given repeated
letters.
Int.: Okay, good.
Carson: So, I mean, yeah, it doesn’t really matter what the letters are as long as they’re as
those ratios.
This exchange shows evidence that Carson engaged in a reflected abstraction.
Carson recognized that the eq. 5!/(3!2!) was the same as the current situation he was
in, and he also made a more general statement that indicated that he realized a
broader phenomenon. Specifically, Carson understood that although in the
Passwords Activity, they were counting AB passwords with a certain number of As,
this activity was actually representative of a type of activity that could be done with
any two kinds of characters. Carson abstracted a process/operation (arranging two
kinds of characters) across multiple situations – one involving counting the number
of ways to arrange letters in a given word (which he had solved previously) and one
involving determining the number of AB passwords with three As and two Bs.
Furthermore, Carson was consciously aware of this similar structure and could
describe it to the interviewer. Carson’s articulation of an isomorphism across the
two situations (the fact that both cases entail the same process of arranging charac-
ters with repetition) suggests that he had engaged in a reflected abstraction.
Discussion
Our aim in this chapter has been to describe and provide illustrative examples of the
levels of abstraction as described by Piaget. Below, we discuss the standards of
evidence we find useful for guiding our thinking when examining students’ abstrac-
tions. We then describe the cyclical and interrelated nature of different types of
abstraction and conclude by offering some final thoughts on the ways in which
distinguishing between types of abstraction can be useful to researchers.
Standards of Evidence
Having provided a number of data examples demonstrating various types of abstrac-

tion, we now reflect back across those examples to summarize what we take as
standards of evidence for deciding how to classify an abstraction by type. To clas-
sify an abstraction as empirical, it must be of properties of objects that are, to the
actor, inherent in the objects themselves. Generalizing about properties such as
color, weight, size, or, in the case of Mario, the amount of water leaving a faucet
would all be potential examples of empirical abstractions. Once there is evidence
202 A. Ellis et al.
that the actor is relying on coordinations of their physical or mental actions exerted
on the objects, we no longer consider the abstraction to be empirical. It is now a
pseudo-empirical abstraction, even if one is, for instance, making generalizations
about (or classifying objects based on) properties that appear to be inherent to the
object, from the actor’s perspective. In some cases, such as with Tyler’s listing activ-
ity, we adopt Moore’s (2014) stance and consider abstractions to be pseudo-
empirical when they stem from the product of activity, rather than from coordinations
of actions, regardless of whether or not one relies on sensorimotor material. By
expanding what counts as pseudo-empirical abstractions in this manner, we open up
possibilities for conducting more nuanced analyses of students’ thinking that is
common in school mathematics, such as the type of pattern finding we saw with
Angelo, as well as in higher-level mathematics.
When determining whether a student has made a reflecting abstraction, we look
for evidence of them coordinating actions or operations, rather than relying solely
on the outcomes of their actions or operations. In research contexts, this typically
occurs when students engage in multiple tasks over time, such as in teaching experi-
ment settings, because it offers opportunities for us to observe students’ scheme or
operation construction. Specifically, we can look for evidence that students have
engaged in similar coordinations across tasks in such a manner that they have ulti-
mately stripped away specific details from any individual experience and instead
have created a more general structure, such as in Tyler’s activity when he used 1 s
and xs to represent any two-digit code. As we noted above, some researchers only
require that the scheme be one that is composed of operations more advanced than
the prior operations being used (e.g., Campbell, 2001), which is our standard of
evidence as well. For instance, we saw evidence of more advanced operations when
Willow shifted from grouping sets of 1.5 cm2 to coordinating the number of 1.5cm2s
with the number of 1 cms. It is worth noting that others, such as Norton (2018), hold
a higher standard by requiring the presence of operations arising through a coordi-
nation of actions within a reversible and composable system.
When considering whether an abstraction is reflected, we adopt the criterion
Piaget (2001) often used, namely, that one has compared commonalities across dif-
ferent activities and discerned common factors. Such comparisons, as Piaget (1976)
noted, require the actor to explicitly reflect on their coordinations, a reflection of
thought on itself; in this manner, we can infer conscious awareness. We saw evi-
dence of this when Carson saw the Passwords Activity as involving the same opera-
tions as a prior class of arrangement problems, and he was able to articulate that
both were simply the process of arranging two kinds of characters, represented as
5!/(3!2!). We note that this criterion may also be less strict than what other research-
ers require (e.g., Tallman, 2021). For a more elaborated account of reflected abstrac-
tion specifically, see Tallman and O’Bryan (Chap. 8). In general, evidence for
reflecting and reflected abstractions can be challenging to observe and may require
gathering data over time in order to document changes in students’ activity.
The Cyclical Nature of Abstraction
In this chapter, we have introduced a somewhat artificial separation of each type of

abstraction in order to discuss each in turn; in actuality, however, the processes of
abstraction are not so neatly separated. For instance, we have discussed that empiri-
cal abstractions presuppose prior reflective abstractions (Montangero & Maurice-
Naville, 2013). As an example, for Mario to recognize changes in water temperature
via an empirical abstraction, he must already have constructed schemes via reflec-
tive abstraction by which to describe temperature. The necessity of reflective
abstraction for empirical abstractions both underscores Piaget’s unease with the
term “simple” while also highlighting the cyclical nature of the different forms of
abstraction in general.
There are also cyclical relationships between the different types of reflective
abstraction; sometimes, these relationships are opaque, and it can be difficult to
determine precisely what kind of abstraction has occurred. Both Kendis and
Camila’s repeated experiences with the Faucet Task and Tyler’s list-making activity
highlight that pseudo-empirical abstractions can lay an important foundation for
future reflecting (and reflected) abstractions. By repeatedly engaging in pseudo-
empirical abstractions, the students were able to develop operations that ultimately
supported further abstractions based on coordinations of actions. With the Passwords
Activity, for instance, the designed set of tasks intentionally grew in complexity and
generality. This supported students to initially engage in systematic listing of out-
comes and filling out tables, offering opportunities to reflect on the results of their
listing activity via pseudo-empirical abstraction. But then, as the complexity of each
task grew, students were able to reflect more generally on their operations of arrang-
ing a certain number of types of characters, no longer focusing on the specifics of
what the characters were. This afforded reflecting and reflected abstractions that
supported students’ understanding of combinatorial listing strategies, as well as
combinatorial justifications for the number of passwords with certain constraints.
It is important to include two caveats here. Firstly, it is not our intention to imply
that every reflecting abstraction presupposes a pseudo-empirical abstraction. One
can certainly engage in reflecting abstraction without relying on a prior pseudo-
empirical abstraction. The relationship between the different types of abstraction
can be complex, and it is not necessarily the case that individuals shift in some
ladder-like fashion from empirical to pseudo-empirical to reflecting to reflected
abstractions. Our second caveat is that a pseudo-empirical abstraction does not nec-
essarily need to lead to subsequent reflecting or reflected abstractions. As we saw
with Angelo’s determination of a rule based on a pattern of outcomes, students may
complete individual tasks – or entire series of tasks – without ever making reflecting
(or reflected) abstractions.
Repeatedly engaging in different situations that elicit schemes grounded in
reflective abstractions can, in turn, support students in comparing their activity
across situations to ultimately form reflected abstractions. Camila’s description of
two situations that could produce the same graph via different emergent traces is
204 A. Ellis et al.
one example of how a student can build on prior pseudo-empirical and reflecting
abstractions in order to produce a reflected abstraction. Operations that were ini-
tially instruments of calculation, such as creating a trace of a point as a representa-
tion of covarying quantities, then became differentiated objects of thought in their
own right, enabling Camila to identify different scenarios that would result in the
same outcome. As such, reflected abstractions are powerful in that they can support
students in formalizing their own mathematics. Reflected abstractions (in contrast
to other types of abstraction) do necessarily entail prior reflective abstractions, even
though it may be challenging or even impossible for a researcher to tease apart those
myriad prior abstractions. This is the case because reflected abstractions are defined
as conscious products of reflective abstractions (Piaget, 1976). Repeated activity
that elicits meanings grounded in reflecting abstractions, in particular, can be par-
ticularly powerful for supporting students to ultimately develop reflected abstrac-
tions (Oehrtman, 2008).
The Value of Abstraction as a Construct
How can identifying and characterizing instances of abstraction be useful to our

goals as researchers? Adopting the lens of abstraction, and in particular, character-
izing the processes of projection and reorganization, offers a powerful mechanism
to explain students’ learning over time. Certainly, abstraction and other aspects of
Piaget’s genetic epistemology are not the only possible ways to do this, but we find
that it provides a meaningful structure to organize our thinking, as researchers, as
we build second-order models (Steffe et al., 1983) of students’ mathematics. These
models, in turn, enable us to make theory-driven predictions about students’ under-
standing and behavior. For instance, consider Willow and Angelo’s activity with the
Growing Rectangle Problem. Both students produced the same general statement,
A = 1.5 L. Examining the nature of each student’s abstraction, however, provided
important insights into their understanding, as well as the likely generativity of their
generalization. Generative generalizations are ones that can be extended to accom-
modate new cases and can be justified through deductive arguments (Ellis et al.,
2022b). By identifying Willow’s abstraction as reflecting, we could predict that she
would likely be able to justify her general statement, as well as use it to produce new
length-area pairs, even if we provided non-integer length values. Correspondingly,
we could predict that Angelo might need additional support in justifying his general
statement beyond an empirical argument, using it to predict new pairs, or adjusting
it to respond to a different-sized rectangle. Across our data sets, we have found that
generalizations based on reflecting abstractions are more generative than the ones
that emerge from pseudo-empirical abstractions. This, in turn, can inform our design
process by constructing sequences of tasks that can engender reflecting abstractions.
This was the case in our work with Kendis and Camila. Conceiving a graph as a
trace is a non-trivial concept that is seldom intentionally developed in US school
mathematics (Thompson & Carlson, 2017). However, such an understanding is pos-
sible, even for middle-school students, if they are intentionally supported in
gradually engaging in higher levels of abstraction (Paoletti et al., 2021). For instance,
we anticipated that a series of pseudo-empirical abstractions could support students
in subsequently developing the necessary reflecting abstractions to ultimately con-
ceive and anticipate graphs as representing dynamic traces. In the case of Kendis
and Camila, Kendis’s conceiving of a point as a multiplicative object (Lee, 2016;
Saldanha & Thompson, 1998) via a pseudo-empirical abstraction laid the founda-
tion for her eventually engaging in a reflecting abstraction to anticipate the point as
a representation of covarying quantities. Then, in reflecting on their operations, both
students ultimately became consciously aware of their understanding that graphs
are traces representing two covarying quantities, in particular contexts. This process
of reflected abstraction then enabled the students to more generally conceive of
graphs as representations of covariation, even in novel contexts. For example,
Paoletti et al. (2021) described Kendis and Camila graphically representing a rela-
tionship between the side length and the area of a triangle in the sessions immedi-
ately following the Faucet Task. Hence, their reflected abstraction in one context
supported their graphing activity in a novel context.
Similarly, building on our findings from the Growing Rectangle Problem, we
developed a sequence of tasks that would foster shifts from pseudo-empirical to
reflecting abstraction. We did so by encouraging students to reason with multiple
growing rectangles with different height values and ultimately with growing rect-
angles in which they had to determine their own height values. Students who ini-
tially reasoned as Angelo did, generalizing based on the outcome of observed
patterns, began to reflect on their activity of coordinating growth in length with
growth in area for different rectangles. Doing so with new rectangles with unspeci-
fied height values further encouraged this form of reflection, ultimately supporting
the development of a constant rate of change (Ellis et al., 2020). As discussed ear-
lier, we also designed the Passwords Activity to engender opportunities to engage in
repeated abstractions within and across tasks, supporting students in making
pseudo-empirical abstractions that could ultimately be leveraged to foster reflecting
and reflected abstractions.
Using the construct of abstraction provides a level of precision in characterizing
students’ learning and cognition, and it allows us to go beyond merely describing
behavior. It offers a way to make sense of what is at the core of observable distinc-
tions (such as the distinction between result-pattern generalization and process-
pattern generalization (Harel, 2001)) and explain the cognitive mechanisms
responsible for those distinctions.
Piaget’s notion of abstraction is just one construct of many to explain learning.
Models of learning can offer insight into how to design better task sequences, as
well as inform our thinking about powerful instructional practices that can foster
meaningful mathematical engagement. Characterizing the nature and content of stu-
dents’ abstractions when reasoning about particular mathematical ideas has proved
helpful, not only in making sense of students’ thinking in the moment but also in
then being able to predict future reasoning and responsively craft subsequent activi-
ties to more effectively support conceptual development. By understanding stu-
dents’ processes of abstraction, we are able to better understand the nature of
teaching and learning mathematics.
206 A. Ellis et al.
References
Bäck, A. (2006). The concept of abstraction. The Society for Ancient Greek Philosophy
Newsletter, 376.
Campbell, R. (2001). Reflecting abstraction in context. In J. Piaget (Ed.), Studies in reflecting
abstraction (pp. 1–27). Psychology Press.
Dubinsky, E. (1991). Constructive aspects of reflective abstraction in advanced mathematics. In
Dubinsky, E. (2002). Reflective abstraction in advanced mathematical thinking. In Advanced
mathematical thinking (pp. 95–126). Springer.
Ellis, A. B., Ely, R., Tasova, H., & Singleton, B. (2020). Scaling continuous variation: Supporting
students’ algebraic reasoning. Educational Studies in Mathematics, 104(1), 87–103.
Ellis, A., Lockwood, E., & Ozaltun-Celik, A. (2022a). Empirical re-conceptualization: From
empirical generalizations to insight and understanding. The Journal of Mathematical Behavior,
65, 100928.
Ellis, A. B., Lockwood, E., Tillema, E., & Moore, K. (2022b). Generalization across multiple
mathematical domains: Relating, forming, and extending. Cognition and Instruction, 40(3),
351–384.
Gallagher, J., & Reid, D. (1981). The learning theory of Piaget and Inhelder. Wadsworth, Inc.
Harel, G. (2001). The development of mathematical induction as a proof scheme: A model for
DNR-based instruction. In S. Campbell & R. Zaskis (Eds.), Learning and teaching number
theory (pp. 185–212). Ablex.
Kant, I. (1992). Lectures on logic. Cambridge University Press. (Original work published in 1800).
Kant, I. (2003). Critique of pure reason (M. Weigelt, Trans.). Penguin Classics. (Original work
published in 1781).
Laurence, S., & Margolis, E. (2012). Abstraction and the origin of general ideas. Philosopher’s
Imprint, 12, 1–22.
Lee, H. Y. (2016). Just go straight: Reasoning within spatial frames of reference. In M. B. Wood,
E. E. Turner, M. Civil, & J. A. Eli (Eds.), Proceedings of the 38th annual conference of the
North American Chapter of the International Group for the Psychology of Mathematics
Education (pp. 278–281). ASU.
Locke, J. (1975). An essay concerning human understanding. Clarendon Press. (Original work
published in 1690).
Lockwood, E., & Reed, Z. (2016). Students’ meanings of a (potentially) powerful tool for general-
izing in combinatorics. In T. Fukawa-Connelly, K. Keene, & M. Zandieh (Eds.), Proceedings
for the nineteenth special interest group of the MAA on research on undergraduate mathemat-
ics education. West Virginia University.
Lockwood, E., & Reed, Z. (2020). Defining and demonstrating an equivalence way of thinking in
enumerative combinatorics. Journal of Mathematical Behavior, 58. https://doi.org/10.1016/j.
jmathb.2020.100780
Lockwood, E., & Reed, Z. (2021). Using an actor-oriented perspective to explore one under-
graduate student’s repeated reference to a particular problem in a combinatorial context. In
C. Hohensee & J. Lobato (Eds.), Transfer of learning: Progressive perspectives for mathemat-
ics education and related fields (pp. 173–202). Springer.
Moessinger, P., & Poulin-Dubois, D. (1981). Piaget on abstraction. Human Development, 24(5),
347–353.
Montangero, J., & Maurice-Naville, D. (2013). Piaget or the advance of knowledge: An overview
and glossary. Laurence Erlbaum Associates.
Moore, K. C. (2014). Signals, symbols, and representational activity. In L. Hatfield, K. Moore, &
L. Steffe (Eds.), Epistemic algebraic students: Emerging models of students’ algebraic know-
ing (Vol. 4, pp. 211–235). University of Wyoming.
Moore, K. C. (2021). Graphical shape thinking and transfer. In C. Hohensee & J. Lobato (Eds.),
Transfer of learning: Progressive perspectives for mathematics education and related fields
(pp. 145–172). Springer. https://doi.org/10.1007/978-3-030-65632-4_7
Norton, A. (2018). Frameworks for modeling student’ mathematics. The Journal of Mathematical
Behavior, 52, 201–207.
Oehrtman, M. (2008). Layers of abstraction: Theory and design for the instruction of limit con-
cepts. In M. P. Carlson & C. L. Rasmussen (Eds.), Making the connection: Research and
teaching in undergraduate mathematics education (pp. 65–80). Mathematical Association of
America.
Paoletti, T. (2019). Supporting students’ understanding graphs as emergent traces: The fau-
cet task. In M. Graven, H. Vankat, A. A. Essien, & P. Vale (Eds.), Proceedings of the 43rd
conference of the International Group for the Psychology of Mathematics Education (Vol. 3,
pp. 185–200). PME.
Paoletti, T., Corven, J., & Gantt, A. L. (2021). Supporting middle-school students’ developing emer-
gent graphical shape thinking. In D. Olanoff, K. Johnson, & S. M. Spitzer (Eds.), Proceedings
of the forty-third annual meeting of the North American Chapter of the International Group for
the Psychology of Mathematics Education (pp. 499–508). PME-NA.
Piaget, J. (1928). Judgment and reasoning in the child. Harcourt, Brace, & Co.
Piaget, J. (1950). Introduction à l’épistémologie génétique (Vol. 1, II, & III). PUF.
Piaget, J. (1966). Mathematical epistemology and psychology. Gordon & Breach. (Original work
published in 1961).
Piaget, J. (1967). The child’s conception of space. W.W. Norton & Company, Inc. (Original work
published in 1948).
Piaget, J. (1976). The grasp of consciousness: Action and concept in the young child (translated by
S. Wedgewood). Harvard University Press.
Piaget, J. (1977). Recherches sur l'abstraction réfléchissante (Vol. I & II). Presses univ. de France.
Piaget, J. (1980). Adaptation and intelligence: Organic selection and phenocopy. University of
Chicago Press. (Original work published in 1974).
Piaget, J. (2001). Studies in reflecting abstraction. Psychology Press.
Piaget, J., & Garcia, R. (1983). Psychogénèse et histoire des sciences. Flammarion.
Saldanha, L., & Thompson, P. W. (1998). Re-thinking co-variation from a quantitative perspective:
Simultaneous continuous variation. In S. B. Berensah & W. N. Coulombe (Eds.), Proceedings
of the annual meeting of the psychology of mathematics education – North America (Vol. 1,
pp. 298–303). North Carolina State University.
Simon, M. A. (2006). Key developmental understandings in mathematics: A direction for inves-
tigating and establishing learning goals. Mathematical Thinking and Learning, 8, 359–371.
Simon, M. (2016). An approach to the design of mathematical task sequences: Conceptual learning
as abstraction. PNA, 10(4), 270–279.
Simon, M., Tzur, R., Heinz, K., & Kinzel, M. (2004). Explicating a mechanism for concep-
tual learning: Elaborating the construct of reflective abstraction. Journal for Research in
Mathematics Education, 35, 305–329.
Singleton, B., & Ellis, A. B. (2020). Why multiply? Connecting area measurement to multiplica-
tive reasoning. Mathematics Teacher: Learning and Teaching PreK-12, 113(10), e37–e42.
Steffe, L. P. (1991). Operations that generate quantity. Learning and Individual Differences,
3(1), 61–82.
Tallman, M. A. (2021). Investigating the transformation of a secondary teacher’s knowledge of
trigonometric functions. The Journal of Mathematical Behavior, 62, 100869.
208 A. Ellis et al.
von Glasersfeld, E. (1982a). An interpretation of Piaget’s constructivism. Revue Internationale de
Philosophie, 36(4), 612–635.
von Glasersfeld, E. (1982b). Subitizing: The role of figural patterns in the development of numeri-
cal concepts. Archives de Psychologie, 50, 191–218.
von Glasersfeld, E. (1987). Learning as constructive activity. In The construction in knowledge
contributions to conceptual semantics (pp. 307–333). Intersystems Publications.
rience and of Piaget’s approach. In L. P. Steffe (Ed.), Epistemological foundations of math-
ematical experience. Springer.
von Glasersfeld, E. (1995). Radical constructivism: A way of knowing and learning.
Routledge Falmer.
von Humboldt, W. (1975). Werke (Vol.7, part 2). Leitmann. (Original published in 1907).
Chapter 7
Groups and Group-Like Structures
Anderson Norton
In the introduction to this book, we noted that Piaget’s genetic epistemology adopts
a Kantian perspective in that it explains how cognitive (rational) structures shape
empirical experiences to form knowledge of the world. Although Kant’s (1998)
rational empiricism greatly influenced Piaget, the two thinkers diverged when it
came to assumptions about innate structures. Specifically, Kant had taken time,
space, and number for granted, and in this, he was not alone:
Our knowledge of the first principles, such as space, time, motion, and number, is as certain
as any knowledge we obtain by reasoning. As a matter of fact, this knowledge is provided
by our hearts and instinct is the necessary basis on which our reasoning has to build its
conclusions. (Blaise Pascal, 1966/1670 (in Pensées, line 110)).
In contrast, Piaget took the construction of time, space, number, and even logic,
as research foci. He learned from children how those structures emerge through
human activity, beginning with reflexes and sensorimotor activity. With a few close
colleagues—especially Barbel Inhelder—he reported these results in separate books
on each topic (Inhelder & Piaget, 1969; Piaget, 1969/1946; Piaget & Inhelder,
1967/1948; Piaget & Szeminska, 1952).
It was important to Piaget’s genetic epistemology that he study children because,
as evident in Pascal’s quote, we, as adults, take for granted many of these primitive
structures. These early constructions are so fundamental to our organizations of
experience that we assume their certainty and necessity. They are the logico-
mathematical structures from which we build models of the worlds we experience:
Mathematics may be defined as the study of shape and number, or as the science of patterns
(Resnik, 1981; Steen, 1988). Piaget dug deeper into questions of shape, number, and pat-
terns in general, by considering their psychological underpinnings. He demonstrated the
A. Norton (*)
https://doi.org/10.1007/978-3-031-47386-9_7
210 A. Norton
unity of mathematics by identifying commonalities in their structures. Specifically, he

defined mathematics as the coordination of reversible and composable mental actions. That
is, he defined it as an organization of operations: ‘An operation may be defined as an action
which can return to its starting point, and which can be integrated with other actions also
possessing this feature of reversibility’. (Piaget & Inhelder, 1967/1948, p. 36)
In building models of children’s mathematics, Piaget described the organization of

operations within structures. In this chapter, we make a distinction between two
kinds of structures used repeatedly in Piaget’s epistemology of mathematics.
Chapter 3 addressed the first kind of structure, that of a scheme, which Piaget used
to describe how children sequence operations in service of a goal. These structures
have proved fruitful within research in mathematics education, across mathematical
domains from counting to calculus. The second kind of structure—groups and
group-like structures—has received less attention in mathematics education
research. As defined in detail later in the chapter, a group is a mathematical structure
that describes how elements of a set can be combined. Piaget used groups to describe
how a set of mental actions might be combined and organized as reversible and
composable (qua) operations.
When Piaget referred to a group, he referred to the formal mathematical structure
studied in abstract algebra. However, as he never tired of reminding his readers, he
did not assert that students are aware of such a structure: “Let us note that at these
[unformalized] levels, even if the operations used are more and more conscious, the
structured wholes remain completely alien to the subject’s conscious reflection”
(Beth & Piaget, 1966, p. 246). After all, both schemes and groups are researcher
constructs used to build models of students’ mathematics (Glasersfeld & Steffe,
1991). We infer them from students’ behaviors to explain how they might reason.
We rely on them because they prove useful in explaining and predicting how stu-
dents might operate mathematically.
For example, Piaget and Inhelder (1967/1948) relied on a “group of displace-
ments” to describe how children construct space on the basis of their own activity.
Through self-locomotion (e.g., crawling), children learn to alter the worlds they
perceive by moving within them. They can continue their motion in a given direc-
tion or move in a new direction, composing those two displacements to reach some
new vantage point. They can also reverse their motion to return to a starting point.
These two properties of motion—composability and reversibility—satisfy the pri-
mary conditions of a mathematical group. Piaget argued that this group structure
explains how children build the totality of space, in which to move and to posit
objects even when removed from perceptual experience.
The purpose of this chapter is to examine how Piaget used groups and group-like
structures to model mathematical development as a coordination of mental actions.
By extension, we consider ways that mathematics education researchers can use
such structures today, not only to build models of children’s mathematics in various
domains but also to describe the nature of mathematics itself. The chapter begins by
distinguishing the general structure of a group from that of a scheme and by eluci-
dating their relationship. It proceeds to introduce formal properties of groups and
group-like structures Piaget (1972a) called “groupings.” We elaborate on examples
7 Groups and Group-Like Structures 211
of “the splitting loope” (Wilkins & Norton, 2011) and “the splitting group” (Norton
& Wilkins, 2012) to demonstrate how group-like structures have, in fact, proved
useful in mathematics education. After noting formal connections to mathematical
programs carried out by Felix Klein and the Bourbaki, the chapter closes with spe-
cific suggestions for applying groups and group-like structures to future research in
Two Kinds of Structure
Glasersfeld (1995) described Piagetian schemes as three-part structures composed

of a recognition template, a sequence of actions, and a goal. For example, consider
the iterative fraction scheme (IFS), which renders fractions “numbers in their own
right” (Hackenberg, 2007). When a student operating with an IFS sees a fraction—
even an improper fraction, like “7/5”—that fraction symbolizes a sequence of
actions they could perform (Tzur, 1999). Specifically, they can assimilate 7/5 as
seven iterations of the unit fraction, 1/5, which results from partitioning a whole
into five equal parts. Moreover, they can reverse that sequence of actions to repro-
duce the whole from 7/5 of it. Figure 7.1 illustrates a simplified schematic model of
the IFS.
Glasersfeld (1995) was well aware that the three-part model oversimplifies mat-
ters (as all models do); it fits better with sensorimotor schemes than operative
schemes (see Chap. 3). In a sensorimotor scheme, actions are carried out in sequence,
physically or in imagination. In an operative scheme, the three-part structure, along
with all of its actions, collapses into a single logico-mathematical concept:
A schema of action is, in fact, only the form of a series of actions that take place succes-
sively without a simultaneous perception of the whole. Reflective abstraction, on the other
hand, upgrades it to the form of an operational schema, that is, of a structure such that, when
one of the operations is used, its combination with others becomes deductively possible
through a reflection going beyond the momentary action… and these operations can sooner
or later be carried out symbolically without any further attention being paid to the objects
which were in any case ‘any whatever’ from the start. (Beth & Piaget, 1966, p. 234)
Reflective abstraction is the subject of Chap. 6, so we won’t elaborate on its role

here except to emphasize that it is the process by which we “upgrade” actions to
reversible and composable operations, so it is inherently tied to the construction of
group structures. Consider the remainder of Beth and Piaget’s comments with
regard to the IFS, wherein improper fractions, like 7/5, become mathematical
Fig. 7.1 Three-part model of a scheme

212 A. Norton
objects (numbers in their own right). The sequence of actions used to produce 7/5
(partitioning a whole into five equal parts and iterating one of them seven times) no
longer needs to be carried out, as 7/5 symbolizes them all at once. This scheme is
called operational because its actions have become interiorized as operations.
Actions become interiorized as operations when we organize them within groups
or group-like structures. Therein, operations achieve reversibility and composabil-
ity. Composition of actions within a group does not depend on any particular
sequence of actions or any particular scheme. All possible compositions and all
inverse operations are determined by the structure of the group. Note here that an
action could serve as its own inverse (e.g., reflecting the plane over a line), but gen-
erally, actions have other actions as their inverses (e.g., partitioning and iterating or
rotating the plane about a point in either of two directions). Moreover, in mathemat-
ics, we are generally interested in composing actions with many other actions to
produce large sets of mathematical objects (e.g., producing all isometries of the
plane by composing reflections over various lines). So, to operationalize an action,
we generally need a group that contains other actions (including the trivial action of
doing nothing).
Throughout this chapter, “group” refers to a formal mathematical structure, often
as a structure for modeling students’ mathematical ways of operating. The structure
of groups pertains to the organization of logico-mathematical operations them-
selves, an organization that takes time and requires experience and reflection. Within
groups, operations achieve simultaneity and logical necessity. Simultaneity refers to
the sense that the operations do not need to be carried out in sequence. Their orga-
nization depends on the structuring of the group as a whole. As we will see in the
next section, logical necessity depends upon the reversibility and composability of
the operations.
Schemes achieve their own reversibility thanks to the reversibility and compos-
ability of their operations. For example, the IFS is a reversible scheme because its
principal operations are reversible and composable. Specifically, partitioning and
iterating become reversible operations within a group-like structure called the split-
ting loope (Wilkins & Norton, 2011). This structure prescribes ways partitions and
iterations might be composed with themselves and one another. We return to this
example later in the chapter after carefully defining groups and their properties.
Groups
Groups serve as fundamental structures both in mathematics and in Piaget’s episte-

mology. They frame Piaget’s characterizations of logico-mathematical operations in
particular. Indeed, the appearance of groups in psychology as well as mathematics
is foundational to his epistemology of mathematics. Piaget explains the coincidence
as a consequence of the nature of mathematics itself. We have, throughout the his-
tory of mathematics, been operating within laws of reversibility and composability,
and we “discovered” groups when we became explicitly aware of these ways of

operating:
…for it is only at the end of a sufficiently long series of reflective abstractions that the sub-
ject discovers the most profound characteristic of operations: that of being connected
together in structures which have their own laws of totality. That is why we had to wait for
E. Galois to discover the group concept which Viète or Descartes, nevertheless, constantly
used unconsciously in their algebra. (Beth & Piaget, 1966, p. 294)
A formal definition of group did not appear until the 1850s,1 but the group structure
was implicit in the work of mathematicians for many centuries and not just in alge-
bra. As so happens, we can characterize the field of abstract algebra as the study of
our own mathematical ways of operating across various domains, from number
theory to geometry and complex analysis. For example, when mathematicians of the
Italian Renaissance encountered imaginary roots, they implicitly understood that
accepting them as solutions of equations meant including them within a closed sys-
tem for adding and multiplying them with other numbers. Bombelli’s use of a + bi,
in particular, described an extension from real numbers to complex numbers that,
today (thanks to abstract algebra), we recognize as a field extension.
Among the structures of abstract algebra (e.g., rings, fields, and ideals), the struc-
ture of groups is one of the most commonly recognized. The near ubiquity of groups
in mathematics owes to the simplicity of their general structure. A set of mathemati-
cal actions or objects only need to satisfy a short list of conditions to constitute
a group:
1. Closure under composition: Every element of the set can be composed with any
other element of the set to produce some element of the set; this composition
could take the form of addition, multiplication, function composition, or any
other binary operation, so long as it results in another element of the set.
2. An identity element: There is some element of the set that, when composed with
any other element of the set, yields that other element of the set; in the real num-
bers under addition, this element is 0, and under multiplication, 1.
3. Reversibility: Every element of the set has an inverse element in the set, such that
their composition yields the identity element (e.g., 3 × 1/3 = 1).
4. Associativity: Given the composition of any three elements, the order in which
the (binary) composition is performed does not matter (i.e., (a × b) × c = a
× (b × c)); this property should not be confused with commutativity, which con-
cerns the order of the elements themselves (rather than the order of composition)
and is not a necessary property of groups.
We elaborate on each of these conditions/properties, both mathematically and psy-
chologically, in separate subsections below. Familiar examples of organized actions
that we can model with a group structure include the rotations of a wheel and
1
Historians credit Arthur Cayley for providing the first definition of an abstract group, in 1954. See
Kleiner (1986) for a thorough and engaging review of its history, including contributions from
Viète, Descartes, and (especially) Galois.
214 A. Norton
manipulations of the Rubik’s cube. Formal mathematical examples of groups

include isometries of the plane, permutations of an ordered set, positive rational
numbers under multiplication, and n-dimensional vectors under vector addition. As
noted by Beth and Piaget (1966) in the quote above, the group structure formalized
in the nineteenth century had implicitly permeated mathematical thinking for centu-
ries before, from number theory, to geometry, to algebra. The apparent ubiquity of
groups in mathematics underscores the simplicity of the structure and the unity of
mathematics as a product of human thought.
Closure
Closure refers to composability. Not only does it imply that any two elements in the
group can be composed but also that the result of their composition is an element of
the group. For example, consider the group of integers under addition. Whenever we
add two integers, their sum is another integer. In fact, closure under addition is what
defines the integers as a set of numbers. “Need we remind the reader that a whole
number exists, psychologically as well as logically (in spite of Russell2), only by
virtue of being an element of a sequence of numbers (engendered by the operation
+1)” (Piaget, 2001/1947, p. 39). This perspective has roots in the work of French
mathematician Henri Poincaré (1952/1905): “Mathematicians do not study objects
but relations between objects” (p. 40). We can observe its lasting influence in mod-
ern philosophies of mathematics3 and even policy documents, such as those pro-
duced by the National Research Council: “numbers do not exist in isolation” (NRC,
2009, p. 30).
Beginning from units of 1 and − 1, we can define every integer through iterative
addition of those units: 5 is five iterations of 1, and − 7 is seven iterations of −1. We
rely on such recursion to produce an infinite set of integers. Formally, we say that 1
and − 1 generate the group of integers under addition.
Psychologically, we can think about the closure property as describing a kind of
“wholeness” (Piaget, 1970, p. 6). Students construct whole numbers as nested
sequences. Whole numbers are whole, not only in the sense that each number in the
sequence consists of undivided units (1 s) but also because combining any two num-
bers in the sequence results in another number in the sequence.
As with all mathematical objects, we construct numbers as coordinations of
action and then transform them through further action. Binary operations, like addi-
tion, represent this action on objects as a composition of two like objects. For exam-
ple, what we symbolize in writing 2 + 2 is the composition of the actions that define
2
Here, Piaget is referring to Bertrand Russell’s circular definition of number, which Piaget (1971)
critiques in Genetic Epistemology (see pp. 36–37).
3
“We are not given mathematical objects in isolation but rather in structures. That 13 is a prime
number is not determined by some internal property of 13 but rather by its place in the structure of
the natural numbers” (Resnick, 1981, p. 529).
2; namely, 1 and 1 (two iterations of the unit 1). The composition yields 1 and 1, and
1 and 1, or four iterations of 1 (also known as 4).
Note that whole numbers under addition do not form a group because they lack
reversibility until students extend them, as directed quantities, in two directions:
positive and negative (Ulrich, 2012; Wessman-Enzinger, 2019). Students might
count backward from 4, but for closure, when they reach 0, they would need to pro-
duce negative numbers. Moreover, they would need to reconcile these new, negative
numbers with the familiar whole numbers, to form a group.
If we understand mathematical objects as coordinations of action, and if we
understand composition of mathematical objects as a composition of the actions
that define those objects, then closure refers to the scope of possible coordinations
of action. It defines a space for operating. In the case of number, this space consists
of an entire number system defined by our ways of operating (e.g., the set of inte-
gers, defined by iterations of a unit in either of two directions: 1 or − 1). In the case
of geometry, this space could comprise isometries of the plane, all linear transfor-
mations of n-dimensional space, or all possible projections of the plane. As we will
see toward the end of this chapter, Felix Klein (1893) formally classified geometries
in this way, based on groups of transformations.
In the example of integers, we took addition as the binary operation for compos-
ing elements in the group. As formal operations, binary operations differ from many
of the (logico-mathematical) operations Piaget described. Whereas Piaget defined
operations as reversible and composable mental actions, binary operations refer to
ways we might continue or combine the coordinations of action that define a pair of
mathematical objects. In this, logical-mathematical operations and binary opera-
tions have the following in common: “Operations are a continuation of actions; they
express certain forms of co-ordination which are general to all actions” (Inhelder &
Piaget, 1969, p. 291).
In integer addition, we continue the iteration of units (1 or − 1), as described
above. In integer multiplication, we continue a transformation of units. We can
transform the unit 1 into any composite unit, say 3. We can then distribute the units
of 1 within a second composite unit, say 5, over the units of 1 in that first composite
unit: 3 × 5. This is what Steffe (1992) referred to as a unit coordination. It is equiva-
lent to transforming each of the three units of 1 in the first composite unit into the
second composite unit, 5. Several mathematics education researchers have used the
iteration, distribution, and transformation of units to describe such recursive behav-
ior in students’ mathematics (e.g., Confrey & Smith, 1995; Davydov, 1992;
Steffe, 1992).
In the case of finite groups, we can represent closure with a Cayley table, which
shows all possible combinations of elements from the set. For example, consider the
group of symmetries of a rectangle. This group is formed by the four linear transfor-
mations of the plane that leave the rectangle fixed (in its original position in the
plane). They include reflections over the two lines of symmetry shown on the left
side of Fig. 7.2 (rm and rn), a 180-degree rotation about their intersection (ρπ) and a
trivial transformation (i.e., doing nothing or, equivalently, rotating 360 degrees).
The possible combinations of these symmetries are shown in the Cayley table on the
216 A. Norton
Fig. 7.2 Symmetries of the rectangle
right side of Fig. 7.2. Closure is represented in the Cayley table in that no combina-
tion of the four symmetries produces anything but one of those four symmetries.
Identity and Reversibility
In addition to the composability that closure describes, reversibility is the key crite-
rion of logico-mathematical operations, but formally speaking, we could not define
reversibility without an identity element (e.g., the trivial transformation in the
example above). We might think about the identity element as the starting point to
which we return when we compose an element and its inverse. If we think about
these elements as actions (or operations), the identity itself is an action, just one that
has no effect when composed with other actions. Once we have identified an iden-
tity element, we can define inverse elements as pairs of elements (possibly the same
element, as its own inverse) whose composition has the same null effect as the
identity element.
Psychologically, reversibility provides a kind of balance, or stability, to our ways
of operating: “Reversibility is the very criterion of equilibration” (Piaget, 2001/1947,
p. 12). It guarantees that, when acting on an object, we can always return to the
starting point. As such, it renders mathematics completely reliable.
In science, reliability is repeatability: the possibility of returning to a starting
point, repeating the same sequence of actions, and achieving the same result.
However, science can never achieve perfect reliability because the conditions of an
experiment can never be replicated with perfect precision. In contrast, owing to the
closure and reversibility of its actions, mathematics provides for “complete com-
pensation” (Piaget, 1985/1975, p. 133). It “anticipates all transformations and pre-
corrects errors” (p. 133) by never introducing anything but compositions of
reversible actions.
Piaget (1985/1975) distinguished three forms of reversibility: “any operation
always involves relationships of inversion, of reciprocity, or of correlativity with
certain other operations” (p. 133). The first form (inversion) corresponds with the
inverse criterion of groups. Reciprocity refers to ordering relations, as in A
< B. According to Piaget, reciprocity of this relation can take on any of three sub-
forms: A > B, B < A, and B > A (in Beth & Piaget, 1966). Both inversion and reci-
procity appear in Piaget’s research on children’s construction of number, and their
coordination led Piaget to introduce “groupings,” discussed in a later section. The
final form (the correlative) appears only in Piaget’s INRC group—a meta-group that
we will examine in a still-later section.
Math education researchers have recognized the importance of developing
reversible concepts across K-12 mathematics (Greer, 2011; Hackenberg, 2010;
Simon et al., 2016). Although we can identify examples of non-reversibility in for-
mal mathematics, such as multiplying by 0 or non-invertible matrices, reversibility
re-emerges when we consider the mental actions that undergird them (Norton, 2016).
Associativity
Formally, associativity refers to the property that if f, g, and h are elements of a

group, then (h∘g)∘f = h∘(g∘f). We tend to take this property for granted, and we can
when considering functions, mappings, and other transformations. In those cases,
the property only stipulates that one transformation picks up where the prior one
leaves off. Consider the illustration in Fig. 7.3.
From the identity element, i, in the lower left corner of the figure, we can follow
any of three paths in composing h∘g∘f. First, we could move up by way of f, to f∘i,
which is just f; then move diagonally by g, to get g∘f; and finally, move up by h to
get h∘g∘f. But, by closure, h∘g and g∘f are themselves transformations, and it should
not matter which one we substitute into h∘g∘f. Indeed the diagram shows that it does
not matter whether we substitute for h∘g, taking the shortcut from f to h∘g∘f or
whether we substitute for g∘f, taking the shortcut from i to g∘f. Either way, we end
up at h∘g∘f.
Looking at Fig. 7.2, we can see why Piaget (2001/1947) described the property
of associativity as an independence of path. He probably had in mind his group of
displacements, which he often used as an example. In that example, paths refer to
paths traveled in space, and shortcuts, or “detours,” refer to alternative paths that
lead to the same destination (Beth & Piaget, 1966). In a composition of displace-
ments, one displacement always picks up where the prior displacement left off.
Fig. 7.3 Independence

of path
218 A. Norton
Fig. 7.4 Twenty-five

marbles
Associativity, as independence of path, explains why we often find multiple solu-

tions to a single mathematical problem. In a study of fourth-grade students in the
USA and Japan, Silver et al. (1995) found that children readily produce multiple
solutions to enumeration tasks. The researchers asked students to determine the
number of marbles shown in Fig. 7.4 in as many different ways as they could.
Students ordered, subdivided, and grouped the marbles in various ways and then
counted, added, and multiplied to reach the same result: 25. Interestingly, the
researchers found that children in Japan relied more on multiplication in their solu-
tions, whereas children in the USA relied more on addition. This finding buttresses
another Piagetian idea: that children use the mathematical structures they have
available to assimilate tasks, including the patterns they see.
Multiple solution paths are similarly evident in solutions to probability problems
wherein solutions involve determining numbers of possible combinations. Likewise,
in algebra, we can manipulate equations variously to solve them, and in geometry,
we can construct squares in numerous ways. In each of these domains, we ulti-
mately rely on the associativity of mental actions that define mathematical objects.
Group-Like Structures
In some of the examples given heretofore, group elements have been described as
actions or operations. In other examples, they have been described as numbers. This
is not problematic so long as we recognize that numbers themselves are the products
of mental actions, such as unitizing and iterating, as specified by Steffe (1992). As
such, the combination of two numbers constitutes a combination of the mental
actions that define them; just consider the prior example of 2 + 2 = 4. However, in
describing the group-like structure of “groupings,” Piaget sometimes referred to
elements as classes (sets of objects that share a property). These classes are not
necessarily numbers, and their status as mathematical objects (arising from the
coordination of actions) is not clearly specified. They appear more closely related to
the development of set theory, or logic (see Chap. 10), than mathematics. On the
other hand, logic and mathematics are intertwined within Piaget’s logico-
mathematical operations.
Whereas groups satisfy four criteria (closure, identity, reversibility, and associa-
tivity), some algebraic structures satisfy only a subset of these criteria or other
weaker criteria. For example, consider the set of positive rational numbers, closed
under the binary operation of division. There is an identity element, 1, and every
element of the set has itself for an inverse (e.g., 5/7÷5/7 = 1). However, this structure
does not satisfy the associativity criterion because, for example,
(5÷7)÷(5/7) ≠ 5÷(7÷5/7). Instead, it satisfies a weaker condition known as the Latin
squares property. In the case of finite sets, the Latin squares property specifies that
every element appears exactly once in each row and column of the Cayley table
(think Sudoku), as it does in the table on the right side of Fig. 7.2.4 This structure is
known as a loop.
Piaget consistently included composability and reversibility (with the implicit
inclusion of an identity) in his descriptions of logico-mathematical operations, but
he often neglected associativity. It could be that he took this property for granted, as
we are wont to do, but he also wanted to broaden his consideration of algebraic
structures to include non-associative group-like structures. Specifically, he loosened
the associativity criterion for the grouping structure.
Properties of Groupings
Readers might find it difficult to understand Piaget’s concept of groupings, for a few
reasons. First, unlike the group structure, which is central to abstract algebra, the
grouping structure “is essentially only of psychological interest, and this is due to
its elementary character as well as to its own restricted nature” (Beth & Piaget,
1966, p. 172); it seems to have arisen to accommodate Piaget’s models of children’s
construction of number (Piaget & Szeminska, 1952). Second, although groupings
appear in many of Piaget’s books, the most comprehensive definitions appear in
books not yet translated to English, especially Essai de Logique Opératoire (1972a).5
Finally, as a few critics have noted, even when he does elaborate on his meaning,
Piaget sometimes mischaracterizes or ambiguously describes their formal logic.
Piaget’s distinction between groupings and groups aligns with his distinction
between concrete operations and formal operations: groupings describe the pair-
wise combination of concrete operations, and groups describe the complete combi-
natorial system of formal operations. Whereas the integers form a group under the
operation of addition, they form only a grouping under ordering relations (2 is less
than 3) and class inclusions (even numbers are contained in the class of integers).
4
Note that all finite groups satisfy the Latin squares property because they are associative, which
is an even stronger condition. We can use associativity and the other properties of a group to prove
that every element in a finite group appears exactly once in each row and column of the Cayley table.
5
Among books that have been translated to English, Psychology of Intelligence (2001/1947) and
Mathematical Psychology and Epistemology (1966) stand out, so those are used as chief references
on groupings.
220 A. Norton
Orderings and class inclusions form a lattice, which describes the nested relation-
ships of numbers or classes. Piaget (1972a, p. 92) designed the grouping structure
as a hybrid of a lattice and a group:
So the problem is the characterization of a structure that reconciles the proper
reversibility of a group and a system of nestedness, properly found in the lattice. It’s
this double existence that fills the notion of a “groupement” [grouping]. One can, in
effect, conceive of the grouping as a lattice rendered reversible thanks to a game of
dichotomies or complementary hierarchies (e.g., a set A and its complement A’
within a larger set B).
The distinction between groups and groupings is evident in the properties Piaget
specified for groupings, in comparison to the four criteria for groups, listed above.
The following five properties of groupings appear in The Psychology of Intelligence
(2001/1947):
1. Combinativity (composability): Any two concrete operations in the grouping can
be combined (or composed) to form a new operation in the grouping. This prop-
erty resembles the closure criterion for groups, but note the distinction in lan-
guage here: combining a pair of operations generates a new operation in the
grouping. Because groups anticipate all possible combinations, the composition
of elements within a group produces another element of the group. In other
words, the group is a complete structure, but the grouping structure is actively
built up “step by step” by coordinating pairs of concrete operations (Beth &
Piaget, 1966, p. 173).
2. Reversibility: Here, reversibility refers to reversing the composition under the
binary operation. Combined concrete operations can be separated again by
reversing the binary operation used to combine them. For example, if we com-
bine the concrete operations that determine the ordering relations A < B and
B < C, we get A < C, but we can separate A < C again into its constituent rela-
tions, A < B and B < C. Likewise, we might take the sum of two numbers, as in
5 + 2 = 7, but we can also separate them again.
3. Identity: Related to the previous property, and like groups, groupings have a kind
of reversibility of their elements, but this reversibility can take different forms.
For example, ordering relations are reversed by reciprocity, which can take on
any of three sub-forms. Specifically, the reciprocal of the ordering relation, A
< B, can be expressed as B < A, A > B, or B > A; depending on whether the ele-
ments (A and B) are reversed, the relation (< or >) between them is reversed or
both. When reversibility takes the form of inversion, as it does in groups, Piaget
described it as a kind of “annulment.” One element of the grouping cancels out
another, as in +5–5 = 0. In other words, the concrete operations that define +5
and − 5 (five iterations of a unit in the positive or negative direction, respec-
tively) combine to have no net effect (the identity).
4. Associativity: Piaget specified that the combination of concrete operations should
be associative, but later, in Mathematical Psychology and Epistemology, he
made the following distinction: “A group is associative, whilst the associativity
of a grouping is restricted to the combinations of distinct terms; (A + A)-A is not
identical with A+(A-A)” (Beth & Piaget, 1966, p. 174). We elaborate on this
distinction in explaining the fifth property.
5. Tautology (idempotence): To understand how (A + A)-A might not equal to
A+(A-A), thus violating associativity, consider the way we combine classes. The
class of squares combined with itself is simply the class of squares: A + A
= A. This property is called idempotence, though Piaget sometimes referred to it
as ‘tautology’.6 Idempotence appears more formally in set theory, where the sum
of two sets is their union, and any set in union with itself is simply the original
set. When we take the union of two distinct sets, A + B, we generally produce a
new set, except when A is contained in B (or vice versa). In that case A + B=B,
so (A + B)-B=B-B = 0 ≠ A = A + 0 = A+(B-B); thus, the non-associativity. Note
that, in a group, only the identity element would be idempotent.
Recall that Piaget’s research on children’s construction of number shaped his group-
ing construct. His models of that construction consisted of classes and ordering
relations (Piaget & Szeminska, 1952). Due to idempotence (or “tautology”), num-
bers cannot be constructed based on classes alone. After all, if A + A = A, there is
no iteration and, thus, no production of new numbers (think 1 + 1 = 1). To construct
numbers, children need to make distinctions between them, relying on ordering
relations, as well as classes. If we make each iteration of 1 distinct by ordering it
within a sequence, we can break the tautology. Through ordering, we have 1 + 1 = 2,
where 2 > 1, but at the same time, the class 1 is contained in the class 2. Although
Piaget’s models of children’s construction of number have been refined, most nota-
bly by the work of Les Steffe (1992), the point here is that classes and ordering
relations played particularly important roles in Piaget’s models and his conceptual-
ization of groupings.
In all, Piaget (2001/1947) specified eight elementary groupings corresponding to
three binary dimensions: additive versus multiplicative, symmetrical versus
asymmetrical,7 and classes versus relations. Of these eight groupings, we have con-
sidered two:
• Class inclusion: additive, asymmetrical classes
• Ordering relations: additive, asymmetrical relations
These two groupings are additive because they grow through additive operations
like taking the union of classes or iterating a unit of 1. They are asymmetrical
because they build in one direction. The first describes classes or sets of objects
sharing one or more properties, and the second describes relationships between
objects.
6
Even within the same text, Piaget (1972a) sometimes referred to this relationship as idempotent
(p. 90) and in other paces referred it to tautology (p. 97). It is not clear why.
7
In the case of multiplicative relations, Beth and Piaget (1966) characterized this dimension as bi-
univocal (one to one, like a multiplication table) verses co-univocal (many to one, like tree
branching).
222 A. Norton
The synthesis of pairs of elementary groupings generates various logico-

mathematical structures. We have already noted one example: the construction of
natural numbers as a synthesis of additive asymmetrical classes and additive asym-
metrical relations. When its elements become organized as directed quantities, this
structure becomes complete and “no longer a qualitative grouping but the group” of
integers under addition (Piaget, 2001/1947, p. 50). Investigating each of the elemen-
tary groupings and their various syntheses would go beyond the scope of this chap-
ter (see Beth & Piaget, 1966, pp. 174–183 for a detailed discussion). Instead, we
turn to a critique of the grouping construct itself.
A Critical Analysis of Groupings
The most comprehensive critique of Piaget’s grouping construct comes from Erich
Wittmann, in the form of an Educational Studies in Mathematics article published
in 1973. Erich Wittmann earned a PhD degree in mathematics from the University
of Erlangen—not a trivial fact, as we will see—completing a dissertation on group
theory. He became a pioneer for mathematics education as a research field in
Germany. His critique begins, “In spite of the high value of Piaget’s theory, certain
quite serious deficiencies should not be overlooked” (Wittmann, 1973, p. 125).
Wittmann accepted the five properties Piaget laid out for groupings, as described
above, but he defined those properties and the grouping itself more formally.
Specifically, he defined concrete operations as changes in states: (A, B) represents
the concrete operation that transforms state A into state B. For example, a displace-
ment can be represented by a transformation from position A to position
B. Combining (or composing) the two operations (A, B) and (B, C) yields the opera-
tion (A, C), the transformation from A to C. Such a composition can be reversed, or
decomposed, as Piaget described. Relatedly, each concrete operation (A, B) has an
inverse operation (B, A), such that their composition yields the identity (A, A).
Wittmann characterized the transformations (concrete operations) themselves as
“trivially associative” because, as illustrated in Fig. 7.3, all transformations are.
However, one state could be absorbed into another. For example, both class inclu-
sions generate a lattice wherein state A could be contained within state B. We might
combine the states themselves by considering their union. In the case where state A
is contained within state B, this union is B. By combining states in this way, we also
have that A union A is A (A + A = A). This distinction cleared up the apparent ambi-
guity in Piaget’s language as to whether groupings are associative. The composition
of operations is associative, but the class inclusions and ordering relations of states
may not be. Specifically, if A + A = A, A + A-A could be either (A + A)-A = A-A = 0
or A+(A-A) = A + 0 = A.
In affirmation of Piaget’s grouping construct and its five properties, Wittmann
provided several examples of their application in understanding students’ logico-
mathematical reasoning. For instance, in working with Cuisenaire rods, students
might rely on a grouping structure to reason through various ways to break down the
long, orange bar (10 units long). They might replace the orange bar with two brown
bars (5 units long) and then proceed to break down one of the brown bars into two
red bars (2 units long) and one tan bar (the unit bar). Each of these changes in states
(e.g., from the orange bar to two brown bars) represents a concrete operation of
breaking apart. Students might not anticipate all possible partitionings of the orange
bar (this is a challenging combinatorics problem), but they can imagine performing
them step by step, and they can reverse each step by uniting bars into longer bars.
In general, the grouping structure models how students might begin coordinating
operations that are not yet formal: concrete operations, which have not yet been
organized in a group structure. However, grouping structures, as defined by Piaget
and refined by Wittmann, are not the only structures that we can use to model con-
crete operations. In the next section, we will consider an example of another group-
like structure.
The Splitting Loope and the Splitting Group
We have mentioned the partitioning and iterating pair as an example of mutually

reversible operations: one operation undoes the other, yielding the identity opera-
tion (doing nothing). For example, given a whole, a student might partition it into
five equal parts; then, they might iterate any one of those parts five times to repro-
duce the whole. Before students have organized partitioning and iterating as inverse
operations, they might not anticipate this result. Consider the following task, refer-
ring to the bar illustrated on the left side of Fig. 7.5: “If this bar is five times as big
as your bar, show me your bar.”
The task calls for the student to produce a bar that, when iterated five times,
would produce the given bar. To produce the other bar, a student might partition the
given bar into five equal parts. Thus, the student would need to anticipate a revers-
ible and composable relationship between partitioning and iterating. In fact, many
students who solve this task can do so by making a single mark, as shown on the
right side of Fig. 7.5. They understand that the five iterations of the smaller part
would necessarily produce the other four partitions of the whole. This kind of rea-
soning is called splitting.
Fig. 7.5 Splitting task

224 A. Norton
Steffe (2002) described splitting as the simultaneous coordination of partitioning

and iterating. This description fits a kind of development available to students work-
ing with concrete operations: “the fusion into a single act of anticipations and retro-
spections—which is the basis of operational reversibility” (Piaget, 1972b, p. 35). It
is the kind of fusion that group-like structures can explain by specifying the various
compositions of partitioning and iterating that a child might anticipate before acting
and outside of the context of any particular scheme.
Steffe (2002) introduced splitting to explain how students can iterate a unit frac-
tion beyond the whole and maintain the relationship between the resulting improper
fraction and the whole. Earlier in this chapter, we presented the example of iterating
1/5 seven times to produce 7/5. Students operating with an IFS can maintain rela-
tionships between 1/5, 7/5, and the whole. These relationships define 7/5 as a num-
ber measured not only in units of 1/5 but also in units of 1.
The challenge of conceptualizing improper fractions as numbers is indicated by
students’ reluctance to work with improper fractions at all. After all, as one fourth-
grade student put it, “How can a fraction be bigger than itself” (Olive & Steffe,
2002, p. 428)? To avoid confusion, teachers sometimes encourage students to imme-
diately convert improper fractions to mixed numbers.
Steffe (2002) hypothesized that splitting—the simultaneous coordination of par-
titioning and iterating—provided students with the operational power to construct
the IFS: “Upon the emergence of the splitting operation, I regard the partitive frac-
tional scheme as an iterative fractional scheme” (p. 299). However, some students
who split (i.e., can meaningfully solve tasks like the one illustrated in Fig. 7.5)
struggle to conceptualize improper fractions (Hackenberg, 2007). Such students
might engage in reversible reasoning, producing wholes from unit fractions and unit
fractions from wholes, but they do not conceptualize 7/5 as seven iterations of a unit
that, when iterated five times, would produce the whole. These students might even
produce a unit fraction from a given improper fraction, or vice versa, but have not
yet conceptualized improper fractions as numbers (Norton & Wilkins, 2010). In
other words, they can split in various fraction contexts but have not yet constructed
an IFS. Hackenberg (2007) revised Steffe’s splitting hypothesis to explain this
disparity.
Both Steffe (2002) and Hackenberg (2007) viewed splitting as fundamental to
the construction of IFS, but Hackenberg found that IFS demands more. All four of
the sixth-grade students in Hackenberg’s study could split, but only two of them
could also assimilate three levels of units (see Chap. 11). In the context of working
with improper fractions, these two students could conceptualize improper fractions
like 7/5 as a three-level structure: 7/5 as a unit composed of seven units of 1/5, five
of which comprise the whole (5/5). These students’ ways of operating enabled them
to transcend the whole, and they readily constructed an IFS. Thus, Hackenberg
(2007) revised Steffe’s splitting hypothesis as follows:
My revision of that hypothesis is that students can construct the splitting opera-
tion without also interiorizing the coordination of three levels of units, and this
interiorized coordination of three levels of units appears to be necessary for the
construction of improper fractions and, therefore an iterative fraction scheme.

(p. 46).
Whereas splitting provides for the reversibility of partitioning and iterating oper-
ations, it falls short in constructing IFS because, until students interiorize three lev-
els of units, they do not maintain the various units produced through partitioning
and iterating. For instance, students might iterate 1/7 of a whole nine times to pro-
duce 9/7, and they might understand that they can reverse this operation by parti-
tioning the result into nine equal parts, but in so doing, they might lose track of the
whole and conceptualize 9/7 as 9/9. IFS describes ways students might use parti-
tioning and iterating to produce and maintain the three levels of units in an improper
fraction (the unit fraction, the non-unit fraction, and the whole). However, as noted
at the start of this chapter, the scheme structure does not describe the coordination
of operations themselves within reversible and composable systems. For that, we
rely on a group-like structure we call the splitting loope (Wilkins & Norton, 2011).
As previously mentioned, groups are not the only group-like structures. Loops,
for example, share most properties of groups but lack associativity. Here, we intro-
duce a similar structure, which we call a “loope.” Like loops, loopes are closed
under a binary operation, have an identity element, have an inverse for every ele-
ment, and are not necessarily associative. In contrast to loops, loopes do not even
necessarily satisfy the Latin squares property. Specifically, consider the structure of
the splitting loope presented in Table 7.1.
Within the splitting loope, partitioning and iterating are organized as inverse
operations, but their composition is not associative, nor does it even satisfy the Latin
squares property. We can see the violation of both properties in Table 7.1, wherein
P appears twice in the third column and I appears twice in the second row. Owing to
their qualitative character as concrete operations,8 partitioning and iterating are
idempotent, meaning PP=P and II=I. As a consequence of this idempotence, (PP)
I ≠ P(PI): the left side of the inequality yields (PP)I=PI=1, and the right side yields
P(PI) = P1 = P.
Although students might work out the product of two partitions or two iterations
in particular contexts, they might also lack a structure for assimilating the three
levels of units involved. Thus, they might only anticipate the product of two parti-
tions as a partition and the product of two iterations as an iteration. For example,
they might understand that a third of a fifth makes a smaller part without coordinat-
ing how each third of each fifth produces a fifteenth of the whole. In Wittmann’s
Table 7.1 The splitting loope

Composition Identity (1) Iterating (I) Partitioning (P)
Identity (1) 1 I P
Iterating (I) I I 1
Partitioning (P) P 1 P
8
Piaget (2001/1947) described the groupings of concrete operations as having a qualitative (not yet
quantitative) character.
226 A. Norton
(1973) terms, we might say that these students do not make distinctions between
certain states. Thus, the concrete operations that transform those states are “tauto-
logical,” and we have a non-associative structure. In Hackenberg’s (2007) terms, the
splitting loope explains how partitioning and iterating might be coordinated as
inverse operations that still lack the power for constructing an IFS, providing a theo-
retical basis for her revision of Steffe’s splitting hypothesis.
Beyond the mutual reversibility of partitioning and iterating operations, the IFS
additionally requires the coordination of three levels of units. This development cor-
responds with the associativity of the splitting group (Norton & Wilkins, 2012),
wherein products of compositions of partitions and iterations transcend their quali-
tative character and consistently include quantitative distinctions. Within the split-
ting group, every composition of partition and an iteration generates a specific
fraction, even if a fraction can be generated in different ways (e.g., see 1/2 and 2/4
within Table 7.2). In turn, the product of two positive rational numbers can be
expressed as—and determined by—a composition of four partitions/iterations,
which can always be reduced to a composition of a partition and an iteration, as
expressed in Table 7.2. For example, we can represent 3/4 × 2/3 as P4I3P3I2. Relying
on the commutativity of partitioning and iterating operations (not a property shared
by all groups or operations), we have I3I2P4P3. Associating the iterations and parti-
tions, we have (I3I2)(P4P3) = I6P12, which is 6/12 or 1/2.
Like integers, fractions do not exist in isolation but within a system for compos-
ing them. With the splitting group, we achieve a structure isomorphic to the positive
rational numbers under multiplication but defined in terms of students’ coordina-
tions of their own mental actions. In the splitting group, we also have the multiplica-
tive analogue to the group of integers under addition. In fact, those two groups are
also isomorphic. Research on students’ construction of integers suggests that it, too,
relies on the coordination of three levels of units (Ulrich, 2012). However, the par-
ticular mental actions coordinated within each structure differ. In the splitting group,
partitioning and iterating are coordinated to support multiplicative reasoning. In
integer addition, iteration in two different directions (“directed quantities”) is coor-
dinated to support additive reasoning.
Piaget described the construction of number as a synthesis between two group-
ings: the grouping of class inclusion and the grouping of ordering relations. In elab-
orating on Piaget’s models, unit coordination was Steffe’s major contribution. In her
revision of Steffe’s own splitting hypothesis, Hackenberg (2007) further demon-
strated the central importance of unit coordination in children’s construction of
Table 7.2 Generating the splitting group

I1 I2 I3 I4 I5
P1 1 2 3 4 5
P2 1/2 1 3/2 4/2 5/2
P3 1/3 2/3 1 4/3 5/3
P4 1/4 2/4 3/4 1 5/4
P5 1/5 2/5 3/5 4/5 1
number. Therein, we find the key distinction between the splitting loope and the
splitting group. The operations of the IFS become reversible within the splitting
loope but only become associative within the splitting group.
We use the example of the splitting loope and splitting group to demonstrate dif-
ferent ways that operations might be organized within reversible and composable
systems. At the same time, this distinction demonstrates the dependence of schemes
(e.g., the IFS) on the organizations of their operations. A scheme becomes revers-
ible when its operations become reversible within a structure that we might model
with a group-like structure. For example, the splitting loope describes the organiza-
tion of partitioning and iterating operations as inverses of one another. Groups
impose additional structure. In the case of the splitting group, the additional struc-
ture of associativity enables students not only to compose partitions and iterations
of various sizes but also to coordinate the units they produce. What they produce is
isomorphic to the group of all positive rational numbers under multiplication. By
using groups to model the coordination of students’ own operations, we see how
that formal mathematical structure arises from students’ own activity.
Genetic Roots in Mathematics
Well into the nineteenth century, Western mathematicians generally operated under
a Platonist paradigm inherited from the ancient Greeks, pretending to study celestial
mathematical objects whose existence did not depend on human labor. We can see
evidence of this perspective in the opening quote from Blaise Pascal. However,
around the turn of the twentieth century, debates arose around the foundations of
mathematics, with mathematicians like Brouwer taking a constructivist stance not
so different from the one presented here (Mancosu, 1998). In this section, we con-
sider other indirect contributions by mathematicians to Piaget’s epistemology of
mathematics. These contributions concern the special role the group structure plays
in mathematics, and they include the work of Felix Klein, Emmy Noether, and the
Bourbaki.
The Erlangen Program
Upon joining the faculty at the University of Erlangen in 1872, Felix Klein intro-
duced a new program for classifying geometries. This program came on the heels of
two new developments in Western mathematics: the invention (some will say dis-
covery) of non-Euclidean geometries and the formalization of the group structure.
Klein (1893) decided that geometry had focused too much on geometric figures
(e.g., triangles) and not enough on the operations that produce and transform them.
His Erlangen program foregrounded such transformations, highlighting the dynamic
nature of geometry. At the same time, it allowed him to classify various geometries
228 A. Norton
(e.g., Euclidean geometry and projective geometry) based on groups and subgroups
of these transformations.
Consider, for example, the principal group, which describes all transformations
of the plane that leave Euclidean figures invariant. These transformations include
isometries of the plane (reflections, rotations, translations, and their various combi-
nations), as well as dilations. Note that dilations are included because, although
absolute sizes change under such a transformation, relations between angles and
side lengths of geometric figures do not.
Klein (1893) used the principal group to characterize Euclidean geometry. This
group fits as a subgroup within the group of projections, which characterizes projec-
tive geometry. Thus, we see how groups and their relationships as subgroups within
other groups can provide a classification of various geometries. We take this as a
launching point for an expanded Erlangen program that extends beyond geometry
and frames transformations as mental actions.
Consider the simple mental action of reflection. Each reflection has itself as an
inverse action (an involution) and can be composed of other mental actions.
However, composition of reflections is not closed: two reflections can generate
either a translation, when the lines of reflection are parallel, or a rotation, when the
lines of reflection intersect. In fact, we can generate every rotation and translation of
the plane by coordinating a pair of reflections as mental actions. Furthermore, we
can generate every isometry of the plane by coordinating no more than three
reflections.
Rather than delving into the mathematical details of the above claims, we con-
sider their framing within Piaget’s genetic epistemology. Namely, we can generate
the entire group of isometries of the plane by coordinating a simple reversible and
composable mental action and by reflecting on that coordination. Framed in this
way, the group construct is not only useful in formalizing mathematics (e.g., clas-
sifying geometries); it describes the culmination of our psychological abilities to
construct mathematics.
Piaget, Inhelder, and colleagues deserve credit for a paradigm shift in develop-
mental psychology. Above all, they established a framework for conducting empiri-
cal studies of children’s mathematics. However, they were not the first to suggest
reversible and composable actions as the basis for mathematics. Poincaré
(1952/1905) also talked about a group of displacements in Science and Hypothesis.
His views of “space and geometry,” including their basis in children’s motor activ-
ity, seem to have influenced Piaget’s epistemology of mathematics almost as much
as Kant’s rational empiricism.9 In Chap. 4 of the book, Poincaré (1952/1905)
claimed, “the object of geometry is the study of a particular ‘group’; but the general
concept of group pre-exists in our minds, at least potentially” (p. 70). There lies the
heart of Piaget’s epistemology of mathematics.
9
Piaget and Inhelder (1967/1948) acknowledge the influence of both Kant and Poincaré, at the start
of their book, The Child’s Conception of Space.
The year Science and Hypothesis was published, Emmy Noether was extending
Klein’s Erlangen program through her study of algebraic invariants (Rowe &
Koreuber, 2020). Like Wittmann, Noether had taken up graduate studies at the
University of Erlangen, where Klein’s program had a lasting effect. Her work on
algebraic invariants extended the Erlangen program by considering various trans-
formations that leave algebraic relationships unchanged. For example,
x4 + ax3 + bx2 + cx + d = 0 has the same number of real roots when we substitute
x = y-a/4, even though the new equation (in y) will not have a cubed term.10 Noether’s
dissertation included over 300 such invariants, but more general and in three vari-
ables rather than just one.
Taken together—the focus on mental actions and the extension to algebra—we
achieve an expanded Erlangen program that complements Piaget’s genetic episte-
mology. It frames all of mathematics as a coordination of mental actions within
groups for composing and reversing them. Consider again the example of the split-
ting group. Therein, we find fractions defined not as ordered pairs of numbers but as
compositions of the mental actions of partitioning and iterating. As in the Erlangen
program, the focus is on these dynamic operations rather than the static states they
might produce. Similar to the way we can define a square by its symmetries (its
invariance under transformations within a group of isometries), each fraction can be
defined as a composition of partitioning and iterating operations from the splitting
group (e.g., I3P5(1) = 3/5), and is held invariant under further transformation by
partitionings and iterations of the same size (I2P2(3/5) = 6/10 = 3/5).
INRC and the Bourbaki
Piaget was fascinated by the mathematical and scientific developments of his time.
Trained as a biologist, he continually read up on research ranging from neurosci-
ence to general relativity, seeking out connections to his own work. In Genetic
Epistemology, Piaget (1971) recounted a conversation he had with Einstein about
the construct of time: “Einstein himself recognized the relevance of psychological
factors, and when I had the good chance to meet him for the first time in 1928, he
suggested to me that it would be of interest to study the origins in children of notions
of time and particular notions of simultaneity” (p. 7). Piaget took on that task and
published his findings in The Child’s Conception of Time (1969/1946). Later,
Siegler and Richards (1979) reproduced the main result: “both speed and distance
concepts appeared to be mastered well before the concepts of time” (p. 288). This
finding fits nicely with the main premise of Einstein’s special relativity that time is
relative to a frame of reference.
10
Note that even algebraic manipulation relies on algebraic invariance: “manipulating a system of
equations but maintaining its solution set” (Harel, 2008, p. 15).
230 A. Norton
Within advanced mathematics, abstract algebra held Piaget’s attention above all
else. In addition to groups, he adopted constructs like lattices and Boolean algebra
into his work. Piaget justified his use of such constructs not only because of their
power to explain children’s behavior but also because Piaget believed they had psy-
chological roots that underpinned mathematics itself. From his study of children’s
construction of number, Piaget realized that, beyond algebra, he would need addi-
tional structures, such as ordering relations, to investigate the psychological roots of
mathematics.
Piaget settled on three “mother structures”: groups, ordering relations, and topo-
logical structures. Piaget never elaborated much on topological structures, except to
say they “deal with the concepts of neighborhood, limit, and continuity” (Beth &
Piaget, 1966, p. 165). Nevertheless, Piaget seemed to need topological relations to
explain how children learn to partition continuous wholes into parts, as in their con-
structions of geometry and fractions. Interestingly, Piaget found that children attend
to topological relationships (e.g., the closure of a circle or the intersection of lines)
long before attending to geometric properties (e.g., right angles or the number of
sides in a rectangle)—another reversal from historical development, similar to the
relatively late development of group theory noted previously.
Regarding the three mother structures, Piaget (1971) recalled one particularly
affirming event from a conference in Paris:
A number of years ago, I attended a conference outside Paris entitled “Mental
Structures and Mathematical Structures.” This conference brought together psy-
chologists and mathematicians to discuss of these problems. For my part, my igno-
rance of mathematics then was even greater than what I admit today. On the other
hand, the mathematician Dieudonné, who was representing the Bourbaki mathema-
ticians, totally mistrusted anything that had to do with psychology. Dieudonné gave
a talk in which he described the three mother structures, Then I gave a talk in which
I described the structures that I had found in children’s thinking, and to the great
astonishment of us both, we saw that there was a very direct relationship between
these three mathematical structures and the three structures of children’s operational
thinking. (p. 26).
The Bourbaki was a company of French mathematicians who published their
work under the name of a popular general who had served under Napoleon. Their
mother structures preceded category theory in attempting to unify mathematics, not
so different from Klein’s goal for geometry.11 The fact that these three structures
related so closely to those Piaget had identified provided further affirmation of
Piaget’s approach to epistemology. Namely, the study of children’s logico-
mathematical development is the study of the (psychological) foundations of math-
ematics itself.
We have seen how Piaget used groupings to explain how children operating at
the concrete stage might synthesize class inclusion and ordering relations. Piaget
Piaget notes as much in a separate account of the mother structures or “matrix structures” (Beth
11
& Piaget, 1966, p. 164).

also desired a way of explaining how the three mother structures could form a unity
at the formal stage. This was the purpose of his INRC group:
[In groupings] reversibility consists either of inversion (for classes) or reciprocity (for rela-
tions), but the two are not integrated into a single system. Consequently, they do not coin-
cide with the group structure of inversions and reciprocities (the INRC group) and remain
in the state of incomplete groups. (p. 275)
The INRC group is isomorphic to the Klein four-group. This additional connection
between Jean Piaget and Felix Klein is not wholly coincidental. Group theory was
still a young field around the turn of the twentieth century. Both Klein and Piaget
were finding power in its potential to unify mathematics. The Klein four-group is
the group in which all four elements have themselves as inverses (i.e., they are invo-
lutions). This was of mathematical interest to Klein and of psychological interest
to Piaget.
Mathematically, the Klein four-group describes the symmetries of a rectangle
(D2): the identity (do-nothing symmetry), two lines of reflection, and a 180-degree
rotation (see Fig. 7.2). Psychologically, the INRC group describes how three differ-
ent kinds of reversibility (each an involution) from three different psychological
structures can be brought together within a single structure (see Table 7.3).
To illustrate the INRC group and its importance in the formal stage of operating,
Piaget (1971) used the metaphor of a snail (a fitting choice for a researcher who
began his career studying mollusks):
Let us say that we have a snail on a little board. If the snail moves to the right, we
can take that as the direct operation. And the inversion, or negation, would be the
snail moving to the left. But the reciprocal of a move by the snail to the right would
be a move by the board to the left, and then the correlative would be a move by the
board to the right…With respect to an external frame, there are two ways of revers-
ing the snail’s motion: one is for the snail to move back again; the other is for the
board to move. Before children are able to synthesize the two types of reversibility
in a single system, that is, before the age of 11 or 12 years, they cannot resolve
problems of this sort, which require a coordination between two different types of
motion with two possible frames of reference. (pp. 39–40).
This explanation is problematic because, with some training, children as young
as 7 or 8 can solve the snail task (Easley, 1964). If we consider the movements of
the snail and the board as concrete operations (rather than forms of reversibility),
Table 7.3 The INRC group

I N R C
I I N R C
N N I C R
R R C I N
C C R N I
232 A. Norton
there is no reason children of that age would not be able to coordinate them in the
context of the task.
Ascher (1984) provided a separate critique of the INRC group regarding Piaget’s
use of R. As we have noted, Piaget had ambiguously referred to three possible sub-
forms of reversibility among ordering relations. Generally, he meant for R to reverse
the implication p➜q (p implies q) so that R(p➜q) = q➜p. In at least one place,
Piaget specified this reversal in terms of Boolean algebra, with its 16 binary propo-
sitional operators: “We may call reciprocity the transformation that consists in
negating the propositions that compose a binary operation” (Piaget, 1950, p. 144, as
cited in Ascher, 1984). As Ascher points out, there are two automorphisms of the 16
binary propositional operators that perform the reversal, from p➜q to q➜p, and the
one Piaget specified, by negation, was the wrong one. “A group isomorphic to D2
does appear in quite an essential way whenever one considers structure-preserving
transformations of algebras of relations. This group, however, is not INRC” (p. 300).
Ascher instead refers to the group where R is replaced by the other automorphism,
still transforming p➜q to q➜p.
With Ascher’s modification of R, the INRC group neatly brings together two
forms of reversibility into a single group. These two forms, N and C, come from
groupings of class inclusion and ordering relations, respectively—the two group-
ings that come together in children’s construction of number. Note that C does not
correspond with a form of reversibility inherited from any particular kind of group-
ing. Rather, it is forced on the INRC group by closure (the group must contain N
composed with R). This means that INRC accounts for only two of the three mother
structures, leaving topology behind. We can imagine Piaget would have liked the
three non-trivial operations, NRC, to correspond to each of the three mother struc-
tures. However, as we have noted, Piaget never had much to say about topology, and
in particular, he never specified what form of reversibility it might take on except
through vague reference to “proximity and separation” (Piaget & Inhelder,
1967/1948, p. 149).
pplying Mathematical Structures to Mathematics

A
Education Research
Piaget earned a lasting reputation for his ground-breaking research in developmen-

tal psychology. All along, he pursued a primary interest in epistemology, particu-
larly an epistemology of mathematics (see especially Beth & Piaget, 1966). Within
that epistemology, groups played the central role. In particular, they describe the
reversibility and composability of actions as logico-mathematical operations. As
such, groups and group-like structures help explain what distinguishes mathemati-
cal knowledge from other forms of knowledge, including the apparent certainty,
perfect reliability, and unity of mathematics. These distinctions are important
because they help us define our field (mathematics education) and identify some of
the cognitive factors unique to it. In the remainder of this section, we consider the
potential for more specific applications of groups in mathematics education research.
Returning to a point raised at the start of the section on “group-like structures,”
Piaget did not always carry out his program clearly. Specifically, if mathematical
objects arise from a coordination of mental actions, organized with group-like struc-
tures, and if logic and mathematics are of the same nature, then we should be able
to unambiguously identify the mental actions that comprise classes or sets. Piaget
(1972a) began this work in his Essai de logique operatoire: “As we have already
seen, an equivalence relation expresses, like any relation, a quality in understanding.
The corresponding extension, otherwise called the field of this relation, constitutes
a class” (p. 78). That passage suggests that we can define classes and sets based on
equivalence relations, and Piaget goes on to identify the union of sets as an opera-
tion that we can apply to them. However, this approach relies heavily on formal
notions from set theory. Understanding how children construct sets on the basis of
their own mental actions across mathematical contexts and how they might learn to
abstract the logic of sets falls within the purview of mathematics education research
(see Dawkins & Norton, 2022).
The scheme structure is already prevalent in mathematics education research that
adopts a constructivist perspective (see Chap. 3). Schemes describe ways that stu-
dents employ mental actions (operations) in service of solving mathematical tasks.
As researchers, we use schemes to explain students’ problem-solving behavior and
to simulate their mathematical realities. For example, we might use the IFS to
explain how students can produce a 7/5 from a whole and also reverse that way of
operating to reproduce the whole from 7/5. However, our models of students’ math-
ematics remain incomplete until we account for the reversibility and composability
of operations themselves.
In line with Klein’s (1893) Erlangen program, consider the role of groups in
geometry. If a student can imagine the action of reflecting the plane over a line, can
they also imagine the result of composing two such reflections? Formally, the com-
position results in either a rotation about the lines’ point of intersection or a transla-
tion. Can students then compose a pair of rotations about two different points?
Ultimately, through the coordination of their own mental actions, students might
generate the group of isometries of the plane, and the group structure informs us, as
mathematics educators, about that potential. The formal claim that every isometry
of the plane is the composition of at most three reflections then becomes an oppor-
tunity for student empowerment. Isometries are not things to learn but rather opera-
tions to construct through the coordination of their own mental actions.
Questions like those posed above suggest task design. If we understand what
operations students have available and how students might compose them with
other operations, we can design tasks to provoke those coordinations. Furthermore,
if we understand how students have coordinated operations within groups and
group-like structures, we can also imagine the schemes they might construct: the
student’s “zone of potential construction” (Steffe & D’Ambrosio, 1995). Returning
to the splitting hypothesis (Steffe, 2002), for example, Hackenberg’s revision (2007)
suggests that the IFS would not be in the zone of potential construction for a student
234 A. Norton
until they organized partitioning and iterating operations within a system that is not
only reversible and composable but also associative.
Thus, groups and group-like structures can inform task design in two ways: (1)
tasks designed to provoke the coordination of actions/operations themselves within
organized systems that we might model with groups and group-like structures and
(2) tasks designed to leverage the organization of operations in support of students’
constructions of new schemes, within their zones of potential construction. Groups
inform tasks of the first kind by applying our own knowledge of mathematics to
imagine ways students might coordinate the actions/operations they already have
available. How might they compose these actions, in activity and in the context of
solving a task? Groups inform tasks of the second kind by helping us recognize
students’ mathematics as mathematics and imagining new ways for them to lever-
age their mathematical power by constructing new schemes for operating.
Owing to their ubiquity in mathematics, groups, and group-like structures can be
used to model the organization of operations across all domains of mathematics. We
have considered examples from fractions and geometry wherein groups provide
useful models of students’ mathematical power that can also inform task design.
But in comparison to schemes, group-like structures have been underutilized in
mathematical education research, even in those domains. Until we gain a better
understanding of students’ organizations of operations within reversible and com-
posable systems, we will be missing a key aspect of students’ mathematics, includ-
ing what renders it a unified body of knowledge that we can rightly call mathematics.
Summary
Piaget’s (1970) structuralism accounts for the construction of fundamental struc-

tures (like number and space) often taken for granted, even within Kant’s rational
empiricism. These logico-mathematical structures have two operational proper-
ties—reversibility and composability—that, alone, nearly define a group. In
researching children’s logico-mathematical ways of operating, Piaget focused on
how these structures come to be and how they ultimately culminate in groups of
operations. He needed group-like structures (groupings) to explain this formation at
the concrete operational stage.
Children operating with concrete operations, beginning around age 7 or 8, read-
ily compose mental actions and learn how to reverse them, but they do not anticipate
all possible combinations of those actions, nor do those actions form a totality. The
grouping construct describes a structure that comes to be through successive coor-
dinations of action: these children can combine operations in various ways to exper-
iment with them without anticipating the whole structure they will form.
Furthermore, Piaget and Szeminska’s (1952) research on children’s construction of
number demonstrated that the complete coordination of action might require stu-
dents to unite two different groupings (e.g., class inclusion and ordering relations).
If those structures should be unified, an additional structure would be required to
show how their different forms of reversibility relate to one another. The Klein four-
group provided such a structure. Still, it falls short of accomplishing what might
have been Piaget’s ultimate epistemological goal: to unite all of mathematics within
a single meta-structure.
In accord with the Bourbaki, Piaget (1971) identified three mother structures for
all of mathematics: algebraic structures, ordering relations, and topology. With the
N from algebra and the R from ordering relations, NR = C is forced into the INRC
group (isomorphic to the Klein four-group), and C does not correspond in any dis-
cernable way with reversibility in topology. Thus, the INRC group may describe a
unity of class inclusion and ordering relations, helping to explain children’s con-
struction of number, but it does not provide the kind of meta-structure for which
Piaget seemed to hope. Instead, the INRC group, as well as the three mother struc-
tures, might be examples where Piaget fell into the allure of fascinating develop-
ments of his time.
Critiques notwithstanding, Piaget’s focus on reversibility and composability
does accomplish a great deal in describing the unity of mathematics while explain-
ing its psychological development. As mathematicians like Klein (1893) and
Poincaré (1952/1905) had suggested, groups describe the kind of totality, or whole-
ness, that might arise from a complete coordination of our own mental actions. This
is what Piaget (1972b) meant in describing a formal operational stage, beginning
around age 11 or 12. The main takeaway for us, as mathematics educators, is that
groups and group-like structures have proved valuable in describing mathematical
development, even if we leave behind Piaget’s more ambitious claims and have to
overlook some of his mathematical mistakes.
Earlier in the chapter, we elaborated on an example of how students’ mathemat-
ics might develop toward a complete structure for composing and reversing mental
actions and how the transition, from a group-like structure to a group, models that
development. Moreover, that group is isomorphic to the formal mathematics that
students construct: the positive rational numbers under multiplication. We refer to
the transition from the splitting loope to the splitting group (Norton & Wilkins,
2012; Wilkins & Norton, 2011).
The splitting loope describes the kind of progressive coordination of mental
actions that students at the stage of concrete operations might perform. Through
these coordinations, they develop partitioning and iterating as inverse operations.
However, the coordinations are incomplete, as manifested by the non-associativity
of the loope. For example, a partitioning of a partitioning might constitute only
some kind of partitioning for the student—a qualitative character of operations that
is not completely quantified until the student can coordinate the three levels of units
involved. Here, we point to an important distinction between Steffe’s (2002) origi-
nal splitting hypothesis and Hackenberg’s (2007) revision of that hypothesis. The
splitting loope and splitting group capture this distinction in that the former struc-
ture includes only reversibility of the operations, whereas the latter structure also
includes associativity (Norton & Wilkins, 2012).
The splitting loope/group example illustrates the importance of group-like struc-
tures in modeling students’ mathematics. Whereas schemes model the sequences of
236 A. Norton
mental actions that undergird a concept, groups and group-like structures model the
coordination of mental actions themselves as operations. As we have argued,
schemes like the IFS inherit their reversibility from such structures.
Whether fractions, shapes, or algebraic expressions, mathematical objects do not
exist in isolation but, rather, within structured wholes governed by reversibility and
composability. And as mathematicians like Poincaré, Klein, and Noether have
taught us, these objects of study are not static figures but dynamic operations or
transformations. By treating operations and coordinations thereof as mathematical
objects, we open new possibilities for operating on them:
The whole of mathematics may therefore be thought of in terms of the construc-
tion of structures, ... mathematical entities move from one level to another; an oper-
ation on such ‘entities’ becomes, in its turn, an object of the theory, and this process
is repeated until we reach structures that are alternately structuring or being struc-
tured by ‘stronger’ structures. (Piaget, 1972b, p. 70, as cited in Tall et al., 2000).
To various degrees, the basis for logico-mathematical operations is common to
the biology of all living beings by virtue of their acting, growing, and reproducing
in an otherwise unknowable world. Through stages of development, humans become
aware of the logic of their actions, but we often attribute this logic to the structure
of the universe rather than biological structures that govern our own psychology via
our actions in that world. We can even formalize those structures as rules of formal
logic and mathematical systems that appear as God-given, immutable facts that we
have discovered. At least in Piaget’s epistemology, they arise from awareness of our
own psychological ways of operating based on the neurology of living.
“The day when biology will be entirely mathematized, if ever such a thing is possible, we
will certainly see if the equations of protoplasm, and consequently protoplasm itself, result
from our mind or if our mind with its equations results from protoplasm. Perhaps on that
day, psychology will be sufficiently advanced to show mathematicians supporting the first
of these propositions and biologists supporting the second… that they say almost the exact
same thing… but only psychologists will truly know why!” (Piaget, 1971, p. 120).
References
Ascher, E. (1984). The case of Piaget’s group INRC. Journal of Mathematical Psychology, 28,
282–316.
Beth & Piaget (1966). Mathematical epistemology and psychology (Trans. W. Mays). Gordon
and Breach.
Confrey, J., & Smith, E. (1995). Splitting, covariation, and their role in the development of expo-
nential functions. Journal for Research in Mathematics Education, 26(1), 66–86.
Davydov, V. V. (1992). The psychological analysis of multiplication procedures. Focus on Learning
Problems in Mathematics, 14(1), 3–67.
Dawkins, P., & Norton, A. (2022). Identifying mental actions necessary for abstracting the logic of
conditionals. Journal of Mathematical Behavior, 66, 100954.
Easley, J. A. (1964). Comments on the INRC group. Journal of Research in Science Teaching, 2,
233–235.
Greer, B. (2011). Inversion in mathematical thinking and learning. Educational Studies in

Mathematics, 79, 429–438.
sion of the splitting hypothesis. The Journal of Mathematical Behavior, 26, 27–47.
Cognition and Instruction, 28(4), 383–432.
Harel, G. (2008). What is mathematics? A pedagogical answer to a philosophical question. In
R. B. Gold & R. Simons (Eds.), Current issues in the philosophy of mathematics from the per-
spective of mathematicians. Mathematical American Association.
Inhelder, B., & Piaget, J. (1969). The early growth of logic in the child (E. A. Lunzer & E. Papert,
Trans.). Norton.
Kant, I. (1998). Critique of pure reason (Trans. P. Guyer & A. Wood). Cambridge University Press.
(Original work published in 1781).
Klein, F. (1893). A comparative review of recent researches in geometry. (Trans. M. W. Haskell).
Bulletin of the American Mathematical Society, 2(10), 215–249.
Kleiner, I. (1986). The evolution of group theory: A brief survey. Mathematics Magazine, 59(4),
195–215.
Mancosu, P. (1998). From Brouwer to Hilbert: The debate on the foundations of mathematics in
the 1920s. Oxford University Press.
National Research Council. (2009). Mathematics learning in early childhood: Paths toward excel-
lence and equity. National Academy Press.
Norton, A. (2016). (Ir)reversibility in mathematics. Proceedings of the thirty-eighth annual meet-
ing of the north American chapter of the International Group for the Psychology of mathemat-
ics education. University of Arizona.
Norton, A., & Wilkins, J. L. (2010). Students’ partitive reasoning. The Journal of Mathematical
Behavior, 29(4), 181–194.
Education, 43(5), 557–583.
Olive, J., & Steffe, L. P. (2002). The construction of an iterative fractional scheme: The case of joe.
Journal of Mathematical Behavior, 20, 413–437.
Pascal, B. (1966). Pensées (A. J. Krailsheimer, Trans.). Penguin (Original work published in 1670).
Piaget, J. (1969). The child’s conception of time (A. J. Pomerans, Trans.). Ballantine (Original
work published 1946).
Piaget, J. (1970). Structuralism (C. Maschler, Trans.). Basic Books (Original work published 1968).
Piaget, J. (1971). Genetic epistemology (E. Duckworth, Trans.). Norton.
Piaget, J. (1972a). Essai de Logique Opératoire (2nd ed.). Dunod. (Original work published
in 1949).
Piaget, J. (1972b). Principles of genetic epistemology (W. Mays, Trans.). Routledge & Kegan Paul
Piaget, J. (1973). The child and reality: Problems of genetic psychology. (A. Rosin, Trans.).
Grossman Publishers.
Piaget, J. (1985). The equilibration of cognitive structures: The central problem of intellectual
development (T. Brown & K. J. Thampy, Trans.). University of Chicago Press (Original work
published in 1975).
Piaget, J. (2001). The psychology of intelligence. London: Routledge (M. Piercy & D. E. Beryne,
Trans.). Routledge Classics (Original work published 1947).
Piaget, J. & Inhelder, B. (1967). The child’s conception of space (F.J. Langdon & J.L. Lunzer,
Trans.). Norton (Original work published in 1948).
Piaget, J., & Szeminska, A. (1952). The child’s conception of number. Routledge and Kegan Paul.
Poincaré, H. (1952). Science and hypothesis. Dover (Original work published in 1905).
Resnik, M. D. (1981). Mathematics as a science of patterns: Ontology and reference. Nous,
529–550.
238 A. Norton
Rowe, D. E., & Koreuber, M. (2020). Proving it her way: Emmy Noether, a life in mathematics.
Springer.
Siegler, R. S., & Richards, D. D. (1979). Development of time, speed, and distance concepts.
Developmental Psychology, 15(3), 288.
Silver, E. A., Leung, S. S., & Cai, J. (1995). Generating multiple solutions for a problem: A com-
parison of the responses of US and Japanese students. Educational Studies in Mathematics,
28(1), 35–54.
Simon, M. A., Kara, M., Placa, N., & Sandir, H. (2016). Categorizing and promoting reversibility
of mathematical concepts. Educational Studies in Mathematics, 93(2), 137–153.
Steen, L. A. (1988). The science of patterns. Science, 240(4852), 611–616.
Individual Differences, 4(3), 259–309. https://doi.org/10.1016/1041-6080(92)90005-Y
Steffe, L. P. (2002). A new hypothesis concerning children’s fractional knowledge. The Journal of
Mathematical Behavior, 20(3), 267–307.
Steffe, L. P., & D’Ambrosio, B. S. (1995). Toward a working model of constructivist teaching: A
reaction to Simon. Journal for Research in Mathematics Education, 26, 146–159.
Tall, D., Thomas, M., Davis, G., Gray, E., & Simpson, A. (2000). What is the object of the encap-
sulation of a process? The Journal of Mathematical Behavior, 18(2), 223–241.
Tzur, R. (1999). An integrated study of children’s construction of improper fractions and the teach-
er’s role in promoting that learning. Journal for Research in Mathematics Education, 30(4),
390–416.
Ulrich, C. L. (2012). Additive relationships and signed quantities. Doctoral dissertation, University
of Georgia).
von Glasersfeld, E. (1995). Radical constructivism: A way of knowing and learning.
Routledge Falmer.
von Glasersfeld, E., & Steffe, L. P. (1991). Conceptual models in educational research and prac-
tice. The Journal of Educational Thought, 25(2), 91–102.
Wessman-Enzinger, N. M. (2019). Integers as directed quantities. In A. Norton & M. Alibali
(Eds.), Constructing number (pp. 279–305). Cham.
Wilkins, J. L. M., & Norton, A. (2011). The splitting loope. Journal for Research in Mathematics
Education, 42(4), 386–406.
Wittmann, E. (1973). The concept of grouping in Jean Piaget’s psychology—Formalization and
applications. Educational Studies in Mathematics, 5, 125–146.
Chapter 8
Michael A. Tallman and Alan E. O’Bryan
Introduction
Piaget’s genetic (developmental) epistemology (Piaget, 1970, 1977) is a theory

addressing the formation and meaning of knowledge. Piaget explained, “Genetic epis-
temology attempts to explain knowledge and in particular scientific knowledge, on the
basis of its history, its sociogenesis, and especially the psychological origins of the
notions and operations upon which it is based” (1970, p. 1). Piaget’s theory arose out
of his concern to develop a viable model of how individuals construct stable realities
from the flow of their subjective experience (von Glasersfeld, 1995, p. 57).
Piaget focused his nearly six decades of psychological research on elaborating
the general mechanisms of cognitive development—that is, a theory of how a cog-
nizing subject comes to know their world and construct increasingly coherent cog-
nitive structures that are viable with their experiential reality. Piaget ultimately
aspired to address fundamental questions in epistemology by employing the scien-
tific, rather than philosophical, method. His self-identification as a genetic episte-
mologist reflected his expectation that conducting scientific analyses on the
psychological processes of children (including but not limited to the child’s concep-
tion of language, judgment, morality, causality, number, quantity, movement, speed,
time, space, geometry, and chance) might provide insight into the nature and devel-
opment (i.e., “genesis”) of knowledge structures generally. While Piaget’s approach
was multidisciplinary—making a strategic use of concepts in logic, mathematics,
M. A. Tallman (*)
Department of Mathematics, Oklahoma State University, Stillwater, OK, USA
e-mail: michael.tallman@okstate.edu
A. E. O’Bryan
School of Mathematical and Statistical Sciences, Arizona State University, Tempe, AZ, USA
e-mail: obryan_44@yahoo.com
https://doi.org/10.1007/978-3-031-47386-9_8
240 M. A. Tallman and A. E. O’Bryan
psychology, and the natural sciences—the core of his theory originated from the
parallelisms he drew between biological and psychological processes, namely, the
foundational principles of adaptation and organization (Gallagher & Reid, 2002).
In contrast to the empiricist notion that knowledge results from a simple register-
ing of experience, and the innatist or nativist position that knowledge derives from
nervous system maturation, Piaget’s biological perspective of cognitive develop-
ment emphasized the interaction between the knowing subject and their environ-
ment.1 Cognition, according to Piaget, is an instrument of adaptation instead of the
source of objective knowledge; cognitive structures evolve to enable individuals to
establish a fit, or equilibrium, between their conceptual model of reality and the
reality they experience (von Glasersfeld, 1995, p. 59). Accordingly, central to
Piaget’s theory and consistent with its biological underpinnings is the action of the
subject, defined broadly to encompass all movement, thought, or emotion that
responds to a need (Piaget, 1967, p. 6). Genetic epistemology, being fundamentally
interactionist, is based on the premise that cognizing subjects create, organize, and
act within their experiential world. The whole of Piaget’s theory aligns with the
premise that reality is a subjective construction; an individual’s experience of “real-
ity” is necessarily a product of neural processing unique to the individual, which
cannot be transcended to enable an experience of reality as it “really is.” Damasio
(1994) summarized this point cogently:
our very organism rather than some absolute external reality is used as the ground reference
for the constructions we make of the world around us and for the construction of the ever-
present sense of subjectivity that is part and parcel of our experiences. (p. xvi)
An individual’s knowledge of the world is “correct” not because it resembles exter-

nal reality, but because the individual has no reason to believe it doesn’t. We expect,
for example, that our perception of reality would be substantially different had the
human eye evolved to perceive a wider range of the electromagnetic spectrum or if
our ears were capable of discerning audio frequencies outside the 20 to 20,000 Hz
range. How different might the world seem if we could “see” x-rays and gamma
rays or if we could “hear” infrasonic waves? A knowing subject’s experience of
reality is necessarily constrained by both the physical limitations of their sensory
organs and idiosyncrasies of the neural system through which one becomes cogni-
zant of sensory data. These biological suppositions justify the two foundational
premises of radical constructivism, inferred and articulated by Ernst von Glasersfeld
based largely on Piaget’s work:
1. Knowledge is not passively received either through the senses or by way of com-
munication; knowledge is actively built up by the cognizing subject.
2. The function of cognition is adaptive, in the biological sense of the term, tending
towards fit or viability; cognition serves the subject’s organization of the experi-
ential world, not the discovery of an objective ontological reality. (von
Glasersfeld, 1995, p. 51)
1
Because Piaget neither considered cognitive development a process determined by environmental
pressures nor biological maturation, his theory is often referred to as a middle-ground position,
third possibility, or tertium quid (Gallagher & Reid, 2002, p. 23).
8 Reflected Abstraction 241
The principle that knowledge derives from a subject’s interaction with their envi-
ronment does not explain how the subject constructs knowledge. For this, Piaget
elaborated the concept of equilibration, the mechanisms of which are assimilation
and accommodation. Briefly defined, equilibration is the self-regulatory process by
which an individual actively compensates for external disturbances (Piaget &
Inhelder, 1969); assimilation is the process whereby a subject incorporates experi-
ences into existing cognitive structures, and thus consists of the meanings the sub-
ject holds, and accommodation entails the modification of an individual’s cognitive
schemes to enable their assimilation of novel experiences. We have noticed several
mathematics education researchers conducting their work from a radical construc-
tivist or Piagetian perspective invoking assimilation, accommodation, and equili-
bration (and perhaps perturbation) to describe students’ learning or activity. While
precise descriptions are certainly of value, it is crucial to recognize that this triad of
theoretical constructs has limited explanatory potential by itself without specifying
the cognitive mechanisms entailed.
A deeper analysis of knowing subjects’ cognitive activity is afforded by the
instruments of assimilation and accommodation that Piaget defined. Assimilation
and accommodation, and thus equilibration, rely heavily on the notion of abstrac-
tion, of which Piaget distinguished five varieties: empirical, pseudo-empirical,
reflecting, reflected, and metareflection.2 Piaget explained that higher forms of
knowledge derive from abstractions of the subject’s actions and the results of apply-
ing them in specific contexts. As Piaget’s scholarship progressed, he developed a
deeper appreciation and understanding of the utility of equilibration and abstraction
for providing a functional explanation for the construction of knowledge—so much
so that by the end of his academic career, these two constructs were at the center of
his genetic epistemology.
The primary focus of our chapter is to describe reflected abstraction and discuss
its implications for empirical research into the teaching and learning of mathemat-
ics. For contrast, we begin by briefly defining and illustrating the more basic forms
of abstraction that Piaget distinguished.3 We then offer a discussion of reflected
abstraction and describe its relation to relevant theoretical constructs within Piaget’s
genetic epistemology, and to the mathematics education research literature gener-
ally. With this background established, we outline various implications of reflected
abstraction for students’ learning of mathematics as well as teachers’ construction
of pedagogical content knowledge (Shulman, 1986). We then consider strategies for
promoting reflected abstractions and propose explicit standards of evidence required
to support valid claims about whether an individual has constructed schemes at the
level of reflected thought.4 We conclude by outlining pertinent questions that
mathematics education researchers might pursue regarding either the necessary
2
We note that pseudo-empirical, reflecting, reflected, and metareflection are non-mutually exclu-
sive forms of abstraction.
3
Ellis, Lockwood, and Paoletti (this volume) provide thorough descriptions of empirical, pseudo-
empirical, and reflecting abstraction.
4
A reflected abstraction is a scheme constructed at the level of reflected thought.
conditions for promoting reflected abstractions or the implications of reflected

schemes for students’ mathematical activity and teachers’ instructional practices.
Some readers might consider parts of this chapter repetitive with other contribu-
tions in this volume, or might regard this chapter as being more about genetic epis-
temology in general, with reflected abstraction positioned as one among its many
important constructs. As a dialectical thinker, Piaget developed an empirically
grounded epistemology consisting of an elaborate network of interconnected ideas.
We find it impossible to discuss only one concept without considering its relation to
other elements of Piaget’s theory. Ultimately, we feel that understanding the signifi-
cance of reflected abstraction requires comprehending its connection to other ele-
ments of the broader theory and how it contributed to Piaget’s pursuit of scientific
answers to the central questions of epistemology.
Piagetian Abstraction
Empirical Abstraction
Empirical abstractions result from extracting properties from any source considered
external by the subject, and thus are the source of exogenous knowledge. In empiri-
cal abstraction, the subject constructs knowledge that originates in their perception
of exogenic objects (objects other than the self; see Fig. 8.1). Crucially, the charac-
teristics the subject extracts are not from the object but rather from the subject’s
perception of the object. External entities only have properties in the context of an
individual’s experience of them; such properties are not inherent to the objects
themselves. A passive subject is, therefore, unable to extract features of objects, and
the properties the subject is capable of observing are constrained by the cognitive
structures that control their perceptive activity. Since the information one is able to
abstract from objects is restricted by their own subjective ways of perceiving,
empirical abstractions rely on the cognitive structures constructed and refined
through previous reflecting abstractions.5
To illustrate empirical abstraction in a mathematical context, consider a student
with no formal experience with angle measure attempting to solve the task
Fig. 8.1 Empirical

abstraction
5
See the discussion of reflecting abstraction later in this chapter.
Fig. 8.2 Task that supports empirical abstraction

Note. This task is intentionally imprecise to allow an individual with limited knowledge of
angle measure to assimilate and respond to it
displayed in Fig. 8.2. Although the student has no conceptual knowledge of what it
means to measure an angle, the student would likely be able to respond sensibly, if
not correctly, to this (admittedly vague) task. The student might examine the three
geometric objects—which they recognize as existing independently of their thought
and action—and abstract “openness” as a property of these angles so that it is pos-
sible to order their “size” according to this characteristic. Upon doing so, the student
might recognize that Angle A has the smallest amount of “openness,” while Angle B
has the largest amount of “openness.” It is noteworthy that, via empirical abstrac-
tion, this task is accessible even without the student clarifying what it means to
measure “openness.” Completing the task as described exemplifies an empirical
abstraction since the student perceives the property of “openness” as a feature of
two rays extending from a common endpoint that exists independent of their actions.
Abstracting “openness” does not require a sophisticated scheme for angle measure,
and, in fact, the student’s conception of “openness” as an attribute vaguely associ-
ated with the space between the angle’s rays is certainly inexact. Many students, in
fact, are unlikely to progress beyond this level of abstraction for the concept of
angle measure if the curricula and instruction they experience do not support con-
ceptions based in quantitative reasoning (Moore, 2010, 2014; Moore et al., 2016;
Tallman & Frank, 2020; Thompson, 2008). Students instructed to measure angles
by laying down protractors and identifying the number to which the terminal ray
points are following a dictated procedure that they incorporate into their empirical
abstraction of the attribute being quantified—larger protractor readings correspond
to angles with “more openness.” Students operating with such indexical meanings
do not conceptualize angle measures as quantifications of a particular measurable
attribute.
Indications of empirical abstraction related to angle measure may include the
following common behaviors: (1) when asked how to measure an angle, a student
responds “with degrees” or “with radians,” yet cannot explain aspects of this mea-
surement system, such as why differently sized protractors can be used to produce
the same angle measure; (2) they express confusion when confronted with congru-
ent angles that have their rays drawn vastly different lengths (or if one of the two
rays is drawn much longer than the other); (3) they associate an angle’s measure
with the geometric area bounded by two rays extending from a common endpoint;
(4) they struggle to identify a precise attribute for which radians and degrees are
appropriate units of measure; or (5) they explain that radian measures are given in
multiples of π, while measures in degrees do not include π. These behaviors reflect
the limitations of meanings abstracted from exogenic objects alone. Empirical
abstractions, having their origin in observable properties of objects, are the basis of
figurative schemes—a topic to which we later return.
Pseudo-Empirical Abstraction
Pseudo-empirical abstractions result from an individual abstracting properties of

objects that have been modified by or created through their actions (Piaget, 1977).
In pseudo-empirical abstraction, the individual neither disassociates the action from
the object nor the object from the action; the action and abstracted property of the
object remain undifferentiated. What the subject abstracts, then, is not simply a
property of the object, as in empirical abstraction, nor the action that produces or
modifies the object, as in reflecting abstraction (described in the next section), but a
property of the object that results from and, crucially, represents (in the sense of
signifying) the action of the subject (see Fig. 8.3). That is, in pseudo-empirical
abstraction, the subject conceives the properties of an object as the outcome of hav-
ing acted upon it and as encapsulating these actions so that the properties the subject
abstracts capture both the action and the result of acting. In this way, pseudo-
empirical abstraction relies on what Piaget called the semiotic function, or the sub-
ject’s capacity to re-present their experience (described in more detail in a later
section). The individual, when engaged in pseudo-empirical abstraction, considers
the object, or a momentary property of it, as a symbol6 that captures both the object
and action that produced or modified it so that the individual can then operate on
these symbols without having to engage in the actions involved in constructing or
transforming the object.
Fig. 8.3 Pseudo-empirical

abstraction
6
We use symbol to refer to any object that evokes in an individual an “abstracted representation” of
experience and associated meanings (von Glasersfeld, 1995, p. 99). The object may be a stimulus
in one’s sensory field (i.e., exogeneous) or a representation of experience (i.e., endogenous).
Fig. 8.4 Task that supports pseudo-empirical abstraction

Note. Each red dot is equidistant from the vertex of the respective angle. The student is
expected to move the red dot by using Geometer’s Sketchpad (Jackiw, 2011), a geometric visu-
alization program
Fig. 8.5 Example of pseudo-empirical abstraction
To illustrate pseudo-empirical abstraction, consider again a hypothetical student

engaged with the activity stated in Fig. 8.4, which is an extension of the task pre-
sented in Fig. 8.2. In this activity, the student interacts with a virtual manipulative
that allows them to grab a point on the terminal ray and move it along a circu-
lar path.
In response to the task, the student may have produced something like the angles
displayed in Fig. 8.5. After having rotated the terminal ray of each angle by drag-
ging the red dots, the student may abstract the property that each red dot traces out
an arc and that larger angles are created when a longer arc is produced (assuming
equal distances between the angles’ vertices and the rotated point). These abstrac-
tions are pseudo-empirical because the arc is a property that emerges only from the
individual’s action, and thus represents having rotated the terminal ray. As a pseudo-
empirical abstraction, the student is not (yet) positioned to conceptualize this prop-
erty as existing independently of the action that produced the subtended arcs.
Moreover, since the student has not (yet) differentiated the arcs from the action that
created them, the student is not (yet) prepared to understand angle measure as an
equivalence class of subtended arc lengths measured in some unit proportional to
the circumference of the circle containing any particular subtended arc. The student
is also not (yet) equipped to think about any angle as having been produced through
such a rotation without performing (either in action or thought) the rotation for each
new angle.
Crucially, after having constructed this pseudo-empirical abstraction, the subject
recognizes that the abstracted properties result from and represent the sequence of
actions through which they were created or modified. In other words, the subject
does not conceive the abstracted property as existing independently of the action
they engaged in to construct or transform it. The failure to dissociate the property of
the object that the individual abstracts from the actions they engaged in to create or
modify it is the distinguishing feature of pseudo-empirical abstraction. We stress
that pseudo-empirical abstractions need not always be correct or developmentally
productive.7 Their potential value for students’ learning is determined by the extent
to which the differentiation of an action on an object from the property it produces
might enable students’ construction of an operative scheme to which a class of
objects can be assimilated with immediate implications for students’ subsequent
activity.
Reflecting Abstraction
Reference to abstraction reflechissante (reflecting abstraction) did not appear in

Piaget’s writings until 1950, nearly 30 years into his psychological research pro-
gram. The introduction of this construct was motivated by Piaget’s efforts to convey
the limitations of empirical abstraction in particular, and empiricist theories of the
origins of mathematical concepts generally (Piaget, 2001, p. 9). In contrast to
empirical and pseudo-empirical abstraction, reflecting abstraction involves the sub-
ject’s reconstruction on a higher cognitive level of the coordination of actions from
a lower level, and results in the development of logico-mathematical knowledge, or
schemes at the level of operative thought (Chapman, 1988; Piaget, 1971). Reflecting
abstraction is thus an abstraction of actions and occurs in three phases: (1) the dif-
ferentiation8 of an action from the effect of the action, (2) the projection of the
action from the level of material action to the level of representation9, and (3) the
reorganization that occurs on the level of representation of the action projected
from the level of material action (see Fig. 8.6).
Actions must be differentiated from their effects before a subject can project
these actions to the level of representation or the level of thought (this is the
7
Consider, for example, the common abstraction that a product is always greater than its factors
(i.e., “multiplication makes bigger”).
8
Piaget encapsulates differentiation within the projective aspect of reflecting abstraction. We make
differentiation explicit because it is one of the key distinctions between reflecting abstraction and
pseudo-empirical abstraction.
9
Throughout this chapter, we use “representation” in a manner consistent with von Glasersfeld’s
(1995) interpretation of how Piaget applied the term: “For Piaget, representation is always the
replay, or re-construction from memory, of a past experience and not a picture of something else,
let alone a picture of the real world” (p. 59).
Fig. 8.6 Reflecting abstraction
conceptual domain of mental re-presentations of actions). Additionally, the subject

must coordinate the actions that produced the effect before they can be projected
and represented on this higher cognitive level. Once a subject differentiates actions
from their effect and coordinates them, the subject can project these coordinated
actions to the level of representation where they are organized into cognitive struc-
tures, or schemes, of internalized actions and operations.
To exemplify reflecting abstraction, imagine that the student we discussed respond-
ing to the tasks in Figs. 8.2 and 8.4 had continued to think about creating angles and
comparing their sizes, as prompted by the task displayed in Fig. 8.7. This task promotes
a reflecting abstraction by providing an opportunity for the student to differentiate the
action of rotating a point on the terminal ray from its effect of producing an angle sub-
tending an arc. By prompting the student to justify their ordering, the task encourages
the student to reflect on the physical actions they engaged in while solving the task in
Fig. 8.4, namely, moving the red dot to create angles with a particular amount of “open-
ness” and to generally associate an angle’s measure with a subtended arc length.
The student might approach the task by examining each of the angles in Fig. 8.7
and imagine having moved a point on the terminal ray (equidistant to the vertex) to
create the respective angle. If asked to compare the size of the angles in Fig. 8.7, the
student can now explain their choices by comparing the arc lengths produced as
they imagine the angle being formed by rotating the terminal ray.10 The student’s
ability to do this suggests that a reflecting abstraction has occurred. The student’s
conception of the six angles as the result of having rotated a point on the terminal
ray demonstrates their capacity to dissociate this action from the effect it produces
10
For this to be an appropriate method for comparing the “size” (i.e., non-quantitative comparison)
of the angles, the student must recognize that the selected point on the terminal must be the same
distance away from the vertex of each angle—a fact that the student had the opportunity to pseudo-
empirically abstract while engaged in the activity displayed in Figure 8.4.
Fig. 8.7 Task that supports reflecting abstraction
(an angle with a particular amount of openness that subtends an arc generated by
tracing a point on the terminal ray as it rotates). That the student can perform men-
tally an action that they engaged in physically on a previous task (see Fig. 8.4) sug-
gests that the student has projected this action to the level of thought. In other words,
the student has constructed a mental representation of the physical action of moving
a red dot on the terminal ray of an angle to create an angle with a specific amount of
“openness.” This mental representation of coordinated physical actions distin-
guishes reflecting abstraction from the more basic abstractions previously discussed.
As with pseudo-empirical abstractions, the cognitive structures that result from
reflecting abstractions might not be “correct” from the perspective of an expert. This
is particularly plausible in the realm of mathematics, where the objects on which
one acts are mental constructions. The mathematical environment, therefore, does
not typically “push back” the same way as the physical world. Moreover, without
appropriate instructional and curricular support, learners do not always engage in
critical thinking and self-regulatory behaviors to provide their own feedback on the
products and processes of their mathematical reasoning.
What Is Reflected Abstraction?
When an individual becomes cognizant of the coordinated actions projected to the

level of representation through reflecting abstraction, we say that the resulting
abstraction becomes reflected, or that the organization of internalized actions exists
at the reflected level of thought (i.e., the “plane of thematization” (Piaget, 2001,
p. 51) where projected actions become the objects of mental operations). The
reflected level is a higher cognitive plane to which one’s reasoning (as opposed to
actions, as in the case of reflecting abstraction) is projected. Piaget’s distinction
between reflecting and reflected abstraction suggests that conscious knowledge is
not a byproduct of reflecting abstraction alone. Whereas empirical, pseudo-
empirical, and reflecting abstractions are all constructive processes, reflected
abstraction is a product of reflecting on the projected actions from previous11 reflect-
ing abstractions, which results in the knower’s awareness of these internalized
actions. Piaget (2001) stated the distinction between these related forms of abstrac-
tion clearly, though not always consistently: “Reflecting abstractions constitute pro-
cesses, and reflected abstractions constitute becoming conscious of their results”
(p. 205); “we call the result of reflecting abstraction reflected when it has become
conscious, regardless of its developmental level” (p. 303); “reflected abstractions
are conceptualized objects of representation or consciousness” (p. 67); and “the
‘reflecting’ process does not immediately lead to a ‘reflected’ result” (p. 205).
Consistent with the description of reflected abstraction as a conceptual object—a
type of developmental destination—Piaget (2001) explained, “There is reflecting
abstraction as a constructive process, and there is its retroactive thematization,
which then becomes a reflection on the reflection. In these cases, we will speak of
‘reflected abstraction’ or ‘reflective thought’” (p. 31). “Retroactive thematization”
is the process of an action becoming an object of thought—not just an instrument of
transformation—that results in the knower’s cognizance of projected action coordi-
nations. Thus, reflected abstractions enable an individual to explicitly formulate the
results of prior reflecting abstractions. Reflected abstraction, then, describes a
scheme constructed at the reflected level of thought through reflection on the results
of prior reflecting abstractions. These reflected schemes possess an essential char-
acteristic (cognizance of projected action coordinations, decontextualized and
applicable to a generalized class of objects) that endows the knower with the capac-
ity to explicitly formulate meanings and strategically apply them in a range of novel
contexts. Indeed, Piaget (2001) most often associated reflected abstraction with
behavioral expressions of such capacities. For example, in describing reflected
abstraction as a “detailed description of characteristic actions” (ibid., p. 167), Piaget
associated reflected schemes with the aptitude to explicitly formulate meanings
entailed and ways of reasoning supported by the scheme.
To clarify the nature of reflected abstraction, consider the following example
from Piaget (2001). The experimental context involved the interviewer constructing
a “wall” of blue five-unit blocks and the child constructing a “wall” of red one-unit
blocks. The child responded to several tasks including but not limited to (1) deter-
mining the number of blocks contained in a red wall equal in length to the inter-
viewer’s blue wall, (2) comparing the length of red and blue walls each composed
of an equal number of blocks, and (3) comparing the length of red and blue walls
11
Regarding the décalage, or temporal delay, between reflecting and reflected abstraction, Piaget
(2001) explained, “The subject becomes conscious of the result of his acts—which requires a
simple static read-off—before becoming conscious of their mechanism and their exact unfold-
ing—which requires the reconstruction of a process” (p. 191).
constructed from an indefinite number of simultaneous additions of red and blue

blocks. We highlight Piaget’s interpretation of the conceptual demands of a subse-
quent task prompting the child to draw a red line representing the length of a red
wall containing the same number of blocks as a blue wall, not actually constructed
but represented by a blue line the interviewer had drawn. Piaget (2001) explained
that to correctly respond to this task:
a superior form of abstraction is needed: a reflected abstraction, which we might also call a
reflexive abstraction, because it works by reflection on particular reflections. Reflected
abstraction would reunite the additions into an addition of additions; i.e., a multiplication
B − R = N(b − r) or even a proportion B/R = b/r. (pp. 42–43)12
Success on the task requires the child to conceive the blue line as a representation of
an indefinite number (N) of blue blocks, each of which is five times as long as a red
block. Based on this conception, the child can reason additively by imagining iterat-
ing the difference between the length of a blue block (b) and the length of a red
block (r) an indefinite number of times determined by the number of blocks (N)
contained in the blue wall, thus constructing “an addition of additions; i.e., a multi-
plication B − R = N(b − r)” (ibid., p. 43). This construction entails re-presenting
actions carried out in response to the previous tasks and performing operations
(iterations of additive comparisons) on these mental re-presentations. In this con-
text, we interpret Piaget’s use of “reflection on particular reflections” to refer to the
child’s engagement in these conceptual operations, which necessitates the child’s
awareness of the actions projected to the level of representation through prior
reflecting abstractions. In other words, for the child to envision an iteration of addi-
tive comparisons of the length of blue and red blocks, their prior additive compari-
son of a reference pair of one blue and red block must be “differentiated by projection
onto a thematized plane” where it may become an object of thought (Piaget, 2001,
p. 51). Abstracting an invariant multiplicative relationship, or a proportion, between
the length of the blue wall (B) and the length of the red wall (R), each comprised of
an indefinite but equivalent number of blocks similarly requires performing mental
operations on the products of previous reflecting abstractions projected to the
reflected level of thought. In essence, the task necessitates a reflected abstraction
since a successful response involves the child in “reconstituting the reasoning they
engaged in to solve the problems they were given, or at representing this reasoning
to themselves” (Piaget, 2001, p. 107).
Piaget’s interpretation of the conceptual demands of this task reveals that
reflected abstraction involves more than cognizance of the differentiated actions a
subject projects to the level of representation. Reflected abstraction penetrates the
structure of the scheme itself, affecting its organization and internal coherence.
Mental operations performed on the products of reflecting abstractions enable gen-
eralization and contribute to an “overarching integration” of the reflected scheme
(Piaget, 2001, p. 197).
12
B and R, respectively, represent the length of the blue and red walls; b represents the length of a
blue block and r represents the length of a red block.
Supporting Retroactive Thematization
Reflected abstractions are evinced by an individual’s capacity to explicitly formulate

meanings enabled by assimilation to a scheme (i.e., organization of internalized
actions projected to the level of representation through reflecting abstraction). To
assess whether an individual has engaged in the reflective activity that results in their
becoming conscious of the products of prior reflecting abstractions, Piaget and his
collaborators often prompted interview subjects to make explicit comparisons between
two similar tasks to discern the extent to which subjects were capable of identifying
analogies between them. Piaget (2001) explained that these comparisons “require
reflected abstraction that teases out the common structure of the tasks” (p. 216).
Piaget tested for reflected abstraction by asking subjects to make explicit com-
parisons between similar tasks. But how might reflection on the products of prior
reflecting abstractions be supported? Is reflected abstraction only assessed by an
individual’s ability to identify action coordinations required by two analogous tasks,
or might such comparisons also engender the precise reflective activity that results
in retroactive thematization? Although the distinction between testing for reflected
abstraction and provoking cognizance of actions coordinations projected to the level
of thought through reflecting abstraction is often blurred in Piaget’s writings, we
find clarity in his description of “reflection on reflections” (i.e., reflected abstrac-
tion) as a comparison between objects: that is, as a mental operation on the products
of prior reflecting abstractions that results in a reflective reorganization of these
products (Piaget, 2001, p. 52). It is the act of deliberately operating on the products
of prior reflecting abstractions that brings the actions and operations organized
within cognitive structures into the subject’s conscious mind. To knowingly operate
on actions at the level of representation suggests that one has symbolized actions at
this higher cognitive level. Reflected abstraction, like pseudo-empirical abstraction,
thus relies on the semiotic function (described in a later section). The knowing sub-
ject symbolizes coordinated actions at the level of representation to reify the mate-
rial actions the symbol represents into a form they can use as an object of thought at
the reflected level. On this thematized level of thought, the subject can consciously
manipulate these symbols (i.e., generalized mental images) without having to re-
enact the coordinated actions they signify. The application of the semiotic function
is thus the mechanism by which retroactive thematization occurs. Performing oper-
ations on the symbols the subject constructs to represent coordinated actions at the
level of representation results in increasingly organized and refined cognitive struc-
tures to which one can purposefully assimilate novel experiences. We demonstrate
the purposeful assimilation afforded by reflected schemes in the following example.
Example of Applying a Reflected Scheme
Assessing the validity of unconventional tools for measuring angles, such as

the funky protractors introduced by Hardison and Lee (2020), is a task that
requires reflected abstraction. Consider the hypothetical student discussed
previously tasked with justifying which of the funky protractors displayed in

Fig. 8.8 are valid tools for measuring angles and abstracting the general fea-
tures of an unconventional protractor that make it a valid tool for measur-
ing angles.
After having engaged in the empirical, pseudo-empirical, and reflecting
abstractions supported by the tasks, respectively, displayed in Figs. 8.2, 8.4,
and 8.7, the student is positioned to assess the validity of the four funky pro-
tractors in Fig. 8.8 by considering whether equal changes in the measures indi-
cated on each protractor correspond to uniform amounts of rotation of the
angle’s terminal ray, which may be represented by successive arc lengths traced
out by an arbitrary point on the rotating ray. While considering Protractor A,
for example, the student might construct a mental image like the one displayed
in Fig. 8.9. Specifically, the student might envision a sequence of terminal rays
that intersect the indices of the protractor and then assess whether the arcs (of
equal radii) subtended by adjacent rays are equal in length. Indeed, after ana-
lyzing Protractors B, C, and D, the student may abstract this property as the
essential characteristic of any valid instrument of angle measure. This is an
example of a reflected abstraction because it relies upon the product of the
previous reflecting abstraction (i.e., conceiving the length of a subtended arc as
a quantity whose measure is in direct correspondence with an angle’s “open-
ness”) being an object of thought.
Fig. 8.8 Task that requires reflected abstraction. (Hardison & Lee, 2020)
Note. The location of the angle’s vertex is indicated by the large dot on each funky protractor
Fig. 8.9 Engaging in reflected abstraction to assess the validity of funky protractor
elation of Reflected Abstraction to Other

R
Piagetian Constructs
The Semiotic Function and Representational Thought
We previously stressed that reflected abstraction relies on the semiotic function.

Fully understanding this dependence requires a discussion of the origins of the
semiotic function and its contribution to representational thought.
Regarding reflection, von Glasersfeld (1995) endorsed the following assertion by
von Humboldt: “In order to reflect, the mind must stand still, for a moment in its
progressive activity, must grasp as a unit what was just presented, and thus posit it
as object against itself” (von Humboldt, 1907, p. 581; quoted in von Glasersfeld,
1995, p. 90). Central to von Glasersfeld’s conceptualization of reflection is the con-
cept of re-presentation. Thus, “to posit an experience against itself” involves remov-
ing the experience from its temporal placement to re-present it as a closed entity
(von Glasersfeld, 1995, p. 91). For von Glasersfeld, re-presentation (and this is the
reason for the hyphen) is a kind of re-construction of a prior sensory experience
(from the re-presenter’s perspective),13 not a copy of some aspect of objective real-
ity. Representational thought, therefore, involves performing mental actions and
operations on reified re-presentations of experience (i.e., images—see Thompson,
Byerley, and O’Bryan, this volume).
von Glasersfeld (1995) argued that the abstractions of general ideas from sensory
experience serve a classificatory function:
13
Piaget and Duckworth’s (1968) book On the Development of Memory and Identity necessitates
this qualification.
I suggest that we are quite able to abstract general ideas from experience. We do this by
substituting a kind of place-holder or variable for some of the properties in the compound
sensory structures we actively build up to form particular things from the flow of experi-
ence. … [This is what] enables us to recognize items that we have never seen before, as
exemplars of a familiar kind. (p. 93)
Logical operations are performed on the signifiers (“place-holder or variable” in

von Glasersfeld’s usage), the individual constructs to represent14 objects, actions,
and experiences. Prior to the construction of these signifiers, the only objects of a
knowing subject’s thought are those within their immediate sensory field. The
acquisition of the semiotic function, therefore, is characteristic of the child’s pro-
gression from the sensorimotor stage of development to the preoperational stage.
The ability to mentally represent objects and events liberates the child’s thought
from its reliance on sensorimotor action and facilitates the child’s ability “to locate
himself in a space that goes beyond perceptual space and in a field of time that per-
mits the coordination of past, present, and future” (Gruber & Vonèche, 1977,
p. 485). The affordances of representational thought, and the acquisition of the
semiotic function on which it depends, suggest that the child’s capacity to construct
mental representations of objects and events not perceived constitutes, in Piaget’s
words, “the most decisive turning point in the mental development of the child”
(ibid., p. 508) because of its potential to create objects of thought detached from
concrete supports and ultimately to facilitate an individual’s dissociation of the form
of their thought from its content—an essential feature of reflected abstraction.
So how does a knowing subject develop ways to represent the world through
signs and symbols15? Piaget proffers imitation as central to the child’s acquisition of
the semiotic function. Imitation, as a type of representation in action, indicates a
transitional phase between the child’s sensorimotor activity and representational
thought. Piaget distinguished five behavior patterns that imply the representation in
thought of an object or event not immediately perceived, the first four of which are
based on imitation: (1) deferred imitation, (2) symbolic play, (3) drawing, (4) men-
tal image, and (5) verbal evocation. Deferred imitation marks the beginnings of
representational thought because it involves the child’s representation in action of
an event not present. In deferred imitation, the imitated action becomes a signifier
of an event the child does not directly perceive—as when the first author’s two-year-
old son, Everett, imitates his daycare teacher by gathering pen, paper, and tablet and
instructing his sister, mother, and father to gather by the back door so he can take
attendance and lead the “class” outside to play.
While deferred imitation serves the purpose of adapting the child to reality (i.e.,
deferred imitation is accommodative), symbolic play serves to assimilate reality to
the child. The child makes use of the semiotic function in symbolic play by creating
symbols in the form of gestures and objects that represent those aspects of the
14
Throughout this chapter, we use “represent” (without the hyphen) to refer to the action of reify-
ing a re-presentation of experience in the form of a symbol.
15
Piaget referred to symbols as “internalized imitations,” which are necessarily grounded in the
knowing subject’s actions (Ginsburg & Opper, 1988, p. 74).
child’s experience that cannot be assimilated using language alone (Gruber &
Vonèche, 1977, p. 494). Drawing is a transitional behavior pattern between sym-
bolic play and mental imagery. Unlike symbolic play, the signifier present in the
behavior of drawing is not an action but nevertheless derives from the child’s per-
ception. Mental images are the first form of behavior implying representational
thought that need not ensue from perception. Moreover, mental images are a conse-
quence of internalized imitations and are thus dissociated from sensorimotor activ-
ity. Verbal evocation, or language, is the only behavior indicative of representational
thought that is not imitative, and thus the signifiers of language bear no resemblance
to the things signified but are instead conventional and learned through socialization.
According to von Glasersfeld (1995), symbols, particularly language, enable
more elaborate representations. Symbols endow the individual with the means to
reify the meanings the symbol represents into a form that can be used as the object
of thought on a higher cognitive plane (i.e., the reflected level of thought). On this
higher plane, the individual can manipulate symbols independently of reproducing
their associated representation. This is what von Glasersfeld referred to as the
“pointing function of symbols” (p. 108). Symbols, as representing meaning, can be
taken as the content of thought without the knowing subject having to decompress
the entirety of the signified meaning. It is in this spirit that von Glasersfeld (1995)
understood the meaning of “reflection” in the context of reflective abstraction:
reflection “should be interpreted as projection and adjusted organization on another
operational level in the case of reflective abstraction” (p. 107).
von Glasersfeld (1995) recognized that abstraction, reflection, and representation
“interact on various levels of mental operating” (p. 109). On one level, representa-
tions serve to organize the content of sensorimotor experiences, considered exoge-
nous by the representer. These representations, facilitated by the use of the semiotic
function, become the content of operations on a higher conceptual plane. On this
higher level, symbols serve as pointers to the content of the representation without
necessitating the evocation of such content in order to treat the representation as
content in itself and thus to operate upon it. While the representation is a form that
serves to organize exogenous content at lower levels of mental operating, the know-
ing subject constructs it as exogenous content on a higher level and associates with
it a symbol.
Piaget’s assertion that knowledge is not a reproduction of reality suggests that
objects of thought are not copies of real objects or events but are signifiers con-
structed by the individual to represent real objects or events. The central problem
Piaget addressed in his pursuit of elaborating a theory of the origins of knowledge,
then, was “to describe the relationship between imaginal symbolism and the preop-
eratory or operatory mechanisms of thought” (Gruber & Vonèche, 1977, p. 499).
This focus is very much evident in the role of reflected abstraction in Piaget’s
genetic epistemology, as we discuss in the following subsections relating reflected
abstraction to schemes, equilibration, imagery, and figurative and operative modes
of thought.
Schemes16 and Equilibration
Piaget endorsed the basic assumption that the primary function of the human brain
is to bring meaning and coherence to an individual’s experiences, and it often does
so in ways that are beyond the individual’s conscious awareness. Schemes (what
they are and how they are created and modified) are “theoretical construct[s] that we
impute to individuals to explain their behavior” (Thompson et al., 2014, p. 9). They
are the basic cognitive units related to knowledge and what it means to understand
something. We rely on Thompson et al.’s (2014) description of a scheme as a start-
ing point for our discussion: “Schemes are organizations of mental activity that
express themselves in behavior, from which we, as observers, discern meanings and
ways of thinking” (p. 9).
Piaget generally defined a scheme as “whatever is repeatable and generalizable
in an action” (1970, p. 42). Specifically, Piaget described a scheme as “the structure
or organization of actions as they are transferred or generalized by repetition in
similar or analogous circumstances” (Piaget & Inhelder, 1969, p. 4). Schemes are,
therefore, cognitive entities that result from a knowing subject’s disposition to con-
struct meaning and make sense of their experiences. Schemes serve to organize the
individual’s reality and impose order on an ever-changing world by equipping the
individual with the conceptual tools to systematically act on their environment and
expect particular outcomes.
Assimilation to a scheme involves “treating new material as an instance of some-
thing known” (von Glasersfeld, 1995, p. 62, italics in original). What an individual
can assimilate is entirely dependent upon their current cognitive schemes. “In short,
assimilation always reduces new experiences to already existing sensorimotor or
conceptual structures” (von Glasersfeld, 1995, p. 63); what an individual already
knows determines what they may come to know.
If an individual has no reason to believe that their model of the world is inaccu-
rate, they will remain in a state of cognitive equilibrium, in which their conceptual
structures are viable with their experiential reality. Such a state is preserved so long
as the individual can assimilate new experiences to existing schemes. Often in the
flow of experience one’s perception of a particular situation induces them to employ
an action that produces an effect that violates the actor’s expectation. In this case,
the scheme governing the perception, action, and expectation is not viable with the
individual’s experiences. When a knowing subject does not possess an appropriate
assimilatory scheme to inform their actions when confronted with a novel stimulus
or experience, a state of cognitive perturbation or disequilibrium results. This state
can induce a need for the individual to accommodate their scheme to the stimulus or
experience by modifying the scheme, or constructing a new one, so that their con-
ceptual structures remain viable with their experiential reality.
Piaget’s concepts of functional and generalizing assimilation stress that indi-
viduals possess a tendency to exercise their schemes and extend the class of objects
16
Tillema (this volume) provides a thorough description of scheme from a Piagetian perspective.
to which they may be applied (Ginsburg & Opper, 1988, pp. 30–31). The energetic
propensities of functional and generalizing assimilation imply that assimilation, and
inevitable accommodation, occur naturally. Piaget regarded cognitive development
as a process that inclines toward a balance, or equilibrium, between assimilation and
accommodation. Hence, his most general description of cognitive development was
an account of the equilibration of cognitive structures, the complexity of which
depends on the type of structures involved. Equilibration also varies in detail
depending on whether structures are assimilating and accommodating the physical
environment or other cognitive structures possessed by the knowing subject. Basic
equilibration pertains to a subject’s interaction with physical stimuli; more refined
forms pertain to internal adjustments and integrations within the knowing subject.
But none of these forms of equilibration explain how a cognizing subject might
come to know about their own processes of knowing or about coordinations of their
actions. How might an individual build new cognitive structures that are about exist-
ing structures (as when one constructs formal operations, which Piaget regards as
operations on concrete operations)? That is, how does an individual construct
knowledge about their own knowledge? Addressing this question brings us back to
a point we made earlier in the chapter. We claimed that the triad of constructs assim-
ilation, accommodation, and equilibration is often applied in mathematics educa-
tion research to identify a subject’s behaviors; they alone do not provide a full
explanatory framework for how learning occurs. This is the role of abstraction in
genetic epistemology.
The previous discussion of the various forms of Piagetian abstraction reveals that
the nature of the cognitive schemes that characterize logico-mathematical knowl-
edge is an organization of internalized (mental) actions constructed at the level of
representation through the process of reflecting abstraction and refined through fur-
ther projection of reasoning to the reflected level of thought.17 Actions, as men-
tioned earlier, include “any thought, emotion, or movement that serves to satisfy a
need” (Piaget, 1967, p. 6), where the “need” refers to a goal not currently met.
Actions can exist at various levels, from sensorimotor reflexes to intelligent thought.
For example, a person may sneeze to relieve irritation (the need is relieving irrita-
tion and the action is sneezing), and this is a reflex. On the other hand, as an exam-
ple of intelligent thought based on a reflected abstraction, a student might interpret
the expression sin(θ) as a representation of the distance of the terminal point18 of an
angle above the horizontal diameter of a circle centered at the vertex of an angle,
measured in units of the circle’s radius. This interpretation relies upon constructing
sin(θ) as a symbol representing the following sequence of coordinated actions and
17
It is possible for figurative schemes to emerge via empirical and pseudo-empirical abstractions,
but such schemes are, by definition, non-operative (see the discussion of figurative and operative
knowledge structures below). Logico-mathematical knowledge, on the other hand, must be opera-
tive, meaning that the schemes it entails are composed of reversible mental actions that can be
applied to a generic class of objects without regard for an initial state.
18
The terminal point of an angle in standard position is the intersection of the terminal ray and a
circle centered at the angle’s vertex.
abstractions: (1) rotate an angle’s terminal ray so that the angle subtends an arc that
is θ times as large as the arc’s radius; (2) pseudo-empirically abstract the terminal
point’s distance above the horizontal diameter of the circle containing the subtended
arc as a measurable attribute (i.e., a quantity); and (3) multiplicatively compare this
distance to the radius of the subtended arc to quantify the value of sin(θ). These two
examples illustrate that every action produces an effect and are thus transformations
from one state to another. It is worth emphasizing again that the relationship between
actions and effects (and whether an action and its effects are decoupled) is funda-
mental to the process of abstraction and the distinction between the various forms
of abstraction that Piaget defined.
From a Piagetian perspective, to “know” something is to have a scheme with
which one might assimilate experiences of it. Piaget (1967) explained, “To know an
object implies its incorporation in action schemes, and this is true on the most ele-
mentary sensorimotor level and all the way up to the highest logical-mathematical
operations” (p. 17). To speak of one’s knowledge of a particular concept or idea,
then, is to speak of their scheme of meaning, or perhaps a network of related
schemes of meaning, pertaining to the concept or idea. Similarly, to speak generally
of one’s mathematical knowledge is to reference the aggregate collection of such
schemes for mathematical concepts and ideas.
In our discussion of scheme, the keyword is organization. Of course, the actions
being coordinated matter significantly, but the scheme is the organization of inter-
nalized actions. Identifying weak connections or seemingly random associations
that come to an individual’s mind related to specific stimuli does not indicate the
presence of a scheme and would not suffice for modeling a possible scheme.
Constructing a model of another’s scheme involves gathering evidence about the
predictable implications that arise when the scheme is active.
Imagery19
Images are things that come to mind. They are fragments of a person’s past experi-
ences. Thompson, Byerley, and O’Bryan (Chap. 5) provide a thorough description
of the role of imagery in genetic epistemology. We have nothing to add to their
work, but we feel it is worth highlighting several points they make. They argue that
“images of having reasoned are the foundation for reflection and scheme develop-
ment” (p. 164) and define imagery and its connection to scheme construction as
follows:
In an act of re-presentation, a person recalls having an experience. The experience might
have been having a thought or feeling, having done something in some context, having
interpreted something in a particular way, recalling having recalled something, or other acts
of recollection. In general, by “image” we mean the recollection of experience. […] Images
19
Thompson, Byerley, and O’Bryan (this volume) provide a thorough description of imagery
from a Piagetian perspective.
enter our definition of scheme in three ways: Images can be contexts that activate a scheme,
they can be waypoints in a scheme’s activity, and they can be anticipations of a scheme’s
result within the context of the scheme’s activation. (p. 130)
Constructing schemes requires individuals to reflect on their reasoning, where re-

presented “images of having reasoned become students’ objects of reflection” (ibid.,
p. 141).
Thompson et al. go on to describe different levels of imagery “differentiated by
the ways images are integrated into individuals’ reasoning and the types of reason-
ing into which they are integrated” (ibid., p. 135). The two levels most germane to
this chapter are the figurative level of imagery and the operative level of imagery.20
Figurative imagery relates to an:
activity sequence [that] results in a new state, but upon arriving at a new state it is simply
the end of the activity sequence. It is not a goal toward which the student strove and thus the
end state is not an anticipated result of the activity, and the actual result does not, to the
student, entail an image of the activity leading to it. (ibid., p. 136)
Operative imagery, on the other hand, “serve(s) as arbitrary ‘momentary states’ in a

scheme’s implementation” (ibid., p. 139). The imagery is no longer critical for the
individual’s reasoning but can be selected with intention as a symbolically conve-
nient tool to aid reasoning.
In discussing students’ thinking about angle measure, we examined how a stu-
dent may pseudo-empirically abstract the idea of angle measure as a comparison of
arc lengths produced after having rotated one ray in each of several angles (see our
discussion in Figs. 8.4 and 8.5). The student creates the arcs by rotating angles’
terminal rays, but the existence of such arcs is not differentiated from the student’s
action that produced them. The imagery involved is a result (three red arcs) with
lengths that can be compared. This imagery (consistent with the description of fig-
ural imagery) is integral to the student’s pseudo-empirical abstraction and is tied to
the context in which the activity occurs (i.e., it is dependent on the figural aspects of
the context in which the student is working). Working with angle measures in other
imagistic contexts (such as comparing the sizes of the three angles in a triangle) is
unlikely to be assimilated in such a way that the student conjures the image of arc
lengths produced after rotating one ray in an angle. A student shifting between the
two imagistic contexts might display what, to an observer, seems like inconsistent
meanings and imagery but that are not inconsistent to the student since different
imagistic contexts are associated with different images of having reasoned.
In Fig. 8.8, we introduced Hardison and Lee’s (2020) funky protractor task. If the
student has constructed a reflected abstraction for the meaning of angle measure,
then they have images of having reasoned about angle measure as an equivalence
class of arc lengths for all circles centered at an angle’s vertex. The student may then
spontaneously (i.e., not under direction of another) draw a succession of rays from
the large dot on each funky protractor through the angle measure indices on the
protractor, draw a circle centered at the large dot, and assess whether equal arc
20
Also, see the next section where we discuss figurative thought and operative thought.
lengths are subtended by adjacent terminal rays. The imagery that supports this
reasoning is operative (i.e., not dependent on the figurative elements of a specific
funky protractor). On the other hand, students incorrectly complete this task by, for
example, claiming that funky protractor C is a valid tool because the angle measure
marks on the isosceles triangle appear to be equally spaced or that only half of funky
protractor D could be used as a valid tool relying on figurative imagery associated
with laying down protractors to measure angles.
Imagery extends beyond objects and experiences to schemes themselves. It is
thus possible to have an image of a scheme. As a scheme develops, it becomes pos-
sible to carry out that scheme as an action—something implemented without effort,
often reflexively. If the scheme reaches a certain level of sophistication, the indi-
vidual can gain conscious access to the scheme and the internalized actions orga-
nized within it. That is, the knowing subject can think about the scheme, talk about
its characteristics, and apply it strategically, as exemplified by the reasoning required
to respond successfully to the task in Fig. 8.8. At this point, the individual has an
image of the scheme, which can be very powerful. When the scheme has become
reified as an image, it facilitates the individual’s development of a way of reasoning
that comes to mind easily when the scheme is triggered and can be purposefully
implemented in quite flexible and general ways. Reflected abstraction is present
whenever an image of a scheme becomes an object of conscious thought, after
which its characteristics can be abstracted from the class of objects and experiences
to which the scheme applies. This is the origin of explicit ways of reasoning that
enable a student to experience mathematics as a coherent body of ideas. We return
to the role of reflected abstraction on students’ development and application of
mathematical ways of thinking in our discussion of the relationship between
reflected abstraction and Harel’s theoretical framework, DNR-based instruction in
mathematics.
Figurative and Operative Modes of Thought21
Reflecting and reflected abstractions contribute to the construction, modification,

refinement, and integration of operative structures. To clarify the dependence of
operative modes of thought—and logico-mathematical schemes in particular—on
reflected abstractions, we discuss the distinction between figurative and operative
knowledge.
Piaget (1970) contrasted aspects of two types of thought, operative and figura-
tive, as follows:
The figurative aspect is an imitation of states taken as momentary and static. In the cogni-
tive area the figurative functions are, above all, perception, imitation, and mental imagery,
21
Moore, Stevens, Tasova, and Liang (this volume) provide a detailed description of figurative
and operative modes of thought.
which is in fact interiorized imitation. The operative aspect of thought deals not with states
but transformations from one state to another. For instance, it includes actions themselves,
which transform objects or states, and it also includes the intellectual operations, which are
essentially systems of transformation. (p. 14)
The operative aspect of thought relates to how a knowing subject structures their
experiences by assimilation of figurative material (Müller, 2009, p. 223). Operative
and figurative thought are, therefore, complementary and interdependent. Knowing
subjects possess the figurative functions of perception, imitation, and mental imag-
ery that produce the material for assimilation and the objects on which operative
schemes perform actions and transformations. On the other hand, the contents and
organization of an individual’s operative structures influence what they can assimi-
late through these figurative functions. All of this means that, for Piaget, knowledge
is essentially operative. Figurative material is assimilated to and controlled by oper-
ative structures, enabling the knowing subject to make strategic decisions while
engaged in goal-oriented activity and to anticipate specific outcomes of their actions
(Thompson, 1985; von Glasersfeld, 1995).
Thompson (1985) expanded the definition of figurative aspects of thought in
explaining how operative schemes control mathematical reasoning processes. In
doing so, he broadened what constitutes figurative thought by including more than
constructions derived from perception, imitation, and mental imagery:
When a person’s actions of thought remain predominantly within schemata associated with
a given level (of control), his or her thinking can be said to be figurative in relation to that
level. When the actions of thought move to the level of controlling schemata, then the think-
ing can be said to be operative in relation to the level of the figurative schemata. That is to
say, the relationship between figurative and operative thought is one of figure to ground.
Any set of schemata can be characterized as figurative or operative, depending upon
whether one is portraying it as background for its controlling schemata or as foreground for
the schemata that it controls. (p. 195)
We previously stated that operative thought develops through reflective abstrac-

tions: reflecting abstractions are involved in the construction and modification of
schemes with which figurative material can be assimilated and transformed, and
reflected abstraction is the product of these schemes being organized in ways that
enable the individual to become cognizant of the coordinated actions organized
within them. There are many affordances of constructing operative mathematical
schemes through reflecting and reflected abstractions. For instance, the control over
figurative material by operative structures allows one to anticipate the outcome of
applying a scheme and to make propitious decisions in relation to prior events and
current circumstances (Thompson, 1985, p. 194; von Glasersfeld, 1995, p. 65).
Piaget’s (1970) description and Thompson’s (1985) generalization of the opera-
tive form of cognition clarify how constructing schemes that foreground the opera-
tive mode of thought is essential to acting with and performing operations on
mathematical objects based on their meaning for the actor. If the figurative aspects
of a student’s thought are uncontrolled by operative structures—that is, if a student’s
conception of a mathematical idea is dominated by their representation of figurative
material without this material being the object of action or transformation—then
they are unlikely able to assimilate novel figurative material and experiences. If, on
the contrary, a student has constructed operative structures that control their action
on and transformation of figurative material, then they are capable of reasoning
productively about unfamiliar mathematical problems and situations.
We borrow an example of the contrast between figurative and operative modes of
thought from Tallman and Uscanga (2020) and use this example to comment on
how the strategic application of operative structures relies upon reflected abstrac-
tions. Consider how two students might respond to the task of defining a function to
represent the relationship between quantities represented graphically, one in
Cartesian coordinates and another in polar coordinates (Fig. 8.10).
Both graphs represent linear functions, a property that can be empirically
abstracted when considering the graph in Cartesian coordinates (its straightness can
be perceived) and pseudo-empirically abstracted from the graph represented in
polar coordinates (the graph is a product of an action: using a graphing technology
to represent an algebraic relationship that corresponds to a general symbolic form
(e.g., [output variable] = [“slope”] × [input variable] ± [constant]) the student asso-
ciates with linear functions). Constructing a function definition for the relationship
between the input and output quantities of the two functions first requires recogniz-
ing that each function is linear and then determining the constant rate of change and
initial value of each.
The familiar idiom “rise-over-run” is suggestive of a meaning for rate of change
based in and constrained to figurative material, where “rise” refers to the vertical
change between two points on the Cartesian graph (often with integer ordinate val-
ues) and “run” refers to the corresponding horizontal change, both measured in
units determined by the respective axes of the coordinate system. This conception is
dominated by figurative thought because of its reliance on sensorimotor experience
(i.e., moving up and over) and representation of perceptual material (i.e., a trace of
sensorimotor movements that, respectively, represent “rise” and “run”). Moreover,
this meaning is constrained to a particular representation, which is an essential
Fig. 8.10 Linear functions represented in Cartesian and polar coordinates. (Tallman & Uscanga,
2020, p. 205)
feature of conceptions that foreground figurative aspects of cognition (Moore et al.,

2019). A student who operates with this meaning will respond to the task of deter-
mining the constant rate at which y varies with respect to x by identifying locations
where the graph intersects lattice points of the Cartesian coordinate plane and then
computing the ratio of the vertical change (rise) and horizontal change (run) between
these points. Although the student might recognize that the selection of points that
determine these directed lengths is arbitrary, they do not conceptualize the resulting
ratio as a rate since, for the student, it has no meaning in terms of the constrained
covariational relationship between x and y.22 Rather than unhesitatingly trying to
leverage this “rise-over-run” conception of rate of change to the linear function
represented in polar coordinates, the student will likely notice perceptual novelties
that they are not positioned to accommodate (e.g., the gridlines are circles centered
at and lines passing through the origin; the graph fails the “vertical line test”).
For contrast, consider a student who conceptualizes constant rate of change as a
proportional relationship between corresponding changes in the continuously vary-
ing measures of a function’s input and output quantities. Rather than being con-
strained by perceptual material or sensorimotor experience, these aspects of
figurative thought are controlled by operative structures of action and transforma-
tion. Although the student who possesses this operative conception might demon-
strate similar behaviors as the student whose understanding for rate of change is
∆y
limited to “rise-over-run,” what distinguishes the result of their computation is
∆x
that it represents the constant of proportionality relating the measures of covarying
magnitudes ∆y and ∆x; it is not the direct application of a mnemonic device.
Importantly, the student who possesses such an operative conception will also rec-
ognize perceptual novelties of the polar graph but will nonetheless expect that their
meaning is sufficient for coping with them to sensibly determine the constant rate at
which the output quantity varies with respect to the input quantity. This, of course,
requires a robust understanding of the polar coordinate system (Moore et al.,
2014)—including what it means to measure an angle in radians—and a particular
way of conceptualizing the covariation of the input and output quantities in this
context (Thompson & Carlson, 2017). Equipped with such an understanding and
way of reasoning, the student can coordinate changes in an angle’s measure (θ) with
corresponding changes in radius (r) to determine the constant rate at which r varies
with respect to θ), as shown in Fig. 8.11.
This example demonstrates the affordances of having constructed an operative
scheme for constant rate of change through the processes of reflecting and reflected
abstraction to which the student can purposefully assimilate the novel task of con-
structing a symbolic definition for the function represented in polar coordinates.
22
A ratio is a multiplicative comparison of the measures of two constant (non-varying) quantities,
while a rate defines a proportional relationship between varying quantities’ measures (Thompson
& Thompson, 1992). Constructing a rate therefore involves images of smooth continuous variation
(Thompson & Carlson, 2017), as well as the expectation that as two quantities covary, multiplica-
tive comparisons of their measures remain invariant.
Fig. 8.11 Constant rate of change in polar coordinates. (Δr = 2Δθ) (Tallman & Uscanga,
2020, p. 207)
Note. The length of the green arc representing Δθ is measured in units of the radius r
Assimilating the task as described depends on the student’s awareness of their own
meaning for constant rate of change; assessing the linearity of the function by iden-
tifying and multiplicatively comparing several corresponding changes in the input
and output quantities of the function represented in polar coordinates is motivated
by the student’s awareness of the actions and operations entailed in their meaning
for constant rate of change.
elation of Reflected Abstraction to Theoretical Constructs

R
within Mathematics Education Research
In this section, we discuss how reflected abstraction is the basis for, or is related to,
other ideas in the field used to study students’ mathematical thinking and learning
and thus how reflected abstraction helps make sense of and potentially unite other
theoretical frameworks. We begin by describing Dubinsky’s (1991) APOS theory
and compare his use of reflective abstraction to characterize how students construct
mathematical knowledge with our understanding of Piaget’s description of reflect-
ing and reflected abstraction. We then outline Harel’s (2007, 2008a, 2008b) DNR-
based instructional theories, specifically the duality principle describing the
interconnectedness of ways of understanding mathematical ideas and general ways
of thinking about mathematics. We next discuss the theories of quantitative reason-
ing (Thompson, 1988, 1990, 1993, 1994, 2011) and covariational reasoning (Carlson
et al., 2002; Saldanha & Thompson, 1998; Thompson & Carlson, 2017) as ways of
thinking in the spirit of Harel’s definition of the term. We end by describing Lobato’s
(2003, 2006, 2008, 2012; Lobato et al., 2012; Lobato & Siebert, 2002) framework
of actor-oriented transfer which connects to Harel’s and Thompson’s (and their col-
leagues’) work through its emphasis on respecting and modeling students’ idiosyn-
cratic meanings for mathematical ideas. Throughout all of the latter three research
programs we discuss, reflected abstraction is key to understanding how students’
mathematical meanings and reasoning patterns develop and as a lens for analyzing
and modeling student thinking.
APOS Theory
Dubinsky (1991) devised a theoretical framework of mathematical knowledge and

its construction based on Piaget’s notion of reflective abstraction. For Dubinsky, and
consistent with Piaget’s distinction between figurative and operative modes of
thought, reflective abstraction is the means by which form or process is divorced
from sensorimotor content and converted into objects of thought that the subject is
then able to act upon to form new constructions (1991, p. 98). Dubinsky explained,
“For us, reflective abstraction will be the construction of mental objects and of men-
tal actions on these objects” (ibid., p. 102).
Dubinsky emphasized the constructive aspect of reflective abstraction,23 which
led him to define the following five varieties of construction, each of which consti-
tutes a reflective abstraction: (1) interiorization, (2) coordination, (3) encapsula-
tion, (4) generalization, and (5) reversal. Together these five forms of construction
comprise a theoretical framework that Dubinsky argued can be applied to develop
genetic decompositions of mathematical concepts. A genetic decomposition is a
description of a particular mathematical idea together with how one might make the
constructions necessary to achieve a mature understanding of the idea (Dubinsky,
1991, p. 96). The purpose of developing a genetic decomposition is to inform
instructional design in ways that encourage students to engage in the precise reflec-
tive abstractions, and thereby constructions, necessary to achieve an intended under-
standing of a specific concept.
Interiorization, the first of the five varieties of construction in reflective abstrac-
tion, involves the internal representation of an action or process. Coordination, the
second type of construction, entails composing two or more actions or processes to
23
Dubinsky (1991) used the term “reflective abstractions,” and so we follow his lead in this section.
Piaget used “reflective abstraction” generally to describe abstraction that is not empirical or
pseudo-empirical. Its use does not distinguish between reflecting abstractions, the construction of
reflected schemes, and meta-reflection. Dubinsky does not himself distinguish between these more
nuanced characterizations of abstraction in his theory, although we would argue that his use of
“reflective abstraction” might be better interpreted as “reflecting abstraction” consistent with how
we’ve described the construct in this chapter.
develop a new one. The third form, encapsulation, requires the conversion of
dynamic actions or operations into objects that can then be acted or operated upon
on a higher plane of thought. Generalization, the fourth form, occurs when one is
able to widen the applicability of existing schema,24 thus expanding the set of expe-
riences these schema are equipped to assimilate. Finally, the last type of construc-
tion, reversing, involves the construction of a new process that reverses some initial
process. Dubinsky applied these five forms of construction in reflective abstraction
to explain how one is able to construct new mathematical objects, processes, and
schema (ibid., p. 103).
In Dubinsky’s usage, schemes are dynamic activities composed of objects and
processes (1991, p. 106). A scheme is dynamic in the sense that new objects and
processes are being constructed out of existing ones via the five forms of construc-
tion in reflective abstraction that Dubinsky defined. When a subject interiorizes an
action performed on objects, the action becomes a process. New processes may also
arise out of coordination and/or reversal of existing processes. Encapsulating a pro-
cess transforms it into an object capable of being acted upon, thereby expanding the
applicability of the existing scheme, and thus constituting a generalization.
Figure 8.12, reproduced from Dubinsky (1991), illustrates the way a subject acts
upon the existing objects and processes of a scheme via the five forms of construc-
tion in reflective abstraction to construct new objects and processes.
One of the challenges with linking genetic epistemology and APOS theory is the
somewhat cursory application of Piaget’s framework to justify the five constructive
processes within APOS theory. The value of a theoretical framework lies is in its
potential to explain phenomena, not simply label them. But in describing the
five constructive abstractions, APOS theory supplies theoretical constructs
Fig. 8.12 Schemas and their construction. (Dubinsky, 1991, p. 107)
24
The terms “schemes,” “schemas,” and “schemata” have appeared in English translations of
Piaget’s work as plurals of “scheme.”
(interiorization, coordination, encapsulation, generalization, and reversal) in place

of detailed explanations of the cognitive mechanisms they entail, including the con-
nections between them. Thus, while some consider APOS theory an extension of
Piaget’s genetic epistemology, we perceive it as more of a pragmatic simplification
of his ideas due to its lack of specification of the explanatory mechanisms of reflec-
tive abstraction that Piaget inferred from his decades-long psychological research
program.
In addition, some researchers (e.g., Thompson, 2002) have suggested that genetic
decomposition, while useful for specifying distal learning goals, falls short of being
sufficient to inform teachers' instructional design and guide their in-the-moment
interactions with students because, in contrast with the related idea of conceptual
analysis (Thompson, 2008), genetic decompositions are not:
grounded in conceptual experience—to make clear that we’re talking about someone hav-
ing something in mind—that what they have in mind derives from and is constantly
informed by imagery, emotion, intention, etc. … [and] without grounding [what we want
students to learn] in conceptual experience it loses much of its power when attempting to
describe how someone might learn it. (Thompson, 2002, p. 196)
Thus, although Dubinsky argued that a genetic decomposition is useful for under-
standing how to design learning experiences that foster students’ engagement (via
reflective abstraction) in the five constructive processes defined within APOS the-
ory, others have questioned whether working backward from the mature under-
standing of an expert fully respects the actual development that must occur as
individuals move from immature, naïve conceptions of mathematical ideas toward
more sophisticated understandings. Attention to the cognitive mechanisms of
reflecting and reflected abstraction that Piaget (2001) described, and their depen-
dence on the current cognitive structures of the learner, might attenuate this criti-
cism of APOS theory to enable mathematics educators to design learning experiences
for students that, as Dewey (1902) advocated, maintain fidelity to both the content
of the academic discipline and the current and potential experience of the learner.
Harel’s Duality Principle
Harel (2007, 2008a, b) offered a conceptual framework (called DNR-based instruc-

tion in mathematics or just DNR) for curriculum and instructional design that pro-
vide students with rich, engaging, and powerful mathematical experiences. The
foundation of his framework consists of three primary principles: (1) the duality
principle, (2) the necessity principle, and (3) the repeated-reasoning principle. In
service of the central goals of this chapter, we omit a discussion of intellectual need
and focus on the other two principles and their connections to reflected abstraction.
Explaining the duality principle requires attention to supporting theoretical con-
structs within DNR. Harel (2007) acknowledged that mathematical learning entails
a subject’s engagement in mental activity, including but not limited to defining,
conjecturing, symbolizing, predicting, and anticipating. However, the DNR frame-

work focuses on “the product and character of a mental act … for it delineates the
content of the cognitive objectives at which DNR-based instruction aims as well as
the knowledge held by the learner” (p. 265). The former is associated with what
Harel called a way of understanding and the latter a way of thinking. A way of
understanding is a “product of a mental act” (Harel & Koichu, 2010, p. 117) that
corresponds with the individual’s meaning of that act. Harel and Koichu (2010)
illustrate various ways of understanding by discussing potential meanings of divid-
ing 4/5 by 2/3, such as seeing the result as the product of 4/5 and 3/2 (the reciprocal
of 2/3), the value of x that satisfies (2/3)x = 4/5, or how many times 2/3 “goes into”
4/5. Each of these results reflects a different way of understanding division of frac-
tions and is a product derived from the application of individuals’ idiosyncratic
schemes. Note that a way of understanding need not correspond to common or
accepted mathematical ideas.
A way of thinking, on the other hand, describes the character of an individual’s
thinking in relation to a mental act or combination of mental acts. For example, in
solving a variety of novel problems, a student may search task statements for key-
words, identify surface-level characteristics to cue the enactment of familiar proce-
dures, or try to solve a simpler version of a problem (Harel, 2007). These are all
ways of thinking about the constellation of mental acts entailed in problem-solving.
Designating something as a way of thinking does not imply anything about its cor-
rectness or its apparent utility from an observer’s perspective. Engaging in a “guess
and check” strategy could be a student’s way of thinking about solving novel math-
ematics problems, a method that an instructor might consider of limited utility. A
way of thinking develops through the reasoner experiencing the affordances of their
thinking in a range of contexts the reasoner perceives as dissimilar.
The duality principle describes the reciprocal relationship between ways of
understanding and ways of thinking. Harel (2007) argued, “Students develop ways
of thinking only through the construction of ways of understanding, and the ways of
understanding they produce are determined by the ways of thinking they possess”
and added, “not only do these two categories of knowledge affect each other but a
change in one cannot occur without an appropriate change in the other” (p. 11).
Harel’s point is, in part, intended to highlight the ways that common instructional
methods neglect supporting students in developing meaningful ways of thinking
(which, in turn, limits their ability to construct meaningful ways of understanding).
Ideally, instruction would support students’ development of robust, flexible, con-
nected, and consequential ways of understanding a variety of mathematical ideas
from which students might become cognizant of general ways of thinking that guide
their effective mental activity. In addition, instruction should support students' stra-
tegic application of powerful ways of thinking in the service of supporting their
construction of more robust, flexible, connected, and meaningful ways of under-
standing mathematical ideas. Such purposeful enactment of explicit ways of think-
ing presupposes cognizance of their essential features, and thus is dependent upon
having projected one’s reasoning to the reflected level of thought.
If a student can develop productive ways of understanding a variety of mathe-

matical ideas and become aware of the coordinated actions organized within their
cognitive structures, then they are positioned to abstract common features of their
mathematical schemes and to reflect on how particular ways of thinking contributed
to their construction. This kind of reflective activity is indispensable for supporting
an aptitude for flexible problem-solving because it equips the learner with an image
of how, and in what circumstances, they might leverage appropriate ways of think-
ing when confronted with a novel mathematical task.
A student’s way of understanding an idea connects with the assumption within
genetic epistemology that statements about a subject’s understanding of something
are claims about the subject having assimilated an experience, and what we call the
subject’s meanings refers to the implications (actions, expectations, connections,
etc.) that result from the assimilation (Thompson et al., 2014). That is, “schemes
provide the ‘way’ in Harel’s ‘way of understanding’” (ibid., p. 12) since schemes
consist of the meaning individuals possess for the processes they enact and the
products of reasoning they produce. Ways of understanding emerge from abstract-
ing reasoning from mathematical experiences. Students must recurrently engage in
similar reasoning, formulating the knowledge repeatedly until they construct a sta-
ble operative scheme for a particular mathematical idea. This is Harel’s repeated
reasoning principle. It must be noted, however, that what the student attends to is
critical. Repeatedly practicing solution methods is very different from recurrently
reasoning about genuine problems and working to communicate and justify that
reasoning. Harel envisioned engaging in repeated reasoning as an essential compo-
nent of meaningful mathematics learning.
From the perspective of genetic epistemology, repeated reasoning (including
reflecting on that reasoning) provides opportunities for a learner to abstract proper-
ties of their actions dissociated from the effects they produce, and to thus engage in
reflecting abstraction. As a result, students may develop stable schemes of particular
mathematical ideas. Reflected abstractions result when the individual reflects on the
products of previous reflecting abstractions (their schemes), reorganizing them and
becoming cognizant of their meanings so they can be applied intentionally and stra-
tegically to reason about novel mathematical tasks and contexts. Sophisticated ways
of thinking, such as quantitative reasoning (see the next section), develop through
repeated reasoning and projecting this reasoning to the reflected level of thought.
Students who are aware of the elements of their reasoning about specific mathemati-
cal ideas are poised to compare that reasoning and recognize generalities (i.e., ways
of understanding impact ways of thinking an individual constructs). This can prompt
a further reorganization of their reflected schemes such that ways of thinking emerge
that can be purposefully applied to solve novel problems. Ways of thinking thus
employed to make sense of new mathematical ideas conditions what the individual
notices and abstracts from those learning experiences, and thus impacts the new
schemes the individual constructs (i.e., ways of thinking impact the development
and refinement of ways of understanding).
For example,25 imagine that a student is supported in analyzing various function

relationships by creating tables of values and studying the patterns that emerge (and
then exploring how those same patterns are apparent in graphical and formulaic
representations of the same relationships). Through analysis of many examples
within each function family, the student may abstract general properties of each
(e.g., linear functions are relationships where corresponding changes in the inde-
pendent and dependent quantities are proportional; exponential functions are rela-
tionships where the dependent values are proportional for fixed changes in the
independent quantity (with the constant of proportionality dependent on the interval
size chosen)). These abstractions result in schemes for function classes grounded in
an analysis of what remains invariant in how pairs of quantities covary. If the stu-
dent is consciously aware of their meanings of these ideas (i.e., the student pos-
sesses reflected schemes) and sees how their meanings emerged through the
common practice of constructing tables of values, looking for patterns in how quan-
titates’ values change together, and considering how the same patterns can be repre-
sented graphically and formulaically, the student might interpret their covariational
reasoning as a general process that is useful for analyzing any mathematical rela-
tionship they encounter. If so, the student might construct this process as a way of
thinking about understanding mathematical relationships (ways of understanding
impact ways of thinking). When the student encounters a novel relationship in the
future, the student might apply this way of thinking on their own to make sense of
the situation. The properties the student notices and how they come to understand
this new situation are conditioned by their way of thinking about how to make sense
of mathematical relationships (ways of thinking impact ways of understanding).
Quantitative26 and Covariational Reasoning
Thompson’s (1988, 1990, 1993, 1994, 2011) theory of quantitative reasoning

describes a way of conceptualizing mathematical relationships and representations.
Quantitative reasoning is a way of thinking about a context such that the individual
conceptualizes measurable attributes (quantities) and mentally forms an organized
structure of relationships between measurable attributes. Thompson emphasizes
that quantities and quantitative structures exist in an individual’s mind and are thus
idiosyncratic to the individual. How an individual conceptualizes quantities and the
relationships between them affords or constrains the individual’s capacity to reason
within a situation (Smith & Thompson, 2008; Thompson, 1994).
Central to quantitative reasoning is the process of quantification or the process of
developing a measurement scheme for producing a quantity’s value(s) as a
25
This example is related to quantitative reasoning and covariational reasoning (see the next
section).
26
Boyce (this volume) provides a detailed description of quantitative reasoning.
multiplicative relationship between the signed magnitude of the quantified attribute

and the magnitude of the compatible attribute of a particular unit of measure.
Quantification is vitally important for mathematical modeling because “[i]t is in the
process of quantifying a quality that [the quality] becomes truly analyzed”
(Thompson, 1990, p. 5). Some quantities, such as torque, the force of a collision, or
the speed of a moving object, are conceptualized through a quantitative operation,
a mental operation of comparison or coordination of other conceptualized quantities
(Thompson, 1990, 1994, 2011). Such quantities (called intensive quantities) cannot
be measured directly the same way the length of a table can be measured in linear
units. Quantifying intensive quantities depends upon the quantitative operations the
individual conceptualizes (Johnson, 2015; Moore, 2010; Piaget & Duckworth,
1968; Schwartz, 1988; Simon & Placa, 2012; Thompson, 1990; Thompson et al.,
2014). As someone reasons about a context, they may distinguish between quanti-
ties with fixed magnitudes and quantities with varying magnitudes. Covariational
reasoning refers to reasoning about how a quantity’s magnitude (and measurement)
varies in tandem with another quantity’s magnitude (and measurement), which
involves holding in mind a continuous image of the two quantities’ magnitudes (or
measurements) as they change together (Carlson et al., 2002; Saldanha & Thompson,
1998; Thompson & Carlson, 2017).
In explaining the characteristics of quantitative and covariational reasoning,
what some researchers overlook is the emphasis that these are ways of thinking in
the spirit of Harel’s definition of the term (as opposed to simply being descriptions
of types of reasoning). While it is possible to point to an individual’s specific rea-
soning in a specific context and characterize the degree to which that reasoning is
consistent with elements, or distinct forms, of quantitative or covariational reason-
ing, that is not Thompson’s (sole) vision for the affordances of his theory. Rather,
the propensity to reason quantitatively allows an individual to impose structure on
and coherence to the relationships within a context. Individuals will intentionally
and strategically reason quantitatively about a situation when they appreciate the
utility of such reasoning and have the expectation that doing so will be productive
in solving the task at hand. For example, a person whose quantitative reasoning is
robust enough to function as a way of thinking consistently expects that a numerical
value is the result of applying a measurement process to some attribute (making a
multiplicative comparison to a unit of measure) and that the value is dependent on
the unit chosen such that changing the unit of measure changes the numerical value
representing the attribute’s magnitude without the attribute itself changing
(Thompson et al., 2014). In addition, an individual with the proclivity to reason
quantitatively will search for what remains invariant in a situation as two (or more)
quantities change in tandem. They might begin a modeling task by asking, “What
quantities have values that can change? Which do not? Is there an invariant pattern
in how a pair of covarying quantities change together?” The answers to such ques-
tions are critical to mathematical modeling because by seeking what remains invari-
ant in a dynamic situation (which is often an intensive quantity), one can generate,
justify, and operate with meaningful representations of these quantitative relation-
ships (Carlson et al., 2002; Thompson & Carlson, 2017).
As we stressed in our discussion of Harel’s duality principle, constructing quan-

titative reasoning as a way of thinking and applying it intentionally requires reflected
abstraction. A proneness to reason quantitatively develops27 as students recognize
the utility of quantitative reasoning across many mathematical contexts (e.g., facil-
ity with organizing information in word problems, successfully reasoning about
linear contexts by holding a constant ratio of the changes in two covarying quanti-
ties, understanding growth factors of exponential functions as relative size measure-
ments of two dependent values separated by a fixed change in the independent
variable across the function’s domain, etc.). That is, students develop ways of
understanding that rely on quantitative reasoning and begin to generalize aspects of
their solution methods as a general way of thinking. Conversely, the inclination to
reason quantitatively becomes a lens through which new situations are examined,
and ways of understanding develop for making sense of new mathematical ideas as
per Harel’s duality principle.
Lobato’s Actor-Oriented Transfer
Students’ transfer of learning from one setting to another is an important focus in

mathematics education research. Traditional cognitive accounts of students’ trans-
fer of learning focus on “the surface/structure distinction” (Lobato & Siebert, 2002,
p. 88). Researchers studying transfer often focuses on students’ work in paired tasks
with similar structures (such as a similar solution method) but different surface
details (such as the context in which the problem is set). For example, determining
a missing angle measure in a right triangle using trigonometric ratios and finding the
angle of inclination for a line passing through the origin and a point in the first quad-
rant can be solved by similar solution methods but might be thought of as differing
in their surface-level details. Researchers with this perspective routinely fail to find
evidence that students have transferred learning when surface features change.
Lobato and colleagues (Lobato, 2003, 2006, 2008, 2012; Lobato et al., 2012;
Lobato & Siebert, 2002) challenged this traditional perspective and argue that trans-
fer occurs more often than previously reported—the issue is that researchers need to
rethink their perspective when searching for learning transfer. They presented a dif-
ferent conceptualization, which they termed actor-oriented transfer (AOT) or “the
personal construction of relations of similarity between activities, or how ‘actors’
see situations as similar” (Lobato & Siebert, 2002, p. 89). The distinction Lobato
emphasized is between perspectives—the observer versus the actor. In observer-
oriented cognitive transfer studies, researchers seek evidence of skills learned in one
setting applied in another setting that the researcher considers to have shared struc-
tural elements. In actor-oriented transfer studies, the focus shifts to “scrutinizing a
given activity for any indication of influence from previous activities and by
27
See our example in the previous section.
examining how people construe situations as similar” (Lobato & Siebert, 2002,
p. 89). Actor-oriented transfer assumes that transfer across contexts occurs when the
cognizing individual recognizes some inherent “sameness” regardless of similari-
ties anyone else sees (or does not see) between the contexts.
Lobato and Siebert (2002) detail a study where one student (Terry) did not apply
the general slope formula to determine a wheelchair ramp’s steepness despite partici-
pating in lessons during which the instructor introduced the slope formula. Observer-
oriented transfer analysis would group these two techniques into the same structural
family and treat the differing context as a surface feature of the problem. However,
Lobato and Siebert’s interviews with Terry showed that his conceptualizations of the
quantitative relationships in the problem and his conception of different ways to
think about measuring steepness demonstrated that the context was far from a sur-
face feature to Terry—it was fundamental to what he noticed and how he reasoned
about the problem. In another study (Lobato, 2008), students struggled to determine
the slope of a playground slide with a top horizontal platform because they did not
select the correct lengths for the vertical and horizontal displacements of the slide
itself. In her analysis, Lobato points out that classroom lessons emphasized a stair-
case metaphor for counting these displacements for purposes of calculating a slope
measurement. With this knowledge, Lobato reanalyzed student responses and argued
that students were attempting to apply this metaphor to the slide problem—they were
using the entire horizontal and vertical dimensions of the entire structure (like they
did with staircases) and perhaps were looking specifically for step-like shapes to help
them determine the displacements for their calculations. From this perspective, stu-
dents did transfer their learning from one setting to another. The issue, which tradi-
tional cognitive accounts of transfer often overlook, is that transfer of learning can be
seen even when the resulting conclusion is incorrect. In fact, such information is
quite useful for curriculum designers and teachers because understanding the
ways that certain early learning scenarios might introduce barriers or confusion in
later learning can improve lessons and clarify learning trajectories.
Lobato (2006) and Lobato et al. (2012) acknowledge that AOT theories are con-
sistent with and based on genetic epistemology, in particular, the process of reflec-
tive abstraction and the notion of assimilation to cognitive schemes. Lobato (2006)
noted how “reflective abstraction is a constructive process in which the regularities
abstracted by the learner are not inherent in the situation, but rather are a result of
personal structuring related to learner’s goals and prior knowledge” (p. 441). Rather
than providing a detailed account of the mechanisms of reflective abstraction and
connections with learning and transfer, however, Lobato discussed how she adapted
the notion of reflective abstraction to account more for the social environment of the
classroom in transfer studies. But it is useful for our purposes to highlight how
thinking about the mechanisms of reflective abstraction can help us think about the
issue of students transferring their learning from one context to another. When a
student encounters a problem and makes a link to prior experiences, it is an instance
of assimilation. Those prior experiences are encapsulated within schemes, and the
scheme’s contents dictate the actions and expectations associated with the goal of
reasoning through the problem based on assimilating it to that scheme.
This last comment is important and is at the heart of AOT and Lobato’s
perspective that we can see transfer occurring even when solution methods are
incorrect or inappropriate from an expert’s perspective. Whatever scheme is
activated through assimilation indicates what other experiences the individual
interprets as similar to the current context. Thus, these are the experiences and
images of having reasoned that serve as the model for how the student will
reason about and approach the current task. If the student’s behaviors indicate
a need for figurative elements about which to reason, then the scheme to which
the experience is assimilated is figurative (and was constructed from a pseudo-
empirical abstraction). For example, when Lobato described students needing
to visualize or create a stairstep/staircase on the slide task to determine the
slide’s slope, the students evoked images of staircases to guide their reasoning
for determining a slope. What they abstracted was not the more sophisticated
notion of a ratio of changes but rather the figurative image of drawing stairs.
Developing an operative scheme for slope from activities of drawing stair-
cases would require a reflecting abstraction such that the individual constructs
a meaning for what work the staircase metaphor accomplishes (organizing
displacements in two dimensions for comparison). A reflected abstraction
would result from further refinement of the scheme such that either (a) the
metaphor can be applied strategically, purposefully, and flexibly across tasks
assimilated to that scheme with expectations for the utility of employing the
associated actions of determining displacements and generating ratio com-
parisons of those displacements or (b) coordinating the relative size of two
displacements becomes disassociated from the need to think about staircases
to quantify slope.
Actor-oriented transfer highlights the notion of abstraction and individuals’
construction of meaning, which is at the heart of genetic epistemology. From the
AOT perspective, the critical focus is what a person extracted from a given
learning experience (through any variety of abstraction) and to what scheme the
individual assimilates new tasks and experiences. When a researcher observes
evidence that a context or problem has been assimilated to a scheme constructed
in prior learning experiences, it is then possible to explore whether the student
has constructed a reflected abstraction. Students’ capacity to explain the con-
nections between past and current experiences and to strategically and purpose-
fully apply similar reasoning across contexts evinces of the kind of organized
control over one’s reasoning that is a hallmark of operative schemes constructed
at the reflected level of thought. Lobato’s AOT perspective reminds us as teach-
ers, curriculum designers, and researchers to focus our attention on what stu-
dents actually do to abstract from their learning experiences (not just the
meanings we hope they construct), to think carefully about the implications of
both the meanings we hope they develop and the meanings they seem to apply
in solving problems, and that learning experiences must be designed intention-
ally to foster reflection on the individual’s reasoning process to support reflect-
ing and reflected abstractions.
Implications of Reflected Abstraction
We conclude by describing the implications of reflected abstraction for students’

mathematical learning, teachers’ construction of pedagogical content knowledge,
and research in mathematics education.
I mplications for Mathematics Education Research

on Student Learning
When reflected abstraction is understood in connection to the other ideas within

Piaget’s genetic epistemology, we can identify principles for engendering reflected
abstraction and specify a standard of evidence for making valid claims about a sub-
ject’s construction of schemes at the reflected level of thought. Piaget (2001) rou-
tinely stressed that reflected abstraction is most effectively provoked when a
knowing subject is prompted to retroactively compare their activity while engaged
in seemingly different tasks that require the subject’s enactment of the same mean-
ing (i.e., invoke the same scheme resulting from a prior reflecting abstraction). An
individual’s ability to recognize the common actions and operations required by two
or more tasks exhibits their construction of a reflected scheme.
To illustrate this standard of evidence, consider the response provided by an
undergraduate student and prospective secondary mathematics teacher (Samantha)
who participated in a teaching experiment conducted by the first author (Tallman &
Weaver, 2018). The purpose of the teaching experiment was to explore the peda-
gogical implications of Samantha having constructed a reflected scheme for con-
stant rate of change. During the final three teaching episodes, the interviewer asked
Samantha to compare and contrast the reasoning required to solve 12 pairs of tasks
from the instructional sequence that guided the teaching experiment.
Tallman and Weaver asked Samantha to articulate her comparison in writing and
then respond to clarifying questions about what she had written. Although the tasks
in the instructional sequence varied substantially in terms of the information pro-
vided and the quantity Samantha was asked to compute or represent, all tasks in the
sequence could be solved by leveraging a meaning for a constant rate of change as
a proportional correspondence between changes in covarying quantities’ measures.
By prompting Samantha to compare the thinking required to reason productively
about pairs of tasks from the instructional sequence, all of which she had previously
completed, Tallman and Weaver intended to promote Samantha’s construction of a
reflected scheme for constant rate of change. These comparisons required Samantha
to perform mental operations on the products of her prior reflecting abstractions.
Continually re-enacting, comparing, and contrasting the reasoning she engaged in
brought her conceptual activity into her conscious mind and resulted in Samantha’s
rate of change scheme becoming progressively coherent and refined. We share one
illustrative example.During the sixth teaching episode, the interviewer prompted
Samantha to compare and contrast the reasoning required to solve the following
two tasks:
1. While training for a triathlon, Clara rides her bike on a long, straight road at a
constant rate. She passes under a stoplight while riding. How long does it take
Clara to ride 17 miles from the stoplight if she rides at a constant speed of
13 miles per hour?
2. A swimming pool that is only partially full has 10,800 gallons of water. Water is
added to the swimming pool 10:00 AM at a constant rate of 24 gallons per min-
ute. How long will it take to add 1725 gallons of water to the pool?
Discussing the meaning of rate of change that might enable students to solve these
two problems, Samantha explained:
I think they have to think about for any, like, (long pause) for any change in one quantity,
how that relates to the other quantity. They need to think about, um, just because your
change in distance is, you know, 13 times as large as your change in time, also thinking
about it works the other way too. Your change in time also has a relationship to your change
in distance. It’s 1/13th. … Like being able to go both directions.
The interviewer asked Samantha to express her point in the general context of “a
varies at a rate of r with respect to b.” She clarified:
They would need to understand that it varying at a constant rate means that like any change
in a is going to be r times as large as the change in b. … They have to think the change in a
is r times as large as the change in b or change in b is 1/rth the change in a. … They have to
be able to think about both of those.
After having recognized that analogous thinking is required to solve all 12 pairs of
tasks from the instructional sequence, Tallman and Weaver prompted Samantha to
summarize in writing the meaning for rate of change that might enable a student to
reason productively about all tasks in the instructional sequence. We include her
response in Fig. 8.13.
As Samantha’s written statement in Fig. 8.13 demonstrates, at the conclusion of
the teaching experiment, she was able to express her conception of a constant rate
of change as a proportional relationship between changes in covarying quantities’
measures and articulate the mental imagery, actions, and operations that comprised
her meaning for constant rate of change.
Fig. 8.13 Samantha’s expressed meaning for a constant rate of change. (Tallman & Weaver,
2018, p. 480)
The case of Samantha demonstrates the grasp of consciousness that results from
having constructed a reflected scheme, but also exemplifies an experimental context
in which an observer can reliably infer that the subject has reflected on the products
of prior reflecting abstractions. Qualitative methodologies compatible with con-
structivist epistemology, particularly teaching experiments (Steffe & Thompson,
2000), facilitate these insights while maintaining high levels of scientific rigor.
I mplications for Supporting Students’ Learning

in Teaching Contexts
In this section, we discuss the implications of reflected abstractions for students’

learning, including implications for teachers designing curricula and learning expe-
riences to support students’ construction of robust mathematical meanings. First,
we begin with the assumption that our goal is for students to do more than imitate
teachers’ prescribed solution methods and memorize a wide array of mathematical
facts. Instead, our aim is to develop students’ flexible mathematical thinking. That
is, we intend to foster the development of students’ operative thought relative to
mathematical situations. This means that students develop the ability to reason
about the ideas of the course at the level of reflection, i.e., that they are acting with
a level of controlled, conscious awareness of their reasoning (Thompson, 1985).
It is in this spirit that Thompson (1985) argued that “a constructivist mathematics
curriculum” must be problem-based. Note that the idea of a “problem” (to a con-
structivist) means that the task at hand is one with a solution that is non-routine for
the individual. Additionally, “in developing a scheme, a student must develop
images of having reasoned in particular ways, meaning she must develop recollec-
tions of ‘momentary states’ in having reasoned” (Thompson et al., 2014, p. 10).
Students must repeatedly construct specific meanings for their conceptions to
become stable (Harel, 2007, 2008a, b; Thompson, 2013). That is, an individual con-
structs a scheme related to a mathematical idea only when they attempt to reason
with and about that idea and has opportunities to reflect on those experiences.
Students do not construct powerful, flexible schemes for mathematical ideas quickly
or easily. They construct schemes over long periods of time with repeated opportu-
nities to develop related imagery, to abstract properties of their reasoning, and to
organize their understanding of related mathematical ideas via reflected abstraction.
Thus, students’ mathematical experiences must be strategically guided by les-
sons, units, and courses that afford coherent connections across mathematical topics
and allow students multiple opportunities to construct meanings for important ideas.
Additionally, the tasks students encounter must provide them with genuine oppor-
tunities to develop plans, enact those plans, and reflect on their experiences. In
designing curricula, it is simply not sufficient to curate a series of examples and
exercises, even if the curriculum designers themselves believe that the activities
unfold in a logical, structured manner. What matters is the connections students
make and what students abstract (if anything) from the experiences promoted by
curricula and instruction. In a teacher-centered mathematics classroom where les-
sons are predominantly presented as lectures, the teacher does the thinking while
students are positioned only to engage in empirical and pseudo-empirical abstrac-
tions that are insufficient for their construction of operative mathematical schemes.
In the long run, this removes most opportunities for students to engage in reflective
abstraction and develop transferable mathematical reasoning abilities.
It is understandable why procedural instruction might not seem problematic for
teachers who witness many of their students perform acceptably on assessments of
mathematical skills. Unless those teachers either follow students into future math-
ematics courses or ask students to apply their knowledge in novel contexts, it is easy
to miss how short-lived those successes might be. In many cases, students learning
to mimic procedures through memorization are making no abstractions at all and
are not engaging in anything resembling genuine mathematical reasoning. At best,
some students might be seeing patterns in their work and making pseudo-empirical
abstractions from those patterns. But, in general, what they learn is highly compart-
mentalized and dependent on the figurative elements of the tasks they encounter.
For example, students trained to work with slope as “rise over run” and who calcu-
late slopes or draw the graphs of linear functions by counting off horizontal and
vertical displacements are not positioned to answer questions like, “If the constant
rate of change of y with respect to x is ¾, and if (x, y) = (3, 7) is a point on the line,
where does the line cross the y-axis?” Seeing “¾” as communicating two displace-
ments (a rise of 3 and a run of 4) and attempting to count off those displacements
will not successfully answer the question. Similarly, Lobato and Thanheiser (2002)
showed that students who were taught to calculate the slope of a line passing through
two points did not necessarily apply this skill when encountering tasks requiring
them to find slopes in other contexts (such as the slope of a wheelchair ramp).
Finally, memorizing procedures for calculating slopes and sketching graphs of lin-
ear functions in the Cartesian plane are poor preparations for working with a con-
stant rate of change in other coordinate systems (see the discussion around Fig. 8.11).
If classroom experiences are not supporting reflecting abstraction, students are not
likely to build schemes that facilitate stable and predictable reasoning with mathe-
matical ideas. Furthermore, if teachers are not providing opportunities for students
to reflect on their meanings, become aware of their reasoning, and make conscious
connections between their reasoning in different contexts (i.e., supported in con-
structing reflected abstractions), then students are not likely to develop the kind of
flexible and conscious meanings and generalized ways of thinking that enable them
to assimilate novel experiences to previously constructed ways of reasoning.
Let us return to Lobato’s AOT theory. Lobato emphasized that what students do
in situations is dependent on the way they have come to think about certain ideas
(the meanings they have constructed), and when we see them behave in “incorrect”
ways, we can find the roots of those behaviors in past learning experiences. This
means we can use this feedback to think carefully about our lessons, instructional
goals, learning trajectories, and so on to help students focus their attention on salient
features of mathematical objects and processes and abstract the conclusions we
expect will be productive for them in a variety of mathematical contexts. However,

as Lobato and Siebert (2002) emphasize, “simply trying to ‘teach for understand-
ing’ in the hopes that students will transfer that understanding is far too general a
guide to be useful in designing instruction” (p. 113). Thompson (1985) echoed
this point:
We must make explicit the nature of the knowledge that we hope is constructed and make a
case that the chosen activities will promote its construction. How one characterizes the
knowledge to be constructed, and how the selected activities might lead to its construction,
will greatly influence the choice of activities.
The general stance taken here is that anything we might wish to call mathematical
knowledge is a structure of thinking—the structure is a structure of processes. Also, it is
assumed that mathematical structures arise from abstracting the invariant features of one’s
thinking in problematic situations … The task of the curriculum developer is to select prob-
lematic situations that provide occasions for students to think in ways that have a generative
power in regard to the objectives of instruction. (p. 192)
Finally, we remind readers of an important implication of how knowledge develops

from the perspective of genetic epistemology:
Constructivism says that whatever sense people make of their experience, they construct
that sense themselves—regardless of what anyone else does to influence it. What a person
actually ends up knowing can be influenced by what others do, but communication happens
only through interpretation … It can be appropriate for teachers to describe explicitly to
their students the understandings they hope their students will have. But after so describing
those understandings, and after teaching for them, teachers should not presume that what
students understand is what was intended. Instead, they should listen for cues as to what
sense students have made of what was said or done, including asking for students’ interpre-
tations of it. (Thompson, 2002, p. 193)
In other words, teachers should not assume that students who participate and pay
attention walk away with the meanings and skills intended (Dorko, 2019; Lew et al.,
2016). The reality may be quite different. In the next section, we discuss the knowl-
edge and perspective that enables teachers to effectively plan and deliver effective
lessons. See also Carlson, Bas-Ader, O’Bryan, and Rocha (Chap. 9) for a detailed
discussion of teachers decentering relative to student thinking during instruction.
I mplications for Researching and Supporting Teachers’

Pedagogical Content Knowledge
In this section, we discuss the important role of reflected abstraction in teachers’

development of pedagogical content knowledge (PCK), which refers to the charac-
ter of disciplinary knowledge needed for enacting effective pedagogies that support
students' construction of in constructing productive meanings.
Shulman (1986, 1987) proposed a theoretical framework for how content-related
knowledge is organized in the minds of teachers. Central to this framework is the
notion of PCK. Inspired by Dewey’s (1902) proposal for educators to “psychologize
the subject matter,” Shulman popularized the notion that pedagogical knowledge is
shaped by one’s comprehension of academic content. Shulman’s conception of

PCK encapsulated Dewey’s proposal for educational psychologists to explicate the
experiential basis of the facts, concepts, and ways of reasoning that comprise the
subject matter of a discipline and to identify the capacities and proficiencies that
these facts, concepts, and ways of reasoning enable (Dewey, 1902). To Dewey, psy-
chologizing the subject matter meant to explore the experiential reality of learners
as they engage with academic subjects and to characterize this reality in psychologi-
cal terms. Psychologizing the subject matter is a cognitive activity that results in
one’s capacity to provide the conditions for students to engage in the experiences
necessary to stimulate their intellectual growth in a particular direction. The act of
psychologizing the subject matter is fundamentally an accommodation of cognitive
schemes motivated by one’s consideration of the experiential basis for the content
of an academic subject.
Conceptions of PCK consistent with its Deweyan foundations are based on the
notion that the psychologization of subject matter is the essential process by which
content knowledge is endowed with pedagogical utility (e.g., Silverman &
Thompson, 2008; Tallman, 2015, 2021). From this perspective, subject matter
knowledge, once psychologized, necessarily enables effective pedagogical action.
This conception of PCK prioritizes educators' identification of the features of psy-
chologized subject matter schemes that facilitate pedagogical efficacy, as well as
understanding how these characteristics might be engendered in pre- and in-service
teacher education. From this perspective, a mathematics teacher educator is respon-
sible for supporting pre-service teachers’ construction of mathematical knowledge
structures with specific characteristics that permit insight into the experiential basis
of productive mathematical understandings and requisite ways of thinking. This
(pedagogical) content knowledge might then facilitate teachers’ implementation of
instructional actions to purposefully engage students in the experiences necessary to
promote their construction of targeted conceptions.
We propose that constructing mathematical schemes at the reflected level of
thought is essential for teachers’ development of PCK. Thompson (1985) expressed
a similar imperative, writing that “Ideally, teachers should not only possess the cog-
nitive structures that the curriculum aims for in the students but possesses them at a
reflected level. They should understand the subject matter both intuitively and for-
mally” (p. 222). The awareness of the contents of one’s mathematical schemes
entailed in reflected abstraction is the essential characteristic of a teacher’s mathe-
matical knowledge that facilitates their pedagogical potential (Silverman &
Thompson, 2008; Tallman, 2015, 2021; Tallman & Weaver, 2018; Thompson, 1985,
p. 222–23). This awareness motivates a teacher’s attention to epistemology, neces-
sitates their construction of epistemic ways of understanding (Liang, 2020), and
promotes their consideration of consequential ways of thinking and reasoning
(Tallman & Frank, 2020).
In its most general description, a mathematics teacher is responsible for provid-
ing opportunities for students to engage in the conceptual activity required for their
construction of productive meanings. Accomplishing this goal demands that the
teacher’s actions be deliberately informed by an understanding of the cognitive
mechanisms of mathematics learning so that these mechanisms can be purposefully

engendered through instruction. This essential obligation cannot be fulfilled if a
teacher is aware only of the behavioral capacities that result from applying their
mathematical schemes in specific contexts. A teacher operating with this type of
awareness seeks to support students in becoming fluent at engaging in the sequence
of actions by which they can successfully complete routine tasks. Alternatively, a
teacher cognizant of the mental actions and operations that comprise their mathe-
matical schemes is positioned to reflect on the conceptual process by which students
might construct similar conceptions (Tallman, 2015).
Effective instruction additionally necessitates that the teacher’s actions be
informed by an image of how students might develop particular mathematical con-
ceptions. Constructing this image entails (1) clarifying what it means to understand
a specific mathematical idea; (2) discerning the actions, abstractions, and general-
izations in which a student must engage to construct this meaning; and (3) designing
curricular artifacts to provoke such cognitive activity. These three components are
consistent with Simon’s (1995) articulation of a hypothetical learning trajectory
and Thompson’s (2008) description of conceptual analysis. Engaging in conceptual
analysis to construct a hypothetical learning trajectory for a mathematical idea
relies upon one’s awareness of the mental actions and conceptual operations that
comprise their scheme for the idea. Moreover, the second and third aspects of a
hypothetical learning trajectory necessitate a teacher’s consideration of students’
potential experiences as they engage with and progress through a curriculum
designed to promote particular conceptual activity, thus entailing an explication of
the experiential basis of targeted mathematical conceptions, which comprises the
first aspect of a hypothetical learning trajectory (Thompson, 1985). Conceptual
analysis is, therefore, the means by which one psychologizes mathematical subject
matter and, as previously emphasized, psychologizing the subject matter is the pro-
cess that endows content knowledge with pedagogical efficacy. Constructing
reflected abstractions is the essential cognitive mechanism that facilitates a teach-
er’s psychologization of mathematical subject matter, and is thus foundational to
constructing PCK.
Another potential implication of a teacher having constructed reflected abstrac-
tions is that the ensuing awareness of the conceptual contents of their mathematical
schemes motivates and enables their construction of second-order models28 (Steffe
et al., 1983) of students’ mathematical meanings. Epistemic ways of understand-
ing29 (Beth & Piaget, 1966) are generalizations of these second-order models, the
28
Steffe and Thompson (2000) distinguished first- and second-order observers as follows: “first-
order observers address what someone understands, while second-order observers address what
they understand about what the other person could understand” (p. 303, italics in original).
29
Thompson (2002) defined an epistemic student not as “anyone in particular. Rather, we speak
generically of a person, as in ‘Suppose a person understands fractions as ‘so many out of so
many …” Images of this type allow us to propose ways of thinking that are not specific to any one
person, yet they are not ‘disembodied cognizing’” (p. 195). These images are what we mean by
epistemic ways of understanding. Developing images of multiple ways of understanding students
construction of which requires a teacher to maintain a particular orientation to their

interactions with students. Specifically, a teacher must be committed to inferring the
conceptual operations that inform students’ language and actions and to engage
them in experiences that make such insights possible, robust, and viable. Such stra-
tegic interaction is informed by one’s awareness of the mental actions and opera-
tions that comprise their mathematical schemes, as this cognizance serves as a basis
of comparison for how students might understand a mathematical idea.
Models of epistemic ways of understanding additionally empower a teacher to
design instruction with an anticipation of how their conceptual goals for students’
learning might result from a particular progression of cognitive experiences. This
capacity resonates with how Dewey (1902) described an educator’s essential com-
mitment: a teacher “is concerned with the subject matter of the science as represent-
ing a given stage and phase of the development of experience” (p. 30, 1902, emphasis
in original). Constructing epistemic ways of understanding is thus necessary for
developing hypothetical learning trajectories, particularly their second and third
components.
As we previously described, Harel’s duality principle suggests the process by
which ways of thinking might mature: “Students [or teachers] develop ways of
thinking through the production of ways of understanding, and, conversely, the
ways of understanding they produce are impacted by the ways of thinking they pos-
sess” (Harel, 2008b, p. 899). The cognitive processes that underlie the duality prin-
ciple, however, are insufficiently specified to inform interventions intended to
advance teachers’ ways of thinking or to make them consciously aware of their criti-
cal role in supporting students’ construction of coherent ways of understanding.
Piaget’s notion of reflected abstraction might attenuate this limitation. If a teacher
can develop productive ways of understanding specific mathematical ideas and
become cognizant of the mental actions and operations that constitute them, then
they are positioned to abstract common features of their mathematical schemes and
to recognize how particular ways of thinking contributed to their construction and/
or facilitate their application. This kind of reflective activity is indispensable for
teachers because it equips them with an image of how they might leverage specific
ways of thinking in their teaching to support students’ construction of coherent
mathematical meanings.
inal Comments on Genetic Epistemology’s Place

F
in Educational Psychology and Mathematics Education
The version of Piaget’s genetic epistemology imported into the Anglophone acad-
emy was, at best, a simplified account of the elaborate theory Piaget spent decades
developing and refining (von Glasersfeld, 1995, 2007). As a result of his dense and
in a class could have is highly useful in helping teachers react productively in the moment by help-
ing them recognize the potential way of understanding a student could have that is manifesting in
particular behaviors.
convoluted writing style, Piaget bears some responsibility for the selective interpre-
tation (and often misinterpretation) of his theory by American educational psychol-
ogists, who were far more interested in his stages of development (i.e., sensorimotor,
preoperational, concrete operational, and formal operations) and much less con-
cerned with the cognitive mechanisms that facilitate a knowing subject’s transition
from one stage to the next. Indeed, detailing these cognitive mechanisms was cen-
tral to Piaget’s epistemological ambitions. The dominant paradigms within the
American academy prevented scholars from appreciating the value of the central
constructs within Piaget’s theory. Thorndike’s connectionism, for instance, was far
more influential to the practice and scholarship of education in the 1920s—when
Piaget began his psychological research program—than was Dewey’s psychology
of school subjects. Dewey’s psychology was compatible with Piaget’s genetic epis-
temology in terms of its fundamental premises and objectives, which included a
shared commitment to elevating the experience of the learner to the status tradition-
ally afforded in educational research to the canonical subject matter of the curricu-
lum. Consistent with the epistemological aims of Piaget’s research program, Dewey
was interested in explicating the experiential basis of the facts, concepts, and ways
of reasoning that comprise the subject matter of a discipline, and to identify the
capacities and proficiencies that these facts, concepts, and ways of reasoning enable
(Dewey, 1902).
In contrast, Thorndike’s application of the universal principles of his connection-
ist theory of learning to academic content areas conflicted with Dewey’s belief that
the subject matter of the curriculum achieves its meaning and significance only with
reference to the past, present, and potential experience of the learner, and that any
psychology of school subjects must therefore “[a]bandon the notion of subject-
matter as something fixed and ready-made in itself, outside the child’s experience”
(Dewey, 1902, p. 16). Thorndike sought to establish his psychology as a scientific
enterprise by aspiring to conduct his research with an experimental precision char-
acteristic of the natural sciences. This commitment, operationalized in the form of
reducing the complexity of thought and reason to associations between sensations
and impulses, established a foundation for the neobehaviorism of Edward Tolman,
Clark L. Hull, and Edwin R. Guthrie and later the radical behaviorism of John
B. Watson and B. F. Skinner (Kilpatrick, 1992). Collectively, these antimentalistic
paradigms were in conflict with Dewey’s psychology and Piaget's genetic episte-
mology, and not only contributed to the dormancy of Dewey’s ideas for half a cen-
tury but also created an academic culture skeptical that qualitative analyses of the
cognitive processes of children can be scientifically rigorous (Shulman, 1974).
In its motivation to establish a scientific approach to psychology, behaviorism
supplied experimental methods and an analytic focus that would influence the trajec-
tory of educational research from the 1950s through the 1970s. The radical behavior-
ists’ critique of Titchenerian introspection as a tool for the qualitative analysis of
conscious states—and as a valid source of psychological knowledge generally—
combined with their strict focus on the objective data of behavior, contributed to the
broad appeal of classical experimental designs in the early research on teaching and
learning (Shulman, 1970). The actions exhibited by students and teachers in response
to instructional, pedagogical, or environmental treatments comprised the primary

analytical unit in this anti-cognitive genre of educational research. Within the behav-
iorist paradigm, there was no relevance for the central ideas of Piaget’s genetic epis-
temology, especially opaque cognitive processes like reflective abstraction.
von Glasersfeld (2007) cogently expressed concern about the long-term detri-
mental effects of behaviorism within the American academy:
For about half a century behaviorists have worked hard to do away with ‘mentalistic’
notions … It is up to future historians to assess just how much damage this mindless fashion
has wrought. … Since behaviorism is by no means extinct, damage continues to be done.
(von Glasersfeld, 2007, p. 13)
This continuing damage includes a reluctance among many mathematics education

researchers to employ highly abstract theoretical constructs within Piaget’s genetic
epistemology to study various phenomena related to the teaching and learning of
mathematics. Latent anxieties about the scientific respectability of educational
research—which throughout the twentieth century contributed to the broad appeal
of Thorndike’s connectionism and the science of behavior it helped popularize—
continue to contribute to a disincentive for mathematics educators (especially new
researchers) to pursue the difficult path of understanding Piaget’s insights about
learning and applying this understanding in pursuit of the design-science enterprise
of our field (Cobb, 2007). A century after Piaget began his scientific approach to the
fundamental questions of epistemology, research in mathematics education based
on the central insights of Piaget’s theory remains the road less traveled. However,
we have found Piaget’s genetic epistemology, and specifically reflected abstraction,
has significant affordances for mathematics educators' pursuit of important ques-
tions, such as: What meanings do students construct while engaged in mathematical
experiences? What are the implications of these meanings for future learning? What
connections do students make between prior and current learning experiences? How
can we best support students in making connections between learning experiences
and coming to understand mathematical ideas in ways that allow them to flexibly
adapt that reasoning to novel contexts? In what ways does constructing mathemati-
cal schemes at the reflected level of thought enable teachers’ effective implementa-
tion of pedagogical practices?
There are several epistemological positions on offer to the mathematics educator
deriving from theoretical traditions such as radical constructivism, behaviorism,
sociocultural theory, situated cognition, cognitive information processing theory,
cognitive psychology, experimental psychology, embodied cognition, social cogni-
tive theory, and social constructivism. The epistemological orientations that ensue
from these and other theoretical perspectives serve the same purpose: to orient and
constrain the types of explanations of particular phenomena a researcher constructs
(Thompson, 2002, p. 192). It is in this sense that we have experienced Piaget’s
notion of abstraction as valuable. Reflecting abstraction as a theoretical construct
that explains the construction of operative cognitive structures, and reflected abstrac-
tion as a construct that describes the characteristic of schemes that enable purpose-
ful reasoning afforded by the knower’s cognizance of the coordinated actions
organized within these schemes, guides both the explanations a researcher gives to
account for learning and the experiences one designs to support learning.
Acknowledgements Much of what we express in this chapter has its origins in conversations we
have had with Patrick Thompson. Throughout this chapter, we acknowledge where our writing was
informed by or connects with Pat Thompson’s published work. However, these references do not
adequately represent his influence on our understanding of Piaget’s theory. We are indebted to Pat
for his many years of helping us navigate the complexities of Piaget’s genetic epistemology. This
chapter would not have been possible without our benefiting from his expertise.
References
Beth, E. W., & Piaget, J. (1966). Mathematical epistemology and psychology (W. Mays, Trans.).
D. Riedel.
Carlson, M., Jacobs, S., Coe, E., Larsen, S., & Hsu, E. T. (2002). Applying covariational rea-
soning while modeling dynamic events: A framework and a study. Journal for Research in
Chapman, M. (1988). Constructive evolution: Origins and development of Piaget’s thought.
Cambridge University Press.
Cobb, P. (2007). Putting philosophy to work: Coping with multiple theoretical perspectives. In
F. K. Lester (Ed.), Second handbook of research on mathematics teaching and learning (Vol. 1,
pp. 1–38). Information Age Publishing.
Damasio, A. (1994). Descartes’ error: Emotion, reason, and the human brain. Penguin Books.
Dewey, J. (1902). The child and the curriculum. University of Chicago Press.
Dorko, A. (2019). Professors’ intentions and student learning in an online homework assignment.
In A. Weinberg, D. Moore-Russo, H. Soto, & M. Wawro (Eds.), Proceedings of the 22nd
annual conference on research in undergraduate mathematics education (pp. 172–179).
Dubinsky, E. (1991). Reflective abstraction in advanced mathematical thinking. In D. Tall (Ed.),
Advanced mathematical thinking (pp. 95–123). Kluwer.
Gallagher, J. M., & Reid, D. K. (2002). The learning theory of Piaget and Inhelder. iUniverse.
Ginsburg, H., & Opper, S. (1988). Piaget 's theory of intellectual development. Prentice-Hall.
Gruber, H. E., & Vonèche, J. J. (1977). The essential Piaget: An interpretative reference and guide.
Routledge and Kegan Paul.
Harel, G. (2007). The DNR system as a conceptual framework for curriculum development and
instruction. In R. Lesh, J. J. Kaput, & E. Hamilton (Eds.), Foundations for the future in math-
ematics education. Routledge.
Harel, G. (2008a). DNR perspective on mathematics curriculum and instruction part I: Focus on
proving. Zentralblatt für Didaktik der Mathematik, 40, 487–500.
Harel, G. (2008b). A DNR perspective on mathematics curriculum and instruction, part II: With
reference to teacher’s knowledge base. Zentralblatt für Didaktik der Mathematik, 40, 893–907.
Harel, G., & Koichu, B. (2010). An operational definition of learning. Journal of Mathematical
Behavior, 29, 115–124.
Hardison, H. L., & Lee, H. Y. (2020). Funky protractors for exploring angle measure. Mathematics
Teacher: Learning and Teaching PK-12, 113(3), 229–232. https://doi.org/10.5951/
MTLT.2019.0214
Jackiw, N. (2011). The geometer’s sketchpad v5. 0. Key Curriculum Press.
Johnson, H. L. (2015). Secondary students’ quantification of ratio and rate: A framework for reason-
ing about change in covarying quantities. Mathematical Thinking and Learning, 17(1), 64–90.
Kilpatrick, J. (1992). A history of research in mathematics education. In D. A. Grouws (Ed.),
Handbook of research on mathematics teaching and learning (pp. 3–38). National Council of
Teachers of Mathematics.
Lew, K. M., Fukawa-Connelly, T., Mejía-Ramos, J. P., & Weber, K. (2016). Lectures in advanced
mathematics: Why students might not understand what the professor is trying to convey.
Liang, B. (2020). Theorizing teachers’ mathematical learning in the context of student-teacher
interaction: A lens of decentering. In S. S. Karunakaran, Z. Reed, & A. Higgins (Eds.),
Proceedings of the 23rd annual conference on research in undergraduate mathematics educa-
tion (pp. 742–751). Boston.
Lobato, J. (2003). How design experiments can inform a rethinking of transfer and vice versa.
Educational Researcher, 32(1), 17–20.
Lobato, J. (2006). Alternative perspectives on the transfer of learning: History, issues, and chal-
lenges for future research. The Journal of the Learning Sciences, 15(4), 431–450.
Lobato, J. (2008). When students don’t apply the knowledge you think they have, rethink your
assumptions about transfer. In M. Carlson & C. Rasmussen (Eds.), Making the connec-
tion: Research and teaching in undergraduate mathematics (pp. 289–304). Mathematical
Association of America.
Lobato, J. (2012). The actor-oriented transfer perspective and its contributions to educational
research and practice. Educational Psychologist, 47(3), 1–16.
Lobato, J., Rhodehamel, B., & Hohensee, C. (2012). “Noticing” as an alternative transfer of learn-
ing process. Journal of the Learning Sciences, 21(3), 1–50.
Lobato, J., & Siebert, D. (2002). Quantitative reasoning in a reconceived view of transfer. Journal
of Mathematical Behavior, 21, 87–116.
Lobato, J., & Thanheiser, E. (2002). Developing understanding of ratio as measure as a foundation
for slope. Making sense of fractions, ratios, and proportions, 162–175.
Moore, K. C. (2010). The role of quantitative reasoning in precalculus students' learning central
concepts of trigonometry (unpublished doctoral dissertation). Arizona State University.
Moore, K. C. (2014). Quantitative reasoning and the sine function: The case of Zac. Journal for
Moore, K. C., LaForest, K. R., & Kim, H. J. (2016). Putting the unit in pre-service secondary
teachers' unit circle. Educational Studies in Mathematics, 92(2), 221–241.
Moore, K. C., Paoletti, T., & Musgrave, S. (2014). Complexities in students’ construction of the
polar coordinate system. The Journal of Mathematical Behavior, 36, 135–149.
Moore, K. C., Stevens, I. E., Paoletti, T., Hobson, N. L., & Liang, B. (2019). Pre-service teachers’
figurative and operative graphing actions. Journal of Mathematical Behavior, 56, 100692.
Müller, U. (2009). Infancy. In U. Müller, J. Carpendale, & L. Smith (Eds.), The Cambridge
companion to Piaget (Cambridge companions to philosophy) (pp. 200–228). Cambridge
University Press.
Piaget, J. (1967). Six psychological studies. Random House.
Piaget, J. (1970). Genetic epistemology. W. W. Norton & Company Inc.
Piaget, J. (1971). Biology and knowledge. University of Chicago Press.
Piaget, J. (1977). Psychology and epistemology: Towards a theory of knowledge. Penguin Books.
Piaget, J., & Duckworth, E. (1968). On the development of memory and identity (Vol. 2). Clark
University Press.
Saldanha, L., & Thompson, P. W. (1998). Re-thinking covariation from a quantitative perspective:
Simultaneous continuous variation. In Paper presented at the annual meeting of the psychology
of mathematics education-North America. North Carolina State University.
Schwartz, J. L. (1988). Intensive quantity and referent transforming arithmetic operations. In
J. Hiebert & M. Behr (Eds.), Number concepts and operations in the middle grades (Vol. 2,
pp. 41–52). National Council of Teachers of Mathematics.
Shulman, L. S. (1970). Reconstruction of educational research. Review of Educational Research,
40(3), 371–396.
Shulman, L. S. (1974). The psychology of school subjects: A premature obituary? Journal of
Research in Science Teaching, 11(4), 319–339.
Shulman, L. S. (1986). Those who understand: Knowledge growth in teaching. Educational

Researcher, 15(2), 4–14.
Shulman, L. S. (1987). Knowledge and teaching: Foundations of the new reform. Harvard
Educational Review, 57(1), 1–22.
Silverman, J., & Thompson, P. W. (2008). Toward a framework for the development of mathemati-
cal knowledge for teaching. Journal of Mathematics Teacher Education, 11, 499–511.
Simon, M. (1995). Reconstructing mathematics pedagogy from a constructivist perspective.
Journal for Research in Mathematics Education, 26, 114–145.
Simon, M. A., & Placa, N. (2012). Reasoning about intensive quantities in whole-number
multiplication? A possible basis for ratio understanding. For the Learning of Mathematics,
32(2), 35–41.
Smith, J. P., III, & Thompson, P. W. (2008). Quantitative reasoning and the development of alge-
braic reasoning. In J. J. Kaput, D. W. Carraher, & M. L. Blanton (Eds.), Algebra in the early
grades (pp. 95–132). Lawrence Erlbaum.
and essential elements. In A. E. Kelly & R. A. Lesh (Eds.), Handbook of research design in
mathematics and science education (pp. 267–307). Erlbaum.
Tallman, M. (2015). An examination of the effect of a secondary teacher’s image of instructional
constraints on his enacted subject matter knowledge. Unpublished Ph.D. dissertation, School
of Mathematical and Statistical Sciences, Arizona State University.
trigonometric functions. Journal of Mathematical Behavior, 62, 100869.
Tallman, M. A., & Frank, K. M. (2020). Angle measure, quantitative reasoning, and instructional
coherence: An examination of the role of mathematical ways of thinking as a component of
teachers’ knowledge base. Journal of Mathematics Teacher Education, 23(1), 69–95.
Tallman, M. A., & Uscanga, R. (2020). Managing students’ mathematics anxiety in the context
of online learning environments. In J. P. Howard & J. F. Beyers (Eds.), Teaching and learning
mathematics online (pp. 189–216). CRC Press.
Tallman, M. A., & Weaver, J. (2018). Reflected abstraction as a mechanism for developing peda-
gogical content knowledge. In T. E. Hodges, G. J. Roy, & A. M. Tyminski (Eds.), Proceedings
of the 40th annual meetings of the north American chapter of the international group for the
psychology of mathematics education (pp. 476–483). University of South Carolina & Clemson
University.
cal problem solving: Multiple research perspectives (pp. 189–243). Earlbaum.
Thompson, P. W. (1988). Quantitative concepts as a foundation for algebra. In M. Behr (Ed.),
Proceedings of the annual meeting of the North American chapter of the international group
for the psychology of mathematics education (Vol. 1, pp. 163–170).
Thompson, P. W. (1990). A theoretical model of quantity-based reasoning in arithmetic and alge-
bra. Center for Research in Mathematics & Science Education/San Diego State University.
Thompson, P. W. (1994). The development of the concept of speed and its relationship to concepts
learning of mathematics (pp. 181–234). SUNY Press.
K. Gravvemeijer, R. Lehrer, B. V. Oers, & L. Verschaffel (Eds.), Symbolizing, modeling, and
tool use in mathematics education (pp. 191–212). Kluwer.
foundation of mathematics education. Proceedings of the annual meeting of the International
Group for the Psychology of Mathematics Education, 1, 45–64.
Thompson, P. W. (2011). Quantitative reasoning and mathematical modeling. In L. L. Hatfield,

S. Chamberlain, & S. Belbase (Eds.), New perspectives and directions for collaborative
research in mathematics education (pp. 33–57). University of Wyoming.
magnitudes: An hypothesis about foundational reasoning abilities in algebra. In K. C. Moore,
L. P. Steffe, & L. L. Hatfield (Eds.), Epistemic algebra students: Emerging models of students’
algebraic knowing. (WISDOMe Monographs) (pp. 1–24). University of Wyoming.
Thompson, P. W., & Thompson, A. G. (1992, April). Images of rate. In Paper presented at the
annual meeting of the American educational research association.
von Glaserfeld, E. (1995). Radical constructivism: A way of knowing and learning. RoutledgeFalmer.
von Glasersfeld, E. (2007). In M. Larochelle (Ed.), Key works in radical constructivism. Sense.
von Humboldt, W. (1907). Werke (Vol. 7). Leitmann.
Chapter 9
The Construct of Decentering in Research
on Mathematics Learning and Teaching
Marilyn P. Carlson, Sinem Bas-Ader, Alan E. O’Bryan, and Abby Rocha
According to Zuberbühler (2018), human communication is a social activity that

requires the combined effort of at least two participants who consciously and intention-
ally cooperate to construct the meaning of their interaction. Humans utilize multiple
modalities during communication, including the spoken word, drawings, gestures,
written symbols, etc. A key feature of intentional communication is that the person act-
ing attempts to draw another’s attention to what they consider relevant entities, both
real and imagined. The intentionality of the one doing the communication is to convey
their meanings through speaking and actions. A goal of the one receiving the commu-
nication then is to understand the meanings being conveyed. The mental processes
involved in understanding the speaking and actions of another person from the perspec-
tive of that person involve imagining how they might have been thinking. Steffe and
Thompson (2000) classified actions to build a mental model of another’s thinking from
the perspective of the other as a form of reflection they called decentering.
Characterizing communication involves describing the model each person con-
structs of the other’s thinking and the assumption that the speaker utters phrases that
are coherent and convey the meaning that the speaker intends. Musgrave and Carlson
M. P. Carlson (*) · A. E. O’Bryan

School of Mathematical and Statistical Sciences, Arizona State University, Tempe, AZ, USA
e-mail: mcarlson@asu.edu; alan.obryan@asu.edu
S. Bas-Ader
Faculty of Education, Department of Mathematics and Science Education, Istanbul Aydin
University, Istanbul, Turkey
e-mail: sinembas@aydin.edu.tr
A. Rocha
STEM Learning Center, University of Arizona, Tucson, AZ, USA
e-mail: abbyrocha@arizona.edu
https://doi.org/10.1007/978-3-031-47386-9_9
290 M. P. Carlson et al.
(2017), in a study of precalculus instructors’ meanings1 for the idea of average rate of
change, reported that instructors with weak meanings for the idea of average rate of
change had difficulty expressing a meaning that was coherent and clear enough for
others to understand. Other studies (Baş-Ader & Carlson, 2022; Carlson & Baş-Ader,
2019; Rocha, 2021, 2022; Rocha & Carlson, 2020; 2021, 2022 Teuscher et al., 1982)
report data that illustrate how a teacher’s image of understanding and learning an idea
influences a teacher’s actions, including how they respond to students’ statements and
questions. Efforts to study and characterize communication between an instructor and
a student should then consider how a teacher’s meaning for the idea under discussion
influences a teacher’s actions. Making sense of a student’s thinking will also require
that the instructor put their own understanding aside and try to understand how the
student might be thinking to have said or written what they did. Such an instructor
would be classified as engaging in decentering or acting in a decentered way. In con-
trast, if an instructor acts as if students’ thinking is identical to their own or discounts
students’ contributions to interactions, then we say the instructor is acting in a non-
decentered way (Steffe & Thompson, 2000). It is also important to consider if
researchers studying teaching or student learning are engaging in decentering. Are
they trying to model their subjects’ thinking, or are they only focused on judging the
degree to which their subjects’ meanings align with their own?
This chapter traces the historical roots of the construct of decentering. We follow
by providing an overview of the construct of decentering and its theoretical framing
in the context of mathematics education research. We then discuss the uses of decen-
tering for studying student thinking and illustrate the symbiotic relationship between
a researcher’s mathematical meanings for understanding and learning an idea and a
researcher’s decentering actions. We further describe what our data has revealed
about the role of decentering and the process by which a teacher might develop ways
of thinking for teaching an idea. We introduce a framework that illustrates the pro-
cess of acquiring ways of thinking for teaching an idea. We follow by discussing our
studies of teacher decentering and elaborate five levels of mental actions that provide
a more fine-grained description of the mental actions teachers exhibited in our inves-
tigations of teachers’ decentering when interacting with one another.
Theoretical Background, Framing, and Connections
ecentering’s Origins and Adaptation for Use in Mathematics

D
Education Research
In this section, we provide an overview of how the idea of decentering was origi-
nally observed in Piaget’s early work and how the construct was adapted for use in
mathematics education research.
1
We adopt Thompson et al.’s (2019) use of the word “meaning” in place of “understanding” to
refer to the “space of implications of an understanding,” including the images and actions that are
available, and thus influence the possible ways in which the individual might act.
9 The Construct of Decentering in Research on Mathematics Learning and Teaching 291
The Origins of Decentering in Piaget’s Genetic Epistemology
Piaget introduced the construct of decentering when studying children’s cognitive,

affective, and social development2 as they progressed through three developmental
stages. In his theory, decentering referred to a child’s gradual progress away from
egocentrism and their growing ability to coordinate multiple aspects of an object or
situation simultaneously (multiple centrations) (Piaget, 1965, 1995). During the sen-
sory-motor stage of development (the period between birth to approximately 2 years),
the child interacts directly with people and objects that are physically present. The
child is unable to mentally construct an image of another when they aren’t physically
present since the child lacks the semiotic function (the ability to create representa-
tions) in this stage. However, mental development during this stage provides a basis
for the child’s future intellectual development (Piaget, 1965, 1995; Piaget & Inhelder,
1966, 1973). In particular, through a general decentering process, the child structures
his universe during this stage. As Piaget and Inhelder (1966, 1973) explained:
In the course of the first eighteen months, however, there occurs a kind of Copernican revo-
lution, or, more simply, a kind of general decentering process whereby the child eventually
comes to regard himself as an object among others in a universe that is made up of perma-
nent objects (that is, structured in a spatio-temporal manner) and in which there is at work
a causality that is both localized in space and objectified in things. (p. 13)
As children transition to the pre-operational stage (approximately ages 2 to 7 years),

they can conceive of objects outside of their immediate experience (von Glasersfeld,
1995). Piaget’s experiments (e.g., viewing a mountain from another position) fur-
ther revealed that it is during the pre-operational stage that children develop the
semiotic function and thus begin to construct mental representations or representa-
tional thought and progress to using symbols and signs to evoke objects when they
are not physically present (Piaget & Inhelder, 1966, 1973). Prior to their transition
to a decentered state, children are egocentric (they struggle to distinguish others’
viewpoints, perspectives, and perceptions from their own – their world is strictly
about themselves), their reasoning is dominated by figurative, perceptual features of
situations (e.g., they do not understand how properties of an object like its volume
might be preserved even if its appearance changes), and they tend to focus only on
one aspect of an object or situation at a time (Piaget, 1965, 1995; Piaget & Inhelder,
1966, 1973).
As children decenter, they begin to understand and appreciate that others per-
ceive the world in unique ways and have independent feelings, motivations, and
opinions. It is through decentration that the child introduces reciprocity among the
diversity of points of view along with differentiating among them. Decentration also
allows for the ability to consider their actions as reversible and to understand that
some properties of objects are conserved under transformations. Children’s acquisi-
tion of notions of conservation exemplifies this sort of constructive processes
2
In Piaget’s theory of constructivism, cognitive, affective, and social development of a cognizing
subject are inseparable (Piaget & Inhelder, 1966, 1973).
(Ginsburg & Opper, 1988; Piaget, 1945, 1952). In an experiment about the conser-
vation of continuous quantity, the liquid in container A is poured into a narrower
container B or wider container C (Piaget & Inhelder, 1966, 1973). Younger children
describe the volume of liquid in container A increasing when poured into container
B or decreasing when poured into container C. This is because they focus only on
the figurative image of the liquid’s height in each container. They do not conceptual-
ize their actions of pouring as transformations of the state of the liquid nor as an
action that could be reversed (Ginsburg & Opper, 1988; Piaget & Inhelder, 1966,
1973). Older children, however, understand that when the liquid is poured from
container A into container B it could simply be poured back into container A and
that these actions do not alter the liquid’s volume. The child decenters these trans-
formations from the child’s actions, and thus the child is no longer constrained by
pre-operational thought nor dependent on specific figurative representations (Piaget
& Inhelder, 1966, 1973).
Adapting Decentering for Use in Mathematics Education Research
Many researchers (e.g., Baş-Ader & Carlson, 2022; Carlson et al., 2007; Steffe &
Thompson, 2000; Thompson, 2013), including us, have adopted the term “decenter-
ing” as a construct for studying mathematics teaching and for researching student
learning that extends its meaning as “a shift away from egocentrism” and “the
increasing ability to coordinate centrations in one’s reasoning” that occurs during
childhood development. Earlier in the paper, we defined “decentering” as attempt-
ing to understand another’s actions from the perspective of the other by imagining
how the other person might have been thinking. On the other hand, acting in a non-
decentered way is characterized by imputing one’s own reasoning, goals, beliefs,
worldview, or understandings onto another person to explain their actions.
Thus, it is useful to understand our use of “decentering” to be synonymous with
the notion of modeling another person’s thinking in such a way as to hypothesize
meanings, beliefs, goals, and so on that would logically result in that person’s
observable actions even if (or especially if) those meanings, beliefs, goals, and so on
differ from our own. Teachers decenter when they work to understand a student’s
meaning and reflect on what meaning a student constructed and then leverage their
image of a student’s meaning to inform their subsequent instructional actions. A
researcher’s decentering actions are similar, although the researcher typically has
more robust schemes for the idea under investigation and thus is better equipped to
understand and characterize a student’s thinking. They are also typically more sin-
gularly focused on building a model of a student’s thinking and/or testing a hypoth-
esis for advancing a student’s meaning for an idea that is a focus of a study. We will
continue to clarify the uses of decentering by providing concrete examples of how
teachers and researchers have used decentering in their work.
onnections Between Decentering and Other Theoretical

C
Constructs Used in Mathematics Education
In this section, we describe the theoretical foundations and connections between the
decentering construct as we define it and Piaget’s constructs of reflecting and
reflected abstraction, first- and second-order models for individuals’ reasoning, and
mathematical knowledge for teaching (MKT).
eflecting and Reflected Abstraction and Their Connection to Decentering

R
in the Context of Teaching
According to von Glasersfeld (1991), reflection “allows us to step out of the stream
of direct experience, to re-present a chunk of it, and to look at it as though it were
direct experience while remaining aware of the fact that it is not” (p. 90). It follows
from this conceptualization that the reflection process has two main components:
representing and abstraction. The mental act of considering a chunk of experience
and isolating it from what came before and what follows is a simple kind of abstrac-
tion, and representing is the mental act of treating the abstraction as an object in and
of itself. Representation of a previous experience then refers to the mental actions
that reconstruct the experience, thus making it conscious for the individual (von
Glasersfeld, 1991).
Piaget adopted and extended the idea of reflection in his theory of abstraction
(Simon et al., 2004; Tallman, 2021; von Glasersfeld, 1995) Among the four kinds of
abstraction (von Glasersfeld, 1991), reflecting and reflected abstraction are particu-
larly relevant to decentering.3 Reflecting abstraction is the process that “involves an
individual’s reconstruction on a higher cognitive level of the coordination of actions
from a lower level” (Tallman, 2021, Reflecting Abstraction section, para. 12), such as
the reconstruction of physical actions at the level of mental representations of those
actions. This reconstruction requires isolating those actions from their effects so that
the actions can be generalized, and this process is the source for the construction of
new knowledge or conceptions (Simon et al., 2004; Tallman, 2018). That is, the result
of reflecting abstraction is a scheme of meanings.4 It is worth noting that an individual
can have a scheme of meanings but lacks conscious awareness of the scheme to the
extent that they are unable to explain it or discuss it in general terms or consciously
manipulate symbolic representations of the scheme’s actions without needing to
3
See Ellis, Lockwood, and Paoletti (Chap. 6) and Tallman and O’Bryan (Chap. 8) for a more
detailed discussion of reflecting and reflected abstraction.
4
We often find it useful to remind readers that all types of abstractions in Piaget’s theory have the
ability to “go wrong.” That is, depending on what an individual attends to and takes away from
experiences dictates what becomes abstracted, and it is quite common for someone to develop a
scheme that, from an observer’s perspective, contains faulty reasoning and incorrect information
or is incomplete.
reimagine the coordinated actions they represent (Piaget, 1977, 2001). Conscious
awareness of one’s own reasoning comes with reflected abstraction. Tallman and
O’Bryan (Chap. 8) describe reflected abstraction as follows:
Thus, reflected abstractions enable an individual to explicitly formulate the results of prior
reflecting abstractions. Reflected abstraction, then, describes a scheme constructed at the
reflected level of thought through reflection on the results of prior reflecting abstractions.
These reflected schemes possess an essential characteristic (cognizance of projected action
coordinations, decontextualized and applicable to a generalized class of objects) that
endows the knower with the capacity to explicitly formulate meanings and strategically
apply them in a range of novel contexts. Indeed, Piaget (1977, 2001) most often associated
reflected abstraction with behavioral expressions of such capacities. For example, in
describing reflected abstraction as a “detailed description of characteristic actions” (p. 167),
Piaget associated reflected schemes with the aptitude to explicitly formulate meanings
entailed and ways of reasoning supported by the scheme. (p. 249)
Reflecting and reflected abstraction are thus critical constructs in decentering in

mathematics education research and with respect to decentering during teaching. It
is first necessary that an individual has constructed schemes for the mathematical
ideas at hand via reflecting abstraction. The individual then must have constructed a
reflected scheme as a reorganization of previously constructed schemes so that their
mathematical meanings for the ideas are coherent, organized, and refined and have
been brought into the individual’s conscious awareness. This involves reflecting on
one’s own reasoning, comparing one’s reasoning in one context to the reasoning in
another context, considering additional implications of that reasoning, etc. The pro-
cess of comparing the characteristics of one’s reasoning to the characteristics of
another’s reasoning necessarily requires that the individual has developed or is
developing conscious awareness of their own reasoning [i.e., requires reflected
abstraction].
First- and Second-Order Models and Their Connection to Decentering
Thompson (2002) summarized the important distinctions between first- and second-
order observers and between first- and second-order models initially proposed by
Steffe et al. (2009). To first understand these distinctions, Thompson emphasizes
that “any description of affairs must be done at a level of monitoring that is above
what is being described” (p. 303). For example, a researcher describing children’s
thinking within a research study is “not just acting from an image of how students
think. She is monitoring how she understands their thinking, and she has connected
it with ways of thinking about [related mathematical ideas]” (p. 303). Researchers
and teachers working with students could be actors in the situation, or they could be
acting as observers even while being part of the interaction:
If [researchers/teachers] do not reflect on aspects of an interaction that are contributed by
students, if they are fused with the situation as they have constituted it, then they are actors
in the interaction, not observers. To be an observer of another while involved in interaction,
one necessarily moves to a level of reflection. When this happens, it adds a dimension to the
social interaction — the observer is now acting purposefully and thoughtfully, using the
social interaction instrumentally. (Thompson, 2002, p. 303)
Assuming a researcher or teacher is acting as an observer (i.e., they are reflecting on

students’ contributions), they can be acting as first- or second-order observers. First-
order observers acknowledge that students can have different reasoning from the
observer and are oriented to develop an image of that reasoning. Second-order
observers are additionally oriented to consider the implications of the student’s rea-
soning and to consider if other ways of reasoning might be more beneficial. Put
another way, “first-order observers address what someone understands, while
second-order observers address what they understand about what the other person
could understand” (Thompson, 2002, p. 303).
First-order models are “models the observed subject constructs to order, compre-
hend, and control his or her experience (i.e., the subject’s 'knowledge')” (Steffe
et al., 2009, p. xvi). That is, first-order models can be thought of as someone’s per-
sonal understanding of a mathematical idea, problem solution, scenario, etc. It
describes how that person thinks about the mathematics at hand. Second-order mod-
els are “models observers may construct of the subject’s knowledge in order to
explain their observations (i.e., their experience) of the subject’s states and activi-
ties” (Steffe et al., 2009, p. xvi). In constructing a second-order model, a teacher or
researcher “[puts themself] into the position of the student and attempts to examine
the operations that [they] (the researcher) would need and the constraints [they]
would have to operate under in order to (logically) behave as the student did”
(Thompson, 1988, p. 159). The result of this construction is the researcher’s/teach-
er’s second-order model of the subject’s thinking that explains the researcher’s/
teacher’s observations of the student’s states and activities.
These distinctions are important for understanding whether teachers’ actions
indicate them operating in decentered or non-decentered ways and, if evidence of
decentering exists, how to characterize the nature of the decentering and its possi-
ble implications for supporting students’ learning. Teachers acting in non-
decentered ways (i.e., operating solely as actors in a situation) are not acting
reflectively and rely solely on their first-order models (their meanings) to guide
pedagogical decisions (Teuscher et al., 1982). “For instance, a teacher who plans a
lesson based on how she learned an idea is using her first-order model to plan the
lesson” (Baş-Ader & Carlson, 2022, p. 101). It is also possible for researchers to
judge data based on their first-order models (their meanings for ideas and methods
for solving problems). For example, Lobato and her colleagues (Lobato, 2006,
2008, 2012, 2012; Lobato et al., 2017; Lobato & Siebert, 2002) point out that tra-
ditional learning transfer studies often fail to detect transfer precisely because they
rely on students seeing the same connections between problem situations that the
researchers see, while variations in students’ prior experiences and how they con-
ceptualize what they see get overlooked. Teachers or researchers operating from
first-order models are constrained to using their own thinking and meanings to
make sense of students’ activities.5
On the other hand, decentering requires that the individual is acting reflectively
(i.e., acting as a first- or second-order observer in a situation and considering stu-
dents’ contributions to the interaction). Decentering actions are then driven by the
propensity to construct second-order models of students’ thinking and, ideally, to
use these models to inform a teacher’s or researcher’s actions. The nature of the
decentering actions and their usefulness, however, depend on whether the teacher/
researcher positions himself as a first- or second-order observer.
As necessary as they are, second-order models of knowing made by a first-order
observer provide only weak guidance to a teacher, developer, or researcher. The only
thing they can draw from them is that what students end up knowing comes out of the
sense they make of teaching and bears no necessary relationship with what the teacher,
developer, or researcher intended. Second-order models of knowing made by a sec-
ond-order observer, however, can provide strong guidance for the teacher, developer,
or researcher who has developed a vocabulary and system of constructs to describe
students’ conceptual schemes together with transformations, reorganization, or other
modifications in them (Thompson, 2002, p. 304).
A teacher or researcher constructs a second-order model through decentering
actions if they wonder, “How might the student be thinking to act as they did?”
which necessarily involves interpreting a student’s actions through a model of the
student rather than their (the teacher’s/researcher’s) own cognitive schemes.
However, the ultimate usefulness of this model depends on the degree to which the
observer possesses a rich, coherent system of meanings and ways of thinking such
that the teacher or researcher can reflect not just on how a student understands an
idea but also on the implications and usefulness of that understanding for the stu-
dent’s future mathematical experiences. In other words, the nature of the teacher’s/
researcher’s first-order model matters in providing capacity for constructing second-
order models of others’ meanings and considering the implications of those mean-
ings. We will demonstrate examples of this in later sections.
Later in this chapter, we discuss a framework developed by Baş-Ader and Carlson
(2022) that characterizes five levels of decentering actions exhibited by teachers
when interacting with students when teaching. We hypothesize and describe con-
nections between these decentering actions and teachers’ mathematical meanings
for teaching (MMT), their level of awareness of their own mathematical meanings,
and how they position themselves as observers while interacting with their students.
5
We want to emphasize that we do not intend to convey that first-order models are not useful or
important for decentering. In later sections we discuss the nature of a teacher’s/researcher’s first-
order models that are necessary for and support constructing second-order models. The teacher’s/
researcher’s first-order model is used to comprehend and control their experiences (including
experiences of interacting with students or research subjects). A person cannot abandon their first-
order model because it enables and constrains what they can observe, notice, and reflect on when
engaging with students or research subjects. We however caution assuming that one’s first-order
model characterizes how others think.
Mathematical Knowledge for Teaching and Its Connection to Decentering
After having introduced the idea of teachers acting as first- or second-order observ-
ers in an interaction and the importance of constructing second-order models of
students’ thinking, it is natural to wonder what enables researchers and teachers to
construct these models. In this section, we discuss the kind of specialized knowl-
edge necessary for constructing second-order models of students’ thinking and
positioning oneself as a second-order observer while doing so.
Key Developmental Understanding (KDU) and Pedagogical Understanding
Silverman and Thompson (2008) “propose a theoretical framework that extends a

constructivist perspective to include the development of MKT [Mathematical
Knowledge for Teaching]” to address the need for a theoretical foundation for
studying teaching and teacher development (p. 501). Silverman and Thompson
argue that content knowledge alone is insufficient for ensuring high-quality teach-
ing. Instead, a teacher’s MKT is based first in what Simon (2006) called a key devel-
opment understanding (KDU) or “a way to think about understandings that are
powerful springboards for learning, and hence are useful goals of mathematics
instruction” (Silverman & Thompson, 2008, p. 502). Developing a KDU involves
reflective abstraction whereby new knowledge is constructed and connected to
existing knowledge through learning experiences designed to promote those under-
standings (perhaps in teacher preparation courses or professional development set-
tings). Having a KDU is not enough, however. MKT develops when a teacher
conceptualizes the pedagogical power of their KDU.
As a teacher thinks about the content to be taught, she envisions a student (other
than the teacher) working through the material, easing through some problems and
stumbling over others. The entire time, the teacher must ask herself “What must a
student understand to create the understanding that I envision?” and “What kinds of
conversations might position one to develop such understandings?” The prospective
teacher must put herself in the place of a student and attempt to examine the opera-
tions that a student would need and the constraints the student would have to operate
under in order to (logically) behave as the prospective teacher wishes a student to
do. This is reflective abstraction [in particular, reflected abstraction].
A KDU might be viewed as a pedagogical action, where action is used in the
Piagetian sense. Teachers are engaged in pedagogical actions when they wonder,
“What might I do to help students think like what I have in mind?” Their question
is posed in a domain-specific manner, such as “How might I help my students think
about logarithms as an accelerated condensing and recoding of the number line?”
The development of MKT involves separating one’s own understanding from the
hypothetical understanding of the learner (Steffe, 1983). When a person views a
pedagogical action as if she is not an actor in the situation (even though she is), and
when the person can separate herself from the action (and thereby reflect on it), the
pedagogical action has been transformed into a pedagogical understanding

(Silverman & Thompson, 2008, p. 508).
There are a few important points of emphasis from this passage connecting
MKT, reflective abstraction, and decentering. Silverman and Thompson
(2008) describe how a KDU with pedagogical potential gains pedagogical power
when a teacher reflects on their students’ current meanings while hypothesizing
activities and interventions they conjecture for advancing students’ meanings. A
teacher is decentering as they “separate one’s own understanding from the hypo-
thetical understanding of the learner,” but it differs from decentering actions that
involve a teacher constructing a second-order model of a student’s thinking while
directly interacting with a student.6 A teacher’s images of students’ thinking relative
to an idea emerge from one-on-one interactions with individual students. As a
teacher (or researcher) engages with different students, she might emerge with
images of multiple hypothetical students. Thus, when we talk about decentering in
this paper, we are not talking solely about decentering actions that one engages in
when interacting directly with another. Decentering can occur in all phases of plan-
ning for, engaging in, and reflecting on the result of instruction (or research) with
students. Silverman and Thompson also emphasize that the development of MKT
involves teachers engaging in reflecting abstraction on their KDU (the product of
which is a reflected abstraction or a reflected scheme) and that this process is moti-
vated by attempts at decentering.
pistemic Students Emerge from Conceptual Analysis

E
and Second-Order Models
Conceptual analysis, described by von Glasersfeld (1991) and elaborated by Steffe

(1983) and Thompson (2002, 2011), addresses teachers’/researchers’ “need to
describe what students might understand when they know a particular idea in vari-
ous ways” (Thompson, 2011, p. 43). The goal of conceptual analysis is to hypoth-
esize the mental operations a person engages in as they reason about a situation
based on their personal image of that situation. von Glasersfeld used conceptual
analysis both to generate models of how others understand certain ideas and to gen-
erate meanings for ideas that would be beneficial for students in reasoning about
those ideas in powerful ways (and why those meanings are especially beneficial).
Thompson (2011) writes that:
Finally, as illustrated in this papers [sic] first part, conceptual analysis can be employed to
describe ways of understanding ideas that have the potential of becoming goals of instruc-
tion or of being guides for curricular development. It is in this regard that conceptual
6
Note, however, that images of these hypothetical students are often formed from past experiences
working with actual students in addition to engaging in conceptual analysis. We will discuss these
connections in the next section.
a nalysis provides a method by which to construct and test a foundation of mathematics

education in the same way that people created a foundation of mathematics.
In summary, conceptual analysis can be used in four ways:
(1) in building models of what students actually know at some specific time and what they
comprehend in specific situations,
(2) in describing ways of knowing that might be propitious for students’ mathematical
learning, and
(3) in describing ways of knowing that might be deleterious to students’ understanding of
important ideas and in describing ways of knowing that might be problematic in specific
situations.
(4) in analyzing the coherence, or fit, of various ways of understanding a body of ideas.
Each is described in terms of their meanings, and their meanings can then be inspected
in regard to their mutual compatibility and mutual support.
I find that conceptual analysis, as exemplified here and practiced by Glasersfeld,
provides mathematics educators with an extremely powerful tool. It orients us to
providing imagistically-grounded descriptions of mathematical cognition that cap-
ture the dynamic aspects of knowing and comprehending without committing us to
the epistemological quagmire that comes with low-level information processing
models of cognition (Cobb, 1987; Thompson, 1990). Conceptual analysis provides
a technique for making concrete examples, potentially understandable by teachers,
of the learning trajectories that Simon (1995) calls for in his re-conceptualization of
teaching from a constructivist perspective and which Cobb and his colleagues
employ in their studies of emerging classroom mathematical practices (Cobb,2000;
Gravemeijer, 1994; Gravemeijer et al., 2014). In addition, when conceptual analysis
is employed by a teacher who is skilled at it, we obtain important examples of how
mathematically substantive, conceptually-grounded conversations can be held with
students (Thompson, 2011, p. 44–45)
We include this entire quote because Thompson’s description of conceptual anal-
ysis and its uses captures its relationship with decentering. Conceptual analysis is
the mechanism by which teachers/researchers engage in aspects of decentering
(even when no students are present) and involves the same mental activity involved
in decentering in the moment of interaction with students/subjects.
A primary goal of mathematics education researchers employing a radical con-
structivist’s7 theoretical perspective is to model students’ thinking, usually about
particular mathematical ideas. To do so, these researchers often go through a multi-
step process. The early phases involve synthesizing related literature on previous
studies, reflecting on past work with students, and performing aspects of conceptual
analysis about the idea (focusing on uses 2, 3, and 4 from Thompson’s (2011) list).
What emerge from this work are generalized models of students’ thinking called
epistemic students (Thompson, 2002).8 An epistemic student is a description of a
7
Mathematics educators leveraging radical constructivism as a background theory take seriously
that (1) students’ mathematical reality is distinct from the researchers’ and (2) researchers have no
direct access to students’ mathematical reality.
8
Thompson (2002, 2008) discussed the notion of an epistemic person (student) as a way of think-
ing about a particular idea. It is built from Piaget’s (1981,1987) notion of an epistemic subject.
way of thinking about a particular idea that we could hypothetically attribute to a

person in the way we might say, “Imagine that a student thinks about angle measure
as reporting the area of a region bounded by an angle and a circle centered at the
angle’s vertex.” The researcher constructs images of multiple epistemic students,
considers the implications of each way of thinking about the idea and what observ-
able behaviors might indicate that someone possesses a similar way of thinking
about the idea, and, depending on the research study, may go on to consider how to
scaffold learning experiences that support students in having the opportunity to con-
struct a particular way of thinking about the idea the researcher believes to be pow-
erful. Engaging in this activity involves a type of decentering activity. The epistemic
students generated depend to a large degree on the researcher’s own first-order
model of the mathematics at hand and represent the researcher’s initial attempts to
generate models of others’ thinking related to the idea.
Researchers then typically engage in work with actual students, such as conduct-
ing clinical interviews (Clement, 2000) or teaching experiments (Steffe &
Thompson, 2000). While interacting with subjects in these settings, the researcher’s
goal is to generate conjectures about subjects’ observable actions (utterances, draw-
ings, gestures, etc.) and the nature of the subjects’ meanings that might explain
these actions. These second-order models that emerge are the result of the research-
er’s decentering actions and inform subsequent conjectures of the student’s think-
ing. As researchers repeatedly decenter when interacting with students in the context
of their learning or using an idea (e.g., average rate of change, exponential growth),
the researcher’s models of students’ thinking become more viable, and their images
of students’ ways of thinking (epistemic students) become more refined. It is
through the successive refinements of a theoretical model that it becomes more
stable, although, as von Glasersfeld and Steffe (1991) have noted, we cannot arrive
at absolute certainties, and “The most one can hope for is that the model fits what-
ever observations one has made and, more importantly, that it remains viable in the
face of new observations” (p. 98).
It is worth noting that a teacher’s decentering actions can similarly lead to the
teacher constructing images of epistemic students. In Silverman and Thompson’s
(2008) description of how a teacher’s KDU gains pedagogical power, they are
describing the same mental actions involved in constructing images of epistemic
students. These images provide the foundation for instructional design and guidance
for leading classroom conversations. When engaged in instruction, the teacher’s
epistemic students help guide the teacher’s decision-making (such as what ques-
tions to ask or how to follow up on student errors). However, if teachers are acting
as first- or second-order observers while engaged in teaching, it is likely that they
will routinely recognize novel student reasoning or encounter unexpected difficulty
in supporting students’ learning of some mathematical idea. The second-order
model a teacher constructs when teaching and reflecting on specific classroom
Piaget distinguished three constructs: the individual (an actual person), the psychological subject
(an abstract person described in terms of psychological traits), and the epistemic subject (a coher-
ent collection of ways of knowing).
interactions can lead to the teacher refining and adapting their images of students’
thinking relative to an idea and when interacting with specific tasks related to that
idea. These reflections may lead to the teacher adjusting her teaching practices and
lessons according to these updated images of students’ ways of thinking about an
idea. We further claim that a teacher’s first-order model of the mathematical idea
under consideration/discussion influences the nature of the teacher’s image, the use-
fulness of those images, and their ability to notice and reflect on elements of stu-
dent’s reasoning that can feed into the kinds of refinements in their teaching practice
we have described. We will return to these ideas later in this chapter and propose a
framework that describes the uses of decentering during teaching and its role in a
teacher’s development of mathematical meanings for teaching and mathematical
ways of thinking for teaching specific ideas.
Clinical Interview Methodology
Mathematics education researchers have devised data collection methods for gener-
ating artifacts to inform their construction of a model of a student’s thinking. One
data collection method is clinical interviews, first used by Piaget in his early experi-
ments investigating children’s cognitive development. This method was later
expanded to include open-ended interviews and think-aloud protocols
(Clement, 2000).
The Role of Decentering in Clinical Interview Data Collection
The over-arching goal of a clinical interview is to reveal information about a sub-

ject’s mental processes “at the level of a [subject’s] authentic ideas and meanings”
(Clement, 2000, p. 547). However, as researchers’ knowledge base and inquiries
have evolved, so have the field’s data collection methods. As one example, Clement
(2000) formalized two general categories of clinical interviews, generative and con-
vergent. The goal of a generative clinical interview is to produce theoretical descrip-
tions of mental structures and processes that reveal nuanced and diverse ways of
thinking. Researchers in mathematics education commonly use generative clinical
interviews in the early phases of exploring student thinking relative to a specific
mathematical idea (e.g., rate of change, angle measure, function composition, accu-
mulation). Their analysis of the clinical interview data generates a conjectured
description of a subject’s thinking (a second-order model developed via conceptual
analysis). A researcher’s coding activity involves their use of (or development of)
theoretical constructs to label the subject’s utterances, explanations, and actions as
a basis for their claims about what the student might have been thinking. Convergent
clinical interviews seek to apply the theoretical models developed from generative
clinical interviews to code new or larger data sets. While the goal in convergent
clinical interview studies is to have codes developed that allow high levels of
agreement across independent coders, decentering does come into play when cer-
tain subjects or data sets turn out to be difficult to code reliably. Such instances
indicate a need to revisit or conduct new generative studies to update models of
student thinking (which necessarily involves conceptual analysis and decentering).
Clement’s (2000) discussion of clinical interviews includes detailed descriptions
of the various types of clinical interviews and their purposes. He also notes that the
theoretical descriptions generated by a researcher’s clinical interview data analysis
are influenced by the researcher’s current conceptions and theories. However,
descriptions of methods for collecting clinical interview data rarely discuss how a
researcher determines the interview task(s) or develops the line of questioning used
during data collection and, more generally, how a researcher’s theoretical perspec-
tive on what is entailed in learning an idea impacts the model that the researcher
constructs of the subject’s thinking. This observation led us to ponder the under-
standings and thinking that led to Piaget’s task design and his line of probing that
produced novel insights into children’s thinking. More generally, we invite the
reader to consider the mechanisms for building a model of another’s thinking – what
is the nature of the researcher’s meanings and ways of thinking relative to the idea
being studied that enables a researcher to notice and model important aspects of a
research subject’s thinking? What thinking is involved in putting aside one’s own
way of thinking about an idea and viewing a student’s actions in responding to a
problem from the student’s perspective (decentering)?
Examples of Decentering in Mathematics Education Research
In this section, we provide two examples of how researchers’ decentering actions

and engaging in conceptual analysis influenced their work, including what they
noticed, what they reported, and the connection between their first-order knowledge
and the second-order models they constructed of subjects’ thinking. As we discuss
these examples, we illustrate connections among and uses of the various theoretical
constructs already discussed.
xample 1: How a Researcher’s Meaning for Rate of Change

E
Informed Data Collection and Data Analysis
Thompson (2016) describes his observation of a ninth-grade teacher as she is teach-

ing a lesson on the point-slope and point-point formulas for linear functions.9
Thompson’s characterization of the teacher’s interactions reveals the role of his
9
Thompson conveyed that the teacher had attended a professional development workshop he led,
and the teacher had subsequently decided to use a task similar to one that he had used during the
meaning for constant rate of change and his use of decentering in describing the
teacher’s actions.
Thompson (2013) reports that the teacher first plotted the two points (3, 1) and
(7, 4) on a coordinate axis. This action was followed by the teacher sketching line
segments to illustrate her claim that “the function goes over 4 and up 3.” Thompson
further conveyed that the teacher attempted to locate the y-intercept of the line by
moving to the left four units from the point (3, 1). He then noted that while the
teacher was constructing the line (see Fig. 9.1), she stopped mid-sentence after
uttering the phrase, “If we go 4 to the left….” After a long pause, the teacher stopped
her lesson and proceeded to discuss the homework assignment for the next day.
Thompson (2013) provided his fine-grained description of the teacher’s actions
(including utterances and drawings) used to characterize her thinking. According to
Thompson:
She saw the change in x as a chunk. This was unproblematic in the case of one point.
However, her chunk in this problem did not place her at x = 0 as she wished. Second, her
meaning for slope was “rise over run”, where rise and run were both chunks. Third, her
computation of slope, not evident in this excerpt but made clear later, was of a procedure
that produced a number that is an index of a line’s “slantiness”. Division did not produce a
quotient that has the meaning that the dividend is so many times as large as the divisor—3/4
as a slope was not a number that gave a rate of change. It gave a “slantiness”. Fourth, her
meaning for rate of change entailed neither smooth variation nor proportionality. It was
more akin to her meaning for slope—two things changing in chunks. These meanings not
Fig. 9.1 An example provided by Thompson (2016, p. 84)
workshop (P. Thompson, personal communication). She decided to use the task the day he was
visiting their school so he could observe her lesson. Note that Thompson (2016) explains that his
intention with this task in the workshop was for teachers to determine the function’s constant rate
of change and use this to coordinate proportional changes in x and y to determine the function’s
vertical intercept. Such a solution method requires multiplicative reasoning, which contrasts with
the additive “rise over run” reasoning many teachers were employing.
only failed to provide Sandra a connection between her current setting (two points) and
prior method, they led her down the dead-end path she followed. (Thompson, 2013,
pp. 81–82)
We claim that Thompson’s (2013) ability to characterize the teacher’s thinking

(construct the second-order model he described) was based on both the actions
and utterances that Thompson observed and his interpretation of these actions,
and both of these were influenced by his first-order model of what is entailed in
understanding and learning the idea of constant rate of change (including ideas
about variation, proportionality, slope and rate of change, the meaning of division,
and so on).10 Furthermore, it was similar to second-order models of teachers’
thinking in the professional development workshop that prompted him to create
the task in the first place (P. Thompson, personal communication). Thompson
described that many of the teachers’ images of slope entailed an amount of rise
and an amount of run, where rise and run represented two things changing in
chunks, and they were therefore relying on additive reasoning to think about
slope. Thus, Thompson deliberately created a scenario such that teachers attempt-
ing to leverage this image would encounter difficulties completing the task. This
suggests that Thompson was alert to the limitations of this way of thinking and
was intentional in designing the task to reveal and support students and teachers
in confronting this conception of slope.
Thompson (2013) went on to describe a way of thinking about constant rate of
change that would have been more useful for Sandra: “Had Sandra reasoned pro-
portionally and with smooth continuous variation, she might have said ‘… over
3/4 of 4 and up 3/4 of 3’. That would have given her the graph’s y-intercept”
(p. 81). Note that Thompson’s analysis involves multiple aspects of conceptual
analysis (building models of a subject’s thinking and describing ways of knowing
a mathematical idea that might be either powerful or problematic) and indicates
that he positioned himself as a second-order observer while constructing his sec-
ond-order model of the teacher’s thinking (focused not just on how her thinking
was different from his but also the implications of that thinking). His suggestion
of a more powerful way of thinking about constant rate of change was based on
the thinking he considered at the time to be productive, including an image of the
two quantities varying together smoothly and continuously and the recognition of
a proportional relationship between the changes in the two quantities and all that
proportionality entails. Finally, we point out that Thompson’s first-order model of
10
We don’t have the space to fully unpack all that Thompson and others have written about stu-
dents’ and teachers’ meaning for these ideas (e.g., Byerley et al., 2012; Castillo-Garsow, 2010;
Johnson, 2015, 2018; Thompson, 1994b, 1994, 2000, 2011; Thompson & Carlson, 2014;
Thompson & Thompson, 2007, 1994). However, Thompson’s interpretations of the teacher’s
actions revealed aspects of his image (at the time of reporting) of the ways of thinking that he
deemed to be most productive (and less productive). For example, his conceptualization of the dif-
ference between thinking about variations happening in chunks compared to smooth transitions
where the varying quantities takes on all possible values within an interval oriented his attention to
how the teacher was conceptualizing variation and the implications of her “chunky thinking.”
constant rate of change influenced the data he presented to make arguments in the
paper about the importance of MKT for creating mathematical experiences where
students have a chance to learn mathematical ideas that are powerful and will
serve them well in the future.We invite the reader to consider questions such as the
following. What led to Thompson’s (2013) construction of the task and character-
ization of the teacher’s thinking? What meaning of constant rate of change did
Thompson deem to be productive? In particular, what was Thompson’s first-order
knowledge of constant rate of change that resulted in him being alert to the
“chunky thinking” that he observed? How did Thompson’s first-order knowledge
influence the second-order model he built of the teacher’s thinking? We further
invite the reader to reflect on what guides your own task design and data collec-
tion. How does your first-order model for understanding and learning an idea
influence what questions you ask, what you notice, and how you interpret and
report your data?
xample 2: Researcher Decentering in a Teaching Experiment

E
on Logarithms
In this section, we provide an example of researchers’ iterative efforts to understand

and characterize what is entailed in understanding the idea of logarithm, and we
highlight how the products of the researchers’ decentering actions advanced their
first-order models of what is entailed in understanding and learning the idea of loga-
rithm. We also illustrate how the researchers’ first-order models for teaching the
idea of logarithm evolved during, and as a result of, their decentering actions. Their
conceptual analysis included fine-grained descriptions of students’ thinking relative
to understanding the idea of logarithm and then shaped the researchers’ hypotheti-
cal learning trajectory and subsequent teaching experiment design and data analy-
sis. Their first-order models (including images of students’ thinking) also influenced
their subsequent decentering actions, including the questions they posed and the
models they built. We begin by discussing the purpose and design of teaching exper-
iments in cognitively focused mathematics education research.
I terative Models of a Student’s Thinking Informs Teaching

Experiment Design
A teaching experiment involves a series of teaching episodes that comprise a teach-

ing agent, one or more students, a witness, and a recording method. Teaching exper-
iments differ from clinical interviews in that they involve experimentation with the
ways and mechanisms of influencing students’ mathematical understandings (Steffe
& Thompson, 2000). Teaching experiments allow a teacher-researcher to generate
situations of learning systematically and to test conjectures and local hypotheses
about students’ mathematical meaning for a specific idea and how the student’s
meaning is constructed (Steffe & Thompson, 2000). The design of a teaching exper-
iment begins with the researcher describing the mental actions and ways of thinking
she conjectures to be essential for understanding the idea that is the focus of the
study (i.e., performing a conceptual analysis; Thompson, 2011). A researcher might
generate her initial conceptual analysis by reviewing the research literature related
to understanding and learning the idea while leveraging her prior decentering
actions and images of students’ thinking when learning that idea. In subsequent
studies focused on this idea, the researcher generates new images of students’ think-
ing that lead to modifications of her conceptual analysis. If little is known about
what is entailed in understanding and learning the idea that is the focus of the inves-
tigation, it is common for researchers to conduct a sequence of small-scale investi-
gations (e.g., exploratory teaching interviews11) with several students prior to
conjecturing a trajectory and activities to support students in learning the idea in
ways the researcher hypothesizes will be useful for them.
xploratory Teaching Interviews Lead to Advancements in a Researcher’s

E
First-Order Model
Weber (2002) demonstrated that stating and attempting to explain the common con-
version equation that links a logarithmic statement and its equivalent exponential
statement, logb(x) = y ↔ by = x, were not effective in supporting students in under-
standing of what the statement, y = logb(x), conveys—in particular, that y represents
the number of factors of b in the product that equals the value of x. Kuper (2020)
leveraged these findings in their design of tasks for a series of initial exploratory
teaching interviews aimed at producing models of students’ meanings for exponen-
tial relationships and logarithms that might inform her first-order model for stu-
dents’ learning of the idea of logarithmic function and understanding logarithmic
properties conceptually. Kuper began by probing students’ thinking when respond-
ing to the bacteria growth task, which states: “The population of bacteria in a culture
t days since the bacteria started growing can be determined by the formula
P(t) = 7(2)t. How many days have passed when the bacteria population is 224?”.
Results from performing a retroactive conceptual analysis on the data led Kuper
and Carlson (2020) to identify common ways of thinking used by multiple students.
Some students viewed P(t) as the name of the function instead of a representation of
the bacteria’s population and the expression 7(2)t as instructions for calculating a
value. These conceptualizations did not support the students in viewing the expres-
sion 7(2)t as defining the co-varying relationship between the population’s bacteria
P(t) and the corresponding values of t.
11
See Steffe and Thompson (2000) and Castillo-Garsow (2010) for a description of exploratory
teaching interviews.
Kuper and Carlson (2020) further report that it was common for students to
exhibit a calculational conception of solving the eq. 224 = 7(2)t. That is, they
described their solution process as following a “solving for” algorithm of isolating
a variable. Consequently, the students did not attend to the quantitative meaning of
the calculations and expressions produced when constructing their solution. For
example, their first step, dividing 224 by 7 to produce a result of 32, was viewed as
a calculation disconnected from the context (i.e., students could not explain what 32
represented in the context). We can contrast this with a quantitatively focused solu-
tion approach where the steps and expressions emerge through a process of deter-
mining the values of important quantities in the situation.12 This way of thinking
would develop the ratio 224/7 as a relative size comparison of the final number of
bacteria and the initial number of bacteria (with the result, 32, being the measure of
that relative size). From this conceptualization, writing the eq. 32 = 2t could be
explained as representing the fact that the number of bacteria must grow to become
32 times as large as the original number in two ways: (i) as the result of making a
relative size measurement and (ii) as the result of doubling the initial number of
bacteria some number of times (in this case five times). (It is worth noting that
Kuper (2020) gained clarity in the utility of this way of thinking about the meaning
of the output of a logarithmic function, which impacted her first-order model of
exponential functions and logarithms, during analysis of her interview data.) Since
most students did not exhibit this quantitative approach when constructing their
solution, their justifications for the solution included statements like, “t = 5 is the
exponent on 2 that gives an answer of 32” and the answer 5 is the result of “taking
the log of both sides.” They also observed students conceptualizing the solution
(t = 5) as being the unknown value that satisfies the original statement 224 = 7(2)t
because it produces the correct numerical value. We contrast this rather superficial
meaning for the equation’s solution with other possible meanings, such as seeing t
as representing a varying number of days since the first bacteria population mea-
surement was taken and imagining the value of t varying with the bacteria popula-
tion; then viewing the solution to 224 = 7(2)t as the value of t that corresponds to the
number of bacteria reaching 224.
While Kuper (2018) had initial hypotheses about meanings for logarithms that
would be productive for students understanding the idea of logarithm, her decenter-
ing actions led to her constructing new images of students’ thinking (epistemic stu-
dents). In particular, she gained clarity about students’ conceptions of function and
function notation and their conception of what the output value of a logarithmic
function represents. It is also noteworthy that the researcher’s decentering actions,
including her reflecting on a student’s thinking and efforts to create tasks to advance
12
See O’Bryan (2018, 2020) and Carlson et al. (2003) for more information about emergent symbol
meaning, a way of thinking about calculations, expressions, and formulas where the order of oper-
ations and structure emerges from the way the individual has conceptualized relationships between
quantities in the situation as opposed to viewing them merely as steps or parts of a calculation
algorithm.
a student’s thinking, led to adaptations to her first-order model and subsequent

refinements of her conceptual analysis and instructional materials.
After their analysis of this initial interview data and after reflecting on ways of
thinking they conjectured would advance students’ thinking, Kuper and Carlson
(2020) conducted additional exploratory teaching interviews to test their hypothe-
ses. They devised tasks to support students in understanding a ratio as a relative size,
a quotient as a measure of the relative size of a ratio’s numerator and the denomina-
tor, and the usefulness of these meanings on a student’s conception of the expres-
sion log2(32) as expressing the number of times some value needed to be doubled to
increase that value by a factor of 32. The researchers’ analysis of the exploratory
teaching interviews with the bacteria task revealed that, as students shifted to inter-
pret t = 5 as representing the number of doublings required for the initial number of
bacteria to grow by a factor of 32, students began to coordinate the number of dou-
bling periods (5) and the relative size of the final number of bacteria and the initial
number of bacteria. Put another way, if some value is doubled every day for 5 days,
then the period over which the doubling occurred is 1 day, and the result of doubling
a value each day for 5 days would produce a new value that is 25 (or 32) times as
large as the original value.
To illustrate another foundational understanding, after exploring students’ con-
ceptions of repeatedly doubling and repeatedly tripling some value, including the
ability to explain what the results of the repeated doubling or repeated tripling rep-
resented, Kuper and Carlson (2020) observed that the fluency students acquired did
not extend to determining and representing repeatedly increasing some value by
other factors (e.g., 1.5 or 17). Kuper and Carlson’s analysis of these interactions led
to their hypothesizing that introducing the term tupling would lead to students’ gen-
eralizing their fluency in reasoning about doubling and tripling to non-integer num-
bers and result in their conceptualizing and speaking about tupling a number 1.25
times.13 This observation was not only transformational relative to their hypotheses
about productive meanings for students. It also became a new way of thinking about
exponentiation for the researchers because it provided clarity in their own reasoning
about exponentiation and productive ways of thinking about the connections
between exponential expressions and logarithmic expressions (i.e., it profoundly
changed their first-order models of exponential and logarithmic functions).14
We have provided these descriptions of two exploratory teaching interviews to
highlight the subtle and novel insights the researchers gained about students’
13
In a later section, we described a hypothetical trajectory for introducing the idea of tupling.
14
See Kuper and Carlson (2020) for a detailed analysis of the development of the tupling idea and
term and a conceptual analysis of its relationship to how the researchers thought about related
mathematical ideas. For purposes of this chapter, it is enough to note the following descriptions.
“If the value of a quantity becomes m times as large, we say the quantity’s value m-tuples…An
m-tupling is an event in which the value of a quantity becomes m times as large…An m-tupling
period is the amount of change in the independent quantity of an exponential function needed for
the dependent quantity of the exponential function to become m times as large…[An] Exponent
(on a value, b)…[is] [t]he number of elapsed b-tupling periods” (Conceptual Analysis section,
Table 1).
thinking and learning that were only possible by the researchers taking seriously
their commitment to model students’ thinking (i.e., generating second-order models
of students’ thinking and reflecting on these models to inform modifications to a
researcher’s first-order model).
irst-Order Models, Decentering, Second-Order Models, and Conceptual

F
Analysis Inform Task Design
A researcher’s models of students’ thinking of an idea guide the researcher’s task

development and conceptual scaffolding. In the prior section, we described the
researcher’s conception of a student’s impoverished conception of ratio and quo-
tient (the researcher’s second-order model) and noted that subsequent exploratory
teaching interviews revealed students not initially conceptualizing a quotient as a
measure of the relative size of a ratio’s numerator and denominator (Kuper &
Carlson, 2020). The researchers’ awareness of students’ calculational orientation
and its role in obstructing advances in students’ understanding of what expressions
like log2(32) represent (e.g., the number of doublings that results in an initial value
growing by a factor of 32) informed the design of new tasks aimed at fostering stu-
dent reasoning about the relative size of two quantities’ values and for representing
the repeated tupling of some initial value. As the researchers reflected on the think-
ing they had observed, in relation to the thinking they deemed to be productive, we
claim they were engaged in decentering – they were bringing to mind students’
thinking and deciding on specific actions they hypothesized (e.g., showing an ani-
mation and deciding on specific questions to pose) for supporting advances in stu-
dents’ reasoning.
As an example of the impact of this reflection on decentering (meta-reflection),
Kuper and Carlson (2020) proposed to initially ask students to construct a cactus.
This was followed by a request to construct a second cactus to represent a doubling
of the height of the first cactus (Table 9.1). The researchers also elected to introduce
the term two-tuple to refer to a double in a cactus height. The rationale for this
choice was to support students in developing a meaning for the idea of tupling in a
familiar context as a foundation for future engagement in activities to generalize
this meaning to conceptualize an 8-tupling, 4.5-tupling, and eventually an x-tupling.
We claim that the decentering actions of the researchers contributed to their con-
structing ways of thinking about teaching the idea of logarithm. These included the
researchers conceptualizing specific approaches (see Task Column in Table 9.1) for
engaging students in conceptualizing the relative size of two quantities’ values as a
foundational way of thinking for conceptualizing tupling and considering how
many two-tuplings of some value would result in a final value of some quantity
becoming 32 times as large as its initial value. We further claim that the products of
the researchers’ decentering actions led to advances in their second-order models of
specific students’ thinking and their resulting epistemic students (generalized
images of students’ ways of thinking) related to learning the idea of logarithm. The
researchers’ updated hypothetical learning trajectory was similarly refined to
Table 9.1 Sequence of tasks for learning the tupling language

Rationale for task design and
Stage Task Possible responses sequencing
1 Draw a cactus. Suppose The student might measure This task introduces the new
this cactus doubles, the height of the cactus the tupling language while also
which we refer to as student initially drew and using the colloquial term
two-tuples, in height. use that height to draw a “doubles.” Students are
Draw the resulting cactus new cactus that is as tall as encouraged to participate in the
after this growth. The two copies of the initial activity of drawing the initial
resulting cactus is how cactus and resulting cactus
many times as tall as the The resulting cactus is two The effect of the activity that
starting cactus? times as large as the initial students may notice is that the
Also consider: three- cactus resulting cactus is two times as
tuples or triples tall as the initial cactus
2 Suppose the cactus you The student might measure This task uses the new tupling
drew is eight-tuples the height of the initial language while also providing
(octuples) in height. If cactus drawn and draw a the colloquial term “octuples.”
needed, draw the new cactus that is as tall as The student may choose to
resulting cactus after this
eight copies of the initial participate in the activity of
growth. The resulting cactus. Or the student might drawing the initial and resulting
cactus is how many times reflect on his thinking in the cactus if he experiences
as large as the starting previous task and answer difficulties
cactus? the question without A student that reflects on the
Also consider: 4.5-tuples needing to draw the activity-effect relationship
resulting cactus might imagine drawing the new
The resulting cactus is eight cactus and conclude that the
times as large as the initial new cactus will be eight times
cactus as tall as the initial cactus
3 Suppose a cactus is The student might describe This is the first task in the
57-tuples in height. If a process of measuring 57 sequence where participating in
you were to draw the copies of the original cactus the activity of drawing the
resulting cactus, what to arrive at the resulting resulting cactus would be too
would we observe? (Hint: cactus’ height tedious. Instead, we designed
The resulting cactus is The resulting cactus will be this task to perturb students
how many times as large 57 times as large as the who relied on the activity
as the starting cactus?) initial cactus aspect of the task to reflect on
Also consider: the activity-effect relationships
1000-tuples of the previous tasks and
imagine the result of the
activity as if the actions were
performed
4 Suppose the cactus The resulting cactus will be This task does not provide a
provided x-tuples in x times as large as the initial numerical value for the student
height. If you were to cactus to work with. Rather, the
draw the resulting cactus, student is expected to make a
what would we observe? generalization using the tupling
(Hint: The resulting language to describe an
cactus is how many times arbitrary number of tuples
as large as the starting Students who struggle to make
cactus?) this generalization may need to
attempt more tasks like those in
Stages 2 and 3 to engage in
reflection of the activity-effect
relationship
include anticipated ways of thinking observed in prior exploratory teaching inter-

views. These updates resulted from advances in the researchers’ first-order model
for understanding and learning the idea of logarithm. These advances in the research-
ers’ first-order model also led to more effective decentering (see Table 9.1, column
2) and more targeted probing (see Table 9.1, column 1) during subsequent explor-
atory teaching interviews.
Advances in the researchers’ image of student thinking led to advances in the
researchers’ trajectory for reasoning about the relative size of two quantities’ values
(see Table 9.1, column 3)—in particular, the researchers were more anticipatory of
student thinking and thus better prepared to respond productively to thinking that
students presented. To view the tasks and ordering for other ideas in the hypothetical
learning trajectory for supporting students’ understanding of the idea of tupling, and
gaining fluency in using the idea to describe what the dependent quantity of a loga-
rithmic function represents, see Kuper and Carlson (2020).
The summary of the researchers’ hypothetical learning trajectory in Table 9.1
reflects how the researchers came to think about exponential growth as a tupling
process, a logarithmic expression as representing a number of tupling periods for a
quantity to become some number of times as large as its original value, and the
meaning of the relationship between exponential and logarithmic functions that fol-
lows from these understandings. The researchers then considered their images of
epistemic students, grounded in second-order models of students they had worked
with, and imagined the kinds of tasks, questions, and interventions necessary for
perturbing and advancing the thinking of students who present these hypothetical
ways of thinking. For example, the learning trajectory in Table 9.1 shows the
researchers’ expectation that drawing diagrams representing the relative sizes of the
cactus before and after growing, connected to common language such as “doubling”
and “tripling,” would provide objects of reflection for making sense of the tupling
process and developing a meaning for m-tupling as a general idea. Their images of
epistemic students also guided them to imagine some students remaining reliant on
figurative activity, such as drawing diagrams and not transitioning to coordinating
the multiplicative growth of the cactus height and elapsed time more generally. This
led the researchers to design tasks that would become increasingly difficult without
abstracting the process of tupling to construct a mental image of the general rela-
tionship between elapsed time and the cactus’s height. The learning trajectory in
Table 9.1 became the foundation for a teaching experiment with one student.
Modeling Student Thinking in the Context of a Teaching Experiment
Kuper and Carlson (2020) implemented their learning trajectory in a subsequent

teaching experiment conducted with one student. In the initial phase of their teach-
ing experiment, Kuper and Carlson aimed to support the student in conceiving of
logarithms as functions that determine the number of b-tupling periods needed for
some value to grow by a factor of m (m-tuple) (e.g., log2(8) represents the number
of doubling periods needed for a value to eight-tuple or become eight times as
large). To facilitate this, Kuper and Carlson engaged the student in a series of tasks
that prompted her to examine the growth of a mystical cactus, “Sparky the Saguaro.”
During each teaching episode, the researchers continually updated their model of
the student’s thinking as they interacted with the student. They included the follow-
ing vignette from an interaction with the subject, Lexi.:
In an attempt to determine the 3-week growth factor, Lexi began by noting
Sparky’s initial height of one foot at week zero and then claimed, “three time(s)–
no, every week it’s doubling, or times two for the height. So to get to week three,
you’d say it’s like, you wouldn’t say 6 times as large – that wouldn’t make sense. I
feel like you would say 3 times as large – that doesn’t make sense either.” This
response suggests that Lexi first considered multiplying the 1-week growth factor
(2) by the number of elapsed weeks (3) to calculate the 3-week growth factor.
However, she quickly ruled out that option and looked to other values appearing in
the situation. Lexi then appeared to observe the height of the cactus 3 weeks after its
purchase and eventually concluded that at the end of week 3, Sparky would be eight
times as tall as the initial Sparky. However, there was no evidence to suggest that
Lexi had contemplated the relationship between the 1-week growth factor (2) and
the number of weeks elapsed (3) as an approach for conceptualizing the 3-week
growth factor (8). In particular, although Lexi noted that Sparky was doubling in
height every week, her responses and attention to the heights of the cacti suggest
that she had not yet conceptualized that doubling Sparky’s height 3 weeks in a row
is the same as growing by a factor of 23 or 8 (Kuper & Carlson, 2020, "Teaching
Experiment #1" section, para. 27).
Through a combination of decentering in the moment and performing a concep-
tual analysis of the interview data after the fact, Kuper and Carlson (2020) con-
structed an initial model of Lexi’s thinking and the implications of her thinking
through the lens of the constructs characterized in their conceptual analysis. This
led to follow-up work with the student to establish a more fine-grained description
of her meanings that would inform modifications to the researchers’ first-order
models, their images of epistemic students, and their hypothetical learning trajec-
tory. In the next teaching episode, Lexi was able to note that the doubling period for
the cactus’s height was 1 week, the four-tupling period was 2 weeks, and the eight-
tupling period was 3 weeks, but when asked to determine the three-tupling period,
Lexi predicted it to be 1.5 weeks (halfway between 1 and 2 weeks since three is
halfway between 2 and 4). To determine how Lexi thought about her answer, they
asked her to determine the nine-tupling period (which, if her answer was correct and
she understood that her answer provided the growth factor for a change in time
elapse of 1.5 weeks, would mean that the nine-tupling period was 3 weeks). Rather
than build off of her previous work, Lexi predicted that the nine-tupling period
would just be a bit longer than the eight-tupling period, first guessing 3.5 weeks and
then revising to 3.25 weeks. Kuper and Carlson write that “During the retrospective
analysis of the third teaching episode, we hypothesized that Lexi’s difficulties with
the aforementioned ideas were due to her not understanding that an A-tupling fol-
lowed by a B-tupling had the same overall effect as an AB-tupling” ("Teaching
Experiment #1" section, para. 29). This was an important additional insight that,
once again, contributed to modifications to the researchers’ first-order models of the

ideas under study, a refinement of their images of epistemic students, and contrib-
uted to revisions in their hypothetical learning trajectory. As a final comment, Lexi
finally developed a meaning for the idea that a quantity that m-tuples and then
n-tuples grows by a factor of mn and this “was essential for Lexi as she attempted
questions requiring the use of the first logarithmic property
[logb(m) + logb(n) = logb(mn)]” ("Teaching Experiment #1" section, para. 31).
Comments on Examples 1 and 2
Examples 1 and 2 provide illustrations of several key ideas related to decentering.

In Thompson’s (2016) report, the task he designed in the professional development
workshop, which the teacher opted to use with her students, was in response to
decentering in the moment of working with the teachers. The second-order models
he constructed of the teachers’ meanings aligned with images of epistemic students
he possessed from prior research, reading, and conceptual analysis. He designed the
task such that it would be particularly challenging for teachers with the ways of
thinking about constant rate of change he hypothesized the teachers possessed, thus
introducing a perturbation that would necessitate advances in the teachers’ mean-
ings. As he observed the teacher with her students, his first-order model influenced
Fig. 9.2 The relationship between first-order models, decentering, conceptual analysis, second-
order models, and epistemic students. (Note: *Recall that conceptual analysis can be used in mul-
tiple ways, each contributing differently to this process)
both his decentering in the moment (what he noticed and how he responded to what
he noticed) and during his retrospective conceptual analysis of the teaching episode
and discussing the implications of the teacher’s meanings.
Kuper’s (2020) and Kuper and Carlson’s (2020) work illustrates how products of
researchers’ decentering actions (second-order models) advanced their first-order
model of what is entailed in understanding and learning the idea of logarithm.
Kuper’s decentering actions when conducting a sequence of exploratory teaching
interviews with different students were instrumental in the construction of epis-
temic students (various ways of thinking about logarithms). They further illustrated
how the insights gained from these investigations led to updates in the researchers’
first-order model to include (i) ways of thinking about learning the idea of logarithm
that were novel to them, (ii) new images of students’ thinking when interacting with
tasks designed to perturb their thinking, and (iii) new ways of thinking about teach-
ing the idea of logarithm meaningfully to students. They claim that these insights
led to refinements of the first author’s subsequent decentering actions, including the
questions she posed and the nature of the models she built.
Figure 9.2 represents our image of the relationship between many of the con-
structs and processes we have discussed and exemplified thus far in the chapter.
Due to the cyclical and interdependent nature of many of the constructs and
processes, there is no single “entry-point” to thinking about the relationship
between first-order models, decentering, conceptual analysis, second-order
models, and epistemic students. However, we will start our discussion with
first-order models (an individual’s meanings and ways of thinking about a
mathematical idea). As we have discussed and Thompson (2016) demonstrated,
a first-order model impacts what someone is primed to notice and thus the
nature of their decentering actions. Someone’s first-order model also supports
the multiple facets of conceptual analysis (modeling a student’s/subject’s
thinking and considering the implications of particular meanings), and concep-
tual analysis can produce insights that feed back into and influence the first-
order model. Reflection on the results of decentering actions and conceptual
analysis can both support the construction of second-order models of students’/
subjects’ thinking, which, as demonstrated by Kuper (2020) and Kuper and
Carlson (2020), can also influence a researcher’s first-order models. Engaging
in decentering, conceptual analysis, and model-building all feeds into the
development and refinement of images of epistemic students (the researcher’s
image of generalized ways that students think about an idea). Images of epis-
temic students become incorporated in the teacher’s/researcher’s first-order
model and are helpful for performing conceptual analysis and constructing
second-order models because they are familiar ways of thinking a teacher/
researcher might recognize and be prepared to leverage during subsequent
exploratory teaching interviews.
Uses of Decentering When Studying Teachers and Teaching
In the prior section, we illustrated how a researcher’s first-order model of an idea

can influence a researcher’s decentering actions (to create second-order models
of students’ thinking, perform conceptual analysis, and construct images of epis-
temic students) and how a researcher’s decentering actions in the context of
interacting with a subject about an idea can influence the researcher’s first-order
model of learning that idea. In this section, we make a case that advancing an
instructor’s teaching practice to be more meaningful, coherent, and responsive to
students’ thinking necessarily requires a similar symbiotic interaction between
an instructor’s mathematical meanings and decentering actions (including decen-
tering in the moment or in retrospect to generate second-order models of stu-
dents’ thinking, when performing conceptual analysis and constructing images
of epistemic students). We then illustrate uses of decentering for studying
teacher-student interactions and conclude by providing an overview of a frame-
work that emerged from studying teachers in the context of their interacting with
students in a classroom setting.
laborating Our Meaning for MMT and Its Symbiotic

E
Relationship with Decentering
In our descriptions of a teacher’s mathematical knowledge for teaching (MKT), we

adopt Thompson’s (2016) use of the term mathematical meaning when referencing
the organization of an individual’s actions, images, and other meanings associated
with an idea, also referred to as a scheme.15 It is through repeated reasoning, reflec-
tion, and reconstruction that an individual constructs schemes to organize experi-
ences in an internally consistent way (Piaget & Garcia, 1987, 1991; Thompson,
2016; Thompson et al., 2019). Thompson extended the construct of mathematical
meaning when defining mathematical meanings for teaching (MMT) by including
characterizations of a teacher’s actions that are related to teaching. Our character-
ization of a teacher’s MMT for an idea enabled us to make inferences about how a
teacher organizes her experiences with an idea that might reveal the teacher’s image
of what is entailed in understanding, learning, and teaching an idea.
Silverman and Thompson (2008), as discussed earlier in this chapter, claim that:
A teacher has developed knowledge that supports conceptual teaching of a particular math-
ematical topic when he or she (1) has developed a KDU within which that topic exists [e.g.,
proportionality as a way of thinking for understanding constant rate of change; covaria-
tional reasoning as a way of thinking for understanding growth patterns in quantitative
relationships], (2) has constructed models of the variety of ways students may understand
the content (decentering) [through decentering and performing a conceptual analysis]; (3)
15
Tillema and Gatza (Chap. 3) provide a detailed discussion of scheme.
has an image of how someone else might come to think of the mathematical idea in a simi-
lar way; (4) has an image of the kinds of activities and conversations about those activities
that might support another person’s development of a similar understanding of the mathe-
matical idea; (5) has an image of how students who have come to think about the mathemat-
ical idea in the specified way are empowered to learn other, related mathematical ideas.
(Silverman & Thompson, 2008, p. 508)
We view these phases as a general framing of the processes and products of devel-
oping MMT as a specific idea.
What is relevant for our work then is to describe the mechanisms for advancing
an instructor’s MMT in ways that might lead to the instructor constructing an image
of students’ thinking and spontaneously leveraging that image when interacting
with students, planning a lesson, or reflecting on the impact of their instructional
actions on students’ thinking. We classify images of teaching that include ways of
thinking about students’ ways of learning an idea as a way of thinking about teach-
ing that idea. A way of thinking about teaching an idea then necessarily includes (i)
how the teacher reasons about and understands the idea (the teacher’s first-order
model) and if they have engaged in decentering relative to students’ thinking about
that idea and (ii) their second-order models of students’ thinking about and learning
that idea. As a teacher’s image of students’ mathematics (the teacher’s second-order
models) becomes more stable through reflecting on acts of decentering, we consider
Fig. 9.3 A framework for the development of ways of thinking about teaching an idea. (Note:
*The act of teaching provides feedback on ways of thinking about teaching an idea and new oppor-
tunities to engage in decentering, conceptual analysis, refine first-order models, and refine images
of epistemic students)
these teacher images of students’ generalized ways of thinking (epistemic students)

as informing (and being assimilated into) a teacher’s first-order model or mathemat-
ical meaning for teaching that idea (Figs. 9.2 and 9.3) such that the first-order model
then contains multiple ways of thinking about the idea.16 To further clarify, we claim
that a teacher’s MMT for an idea includes (i) habitual actions of making sense of a
student’s thinking when learning an idea, (ii) images of students’ generalized ways
of thinking about that idea (epistemic students), and (iii) images of the effect of prior
instructional moves on students’ thinking. We further claim that a teacher’s scheme
for teaching that idea includes both ways of thinking about learning that idea and
ways of thinking about teaching that idea, in addition to the teacher’s stable way of
understanding the idea. Through perturbations and accommodations to one’s
scheme for knowing, learning, and teaching that idea, the teacher constructs more
connected way of thinking about teaching that idea.
Tallman and Frank (2016) describe how a teacher’s attention to constructing
models of students’ thinking was constrained when the teacher’s MKT did not
include a more generalized way of thinking about what is entailed in measuring an
angle (that any process for measuring an angle involves quantifying the length of
the arc subtended by a circle centered at the angle’s vertex in any unit proportional
to the circle’s circumference). They further claim that “the absence of such an inten-
tion makes a teacher ill-equipped to recognize the possibility that he or she is pro-
moting students’ construction of uncoordinated angle measure schemes” (Tallman
& Frank, 2016, p. 92). They illuminate this claim in the data they present to illus-
trate that a teacher’s way of thinking about the idea of angle measure was tied to
features of a specific context and was absent of a more general conceptualization of
an angle measure as the result of a quantitative comparison. This tendency to employ
specific ways of thinking about angle measure based on the context resulted in the
teacher not recognizing reasoning used in one context as also being applicable in
other contexts.17 In the moment of teaching, the teacher was focused on determining
if students thought about the solution method in the same way as the teacher rather
than considering the students’ contributions and recognizing the validity of other
quantitative comparisons that could produce the correct answer.
Tallman (2018) demonstrated that a teacher’s interpretation of and responses to
a student’s expressed thinking are strongly related to both the teacher’s personal
mathematical meanings and the learning goals a teacher articulates prior to instruc-
tion. This finding further corroborates claims that teachers’ mathematical meanings
for teaching are a primary source of their instructional decisions and actions
(Thompson, 2017). The evidence produced by Tallman is conveyed in his detailed
16
These ways of thinking about the idea may represent both valid and invalid reasoning and/or may
be more or less productive based on the details of a situation.
17
For example, thinking about angle measure as communicating a fraction of the circumference of
all circles centered at the angle’s vertex subtended by the angle compared to thinking about angle
measure as communicating the relative size of the subtended arc lengths of all circles centered at
the angle’s vertex and units of measure proportional to the circles’ circumferences (such as radius
lengths or 1/360th circumference length).
analysis of a teacher’s pre-teaching conceptions, classroom instruction, and stated

learning goals. As an added methodological consideration, Tallman argues for the
value of identifying and characterizing consistencies and discrepancies between
teachers’ personal and enacted mathematical knowledge. As was demonstrated by
Tallman, this approach has the potential to provide insights into “the characteristics
of teachers’ subject matter knowledge that enable its effective transformation in the
context of practice” ("Research on the Enactment" section, para. 8).
he Symbiotic Relationship Between a Teacher’s Meaning

T
for an Idea and Her Decentering Actions
Research findings have demonstrated links between a teacher’s mathematical mean-

ings and her teaching practices, including her goals for student learning, and how a
teacher interprets and responds to a student’s questions (Baş-Ader & Carlson, 2022;
Heid et al., 2012; Marfai, 1992; Musgrave & Carlson, 2017; O’Bryan & Carlson,
2016; Rocha, 2021; 2022; Rocha & Carlson, 2020; Tallman, 2021, 2018; Uscanga
et al., 1995; Zbiek et al., 2014). According to Thompson et al. (1996), and further
substantiated with empirical data by Tallman (2018) and Rocha (2022), “If a teach-
er’s conceptual structures comprise disconnected facts and procedures, their instruc-
tion is likely to focus on disconnected facts and procedures” (p. 416). In contrast, if
a teacher has a coherent set of mathematical meanings, then it is at least possible,
but not sufficient to guarantee, that this teacher will attempt to have her students
construct this same set of coherent mathematical meanings (Thompson et al., 1996).
This claim leads to the following questions. What processes and support tools are
effective for supporting teachers’ development of coherent mathematical meanings?
How can teachers be supported in noticing and utilizing students’ mathematical
meanings within lessons designed to facilitate construction of those meanings?
Based on our prior discussions of decentering and claim of the symbiotic relation-
ship between a teacher’s first-order model and decentering actions, we suggested
further exploration of the role of teacher engagement in decentering actions as an
approach for fostering ongoing advances in a teacher’s first-order model of know-
ing, learning, and teaching an idea.
In our research, we have found that supporting teachers’ development of more
coherent meanings for the ideas they are teaching is a critical first step toward teach-
er’s valuing conceptual learning in their students; however, without the support of
conceptually scaffolded curricular materials and aligned professional development,
the teacher’s ways of thinking about teaching ideas and their actual teaching prac-
tices, including their inquiry into students’ thinking, might not advance (Carlson
et al., 2010a, b). One benefit of such curriculum materials is how their design sup-
ports and encourages students to put their thinking on display, making it more likely
that the teacher will begin to notice discrepancies between how students are think-
ing and how the teacher is thinking. As a teacher begins to puzzle about and make
sense of students’ thinking, they might also begin to notice commonalities across
multiple students’ thinking and weaknesses in their own thinking. Both are impor-
tant in supporting teachers’ decentering activity. We provide an example of this in
the next section.
dvances in a Teacher’s Ways of Thinking About

A
Teaching an Idea
In this section we illustrate how a teacher’s interaction with a student in the context
of discussing a conceptually focused task in the Pathways materials (Carlson et al.,
2021) created a perturbation for the teacher. We further discuss how the teacher’s
focus on expressing the quantitative meaning for the algebraic representation of the
AROC led to advances in the teacher’s personal meaning for the idea of average rate
of change (AROC), mathematical meaning for teaching the idea of AROC, and the
teacher’s ways of thinking about teaching AROC.18
hifts in a Teacher’s Meaning for the Idea of Average Rate of Change

S
(AROC) and Her Ways of Thinking About Teaching the Idea of AROC
Musgrave and Carlson (2017) have reported wide variations in precalculus instruc-
tors’ conceptions of average rate of change. In their study of 19 mathematics gradu-
ate students’ mathematical meanings for average rate of change, most expressed a
f b f a
procedural approach that involved their computing and explaining the
ba
result as representing “rise over run” or the slope of a line between two points. After
a professional development intervention, the researchers detected shifts in the grad-
uate students’ expressed meanings with most associating the idea of average rate of
change with a constant rate of change. However, only eight (of the 19) included a
description of what the constant rate of change represented, e.g., the constant rate of
change over a specified interval of a function’s domain that results in the same
change in the dependent quantity as the non-linear relationship defined by the func-
tion. These findings that even graduate students in mathematics have weak mean-
ings for this foundational idea bring to light the need for professional development
for advancing a teacher’s meanings for ideas that are the focus of the teacher’s
instruction (Musgrave & Carlson, 2017).
As a follow-up to analyzing and reporting data on the instructors’ conceptions of
the idea of AROC, the researchers video-recorded two of the instructors who had
18
The teacher was using conceptually scaffolded instructional materials that are based in research
on learning the course’s ideas. She had attended a 3-day workshop prior to teaching with the mate-
rials and weekly content-focused professional development for a full year while using the materials.
expressed weak meanings for AROC when teaching this idea in their Pathways
Precalculus class. Their analyses revealed one instructor conveying a similar proce-
dural explanation to her students by demonstrating the steps for calculating an aver-
age rate of change while describing the result of the calculation as a “constant rate
of change from 2 to 7.” The results from analyzing a second instructor’s interaction
with a student revealed how an instructor’s decentering actions led to advances in
both the instructor’s mathematical meaning of the idea of AROC and her ways of
thinking about teaching the idea of AROC. In what follows, we describe what
unfolded during the class interaction between this instructor and a student.
Decentering Actions Lead to Advances in a Teacher’s MMT

for Teaching AROC
In this example, the instructor began her lesson by showing her students how to
f 7 f 2
determine the value of for a quadratic function f. After completing the
72
f 7 f 2
calculation, a student asked the instructor what the answer meant.
72
(The instructor later indicated that at that moment she was unsure how to answer the
student’s question. She further conveyed that she had difficulty saying what an
AROC represented in prior semesters, even though the curriculum included a con-
ceptual definition and explanation.) Since the curriculum had a consistent focus on
representing the meaning of symbols,19 the instructor decided to follow up by
exploring how the student was visualizing the symbols in the AROC definition. She
asked the student to draw the graph of f on a whiteboard at the front of the room and
then asked the student to locate the points (2, f(2)) and (7, f(7)) on f’s graph and
illustrate “where you see f(7) – f(2) and 7 – 2 on the graph.” The student constructed
vertical and horizontal lines to illustrate these changes on the graph. With some
prompting the student was able to describe the horizontal line’s length as represent-
ing a change in x from 2 to 7 and f(7) – f(2) as “how much f changed when x
increased from 2 to 7.” The instructor then asked the student to explain how she saw
f 7 f 2
the expression on the graph. The student responded by asking, “Isn’t
72
this just the change in y over the change in x, which is the slope of a straight line?”
The instructor responded by asking the student, “What straight line?” After the stu-
dent constructed a line segment connecting the points (2, f(2)) and (7, f(7)), the
instructor prompted the student to explain what this slope of the line segment repre-
sented. After a long pause, the instructor asked the more specific question, “What
constant rate of change does the slope of this line represent?”
19
For more detail on this focus, see O’Bryan (2018), O’Bryan and Carlson (2016), and Carlson
et al. (2003).
In the instructor’s description during a follow-up interview, it was during this

long pause that she first conceptualized the calculation for the AROC as represent-
ing a constant rate of change over a specified interval of a function’s domain (e.g.,
from 2 to 7) that results in the same gain in the function’s dependent quantity com-
pared to what was gained by the function on that interval. She stated that prior to
this moment she was only trying to recite the textbook description of what an AROC
represents and was struggling to repeat the wording. This comment suggests that
she might also have been perturbed by her prior inability to provide an adequate
explanation and thus had brought the idea to mind at times other than when she was
attending professional development or when she was teaching.
After the long pause, the student responded by describing the result of the calcu-
lation as the constant rate of change from 2 to 7. The instructor asked the student for
clarification by demonstrating on the graph the change in the dependent quantity on
that interval of the domain. We conjecture that the shift in the instructor’s under-
standing during the prior interaction is what motivated her to ask this question. The
student responded by pointing to the vertical segment connecting (7, f(2)) and (7,
f(7)) to represent the value of f(7) – f(2) and answered by saying the slope of the line
segment connecting the two points is the constant rate of change on the interval
from 2 to 7 that gives the correct change in the function value on that interval.
We claim that this exchange is an example of a teacher advancing both her MMT
for AROC and her way of thinking about teaching the idea of average rate of change
in the moment of teaching. We claim that this shift occurred when she was attempt-
ing to understand her student’s thinking. Because of the perturbation, the instructor
experienced when she was unable to provide a response to the student’s initial ques-
tion and her subsequent actions to probe the student’s meaning for the symbols in
the AROC expression, the interaction appeared to advance the instructor’s scheme
for AROC. In the moment of interacting with the student, the instructor conceptual-
ized how the AROC calculation determined the constant rate of change that achieved
the same net change as the function value on the interval of the domain over which
she was determining the function’s AROC.
The data suggest that the instructor’s acts to decenter led to her reorganizing her
scheme for AROC. We conjecture that the perturbations she experienced when
attempting to explain the idea in prior semesters (as noted above) contributed to her
experiencing an accommodation to her scheme that resulted in her advancing her
meaning for AROC. In the follow-up interview with this instructor, she revealed that
she planned to use a graph and the same line of questioning about what the symbols
in the AROC definition represent when teaching the idea of AROC in the future. This
comment suggests that she also experienced a shift in her ways of thinking about
teaching AROC. If the instructor engages in decentering behaviors in future interac-
tions with students when teaching the ideas of AROC, it is likely that both her image
of students’ ways of learning the idea of AROC and her ways of thinking about teach-
ing the idea of AROC will continue to advance. In contrast, if the instructor’s actions
are exclusively focused on conveying her more conceptually oriented way of think-
ing, with no consideration given to her lesson scaffolding, her students’ thinking, or
how she is being interpreted when teaching, her image of students’ thinking and her
image of teaching the idea of AROC will not be likely to advance further.
Characterizing Teacher Decentering
Our claim of the symbiotic relationship between a teacher’s MMT and her decenter-
ing actions extends to our research activity in that we have found that investigations
focused on understanding the nature and development of a teacher’s MMT have led
to advances in our understanding of mechanisms of decentering and vice versa. Thus,
our research has sometimes focused on understanding and characterizing decenter-
ing while not focusing on a teacher’s MMT or how it develops. In this section, we
provide an overview of our investigations that studied teacher decentering and share
our most recent framework for characterizing teacher decentering behaviors that we
observed in teachers during these studies (Baş-Ader & Carlson, 2022; Carlson &
Baş-Ader, 2019; Carlson et al., 2015; Rocha & Carlson, 2020; Teuscher et al., 1982).
Decentering as a Construct for Studying Teachers
We first used decentering (Carlson et al., 2007) to study teachers and teaching in the
context of a research and development project aimed at supporting secondary
precalculus-level teachers in shifting their instruction to be more conceptually ori-
ented (as defined by Thompson et al., 1994), coherent, and supportive in advancing
students’ mathematical thinking of key ideas of precalculus that were documented
to be foundational for learning calculus (e.g., Carlson, 1998; Carlson et al., 2007,
2002, 2001). Our observations of select teachers’ classrooms prior to beginning the
intervention revealed widespread practices of procedurally oriented classes with
procedural explanations, similar to what had been reported in the Trends in
International Mathematics and Science Study (TIMSS) (Stigler & Hiebert, 2008).
The initial intervention included weekly 3-hour professional development meet-
ings20 engaging precalculus teachers in tasks to support their constructing coherent
meanings of the key ideas of precalculus (Carlson et al., 2007).21 Teachers also met
twice per week in professional learning communities (PLCs) of four to seven teach-
ers from the same school. Each PLC had a designated facilitator (one of the
20
The teachers received 3 hours of mathematics education graduate credit and a stipend for
attending.
21
The targeted meanings were based on our current conception of the body of research on knowing
and learning key ideas of precalculus (Breidenbach et al., 1992; Carlson, 1998; Carlson et al.,
2001, 2002, 2015; Carlson, Oehrtman & Engelke 2010; Carlson, Oehrtmen & Teuscher, 2010;
Engelke et al., 2005; Frank, 2017; Kuper, 2020; Kuper & Carlson, 2020; Carlson et al., 2021;
Monk, 2010; Moore, 2012; Moore & Carlson, 2012; O’Bryan, 2018; Strom, 2015; Thompson,
1994a, b, 2000</CitationRef>).
teachers) who led the community in discussing the idea (e.g., rate of change, expo-
nential growth) that was the focus of the prior week’s professional development
meeting. The facilitators received weekly coaching between meetings from a proj-
ect leader who had reviewed a video recording of their PLC meeting from the prior
week to offer suggestions for improving their facilitation abilities, including the
advancement of their colleagues’ thinking and meanings, with the coach highlight-
ing interactions that probed another’s thinking. For example, they posed questions
to promote reflection on unproductive PLC interactions (e.g., “Sharon, is there a
reason you don’t allow Mary to answer questions?”; “Juan, what do you think David
meant when he said…?”; “Dan, are you aware of who is doing most of the
talking?”),22 shared video of interactions in which the facilitator posed questions
aimed at revealing how another PLC member was thinking, and prompted the com-
munity of PLC facilitators to discuss aspects of the interaction.
The results from coding the videos of the professional development and PLC
meetings that were collected over that full semester revealed that the teachers gener-
ally did not attempt to communicate their thinking when speaking and did not make
efforts to understand their colleagues’ thinking (Carlson et al., 2007). In particular,
the teachers’ explanations made vague references to quantities (e.g., “it goes up”),
provided variable definitions that lacked precision (e.g., “d = distance”), and
explained graphs as static shapes. Their explanations included little or no explana-
tion of the thinking they used to approach a problem or construct their solutions.
These findings of the widespread procedurally oriented patterns of interaction
led to our adaption of the norms for the bi-weekly PLC meetings to include a con-
vention that all PLC members attempt to reference quantities when speaking and
explain the thinking that led to their constructions and calculations. To raise teach-
ers’ consciousness about their patterns of speaking, thus making their speaking an
object of reflection, we termed this expectation for their speaking and explaining as
speaking with meaning (Clark et al., 2008). We further negotiated more general
Rules of Engagement23 that included a goal for PLC members to listen to the utter-
ances of their colleagues while attempting to make sense of the meanings they were
conveying, and in instances when they were unclear about what was being expressed,
we conveyed an expectation that they pose a question that might clarify what they
thought was unclear.
The PLC facilitators were charged with taking the lead role in modeling and sup-
porting their colleagues in speaking with meaning and attempting to understand the
meanings being expressed by members of their PLC. The study revealed widespread
22
Over a full year, each of four PLC twice-weekly meetings was video-recorded, with weekly
recording coded by two researchers and then shared weekly with research team (including the
facilitator coach).
23
There were four rules of engagement that emerged as explicit conventions for promoting the
sociomathematical norms of “speaking with meaning” and attempting to make sense of another’s
thinking that were adopted by the teachers as explicit conventions for communication within their
classrooms. We created posters that listed these conventions and the teachers elected to post them
in their classrooms.
advances in the PLC members speaking with meaning and substantial advances in
three of the PLC facilitators’ interest and efforts to make sense of other PLC mem-
bers’ thinking. The various manifestations of decentering we observed are articu-
lated in five decentering moves we identified when analyzing their conversations
(Carlson et al., 2007). Their decentering actions ranged from the facilitator showing
no interest in understanding the thinking or perspective of another PLC member to
the facilitator building a model of another’s thinking and through interaction, build-
ing a model of how she (the facilitator) might be interpreted, and then using both
models to determine her next action.
The analysis of this data further revealed that a facilitator’s ability to pose ques-
tions that resulted in advances in another’s thinking was typically linked to the facil-
itator’s understanding of the idea under discussion. This awareness provided our
first insight into the strong link between a teacher’s capacity to respond productively
to another’s thinking and the teacher’s personal meaning for the idea under discus-
sion. When PLC facilitators had a weak personal meaning for an idea under discus-
sion, they were not equipped to assess the efficacy of another’s thinking or intervene
productively to advance another’s thinking; nor were they able to provide conceptu-
ally oriented explanations. In the authors’ reporting (Carlson et al., 2015), they fur-
ther noted that general inquiries made by a facilitator such as a request for a PLC
member to revoice or extend another PLC member’s explanations sometimes led to
two PLC participants (both who had relatively strong meanings for the idea being
discussed) constructing models of each other’s thinking. In contrast, the facilitator
who showed no interest in understanding her PLC members’ thinking “exhibited
extreme discomfort with the content” (Carlson et al., 2015, p. 847) and, when
probed by the PLC facilitator coach to explain her thinking, consistently responded
by explaining how she obtained her answer. On an occasion when the coach encour-
aged her to ask her colleagues to explain their thinking, she responded by question-
ing why one would be interested in understanding the thinking of another. She
clearly had not adopted the goals of the project. The project leaders later learned
that this PLC facilitator was assigned to be the facilitator because she was the most
senior member of the precalculus instructional team at that school.24
After a year of participating in the intervention, some teachers expressed that
their participation in the professional development and PLC meetings had impacted
the quality of their interactions with students, in particular their attempts to make
sense of their students’ explanations and their ability to pose follow-up questions to
reveal students’ thinking. During the subsequent academic year, the researchers
video-recorded and studied the classroom instruction of four PLC members who
had demonstrated efforts to make sense of their colleagues’ thinking during the
PLCs. The researchers’ analysis of this data (Teuscher et al., 1982) illustrated how
the teacher’s decentering actions (or lack of these actions) assisted (or constrained)
24
We report this result to highlight the importance of initial interest and seeing value in understand-
ing another’s thinking as a desirable attribute of a facilitator of other teachers’ decentering actions.
Going forward in our project, we requested school administrators to identify facilitators who val-
ued the goals of the project.
the teacher in making productive in-the-moment instructional decisions. These

studies established that advances in a teacher’s ability to build a model of a student’s
thinking corresponded with advances in a teacher’s meanings and that advances
toward more conceptually oriented explanations could be fostered by teachers mak-
ing their speaking and listening an explicit focus of personal and collective reflec-
tion. In what follows we describe the products of a third study (Baş-Ader & Carlson,
2022) that built on these findings and produced a more comprehensive Teacher
Decentering Framework, including specific teacher actions that were associated
with five different levels of decentering actions that we observed.
A Decentering Framework for Studying Teaching
Our subsequent professional development work in the Pathways Project continued

our focus on promoting reflective discourse (e.g., speaking with meaning and mak-
ing sense of another’s thinking) as a mechanism for advancing a teacher’s personal
mathematical meanings and teaching practices. In this section, we provide an over-
view of a framework (Baş-Ader & Carlson, 2022) that resulted from analyzing
video data from three 40-student precalculus classes taught by mathematics gradu-
ate students using an early version of the Pathways Precalculus research-based con-
ceptually scaffolded curriculum (Carlson et al., 2021) in a university setting. The
three subjects had taught with these materials for at least 1 year while attending
weekly 90-minute Pathways conceptually focused professional development meet-
ings. In these meetings, the professional development leader held the instructors
accountable for attempting to speak with meaning (Clark et al., 2008) and attempt-
ing to understand their colleagues’ explanations.25 The choice to collect the data in
a context of instructors attending content-focused professional development and
using a conceptually focused curriculum provided the possibility for the teacher-
student interactions to be coherent and meaningful.
In presenting the data, Baş-Ader and Carlson (2022) described the learning
goals for the idea central to the interaction prior to presenting their analysis of the
data; doing so enabled the reader to assess the decentering moves they identified
relative to their effectiveness in advancing the conversation toward a more pro-
ductive mathematical way of thinking. Level 0 of the framework (Table 9.2) is
characterized as showing no interest in the student’s thinking, although the
instructor does prompt the student to share her answer. The instructor might show
concern for the student getting the correct answer but shows no interest in the
thinking the student engaged in to determine her answer. In light of our review of
Piaget’s construct of intellectual egocentrism (the child unconsciously remains
centered on his own actions and own point of view), we extend this construct to
25
In particular, the professional development leader posed questions to the speaker to call attention
to vague descriptions (e.g., “What distance?” or “What does it mean for it to be decreasing?”) and
regularly asked participants to describe how they had interpreted what a colleague had just
described.
Table 9.2 A framework that characterizes a continuum of an instructor’s decentering actions

(Baş-Ader & Carlson, 2022, pp. 106–107)
Mental actions Decentering levels Description of the behaviors
Does not reflect on Level 0: no interest Asks question to elicit the student’s
aspects of the interaction Shows no interest in the answer
that are contributed by student’s thinking, but shows Listens to the student’s answer
the student interest in the student’s answer. Does not pose question aimed at
Assumes that the Makes no attempt to make revealing the student’s thinking:
student’s thinking is sense of the student’s thinking, Poses question that focuses on a
identical to her own but takes actions to get the procedure [e.g., What do you do
Uses first-order model student to say the correct next?]
during interaction to answer Poses question to have the
organize and guide her student echo key phrase or
actions complete steps to get the correct
answer
Does not reflect on Level 1: Interest Poses question to reveal the

aspects of the interaction Shows interest in the student’s student’s thinking [e.g., Why? What
that are contributed by thinking, but makes no attempt does that mean? What does that
the student to make sense of the student’s term represent?], but does not
Recognizes that the thinking. Attempts to move the attempt to understand the student’s
student’s thinking is student to her way of thinking thinking
different than her own without trying to understand or Guides the student toward her way
but does not intentionally build on the expressed thinking of thinking:
analyze the student’s and/or perspective of the Poses question for the purpose
mental actions student of getting the student to adopt
Uses first-order model her way of thinking or has the
during interaction to student echo key phrase
organize and guide her Gives explanation aimed at
actions getting the student to adopt her
way of thinking
Evaluates how the student’s
response compares to her way of
thinking or approach
Reflects on aspects of the Level 2: make sense Poses question to reveal the
interaction that are Makes an effort to make sense student’s thinking and attempts to
contributed by the of the student’s thinking and adopt the student’s perspective
student perspective but does not use Does not use the student’s thinking
Recognizes that the this knowledge in for the purpose of advancing the
student’s thinking is communication student’s current way of thinking:
different than her own Does not pose question aimed at
and attempts to analyze prompting the student to reflect
the student’s mental on or advance her current way of
actions thinking
Builds the second-order Redirects the conversation
model of the student’s toward her way of thinking or
thinking approach
Uses first-order model
during interaction to
organize and guide her
actions
(continued)

Mental actions Decentering levels Description of the behaviors
Reflects on aspects of the Level 3: use Prompts the student to explain the
interaction that are Makes sense of the student’s student’s way of thinking:
contributed by the thinking and perspective and Poses question to gain insights
student makes general moves to use the into the student’s way of
Assumes that the student student’s thinking when thinking
has unique ways of interacting with the student Leverages the student’s expressed
thinking thinking to advance the student’s
Builds and continually understanding of key ideas in the
adjusts the second-order lesson:
model of the student’s Poses question or gives
thinking explanation that is attentive to
Operates from the the student’s thinking and aimed
second-order model of at advancing the student’s
the student’s thinking to understanding of the key idea
make decisions on how Poses question or gives
to act explanation to support the
student in making connections
between different viable ways of
thinking of the key idea
Reflects on aspects of the Level 4: use and adjust Poses question to reveal and
interaction that are Constructs an image of the understand the student’s thinking
contributed by the student’s thinking and Follows up on the student’s
student perspective and then adjusts her response in order to perturb the
Assumes that the student action to take into account both student in a way that extends the
has unique ways of the student’s thinking and how student’s current way of thinking:
thinking the student might be Poses question informed by the
Builds and continually interpreting her utterance student’s current way of thinking
adjusts the second-order and meaning of the key idea
model of both the Gives explanation informed by
student’s thinking and the student’s current way of
how the student might be thinking and meaning of the key
interpreting her utterance idea
Operates from the
second-order model of
both the student’s
thinking and how the
student might be
interpreting her utterance
classify an instructor’s actions as exhibiting intellectually egocentric behavior

when they only exhibit Level 0 decentering actions.
To illustrate this behavior and how we extracted information from our analy-
sis, we present one of the excerpts from Baş-Ader and Carlson (2022) detailing
an interaction between an instructor and student in the context of talking about a
task that prompted students to describe the relative size of two quantities. As
background, the authors conveyed that the students were given two lines of dif-
ferent lengths (60 and 25 inches), and students were expected to perform a mul-
tiplicative comparison of the segments’ lengths. Baş-Ader and Carlson explain
that the curriculum prompted students to determine the length of one segment
using the second segment’s length as the unit of measurement. The reader is also
alerted to the fact that both segments are given in a common unit, so it is possible
for the student to determine the relative size of quantity A in units of quantity B
by performing the calculation (measure of quantity A in the common unit)/(mea-
sure of quantity B in the common unit). Baş-Ader and Carlson explain that the
quotient, k = (measure of quantity A in the common unit)/(measure of quantity B
in the common unit), conveys that quantity A is k times as large as quantity
B. They further note that this pre-algebra-level task is included in Pathways
Precalculus curriculum because the idea of relative size is foundational for
understanding and representing exponential growth and other ideas that appear in
a precalculus course (e.g., rational functions, radian measure).
Excerpt 1, taken from Baş-Ader and Carlson (2022, p. 108), highlights the main
point that the instructor’s (Karen’s) focus was on getting the students to say the cor-
rect answer. She told them what to calculate (“quantity A over quantity B”) and then
prompted them to restate the value of the quotient using the numbers they were
given for the two lengths.
Excerpt 1
1 Karen: If we want to see how big A is in terms of B…Remember, so quantity A
over quantity B, what is that?
2 Student 1: Twenty-five over sixty?
3 Karen: Twenty-five over sixty, good. So, if we write twenty-five over sixty, what
does that represent, again?
4 Student 2: A over B.
5 Karen: A over B, what does that mean like in English sentence?
6 Student 2: How big A is compared to B.
7 Karen: Exactly, how big A is compared to B. So, twenty-five over sixty, we would
say A is equal to twenty-five over sixty whatever B’s length is, okay? If we write
out twenty-five over sixty in a calculator we will get .416 about. So we could say
A is .416 times B.
The instructor listened to students’ answers, but at no point during the exchange
did she ask students to share their thinking. Baş-Ader and Carlson (2022) concluded
by pointing out that the instructor appeared satisfied that a student had finally
repeated the words she had previously expressed. We could classify this instructor’s
interaction as intellectually egocentric behavior since there were no instances of her
exhibiting behaviors to suggest that she was interested in her students’ thinking.
Table 9.2 displays the researchers’ characterization of the decentering levels
(column 2), the associated mental actions (column 1), and the instructor behaviors
identified when coding the videos (column 3). Each row of the table illustrates how
these three categories are linked. As an example, the entries in row 2 characterize
Level 1 decentering as expressing interest in a student’s thinking by asking the stu-
dent to say what a statement, term, etc. represents. The instructor recognizes that the
student’s thinking differs from her own, but she takes no action to understand the
student’s thinking. Instead, the instructor presents her approach and/or thinking and
makes moves to get the student to adopt her way of thinking.
An instructor is classified as exhibiting Level 2 decentering actions if she poses
questions to reveal a student’s thinking and then shows evidence of understanding the
student’s thinking (builds a second-order model) but does not use the student’s think-
ing during interaction. Even though the instructor has constructed an understanding of
the student’s perspective/thinking, her subsequent questions, drawings, and explana-
tions do not build from or leverage the thinking the student displayed. Similar to Level
1, the instructor redirects the conversation to the instructor’s way of thinking. In our
video analyses of teacher-student interactions, our efforts to characterize teacher-stu-
dent interactions necessitated our including constructs to classify mental actions that
would not be considered decentering since the teacher did not attempt to build a model
of a student’s thinking. These framework levels allow for the classification of initial
shifts toward the students’ perspectives, such as showing interest in a student’s
answers (Level 0), showing interest in a student’s thinking, and then taking action to
move the student to the teacher’s way of thinking (Level 1).26
A teacher who is classified as exhibiting Level 2 mental actions is inquiring into
a student’s thinking and engages in mental actions to construct a second-order model
of the student’s thinking. However, she does not use her image of how the student is
thinking to inform future actions. In contrast, both Level 3 and Level 4 classifications
characterize a teacher using her second-order model of a student’s thinking to inform
her instructional actions. In the highest level (Level 4) of the Decentering Framework,
the teacher reflects not only on the student’s thinking but also on her own thinking in
relation to the student’s thinking. This mental action aligns with Liang’s (2003) theo-
retical description of “scheme activity” that she associates with what she calls Type
3 decentering, involving the teacher consciously treating her schemes as objects of
reflection. If a teacher then reflects on her images of students’ ways of thinking about
an idea (epistemic students), we say the teacher is developing ways of thinking about
teaching an idea. As such, we consider Level 4 of our framework to involve not only
teachers’ reflecting (decentering) activity but also their reflections on prior reflecting
(decentering) activity. We also note that a teacher’s actions to reflect on her decenter-
ing actions and their effectiveness in advancing a student’s thinking is constructing
ways of thinking about teaching an idea.
Our experiences of conducting professional development with hundreds of sec-
ondary teachers and college instructors (and a retrospective review of our classroom
video data) support that while most secondary and university teachers claim to be
interested in student thinking, very few in either population initially exhibited even
Level 1 decentering behavior (e.g., Carlson et al., 2015); their focus was primarily
on supporting students in learning methods for obtaining answers. Moreover, during
the first semester of either a secondary teacher or university instructor using the
Pathways cognitively scaffolded instructional materials (Carlson et al., 2021) and
For a detailed description of the mathematical contexts, learning goals, tasks, and researchers’
26
methods for constructing this decentering framework, see Baş-Ader and Carlson (2022).
participating in our weekly professional development, we observed only modest

shifts in the teachers’ interest in their student’s thinking. During the first semester of
their teaching with the Pathways Precalculus materials, the instructors appeared
unable (or unwilling) to consider their students’ thinking while consumed by mak-
ing sense of the course’s ideas.27 After a teacher’s personal meanings for the course’s
ideas became more coherent (Musgrave & Carlson, 2017), professional develop-
ment leaders were more successful in fostering productive discussions about meth-
ods for revealing students’ thinking (Rocha, 2021).
Our approaches for fostering higher-level decentering actions include (i) view-
ing and discussing classroom videos of a teacher’s decentering actions, (ii) asking
the instructors to share instances from their teaching when they made sense of and
leveraged a student’s thinking,28 (iii) designing assessment items that reveal stu-
dents’ thinking, etc. It is not our goal, nor is it realistic, for most teachers to con-
sistently engage in high-level decentering actions when teaching a course with
fixed pacing. However, increasing a teacher’s interest in students’ thinking and
fostering growth in their listening and questioning toward understanding students’
thinking can create the expectation that students’ thinking is relevant. As the
teacher operates with an orientation to understand and affect students’ thinking,
occasional efforts to decenter, particularly in instances when the teacher is not
satisfied with her efforts to affect her students’ learning, can over time lead to a
large cumulative effect on a teacher’s MMT and ways of thinking about teaching
specific ideas.
In one case in which we collected interview and classroom video data from a
few teachers over multiple (as many as three) years, we observed a teacher’s lec-
tures becoming more coherent and anticipatory of students’ thinking (e.g., mak-
ing drawings and providing explanations focused on engendering productive ways
of thinking). We have observed that a teacher can progress to demonstrate higher
levels of decentering actions when supported in doing so—e.g., using cognitively
scaffolded curriculum focused on advancing and revealing students’ thinking,
attending weekly PD aimed at supporting teacher reflection on MMT for ideas,
how they are learned, and how they can be taught. In this sense, we claim that
advancing toward higher-level decentering actions is developmental and emerges
in concert with advances in a teacher’s meanings for the idea under discussion and
opportunities for teachers to reflect on their student’s thinking and the effective-
ness of their instructional choices. As teachers repeatedly engage in decentering
in the context of teaching a specific idea, their ways of thinking about learning
that idea and ways of thinking about teaching that idea can continue to advance.
As we continue to inquire into the symbiotic relationship between a teacher’s
mathematical meanings and decentering actions, we conjecture that studying
27
Recall that our assessment of graduate students’ meaning for precalculus ideas (Musgrave &
Carlson, 2017; Carlson & Baş-Ader, 2019) revealed that even graduate students in mathematics
have weak meanings of precalculus ideas (average rate of change).
28
During one semester the professional development leader began each weekly meetings by asking
each (of 9 = nine) instructors to share instances when they made sense of how a student was think-
ing and then made an instructional move that leveraged the model they had built.
multiple teachers in the context of their teaching a specific idea will expand our
understanding of this relationship while contributing to our understanding of how
ways of thinking about teaching that idea develop.
Concluding Remarks
We have illustrated the important role that decentering plays when modeling a stu-
dent’s thinking during both research and teaching and highlight the symbiotic rela-
tionship that exists between a teacher’s and researcher’s mathematical meanings for
an idea and their decentering actions. In particular, we presented examples of how a
teacher’s/researcher’s first-order model of the mathematical ideas impacts what
they are positioned to notice when decentering, how reflecting on the results of
decentering and performing a conceptual analysis can produce second-order models
of students’ thinking, and how these models might inform the teacher’s/researcher’s
first-order models. We further highlighted how iterating this cycle (Fig. 9.3) can
help teachers/researchers construct stable images of epistemic students. Our exam-
ples illustrate how repeated efforts to decenter when interacting with students (dur-
ing teaching or conducting research studies) help support iterative refinements in an
individual’s meanings for mathematical ideas and images of others’ thinking and
learning an idea. As further illustration of the nature of decentering, we discussed
the Baş-Ader and Carlson's (2022) Teacher Decentering Framework (Table 9.2) that
provides a fine-grained characterization of the mental processes exhibited by teach-
ers as they move from an egocentric perspective to building and leveraging their
second-order model of a student’s thinking during interactions.
It is worth noting that the examples provided in this chapter are from research
studies that focus on quantitative reasoning29 and covariational reasoning30 as foun-
dational ways of thinking that can provide coherence to students’ mathematical
experiences. Students who develop the disposition to reason quantitatively (i.e.,
identify quantities in a situation, and conceptualize how those quantities’ values are
related and vary in tandem) are better positioned to construct algebraic and graphi-
cal representations of mathematical relationships that are personally meaningful to
the one constructing them (see Carlson et al., 2003). O’Bryan and Carlson (2016)
also demonstrated that teachers who value and are committed to supporting their
students in reasoning quantitatively may be more attentive to and interested in the
meanings students express. They may also be more likely to intentionally create
opportunities to reveal how their students are thinking and consider what meanings
29
Quantitative reasoning (Thompson, 1988, 1993, 1994a, 1994b, 2012, 2013) describes a way of
thinking whereby an individual identifies quantities in a situation (measurable attributes of an
object), conceptualizes a measurement scheme for producing values representing the quantity’s
size, and conceptualizes a structure in how various quantities in a situation are related and
interdependent.
30
Covariational reasoning (Carlson et al., 2002; Saldanha & Thompson, 1998; Thompson &
Carlson, 2014) involves the identification of pairs of quantities in a situation that vary and concep-
tualizing the constraints on how those quantities change in tandem.
motivated a student’s construction of specific mathematical representations of quan-

titative relationships.
The examples we presented also highlight the role of cognitive research on task
development (recall the constant rate of change, logarithm, and AROC tasks) and
their role in perturbing and revealing students’ thinking and advancing a teacher’s
mathematical meaning for teaching an idea. Since most teachers do not have the time
or resources to write curriculum or study student learning, researchers must lead the
way in devising and studying processes for supporting shifts in teachers’ mathemati-
cal meanings, ways of thinking about teaching specific ideas, and their habitually
engaging in decentering (when teaching and when planning and reflecting on their
teaching). In our research projects (Musgrave & Carlson, 2017; O’Bryan & Carlson,
2016), we found that researcher-developed professional development and aligned
research-developed curricula can facilitate teacher decentering and advances in a
teacher’s mathematical meanings for teaching ideas. We offer our framework
(Fig. 9.3) for advancing teachers’ ways of thinking about teaching an idea as a gen-
eral theory for others to leverage and refine in the context of devising and studying
mechanisms and tools to support shifts in teachers – shifts that lead to teachers enact-
ing and continually advancing their teaching practices to include a focus on under-
standing and advancing students’ thinking and understanding specific ideas.
As a first line of inquiry for future research, we recommend devising and study-
ing professional development experiences for instructors aimed at supporting
advances in their (i) mathematical meanings for teaching an idea (first-order knowl-
edge), (ii) decentering actions when planning a lesson and implementing the lesson,
and (iii) subsequent meta-reflection on the effectiveness and products of their
decentering actions and instructional choices when teaching the lesson. We hypoth-
esize that professional development with this focus will lead to incremental advances
in a teacher’s instruction to be more supportive of and responsive to students’ think-
ing and learning. We conclude by providing a short list of questions for researchers
to consider:
• As teachers’ (or researchers’) images of epistemic students become more refined,
how does this change what they notice, how they plan lessons, and how they
respond to students during interactions?
• What factors foster advancements in teachers’ interest in and spontaneously
inquiring into a student’s thinking during teaching and after teaching an idea?
What is the interaction between a teacher’s/researchers’ decentering actions and
her mathematical meanings for teaching that idea?
• What factors foster advances in a teacher’s reflection on her prior reflections
(meta-reflection) when planning a lesson and when considering the effectiveness
of a lesson?
• What factors (profession development experiences/tools) support advances in an
instructor’s reflection on her ways of thinking about teaching an idea, including
her pedagogical choices (e.g., if, when, and how to use an applet, choice of tasks
for students to complete, interactions/questions for students, what to say and
write when leading a discussion)?
References
Baş-Ader, S., & Carlson, M. P. (2022). Decentering framework: A characterization of graduate

student instructors’ actions to understand and act on student thinking. Mathematical Thinking
and Learning, 24(2), 99–122. https://doi.org/10.1080/10986065.2020.1844608
Breidenbach, D., Dubinsky, E., Hawks, J., & Nichols, D. (1992). Development of the process
conception of function. Educational Studies in Mathematics, 23(3), 247–285. https://doi.
org/10.1007/BF02309532
Byerley, C., Hatfield, N., & Thompson, P. W. (2012). Calculus student understandings of division
and rate. In S. Brown, S. Larsen, K. Marrongelle, & M. Oehrtman (Eds.), Proceedings of the
15th annual conference on research in undergraduate mathematics education (pp. 358–363).
SIGMAA on RUME. http://sigmaa.maa.org/rume/RUME_XV_Proceedings_Volume_2.pdf
Carlson, M. (1998). A cross-sectional investigation of the development of the function concept. In
E. Dubinsky, A. H. Schoenfeld, & J. Kaput (Eds.), Research in collegiate mathematics educa-
tion. III (Vol. 7, pp. 114–162). American Mathematical Society. https://www.researchgate.net/
publication/35777721_A_cross-sectional_investigation_of_the_development_of_the_func-
tion_concept
Carlson, M. P., & Baş-Ader, S. (2019). The interaction between a teacher’s mathematical con-
ceptions and instructional practices. In A. Weinberg, D. Moore-Russo, H. Soto, & M. Wawro
(Eds.), Proceedings of the 22nd annual conference on research in undergraduate mathemat-
ics education (pp. 102–110). SIGMAA on RUME. http://sigmaa.maa.org/rume/RUME22_
Proceedings.pdf
Carlson, M. P., Bowling, S., Moore, K., & Ortiz, A. (2007). The role of the facilitator in promoting
meaningful discourse among professional learning communities of secondary mathematics and
science teachers. In T. Lamberg & L. R. Wiest (Eds.), Proceedings of the 29th annual meeting
of the North American Chapter of the International Group for the Psychology of Mathematics
Education (pp. 841–848). University of Nevada, Reno. https://www.pmena.org/pmenapro-
ceedings/PMENA%2029%202007%20Proceedings.pdf
Carlson, M., Jacobs, S., Coe, E., Larsen, S., & Hsu, E. (2002). Applying covariational reason-
ing while modeling dynamic events: A framework and a study. Journal for Research in
Carlson, M. P., Larsen, S., & Jacobs, S. (2001). An investigation of covariational reasoning and
its role in learning the concepts of limit and accumulation. In R. Speiser, C. A. Maher, &
C. N. Walter (Eds.), Proceedings of the twenty-third annual meeting of the North American
Chapter of the International Group for the Psychology of Mathematics Education (Vols. 1-2,
pp. 151–159). ERIC Clearinghouse for Science, Mathematics, and Environmental Education.
https://files.eric.ed.gov/fulltext/ED476613.pdf
Carlson, M. P., Madison, B., & West, R. D. (2015). A study of students’ readiness to learn calculus.
International Journal of Research in Undergraduate Mathematics Education, 1(2), 209–233.
https://doi.org/10.1007/s40753-015-0013-y
Carlson, M. P., O’Bryan, A., & Rocha, A. (2023). Instructional Conventions for Conceptualizing,
Graphing and Symbolizing Quantitative Relationships. In G. Karagöz Akar, İ. Ö. Zembat,
S. Arslan, & P. W. Thompson (Eds.), Quantitative reasoning in mathematics and science edu-
cation (pp. 221–259). Springer, Cham. https://doi.org/10.1007/978-3-031-14553-7
Carlson, M., Oehrtman, M., & Engelke, N. (2010). The precalculus concept assessment: A tool for
assessing students’ reasoning abilities and understandings. Cognition and Instruction, 28(2),
113–145. https://doi.org/10.1080/07370001003676587
Carlson, M. P., Oehrtman, M., Moore, K., & O’Bryan, A. (2021). Precalculus: Pathways to calcu-
lus, A problem solving approach (8th ed.). Hayden-McNeil.
Carlson, M., Oehrtman, M., & Teuscher, D. (2010, January 24–26). Transforming the professional
development culture and quality of mathematics and science instruction within a second-
ary school [Paper presentation]. The 2010 Math and Science Partnership Learning Network
Conference, Washington, D.C., United States. https://mspnet-static.s3.amazonaws.com/10_

Carlson.pdf
Carlson, M. P., Smith, N., & Persson, J. (2003, July 13–18). Developing and connecting calcu-
lus students’ notions of rate-of-change and accumulation: The fundamental theorem of cal-
culus [Paper presentation]. The 27th International Group for the Psychology of Mathematics
Education Conference held jointly with the 25th PME-NA Conference, Honolulu, HI, United
States. https://files.eric.ed.gov/fulltext/ED500922.pdf
Castillo-Garsow, Carlos (2010). Teaching the Verhulst model: A teaching experiment in covaria-
tional reasoning and exponential growth [Unpublished doctoral dissertation]. Arizona State
University.
Clark, P. G., Moore, K. C., & Carlson, M. P. (2008). Documenting the emergence of "speak-
ing with meaning" as a sociomathematical norm in professional learning community dis-
course. The Journal of Mathematical Behavior, 27(4), 297–310. https://doi.org/10.1016/j.
jmathb.2009.01.001
Clement, J. (2000) Analysis of clinical interviews: Foundations and model viability. In A. E. Kelly
& R. A. Lesh (Eds.), Handbook of research design in mathematics and science education
Engelke, N., Oehrtman, M., & Carlson, M. (2005). Composition of functions: Precalculus stu-
dents’ understandings. In G. M. Lloyd, M. Wilson, J. L. M. Wilkins, & S. L. Behm (Eds.),
Proceedings of the 27th annual meeting of the North American Chapter of the International
Group for the Psychology of Mathematics Education (pp. 570–577). https://www.pmena.org/
pmenaproceedings/PMENA%2027%202005%20Proceedings.pdf
Frank, K. M. (2017). Examining the development of students’ covariational reasoning in the con-
text of graphing [Unpublished doctoral dissertation]. Arizona State University.
Ginsburg, H. P., & Opper, S. (1988). Piaget’s theory of intellectual development (3rd ed.).
Prentice-Hall.
Heid, M. K., Grady, M., Jairam, A., Lee, Y., Freeburn, B., & Karunakaran, S. (2014). A processes
lens on a beginning teacher’s personal and classroom mathematics. In JJ. Lo, K. R. Leatham,
& L. R. Van Zoest (Eds.), Research trends in mathematics teacher education (pp. 67–82).
Springer, Cham. https://doi.org/10.1007/978-3-319-02562-9_4
Johnson, H. L. (2012). Reasoning about variation in the intensity of change in covarying quantities
involved in rate of change. The Journal of Mathematical Behavior, 31(3), 313–330. https://doi.
org/10.1016/j.jmathb.2012.01.001
Johnson, H. L. (2015). Secondary students’ quantification of ratio and rate: A framework for rea-
soning about change in covarying quantities. Mathematical Thinking and Learning, 17(1),
64–90. https://doi.org/10.1080/10986065.2015.981946
Kuper, E. (2018). Teaching experiments examining students' development of the idea of logarithm
[Unpublished doctoral dissertation]. Arizona State University.
Kuper, E. & Carlson, M. (2020) Foundational ways of thinking for understanding the idea of loga-
rithm. The Journal of Mathematical Behavior, 57, Article 100740. https://doi.org/10.1016/j.
jmathb.2019.100740
Liang, B. (2020). Theorizing teachers’ mathematical learning in the context of student-teacher
interaction: A lens of decentering. In S. S. Karunakaran, Z. Reed, & A. Higgins (Eds.),
Proceedings of the 23rd annual conference on research in undergraduate mathematics educa-
tion (pp. 742–751). SIGMAA on RUME. http://sigmaa.maa.org/rume/RUME23.pdf
Lobato, J. (2003). How design experiments can inform a rethinking of transfer and vice versa.
Educational Researcher, 32(1), 17–20. https://doi.org/10.3102/0013189X032001017
Lobato, J. (2006). Alternative perspectives on the transfer of learning: History, issues, and chal-
lenges for future research. The Journal of the Learning Sciences, 15(4), 431–449. https://doi.
org/10.1207/s15327809jls1504_1
Lobato, J. (2008). When students don’t apply the knowledge you think they have, rethink your
assumptions about transfer. In M. P. Carlson & C. Rasmussen (Eds.), Making the connec-
tion: Research and teaching in undergraduate mathematics (pp. 289–304). Mathematical

Association of America.
research and practice. Educational Psychologist, 47(3), 232–247. https://doi.org/10.108
0/00461520.2012.693353
Lobato, J., Rhodehamel, B., & Hohensee, C. (2012). “Noticing” as an alternative transfer of learn-
ing process. The Journal of the Learning Sciences, 21(3), 433–482. https://doi.org/10.108
0/10508406.2012.682189
Lobato, J., & Siebert, D. (2002). Quantitative reasoning in a reconceived view of transfer. The Journal
of Mathematical Behavior, 21(1), 87–116. https://doi.org/10.1016/S0732-3123(02)00105-0
Marfai, F. S. (2017). Characterizing teacher change through the perturbation of pedagogical goals
[Unpublished doctoral dissertation]. Arizona State University.
Monk, S. (1992). Students’ understanding of a function given by a physical model. In G. Harel &
E. Dubinsky (Eds.), The concept of function: Aspects of epistemology and pedagogy (Vol. 25,
pp. 175–193). Mathematical Association of America.
Moore, K. C. (2010). The role of quantitative reasoning in precalculus students learning cen-
tral concepts of trigonometry (Publication No. 3425753) [Doctoral dissertation, Arizona State
University]. ProQuest Dissertations & Theses Global.
Moore, K. C., & Carlson, M. P. (2012). Students’ images of problem contexts when solving applied
problems. The Journal of Mathematical Behavior, 31(1), 48–59. https://doi.org/10.1016/j.
jmathb.2011.09.001
Musgrave, S., & Carlson, M. P. (2017). Understanding and advancing graduate teaching assistants'
mathematical knowledge for teaching. The Journal of Mathematical Behavior, 45, 137–149.
O'Bryan, A. E. (2018). Exponential growth and online learning environments: Designing for and
studying the development of student meanings in online courses [Unpublished doctoral dis-
sertation]. Arizona State University.
O’Bryan, A. (2020, Spring). You can’t use what you don’t see: Quantitative reasoning in applied
contexts. OnCore: Journal of the Arizona Association of Teachers of Mathematics, 66–73.
O'Bryan, A. E. & Carlson, M. P. (2016) Fostering teacher change through increased noticing:
Creating authentic opportunities for teachers to reflect on student thinking. In T. Fukawa-
Connelly, N. Engelke Infante, M. Wawro, & S. Brown (Eds.), Proceedings of the 19th annual
conference on research in undergraduate mathematics education (pp. 1192–1200). SIGMAA
on RUME. http://sigmaa.maa.org/rume/RUME19v3.pdf
Piaget, J. (1952). Play, dreams and imitation in childhood (C. Gattegno & F. M. Hodgson, Trans.).
W W Norton & Co. (Original work published 1945)
Piaget, J. (1987). Possibility and necessity: Volume 1. The role of possibility in cognitive develop-
ment (H. Feider, Trans.). University of Minnesota Press. (Original work published 1981)
Piaget, J. (1995). Sociological studies (T. Brown, R. Campbell, N. Emler, M. Ferrari, M. Gribetz,
R. Kitchener, W. Mays, A. Notari, C. Sherrard & L. Smith, Trans.). Routledge. (Original work
published 1965)
Piaget. (2001). Studies in reflecting abstraction (R. L. Campbell, Ed. & Trans.). Psychology Press.
(Original work published 1977)
Piaget, J., & Garcia, R. (1991). Toward a logic of meanings (P. M. Davidson & J. Easley, Eds.).
Psychology Press. (Original work published 1987)
Piaget, J., & Inhelder, B. (1973). The psychology of the child (H. Weaver, Trans.). Routledge &
Kegan Paul. (Original work published 1966)
Rocha, A. (2021). The effectiveness of a professional development video-reflection interven-
tion: Mathematical meanings for teaching angle and angle measure. In S. S. Karunakaran
& A. Higgins (Eds.), 2021 Research in undergraduate mathematics education reports
(pp. 264–273). SIGMAA on RUME. http://sigmaa.maa.org/rume/2021_RUME_Reports.pdf
Rocha, A. (2022). The influence of graduate student instructors’ mathematical meanings for teach-
ing sine function on their enacted teaching practices. In S. S. Karunakaran & A. Higgins (Eds.),
Proceedings of the 24th annual conference on research in undergraduate mathematics educa-

tion (pp. 472–480). SIGMAA on RUME. http://sigmaa.maa.org/rume/RUME24.pdf
Rocha, A., & Carlson, M. (2020). The role of mathematical meanings for teaching anddecentering
actions in productive student-teacher interactions. In S. S. Karunakaran, Z. Reed, & A. Higgins
(Eds.), Proceedings of the 23th annual conference on research in undergraduate mathematics
education (pp. 1146–1153). SIGMAA on RUME. http://sigmaa.maa.org/rume/RUME23.pdf
Saldanha L. A., & Thompson, P. W. (1998). Re-thinking covariation from a quantitative per-
spective: Simultaneous continuous variation. In S. Berenson, K. Dawkins, M. Blanton,
W. Coulombe, J. Kolb, K. Norwood, & L. Stiff (Eds.), Proceedings of the twentieth annual
meeting of the North American Chapter of the International Group for the Psychology of
Mathematics Education (Vol. 1, pp. 298–303). https://files.eric.ed.gov/fulltext/ED430775.pdf
Silverman, J., & Thompson, P. W. (2008). Toward a framework for the development of math-
ematical knowledge for teaching. Journal of Mathematics Teacher Education, 11(6), 499–511.
https://doi.org/10.1007/s10857-008-9089-5
Simon, M. A., Tzur, R., Heinz, K., & Kinzel, M. (2004). Explicating a mechanism for concep-
Simon, M. A. (2006). Key developmental understandings in mathematics: A direction for inves-
tigating and establishing learning goals. Mathematical Thinking and learning, 8(4), 359–37.
https://doi.org/10.1207/s15327833mtl0804_1
Steffe, L. P. (1996, December). Radical constructivism: A way of knowing and learning [Review
of the book Radical constructivism: A way of knowing and learning, by E. von Glasersfeld].
Zentralblatt für Didaktik der Mathematik, 28(6), 202–204.
and essential elements. In A. E. Kelly & R. A. Lesh (Eds.), Handbook of research design in
mathematics and science education (pp. 267–306). Lawrence Erlbaum Associates.
Stigler, J. W., & Hiebert, J. (2009). The teaching gap: Best ideas from the world's teachers for
improving education in the classroom. Free Press.
Strom, A. D. (2008). A case study of a secondary mathematics teacher’s understanding of expo-
nential function: An emerging theoretical framework (Publication No. 3304889) [Doctoral dis-
sertation, Arizona State University]. Dissertation Abstracts International.
Tallman, M. A. (2015). An examination of the effect of a secondary teacher's image of instruc-
tional constraints on his enacted subject matter knowledge [Unpublished doctoral disserta-
tion]. Arizona State University. https://doi.org/10.13140/rg.2.1.4871.5363
trigonometric functions. The Journal of Mathematical Behavior, 62, Article 100869. https://
doi.org/10.1016/j.jmathb.2021.100869
teachers’ knowledge base. Journal of Mathematics Teacher Education, 23(1), 69–95. https://
doi.org/10.1007/s10857-018-9409-3
Teuscher, D., Moore, K. C., & Carlson, M. P. (2016). Decentering: A construct to analyze and
explain teacher actions as they relate to student thinking. Journal of Mathematics Teacher
Education, 19(5), 433–456. https://doi.org/10.1007/s10857-015-9304-0
Thompson, P. W. (1982). Were lions to speak, we wouldn't understand. The Journal of Mathematical
Behavior, 3(2), 147–165. http://pat-thompson.net/PDFversions/1982WereLions.pdf
Thompson, P. W. (1988). Quantitative concepts as a foundation for algebraic reason-
ing: Suffi ciency, necessity, and cognitive obstacles. In M. Behr, C. Lacampagne, &
M. Wheeler (Eds.), Proceedings of the annual conference of the International Group
for the Psychology of Mathematics Education (pp. 163–170). http://pat-thompson.net/
PDFversions/1989PMEQuant.pdf
Thompson, P. W. (1990). A theoretical model of quantity-based reasoning in arithme-

tic and algebra. Department of Mathematical Sciences and Center for Research in
Mathematics and Science Education, San Diego State University. h ttp://pat-thompson.net/
PDFversions/1990TheoryQuant.pdf
Studies in Mathematics, 25(3), 165–208. https://doi.org/10.1007/BF01273861
Thompson, P. W. (1994a). The development of the concept of speed and its relationship to con-
cepts of rate. In G. Harel & J. Confrey (Eds.), The development of multiplicative reasoning in
the learning of mathematics (pp. 179–234). SUNY Press. http://pat-thompson.net/PDFversion
s/1994ConceptSpeedRate.pdf
Thompson, P. W. (1994b). Images of rate and operational understanding of the fundamen-
tal theorem of calculus. Educational Studies in Mathematics, 26(2-3), 229–274. https://doi.
org/10.1007/BF01273664
Thompson, P. W. (2000). Radical constructivism: Reflections and directions. In L. P. Steffe &
P. W. Thompson (Eds.), Radical constructivism in action: Building on the pioneering work of
Ernst von Glasersfeld (pp. 412–448). The Falmer Press.
K. Gravemeijer, R. Lehrer, B. van Oers, & L. Verschaffel (Eds.), Symbolizing, modeling and
tool use in mathematics education (Vol. 30, pp. 197–220). Springer. http://pat-thompson.net/
PDFversions/2002DidacticObjs.pdf
foundation of mathematics education. In O. Figueras, J. L. Cortina, S. Alatorre, T. Rojano,
& A. Sepulveda (Eds.), Proceedings of the joint meeting of PME 32 and PME-NA XXX (Vol.
1, pp. 31–49). https://www.pmena.org/pmenaproceedings/PMENA%2030%202008%20
Proceedings%20Vol%201.pdf
Thompson, P. W. (2011). Quantitative reasoning and mathematical modeling. In S. A. Chamberlin,
L. L. Hatfield, & S. Belbase (Eds.), New perspectives and directions for collaborative research
in mathematics education, WISDOMe Monographs (Vol. 1, pp. 33–57). University of
Wyoming. file:///Users/sinembas/Downloads/WISDOMeMonographVol.1.pdf
Thompson, P. W. (2012). Advances in research on quantitative reasoning. In R. Mayes, R. Bonillia,
L. L. Hatfield, & S. Belbase (Eds.), Quantitative reasoning: Current state of understanding,
WISDOMe Monographs (Vol. 2, pp. 143–148). University of Wyoming. https://www.uwyo.
edu/wisdome/_files/documents/thompson.pdf
Thompson, P. W. (2013). In the absence of meaning…. In K. R. Leatham (Ed.), Vital
directions for mathematics education research (pp. 57–93). Springer. https://doi.
org/10.1007/978-1-4614-6977-3_4
Thompson, P. W. (2016). Researching mathematical meanings for teaching. In L. D. English &
D. Kirshner (Eds.), Handbook of international research in mathematics education (3rd ed.,
pp. 435–461). Routledge. http://pat-thompson.net/PDFversions/2015ResMathMeanings.pdf
Thompson, P. W, Carlson, M. P., Byerley, C., & Hatfield, N. (2014). Schemes for thinking with
K. C. Moore, L. P. Steffe & L. L. Hatfield (Eds.), Epistemic algebra students: Emerging mod-
els of students' algebraic knowing, WISDOMe Monographs (Vol. 4, pp. 1–24). University of
Wyoming. file:///Users/sinembas/Downloads/wisdome_4_final.pdf
Thompson, P. W., Carlson, M. P., & Silverman, J. (2007). The design of tasks in support of
teachers’ development of coherent mathematical meanings. Journal of Mathematics Teacher
Education, 10(4-6), 415–432. https://doi.org/10.1007/s10857-007-9054-8
Thompson, A. G., Philipp, R. A., Thompson, P. W., & Boyd, B. A. (1994). Calculational and
conceptual orientations in teaching mathematics. In D. B. Aichele (Ed.), Professional devel-
opment for teachers of mathematics, 1994 yearbook of the NCTM (pp. 79–92). National
Council of Teachers of Mathematics. https://mathed.byu.edu/kleatham/Classes/Fall2010/

MthEd590Library.enlp/MthEd590Library.Data/PDF/Thompson%20(1994)-1051403520/
Thompson%20(1994).pdf
Thompson, P. W., & Thompson, A. G. (1994). Talking about rates conceptually, Part I: A teach-
er's struggle. Journal for Research in Mathematics Education, 25(3), 279–303. https://doi.
org/10.5951/jresematheduc.25.3.0279
Thompson, A. G., & Thompson, P. W. (1996). Talking about rates conceptually, Part II:
Mathematical knowledge for teaching. Journal for Research in Mathematics Education, 27(1),
2–24. https://doi.org/10.5951/jresematheduc.27.1.0002
Uscanga, R., Simmons, C., Tallman, M., & Oehrtman, M. (2019). An exploration of the factors
that influence the enactment of teachers’ knowledge of exponential functions. In A. Weinberg,
D. Moore-Russo, H. Soto, & M. Wawro (Eds.), Proceedings of the 22nd Annual Conference on
Research in Undergraduate Mathematics Education (pp. 621–628). SIGMAA on RUME. http://
sigmaa.maa.org/rume/RUME22_Proceedings.pdf
von Glasersfeld, E., & Steffe, L. P. (1991). Conceptual models in educational research and practice.
The Journal of Educational Thought, 25(2), 91–103. https://www.jstor.org/stable/23767267
Weber, K. (2002). Developing students’ understanding of exponents and logarithms. In
D. S. Mewborn, P. Sztajn, D. Y. White, H. G. Wiegel, R. L. Bryant, & K. Nooney (Eds.),
Proceedings of the twenty-fourth annual meeting of the North American Chapter of the
International Group for the Psychology of Mathematics Education (Vols. 1-4, pp. 1019–1027).
ERIC Clearinghouse for Science, Mathematics, and Environmental Education. https://files.
eric.ed.gov/fulltext/ED471747.pdf
Zbiek, R. M., Peters, S. A., Johnson, K. H., Cannon, T., Boone, T. M., & Foletta, G. M. (2014).
Locally logical mathematics: An emerging teacher honoring both students and mathematics. The
Journal of Mathematical Behavior, 34, 58–75. https://doi.org/10.1016/j.jmathb.2014.01.003
Zuberbühler, K. (2018). Combinatorial capacities in primates. Current Opinion in Behavioral
Sciences, 21, 161–169. https://doi.org/10.1016/j.cobeha.2018.03.015
Chapter 10
Logic in Genetic Epistemology
Introduction and Goals
In this chapter, I shall consider Piaget’s1 use of logic as a topic of investigation in

children’s reasoning and development, as well as his use of formalized logic as a
modeling tool. This focus sets this chapter apart from most others in this section
since it is not directly focused on characterizing a construct from genetic epistemol-
ogy for use in ongoing mathematics education research. A large part of the chapter
shall consist of analyzing Piaget’s use of logic as a modeling tool in key texts, which
should aid the reader in making sense of such texts when reading on their own,
while a helpful contribution (assuming I achieve this goal), facilitating the reading
of Piaget’s texts, is neither the primary goal of the book nor of this chapter. Rather,
my goals for the chapter include the following:
1. I hope to convey to the reader something of the ambition of Piaget’s agenda of
researching logic in reasoning and using logic to model reasoning. This may
stimulate the formulation of ambitious agendas of research on the nature and
development of children’s thinking, specifically for topics as complex and hard
to study as logic.
1
Here, I shall use “Piaget” often to refer to Piaget and his colleagues. This is intended for conve-
nience as I do not mean to downplay his colleagues’ contributions or intellectual merit in the
shared work. In the case of Beth and Piaget (1966), the book was written in two separate sections
by the two authors, though each gave feedback and advice on the refinement of the other section.
Though in quotes from that book I may actually mean Piaget as the primary author, I shall usually
use his name to stand for the research group.
P. C. Dawkins (*)
https://doi.org/10.1007/978-3-031-47386-9_10
340 P. C. Dawkins
2. I intend to use the case study of Piaget’s use of logical formalisms to study chil-
dren’s reasoning to reflect on the use of expert models to study children’s reason-
ing more broadly. I claim this is frequent in mathematics education as our most
readily available tools for interpreting student reasoning are our own mathemati-
cal understandings (whether they be idiosyncratically personal or our under-
standing of institutionalized mathematical meanings that we take to be shared by
experts).
3. I hope to raise a series of questions and observations about Piaget’s use of logic
as a modeling tool that reveals some of what requires explication. While I do not
claim to have the authority to fully resolve questions about his intent (or how it
changed over time), I hope they will provide helpful points of reflection for the
reader regarding how we can use such modeling tools with clarity and self-
awareness. I offer this in pursuit of advancing our science and communication of
that science.
The first question to address is, “Why logic?” There are multiple ways to address
this. Personally, I pursue the relationship of logic and reasoning in Piaget’s work
because it is an area of intense personal interest and ongoing investigation. More
generally though, logic played an interesting and central role in much of Piaget’s
own work. He wrote multiple books explicitly focused on the development of logic
in human reasoning:
• “Logic and psychology” (1953)
• “The Growth of Logical Thinking from Childhood to Adolescence” (1958, with
B. Inhelder)
• “The Early Growth of Logic in the Child” (1964, with B. Inhelder)
• “Mathematical Epistemology and Psychology” (1966, with E.W. Beth)
• “Toward a Logic of Meanings” (1991, with R. Garcia)
Piaget’s interest in logic is somewhat predictable given his interest in mathemati-
cal and scientific forms of knowledge more generally. His focus on operations (see
Moore et al.’s, Chap. 4, this volume) and group structures that arise from revers-
ibility and composability (see Norton’s, Chap. 7, this volume) reflect this mathemat-
ical leaning in his psychology, which gives rise to a natural interest in logic as a
marked feature of mathematical activity. Piaget, like many mathematicians and
logicians, did not understand logic to pertain only to mathematical subject matter.
The scope of logic is a question in need of some clarification, which will naturally
depend upon which tools of logic one considers and how one employs them.
Nevertheless, in Beth and Piaget (1966), Piaget discusses a key observation: Formal
logic is performed by humans (even if mostly adults with specialized expertise),
which puts the question of even the most formalized logic into the realm of psychol-
ogy. However, Piaget’s use of at least the symbols of formal logic to model the
reasoning of much younger students clearly entails a more fundamental set of ques-
tions and hypotheses about the relation between logic and human reasoning. What
precisely were these questions and hypotheses? Further, why does the relationship
between logic and psychology even need such a justification? Is not logic merely a
10 Logic in Genetic Epistemology 341
description of right and rational reasoning, or maybe even a tool for describing (and
critiquing) reasoning that might be considered sub-rational or not aligned to norma-
tive rationality? In what follows, I shall attempt to provide an operational answer to
these questions sufficient to lay a groundwork for the following discussion of
Piaget’s work. Once I have reviewed some concepts from logic to consider their
operation, I shall then turn to consider its relation to psychology, specifically in
Piaget’s research and theory.
he Nature of Logic and Its Possible Relations

T
to Human Reasoning
To clarify what is commonly meant by logic, we shall draw upon the distinction
between syntax and semantics. In discussing language, semantics refers to relation-
ships between the language and the subject of discussion (reference and meaning),
while syntax refers to relationships within the language itself (grammar). The first
feature of logic to note is that it entails comparison or generalization across seman-
tic contexts (various topics that statements describe). When we speak of the logic of
someone’s reasoning, a decision, or an argument, this is generally taken to mean the
structure that might be shared whenever the particular subject in question is changed.
Accordingly, logic is often identified by comparison. “By that line of reasoning,
thus and such must follow.” “If that argument holds, then this other argument would
also hold.” Once we have identified a pattern across enough instances, we may name
the pattern such that new cases are merely grouped into a pre-existing class, such as
modus tollens, post hoc rationalization, circular reasoning, or ad hominem. While
various forms of “logic” may vary in the extent to which they ignore semantic
details within a given context, this notion of generalizing across two or more con-
texts is relatively consistent.
Logic’s generalization across semantic contexts is accomplished and expressed
in a few key ways. First, formalized logic often uses logical variables to stand for
objects, sets, or claims, since it is understood that the subject of logic is whatever is
fixed when these elements vary. In other words, the notation expresses a key assump-
tion about the nature of logical investigation. Piaget made frequent use of logical
variables for claims, such as p and q, or variables for sets, such as A and B. Not only
then do these notations express the implicit commitment to varying the semantic
content in question, but they also afford the comparison between those semantic
domains. Which domains have the same logical structure? Those that are modeled
by the same variable expressions. However, if we think about taking various domains
of discourse and mapping them onto the bare structures of logic, such as p ⊃ q or
A × B, then we should note that this mapping entails a deep loss of information.
While some “structure” is retained, or even made visible, by this representation,
much else is lost in translation. What then does logic express?
342 P. C. Dawkins
There are key concepts that are central to logic and its ostensive ability to express
human argument and possibly even reasoning. First, the primary aspect of a propo-
sition p that is retained by logic is its truth value: true or false. What, however, does
this truth value mean for a proposition? Ostensibly, it could convey whether it cor-
responds to an actual state of affairs, but this notion is only occasionally invoked in
logical discussions. The logical idea of validity dispenses with the need to actually
claim what is true about the world since it requires an argument that the conclusions
are true whenever the hypotheses are true. In this sense, the truth of the hypotheses
is merely given as a starting point. This suggests that truth operates more like a light
switch that can be toggled for some variables at will. Rather than corresponding to
what is, logic provides tools for considering what could be the case for hypotheses
and what must be the case for conclusions as a result. Indeed, a feature that naturally
arises in logic when considering truth values of related propositions is to consider
all of the combinations of truth values (a truth table), regardless of their correspon-
dence to reality. This is a key concept we shall revisit later.
Already, to discuss truth values, I have needed to implicitly introduce some other
key elements in the formal logical framework. First, I have needed to bring in mul-
tiple variables together to consider their internal relationships (“p or q,” “p and q,”
“if p, then q”). Logic operates by relating multiple elements within its scheme to
consider how they interact. Logic somewhat partitions itself off from concerns
about what is true of the world to consider questions about whether a given set of
claims are coherent or contradictory (p and ¬p cannot both be true), whether two
given claims are equivalent (p ⊃ q is equivalent to ¬q ⊃ ¬ p), independent (p ⊃ q
and q ⊃ p are independent), or one entails the other (p and p ⊃ q entail q), etc. By
replacing claims of truth with truth values, logic accomplishes this separation from
description to consider various relations among claims themselves. Indeed, formal
logic is generally explained to operate not on states of affairs at all, but on sets of
statements or other units of language. The actual reference of particular sentences to
states of affairs is replaced in formalized logic by either truth values or truth sets and
false sets. Aristotle’s logic is often called a term logic since it considers categories
such as even number or square that are neither true nor false, but rather have objects
of which they are true and objects of which they are false. Modern logics are often
propositional or predicate logics since they operate on claims that have truth values
such as “15 is an even number” or “x is an even number.” The point is that modern,
formal logic may be understood as a meta-language whose objects of study are units
of language. These units of language both stand in for states of affairs and afford
logic’s dissociation from states of affairs. This is why logic is often described as
syntactic rather than semantic: Formal logic generalizes across semantics using
mere syntax. Logic deals with whatever statements are shared by virtue of their
shared form or shared grammar (specifically connectives and operators like or, and,
if/then, and not).
Second, by discussing hypotheses and conclusions, I have made an implicit ref-
erence to the key logical relationship we may call implication (often notated p ⊃ q).
Implication would seem to provide a point of contact between logic and reasoning.
The logical notation of implication should capture when one claim or piece of infor-
mation follows from another in some sense. We shall need to consider the particular
senses of implication, but it is enough, for now, to say that logic classically took this
to mean “necessarily” or “in all such cases.” This could be a tool to express an actual
inference that someone draws: “Since I believe this, it must also be true that…” It is
this (potential) relationship between logical implication and human inference that
we will turn to in the following sections.
Logic and Psychology: Respectfully Disjoint
Despite the frequent recourse to cognitive terms in early logic (such as Boole’s book
The Laws of Thought), many early logicians clearly denied that logic held any direct
relationship with human reasoning. Frege and Husserl, in particular, rejected this as
psychologism. The logician Beth noted that this was an influential view that led
many to reject any semantic elements of logic:
Semantics also allows the introduction of arbitrary sets and series of formulae, but further
it brings in a conceptual apparatus permitting the study of the meaning of certain symbols
and of the truth or falsity of certain formulae. This step has given rise to vigorous protests,
because it was incompatible with certain widespread concepts according to which, in
particular:
1. The concepts of meaning, truth, and falsity incorporate a psychological element which is an
obstacle to an analysis laying claim to a standard of mathematical exactitude.
2. The concepts now mentioned, as far as they are applied to mathematical symbols and state-
ments, call upon a realist conception of mathematical entities. (Beth & Piaget, 1966, p. 69–70)
This supports my prior claim that the notion of validity can actually operate to
separate logic from any need for correspondence to reality such that truth and falsity
take on somewhat artificial meanings in logic.2 Psychology simply introduces into
logic far too much variance and uncertainty for it to operate as a field of mathemat-
ics, which is the primary origin of modern logic.
Why, then, would Piaget make heavy use of logic in psychology? Some claim
that he did so initially in some state of ignorance about logic. Quine (1960, quoted
in Burman, 2016) is quoted as referring to “Piaget’s persistent and evidently incor-
rigible stupidity over matters of logic.” It may alternatively be the case that Piaget’s
project was largely misunderstood (or some of both). In Piaget (1950), he cautioned
2
It is beyond the scope of this chapter to cover the variety of logical tools and their conformity to
this discussion. I shall focus on the logic most closely associated with Piaget’s own modeling,
which is largely first-order logic. Of this, Schroyens (2010) notes, “The degree of idealization in
classic first-order logic is the prime reason for anti-psychologism. Contemporary logics are much
less idealized (i.e., there is less abstraction, generalization, and/or simplification involved).”
(p. 79). The extent to which other logics may be more suitable as modeling tools of human reason-
ing (as explored by Stenning & van Lambalgen, 2008) is irrelevant to my discussion of
Piaget’s work.
344 P. C. Dawkins
those who wanted to make “thought the mirror of logic,” saying, “reverse the terms
and make logic the mirror of thought, which would restore to the latter its construc-
tive independence” (p. 30). This suggests that his goal all along was to remake the
tools of logic toward a new purpose, which I shall argue he in many ways did, how-
ever, wittingly.
In Piaget’s later work, we get a clearer reflection on his understanding of the
relationship between formal logic and psychology. Piaget engaged in collaboration
with the logician Beth and developed his use of an explicit stance toward logic. In
the collaborative book Piaget wrote with Beth, Piaget affirms psychologism’s
critique:
We likewise condemn psychologism without hesitation, for it shows a confusion not only
of the methods but also of the problems themselves. In effect, if the logical problem, in the
case of a mathematical demonstration consists in discovering under what conditions it can
be accepted as valid, the psychological problem consists only in determining by what men-
tal mechanisms it actually develops in the mathematician’s mind. These two distinct prob-
lems, the one of foundations, the other of causal explanation, correspond, on the other hand,
to two heterogeneous methods, one of deductive analysis and the other of verification or
experiment, so that the failure of all psychologism is easily understandable. (Beth & Piaget,
1966, p. 132)
Piaget was thus happy to affirm that formalized logic did not concern human rea-
soning and reciprocally that the mechanisms of development of human reasoning
are not in some sense subject to the laws of logic. One reason for this separation that
Piaget notes is that the psychologist can choose which aspects of human reasoning
to study and explain.
It is therefore for the psychologist to decide whether, for the requirements of his
interpretations, he will limit himself to the study of fallacious or incomplete argu-
ments, or whether the arguments considered as valid by the logician interest him
also from the viewpoint of actual thought mechanisms… even if the mental mecha-
nisms corresponding to this formalisation are only exemplified by a small elite of
“subjects,” who are logicians considered as living and thinking beings… How can
we explain psychologically the possibility of “pure” logic and mathematics (the
term possibility being taken in the sense of possible realisation and not of possible
validity, and the term “pure” meaning simply independent of all content)? (ibid.,
p. 133).
Thus, Piaget is interested in explaining the mechanisms by which mathematical
logicians can come to reason about logical concepts such as validity void of any
specific content. Piaget’s goals for genetic epistemology included developing a the-
ory of learning rooted in biology that is comprehensive enough to account for all
stages of human development. Piaget understands that the emergence of formal
logic, its independence from psychology notwithstanding, is open to psychological
investigation. It is worth noting that his focus, even in this rarified arena, is upon the
“mental mechanisms [by which] it actually develops.”
Logic and Psychology: A Recurring

Methodological Conundrum
For the present discussion, we shall not follow Piaget’s inquiry into logicians’ rea-
soning, but rather return to the key conundrum of the role of content (or lack thereof)
in reasoning. This is an issue that plagues any psychological investigation that
maintains an interest in logic. Logic compares across contexts or operates on sym-
bols that take the place of any meaningful context. If logic bears any relevance to
human reasoning, then one would expect that human reasoning would be compa-
rable even as we shift contexts or replace contexts with symbols (in the presentation
of a task). Empirically, this is clearly not always the case (e.g., Evans & Over,
2004). Human reasoning shows great sensitivity to variance in context even when
the structure of a task (from an expert’s “logical” standpoint) remains largely the
same. This shows that human reasoning cannot be modeled or explained merely
using logical tools.
However, no one claims that there are no content general trends in human reason-
ing that can be reasonably described (or even explained) using aspects of logical
structure. For instance, two robust and commonly cited phenomena in how people
reason about conditional statements—statements of the form “if…, then…”—are
the if-heuristic and matching bias (Evans & Over, 2004; Inglis, 2006). The if-
heuristic refers to people’s tendency to focus their attention on a conditional state-
ment in cases where the “if” portion of the statement is true. This stands in opposition
to the logical notion of imagining the full truth table of possibilities. Matching bias
refers to the frequent way in which people process a negative claim, such as “not a
flower,” by thinking about the positive category of flowers. Both of these content
general patterns of human reasoning can at least be described in the language of
logic: hypotheses and negations.
Piaget made a similar argument for why logic might be a relevant tool for model-
ing. In Beth and Piaget (1966), he explained,
We immediately see the main difficulty: It is to establish how far the “structures” thus pre-
supposed in the operative mechanism (actions and operations), which is above all opera-
tional as far as the subject’s activities are concerned, really belong to him or are introduced
by the psychologist himself, that is to say, by a subject no. 2 studying subject no. 1 and
projecting on to him his own mental structures… in the subject’s activities there exist struc-
tures in the two-fold sense: (1) of systematic wholes with laws of combination peculiar to
the system as such, and (2) of systems capable of exhibiting the same forms independently
of their different content. (p. 167–169)
Researchers may observe similar behavior across tasks either because the tasks
were designed with a similar categorical structure or because the reasoner engaged
in a similar way of reasoning about the two tasks (whether or not that similarity was
intended or even perceived by the researcher). In either case, it is reasonable to
investigate which aspects of people’s reasoning have such context-independent
structures and under what circumstances they may be observed.
346 P. C. Dawkins
The more fundamental conundrum for studying reasoning to which logic may be
relevant3 is the extent to which a researcher can distinguish the influence of seman-
tic content versus the influence of syntactic form in an individual’s reasoning on a
particular task or set of tasks. At the risk of a slight digression, I shall give some
examples of this issue in practice. Markovits (2004) relayed the findings of one
study, saying:
Young children do have problems making [the modus ponens] inference when contrary-to-
fact premises are used… For example, if given premises such as “if it is raining, then the
grass is dry,” young children show a strong tendency to conclude that “it is raining” implies
that “the grass is wet.” However, when these children are given some external support…
they consistently make the logical modus ponens inference. (p. 323–324)
This “external support” takes the form of telling children somehow that we are play-
ing a game of make-believe in the task. Why would psychologists want children to
draw inferences from “contrary-to-fact premises”? The challenge they face is that to
give children the premise “if it is raining, then the grass is wet” is to affirm what
children already know. In this case, how do we know whether the child’s reasoning
is from the premise (as is understood in logic) or is merely based on prior knowl-
edge (which psychologists of reasoning have sometimes deemed “beliefs-bias,”
Evans, 2002; Evans & Feeny, 2004)? By giving a counterfactual premise, the
researcher can ostensibly distinguish the two. Fortunately, children are intelligent
enough to ignore the researcher’s silly premise and to reason from what they under-
stand about the world. They do not know that the researcher is attempting to play a
logic game in which we (for some reason) only care about the relationships among
statements, not their correspondence to reality (“truth” in the intuitive sense).
What about the researcher’s view of logic has given rise to this very strange idea
that they want children to reason from contrary-to-fact claims in order to demon-
strate logical competence? The real issue is related to the researcher’s need for a
certain kind of evidence, not necessarily the nature of the reasoning to be studied.
Logic, to an expert, regards the form of the argument, not its semantic particularities
or its correspondence to the world. Human reasoning, in contrast, is naturally sensi-
tive to semantic particularities and to their views of the world. The researcher needs
some effective tool for isolating the feature of reasoning they care about, which is
syntactic form. The two natural ways to do this are (1) to provide tasks whose solu-
tions are so counter-intuitive that no one would solve them correctly unless they
were attending merely to syntactic form or (2) to provide a range of activities that
maintain syntactic form and vary semantic detail to see when and how responses
remain invariant (in other words, do people reason the same about “if p, then q” as
we systematically vary the values of p and q?). Both approaches have been tried at
a great scale by psychologists. Piaget exclusively used the latter. Somewhat like the
first approach, Piaget attended to the extent to which students reasoned about states
3
By which we may mean (1) those aspects of reasoning that generalize across semantic context,
(2) reasoning sensitive to syntactic form, or (3) performance on tasks whose intended solution is
understood to depend upon logical structure or principles.
of affairs beyond what they observed in the form of thought experiments. The pri-
mary tool he used to study reasoning related to logic was to compare student reason-
ing across quite different semantic contexts. When we get to reviewing some of his
particular tasks, we shall notice a stark difference between what he took to be the
shared structure of tasks and the shared syntax employed by many later psycholo-
gists. Thus, Piaget faced the same methodological issue that other psychologists
face: In semantically rich tasks, how do we isolate what is “logical” about some-
one’s reasoning from what is semantically particular to it? Piaget’s distinct solution
to this methodological issue speaks to a very different image of what he meant by
logic as compared to many later cognitive scientists.
Extension and Intension
I have argued that logic generalizes across various semantic contexts by mapping
the various kinds of claims that people make into a generalized language. This lan-
guage captures some shared “structure” and also removes much semantic detail.
What is expressed and what is removed when one maps onto the logical notion of
implication (p ⊃ q)? Stated another way, in what sense does logic express that one
claim implies another? Implication is often associated with the sentence structure
“if…, then…,” which I shall use for ease henceforth.
One dichotomy that will help us contrast Piaget’s use of logic with the formal-
ized tools is that between extension and intension. For convenience, we will con-
sider the following example statements:
1. If an integer x is a multiple of 6, then it is a multiple of 3.
2. If an object is made (completely) of metal, then it will conduct electricity.
3. If a person is a politician, then they are a liar.
Intension refers to the meaning of the term or phrase. Regarding statement 1, the
antecedent can be defined as “there exists an integer k such that x = 6k” or “x can be
divided by 6 with no remainder.” Extension refers to the truth set of the property,
meaning {… − 12, −6, 0, 6, 12…}. Classical logic, by replacing “multiple of 6”
with p and “multiple of 3” with q, has no way to capture the relationship between
the intensions. If we merely look at truth values, we can merely say, “There is no
integer x that makes p true and q false.” If we consider truth sets, we can capture the
classical Aristotelian notion that statement 1 is equivalent to “All multiples of 6 are
multiples of 3,” since the truth set of p will be a subset of the truth set of q. This
relates to the extensions of the properties. This is characteristic of logic in that it
captures extensions and not intentions. For instance, an “axiom of extension” in set
theory states that two sets are equal precisely when they share the same elements.
There is a cognitive difference between “the set of all multiples of 3” and “the set of
all sums of three consecutive integers” (3 = 0 + 1 + 2, 6 = 1 + 2 + 3, 9 = 2 + 3 + 4,
etc.) since they have different intensions. The axiom of extension can be taken to
mean this difference is formally irrelevant since the sets’ extensions are prioritized,
348 P. C. Dawkins
and the extensions are indistinguishable (the exact same objects share both proper-
ties). Frege classically tried to address this issue with his distinction between refer-
ence and sense, but this has not really been captured in most systems of formal logic.
Note that an assumption within the logical model of extensions is that both the
antecedent (“if” part) and consequent (“then” part) of each statement have an exten-
sion (independent of each other), and we may compare those extensions to decide if
the statement is true. We may imagine the set of multiples of 6 and confirm that they
are also multiples of 3. We may imagine the collection of objects made of metal and
the collection of objects that conduct electricity and interpret the statement to con-
firm the first is contained in the second. Interestingly, the common definition of the
truth of a conditional in logic focuses more on what the statement does not allow:
an object that is in the extension of the antecedent but not in the extension of the
consequent. This is more often stated that a case making the first part true and the
second part false is a counterexample to the statement (p and not q or more briefly
p ∧ q )4. Interestingly, one of the best-supported modern psychological theories of
how people interpret “if-then” statements claims that people take these statements
to mean that the consequent is highly probable whenever the antecedent is true (see
Evans & Over, 2004). This allows for the empirical finding that people still affirm
these statements even in the presence of at least some known counterexamples. We
leave it to the reader to determine whether statement 3 has known counterexamples
or not. This psychological theory claims that people use statement 3 not to claim
there are no honest politicians, but to state that the extension of “politicians” is
mostly in the extension of “liars.” In this case, the consequent does not follow from
the antecedent “necessarily,” but only “usually.” In mathematics, however, insis-
tence on no counterexamples to statements is conventional.
We should note that there is a certain convenience to the formal logical model
ignoring intensions. If we require these statements to convey a particular relation-
ship between the meaning of the two parts, then we now have a much more complex
and challenging question regarding what they mean, when they are true, and how
people understand them. The problem is that human reasoning does not ignore
intension. People may notice a pattern about metal and electricity without being
able to explain it (a relationship between extensions rather than intensions), but in
many other cases we reason about these claims based on our meanings for the con-
cepts (especially mathematical concepts). We shall consider these possible relation-
ships in the context of Piaget’s tasks in the next section.
Implication and Inference
Consider the following implications reflecting aspects of student reasoning on tasks

Piaget used in his studies of logic (Inhelder & Piaget, 1958; Piaget & Garcia, 1991):
4
The bar is a conventional representation of negation, so q is typically read “not q.”
1. If a rolling ball hits a boundary at a 30° angle, then it will bounce off at a 30°
angle in the other direction.
2. If the rod were longer, then it would be more flexible.
3. If I mix the liquid in the first and third and fifth beakers, then the mixture
turns pink.
4. If I use the hook, then I can lift the monkey out of the box by its tail.
As I shall note later, students did not make these precise statements in Piaget’s stud-
ies, but it is useful to consider these claims in the language of the model of logic (as
Piaget does in various ways).
What are the extensions of these statements in children’s reasoning? A child
might say something like statement 4 only to describe what they just observed,
meaning the extension at that moment is only a single ball. Alternatively (or implic-
itly), it may be anticipated to refer to future balls. Statement 5 has an even less clear
extension since it makes it sound as though the rod must be modified. That is to say,
it expresses an insight into the relationship between length and flexibility (the inten-
sions), but that expression does not compose well with the logical model (rooted in
extensions). We might repair this issue by revising the statement to refer to pairs of
rods, such as, “If two rods differ only in length, then the longer one will be more
flexible.” However, it may be the case that this modification to make the claim have
a clearer extension does not correspond to anything within children’s line of reason-
ing. We shall consider statement 7 much later in the chapter because it has a funda-
mentally distinct structure. The “if” part of that statement does not really have an
extension in the world (a set of objects with the given property), but rather describes
a possibility within the child’s own activity. We include it both because we will
discuss it in a later section and to provide contrast to the structure shared by state-
ments 4–6.
The distinction I intend to point out is that while the logical model attributes
extensions to the two parts of each claim, it is much less clear how extension can or
should be understood with regard to child thinking. I find it helpful to begin from
the standpoint that Piaget does not intend extension to cover all instances to which
the adult might see these statements as applying. For the children engaged in
Piaget’s tasks, extension is empirically operationalized in terms of the set of objects
they choose to interact with in the task (we shall consider some examples later on).
He treats the question “What is the extension of this claim for the child” as part of
the investigation. For older and more advanced reasoners, he noted when the exten-
sion of their generalizations seemed to stretch beyond the scope of the data they had
observed. Indeed, the concepts of assimilation and accommodation (see Chap. 3)
can easily be applied to describe whether new experiences are taken to be within the
extension of a claim the child made.
The main issue beneath extension in student reasoning about implications in
Piaget’s tasks is that (1) children have only seen a very small number of examples
to which any statement applies and (2) they are relying on either inductive general-
ization or some relationship of meaning to extend the scope of the implication
beyond the few cases observed. At the very least, children would need to generalize
350 P. C. Dawkins
based on an assumption that the situation has great regularity either as they search
for some explanation for the implication (statements 4 or 5) or when one is simply
unavailable (statement 6, in which the chemistry-based explanation is not accessible
to the student and is not intended to be). Interestingly, the child’s assumption of
regularity in the situation generally corresponded to a feature that Piaget and his
colleagues built into their research tasks. Piaget did not really discuss task design in
the books on logic.5 However, a quick review of the tasks in Inhelder and Piaget
(1958) shows that they were all built to invite and afford generalizations. These
often corresponded to general laws of physics, such as density determining whether
something floats, the law of equal angles when a ball bounces off a flat boundary,
the oscillation time of a pendulum, etc. In the early chapters, these generalizations
could take the form of implications, while in later chapters, the intended generaliza-
tions had more specific forms, such as proportionality between quantities.
These examples provide a key answer to the question posed above. In what sense
does one claim imply another? In the situations Piaget presented to children, the
answer was often physical causality as regulated by various physical and chemical
principles. This was a very specific setting in which to study the relationship
between logic and reasoning. Piaget does not address the question of whether these
phenomena of reasoning generalize to other settings. Statements 4–6 are not like
statement 1, where the meaning of “multiple of 3” can actually be deduced from the
meaning of “multiple of 6” (i.e., a proof is possible). However, they express more
than a mere generalization, like statement 3. There are physical laws governing the
relationships, even if children engaging in the task will not be fully aware of them.
Why was the choice of generalizable physical relationships an appropriate and
productive choice of context to explore logical reasoning? First, tasks that opera-
tionalize physical laws allow children to gloss over the issue of extension since,
according to the physical principles at work in the tasks, the intended generaliza-
tions from the few observed cases to all future cases were reasonable. Second, they
allowed children to make incorrect or provisional generalizations so that the
researchers could use these predictions to gain insight into children’s reasoning.
This means these task contexts have methodological benefits for the researcher.
Third, it allowed the researchers to not only evaluate and model the children’s rec-
ognition of a generalization but also the quality of their explanations thereof. Finally,
it allowed the children the opportunity to test (or not) various generalizations by
generating further data within the task. Their ability to seek such (counter)-exam-
ples or to respond to those provided by the researcher afford greater insight into
their reasoning.
Piaget explained the assumptions behind his modeling of children’s reasoning in
such settings: “We merely observe that each normal subject… constructs inferences
and understands those of other people and evaluates them as true or false, not only
as regards their agreement with the real world but from the point of view of a certain
5
He did point out features of tasks that distinguished them from others such as the presence of
irrelevant variables as in a pendulum task in Chapter 4 of Inhelder and Piaget (1958).
internal coherence (non-contradiction)” (Beth & Piaget, 1966, p. 154). This duality
between agreement with experience and internal consistency is crucial to under-
standing Piaget’s application of logic as a model. What is more, we must understand
that he made a continual distinction between (1) when children noticed a pattern
that phenomena of type A were always followed by outcomes of type B and (2)
when there was an internal model of the situation that led children to observe/expect
that B followed from A. Soon after the previous quote, he said,
Contrary to the facts of behavior, facts of consciousness are not dependent on most of the
habitual categories applicable to physical reality: substance, space, movement, force, etc.,
and, in a general way, causality. In fact, if states of consciousness develop in time, we still
could not say that they “cause” one another, because they entail one another according to a
mode of connection more noetic or inferential than causal. Their fundamental character
consists, from the cognitive viewpoint, of meanings; and of values, from the affective
viewpoint. Now, a meaning is not a “cause” of another meaning, nor a value of another
value, but they entail each other by means of what we might call, for want of a better expres-
sion, a kind of “naïve” implication, taking this implication in the ordinary sense of “entail-
ing” and not in the technical sense. (ibid., p. 154–155)
Notice here that Piaget is drawing a distinction between the relationships among
understandings that entail one another and the actual relationships of causality. Your
understanding of what it means to be metal may simply be associated with conduct-
ing electricity, or it may actually help to explain it. In either case, these conceptual
links can be distinguished from the actual physical laws by which metal and elec-
tricity interact (your understanding may be inaccurate). The conceptual links are the
subject of psychological investigation, and while children are forming their under-
standing we must remain flexible to make sense of the relationships they conjecture
and test. The researcher modeling children’s reasoning must attend to the relation-
ships of entailment in the child’s reasoning, not merely to matters of causality (as
the researcher understands them).
Considering the context of these “implications” thus reveals a key departure
between Piaget’s use of the conditional p ⊃ q as a modeling tool and its formal
structure. Whereas formal logic attempts to ignore intensions and prioritize exten-
sions, Piaget’s notion of how children would link parts of an implication was firmly
rooted in the child’s meanings for the two parts of the statement. More specifically,
he did not try to separate the child’s conception of the physics at play from the
“logic” of the statement. In Beth and Piaget (1966), he said,
To describe this new structure, we shall for convenience employ the usual notation of the
two-valued logic of propositions, but we stress that this in no way implies, either that the
subject imposes on himself rules equivalent to the logician’s axioms, or that the natural
employment of the operations which we shall write (p ⊃ q), (p ∨ q), etc., conforms to the
logician’s usage (p. 180)
Indeed, he distinguishes the researcher’s model from the knowledge of the subject,
Of course, the subject does not think about the operations he uses and he could not formu-
late them… The structures we are discussing here do not exist as distinct “concepts” in the
subject’s consciousness, but are only manifested in his behaviour. Thus it is the observer
and not the subject who notices and formulates them by referring to a model (ibid., p. 181)
352 P. C. Dawkins
We now have the necessary aspects of logic described to further explore Piaget and
colleagues’ use of logic as a model and their unique transformation of logic for their
purposes. In the following sections, we shall examine some specific tasks that Piaget
used and his analysis thereof to learn about how the model operated for children
while reasoning in physics and chemistry contexts. We remind the reader that we are
examining logic as a central modeling tool to more broadly make sense of how any
cognitive researcher might use formal tools in the service of constructivist model
building.
The Tasks
To begin our analysis of Piaget and colleagues’ use of logic as a model, in this sec-
tion, we shall describe the tasks in the early chapters of Inhelder and Piaget (1958)
and their relation to the model of logic. Each chapter presents an analysis of chil-
dren’s performance on a single task.
In Chap. 1, children were given an apparatus that launched billiard balls onto a
billiard table (the billiards task). The apparatus pivoted around a fixed point of rota-
tion and the balls were launched against one of the boundaries. The key concept to
be discovered in the task is the equality between the angle at which the ball hits the
boundary and the angle at which it rebounds. Piaget modeled this relationship
briefly in terms of the two angular quantities (though the children did not actually
identify numeric measures), but more often discussed it using propositional logic.
To do so, the propositions were all of the form “x = a” or “y = a,” where x, y repre-
sent the angles of approach and rebound and a stands for some particular measure.
Under this rendering, the target understanding was always biconditional in nature
(what is often called “if and only if”).
Chapter 2 reported children’s responses to the task of predicting, describing, and
explaining which objects will float in tubs of water (the floating task). The objects
included a piece of wood, a wire, metal needles, a pebble, a wooden ball, a metal
key, and a nail. The key concepts at play were volume, weight, density, and specific
gravity. Naturally, children often attended first to matters of volume and weight,
which made it hard for them to come to a self-consistent classification. However, the
law that floating depends only on the density of the object as compared to the den-
sity of water provides a “single and noncontradictory relationship.” Clearly, each of
these attributes can be quantified, but none of them is actually measured in the task.
Accordingly, Inhelder and Piaget (1958) did not model the quantities but rather
propositions such as “let p be the assertion that a given object floats and p the
assertion that it does not, q the assertion that its volume is equal to that of a certain
quantity of water, r the assertion that it is lighter than that quantity of water” (p. 41).
This modeling of the quantities in terms of Boolean comparisons matches the argu-
ments given by more advanced children who imagine comparing the objects to the
comparable volume of water. I claim that translating this task into the language of
propositional logic in this way is far from obvious, but it resembles some participant
arguments. As I noted above, this task invites various generalizations from children:
small things float, metal things sink, etc. This provided a rich opportunity for the
researchers to consider how children formulate and test these implications as they
explored the task.
Chapter 3 reported on an experiment in which children were presented with rods
that varied according to four attributes: material, length, thickness, and cross-
sectional shape. These rods were suspended by one end over a pool of water, and a
weight was placed on the opposite end. The weights varied, creating a fifth attribute
of each case. The question posed to students was to describe and explain which rods
were flexible enough to touch the pool of water below (the flexibility task). Inhelder
and Piaget (1958) employed three primary formal tools for modeling student rea-
soning about this task.
First, they described the class relations and inclusions (a common focus in
Piaget’s work). For example, “Let us call A1 the class of rods which are 50 mm or
more long and A1′ that off rods <50 mm; A2 the class of weights of 300 grams or
more and A2′ the that off weights <300 grams… Finally, X will be the class of rods
touching the water and X′ that of rods which do not” (p. 53).6 They use the variable
B to represent the universal class of all such rods, which are then decomposed or
partitioned into the subclasses. Since they used addition to represent the union of
classes, B1 A1 A1 . Piaget used multiplication to represent the simultaneous clas-
sification of cases by different features. According to these conventions, the authors
noted that there are eight subclasses created by the two conditions (length and
weight) and the one outcome (touching the water): (B1) × (B2) × (X+X′) = A1 A2
X + A1 A2 X′ + A1 A2′ X + A1 A2′ X′ + A′1 A2 X+ A′1 A2 X′ + A′1 A′2 X + A′1 A′2
X′ (p. 53). This equation is intended to organize the classes of rods by the various
combinations of attributes.
Second, Inhelder and Piaget (1958) rendered the same basic structure in terms of
propositional variables by simply calling the presence of A1 the proposition p and

the absence F L F L . They thus introduced six such variables for the five attri-
butes of the rod/weight and the outcome regarding touching the water.
Third, the authors briefly provide a notation for varying a specific attribute. For
instance, “if F designates the transition from thicker to thinner and L the transition
from shorter to longer and F and L the inverse transitions” (p. 63) then they express
some children’s recognition of a compensation as p0 .7 Like the example above, this
pushes the boundaries of logical expression as such. The action of making it thicker
does not have a truth value and is thus not a proposition. It also does not directly
relate to an action children took, but rather their comparison of different rods. The
6
Since these letters represent sets or classes, the prime notation may be interpreted as a comple-
ment set.
7
We reproduce the statement as it appears in our translation of the text, though clearly the explana-
tion was meant to have arrows above F and L.
354 P. C. Dawkins
point is that this notation is much more the author’s innovation rather than a simple
application of logical representations. Most importantly, it again corresponds quite
closely to student explanations in a way we shall explore later in the chapter.
In Chap. 4, the authors report on children’s investigation of the swing time of a
pendulum, where they can vary the length of the string, the weight suspended, the
height from which the object drops, and the initial force applied (the pendulum
task). Only the length is relevant to the swing time, so the task focuses on children’s
distinction between relevant and irrelevant attributes in the situation. The logical
representations used in this chapter are similar to the first two in the previous chap-
ter, though classes are used only briefly. A key novelty is that the authors introduce
a new notation to represent two variables that have no underlying relationship
(q × x).
In Chap. 5, children explore how far a ball jumps when rolled down an adjustable
ramp with a small incline at the end to send the ball into the air (the ramp task). The
ramp could be raised or lowered, and the ball could be dropped from different
heights on the ramp. The ball landed in one of a series of numbered slots so the
children could easily track how far it traveled. There were also three different-sized
balls. The goal was for children to identify that the height of release (not the angle
of incline or length of ramp traversed) uniquely determined the distance of travel
(ignoring weight). A notable aspect of this task, like the previous, is that the out-
come was not Boolean in nature (like the floating or flexible rods tasks). There is
little about this task that would lead a researcher to anticipate using propositional
logic to model children’s reasoning about it. Indeed, without any of the variables of
the problem being Boolean, how should one translate this into the relationship
among propositions? In this chapter, Inhelder and Piaget (1958) exclusively used
propositional variables, but they gave them new meanings: “Let us call p0 the state-
ment of the conservation of slope and q0 the statement of variation of this factor;
call q0 and r0 the same statements made in reference to distance; r0 and x0 the
same in reference to height; finally, we can designate by x0 and F L F L the
statements affirming or denying invariance in the result obtained” (p. 91). This
approach combined the propositional variables used in Chap. 3 with the attention to
variation introduced in the expression p.q . This is a clever tool for rendering this
continuous problem Boolean, but it has two other rather important features. First, it
matches Piaget’s broader interest in children’s construction of relationships of
invariance (what can change while the outcome remains the same). Second, as noted
in previous cases it matches closely the way children actually reason about the task
(new trials are interpreted by comparison to previous ones).
In Chap. 11, the authors analyze how children understand balance when they can
place different weight objects in baskets that are suspended on either side of a ful-
crum (the balance task). The baskets can be moved closer or further to the fulcrum
such that the primary variables are weight and distance. Balance is inversely propor-
tional in that twice the weight will balance at half the distance.
Modeling Stages and Stages of Modeling
Piaget and colleagues were fond of identifying stages of children’s development

that, though associated roughly with particular age bands, arose empirically by clus-
tering similar responses to each task. They typically did not describe sampling pro-
cedures or how children were recruited to take part in the studies, in part because it
was assumed that most subjects would pass through the given stages as they aged.
The same subjects were not tracked over the course of years on a single task. The
viability of this approach is implicitly justified by Piaget’s findings that children’s
reasoning followed reasonably similar patterns within certain age bands, and those
patterns recurred across various tasks. As he explained in Beth and Piaget (1966):
We have tried… not to construct the most general possible mental structures, but rather to
discover the most elementary actual structures, which therefore means the most restricted
in their manner of combination and the most specific insofar as tied to their natural function.
In so far as these structures, nevertheless, exhibit a certain degree of generality (however
restricted), this does not arise as a result of a search for generality on the observer’s part, but
from the objective fact that the elementary structures which are achieved or find their equi-
librium in subjects at a determinate stage (about 7–8 years of age) are relatively general at
this particular stage. (p. 171–2)
Indeed, the composition of many of Piaget and colleagues’ books is to cover a range
of studies in a strategic sequence before drawing larger conclusions toward the end.
Piaget argued that the reasoning he studied was logical inasmuch as the patterns of
reasoning recurred across semantically distinct tasks. To the extent that the patterns
of reasoning recurred—as well as the progressions of reasoning and the hypothe-
sized mechanisms of advancement between stages—they had the hallmark quality
of logic: content generality. As we noted above, this is one natural methodological
solution to the conundrum of how to study reasoning that might be called “logical.”
However, unlike many other researchers who study logic, Piaget often used these
same experiments to study how children develop more content-specific notions
(quantities of length, weight, volume, etc.). He could thus productively compare his
content-specific modeling of student understanding with his content general
modeling.
In Inhelder and Piaget (1958), the chapters generally described performance on
each task within each of three stages, each usually divided into two sub-stages.
These stages were shared with other investigations they wrote. Stage I (up to about
7–8 years) children often failed to identify general relationships in the situation and
were not fully perturbed when the researcher provided (what they took to be) coun-
terexamples to the child’s explanation. Indeed, seeming self-contradiction was not
always avoided by children in Stage I. On the floating task, a 4 years old is reported
as having the following exchange8:
8
Inhelder and Piaget (1958) used italics to quote children and non-italics for exposition and inter-
viewer quotes.
356 P. C. Dawkins
Says… of a piece of wood that “it stays on top. The other day I threw one in the water and
it stayed on top.” But a moment later: “Wood? It will swim anywhere.”—“And this one?” [a
smaller piece].—“The little wood will sink.”—“But you told me that the wood would
swim.”—“No I didn’t say so.”… “The pebble?”—“It will sink.”—“Why?”—“Because it
stays on the bottom.” (p. 22)
Piaget next described another 5 years old who predicted that wood would sink;
when it floated on the water, he tried to push it to the bottom. When asked if it might
sink some other time, the child says, “Yes.” The authors note that these children
neither search for a single and noncontradictory explanation nor seek to avoid con-
tradictory classifications of the same object (much less the division of objects into
two fixed categories). Inhelder and Piaget generally do not use logical notation to
describe Stage I thinking except to explain why such an attribution would be inap-
propriate. They note on the floating task that children’s classifications “can be
defined by the following characteristics: (1) the subclasses are not all disjunctive;
(2) they do not all have negatives; (3) they do not allow for grouping either by
simple hierarchy (inclusions and complementaries) or by multiple hierarchy (dou-
ble- and triple-entry tables)” (p. 25). Thus, expressions like the eight-fold decompo-
sition by three attributes portrayed in the last section would be an inappropriate
model of these children’s categories.
Stage II (generally 7–11 years) was generally marked by children formulating
generalizations and an effort to avoid contradiction. Not only did children in Stage
II construct classes that have a more stable and composable structure, but also Piaget
attributes these to a content general way of reasoning that Stage II children have
constructed: “We must insist that the way they achieve this rough restructuring of
the concept of weight—one which avoids the inconsistencies of earlier formula-
tions—cannot be understood without considering the new logical apparatus com-
posed of concrete class and relation operations” (p. 30). These last phrases refer to
decomposing and recomposing groups into mutually exclusive and together exhaus-
tive categories such as “floating” and “sinking” comparable to the expression
B = A + A′. Accordingly, Inhelder and Piaget (1958) often use this set notation to
model these children’s reasoning and rarely use the propositional notation of p’s and
q’s. Children in this stage formulate general claims based on these more stable cat-
egories, but often are unable to formulate a generalization that is adequate to avoid
internal tensions (e.g., in the floating task and generalizations based only on weight
or volume but not density). Piaget is quick to defend the rationality of these explana-
tions even while seeking to understand how children move beyond them into Stage
III. We shall note in a later section at least one key distinction Piaget draws between
Stage II explanations and Stage III explanations.
Stage III (from 12 years onward) corresponded to the emergence of formal oper-
ations (see Chap. 7). It is in Stage III or sometimes substage III-B (usually age 13
and up) that normative solutions with viable explanations arise. Inhelder and Piaget
(1958) associated this with the emergence of new logical tools for formulating these
relationships. At this point, the authors often begin modeling the children’s reason-
ing using propositional variables. In the next section, we shall consider such uses of
the logical model.
Varying Use of Propositional Variable Expressions
In this section, I shall attempt to explicate Inhelder and Piaget’s (1958) varying uses
of propositional expressions from logic to describe children’s reasoning. We shall
note both how these uses subtly vary and their relative (in)-compatibility with logi-
cal expressions as classically used. In discussing how children use cases to compre-
hend nonimplication (in Chapter 2 of Inhelder & Piaget, 1958), the authors noted
that “the subject will find that to state the occurrence of the association p.q 9 is all
that is needed to discard the factor q… it is known, in effect, that p.q p q is the
negation of the implication p ⊃ q: p.q ” (p. 40). The authors’ use of the logical
variables here is reasonable inasmuch as they are claiming to express a content
general relationship that children at this stage are aware of a counterexample falsi-
fying an implication. The relationship, furthermore, is normative in the sense that
p. q q p.q ( p.q is the negation of an implication in standard logic. There
is a subtle gloss here in the notation regarding extension: The left-hand side of the
equation refers to a case, while the right-hand side refers to a generalization.
However, this suppression of quantification is rather common in the logical use of
propositional variables. The final thing to note is that this notation does not refer to
statements or children’s judgment about statements. The notation refers to chil-
dren’s understandings of the situation (the floating task) and relationships therein.
In the same discussion of children’s apprehension of nonimplication, Inhelder and
Piaget (1958) sought to express when children note two possibilities that coexist
with another outcome. They modeled the reasoning of a child who said, “The wood
can be heavy (or light) and it floats” with the expression p.q ( p.q ). In classical
logic, these two expressions (on either side of the equation) are true depending upon
the truth of the propositions p and q. The equivalence refers to all possible arrange-
ments of truth values. Inhelder and Piaget used these expressions quite differently.
First, the authors used or to express alternative possibilities. Or linked two cases in
the situation. The expression q ) is formally true if either one is true or the other is
true (or both). Since q and q ∨ q are negations of each other, p q p.q is a
tautology that says little (it floats or it does not; it is heavy or it is not). The key point
is that the authors used these expressions to refer to states that occur in the situation
and thus are “true.” The expression does not mean p. q is contingent (as in formal
logic), but rather that p and q co-occur (a heavy object can float). The point that this
means p does not imply q stands, but my focus here is to note how even these
expressions that are logically valid are being used in novel ways.
And in the text linked properties to express what occurs in a single case (an object
dipped in the water). p. q does not link two separate claims, but rather provides two
attributes of the same case. They used or, in contrast, to juxtapose distinct cases to
consider how the properties vary (different objects dipped in water). To see how this
9
In this book, the authors often use a period for “and” or conjunction.
358 P. C. Dawkins
already departs from logic, we should note that such and or or expressions cannot
be true or false in this usage. We simply represent a case if it occurs and do not
represent it if it does not. The variables do not have truth values, rather, their appear-
ance or absence expresses their truth value. An implication, on the other hand, can
be true or false. This captures a key idea from the set-theoretic view of these logical
operators. And intersects two sets, or unions them, and an implication (p ⊃ q) states
that one set is a subset of another. Notice the first two produce a new set, but have
no truth value. The third may be accurate or not. However, in formal logic an impli-
cation is equivalent to an or claim, meaning it cannot be distinct in this way. The key
for Inhelder and Piaget (1958) was that their appropriation of the notation allowed
them to express how children interacted with these tasks. In these analyses, expres-
sions with and or or alone generally represent what children observed or imagined
in the task (specific cases). An implication usually expresses some generalization a
child makes (accurate or not). Sometimes the expression was used to portray what
the child actually did; in other cases, it was used to illustrate the structure of the
child’s reasoning more generally.
The interplay between the cases each child considered and their generalizations
comprised the logic in the child’s reasoning. As noted before, if a child wanted to
test a possible implication, they might seek certain cases to falsify it. A key concept
in much of Piaget’s work was how the child’s conception of the task informs their
way of interacting with it. Especially in Stage III, children are not mere empiricists
observing patterns in what does and does not occur (though Stage II children at
times may appear to be so), since the implications are related to intensions and
intensions are part of how the child construes the task.
Syntactic Transformations as Researcher Inferences
To this point, we may be tempted to interpret that Inhelder and Piaget (1958) are
reappropriating logical notation in a way that truly dispenses with the formal logical
theory (by which I mean syntax and syntactic manipulations). We can already jus-
tify why they used this notation, even if in a repurposed way. The authors found it
convenient to use this content general notation because they believed they were
observing patterns of reasoning that generalized across various semantic contexts.
Not only this, they posited that students in Stage II form classes in each task because
they have new “logical apparatus” for doing this, meaning there is a general reason-
ing mechanism at play that becomes employed in the child’s engagement with
the task.
However, a closer examination suggests that the authors did not restrict their use
of logical notation to such a repurposed, descriptive task. An interesting argument
appears when discussing the limitations of Stage II reasoning in which (according
to the authors’ analysis) implications merely report on general patterns, but are not
sufficient to isolate relationships of necessity. Piaget was trying to downplay the
utility of the inductive generalization from “some” cases to “all,” claiming this
allowed a worthless extrapolation: “for p. q ⊃ (p ⊃ q) gives p.q ⊃ x , therefore p.

q ⊃ (p ⋇ q)—i.e., ‘some wooden objects float’ could imply any one of several
results in a particular case” (p. 43). This argument gives a more fundamental role to
the syntax of the model, since the authors felt free to transform a formal expression
of a child’s reasoning according to syntactic rules, whether or not there was any
evidence that these transformations corresponded to children’s observed reasoning.
In Chap. 3, while discussing the Flexibility task, Inhelder and Piaget (1958) pro-
vided a much more curious instance of such transformations of logical expressions.
In the summary discussion at the end of the chapter, Piaget was trying to express
how children reasoned about relationships of compensation, which we will discuss
at more length in the next section. The compensation in question related to when
varying two attributes of the rod simultaneously (e.g., metal, p, and thickness, q)
maintained similar flexibility (the rod touches the water in each case, x). They first
p p
wrote, “ p.q ⊃ x and x p.q p.q or = R ” (p. 65). They explained this
q q
as signifying that the presence of factor p and the absence or diminution of q gave
the same result as the situation vis-versa. Why though did they shift from implica-
tions from factors p and q, which the child could control, to an implication from
touching the water back to the factors? This is a bit strange since there are ostensibly
many other ways to obtain x, so the scope of discussion must be reduced from the
full scope of the task (restricted extension).
In the subsequent discussion, Inhelder and Piaget expressed this in terms of reci-
procity, which is part of the authors’ broader model of proportional reasoning (often
abbreviated INRC). Reciprocity represents reversing the direction of an operation.
They then applied this as an operator on a strange proportion of propositions, written
p.q R p.q . It is unclear what a ratio of propositions means (dividing the absence
of one attribute by the presence of another). Indeed, even if these propositions rep-
resented the classes of rods that realize those possibilities, it does not have a clear
interpretation. The authors explained, “This expression signifies (depending upon
whether it is read diagonally, vertically, or horizontally): (a) that p q R p q ;
(b) that p.q R p.q ; (c) that p. p R q.q ; and (d) that
p q p.q p.q p.q since o = Ro” (p. 66). Here, we again see the authors
extrapolating from a formal expression to various consequences that seem to cor-
respond primarily to syntactic operations rather than empirical claims about what
children were observed to do in the task. Each expression now avoids the strange
ratio structure and, in most cases, we can interpret the reciprocity operator as revers-
ing the claim to create an effect in the opposite direction (more or less flexible).
What though is to be made of the disjunction (entry b)? The disjunction was previ-
ously used to express various possible cases that may occur. In what sense are the
three rods designated by p q p.q p.q p.q the reciprocals of the other
three rods p. p , when indeed two rods are shared between the two expressions? The
final expression is the strangest when p clearly refers to no possible rods (both the
360 P. C. Dawkins
presence and absence of the same attribute). What do we mean by the opposite
direction in that case? This case does not obtain one side from the other by negating
both, since p is not the negation of q.
For our current investigation, our goal is not to decide upon the intended mean-
ing of these expressions. Our goal is to understand how the authors were using logi-
cal notation. Here, we see the invention of novel representations that break any clear
ties to standard logical notation. What is more, we see further evidence that the
authors took certain syntactic transformations of their logical models of children’s
reasoning to be acceptable forms of analysis. This expresses the fact that they do not
see these models as mere descriptions of what children did or said, but they under-
stand these models to have some structural parallel to the underlying modes of rea-
soning. As such, the transformations of syntax are understood to correspond to
cognitive processes or even mechanisms of inference.
We can gain more insight into this viewpoint from Chap. 11, in which the authors
discussed the balance task. In this discussion, it seems Inhelder and Piaget (1958)
understood the INRC group as a content general understanding of proportionality
that arose around Stage III-A (11–12 years), and “this is true in all spheres and not
only in the balance scale experiments” (p. 176). They applied this model to the bal-
ance task assuming p is a fixed increase in weight and q a fixed increase in distance.
p q
The notation q and = refers to diminutions of those qualities on the same side
q p
of the balance, while p′ and q′ represent increases on the other arm. The authors went
through a rather long sequence of symbolic manipulations on these expressions
involving the four transformations of the INRC group, the examination of which
would take us far afield. In one of these later expressions, they interpret nW = n : L
nL n : W
as saying p.q , p.q , p.q , p.q , where W and L are fixed weights and lengths
and n is a fixed constant such as 2. In thiswcase, we see that the “fixed increase”
described above referred to a multiplicative change (doubling or halving) and the
logical proportion appears to be standing in for a numerical proportion. This is a
strange mixture of quantitative expressions with logical ones.
To be forthright, the symbolic expressions in this passage are terribly hard to
interpret and make sense of. Nevertheless, the authors made the following claim:
Formulae (9) to (14) may seem much too abstract to account for the actual reasoning of our
subjects. Actually, this is in part an independent result of the symbolism which we have
introduced; nevertheless this is how proportions are discovered… Thus we are justified in
considering the preceding formulae symbolic expressions of the actual reasoning of our
subjects. (p. 180)
Inhelder and Piaget (1958) thus claimed that their syntactic transformations of these
symbolic expressions correspond to models of the children’s actual processes of
reasoning by which proportionality emerges in a context. This clearly points to a
foundational role that the logical theory played in their modeling work.
I am sympathetic to the reader who are dubious about whether symbolic expres-
sions that are hard for an adult reader (like myself) to interpret could reasonably
represent what pre-teens do to construct proportional relationships. To give the

authors the benefit of the doubt, I note that the propositions here represent manipu-
lations of the balance situation (i.e., children’s physical or mental actions) and the
INRC group is understood to transform those actions. It seems reasonable to infer
that the authors intended to model the ways that the forces on the balance (which the
child perceives) come to interact to achieve balance (in the child’s model in addition
to the physical apparatus). The logic in this interpretation is thus related to the rea-
soner’s coordination of their various interactions with the balance apparatus. As
Piaget explained in Beth and Piaget (1966):
From the viewpoint of the observer and not that of the subject, mental life must be consid-
ered as a system of behavior or conduct (including inner thought, unconscious or conscious,
but in the latter case considered as the interiorisation of actions expressed symbolically, as
possible anticipations, etc.). In such a perspective, the essential character of mental life is
its close connection with our actions, and intelligence itself must be conceived as a system
of operations, that is to say (Df.) of interiorized actions, made reversible and co-ordinated
in the form of “operational structures” exhibiting laws of totality (laws which the observer
can describe in terms of lattices, groups, etc., in short, in the language of abstract algebra).
(p. 155–6)
Note here that Piaget clearly viewed this construal of children’s understanding in
this logical syntax as the researcher’s accomplishment; the child was not being con-
strued as knowing logic or abstract algebra as such. As we noted above, implications
are between the child’s meanings in the situation.
While the authors found a means of rendering the task in communication with
logical expressions (some of their own creation, as best I can tell), this occurred at
a rather complex level of inferred cognitive mechanisms. Testing these hypotheses
about student reasoning cannot be accomplished by presenting them with state-
ments including or, and, and implies, since that is not the intent of the model as
Inhelder and Piaget (1958) were using it. The authors’ claim that “reasoning is noth-
ing more than the propositional calculus itself” (p. 305) has been much maligned,
but often by psychologists attempting to test this claim in ways utterly foreign to the
modeling work in the rest of the book. As noted above, psychologists have exten-
sively studied how people of varying ages interpret conditional claims, but this is
very different from the way that Inhelder and Piaget sought to elucidate how chil-
dren draw and explain general inferences in these complex physical situations. In
the next section, we shall consider one interesting hypothesis central to the early
book chapters that could be explored by modern researchers more adequately.
laborating Possibilities and “All Other Things Being

E
Equal” Reasoning
A striking feature of the tasks reviewed earlier is their relative complexity even
within a rather controlled experimental design. Piaget and colleagues both managed
the tasks to make them somewhat amenable to logical analysis (there are a few key
362 P. C. Dawkins
attributes or features to attend to with some underlying regularity to discover) and

introduced rather high complexity such that children could not play out all of the
possibilities. Especially in the flexibility task, children needed to be selective to test
possibilities that would provide useful information. Piaget focused much attention
on the extent to which children seem aware of and in control over the various pos-
sibilities within these spaces. As noted above, Stage II reasoners had more stable
categories of objects in their thinking about the tasks than did Stage 1 thinkers, so it
makes sense to think about Stage II children’s categorizations via multiple attri-
butes. This was represented by modeling their thinking with the set notation, such
as B = A + A′. The authors called classifying according to two properties at once
logical multiplication, which is why they use that symbol in expressions like
(B1) × (B2) × (X + X′) (where B1 and B2 refer to classes formed by two attributes in
the situation and X refers to the class formed by the outcome such as floating or
bending to touch water).
Regarding the flexibility task, Inhelder and Piaget (1958) explained, “While the
subjects at substage II-A do not use logical multiplications except under the elemen-
tary form of one-to-one correspondences, at substage II-B subjects use double-entry
tables with orders oriented serially in different directions as well as multifactor
groupings” (p. 51). The evidence provided does not suggest that children literally
wrote out such tables. Rather, the authors inferred what possibilities the child was
aware of beyond what they actually observed in the task, as well as how children
organized these possibilities conceptually. In simple cases, such as varying two
attributes (comparing two metals and two lengths), children could show clear evi-
dence of being aware of all four combinations. However, as the authors point out,
the task afforded hundreds of possibilities overall. They explained, “This complete
combinatorial system is precisely the mark of formal thought, for its structure goes
beyond additive or multiplicative groupings of classes and relations… and engen-
ders the structuring of propositional logic” (p. 55). The authors’ key conjecture here
is that the ability to reason about all possible combinations of factors (correspond-
ing roughly to a truth table in logic) is a significant achievement of Stage III
reasoners.
Piaget applied this distinction to explain a very particular difference between
Stage II and Stage III arguments. A child in Stage II may note a compensation rela-
tionship between two rods that vary in two attributes. For instance, one rod that is
shorter and thicker may touch the water the same as one that is longer and thinner.
Children in both stages will notice this, but Stage III students will seek to argue or
observe the cases that link these two varying only one attribute at a time (longer and
thicker or shorter and thinner). In other words, passing from (short and thick) to
(short and thin) to (long and thin) allows the child to observe how each attribute
affects flexibility separately. Inhelder and Piaget point out that this kind of “all other
things being equal” reasoning distinguishes Stage III thinkers from Stage II. In the
imagined double-entry table, this corresponds to moving only along a row or col-
umn, so the influence of one attribute on the flexibility of the rod can be isolated. For
this reason, this chapter is entitled “the operations of the separation of variables.”
Once again, it seems clear that no participant gave evidence of completing the
whole double-entry table for this task with hundreds of possibilities. This way of
modeling participant reasoning represents the author’s extrapolation from students’
more local ways of reasoning. Stage III reasoners showed control over the possible
rods they imagined and discussed in explanation beyond merely the rods they
observed. Furthermore, they could strategically select what attributes a rod would
need to have to help test a relationship or isolate a variable. Inhelder and Piaget
extrapolated from such lines of reasoning that the children were capable of complet-
ing such a table. In this sense, the truth table was another key logical tool they used
to model advances in children’s reasoning. This is similar to many other areas of
their work in which they took something that is the starting point of formal theory
(quantification, seriation, classification, etc.) and showed how it could be under-
stood as the product of a constructive process on the part of the learner.
he Child and Researcher’s Constructions of the Truth Table

T
of 16 Possibilities
On the basis of the combinatorial completeness described in the previous section,

Inhelder and Piaget (1958) claimed that children (of Stage III) can construct rela-
tionships that correspond to any of the possible 16-entry truth tables. This is a claim
that deserves some explication both because of its scope and ambition. Recalling
that truth value in this system really corresponds to possibilities that occur in some
systems; what this means is that children can recognize any of 16 possible relation-
ships between two variables or attributes. If we imagine the two variables or attri-
butes may each take on two values (absent/present; more/less), then that creates four
possibilities ( p.q p.q ). In the context of a task, it is assumed that children are
trying to express which set of these actually occur or are possible. In the course of
reasoning or explaining, they will then make claims that call out some combination
of these. Since each of the four may be included or excluded in such an analysis,
there are 16 ways to combine these to express various types of relationships. Inhelder
and Piaget (1958) claimed (described at length on p. 103) that children are able to
formulate any of these in reasoning. One example is “reciprocal exclusion” between
the attributes p and q, made up of p.q p.q . If a child recognizes that one or the
other attributes must be present, but not both, then they can construct and express
such a relationship.
Stated as such, this may not seem a terribly surprising claim that children can
apprehend and formulate all such relationships between two attributes. It certainly
sounds much less grandiose than claiming that children know formal logic (and it is
far short of that claim). What this claim leaves unclear is the extent to which these
understandings are really the same across contexts (or even what we would mean by
that). Would children always verbalize logically equivalent relationships in the same
way? It is unlikely. Would children judge different relationships that are logically
364 P. C. Dawkins
equivalent in this manner as being similar to one another? That likely depends upon
the context. A key point here is that these logical competencies are not taken to be
conscious on the part of the learner, nor are they taken to be embedded primarily in
their language (despite formal logic’s status as a meta-language). Inhelder and
Piaget did not focus on how children interpreted claims of a given form or how they
drew inferences from particular claims, but rather they sought to model how chil-
dren constructed relationships of necessity in situations that afford them.
The rendering of the child’s work into this logical structure is a non-trivial
accomplishment of the researcher. As I have noted before, some of the tasks the
authors used have little that would make them seem amenable to propositional
logic. The imposition of a propositional system in terms of Boolean states or even
Boolean variation of states is a non-obvious modeling step on the part of the
researchers. If other researchers do not model the task in this way, then they likely
will not identify the logical structure that Inhelder and Piaget recognized in the
child’s reasoning. Understanding the model as they used it opens new avenues for
other researchers to think about children’s reasoning. I called this model ambitious
because it claims to capture huge swaths of human reasoning if only a researcher
knows how to apply the model to a given situation. A key issue not addressed in the
book is the extent to which the findings depend upon the very particular types of
tasks provided to the students. They show some content generality to their findings
across the tasks they present, but retrospectively comparing the tasks reveals a fam-
ily resemblance, as I described before.
What then are the claims I outlined in this chapter that could be tested in future
work? Children at Stage II can organize instances into stable categories (wood/
metal, floating/not-floating) and can thus adequately realize when there is a counter-
example to a possible generalization. Their apprehension of generalizations is
largely empirical. Though Stage II children may note a relationship of compensa-
tion between two variables (shorter and thinner produces the same outcome as lon-
ger and thicker), they do not form “all other things being equal” arguments by
examining cases that vary only one attribute at a time. As such, the authors claim
they have not internally completed the table of possibilities (truth table or double-
entry table). Stage III children can construct the full double-entry table and produce
“all other things being equal arguments.” This allows them to produce more robust
explanations and justifications for their generalizations about the relationships in
each situation. Once the child has organized the attributes in the situation in this
way, they can then recognize relationships that correspond to any of the 16 possible
relationships between two variables p and q. These claims together constitute the
foundation of the claim that children construct logic by Stage III and constitute the
implicit meaning of that logic in the work of modeling children’s reasoning.
Piaget’s books do not address the significant work of the researchers in designing
the tasks and in deciding how to apply the model to children’s reasoning. This
increases the reader’s challenge in interpreting their claims and in applying their
claims in future research.
This leads us to the point of considering the possible limitations and drawbacks
of the logical model as Piaget and his colleagues employed it. First, their adaptation
of logical notation deviated from formal logic in a number of ways. This alone has
led to much misunderstanding about the nature and intent of their models. Most
notably, Piaget did not use logic to model language or children’s interpretations of
language, but rather their meanings and lines of inference. The tasks did not present
children with conditional statements, nor did the researchers rely on students to
produce statements in language consistent with logical theory or convention. This is
in stark contrast with later research paradigms in which the tasks presented students
with conditional claims such as “if it is raining, then the grass is dry” or “if a card
has a vowel on one side, then it has an even number on the other side” (Wason,
1966). Piaget consistently varied the contexts without treating children’s reasoning
about the contexts as “bias” or “error.” Rather he conducted analyses of children’s
performance on these tasks both in terms of children’s construction of specific
quantitative understandings (of weight, density, time, etc.) and children’s construc-
tion of structures that generalized across contexts (logic). To their credit, Piaget and
colleagues did not use logical syntax to model children’s reasoning if that reasoning
does not already exhibit some of the structure assumed within the syntax (i.e., Stage
I reasoners). Also, they subtly shift their use of the notation at various points in ways
that are not fully explained in the texts. Most curiously, Piaget relied on transforma-
tions of formal syntax to extrapolate various expressions and maintained that these
transformations corresponded to lines of reasoning in some strong sense. This
hypothesis is fascinating, but the opacity of the logical notation makes those claims
hard to fully explicate or know how to test empirically. This applies most strongly
to Inhelder and Piaget’s models of how children construct proportional relation-
ships, which are less clearly within the purview of propositional logic in the first
place. The claims listed above such as making “all other things being equal” argu-
ments are comparatively more straightforward to understand. Applying them still
entails many questions about the types of tasks amenable to the model and how the
model should be applied in each task (e.g., must they be physical situations with
well-defined attributes to manipulate? must the propositions in the model corre-
spond to the presence or absence of attributes or to their increase, decrease, or
invariance?). Piaget and colleagues were clearly very ambitious in trying to identify
the origins of logic in human reasoning within these tasks. I anticipate that the exact
scope of the lines of reasoning they observed and the viability of their models could
be fruitfully explored in future mathematics education investigations. To my knowl-
edge, very little research has further explored Piaget’s logical models in a spirit
compatible with his original use. Piaget’s use of syntactic extrapolations seems
somewhat at odds with his commitment to limiting his models to what children
actually construct. However, models of this type can only be productive if they
afford some kind of predictions that support falsifiable hypotheses about the actual
processes of learning.
In the final section, we shall return to one loose end that we left out in a much
earlier section. This shall allow us to consider some developments in Piaget’s much
later work on logic that further inform our understanding of his modeling.
366 P. C. Dawkins
Logic of Meanings: Assimilation as Extension
For convenience, we remind the reader of statement 7 listed above: “If I use the
hook, then I can lift the monkey out of the box by its tail” (Piaget & Garcia, 1991,
Ch 1).10 The task involved the students being seated at a table on which sat a toy
monkey with a looped tail in a box open only on top, a toy cat placed in the open,
and a toy dog in a box open on top and on the side opposite the child. The objects
were all placed out of the child’s reach. The child was provided a set of rods with
various attachments on the end, such as a hook, a bent end, and a rake head. The
child’s task was to get the toys while remaining in their seat. Why do I claim the
implication in statement 7 is different than the others discussed earlier? Statement
7’s implication is one of action (“if I do this, that will occur” or “that is possible”),
and the “meaning” of the hook and the monkey only exist in relationship with one
another. Contrast this with the mathematical implication “if an integer is a multiple
of 6, then it is a multiple of 3”; in that statement, both the antecedent and consequent
have a meaning (and an extension) apart from one another. The same can be said of
the other implications we reviewed thus far (length and flexibility are separate con-
cepts that have their own intensions and extensions), but not so with statement 7.
The relationship between the hook and the monkey primarily exists when I juxta-
pose the two objects in the context of the task. The possibility entailed in the state-
ment is not in the general extension or intension of monkeys or of hooks; indeed,
there is not any obvious relationship between the two apart from the task.
Statement 7 expresses a very different kind of relationship conceptually (even as
compared to the physical tasks used by Inhelder & Piaget, 1958), and yet Piaget
considered it an “implication” in the children’s reasoning. Piaget and Garcia (1991)
were clear about this departure, inasmuch as the book proposes to constitute a new
kind of logic that had only few formal antecedents. They called this a “logic of
meanings,” explaining,
If such a logic of meanings really exists, there is no reason why it should be limited to
propositions or statements, for any action or operation also has meanings. As no action, no
operation, and above all no meaning is isolated but is bound up with many others, there are
implications among the meanings of actions or of operations. Such implications are distinct
although inseparable from the causal aspect, or the actual execution, of actions… But with
respect to practical actions, we must distinguish their causal aspect (the outcome that is
verifiable after the fact) from their anticipation, which is inferential. An action in itself is
neither true nor false, and is evaluated only in terms of effectiveness or usefulness in rela-
tion to a goal. (p. 4)
What does it mean to consider the meaning of actions and how does this depart from
what is commonly expressible in logic? First, this view assumes that there are goals
and intentions involved in action that can imbue objects with meaning (e.g., to take a
toy monkey out of a box). Second, meaning can be bound up in a context, such as the
task the researcher poses to the student. The extension of the implication is in this
sense quite local and provisional. This stands again in stark contrast to the assumption
10
As above, this is not what a child said, but rather the rendering of the child’s inference into the
language of logic.
that logic is what is shared when we vary semantic content. Third, the authors point
out that this logic need have no direct relationship with “propositions or statements.”
In contrast to formal logic that operates on a meta-language that is partitioned off
from actual states of affairs by truth values, Piaget seeks to explicate the implications
entailed in actual processes of activity and reasoning. In particular, the anticipation
that an effect will result from an action (“if I do this, then that will occur”) is an impli-
cation for the reasoner that, whether or not it is true to fact, tells the observing
researcher about the child’s meanings by which they draw such inferences.
There is a more fundamental point to be made here about how the link in mean-
ing in statement 7 is distinct from the other cases we considered. Piaget infers that
the child has made such an inference when the child uses the hook to pick up the
monkey. This meaning is entailed neither by the monkey by itself nor the hook, but
by the combination of the two in the goal-oriented context of the task. Logical infer-
ence is often understood to have a definite direction (p ⊃ q), or it is assumed that an
inference (as a mental action) takes place in time (“if I know this, then I can con-
clude that”). Piaget challenges this notion by positing a reflexivity between an
action and a meaning. To pick something up with a hook is to give it the meaning of
being “hookable.” To roll something is to give it the meaning of being “rollable.” As
Piaget explained, “an object x might be laid on a support y which the subject could
use to draw it back… the relation or action ‘laying down upon’ has acquired the
meaning of a ‘reason.’” (Piaget & Garcia, 1991, p. 5).
Does the property come from the action or does the action recognize the object
as having a property? There is an extent to which this question misses the point of
Piaget’s logic of meanings. One way to think of this is to consider how it turns the
question of extension on its head. We learn from the child the extension of an infer-
ence by simply observing what actions they carry out (physical or mental). The
researcher cannot take their statement of the inference (the researcher’s model of the
child’s inference) and necessarily extend it to what they see as all the relevant
instances. Rather, the researcher must observe the child to see how far the inference
extends based on how they act in the task. This is not to say the researcher should
not investigate whether the child assimilates independently or can accommodate on
request other instances (such as the dog or the cat) to an inference; this is a natural
next step. The key point is to recognize the inference as being the child’s and to
recognize how there is a “logic” of every action. This is part of what it meant for
Piaget to make logic the mirror of thought.
The logic of meanings represented a further expansion of the scope of Piaget’s
model. This would capture a much broader swath of human reasoning, which is
reflected in Piaget and Garcia (1991) by the much broader range of tasks. Children
solve puzzles or try to form tessellations with tiles. Piaget’s section of that text includes
comparatively few symbolic expressions. Pat Thompson (Chap. 5) has picked up this
broader line of investigation by describing someone’s meanings in terms of the impli-
cations they draw from reasoning about something in a particular way (see also
Thompson, 2013). We not only infer about an object based on what we think it means,
but our inferences about something also constitute the ways we make meaning of it.
This is an adaptation of Piaget’s insight that understanding comes from activity that is
especially useful for studying older children and college students (as I do).
368 P. C. Dawkins
Concluding Lessons
What can we learn from analyzing Piaget’s use of logic to study children’s reason-
ing? One of the key features of Piaget’s investigations is to treat things that adults
take for granted (possibly as innate or simply true of the world) as something chil-
dren must construct in activity. As such, there is always a danger in trying to apply
formal models (such as propositional logic) to describe children’s reasoning since
that model may assume much that is not shared by the child. I think one of the most
useful things to note about Inhelder and Piaget’s use of the propositional logical
model is how they avoided using it to discuss the reasoning of Stage I and most
Stage II subjects. Indeed, I have found one of the most fruitful lines of inquiry is to
understand when and why a formal model does not apply to a subject’s reasoning so
I can identify (1) what the model takes for granted that children may not and (2) how
to describe what in the child’s reasoning is incompatible with the model so I can
attend more closely to what is relevant and meaningful to the child.
However, like Piaget, I think we should not dispense with formal models since
they often provide us with ambitious goals for children’s learning. Piaget’s use of
propositional logic matched his intent to study how children construct stable catego-
ries, content general modes of reasoning, and the ability to draw and justify gener-
alizations. For instance, what does it mean for children to (in the sense described
above) construct the complete truth table of certain implications in a situation, and
what lessons can we learn about their cognition by investigating this constructive
process? Piaget clearly does not think that children’s reasoning at younger stages
exhibits such structure, but can this model be coherently applied to certain aspects
of adolescents’ reasoning? Piaget’s studies suggest it can, though I have tried to
demonstrate how his application and explanation were subtle and fraught with pos-
sible misunderstanding.11 As I argued above, Piaget and colleagues adopted fasci-
nating and ambitious research goals, though the scope and robustness of Piaget’s
findings are a matter for further investigation. I think misunderstandings that arose
from Piaget’s use of the model unfortunately hindered many efforts to pursue that
task. Also, Piaget and his colleagues did not always expound on their task design
and modeling activity and how it interacted with the kinds of understanding they
hoped to investigate. To alleviate such issues, future investigations of logic in rea-
soning should work to clarify:
1. The meaning of logic being operationalized in the study
2. The task design in light of this meaning of logic (how this both supports and
delimits the nature of the findings)
11
Indeed, many later scholars have shown how logic models can be misapplied to studies of human
reasoning.
3. The standards of evidence for claiming any understanding generalizes across

semantic content (i.e., is in some sense “logical”)
4. Whether and how the reasoner is aware of any sameness between situations that
corresponds to the researcher’s sense of logical analogy
It also bears noting that what mattered most to Piaget and his colleagues was not the
stages in child reasoning, but rather the means by which children advanced between
stages, meaning the processes of abstraction (see Chaps. 6 and 8). It is my sincere
hope that reflections on those researchers’ processes of investigation, modeling, and
learning about child learning will enhance our abilities to use formal models insofar
as they are useful and to remake those models when necessary to do justice to the
experiences of research subjects from whom we learn.
References
Beth, E. W., & Piaget, J. (1966). Mathematical epistemology and psychology. D. Reidel Publishing
Company.
Burman, J. T. (2016). Piaget’s neo-Gödelian turn: Between biology and logic, origins of the new
theory. Theory & Psychology, 26(6), 751–772.
Evans, J. S. B. T. (2002). Logic and human reasoning: An assessment of the deduction paradigm.
Psychological Bulletin, 128(6), 978–996.
Evans, J., & Feeny, A. (2004). The role of prior belief in reasoning. In J. P. Leighton & R. J. Sternberg
(Eds.), The nature of reasoning (pp. 78–102). Cambridge University Press.
Evans, J., & Over, D. E. (2004). If. Oxford University Press.
Inglis, M. (2006). Dual processes in mathematics: Reasoning about conditionals (Doctoral dis-
sertation). University of Warwick.
essay on the construction of formal operational structures. Basic Books.
Inhelder, B., & Piaget, J. (1964). The early growth of logic in the child. W. W. Norton.
Markovits, H. (2004). The development of deductive reasoning. In J. P. Leighton & R. J. Sternberg
(Eds.), The nature of reasoning (pp. 313–338). Cambridge University Press.
Piaget, J. (1950). The psychology of intelligence. Routledge Classics.
Piaget, J. (1953). Logic and psychology. Manchester University Press.
Piaget, J., & Garcia, R. (1991). Toward a logic of meanings. Lawrence Erlbaum Associates.
Quine, W. V. O. (1960). Letter to Beth, regarding Quine’s reluctance to join the advisory board
of Piaget’s International Center for Genetic Epistemology W. V. Quine Papers (MS Am 2587
[852], Box 31). Houghton Library, Harvard University, Cambridge, MA.
Schroyens, W. (2010). Logic and/in psychology: The paradoxes of material implication and psy-
chologism in the cognitive science of human reasoning. In M. Oaksford & N. Chater (Eds.),
Cognition and conditionals: Probability and logic in human thinking (pp. 69–84). Oxford
University Press.
Stenning, K., & van Lambalgen, M. (2008). Human reasoning and cognitive science. The
MIT Press.
Thompson, P. (2013). In the absence of meaning. In K. Leatham (Ed.), Vital directions for research
in mathematics education. Springer.
Wason, P. C. (1966). Reasoning. In B. M. Foss (Ed.), New horizons in psychology 1. Pelican.
Chapter 11
Students’ Units Coordinations
Amy J. Hackenberg and Serife Sevinc
Introduction
What Is Units Coordination?
Units coordination describes how people organize units to solve problems. People
construct units by using their unitizing operation, a mental action of creating singu-
larities out of their experiences (Steffe et al., 1983a; von Glasersfeld, 1981). Children
begin by compounding sensory experiences, such as sounds, rhythms, and visual
and tactile experiences of objects, into experiential units. For example, hearing bells
ringing can be an impetus to take a single “bong” as a unit. Later, people construct
units from what they imagine, eventually producing abstract units that are not tied
to particular sensory–motor experiences (Steffe et al., 1983b; von Glasersfeld,
1981). Counting abstract units involves the construction of arithmetical units, what
we think of as a discrete 1 (Steffe & Cobb, 1988). Similarly, observing wires strung
between poles as one drives in a car can lead children to constructing experiential
segments of length. That construction can be the basis for abstracting length units
from imagined experience. In turn, abstract length units are the basis for the con-
struction of measurement units for length.
A. J. Hackenberg (*)
e-mail: ahackenb@iu.edu
S. Sevinc
Department of Mathematics and Science Education, Middle East Technical University,
Ankara, Turkiye
e-mail: sserife@metu.edu.tr
https://doi.org/10.1007/978-3-031-47386-9_11
372 A. J. Hackenberg and S. Sevinc
Units coordination arose when a research team was trying to map the ways chil-
dren construct operations (mental actions) and schemes1 underlying whole number
multiplication and division (Steffe, 1992). At that point, Steffe and colleagues had
already created a web of constructs via constructivist teaching experiments that
articulated models of children’s early mathematical thinking and learning (Steffe &
Cobb, 1988; Steffe et al., 1983a). In 1985, the team began a 3-year teaching experi-
ment with six 8-year-old children, where children were taught for one year. At that
time, explaining children’s construction of whole number multiplication and divi-
sion was uncharted territory. As Steffe (1992) says, “There was no categorization
schema that was known to me, nor was there a language that I could use to express
the children’s schemes” (p. 262). Piaget (1965/1949) studied children’s conceptions
of multiplication and identified multiplication as a one-for-two and then one-for-
many coordination. Although that formulation may have been an influence on
Steffe, it was not a guide because Steffe sought to build on the findings of Steffe’s
research team to that point.
Indeed, one of the main goals of the 1985 teaching experiment was to investigate
what role children’s number sequences2 played in children’s construction of multi-
plication and division (Steffe, 1992). Steffe did not assume that the number
sequences would play a role—he purposefully selected pairs of students operating
with three different number sequences: the initial number sequence (INS), the tac-
itly nested number sequence (TNS), and the explicitly nested number sequence
(ENS). Then, he let the children teach him how they assimilated situations that were
multiplicative from his perspective. In other words, he adapted to the children as
they interpreted and solved their problems, working to understand his mathematical
interactions with them and what mental activity could underlie these interactions.
From this experiment, Steffe’s (1992) analysis of Maya, a TNS student, led to his
formulation of units coordination as follows:
For a situation to be established as multiplicative, it is always necessary at least to coordi-
nate two composite units in such a way that one of the composite units is distributed over
the elements of the other composite unit. (p. 264)
Here, a composite unit is the result of taking a collection of discrete items as one
unit. For example, from our perspective, three flowers in a row are both one unit and
three individual items. If there are seven rows with three flowers in each row, finding
the number of flowers in all of the rows involves distributing three 1 s across each
of the seven rows. The INS student featured in Steffe’s study (1992), Zachary,
engaged in this distribution experientially. One example of this experiential distri-
bution was that Zachary tried to track threes but did not know where to stop (p. 265).
1
A scheme is a goal-directed assemblage of operations that includes an assimilated situation, activ-
ity, and a result (Steffe, 2010b; von Glasersfeld, 1995). Please see Steffe (Chap. 2, this volume) and
Tillema and Gatza (Chap. 3, this volume) for more detailed discussion.
2
Number sequences are, essentially, schemes that indicate how children conceive of counting situ-
ations, the activity children engage in to achieve goals in those situations, and how the children
conceive of the results of this activity in relation to their goals. It is outside the scope of this chapter
to provide a review of these sequences, but a concise summary can be found in Ulrich (2015,
2016a); they are also addressed by Steffe in the introduction to this volume.
11 Students’ Units Coordinations 373
As Steffe says, “When counting, he was consumed with organizing his counting
acts into trios and became unaware of the numerosity of the rows” (p. 268). So,
Steffe concluded that Zachary had not constructed composite units of three.
Although he learned to use his fingers to keep track of threes during the teaching
experiment, he did not construct units coordination because he always had to create
the experience of a composite unit in order to count it.
In contrast, TNS student Maya “coordinated units prior to executing the activity
of her units-coordinating scheme the first time [Steffe] worked with her” (Steffe,
1992, p. 279). He posed this problem to her: Here is a red rectangle; six small blue
rectangles fit on it exactly, and two orange triangles fit exactly on a blue rectangle.
How many orange triangles fit on the red rectangle? Maya subvocally said number
words and then said, “Twelve” (p. 279). In explanation, she tapped the table twice
with each of her six fingers, saying, “1, 2; 3, 4; 5, 6; 7, 8; 9, 10; 11, 12” (p. 279). To
solve the problem in this way, Steffe claimed that Maya had to have distributed “her
unit of two across the six units of one prior to counting” (p. 280). In other words,
Maya inserted a composite unit of two into each of the units of her composite unit
of six and further used her composite unit of six to keep track of counting six twos.
Steffe inferred she had to have made this coordination prior to counting in order to
count the way she did. Maya did not have to assemble composite units experien-
tially as Zachary did. As Steffe notes, “It is what goes on before counting that con-
stitutes a true units coordination” (p. 300). This finding, that Maya coordinated units
prior to counting, was the most important one of the teaching experiment because it
was “the most primitive concept of multiplication” (p. 304) that Steffe could envi-
sion at that time. In this chapter, we will articulate features of this concept of multi-
plication, as well as more advanced ones because these ideas about students’
multiplicative concepts have held up over time.
In fact, units coordination operations have become a way of understanding chil-
dren’s thinking that has explanatory power far beyond whole number multiplication
and division. In other words, how students (of all ages) construct and coordinate
units turns out to influence many aspects of students’ mathematical thinking, includ-
ing students’ fractions knowledge (e.g., Boyce & Norton, 2016; Hackenberg &
Tillema, 2009; Norton & Wilkins, 2012; Steffe & Olive, 2010); rational number
knowledge (e.g., Hackenberg & Sevinc, 2022; Olive, 1999); understanding of inte-
gers (e.g., Ulrich, 2012); algebraic reasoning (e.g., Hackenberg, 2013; Hackenberg
& Lee, 2015; Hackenberg et al., 2021a; Tillema & Burch, 2022); proportional rea-
soning (e.g., Aydeniz, 2018; Shin et al., 2020); and combinatorial reasoning (e.g.,
Tillema, 2013, 2014, 2018, 2020).
In this chapter, we define and characterize units coordination, and associated
constructs, as a framework for modeling students’ mathematical thinking. We give
examples of its use in research and discuss why units coordination is useful for mak-
ing second-order models of students’ mathematics.3 Then, we describe standards of
3
Second-order models are constellations of constructs to describe and account for the mathemati-
cal thinking of another person (Steffe et al., 1983b; Ulrich et al., 2014).
evidence for making claims about students’ units coordinations.4 Finally, we discuss
what is known about the numbers of students at different units coordination stages
in elementary and middle school and how students advance across stages.
Definitions and Characterizations, with Examples
What does it mean to distribute one composite unit over the elements of another
composite unit? It means different things depending on the nature of one’s units
(Ulrich, 2015, 2016a). In this section of the chapter, we explain what units coordina-
tion looks like for students at three different stages. Units coordination stages are
based on students’ number sequences. We start with students operating with the
TNS because, as we noted above, that number sequence affords students the con-
ceptual tools to make the most basic units coordination (Steffe, 1992).
First, we make some points about terminology. In this chapter, we use the phrase
“operating at stage 1 of units coordination” to refer to TNS students, abbreviating
that to “students at stage 1” where useful. (In other work, we have used “first multi-
plicative concept” to refer to the same group of students; see Hackenberg & Tillema,
2009). Students operating at stage 2 of units coordination have constructed the ENS,
and students operating at stage 3 of units coordination have constructed the GNS.
We chose to use the term “stage” because a “stage designates a stretch of time
that is characterized by a qualitative change that differentiates it from adjacent peri-
ods and constitutes one step in a progression” (von Glasersfeld & Kelley, 1982,
p. 156). The use of “stage” is appropriate because students operating with different
number sequences have qualitatively different assimilatory operations open to them.
So, there are similarities in how students at a particular stage operate to solve prob-
lems. The stages are epistemic students (Beth & Piaget, 1966)5 in the sense that a
units coordination stage involves specific mental operations; the units at each stage
have particular features, and specific schemes are likely to be constructed with these
operations and units. Another reason that stage is an appropriate term is that students
construct the number sequences in a progression because each subsequent number
sequence includes more complex ways of operating that subsume the ways of oper-
ating in the prior number sequence (Ulrich, 2015, 2016a). That said, the term “stage”
can be off-putting because it may connote slotting students into rigid categories.6
4
In the paper, we will use “students” because most of the research involving units coordination has
been done with students from preschool to eighth grade, although some recent research has been
done with college students (e.g., Byerley et al., 2019). For the use of units coordination in research
about practicing teachers, see Izsák et al. (2012).
5
See the Introduction of this volume for an explanation of epistemic students.
6
In fact, we are not using “stage” to denote structural stages like Piaget’s concrete and formal
operations. We are using stage in the sense that Campbell and Bickhard (1986) do, to denote pat-
terns in people’s ways of knowing that are separated by reflective abstractions. Please see the
Chap. 6, in this volume by Ellis et al.
Fig. 11.1 Vacation Problem
So, we hasten to add that students do not remain at a particular number sequence
or units coordination stage for the same period of time. For example, some students
spend only a couple of months operating with the TNS (e.g., Steffe, 1994), while
other students spend more than one year there (e.g., Steffe & Cobb, 1988; Ulrich,
2016b). In addition, there is variation among students operating at a particular stage
(e.g., Hackenberg & Sevinc, 2021, 2022; Shin et al., 2020). Yet, there are greater
contrasts among students across stages (e.g., Hackenberg & Tillema, 2009;
Steffe, 1992).
To move from one stage to the next requires a major reorganization of mental
activity referred to as interiorization, which involves reflective abstraction (Steffe,
2010b; cf. Ellis et al., Chap. 6, this volume). When students interiorize their units
coordination, students are able to read them into a situation prior to solving a prob-
lem. Because of the major reorganization involved in interiorizing units coordina-
tions, students may remain operating at a particular stage for multiple years in
elementary or middle school. Since units coordinations influence students’ thinking
in so many different mathematical domains, this phenomenon makes understanding
students’ units coordinating activity all the more pressing.
We now provide examples of students at different unit coordination stages. To do
so, we consider the Vacation Problem (Fig. 11.1), which from our perspective,
involves the coordination of two composite units, 7 days and 5 weeks.
As mentioned, for us, a unit is the result of using the unitizing operation, which
involves “abstracting out the ‘one’-ness from some aspect of experience” (Ulrich,
2015, p. 3). A composite unit is a unit of units, and it is also the result of using the
unitizing operation. A person who has constructed a composite unit thinks of seven
days as both one entity and as seven individual items.
To solve the Vacation Problem, students at different units coordination stages
coordinate days and weeks in different ways, and the result of the coordination has
different meanings. We describe how students at different stages typically respond
to this problem.
Students at Stage 1
Students at stage 1 can keep track of both the number of 7 s and the number of 1 s
as they solve the Vacation Problem. They typically count on by ones after known
skip-counting patterns. For example, older students might know the first few 7 s,
whereas younger students may start counting on after 7 as follows: “7, that’s one
[seven]; 8, 9, 10, 11, 12, 13, 14, that’s two [sevens]; …., 29, 30, 31, 32, 33, 34, 35,
that’s five [sevens].” Students are likely to put up fingers to keep track of the number
of sevens counted.
The main reason students at stage 1 can solve the problem this way is that their
number sequence, the TNS, is recursive (Steffe, 1992). This means that these stu-
dents can repeatedly count on to solve problems. For example, Maya demonstrated
this feature of the TNS when she solved a problem of how many marbles were in
two cups, one containing 8 and one containing 7. She counted 7 onto 8 to get 15.
Steffe asked her to put 3 more marbles in the cup containing 7 and asked how many
in that cup. She counted 3 onto 7 to get 10. Then, he asked how many in both cups,
and she counted 3 onto 15 to get 18. To arrive at 18 in this way, she had to take the
3 that she had counted onto 7 to get 10 as counted items that were also countable in
determining how many marbles were in both cups. “This dual function of counted
and countable items indicates that she could take her number sequence as its own
material of operating and, consequently, that it was recursive” (Steffe, 1992, p. 278).
In other words, segments of TNS students’ number sequences are “re-usable” in the
sense that the students can repeatedly count on by those segments. Behaviorally,
this aspect of TNS students’ thinking “carries the force of double counting” (per-
sonal communication, Les Steffe, 2/16/23). Maya did not explicitly double-count,
but she tracked the three items repeatedly. TNS students can use this feature of their
number sequence in repeatedly counting on 7 s in solving the Vacation Problem.
The result of the units coordination in the Vacation Problem for students at stage
1 is a composite unit of 35 singleton units: They take 35 as one entity. However, “the
history of double counting dissipates experientially” (personal communication, Les
Steffe, 2/16/23). What this means is that these students assembled their units coor-
dination with some figurative material—in our example, it was the student’s lived
experience of weeks and days and their relationship. They anticipated inserting a
unit of 7 into each unit of a week, and then, they were able to complete the units
coordination using the recursive property of their number sequence. As they used
these operations, the experience of behaviorally making the units coordination dis-
sipated and that means the student took the result as a unit of units. These units are
equivalent but not identical (Steffe, 2010b; Ulrich, 2015). This means, from the
point of view of the students, that the days are all equal but separate. However,
because they are not identical units, the student feels a need to have 35 singleton
ones to represent 35. If the students saw the units as identical, then the students
would not need 35 ones to represent 35—they would be able to think of 35 as 1
repeated 35 times.
In Table 11.1, we have used figures to depict these ideas. First, students at stage
1 keep track of the number of 7 s counted by using their composite unit of 5 to con-
trol counting by five 7 s, shown as the circled numbers in the first stage 1 figure.
Students reach 35 ones by counting on by ones; older students may skip-count by
known patterns before counting on by ones. The result is 35 singleton ones, where
each one is equivalent to other ones: Each one is shown by the same size unit square.
However, the ones are not identical: Students need to have 35 squares to represent
35. This aspect of number for these students will be clearer via the following con-
trast with students at stage 2.
Table 11.1 Units coordination activity by stage to solve the Vacation Problem
Stage and associated number
sequence Image of units coordinating activity Result of the units coordination activity
Stage 1 A composite unit of 35 ones, where the ones
TNS are equivalent but not identical
Stage 2 A composite unit of 35 ones, where 35 is

11 Students’ Units Coordinations
ENS made from 1 iterated 35 times

five 7 s implicitly, or in activity
(continued)
377
378
Stage and associated number

sequence Image of units coordinating activity Result of the units coordination activity
Stage 3 A composite unit consisting of composite
GNS units:
five 7 s can be rearranged to seven 5 s
35 is made from 7 iterated 5 times or 5
iterated 7 times
A. J. Hackenberg and S. Sevinc
Students at Stage 2
Students at stage 2 solve the Vacation Problem by keeping track of both the number
of 7 s and number of 1 s, so like students at stage 1, they use their composite unit of
5 to control accumulating five 7 s. However, in contrast with students at stage 1,
students at stage 2 can learn to reason strategically to accumulate 7 s, rather than
count by 1 s. For example, to add 7 onto 7, they may think of the second 7 as 3 and
4, adding 3 onto the first 7 to make 10, and then adding the remaining 4. So, they
transform 7 + 7 into 10 + 4, and they can continue to add on 7 s strategically in
this way.7
Students at stage 2 can operate in this way because they have constructed disem-
bedding operations (Steffe, 2010b; Ulrich, 2015). A disembedding operation
involves being able to lift a segment of one’s number sequence out of that sequence
while keeping the whole sequence intact. Students can operate on the segment and
then return it to the sequence. We show this in the first figure in Table 11.1 for stu-
dents at stage 2. In this example, students can take the second composite unit of 7
out of their sequence that consists of 7 and 7 more, break the second 7 into 3 and 4,
and then replace it into the sequence of 7 and 7 more. They know that they have
transformed the 7 and 7 into 7 and 3 and 4 and that together this makes up the result
of two 7 s. And, students at stage 2 can learn to repeatedly engage in this reasoning.
The result of the units coordination in the Vacation Problem for students at stage
2 is a composite unit of 35 singleton units. Like students at stage 1, students at stage
2 take 35 as one entity. In contrast with students at stage 1, these singleton units are
equivalent and identical. This means, from the point of view of the students, that 35
can be made from iterating a unit of 1 35 times (Steffe, 2010b). Steffe (2010b) uses
the phrase iterable units of 1 to describe this construction. For a student who has
constructed an iterable unit of 1, “a number word refers to a composite unit contain-
ing a unit that can be iterated the number of times indicated by the number word”
(Steffe, 2010b, pp. 41–42). The students anticipate this operation of iteration, and
they also anticipate the result of using iteration on the unit of 1, the unit of units
(Steffe, 2009a). This construction means that students at stage 2 have constructed a
multiplicative relationship between 1 and other numbers in their number sequence:
35 is 1 times 35 (Steffe, 2010c, p. 118). In Table 11.1, we have depicted this idea by
the figure that shows units tiled on top of each other. We note that this is not a com-
pletely accurate portrayal of an iterable unit of 1, but it does indicate some of the
compression that an iterable unit of 1 provides: Conceiving of large numbers is
easier because students do not feel a need to have all of the ones to make them;
students just need to repeat 1 the appropriate number of times.
We point out one more contrast with students at stage 1 that is a result of the
construction of iterable units of 1: As students at stage 2 solve the Vacation Problem,
they create the 35 as a unit of five units each implying 7 units, an implicit
7
We are not suggesting that students at stage 2 will automatically demonstrate a solution such as
this one; we are suggesting that students are capable of this type of solution with good instructional
support.
three-levels-of-units structure, and they may have awareness of this structure in

their problem-solving activity. They can create this structure because students’ iter-
able units of 1 can imply composite units (Steffe, 2010b).8 Students at stage 2 can
think of each unit of their composite units of 5 as an iterable unit that implies a
composite unit of 7. To imply a composite unit means that students create a “mini-
mal image” (Steffe, 2009a, p. 12) of each week consisting of a string of 7 days as
they enumerate the five weeks. We have shown this idea with the last figure for
students at stage 2 in Table 11.1. Because of the implicit nature of this coordination,
students at stage 2 do not maintain it as they operate further with 35: For them, 35
becomes a composite unit of singleton, iterable units. This aspect of number for
these students will be clearer via the following contrast with students at stage 3.
Students at Stage 3
As students at stage 3 solve the Vacation Problem, they can construct an explicitly
embedded structure of composite units (Steffe, 2009a): 35 is a unit of 5 units, each
of which contains 7 units. Furthermore, these students can take this structure, a
composite unit containing a sequence of composite units of equal numerosity, as
given prior to operating.9 This explicit structure makes it possible for students at
stage 3 to operate on all five 7 s at once in strategic ways. For example, students at
stage 3 can think of each 7 as 5 and 2 and then operate on the five 5 s and five 2 s.
So, they see five 7 s as five 5 s, 25, and five 2 s, 10 more.
Another way to speak about this explicit structure is that students at stage 3 have
constructed iterable composite units (Steffe, 2010b). This construction means that
students at stage 3 anticipate that a number like 35 can be made from iterating a
composite unit some number of times. So, when students determine a result like five
7 s is 35, the number 35 becomes structured as 7 iterated five times, prior to problem-
solving. The students anticipate the iteration of the composite unit and the result of
the iteration, a unit of composite units of equal numerosity. In short, students at
stage 3 have constructed a multiplicative relationship between a number and com-
posite units that can be iterated to fill out that number (Steffe, 2010b).
The construction of iterable composite units provides great flexibility in solving
problems that are multiplicative from the perspective of an observer (Steffe, 2009a).
We give one example here. Because students at stage 3 anticipate numbers as struc-
tured into three levels of units, they can hold these three-levels-of-units structures
out for reflection and reorganize them into other useful structures. Consider this
extension of the Vacation Problem (Fig. 11.2).
8
Students possess an iterable unit of 1 in the sense that they can use them to structure situations,
just as they would possess particular mental operations. Possession means students can call on the
operation in assimilation.
9
“What it means to take such a structure as a given is that the operations that produce the structure
are evoked, but they are not enacted” (Steffe, 2009a, p. 13).
Fig. 11.2 Extension of Vacation Problem
Students at stage 3 can conceive of the five 7 s (five weeks with 7 days in each)
as 35 and then segment the 35 into a number of 5-day school weeks, arriving at
seven. And, when students at stage 3 arrive at the 35 structured into seven 5 s, they
still maintain the 35 as five 7 s: They can switch back to that view if needed. In
Table 11.1, the first figure of stage 3 units coordination shows 5 units of seven rep-
resenting the 5-week vacation, which resulted in 35 days, the second figure. The
third figure shows the reorganization of these days into seven units of 5. In the final
figure, we use tiled composite units to indicate iterable composite units.
How Have Researchers Used Units Coordination in Research?
In this section, we give examples of how units coordination has been used in
research and why units coordination is useful for researchers to include in second-
order models of students’ mathematics. Because it has been used most extensively
as a basis for understanding students’ fractions knowledge, we begin there and then
move to students’ algebraic reasoning.
Units Coordination and Fractions Knowledge: An Overview
Table 11.2 gives an overview of the ways of operating with fractions that are in the
zone of potential construction (ZPC; Steffe & D’Ambrosio, 1995) of students at
each stage of units coordination.10 We discuss what is known about each epistemic
students’ fractions knowledge, referring to Table 11.2 as we move along.
Units Coordination and Fraction Knowledge: Examples

Stage 1
Because students at stage 1 have constructed composite units, they can fragment a
continuous whole into equal parts (Steffe, 2010a). However, they have not con-
structed a disembedding operation. So, they do not understand that one of those
10
The zone of potential construction is a time-sensitive concept describing the schemes and opera-
tions that a student may construct based on the student’s current schemes, operations, and concepts
(Steffe & D’Ambrosio, 1995).
Table 11.2 Fractions schemes in relation to key features of units coordination stages
Fraction
Number Units possible Multiplicative schemes in
Stage sequence to construct Operations operations in ZPC ZPC
1 TNS Composite Unitizing, iterating, Tracking the Parts
units partitioning coordination of two within
levels of units in wholes
activity fraction
scheme
2 ENS Composite Unitizing, iterating, Operating on Part-whole
units where partitioning, disembedded fraction
units of 1 are disembedding, composite units scheme,
iterable splitting (possible to additively to find the Partitive
construct) result of a units unit
coordination fraction
scheme,
Partitive
fraction
scheme
3 GNS Iterable Unitizing, iterating, Switching between Iterative
composite partitioning, three-levels-of-units fraction
units disembedding, structures scheme
Iterable unit splitting, distributing
fractions
parts can be used to remake the whole through, for example, iteration. This is such
a major constraint on building fractions knowledge that Steffe and Olive chose not
to work on the construction of fraction schemes with children in the third grade who
entered their teaching experiment operating at stage 1 (Steffe, 2009b). Instead, they
worked with students at stage 1 on situations that involved composite units from the
researchers’ perspective. This choice means there has been less written about the
fractions knowledge of students operating at stage 1, compared to the fractions
knowledge of students operating at stages 2 and 3 (Steffe & Olive, 2010). Therefore,
in this subsection, we review the fractions knowledge of the stage 1 middle school
students in Hackenberg’s (2013) study. Although students at stage 1 have not con-
structed disembedding operations, they can and do construct partitioning operations
(Hackenberg, 2013; Steffe, 2010a; Ulrich, 2016b). They also can construct an iterat-
ing operation with lengths.
Partitioning Steffe (2010a) explained that students at stage 1 can project a com-
posite unit of their number sequence into a length to fragment the length into con-
tinuous units. For example, when asked to share a rectangular pizza (bar) equally
with three people, students at stage 1 can use their composite unit of three to project
three equal parts into the bar. In many cases, students experiment to make the
equal parts.
Fig. 11.3 Courtney equally sharing a bar with five people (Hackenberg, 2013, p. 548)
Students at stage 1 have a sense of simultaneous awareness of the parts and the
entire length (Steffe, 2010a, p. 69). So, students at stage 1 will usually work to
accomplish the goal of making equal parts and the goal of trying to exhaust the
whole. In some cases, students at stage 1 may not accomplish both goals. For exam-
ple, seventh-grade student at stage 1 Courtney, from Hackenberg’s (2013) study,
shared a bar equally among five people, but her parts were not equal (Fig. 11.3,
upper bar).
When asked to draw it again more accurately, Courtney asked, “What do you
mean?” (p. 548). The researcher colored two nonequal parts and asked who got
more. Courtney identified that the amounts were not equal. She made two further
attempts to make the five parts more equal, ultimately cutting an extra part off the
bar (Fig. 11.3, bottom bar). So, in her attempt to make equal parts, she ended up not
exhausting the whole.
Iterating All six students at stage 1 in Hackenberg’s (2013) study demonstrated
iterating operations. However, three students did not iterate lengths until engaged in
problems specially designed to support that activity. For example, seventh-grade
student Laura was posed the cut-out bar problem (Fig. 11.4). Thin lines in the pieces
show the way the pieces were to be oriented.
Laura picked out the appropriate part right away. The interviewer asked her to
check her choice by testing it. Laura placed all four different-sized parts on the
given bar When asked whether that tested it, Laura said she guessed so. The inter-
viewer asked her to use her chosen part on the bar to see whether it was a fair and
equal share. Laura then moved her chosen part along the bar. Initially, she found that
five fit and then revised that to six, naming the part “one-sixth” (p. 548). Thus, in
solving this problem, Laura demonstrated an iterating operation with lengths, and
she connected the use of that operation to fractions language for the part.
Lack of a Disembedding Operation As discussed earlier, when adding 7 and 7, a
student who has constructed a disembedding operation can take a 7 out of this
sequence to operate on it while keeping the sequence of 7 and 7 more intact. With
regard to lengths, a student who has constructed a disembedding operation can take
a part out of a whole and keep the whole mentally intact (Olive & Steffe, 2010b).
Since students at stage 1 have not constructed disembedding operations (Ulrich,
2016a; Steffe, 2010b), they construct fractions as parts within wholes (Hackenberg,
2013). For these students, fractions mean the number of parts shaded in relation to
Here is a candy bar and several separate pieces that were made from identical candy bars. (The interviewer places the
pieces on the table, not in order of size.) The candy bar is to be shared fairly among six people. Which one of these pieces
would be one share? How do you know it’s going to be fair?
Fig. 11.4 Cut-out bar problem (Hackenberg, 2013, p. 561).
the number of parts shown in the entire unit—we will refer to this as the unit bar.
However, if asked to draw out a part separately, these students will not mentally
reunite that part with the parts in the unit bar. For example, in Hackenberg’s (2013)
study, Courtney was asked to share a bar fairly among four people. She made four
fairly equal parts. When asked to draw out the share for one person as a separate
piece, she did so. When asked for a fraction name for that part, she said it was one-
third because “you’ve got three pieces left.” The interviewer probed whether the
part could be called one-fourth, and Courtney was unsure. So, Courtney appeared to
mentally cut off one of the four parts but did not disembed it; when she mentally
separated it from the whole, she did not reunite it back into the whole in order to see
it as one of four equal parts.
Stage 2
Disembedding The construction of a disembedding operation at stage 2 is, in some

sense, a great relief for students building fractions knowledge and for researchers
making models of that knowledge! It opens the way for students to conceive of frac-
tions as parts out of wholes, which means students can view a fractional part as both
embedded in a larger unit and as a part “in its own right.” As Steffe (2010a) writes,
the student at stage 2.
can unite any subcollection of elements together and disembed them from the composite
units and constitute it as a composite units in its own right without destroying these ele-
ments in the containing composite units formed by partitioning. This establishes the classi-
cal numerical part-to-whole operation that serves as a fundamental operation in the
construction of fraction schemes (p. 74).
In other words, students at stage 2 have constructed the foundation for viewing a
collection of parts of a quantity as both a separate quantity and as part of that
quantity.
The Equi-Partitioning and Partitive Fraction Schemes When students at stage
2 go to partition a length, they project their composite unit with iterable units of 1
into the length. For example, if a student at stage 2 partitions a length into five equal
parts, she “understands that any one of the units can be used to reconstitute a con-
nected but segmented unit equivalent to the original unit” (Steffe, 2010a, p. 69). In
other words, one part of the five equal parts of the length can be iterated to make a
5-part bar that, if the parts are equal, should be the same length as the original
length. Here, the length units inherit the iterability of the discrete units of 1.11
Steffe (2010a) calls this construction an equi-partitioning scheme: To partition a
length equally into five parts, the student partitions the length into parts, disembeds
a part from the length, and iterates it in a test to see if the part is an equal share. The
student uses partitioning, iterating, and disembedding together—in fact, partition-
ing and iterating are part of the same psychological structure for the students
(Steffe, 2010c).
Equi-partitioning is the basis for the partitive fraction scheme, which Steffe calls
the first genuine fraction scheme (Steffe, 2010c, p. 121). For example, suppose that
students at stage 2 are asked to draw three-fifths of a fruit bar. These students can
partition the bar into five equal parts, disembed one part and iterate that part three
times. They view the result as three out of five parts of the bar. This scheme marks
the beginning of length meanings for fractions, which are not fully realized until
students construct more advanced fraction schemes that are in the province of stage
3. Students at stage 2 encounter other constraints on aspects of fractions knowledge;
we give an example in the next section.
Stage 3
Students operating at stage 3 have the operations to construct nearly all fractions
knowledge. That does not mean the construction of fractions knowledge is easy:
There is considerable work involved. However, many schemes can now be con-
structed in comparison to what is possible at stage 2 (Olive & Steffe, 2010b).
The Iterative Fraction Scheme and Iterable Unit Fractions Researchers have
found that operating at stage 3 is required for constructing fractions as multiples of
unit fractions (Hackenberg, 2007; Hackenberg & Lee, 2015; Olive & Steffe, 2010a).
The main reason is that to think of a number like 9/7 as something that can be
worked with further—or even just as a number that makes sense—a student has to
11
There are always exceptions. For example, fourth-grade student Laura was operating at stage 2
and did not construct a partitive fraction scheme because she did not apply her iterable unit of 1 to
lengths (Steffe, 2010c).
Fig. 11.5 Conceiving of 9/7 as three levels of units
be able to see the 9/7 as nine times 1/7. Furthermore, the student has to see any of
those parts as a length that can be iterated 7 times to make the unit whole or bar, 7/7.
Thus, the student coordinates three levels of units: the 9/7 as a unit of nine units, 1/7
as a single unit, and 7/7 as a unit of seven units (Fig. 11.5). What we have just
described is called the iterative fraction scheme (Olive & Steffe, 2010b). When
students have constructed this scheme, fractions are numbers “in their own right”
(Hackenberg, 2007, p. 27).
In fact, students at stages 1 or 2 will often refuse to draw 9/7 of a unit bar, saying
that it does not make sense. Or, they will draw 7/9, or draw a 9-part bar where the
nine parts are not sevenths of the unit bar (Hackenberg, 2007; Hackenberg & Lee,
2015; Olive & Steffe, 2010b).
Units coordination can shed some light on this issue: As we have mentioned,
students at stage 2 can project composite units with iterable singleton units into
lengths. This structure is a unit of units, or a two-levels-of-units structure. For exam-
ple, students can create five-fifths as a unit of units and three-fifths as a unit of units
(Fig. 11.6). They can even view three-fifths as a unit of three units, where each fifth
is one part of the unit of five units. So, three-fifths can be a three-levels-of-units
structure in the activity of creating it. However, because students at stage 2 do not
maintain three levels of units in further operating, three-fifths becomes solely a unit
of units, which they can compare to five-fifths as a unit of units. This comparison is
a part-to-whole comparison, not a multiplicative relationship between one-fifth and
three-fifths. This limitation of the partitive fraction scheme is not always evident
until students work with fractions larger than one.
Indeed, an iterative fraction scheme marks the point where unit fractions become
iterable units. When unit fractions are iterable units, it means that students antici-
pate the operation of iteration to make the whole and also to make other fractions in
that “family.” For example, consider 9/7 in Fig. 11.5. Students who have constructed
an iterative fraction scheme anticipate that 1/7 can be iterated seven times to pro-
duce 7/7—they do not need to carry out that iteration; 1/7 contains that idea for
them. And, they can use that idea to produce other fractions in the number sequence
of sevenths: Nine-sevenths is nine times 1/7, one hundred sevenths is 100 times 1/7,
Fig. 11.6 Conceiving of 3/5 as two levels of units
etc. So, an iterable unit fraction is the basis for being able to construct any fraction
in the sevenths family as a multiple of 1/7 (within reason). Thus, the iterative frac-
tion scheme is the basis for constructing fractions as measures (Hackenberg &
Sevinc, 2022; Steffe et al., 2014).
Taking Fractions of Fractions (Fraction Multiplication) Iterable unit fractions
and switching between three-levels-of-units structures are useful in the construction
of a scheme for taking fractions of fractions. We give an example of Deborah, a
sixth-grade student at stage 3, in contrast with her partner, Bridget, a sixth-grade
student at stage 2 (Hackenberg & Tillema, 2009).
During an 8-month teaching experiment, Deborah and Bridget worked for about
one month on taking fractions of fractions, starting with unit fractions and progress-
ing to proper and improper fractions (Hackenberg & Tillema, 2009). Our concep-
tual analysis prior to this segment of the experiment indicated that a distributive
operation was important for the construction of a fraction composition scheme (cf.
Olive & Steffe, 2010a). For example, to take 2/5 of 3/4, we conjectured the follow-
ing: A student has to take 2/5 of each of the three one-fourths. Taking 2/5 of 1/4 is
twice 1/5 of 1/4, or twice 1/20. Then, three of the 2/20 will yield the result, 6/20. So,
a student has to distribute taking 2/5 of a part across each of the three one-fourth
parts and view those parts in relation to all four-fourths in the unit bar.
We found that Bridget coordinated parts multiplicatively within the given frac-
tional unit, but not in relation to the unit bar (Hackenberg & Tillema, 2009). In solv-
ing the problem above, 2/5 of 3/4, Bridget thought the answer was 6/15. To solve the
problem, she copied the unit bar, partitioned it into fourths, and colored three-
fourths of it. Then, she made each fourth of the 3/4-bar into five equal parts, for 15
parts total (Fig. 11.7). Then, without making more in the picture, she said, “And
then, that would be fifteen, three, she’d have six.” When questioned, she said “Six-
fifteenths.” So, she appeared to divide the 15 parts by 5, got 3, and iterated that to
make 6 parts as the result. However, rather than see those six parts in relation to the
unit bar, she named them in relation to the 3/4-bar with which she had been working.
The unit bar was in Bridget’s visual field, so it was not a matter of not seeing it.
Because this way of operating happened repeatedly, it indicated to us that the 3/4-
bar that she was operating on became her conceptual world, so to speak. She was
consumed with coordinating the units in that 3-part bar: It was a unit of three parts,
each containing five mini-parts, and then she reorganized it as a unit of five parts
each containing three mini-parts. Because Bridget was operating at stage 2, she did
not maintain the first three-levels-of-units structure when she made the second
three-levels-of-units structure. So, her conceptual world literally did become the
unit of five parts each containing three parts, and it was sensible for her to respond
“six-fifteenths” (p. 11).
The way that Deborah found 4/3 of 11/9 of a yard illuminates her units coordina-
tion and the nature of unit fractions for her (Hackenberg & Tillema, 2009). Deborah
made 11/9 of a yard by iterating 1/9 of a yard 11 times. Then she partitioned only
the first ninth into three equal mini-parts. She pulled out one of those mini-parts and
iterated it 44 times (Fig. 11.8). She knew immediately that one part was “one
twenty-seventh” (p. 13). In contrast, Bridget said it was “one thirty-third,” (p. 13)
naming the part in relation to all of the parts in the bar Deborah had made. In expla-
nation, Deborah said, “Take one out of each of those boxes [the eleven one-ninths]
and that equals eleven and that would be one-third, so I multiplied that by four
‘cause she needs four” (p. 14).
Deborah’s explanation indicates that she was taking 1/3 of 11/9 by taking 1/3 of
each of the eleven one-ninths. Then, she knew that she needed four of those 11-part
bars. The operations she used to create the result indicate that she was squarely
focused on the part as one-third of one-ninth, which was why she knew it had to be
Fig. 11.7 Bridget taking 2/5 of 3/4 (Hackenberg & Tillema, 2009, p. 11)
Fig. 11.8 Deborah taking 4/3 of 11/9 (Hackenberg & Tillema, 2009, p. 14)
one twenty-seventh. And, finally, we note that she only partitioned one of the ninths,
not all of them; partitioning one stood in for partitioning all of the “boxes.” This
activity is a good example of a unit fraction as an iterable unit: One-ninth could
stand in for all of the 11 ninths in the problem, and she could operate on one of them
as if she were operating on all of them. It also shows clearly how Deborah coordi-
nated multiple three-levels-of-units structures: 11/9 was a unit of 11 units, any one
of which could be iterated to make 9/9; 1/3 of 1/9 was a unit embedded in a unit of
nine units of three, and also, it was a part of a unit of four units (thirds), each of
which contained 11 of these units. Finally, 44/27 was also a unit of 44 units, any one
of which could be iterated to make 27/27.
Units Coordination and Algebraic Reasoning: Examples
In this section, we present how students at different stages of unit coordination

engage in basic algebraic reasoning. Like many others, to frame algebraic reason-
ing, we use Kaput (2008), who conceived of algebra as involving two core aspects:
systematic reasoning about generalizations of “regularities and constraints” (p. 11)
and syntactically guided reasoning about those generalizations with conventional
symbol systems. These two core aspects manifest in three strands; the work we
discuss in this section falls into the first strand, “algebra as the study of structures
and systems abstracted from computations and relations, including those arising in
arithmetic (algebra as generalized arithmetic) and in quantitative reasoning”
(p. 11).12
One idea in this strand is developing meanings (Thompson, 2013) for unknowns,
variables, and equations that are used to represent reasoning and generalizing about
numbers and quantities. This idea spans both of Kaput’s (2008) core aspects because
it is about systematically reasoning about generalizations and then operating on and
with conventional algebraic notation. Research on units coordination and algebraic
reasoning is less developed than research on units coordination and fractions knowl-
edge, and so we do not provide a table to summarize findings.
Stages 1, 2, and 3: Quantitative Unknowns and Conjectures About Them
We start by giving an orientation to how we have conducted research on units coor-

dination and meanings for unknowns, variables, and equations that are used to rep-
resent reasoning and generalizing about numbers and quantities. In particular, like
many others, we view students’ quantitative reasoning to be an important basis for
their algebraic reasoning (Ellis, 2007; Kaput, 2008; Olive & Caglayan, 2008; Smith
12
The other two strands are algebra as the study of functions, relations, and joint variation, and
algebra as the application of a cluster of modeling languages both inside and outside of mathemat-
ics (Kaput, 2008, p. 11).
& Thompson, 2008; Stephens et al., 2017). For us, a quantity is a measurable prop-
erty of an object or phenomenon (Thompson, 2011). To conceive of a property as
measurable, a person needs to conceive of a measurement unit and measurement
process (Thompson, 2011). Carrying out the process can produce a value for the
quantity.
Following Steffe and colleagues (Steffe et al., 2014), we conceive of a quantita-
tive unknown as the value of a fixed quantity that is not known but could be deter-
mined. For example, consider an unknown length of a field. The object here is the
field, and the property is the distance from one edge to the other. People conceive of
this distance as a measurable quantity when they can think about it as subdivided
into length units that could be enumerated. When we do not know the number of
units that span the field, the length is an unknown and could be depicted as in
Fig. 11.9.
Findings from Hackenberg (2013), Hackenberg and Lee (2015), Steffe and Olive
(2010) and Steffe et al. (2014) caused us to conjecture that students at stages 2 and
3 could construct quantitative unknowns. A main reason is that students at stages 2
and 3 can construct equi-partitioning schemes (Steffe, 2010c). Therefore, these stu-
dents can imagine a length as a unit of mini-units, where any one of those mini-units
can reconstitute the length. We conjectured this construction would be a foundation
for thinking of an unknown length as a unit consisting of an iteration of mini-units
where the number of those mini-units was not known (Fig. 11.9).
We note that in our studies, students either worked with a single quantitative
unknown measured with two different units (Hackenberg et al., 2021a) or with two
quantitative unknowns (Hackenberg & Sevinc, 2021, 2022). We have argued that
working with two unknowns in this way is conceptually more challenging than
working with single unknowns but not as complex as working with two unknowns
that are varying (Hackenberg et al., 2021a).
Because students at stage 1 have not constructed disembedding operations and iter-
able units of 1, they do not construct equi-partitioning schemes. So, we conjectured
that students at stage 1 would not construct quantitative unknowns. The lack of a
disembedding operation for students at stage 1 has at least one other implication for
their algebraic reasoning: It affects the generalizing that they do; see Hackenberg
(2013) for examples.
Fig. 11.9 Drawing of an

unknown length thought of
as a unit of an unknown
number of mini-units
Stages 2 and 3: Drawings of Quantitative Unknowns
We investigated our conjectures that students at stages 2 and 3 could construct quan-
titative unknowns in three iterative after-school design experiments with 6–9 middle
school students (Hackenberg et al., 2021a, b). Because students at stage 1 elected
not to participate in the experiments, we were not able to test the conjecture about
them directly. In these experiments, Hackenberg engaged students in field trips
around the school where students identified quantitative unknowns and tried to draw
pictures to represent them—including representing the unknown nature.
Students often identified unknown distances, such as the height of the classroom.
To represent this unknown, the students drew segments. However, any segment was
a known length. So, the students grappled with how they could indicate it was an
unknown. Students had a variety of suggestions, and Hackenberg did not experience
differences among the students at stages 2 and 3 in this aspect of their work together
(Hackenberg et al., 2021a). For example, a student at stage 3 said to make parts but
put an x near the segment (Fig. 11.10, left). A student at stage 2 suggested drawing
a bar with a couple of units at each end of the bar and with arrows to show that the
units continued, but the number of them was not known (Fig. 11.10, middle). A
student at stage 3 suggested doing that but using “dot dot dot” to show that the units
continued (Fig. 11.10, right). Students then used these ideas when they drew pic-
tures of quantitative unknowns.
Stages 2 and 3: Equation Writing
Although there were no discernable differences between their drawings of

unknowns, students at stages 2 and 3 differed in how they wrote equations with
these unknowns. Following this work on conceiving of unknowns in the study, stu-
dents related two values for a single unknown (Hackenberg et al., 2021a). One
example is the length of hallway problem (Fig. 11.11).
Fig. 11.10 Teacher

recording of students’
suggestions for how to
draw an unknown (from
Hackenberg et al., 2021a,
p. 6)
Length of Hallway Problem. The length of the school’s downstairs hallway is

unknown. You have two measuring units, a straw length and a pen length. The
pen’s length fits along the straw’s length exactly 2 times. Imagine measuring
the school’s height in straw length, and then measuring it again in pen lengths.
You would get two different values.
a. Draw a picture to show how the measurement units are related.
b. Draw a picture to show the two values for the hallway’s length.
c. Describe how those two values for the hallway’s length are related.
d. Let’s let a letter stand in for the length of the hallway measured in
straw lengths, and another letter stand for the length of the hallway
measured in pen lengths. What equation can you write to show the
relationship between these two unknowns? Explain your equation.
e. Can you write a second equation with your unknowns from (d)?
Explain your equation.
Fig. 11.11 Length of hallway problem
On these problems, five of the seven students at stage 2 wrote equations that
represented the relationship between measurement units rather than the relationship
between the values for the unknown length or height. Students drew pictures; the
one in Fig. 11.12 is from a group of three students at stage 2. Here, the bottom bar
shows the hallway length marked into a few straw lengths, which a seventh-grade
student at stage 2, Elliot, drew. Then, the top two bars were made by the students
together: They tried parts in the middle bar until it looked like it was a bar of straw
lengths, and this turned out to be 35 parts. Then, the students doubled that to get 70
and made the 70-part bar at the top. So, from our perspective, the top two bars are a
pair of values for the length of the hallway.
The teacher and students had a whole class discussion about this problem, with
this picture projected and with H and P defined as the number of straw lengths that
fit in the length of the hallway and number of pen lengths that fit in the length of the
hallway. The teacher asked for equations, and two students at stage 2 suggested
P = 1/2H and 2P = H. We note that, from our perspective, those equations represent
the relationship between measurement units: Two pen lengths make a straw length.
The equations do not represent the relationship between P and H as defined. A stu-
dent at stage 3, Harry, articulated P = 1/2H as “the number of pens in the hallway is
equal to half the number of straws that fit in the hallway” (p. 10). There were mur-
murs of “no” to this from all students. A student at stage 3 was going to offer an
equation, but the teacher asked him to wait because Elliot, a student at stage 2,
raised his hand. Then this exchange occurred (p. 10):
ata Excerpt 1: Elliot and the Teacher Converse About His Ideas
D
About Equations
Elliot: I was going to say, um, see how many pen lengths there are and multiply
it by two.
Teacher: See how many, you want to see how many of these there are [pointing to
top bar in Fig. 11.12]?
E: Yeah.
T: And multiply it by two to get the number of straws?
E: Yes.
T: Okay, that’s Elliot’s idea. But wait a minute. How many, which is going to be
more, a bigger number, the number of pen lengths in the hallway or the number
of straws?
E [immediately]: Pen lengths.
The class then had more discussion about which was a bigger number, the num-
ber of pen lengths or the number of straw lengths, with contributions from all stu-
dents affirming the number of pen lengths was a bigger number. The teacher then
returned to the equations. Elliot said, “I think P equals 1/2 of H” (p. 10). After 3.5
more minutes of discussion, during which another student at stage 3 explained his
support of P ÷ 2 = H, the teacher asked again whether 2P = H or 2H = P was true.
“The first” said Elliot (p. 10).
This data shows how focused Elliot was on writing an equation between mea-
surement units, such as P = 1/2H and 2P = H, even though he agreed immediately
and repeatedly that there were more pen lengths than straw lengths in the length of
the hallway during discussion. Our argument is that Elliot constructed the length of
the hallway as a quantitative unknown, a unit of an unknown number of units as
shown by his initial picture (Fig. 11.12, lowest bar). However, he did not structure
the hallway length as a unit of an unknown number of larger units, each containing
two smaller units. Instead, he assigned a known value, 35, and he doubled that to get
70. These actions with knowns are well within his ways of operating: He could have
Fig. 11.12 Recreation of a student at stage 2 group’s picture (Hackenberg et al., 2021a, p. 9)
made the 70 as a unit of two units of 35, or even as a unit of 35 units, each contain-
ing 2 units. If he thought of 70 in this way, as three levels of units, it was in activity.
That unit structure did not remain for him, and he then viewed both the 70 and 35 as
two different two-levels-of-units structures.
Since the hallway length was not a three-levels-of-units structure for Elliot, he
did not have a structural reason to support 2H = P as an equation. Instead, what
appeared to be salient structurally to him was a straw length as a unit of two pen
lengths, and he represented that in his equations. So, even though he constructed a
quantitative unknown as a composite unit of an unknown number of units, “his
equations expressed one measurement unit as a composite unit of the other mea-
surement unit, a two-levels-of-units structure” (p. 10). Four of the six other students
at stage 2 in the study were similar to Elliot.
When students at stage 3 wrote equations for problems like the length of hallway
problem, they initially also represented the relationship between measurement units
(Hackenberg et al., 2021a). However, with questioning, they revised their equations
and represented the relationship between the values for the unknown length or
height—and they did this consistently.
For example, on the length of hallway problem, like many students at stage 2, the
students at stage 3 conceived of the hallway length as a unit of an indeterminate
number of straw lengths. However, they did more: They also structured the hallway
length as a unit of straw lengths, each of which contained two pen lengths, and they
could hold this structure out for reflection in writing and evaluating equations. This
ability to hold the three-levels-of-units structure out for reflection is, we suggest,
what allowed Harry to articulate P = 1/2H as “the number of pens in the hallway is
equal to half the number of straws that fit in the hallway” (Hackenberg et al., 2021a,
p. 10). That is, he and the other students at stage 3 kept P and H squarely in mind as
the number of pen lengths and straw lengths measuring the hallway, even though the
exact number was unknown. Doing so allowed him and other students at stage 3 to
argue for equations like P ÷ 2 = H and 2H = P.
Stage 3: Reciprocal Reasoning
In our three design experiments, we also explored students’ reciprocal reasoning

with unknowns. Our findings revealed that seven out of nine students at stage 3
constructed reciprocal reasoning with unknowns (Hackenberg & Sevinc, 2022),
while no students at stage 2 constructed it (Hackenberg et al., 2017; Hackenberg &
Sevinc, 2021). Reciprocal reasoning involves measuring quantity A with quantity B
and vice versa; that is, if A is 3/5 of B, then reciprocal reasoning is involved in deter-
mining how to measure B with A (Hackenberg, 2010; Thompson & Saldanha,
2003). We found that reciprocal reasoning involves constructing a two-way relation-
ship between the quantities where either quantity can be the referent unit to measure
the other. We also found that constructing a unit fraction of an unknown as an iter-
able unit was the source of reciprocal reasoning with unknowns, which is in the
province of students at stage 3 (Hackenberg & Sevinc, 2022).
For example, seventh-grade student Martin and eighth-grade student Gabriel,

both students at stage 3, constructed reciprocal reasoning as they worked together
on the fern sunflower height problem (Fig. 11.13) (Hackenberg & Sevinc, 2022).
Our interpretation is that the students switched the referent unit from the fern
height to the sunflower height. Martin and Gabriel used their iterative fraction
scheme to see the fern height as five one-thirds of the sunflower height. So, they
measured the fern height with the sunflower height that they had initially made from
the fern height. Thus, they used their iterative fraction scheme recursively.
Furthermore, they showed evidence of double-naming each part of both heights, as
shown in Gabriel’s picture (Fig. 11.13, right). In other words, each part was both 1/5
of the fern height, 1/5x, and 1/3 of the sunflower height, 1/3y, and either could be
used to measure either height. In short, they constructed a multiplicative relation-
ship between a unit fraction of an unknown and the unknowns, which is the basis for
our claim that they constructed unit fractions of unknowns as iterable units. This
was the root of their reciprocal reasoning.
Yet as mentioned, two of the nine students, despite having constructed iterative
fraction schemes where unit fractions were iterable units, did not construct unit
fractions of unknown as iterable units (Hackenberg & Sevinc, 2022). In other words,
they did not anticipate a multiplicative relationship between a unit fraction of an
unknown, like 1/3y, and an unknown, like x. Instead, they constructed reversible
relationships with whole number multiplication and division (e.g., x ÷ 5 × 3 = y and
y ÷ 3 × 5 = x), and they did not interpret these equations with their iterative fraction
schemes. That said, these two students at stage 3 still constructed more advanced
schemes for relating two unknowns than our most advanced student at stage 2 in the
experiments (Hackenberg & Sevinc, 2021). These findings point to the variety of
student thinking within each stage and, in particular, to variety within students at
stage 3 (cf. Shin et al., 2020).
tandards of Evidence for Making Claims about

S
Units Coordination
As we have shown, students’ units coordination stages are one part of building
second-order models of students. Therefore, assessing units coordination stages
well is important for robust research findings based on second-order models. In
addition, assessment of students’ units coordination stages can support formulation
of learning goals and problem sequences for students in design research.13 Thus,
understanding how to make claims about students’ units coordination is important.
So, we now turn our attention to how to make well-supported claims.
13
It can also support formulation of learning goals and curricular activities in classroom instruc-
tion, a separate but overlapping topic that requires working on how classroom teachers learn to
assess units coordination stages and use the results in their teaching.
A fern and sunflower are growing in

a garden, each of unknown height.
The height of the sunflower in cm is
3/5 the height of the fern in cm. Draw
a picture of this situation and write
two equations for the relations
between two unknown heights.
Fig. 11.13 Fern sunflower height problem (abbreviated) and Gabriel’s drawing
At this point, research on units coordination has been conducted largely in two
ways. First, research on units coordination has been conducted via interviews,
teaching experiments, and design research, in which the focus has been on under-
standing in depth a relatively small number of students (e.g., Hackenberg & Tillema,
2009; Steffe & Olive, 2010). Second, research on units coordination has been con-
ducted via written assessments given to larger numbers of students, in which only
written markings of students are analyzed (e.g., Norton et al., 2015; Ulrich &
Wilkins, 2017; Zwanch & Wilkins, 2021). The development of written assessments
has been a contribution to the field so that larger numbers of students can be assessed
efficiently. However, written assessments alone have a cost in that students’ written
markings provide limited evidence for making claims. Because of this limitation,
standards of evidence for making claims about units coordination based on written
assessments have been articulated carefully (e.g., Norton et al., 2015; Ulrich &
Wilkins, 2017). In our view, making claims from qualitative data has not been artic-
ulated as carefully, in part because these claims are often one piece of a second-
order model, and other features of the model may be the focus of the research report
(e.g., Hackenberg et al., 2021a). Therefore, in this section, we address what is
needed to make sound claims about units coordination from interview data, using an
extended example.
Preliminary Requirements
We mention three requirements prior to discussing our example: good tasks, good
probing during interactions with the student, and quality recording of the interac-
tions and student work. For sake of simplicity, we assume that a researcher is con-
ducting an interview with a single student, but interviews with small numbers of
students together are certainly possible.
Good Tasks
The interview must involve tasks that elicit information about units coordination.
One example is the seven units coordination tasks in the context of lengths (bars)
developed by Norton and colleagues (Norton et al., 2015); these tasks together form
a written assessment that can be used in interview settings.
When working with discrete units, four levels of units from the researchers’ per-
spective are needed to elicit student thinking across a range of units coordination
stages (Norton & Boyce, 2015). We call these tasks “embedded units” tasks, and we
give an example later in this section. A third type of task that is useful is tasks that
switch between units coordinating and unit segmenting (Steffe, 1992).14 We give an
example later in this section.
Good Probing of Students
The interview itself has to involve good probing of students’ thinking (see Steffe &
Thompson, 2000) to elicit how students are constructing the involved quantities. We
give some examples in this section. It is essential to investigate the quantities stu-
dents are making and to be on alert that students’ quantities may not be similar to
the quantities researchers have made. For example, if a student says, “32 cans are in
a box,” it is important not to take for granted that the student is thinking of what the
interviewer is thinking. Without good probing, the nature of students’ quantities and
conflations that students make (from researchers’ perspectives) will not be visible.
Good Records of Students’ Interactions and Work
Finally, quality recording of students’ interactions and work is essential for retro-
spective analysis (Steffe & Thompson, 2000). We recommend having two cameras,
one to record the interaction between interviewer and student and one to record the
students’ written work as it evolves. Recording the written work as it evolves is
essential for making good claims from an interview. These records allow research-
ers to carefully review the whole progression of the student’s work.
Data Analysis and Claims: An Example
Now, we turn to our extended example. Emily was a seventh-grade student who
participated in a classroom design experiment as part of the first author’s research
project. At the start of her seventh grade in September, she participated in an initial
14
Unit segmenting refers to a scheme in which a person conceives of “at least two composite units,
one to be segmented (or fragmented) and the other composite unit to be used in segmenting (or
fragmenting)” (Steffe, 1992, p. 267). It can be seen as a basis for solving measurement division
problems.
interview to assess her units coordination stage and fractions knowledge. She also
completed two written assessments on units coordination, one by Norton and col-
leagues (Norton et al., 2015) and one developed in our project. She was selected to
be a focus student during the proportional reasoning unit that the classroom teacher
and first author designed and co-taught in late October and November; the purpose
of the experiment was to design to differentiate instruction during the unit
(Hackenberg et al., 2023). Because in her follow-up interview in December, Emily
showed evidence of a stage change (Hackenberg et al., 2021c), building a thorough
second-order model of her mathematics was of strong interest to us.
In this section, we aim to portray the beginning of that model building, which is
messy. In contrast, research reports are typically neat and tidy. Most of the wrinkly
ambiguity of actual interactions and analysis of them gets ironed out; uncertainties
get stated in clear ways or simplified. To a great degree, this is necessary for com-
municating with others—and for coming to conclusions for ourselves. However, it
can render invisible the pondering, reaching, groping, debating, weighing, and sit-
ting with uncertainty that goes into data analysis in general and creating a second-
order model in particular. Here, we are going to try to represent some of that by
sketching our re-presentation (von Glasersfeld, 1995) of our experiences analyzing
Emily’s mathematics.
Also, this section demonstrates the difference between the working second-order
models made in the moments of interacting with students and the analytical second-
order models that are made retrospectively (Steffe & Thompson, 2000).
Working Model of Emily’s Mathematics in September
In fall semester of 2017, Emily completed her units coordination assessments on

August 29 and 30. The research team, including consultant Andy Norton and his
graduate research assistant Sarah Kerrigan, analyzed the data in order to provide
guidance for selecting students to be interviewed. Emily was assessed to be at stage
1 based on both written assessments, and she was interviewed on September 11. At
the start of her interview, it seemed clear to the interviewer (the first author) that
Emily was operating at stage 1. A main reason was her work on the tiles task, which
we present shortly. However, her work on the crate task caused the interviewer to be
less certain. Still, based on the results of her written assessments and the work on
the tiles task, the first author made a “final” working assessment that she was operat-
ing at stage 1. And then, the classroom unit on proportional reasoning began about
one month later.
Retrospective Model of Emily’s Mathematics in September
Retrospectively, when we viewed Emily’s initial interview, some aspects of her

thinking indicated stage 1 to us, and some indicated stage 2. And, also, nearly simul-
taneously with this research, Ulrich (2016b) and Ulrich and Wilkins (2017)
Fig. 11.14 Tiles problem
published papers about the students operating with an advanced version of the TNS
(aTNS). Students who operate with an aTNS assimilate with composite units but
have not yet constructed the iterable units of 1 that mark stage 2. So, they are stu-
dents at stage 1 with the advantage of being able to read composite units into prob-
lem situations prior to operating. Thus, we had another issue to grapple with: Was
Emily an aTNS student?
We show two excerpts from Emily’s initial interview, one that shows strong evi-
dence of her being a student at stage 1, and the other that is more ambiguous. In the
first excerpt, the interviewer has just posed the tiles problem (Fig. 11.14).
Emily divided 60 by 6. She got 10 and subtracted 4 to get 6. Then, this interaction
followed.
Data Except 2: Emily’s Work on the Tiles Problem
Teacher: So, what does a row look like of these tiles? Do you think you could draw
a picture of what one row looks like?
[Emily draws one column of 4 squares.]
T: That’s very nice. She’s put down six of those, right? So this is one row. What
would it look like if she put down six of those rows?
[Emily draws two more squares in a second, adjacent column.]
T: Okay, explain that to me.
E: If you already put down four, [then much more energetically], oh six rows? Oh!
[She quickly draws more tiles, one-by-one, in columns of 4 until she has 6
columns with 4 tiles in each.]
T: That changes it, huh? I see. I can see that really clearly, [counting columns] 1,
2, 3, 4, 5, 6 with four tiles in each row. Great. That uses up some amount of
tiles, right?
E: Yes. I should have done six times four.
[Small part omitted during which Emily says 6 times 4 is 24.]
T: 24 means what then?
E: That’s how many she’s put down so far.
T: Does that change anything about what you want to do next?
E: Yes. [She calculates 60–24.]

T: What did you find there?
E: If I’m doing six times four, I got 24 and then if the question is how many more
she has to do, I already have 24, so I did 60 minus 24 and got 36.
T: Does that answer the question of how many more rows that she has to make
with those tiles?
E: Mmmm…
T: Or you’re not sure?
E: I’m not really sure.
T: What does the 36 mean? Is that tiles or rows or what would you say?
E: The question’s how many rows?
T: How many more rows, yes.
E: Maybe tiles because I’m not sure if 36 rows makes sense.
T: It seems like too many rows? Okay. Tell me again why you subtracted here. I
think that was a really good idea.
[Emily explains that at her old school she did a problem where she multiplied and
then subtracted, so she did that here.]
T: The 60 means what again in the problem?
E: 60 means how many she has in the bag.
T: Okay, how many tiles. So, that’s tiles. Then the 24 you told me was how many
tiles here [the 4 by 6 array of squares], right?
E: Yes.
T: You subtracted and you got 36, so that must mean what in the problem then?
E [very quietly, questioning]: Tiles?
T: 36 is tiles? You’re not sure.
E: I’m not sure. Maybe tiles because to me 36 rows doesn’t really make sense.
T: Because it’s too much?
E: Yes, I think it’s too much.
T: Okay. Well, let’s go with that. So, let’s say it’s tiles. How would you figure out
how many rows she can make? There’s four tiles in every row.
E: 36 divided by 6 maybe?
This except shows strong evidence that Emily was operating at stage 1
(Hackenberg et al., 2021c). The initial computation (60 ÷ 6 = 10; 10 − 4 = 6) is not
strong evidence of it: Unfortunately, it is not unusual for students at all stages to try
a computation with numbers that do not have a relationship to the problem from our
perspective. What is evidence for stage 1 is that Emily initially drew only six tiles,
not six rows. She self-corrected that fairly quickly, which means she could think
about a row as a unit of four tiles, or a composite unit of 4. Without the
self-correction, we might not attribute composite units of 4 to Emily, which would

be evidence of operating with the initial number sequence (INS), a stage below
stage 1.
A second piece of evidence for stage 1 is that Emily was persistently unsure of
what the 36 meant once she had computed it. She seemed to be making a judgment
about its meaning based on size rather than structure. That is, she thought that 36
rows would be too much for the problem, perhaps because she had drawn 6 rows
that used 24 tiles. She did not appear to create the 60 as some number of rows of 4
that could be separated into six rows and some unknown number of rows. In fact,
she may not have been structuring 60 as some number of tiles that could be sepa-
rated into 24 tiles and a leftover part. This structuring of 60 into two parts, 24 and
an unknown number of tiles, is possible for students at stage 1 (Ulrich, 2015), but it
is an implicit structure because they have not yet constructed disembedding opera-
tions that are essential for explicit part-to-whole relations (Steffe, 2010b). Emily
subtracted 24 from 60, which means she had a sense of 24 being included in the 60
and that taking that amount away from 60 was relevant to the problem. However, we
interpret her uncertainty about the meaning of the 36 to reflect the lack of explicit
part-to-whole relationships.
Third and probably the most compelling evidence of stage 1 is her suggestion to
divide the 36 by 6 to determine the number of rows of 4 to be made with these 36
tiles. This activity shows that although Emily had constructed four tiles as a com-
posite unit and could use it to determine the number of tiles in a known number of
rows, she did not continue to see composite units of 4 as relevant in segmenting the
36. She suggested dividing by 6, despite our repeated talk about four tiles in a row
and her drawing six columns of four tiles.
Our interpretation is that, like TNS student Maya in Steffe’s (1992) research,
Emily had not constructed a feedback system between units coordinating and unit
segmenting. The units of composite units of 4 that Emily made when she tracked six
4 s were not available to her when she did not know how many 4 s to track. Her sug-
gestion to divide by 6 is evidence that she was making the coordination of six 4 s in
her activity, and it was not available to her in reflection. We note that in our view, her
suggestion to divide by 6 is also evidence that she was not operating with the aTNS
(Ulrich, 2016b) because it does not show evidence of assimilating the problem with
a composite unit of 4. However, we acknowledge that we did not pose tasks to
Emily that would help us determine more definitively whether she was operating
with the aTNS.
With interviewer support, Emily did eventually divide 36 by 4 and found 9 rows.
Then, we posed the crate problem (Fig. 11.15) to Emily.
Emily made an initial drawing of each containing unit separately (Fig. 11.16,
left) and explained her drawing as follows: “This is a box [middle picture] and
inside the box is packages and inside this crate [right] is boxes and inside those
boxes are packages and inside those packages are cans.” The interviewer asked her
to draw a picture that showed that. Emily drew a second picture (Fig. 11.16, right).
At this point, the researcher asked Emily how many cans of juice were in
the crate.
Fig. 11.15 Crate problem
Fig. 11.16 Emily’s initial drawing (left) and second drawing (right) of the crate problem
Data Excerpt 3: Determining the Number of Cans of Juice in the Crate
E: How many cans of juice in a crate?

T: Yes. How could you figure that out?
E: What you could do, well a crate has six boxes. Then in these boxes. [pause]
Maybe multiplication? Like six times eight [writes 6 and 8, pauses] 48?
T: What does that find? Does that find the number of cans?
E: No. [pause] I don’t think.
T: It might. I don’t know.
E: Maybe if you do 48 times 4.
In our experience, students at stage 1 often do not get this far in the crate prob-
lem: They do not successfully draw a picture of the whole crate (from our perspec-
tive) that shows all of the embedded units, even with prompting. In addition, they do
not usually find the number of cans by multiplying 6 by 8 and then by 4, with refer-
ence to the quantities in the problem (some do multiply the numbers listed in the
problem but do not have reasons for doing so).
However, returning to the interview, Emily was not sure if the result of 48 × 4
was right because 192 seemed “like a lot.” So, the interviewer explored with her
what the 48 meant in the problem. Emily confirmed that there were 48 packages by
adding the 8 s one by one, counting on by 1 s. She seemed heartened that she arrived
at 48, noting, “I’ve got that one right, when I did that.” The interviewer asked her if
she thought 48 × 4 determined the number of cans. She said no and suggested,
“Maybe you could do eight times four to get you to 32.” Again, the interviewer
asked her what that meant in her picture, and she said it was the cans of juice.
ontinuation of Data Except 3: Continuing to Determine the Number

C
of Cans in the Crate
T: In the whole crate or in one part of it, or what would you say?
E: That would only be in one part of it.
T: What part would it be?
E: That part. [Points to the first box.]
T: Okay.
E: So, maybe I did get it right when I did 48 times 4 because we need to figure out
how many are in all of the packages, and the only way to figure that out is to
find out how many packages are in all of the boxes which would be 48 and then
do 48 times 4 which is 192.
T: Tell me one more time why do you multiply 48 times 4.
E: I multiply 48 times 4 because 48 is the number of packages in all of the boxes.
So, there’s eight in all of these [gesturing to the boxes] but if you add up all of
these, it’s 48… If you want to figure out, if you only do eight times four, that’s
only 32 which is in one package and you need to figure out how many cans are
in all of the packages, so you do 48 times 4.
Her explanations for why multiply 48 by 4 are fairly clear. She seems to be say-
ing, 48 is the number of packages in all of the boxes, and that is why you multiply
by 4. She contrasts that with 8 times 4, which she seems to be saying is not enough—
she says it is in one package, but she may have meant one box.
Yet, we note that Emily never explicitly says that there are four cans of juice in
each package. So, her reason to multiply remains implicit. In addition, her explana-
tion points to an issue that was about to arise for Emily when working on this prob-
lem: She began to view 32 as the number of cans in a single package, not the number
of cans in eight packages. The following interaction occurred directly after her
explanation above.
Second Continuation of Data Except 3: What Does the 32 Mean?
T: Okay, I see. 32 refers to cans, right, and you said it’s in here?
E: Yes. 32 refers to how many cans are in one of the packages.
T: One package or one box or one what?
E: One package.
T: One package? And these [small rectangles inside the six large rectangles] are
packages, right?
E: Yes, these are packages and four cans are in the packages. So, the cans are in
the packages.
T: So, the 32, I want to just make sure I understand. 32 refers to how many in this
[a box] or how many in one of these little guys [a package]?
E: How many in one of these ones [a package].
T: The blue [box] or the orange [package]?
E: The orange [a package]. If you add up all of these [8 packages in first box], it
would be 42 [she likely meant 32].
T: Okay, 32.
E: Yes, 32 because that’s in one of the, the blue is a package and we only figured
out, all we did was eight times four which is one package times four little juices.
T: Is it one package or eight packages, because I think you have eight in here, right?
E: Oh yeah, eight packages.
Here, Emily for about at least 1.5 minutes, thought that the 32 was 32 cans of
juice in a single package, rather than 32 cans in a box. Although students at stage
2—and even students at stage 3—sometimes conflated the units in this problem, no
student had maintained, under questioning, that there were 32 cans in a single pack-
age. We saw this as evidence that the 32 really was not structured as eight 4 s for
Emily: Once she made the 32 cans, it was a composite unit of 32 that did not refer
to the structure of how she had made it—and therefore what containing unit it
belonged to could vary. So, despite Emily’s use of 48 × 4 to determine the total
number of cans in the crate, this aspect of her interview indicated more evidence
that she was operating at stage 1. Based on this consistent pattern of evidence from
analyzing our entire data set, we concluded that Emily was indeed a student at stage
1 at the start of her seventh grade.
Population Estimates and Stage Changes
Given the influences of units coordination described above, natural questions

include how many students are operating at each stage of units coordination across
ages and grades and how to support students to advance to the next stage.
Steffe (Chap. 2, this volume) estimated percentages of elementary school chil-
dren who are operating with each of his number sequences (Table 11.3). For exam-
ple, he estimated that 40% of first-grade students have yet to construct the INS, and
about 50% are operating with the INS or at stage 1 (We do not have estimates about
how many students are at stage 1). Only about 10% of first-grade students are oper-
ating at stage 2, and none at stage 3. By third grade, about 60% are operating with
the INS or stage 1, while 40% of students are operating at stage 2. Third grade in the
United States is where whole number multiplication and division start to become a
central focus (National Governors’ Association Center for Best Practices & Council
of Chief State School Officers [NGACBP & CCSSO], 2010).
Table 11.3 Estimated percentages of students operating at different number sequences and units
coordination stages, taken from Steffe (Chap. 2, this volume)
Incoming grade Pre-numerical INS or stage 1/TNS Stage 2/ENS Stage 3/GNS
1st 40 50 10 0
3rd 0 60 40 0
5th 0 35 40 25
6th 0 30 30 40
Table 11.4 Percentage of students operating at different number sequences and units coordination
stages, taken from Zwanch and Wilkins (2021)
Grade Pre-numerical/INS Stage 1/TNS Stages 2/ENS and Stage 3/GNS Totals
6 10% (10) 61% (61) 29% (29) 100
7 5.4% (5) 61.9% (57) 32.6% (30) 92
8 6.8% (5) 37.7% (28) 55.4% (41) 74
9 6.7% (4) 43.3% (26) 50% (30) 60
Total 7.4% (24) 52.7% (172) 39.9% (130) 326
In one sense, this is appropriate: As we have discussed, students at stage 1 can

construct the most basic meaning for multiplication (Steffe, 1992). However, third-
grade standards include that students “understand properties of multiplication and
the relationship between multiplication and division” (NGACBP & CCSSO, 2010;
CCSS.MATH.CONTENT.3.OA.B.5). Students at stages 1 and 2 do not construct
multiplication and division as inverses of each other—that does not happen until
stage 3 (Steffe, 1992). So, these estimates have significant implications for how
teachers might work to support these students to become multiplicative reasoners.
In addition, the estimated number of students at stage 1 in grades 5 and 6 is an issue
for the construction of fractions knowledge that is expected during these grades.
Zwanch and Wilkins (2021) conducted a study in which they assessed the num-
ber sequences of 326 students across grades 6–9 in a rural district in the southeast-
ern United States. They used a measure from Ulrich and Wilkins (2017) to assess
number sequences; this measure consists of 25 items that assessed unit construction
and coordination across discrete and continuous contexts. As shown in Table 11.4,
in this district, the percentage of students at stage 1 in grade 6 is double what Steffe
estimated, and the percentage of stage 2 and 3 students combined is about half of
what Steffe estimated. These differences may be due to the fact that the students in
Zwanch and Wilkins (2021) study come from a low-income area, where students
may have been disadvantaged due to a lack of resources (Hegedus, 2018; Tate, 1997).
Finally, Acar and Sevinc (2021) also conducted a study to assess the units coor-
dination stages of 139 fifth through eighth-grade students at a middle school in
Istanbul, Turkey. They used Norton et al.’s (2015) 7-item units coordination assess-
ment. As shown in Table 11.5, at fifth grade, the percentages of students at stage 1
are higher than what Steffe (2017) estimated, while the percentages of students at
stage 3 are lower. At sixth grade, the percentages of stage 1 and 2 are both higher
Table 11.5 Percentage of students operating at different units coordination stages, from Acar and
Sevinc (2021), p. 96
Stages of unit coordination
Grade level Stage 1 Stage 2 Stage 3 Total
5 13 (52.0%) 9 (36.0%) 3 (12.0%) 25 (18.0%)
6 10 (47.6%) 10 (47.6%) 1 (4.8%) 21(15.1%)
7 25 (48.1%) 22 (42.3%) 5 (9.6%) 52 (37.4%)
8 14 (34.1%) 25 (61.0%) 2 (4.9%) 41 (29.5%)
Total 62 (44.6%) 66 (47.5%) 11 (7.9%) 139 (100%)
than Steffe’s estimations, while the percentage of students at stage 3 are markedly
lower. In fact, the percentages of students at stage 3 are quite low overall. In com-
parison with Zwanch and Wilkins’ (2021) study, Acar and Sevinc’s (2021) sample
included fewer students at stage 1 across grades. We note that large numbers of
stage 1 and 2 students in middle school grades are a cause for concern because of
the multiplicative reasoning required by many typical mathematical topics in mid-
dle school, such as proportional reasoning. Recent papers address this important
issue (Hackenberg et al., 2023; Lee & Shin, 2021).
Indeed, given that students at stage 1 are constrained to construct fractions as
parts within wholes because of the lack of disembedding operations, a natural ques-
tion is how to support them to become students at stage 2. Because units coordina-
tion stages are not directly controlled by instruction (Steffe, 2017), the extent to
which stage change can be advanced via teaching is uncertain. Norton and Boyce
(2015) have reported a case in which the researchers designed interactions to sup-
port a sixth-grade student at stage 1, Cody, to transition to stage 2. Over 14 sessions,
the researchers engaged Cody in embedded units problems. They used a problem
situation with chips put into cups, and the cups were then put into boxes. For exam-
ple, they put five chips in each cup, two cups in a box, and assigned the chips a value
of 2 cents. Then they asked Cody multiple rapid questions that caused him to think
across units, such as how much a box was worth, how much a cup was worth, how
many more chips are needed to fill a box if you have 14 cents worth of chips, etc.
(p. 216). As a result of this intervention, Cody showed evidence of becoming a stu-
dent at stage 2 with some counterevidence in the coordination of continuous units.
Yet, he clearly demonstrated considerable advances.
Another avenue for supporting stage change for students at stage 1 may be via
work with length units (Hackenberg et al., 2021c). We have already presented the
case of Emily, who was a student at stage 1 at the start of seventh grade. At the end
of the semester, Emily showed strong evidence of having constructed a fraction
scheme that is in the province of stage 2. The research team had not designed
instruction to support this change. However, in trying to account for the change, the
research team pointed to work in the proportional reasoning unit that may have
contributed to this shift. During the unit, Emily’s class worked on determining how
to make two cars travel at the same speed but for different distances and times.
Emily’s drawings of the journeys underwent a change from not showing relative
size to showing it (Hackenberg et al., 2023), and she sustained this change. This
work was likely one factor in her units coordination stage change.
These two case studies of stage change were conducted with middle school stu-
dents. Based on these two case studies, as well as findings from Zwanch and Wilkins
(2021), we suggest that middle school may be a fertile time for students at stage 1
to be supported toward stage 2. Zwanch and Wilkins found that sixth and seventh-
grade students who had constructed an advanced version of the TNS (aTNS; Ulrich,
2016b) were more likely to become ENS students in eighth and ninth-grade than
students who were operating with the “plain” TNS or INS. Since aTNS students
showed progress that TNS and INS students did not show, it is possible that identi-
fying aTNS students in middle school and working to support their stage change
would be quite fruitful. Recall that students operating with an aTNS assimilate with
composite units but have not yet constructed iterable units of 1. This finding indi-
cates that supporting students at stage 1 to assimilate with composite units by mid-
dle school is critical.
Conclusion
In this chapter, we have defined and characterized units coordination, and associ-
ated constructs, as a framework for modeling students’ mathematical thinking.
Units coordination describes and accounts for how students construct and organize
relationships between units. It is especially useful for modeling how students struc-
ture problem situations prior to acting and what operative activity is available to
students once they are working to solve a problem. We showed how students’ units
coordination stages influence students’ construction of fractions knowledge and
algebraic reasoning. We gave an extended example of how to make claims about a
student’s units coordination stage from interview data. And, we reviewed what is
known about units coordination in larger populations, as well as efforts to support
students to advance across stages.
We hope the reader can see how challenging it is to be a student at stage 1 in
upper elementary school and middle school because their ways of thinking are par-
ticularly out of step with goals and topics in standards and curricular materials (e.g.,
NGACBP & CCSSO, 2010; MNE, 2018). Furthermore, it does not appear to be the
case that more time alone will provide a solution, since in a recent study, students at
stage 1 operating with the TNS did not become students at stage 2 during middle
school—only aTNS students made that shift (Zwanch & Wilkins, 2021). Students at
stage 2 have constructed operations and schemes that allow them to navigate upper
elementary and middle school curricular demands better than students at stage 1.
However, as they work on proportional reasoning and other multiplicative topics in
middle school, they, too, face significant challenges because fractions are not mul-
tiplicative in nature for them (e.g., Hackenberg & Sevinc, 2021). And, in our view,
the typical curricular demands may underserve students at stage 3 by not supporting
them to fully develop the power possible with their flexible unit structures.
So, what is needed? In our view, mathematics educators must support upper
elementary and middle school teachers to understand and identify students’ units
coordination stages as they manifest in classrooms. This work requires teachers to
build working second-order models of their students and to act from those models
in their instruction (Bas-Ader & Carlson, 2021; Carlson et al., Chap. 9, this vol-
ume). For students to learn the most at any stage requires supporting them to learn
at their current stage, as well as opening opportunities for them to advance to the
next stage. Although implementing these recommendations is a daunting challenge,
it is also a part of creating a world where all students’ mathematical ways of think-
ing are centered in instruction and where all students have a better chance to make
progress. If we are going to take the findings from units coordination research seri-
ously, this is a critical next step to take.
References
Acar, F., & Sevinc, S. (2021). Investigation of middle school students’ unit coordination levels
in mathematics problems involving multiplicative relations. Elementary Education Online,
20(1), 90–114.
Aydeniz, F. (2018). Investigating elementary pre-service teachers’ distributive reasoning and
proportional reasoning (Doctoral dissertation, Indiana University). ProQuest Dissertations
Publishing.
Bas-Ader, S., & Carlson, M. P. (2021). Decentering framework: A characterization of graduate
and Learning, 24, 99. https://doi.org/10.1080/10986065.2020.1844608
D. Riedel.
Boyce, S., & Norton, A. (2016). Co-construction of fractions schemes and units coordinating
structures. The Journal of Mathematical Behavior, 41, 10–25.
Byerley, C., Boyce, S., Grabhorn, J., & Tyburski, B. A. (2019). Investigating STEM students’
measurement schemes with a units coordination lens. In Annual Conference on Research in
Undergraduate Mathematics Education, Oklahoma City.
Campbell, R. L., & Bickhard, M. H. (1986). Knowing levels and developmental stages. Karger.
Ellis, A. B. (2007). The influence of reasoning with emergent quantities on students’ generaliza-
tions. Cognition and Instruction, 25(4), 439–478.
org/10.1016/j.jmathb.2007.03.002
Cognition and Instruction, 28(4), 1–50. https://doi.org/10.1080/07370008.2010.511565
Hackenberg, A. J., & Lee, M. Y. (2015). Relationships between students’ fractional knowledge and
equation writing. Journal for Research in Mathematics Education, 46(2), 196–243. https://doi.
Hackenberg, A. J., & Sevinc, S. (2021). A boundary of the second multiplicative concept:
The case of Milo. Educational Studies in Mathematics, 109, 177. https://doi.org/10.1007/
s10649-021-10083-8
Hackenberg, A. J., & Sevinc, S. (2022). Middle school students’ construction of reciprocal reason-
ing with unknowns. The Journal of Mathematical Behavior, 65, 100929.
Hackenberg, A. J., & Tillema, E. S. (2009). Students’ whole number multiplicative concepts:
A critical constructive resource for fraction composition schemes. Journal of Mathematical
Behavior, 28(1), 1–18. https://doi.org/10.1016/j.jmathb.2009.04.004
Hackenberg, A. J., Jones, R., Eker, A., & Creager, M. (2017). “Approximate” multiplicative rela-
tionships between quantitative unknowns. Journal of Mathematical Behavior, 48, 38–61.
Hackenberg, A. J., Aydeniz, F., & Jones, R. (2021a). Middle school students’ construction of quan-
titative unknowns. The Journal of Mathematical Behavior, 61, 100832.
Hackenberg, A. J., Creager, M., & Eker, A. (2021b). Teaching practices for differentiating
mathematics instruction for middle school students. Mathematical Thinking and Learning,
23(2), 95–124.
Hackenberg, A. J., Walsh, P. A., & Valero, J. R. (2021c). A case of units coordination stage change
in middle school. In D. Olanoff., K. Johnson, & S. Spitzer (Eds.), Proceedings of the Forty-Third
Annual Meeting of the North American Chapter of the International Group for the Psychology
of Mathematics Education (PME-NA 43), pp. 1281–1286, Philadelphia, October 14–17.
Hackenberg, A. J., Borowksi, R., & Aydeniz, F. (2023). Middle school students at three stages of
units coordination learn to make same speeds. Journal of Mathematical Behavior. https://doi.
org/10.1016/j.jmathb.2023.101085
Hegedus, A. (2018). Evaluating the relationships between poverty and school performance.
NWEA Research. NWEA. https://proxyiub.uits.iu.edu/login?url=https://search.ebscohost.
com/login.aspx?direct=true&db=eric&AN=ED593828&site=eds-live&scope=site
Izsák, A., Jacobson, E., de Araujo, Z., & Orrill, C. H. (2012). Measuring mathematical knowledge
for teaching fractions with drawn quantities. Journal for Research in Mathematics Education,
43(4), 391–427.
Kaput, J. J. (2008). What is algebra? What is algebraic reasoning? In J. J. Kaput, D. W. Carraher,
& M. L. Blanton (Eds.), Algebra in the early grades (pp. 5–17). Lawrence Erlbaum. https://doi.
org/10.4324/9781315097435
Lee, S. J., & Shin, J. (2021). Students’ proportional reasoning and units coordination [Manuscript
submitted for publication].
Ministry of National Education (MNE). (2018). Elementary and middle school mathematics pro-
gram: Grades 1, 2, 3, 4, 5, 6, 7, and 8. Ministry of National Education.
National Governors’ Association Center for Best Practices & Council of Chief State School
Officers. (2010). Common core state standards. Authors. http://www.corestandards.org/wp-
content/uploads/Math_Standards.pdf
Norton, A., & Boyce, S. (2015). Provoking the construction of a structure for coordinating n+ 1
levels of units. The Journal of Mathematical Behavior, 40, 211–232.
Norton, A., Boyce, S., Ulrich, C., & Phillips, N. (2015). Students’ units coordination activity: A
cross-sectional analysis. Journal of Mathematical Behavior, 39, 51–66.
Olive, J. (1999). From fractions to rational numbers of arithmetic: A reorganization hypothesis.
Mathematical Thinking and Learning, 1(4), 279–314.
Olive, J., & Caglayan, G. (2008). Learners’ difficulties with quantitative units in algebraic word
problems and the teacher’s interpretation of those difficulties. International Journal of Science
and Mathematics Education, 6, 269–292.
Olive, J., & Steffe, L. P. (2010a). The construction of fraction schemes using the generalized num-
ber sequence. In L. P. Steffe & J. Steffe (Eds.), Children’s fractional knowledge (pp. 277–314).
Springer.
Olive, J., & Steffe, L. P. (2010b). The partitive, the iterative, and the unit composition schemes. In
L. P. Steffe & J. Steffe (Eds.), Children’s fractional knowledge (pp. 171–223). Springer.
Piaget, J. (1965/1949). The child’s conception of number. Norton.
Shin, J., Lee, S. J., & Steffe, L. P. (2020). Problem solving activities of two middle school stu-
dents with distinct levels of units coordination. Journal of Mathematical Behavior, 59, 100793.
https://doi.org/10.1016/j.jmathb.2020.100793
Smith, J. P., & Thompson, P. W. (2008). Quantitative reasoning and the development of algebraic
reasoning. In J. J. Kaput, D. W. Carraher, & M. L. Blanton (Eds.), Algebra in the early grades
(pp. 95–132). Lawrence Erlbaum.
Steffe, L. P. (1994). Children’s construction of meaning for arithmetical words: A curricu-
lum problem. In D. Tirosh (Ed.), Implicit and explicit knowledge: An educational approach
(pp. 131–168). Ablex.
Steffe, L. P. (2009a). The construction and use of multiplying and dividing schemes: Jason and
Patricia [Unpublished manuscript]. Department of Mathematics and Science Education, The
University of Georgia.
Steffe, L. P. (2009b). The construction and use of multiplying and dividing schemes: Rebecca
and Tanya [Unpublished manuscript]. Department of Mathematics and Science Education, The
Steffe, L. P. (2010a). Articulation of the reorganization hypothesis. In L. P. Steffe & J. Olive (Eds.),
Steffe, L. P. (2010b). Operations that produce numerical counting schemes. In L. P. Steffe &
J. Olive (Eds.), Children’s fractional knowledge (pp. 27–47). Springer.
Steffe, L. P. (2010c). The partitive and the part-whole schemes. In L. P. Steffe & J. Olive (Eds.),
Steffe, L. P. (2017). Psychology in mathematics education: Past, present, and future. In E. Galindo
& J. Newton (Eds.), Proceedings of the Thirty-ninth Annual Meeting of the North American
Chapter of the International Group for the Psychology of Mathematics Education (PME-NA).
Indianapolis. To appear at this url: http://www.pmena.org/proceedings/
Springer-Verlag.
reaction to Simon. Journal for Research in Mathematics Education, 26(2), 146–159.
Steffe, L. P., & Olive, J. (2010). Children’s fractional knowledge. Springer. https://doi.
org/10.1007/978-1-4419-0591-8
Steffe, L. P., Cobb, P., & Richards, J. (1983a). An analysis of children’s creation of countable items
while counting. In L. P. Steffe, E. von Glasersfeld, J. Richards, & P. Cobb (Eds.), Children’s
counting types: Philosophy, theory, and application (pp. 45–76). Praeger Scientific.
Steffe, L. P., Glasersfeld, E. von, Richards, J., & Cobb, P. (1983b). Children’s counting types:
Philosophy, theory, and application. : Praeger Scientific.
Steffe, L. P., Liss, D. R., II, & Lee, H. Y. (2014). On the operations that generate intensive quantity.
In L. P. Steffe, K. C. Moore, L. L. Hatfield, & S. Belbase (Eds.), Epistemic algebraic students:
Emerging models of students’ algebraic knowing (Vol. 4, pp. 49–79). University of Wyoming.
Stephens, A. C., Ellis, A., Blanton, M. L., & Brizuela, B. (2017). Algebraic thinking in the elemen-
tary and middle grades. In J. Cai (Ed.), Compendium for research in mathematics education
Tate, W. F. (1997). Race-ethnicity, SES, gender, and language proficiency trends in mathematics
achievement: An update. Journal for Research in Mathematics Education, 28(6), 652–679.
Thompson, P. W. (2011). New perspectives and directions for collaborative research in mathemat-
ics education. In L. L. Hatfield, S. Chamberlain, & S. Belbase (Eds.), New perspectives and
directions for collaborative research in mathematics education (pp. 33–57). University of
Wyoming.
Thompson, P. W. (2013). In the absence of meaning…. In K. Leatham (Ed.), Vital directions for
mathematics education research (pp. 57–93). Springer.
W. G. Martin, & D. Schifter (Eds.), Research companion to the principles and standards for
school mathematics (pp. 95–113). National Council of Teachers of Mathematics.
Tillema, E. S. (2013). A power meaning of multiplication: Three eighth graders’ solutions of
Cartesian product problems. Journal of Mathematical Behavior, 32, 331–352. https://doi.
org/10.1016/j.jmathb.2013.03.006
Tillema, E. S. (2014). Students’ coordination of lower and higher dimensional units in the con-
text of constructing and evaluating sums of consecutive whole numbers. The Journal of
representation of these problems using arrays. Journal of Mathematical Behavior, 52, 1–20.
Tillema, E. S. (2020). Students’ solution of arrangement problems and their connection to
Cartesian product problems. Mathematical Thinking and Learning, 22, 23–55. https://doi.org/1
0.1080/10986065.2019.1608618
Tillema, E. S., & Burch, L. J. (2022). Using combinatorics problems to support secondary teachers
understanding of algebraic structure. Zentralblatt für Didaktik der Mathematik, 54, 777–793.
https://doi.org/10.1007/s11858-022-01359-1
Ulrich, C. (2012). Additive relationships and directed quantities (Unpublished doctoral disserta-
tion). University of Georgia.
Ulrich, C. (2016a). Stages in constructing and coordinating units additively and multiplicatively
Ulrich, C. (2016b). The tacitly nested number sequence in sixth grade: The case of Adam. Journal
Ulrich, C., & Wilkins, J. L. M. (2017). Using written work to investigate stages in sixth-grade
students’ construction and coordination of units. International Journal of STEM Education, 4,
23. https://doi.org/10.1186/s40594-017-0085-0
von Glasersfeld, E. (1995). Radical constructivism: A way of knowing and learning. Routledge
Falmer. https://doi.org/10.4324/9780203454220
von Glasersfeld, E., & Kelley, M. F. (1982). On the concepts of period, phase, stage, and level.
Human Development, 25(2), 152–160.
Zwanch, K., & Wilkins, J. L. (2021). Releasing the conceptual spring to construct multiplicative
reasoning. Educational Studies in Mathematics, 106(1), 151–170.
Chapter 12
Modeling Quantitative and Covariational
Reasoning
Steven Boyce
Modeling Quantitative and Covariational Reasoning
Instead of de-constructing adult reasoners’ knowledge, Piaget’s (1970a) genetic

approach prioritizes modeling how reasoning develops. Yet, such modeling of math-
ematical reasoning is still the product of the adult reasoner, whose knowledge
frames and continues to develop via its use. Steffe and Wiegel (1996) refer to mod-
eling of others’ mathematics as second-order models. It is in the spirit of second-
order modeling that this chapter is organized. I reflect on my own experiences and
ways of understanding how quantitative and covariational reasoning are described
in the research literature, with particular consideration to how I imagine someone
new to the field of mathematics education research might interpret the language and
terminology. We first focus on contrasting meanings for “quantitative” and “covari-
ational” from everyday meanings.
Meanings for “Quantitative” and “Covariational”
It is important, when learning about constructs for modeling the development of

quantitative or covariational reasoning, to first reflect on one’s initial thinking about
the terminology. I invite the reader to reflect on their own thinking about what quan-
titative means or what it might mean to an educational researcher first encountering
these constructs by considering a problem that arises in the typology of data arising
S. Boyce (*)
Fariborz Maseeh Department of Mathematics and Statistics, Portland State University,
Portland, OR, USA
e-mail: sboyce@pdx.edu
https://doi.org/10.1007/978-3-031-47386-9_12
414 S. Boyce
in quantitative (statistical) methods of social science. The Likert scale (e.g., 1 for
strongly disagree, 3 for neutral, and 5 for strongly agree) is ubiquitous on survey
questionnaires, but there is no consensus as to how responses on items should be
interpreted (Pimentel, 2019). For instance, is the distinction between “strongly dis-
agree” and “neutral” equivalent to the distinction between “neutral” and “strongly
agree?” What about the interpretations of summing results; is a 1 and a 5 equivalent
to two 3s? Such questions depend on researchers’ models for incorporating numeri-
cal structure in modeling phenomena. That is, they regard researchers’ quantitative
reasoning.
When the term quantitative is used in everyday social science, it excludes nomi-
native scales that merely refer to naming categories of phenomena. At a minimum,
“quantitative” requires attending to a meaningful “ordering,” for instance, from
least to greatest or from first to last. The Likert scale includes this ordinal aspect. A
more restrictive “quantitative” meaning includes assigning a unit that permits mul-
tiplicative comparisons, as well as meaningful interpretations of fractional values (a
ratio scale). Quantities such as time (in seconds) or length (in meters) are measured
on ratio scales. Between the ordinal and ratio scales is associating phenomena with
a unit that permits additive but not multiplicative comparisons (an interval scale).
Temperature is an example of a quantity often measured using an interval scale,
such as Celsius. An ambiguity regarding the Likert scale is whether one can inter-
pret the intervals between Likert values (across items or across respondents) as
uniform, the way we consider Celsius degrees when measuring temperature.
Ironically, researchers of quantitative reasoning typically engage in qualitative
methods.
At earlier levels of psychological development, researchers focus on describing
children’s actions and operations when they are reasoning about phenomena for
which adults construct ratio scale interpretations (such as time, volume, area, or
length). The focus of quantitative reasoning research in the early stages of develop-
ment is identifying how learners construct such quantities in tandem with their con-
struction of number and how they transfer reasoning about one quantity to reasoning
about another quantity (Smith & Barrett, 2017).
Later in development, quantitative reasoning research focuses on describing the
ways students reason about a quantity changing (variational reasoning), including
their reasoning about how quantities change together (covariational reasoning). The
term covariational in mathematics education differs from its use in applied statistics.
In statistics, the term covariation refers to correlated variation, whereas in mathe-
matics education research, the term covariation regards corresponding variations.
This is perhaps a subtle but important distinction.
I encourage the reader to consider contrasts in ways of reasoning about a linear
relationship between two quantities represented in a graphical context, depicted in
Fig. 12.1. Figure 12.1a depicts a line of best fit for bivariate numerical data. In
accordance with the statistical meaning of “covariation,” this is the line that repre-
sents the overall trend in the relationships between corresponding numerical values.
Figure 12.1b has dashed line segments intended to represent reasoning about the
12 Modeling Quantitative and Covariational Reasoning 415
Fig. 12.1 Representing statistical covariation (a), covarying quantities (b), and corresponding
values (c)
values of two quantities changing uniformly, between 1 and 2 in the quantity repre-
sented on the horizontal axis and between 1 and 3 in the quantity represented on the
vertical axis (between point A and point B), and similarly between point B and point
C. Figure 12.1c depicts those same three points, but with additional dots, each rep-
resenting a pair of corresponding values, intended to suggest reasoning about filling
in the space between the three points to form a line. The dots and dashes in
Fig. 12.1b, c are intended to impart another consideration important for calculus
learning: Whether and how variables, and changes in variables, are conceived as
varying continuously or discretely.
There are many other considerations for covariational reasoning, including rea-
soning about nonlinear relationships and reasoning in non-graphical contexts (or
with other coordinate systems). Instead of analyzing more sophisticated concep-
tions of covariation, or of quantity, by reducing them to constituent components and
attempting to outline a trajectory of these foundational conceptions that make up
researchers’ knowledge, the genetic epistemologist models distinctions (and com-
monalities) in learners’ conceptual structures as they are developing. It is for this
reason that for modeling the development of quantitative and covariational reason-
ing, consideration of how concepts of motion, space, and numerosity is constructed,
and of particular relevance, as compared to the distinctions in constructs as applied
by quantitative methodologists. It is also why additional terms are useful to describe
learners’ reasoning.
416 S. Boyce
Gross, Intensive, and Extensive Quantities
Piaget found that children tend to reason about different quantities (e.g., area,
length, volume, and time) in similar ways at different stages of their cognitive devel-
opment (e.g., Piaget, 1970b). Piaget identified three stages in children’s reasoning
about measure, which he called gross quantity, intensive quantity, and extensive
quantity. Piaget found these to develop sequentially across a variety of contexts
(see, e.g., Voyat, 1982, for an extensive list of Piaget’s tasks and concise descrip-
tions of distinctions in children’s reasoning).
Reasoning at the gross quantity stage includes aspects of nominative and ordinal
scales in that it is not yet associated with counting actions, but there is an awareness
of size and of distinctions in size. The sources of the distinctions are implicit in the
child’s actions for perceiving the size of an “object”; the “larger” of two objects is
the one that has a further (later) end. For instance, a child might conceive of a child
standing on a chair as being “taller” than an adult standing on the ground. Children
who can manipulate rulers can be taught to align starting points and compare end-
points to compare lengths directly while still reasoning about gross quantity (Barrett
et al., 2017). It is important for the researcher to recognize that the attribute of an
object a researcher conceives of in a situation might not be the same attribute the
student attends to. For instance, when asked which of two jars set upon a table has
more liquid, a child reasoning with gross quantity might state that an object filled to
a greater height has more, irrespective of the jars’ diameters (Piaget, 1954,
Siegler, 2016).
For Piaget, the intensive quantity stage is the beginning of associating a structure
to assign a magnitude, or numerical value, to a quantity. Reasoning about an inten-
sive quantity involves reflecting on the process of focusing attention on the begin-
ning of an object to focusing attention on its end. This process is coordinated with a
counting sequence, that is, with counting 1, 2, 3, … sub-extents (pieces) from start
to finish. Children at this stage can then compare two objects’ sizes by comparing
counts, but they do not necessarily attend to uniformity in the sizes of each piece or
of their contiguity. At the intensive quantity stage, a child might reason that object
A is larger than object B if they conceive of object A including more pieces (even if
object B’s extent is larger, see Fig. 12.2).
When a child’s reasoning about a quantity includes a necessity of uniformity of
sub-extents (units), Piaget would say they had constructed the quantity as extensive.
The uniformity of units allows the child to conceptualize comparisons between
objects, as the extent of the difference between them. Reasoning about an extensive
quantity is akin to thinking of it on (at least) an interval scale. It requires conceptu-
ally extending the units for measuring an object outside of itself, both in the sense
Fig. 12.2 Object A may be larger than object B if reasoning about length as an intensive quantity
Disembedded Unit
Uniformly sized sub-extents (units)
Fig. 12.3 Necessity of uniformity in units and disembedding of units is hallmark of extensive
quantity
of being “beyond” an ending and in the sense of disembedding a unit of measure

from within without modifying the size (see Fig. 12.3).
The necessity of the uniformity is a critical aspect of the development of exten-
sive quantity—it distinguishes the quantitative unit structure from being figurative,
perceptual, or concrete and instead as operational. Piaget did not elaborate further
on extensive quantity that we would associate with understanding a quantity as mea-
sured on a ratio scale. He instead framed multiplicative invariance, consistent within
a ratio scale, as a hallmark of developing proportional reasoning (Piaget &
Inhelder, 1976).
It may seem strange to describe the construction of extensive quantity as follow-
ing the construction of intensive quantity. Piaget described the development of
gross, intensive, and extensive quantities as stages in children’s construction of
extensive quantities. So that the term intensive quantity can apply generally to situ-
ations that scientists may conceive as extensive, intensive quantity in this chapter
should be understood as a quantity whose measure is, for the person, an intensity. It
does not necessarily refer to a ratio of constituent extensive quantities, such as den-
sity as a ratio of volume to mass, or of velocity as a ratio of displacement to time
(Johnson, 2015; Kertil et al., 2019; Simon & Placa, 2012: Steffe et al., 2014;
Thompson, 1994).
The Meaning of Quantity
Piaget’s articulation of stages in the ways children reason about measurement led to
the need for new terminology to describe quantities that children reason about.
Seminal contributions connecting Piaget’s genetic epistemology to quantitative and
covariational reasoning stem from P. W. Thompson and A. G. Thompson’s models
of middle-grade students’ images of ratios and rates (e.g., Thompson, 1993, 1994;
Thompson & Thompson, 1992). Their articulation of speed-distances (Thompson
& Thompson, 1992) is consistent with core principles and utility of genetic episte-
mology in mathematics education research:
418 S. Boyce
1. The researcher is disposed to modeling learners’ mathematical constructs that

are stable and consistent for the learner, even those that may be inconsistent with
researchers’ ways of reasoning about a task.
2. The researcher attends to and seeks out models for ways of operating that could
hypothetically serve as precursors to cognitive processes that could lead to the
construction of more powerful ways of operating.
3. Research questions center on models of students’ ways of understanding.
Identifying and refining epistemic models, and their implications, are paramount.
Before delving further into the construct of speed-distance, I discuss the meaning of
quantity. Thompson (1994, p. 7) described quantity as follows:
A quantity is schematic: It is composed of an object, a quality of the object, an appropriate
unit [or potential units], and a process by which to assign a numerical value to the quality.
Variations in people’s conceptions of a quantity occur in correspondence with variations in
level of development of components within their schemes.
This characterization of quantity is broad in several ways. While it does stipulate

a process for assigning numerical value and its association with a unit, these can be
interpreted differently, in accordance with models of persons’ schemes for number
and for unit. For instance, it is left to interpretation whether the “numerical value,”
or magnitude, of a quantity must be of an interval or ratio type (or of some other
type) explicitly involving units, or if could be an ordinal type.
The term “object” in the definition may or may not be perceptual, as it could
refer to a conceptual object. For instance, in statistics, a sample distribution could
serve as an “object” and (statistical) variation as its quality. This quality could be
associated with variance in order to assign a magnitude, with the value of “0” for the
variance corresponding to no variation and each unit of variation corresponding
with the square of the unit of measure in the distribution. Such aspects depend on
the goals of the person’s reasoning about the situation, rather than just a description
of their available ways of operating. For instance, if a person experiences a need to
compare variations with different means, they might accommodate their scheme for
(statistical) variation to incorporate standard deviation (the square root of the sum
of squared differences from the mean).
There is also a question of whether each of the components of the description of
quantitative reasoning must be explicit in the modeling of the person’s thinking in
order for their reasoning to be considered quantitative. For instance, to what extent
does this characterization exclude “purely” arithmetical schemes, that is, reasoning
about objects that, via processes of reflective abstractions, are no longer conceptual-
ized as associated with particular qualities of objects but are conceptual objects unto
themselves (Steffe & Cobb, 1988)? One way of clarifying this is to consider further
distinctions between arithmetic reasoning and quantitative reasoning. Another is to
distinguish conceiving of quantitative objects from more general reasoning about
quantities. The next section elaborates on these distinctions.
rithmetic Reasoning, Quantitative Reasoning,

A
and Reasoning Quantitatively
The previous section briefly brought up the notion of difference, in the context of
“squared difference” from a mean. Consideration of what it means for a difference
to be a quantity is useful to distinguish between arithmetic reasoning, quantitative
reasoning, and quantitative objects. Ulrich (2012) elaborates on the role of object in
distinguishing between quantitative and arithmetical reasoning in her studies of
middle grades’ students reasoning about signed quantities. The goal of her study
was to engender in students a conception of a “directed difference” as a quantity that
could be acted upon rather than merely the result of an action. She writes,
In the elementary grades, addition is often characterized as increasing the amount in a set,
and subtraction is characterized as taking away items from a set or decreasing the amount
in a set. Hence, for some students, addition and subtraction are actions as opposed to indica-
tions of a sum or difference structure. This can even be true for students who have a fairly
sophisticated understanding of how quantities in sums and differences are related. For
example, all of my students were aware that 24-7 = 17 implies that 7 and 17 are both subsets
of 24 and that, taken together, give a set of size 24. Hence, they understood the necessity of
why 24-7=17 implies 7+17=24 and 24 = 17 + 7. However, even for my participants, the
notation of addition and subtraction does not necessarily indicate the presence of these sum
and difference relationships. Addition and subtraction were still actions that needed to be
carried out (p. 134).
Notice how Ulrich (2012) refers to her students’ understandings of how quanti-
ties within sums and differences are related, despite writing that they were not rea-
soning about the differences (or sums) as quantities. That is, she could ascribe to her
students a units coordinating structure for whole number units (units of 7 and units
of 17 within a unit of 24). For Ulrich’s (2012) students, each of the numerosities 7,
17, and 24 could be considered quantities. But the meaning of the expression “24-7”
was not yet a quantitative object for these students. When a student reasons about
an expression, such as 24-7, they may be reasoning about the numerical result (17)
without reasoning about the quantity of “difference.” For Ulrich’s (2012) students,
the “difference” was an arithmetical operation because the product of reasoning was
a number, but the difference would be considered an action on quantities rather than
a quantitative operation because the product of reasoning was not itself conceived
as a quantity.
The trajectory for constructing an (extensive) quantitative object is to attend to it
as a gross quantity, an intensive quantity, and (if possible) an extensive quantity. The
onset of reasoning about a quantity is attending to a size of an attribute of an object.
When reasoning about a signed difference, if a student begins attending to “got big-
ger” and “got smaller,” they are reasoning about the difference as a gross quantita-
tive object. Continuing to attend to the difference and reflect on their activity can
lead to their reasoning about difference as an intensive and finally extensive quan-
tity. Upon this construction, a student assimilates a difference as an extensive quan-
tity simultaneously with attending to it as a size.
420 S. Boyce
Researchers should not attribute quantitative reasoning to students merely from

students’ use of notation or from their utterances because similar observed behav-
iors may indicate arithmetical reasoning for one student and quantitative reasoning
for another. Moreover, fostering students’ engagement in quantitative reasoning in
a situation can be hindered if the students have previously established strategies for
reasoning arithmetically in that situation. When inferring that a student engaged in
quantitative reasoning, the researcher should identify what quantities, in their mod-
eling of the students’ mathematics, the student reasoned about, including relation-
ships between quantities that they reasoned about (see Moore & Carlson, 2012).
When claiming that a student did not engage in quantitative reasoning, the researcher
should identify how the student was not engaging in reasoning about quantities in
the situation, in particular if they model that the student was instead reasoning
arithmetically.
Reasoning Quantitatively
Distinguishing between “reasoning about quantities,” “quantitative reasoning,” and

“reasoning quantitatively” may be useful at times. Ulrich’s (2012) students “rea-
soned about quantities” in that they acted on quantities. But they engaged in arith-
metical reasoning about the difference. If they had reasoned about the difference as
an object, we could describe it as “quantitative reasoning” because it is regarding
their reasoning about the quantity of difference. Both “reasoning about quantities”
and “quantitative reasoning” are ways of understanding (Harel, 2008). Models of
quantitative ways of understanding include inferences of individuals’ imagery of
objects’ attributes and their extents, as well as their ways of coordinating these
images with numerical conceptions and arithmetic operations.
We can reserve “reasoning quantitatively” to a broader description of an indi-
vidual’s ways of thinking (Harel, 2008)—their tendency or inclination to engage in
quantitative reasoning, to enact their schemes involving quantities across situations
(Tallman & Frank, 2020).
A prominent characteristic of reasoning quantitatively is that numbers and numeric rela-
tionships are of secondary importance, and do not enter in the primary analysis of a situa-
tion. … Quantities, when measured, have numerical value, but we need not measure them
or know their measures to reason about them. You can think of your height, another person’s
height, and the amount by which one of you is taller than the other without having to know
the actual values. (Thompson, 1993, p. 1)
In this formulation, reasoning quantitatively shares aspects with algebraic rea-

soning, as it can involve unknown (or yet-to-be-known) numerical values. However,
quantities can also be considered “more concrete than numbers” (Thompson, 1993,
p. 1). This is because abstract arithmetic units and arithmetic operations are con-
structed via reflection on reasoning about numerosities of some things. Moreover,
when reasoning about continuous quantities, about quantities that have non-discrete
measures, or varying quantities, the accompanying arithmetical units may be
themselves reorganized and coordinated with prior meanings to generalize to new

quantitative contexts. In this way, quantitative and arithmetic reasoning can be con-
sidered to be co-constructed via reasoning quantitatively. Unfortunately, schooling
often impedes such development of reasoning quantitatively, and students instead
learn to reason arithmetically, which can impede students’ generalizations (e.g.,
Ellis, 2007; Stacey & MacGregor, 1997).
Example: Relating Arithmetical and Quantitative Reasoning
To assess students’ quantitative reasoning in a clinical interview setting, it may be
useful to present non-routine but accessible tasks. Particularly for older students
who may have well-established “non-quantitative” ways of thinking about quanti-
ties, responses can reveal their available ways of operating that may be seldom
exhibited in more routine settings. In the excerpt below, I illustrate and analyze the
reasoning of a calculus student, Maura, who engaged with a non-standard measure-
ment task during a video-recorded clinical interview, involving fictitious units of
measure (for further discussion, see Byerley et al., 2019). Maura was given the fol-
lowing prompt:
On an alien planet they could use cacks and lorns to measure length. The length
of one cack is 9/5 times as large as the length of one lorn. Walnut the cat is 89.4
lorns long. What is Walnut’s length in cacks?”
Maura verbalized much of her thinking, both auditorily and in writing.
So nine over five lorns is equal to one cack. [Writes, “9/5 lorns = 1 cack”].
And nine-fifths [writes “9/5 = 1 4 5 ”] is almost two. It’s almost twice as long.
And Walnut is 89.4 lorns long. What is Walnut’s length in cacks.
So [19 second pause without speaking or writing].
[Writes, “89.5 lorns X (1 cack)/(9/5 lorns)”]
[Returns attention to written prompt] So if 1 cack is 9/5 lorns—so it’s approximately two
[writes “~2 below 9/5 in the prompt”].
So if one cack is twice as large, as the length of one lorn. So two lorns equals
a cack. Right. So then you have 89.4 divided by 9/5, which is equal to 89.5 over
one times five over nine.
[Completes the written statement above, writing “= 89.4 / (9/5) = 89.4 * (5/9)”]
[Determines the value of 89.4 times 5 using a standard USA multiplication algorithm].
It’s 447/9. [Writes, “447/9 cacks long”]
This excerpt illustrates relationships between Maura’s arithmetical and quantita-
tive reasoning as she engages with the task. Her first written statement seems to be
algebraic rather than arithmetical or quantitative, a form of reversible reasoning, “A
is X times as large as B means X As is equal to one B” (Hackenberg, 2010;
Hackenberg et al., 2021). It appears she may have been momentarily perturbed by
the fractional value for X. However, her writing of 9/5 lorns = 1 cack, 9/5 = 1 and
4 , and 9/5 ~ 2 suggests that she anticipates resolving this perturbation by trans-
5
forming the relation into an approximate relation involving whole numbers, “two
lorns equals a cack,” for which she had available an image of a quantitative relation-
ship between two lengths. That is, transformation into a whole number relation
seemed to be part of an estimation strategy coming into the situation.
422 S. Boyce
Given the 19-second pause, she seems uncertain of whether to multiply 89.4
lorns by ((9/5 cacks) / 1 lorn) or by (1 lorn / (9/5 cacks)). It is unclear how she rea-
sons that (1 cack) / (9/5 lorns) is the correct form of the ratio. She is satisfied by
experiencing success in interpreting the ratio as consistent with her statement, “So
two lorns equals a cack. Right.” She re-interprets her intermediate goal as deciding
which of (1 cack) / (9/5 lorns) or (1 lorn) / (9/5 cacks) is consistent with the (approx-
imate) equivalence. She indicates proficiency with arithmetic procedures with frac-
tions and decimals (such as invert and multiply for fraction division). But there is
nothing in her utterances or her observable activity that would suggest that she
would have experienced the result of multiplying 89.4 by 9/5 as incorrect. Her leav-
ing the final answer as 447/9 cacks long suggests that including units is important
and that leaving the result as an improper fraction, rather than converting it to a
mixed number, is normative for her.
What quantitative objects, if any, might we attribute to Maura based on this anal-
ysis? Cacks and lorns are each described as lengths in both the prompt and her writ-
ten response [447/9 cacks long]. Both 89.4 and 447/9 seem to be magnitudes of
length, although it is difficult to say what 447/9 means to her. But we might suspect
that the ratio (1 cack / 9/5 lorns) is not yet a quantitative object for her. This is indi-
cated by her way of resolving her uncertainty [indicated the long pause] of whether
the ratio was correct or not. She did not evoke a conception of the (1 cack / 9/5
lorns) ratio as an object describing a scaling between two quantities’ magnitudes,
just as Ulrich’s (2015) students did not indicate constructing a difference between
two integers as a quantitative object. Her ratio either corresponded to “9/5 lorns = 1
cack” or it did not, but there was no indication of her ratio representing a size.
We see partial confirmation for these inferences exhibited in the follow-up task
within the interview, in which Maura was tasked to draw a ruler that has both cacks
and lorns and is at least 6 cacks long. The task also afforded opportunities to distin-
guish quantitative and non-quantitative aspects of Maura’s reasoning. Maura con-
structed her ruler by beginning with drawing an approximate rectangle, upon which
she marked six cacks on the top by first marking an approximate halfway point, then
drawing marks corresponding to one and two, and then the marks corresponding to
four and five. This shows her anticipation of numerical reasoning, of partitioning six
into two groups of three. Next, she wrote numbers one through six at the top of the
rectangle and afterward labeled the top as “cacks” and the bottom as “lorns” (see
Fig. 12.4). Her reasoning about the extent of six cacks suggests that she is, at mini-
mum, reasoning about the extent of cacks as an intensive quantity (as on an ordinal
scale), and given her anticipation of creating equally sized units before labeling with
numerals, she provides an indication of reasoning about a cack as an extensive
quantity.
Maura next prepared to add marks to the bottom of the ruler. She reasoned aloud,
“So for every cack, you have two, almost two, lorns.” Her language suggests antici-
pation of re-presenting the sequence of cacks (“for every cack”) with a correspond-
ing sequence of lorns. She first drew a dashed line beneath her mark for “1” cack,
extending to the bottom of the rectangle. She next drew a mark to the right of that
point and labeled it as “2.” This suggests she is thinking of the ratio “almost two” as
Fig. 12.4 Rendering of Maura’s initial partitioning and labeling
Fig. 12.5 Rendering of Maura’s marking of the second, and then first, lorns
an intensive quantity by which to infer the sequentially of 1 cack and 2 lorns but is
not attending to the extent of the difference. Afterward, she drew a mark for 1 lorn
and labeled it (see Fig. 12.5).
Note that there was no indication of coordinating a fifth of one lorn and the
extent of the difference between the extent of one cack and the extent of two lorns.
Rather, the determination of the mark for two lorns seems to be the result of a gross
quantitative comparison, as the length past 1 cack for two lorns was smaller than
one-fifth of a lorn in her drawing. But there is an indication of reversibility in the
gross comparison: The two lorns should be a little bit bigger than one cack because
one cack is a little bit smaller than (“almost”) two lorns. This reversibility is indica-
tive of her reasoning about “almost two” as an intensive quantity. In contrast, the
mark for “1” lorn is placed approximately halfway between the left end of the rect-
angle and the mark for “2” lorns, a result of more precise partitioning of her length
of two lorns into two equal parts.
Next, Maura focused on marking additional lorns. She seemed perturbed at first
at how to proceed as she sat silent for approximately 10 seconds. I infer that she had
anticipated iterating beyond the second lorn, but because her mark for two lorns
extended past her mark for one cack, she did not know where to start. She appeared
to decide that acting on the difference between one cack and two lorns would be
productive, hovering her pen over that portion of the rectangle and verbalizing “the
one-fifth.” She began to iterate that size to the right of her mark, for two lorns, but
stopped after two iterations. I infer that she had been attempting to determine the
location of the mark for the third lorn, and had anticipated adding more “one-fifths,”
but then realized that she could iterate the size of one lorn instead to determine the
location of the third lorn. She iterated once more, marking the size of four lorns, and
then drew a dashed line connecting her two-cack mark and her four lorn mark, as
shown in Fig. 12.6.
She proceeded to iterate the approximate size from left to right, this time writing
the number of lorns before making the next iteration, ending at 12 lorns (Fig. 12.7).
Afterward, she wrote, in the space to the right of her ruler, “about doubled.”
424 S. Boyce
Fig. 12.6 Rendering of Maura’s marking of the third, and then fourth, lorns
Fig. 12.7 Rendering of Maura’s marking of the remaining lorns
Maura’s iteration of unit sizes of lorns was not very precise in the second half of
her ruler. But the non-uniformity of units in her drawing is not enough indication to
infer that she was not reasoning about the length of the bottom of her ruler as an
extensive quantity. I conjecture that her declining precision may have been partly
due to the interruption after each iteration to write the next numeral. This required
her to make the next mark after having written a numeral, as opposed to directly
following a previous mark. She did indicate an iteration of uniform size in her mark-
ings in the left half of the bottom of the ruler, where she also indicates a correspon-
dence between the length on the top and the length on the bottom. However, her
stopping the iteration at 12 (rather than iterating once more to mark a 13th lorn) and
later her writing “about doubled” to its right suggests that as part of the activity of
labeling numerals, she was anticipating making a final mark labeled as twelve
(“about double” six). Reasoning about finishing by making a mark at twelve may
have influenced her choice of smaller increments for marks 8, 9, 10, and 11. For this
reason, I would describe her placement of the markings in the right half of the ruler
as reasoning about the length as an intensive quantity rather than an extensive
quantity.
Of note, nine lorns was not equated with five cacks in her ruler. The interviewer
brought this issue to Maura’s attention. “Is there a place where some number of
cacks should be the same measure as some other number of lorns?” Here, the inter-
viewer was anticipating that Maura would interpret “some number” as an integer.
Maura replied,
Well if it’s nine-fifths, it probably should be equal at nine and five. I’m guessing.”
[Reviewing her previous writing.] The length of one cack is about twice as large as the
length of a lorn. [7 second pause] So I did do it backwards. The cack is the one that should
be bigger, or have like, the more numbers. The longer I look at this, the more confused I get.
The ruler displayed two extents that Maura conceived as equivalent, in cacks (the
top of the ruler) and lorns (on the bottom of the ruler). She “guessed” that the two
rulers should have corresponding values at 5 and 9, but this was the product of
empirical reasoning, based on her recollection of a pattern holding in similar situa-

tions rather than a logical necessity. She did not indicate reasoning that the extent of
five cacks should be equivalent to the extent of nine lorns because a cack is 9/5
times as large as a lorn. She was reasoning that 5 and 9 should match up because
that is the fraction that appears in the task. Since they did not correspond in the
figure she created, she inferred that she had made a mistake and had the “which is
larger” relation backward (which would imply that 9 cacks should correspond with
5 lorns).
Maura was perturbed by comparing the extents on each side of the ruler she had
made (“longer I look at this, the more confused I get”). The interviewer asked Maura
to revisit what she had previously written about the initial task, and she rewrote that
one cack is approximately two lorns. She then stated, “So the cack is bigger, in size,
but the lorn is bigger, numerically. Like, you have to have more of the lorns to fit
into the cack.” This reasoning is in accordance with her prior reasoning when first
creating the ruler. The perturbation was resolved, as she decided her initial response
was correct. The relationship between the five and the nine not corresponding on the
ruler was left as it “not being to scale exactly.”
One thing that stands out from Maura’s activity with the ruler is that she did not
iterate cacks. She drew an arbitrary length for the ruler, marked its extent as six
cacks, and partitioned to form, and later labeled, the tick marks for the remaining
integer values of cacks. She did iterate lorns, but only units of one lorn. Thus, I
interpret her saying, “You have to have more of the lorns to fit into the cack,” as
referring to a gross comparison between the number associated with the extent of
cacks in her ruler (six) and the number associated with the extent of lorns in her
ruler (about twelve) rather than thinking of “more of the lorns” as referring to 9/5
lorns fitting within a single cack.
Throughout, there was no evidence that Maura was reasoning with 9/5 as a num-
ber other than “almost two.” She did not evidence an understanding of a fraction as
an operator (Steffe & Olive, 2010). Her quantitative reasoning appeared to be con-
strained by her fractional schemes, perhaps, in particular, her meanings for improper
fractions, as she did not reason quantitatively with magnitudes that were not whole
numbers. Yet, in a previous interview task, Maura successfully created the size of a
whole when given a rectangular bar stated as the size of 7/3. This suggested that she
had constructed an iterative fraction scheme, which is an understanding of any frac-
tion as an iteration of a unit fraction, including fractions larger than a whole (see
Steffe & Olive, 2010). If that were the case, her meaning for 9/5 should have
included an understanding of the fraction as 9 iterations of one-fifth, five of which
make the whole. Instead, her conception of 9/5 in the ruler task seemed to be limited
to be of the complement of “one-fifth” within two. Maura may have understood a
fraction such as 9/5 as a measure, an extent resulting from iterating fractional units,
but only with a figurative referent available (Tzur & Simon, 2004).
426 S. Boyce
Modeling Students’ Images of Speed
In Thompson and Thompsons’s (1992) report on speed-distances, they model the concep-
tual structures of students who had not yet constructed multiplicative relationships between
speed, displacement, and time. They describe images of students’ reasoning about speed as
a quantity. The term image has many connotations. Though it often refers to figurative or
iconic mental pictures, I think the term, as used in describing reasoning, may be better
understood in a transformation sense (cf., Vinner & Hershkowtiz, 1980). That is, a person’s
image is the internal output of their mental operations (the output of their schemes), upon
which they can potentially (but not necessarily) act. The word “image” connotes a static
snapshot of reasoning, but each image consists of sub-images, some of which become
apparent simultaneously as part of a structure, and others sequentially, with some in the
foreground and some in the background. Images thus include both figurative and operative
aspects of thinking. The figurative aspect deals with the description of states, concerns
perception, and entails schemata; the operative aspect, on the other hand, provides under-
standing of the pathways from one state to another, referred to schemes of action and
abstraction. (Voyat, 1982, p. 5).
When a researcher attributes an evoked concept image to a student, they are not
necessarily attempting to describe the student’s iconic mental imagery (although
that is not precluded). Rather, they are modeling what is salient for the student in
their interpretation of the student’s reasoning—what the student attends to and how
they attend to it. The extent to which descriptions of operative thinking are included
is associated with the level of abstraction in the researchers’ modeling. The opera-
tive aspects of images include both organizations of actions within images of quan-
tities and organizations of actions connecting images of quantities.
When modeling students’ development of quantitative reasoning involving
abstract (numerical) objects, researchers tend to refer to individuals’ schemes; when
modeling students’ development of quantitative reasoning involving concrete
(quantitative) objects, researchers refer to students’ images (though they may also
describe schemes). One important contrast for quantitative images (or schemes) is
that the result of a scheme, or a component of an image, is not necessarily a quanti-
tative object itself.
Table 12.1 below describes four images of speed, adapted from Thompson and
Thompson (1992). Consider how successive images are distinguished: (1) What
quantities are explicit, and what are implicit, and (2) how do quantities vary?
One main aspect distinguishing the four images of speed described in Table 12.1
is the number of quantities that are modeled to be experienced as part of a reversible
structure (an operative aspect). In Image 1, accumulated time is not a quantitative
object. The individual’s counting of speed-lengths is not connected with accumulat-
ing elapsed time, which is indicated by the unit of time being omitted in the verbal-
ization of the image of speed-distance. In Image 2, elapsed time is a quantitative
object, and there is an association between a number of speed-distances and elapsed
time, but the relationship is not reversible (and not yet simultaneous). The child
becomes aware of the association of elapsed time as a product of their activity of
counting speed-lengths. In Image 3, a change in time is also a quantitative object,
simultaneous with a change in distance. The accumulation of speed-lengths
Table 12.1 Children’s images of speed

Image 1: Speed is Children begin thinking of speed as a distance—how far one goes (in one
a distance and time time unit). The unit of time is implicit in the stated speed-distance. This
is a ratio conception is often revealed in the way they express a speed in context—
“He is going 30 feet,” omitting the time unit (e.g., per second)—or
expressing an elaborated unit, “He goes 45 miles in an hour.” (Thompson
& Thompson, 1992)
Image 2: Speed- Students begin to differentiate speed so as to produce accumulated distance
distances accrue; and accumulated time as two separate quantities. They come to anticipate
time follows that traveling a number of speed-distances produces an identical number of
time-units. However, distance is still predominant in their awareness. That
is, thinking of a number of time-units does not automatically evoke an
image of an identical number of speed-distances
Image 3: [T]his is the first occasion wherein students reason about speed with the
Distance and time seeds of multiplicativity. The multiplicativity in their reasoning is not
accrue numerical. Rather, it is logical—distance and duration co-occur (Piaget,
simultaneously 1970a; Piaget & Inhelder, 1969). Numerically, distance and time still
accrue additively, as in previous images
Image 4: The first image of speed that always entails two levels of simultaneity: the
Speed as a rate simultaneity of accruals and the simultaneity of total accumulations during
accruals. The image of total accumulations of both distance and time
growing simultaneously with accruals of each is similar to the notion of
progressive integration in children’s counting (Steffe et al., 1983), but with
the exception that the integration is continuous and it simultaneously
involves two quantities
Adapted from Thompson and Thompson (1992, pp. 8–11)
coinciding with accumulation of time transforms via reflective abstraction—the

simultaneity goes from being anticipated upon completion of the accumulation of
speed-distances, to anticipated within each increment of accumulation. It is thus in
Image 4 that the simultaneity includes accumulations of distance and accumulations
of time, while maintaining the simultaneity of increments of each.
Notice that the description of variation of quantities in the four images begins
with an absence of variation, with a focus on the endpoint and the beginning point,
without attending to intermediate speed-distances. The development of images of
variation corresponds with the development of simultaneity in the incrementation of
speed (distances) and time. Thus, we see development in both figurative (what they
attend to) and operative (how they organize) aspects of images included in the
modeling.
Variational Reasoning
Piaget’s studies of children’s development of concepts of motion, time, and speed

are fundamental to the constructivist theories of variational reasoning. Within anal-
yses of variational and covariational reasoning, variation is best understood as a
“process of values changing,” and “continuity” is “pertaining to a continuum” rather
428 S. Boyce
than their more specific meanings in calculus, analysis, or statistics. Covariation is

best understood as “correspondence between variations” (cf., Confrey &
Smith, 1995).
The relationship between variational reasoning and the meaning of variable is
often a tension for researchers of quantitative and (co)variational reasoning. The
construct of variational reasoning regards images, concepts, and schemes for vari-
ables’ values changing, and the construct of covariational reasoning regards coor-
dination of related variations. For older students, who have been introduced to
function notation and definitions, meanings for variable as representing universal or
indeterminate yet static values can interfere with analyses of students’ use of lan-
guage when modeling their variational and covariational reasoning. For understand-
ing calculus, it is essential that students understand that variables’ values vary
continuously (Thompson & Harel, 2021).
We see the differences in understandings of variables reflected in the historical
development of the concept of variable and function. Historically, definitions of
functions beginning with Descartes in the seventeenth century referred to a variable
quantity as changing and never constant (Jahnke, 2003). Euler extended this defini-
tion to also include constant quantities, defining a variable to be “an indeterminate
or universal quantity, which includes within itself all completely determined val-
ues” (Euler, 1748, p. 7). For Euler, to say a function was continuous meant that it
was defined by a single formula. Thus, the notion of a “continuously varying quan-
tity” would have referred to a varying quantity whose values were determined by a
single analytic formula. The focus on continuous (or not) regarded the qualities of
the formula, and its operations, rather than the ways the input and output variables
changed. Euler’s definition was dominant until Cauchy prescribed his definition of
continuity in 1821 (Kleiner, 1989, p. 287), which is consistent with the modern defi-
nition of continuity commonly found in calculus and elementary analysis texts.
Models of Continuous Variation
Piaget did not elaborate much on the development of continuous reasoning

(Glasersfeld, 1996). One experiment regarding the construction of the continuum is
described by Voyat (1982, p. 60). In that experiment, five dots are placed in a row,
approximately an inch apart. A child is asked to fill in dots between the given dots
and then reason about what will happen if they continue the process indefinitely.
Voyat describes a child imagining that indefinitely continuing the action of filling in
points would result in a line segment. In other situations, similar reasoning was
attributed to systemic generation and testing of hypotheses in other contexts, such
as probabilistic reasoning (e.g., Piaget & Inhelder, 1976). In the case of filling in
points indefinitely, it is not only the imagining of gaplessness but also that the result
of gaplessness would result in transforming the discrete into the continuous. Such
description of reasoning about the continuum describes figurative thinking in that it
describes differences in states, but not the process, nor image, of the
transformation.
Glasersfeld (1996) offered a model of the conceptual construction of continuous
time that serves as a model of continuous lack of variation. He describes it as the
result of projecting connected and sequential aspects of an individual’s experiential
time onto objects that we perceive as maintaining an identity outside of our flow of
experience. He provides an illustrative example of perceiving the “exact same shirt”
on two consecutive days. Thinking of the shirt as maintaining its identity, its invari-
ance outside of our perceptions, requires projecting a continuous connection during
the gap between the two perceptions.
We have no continuous sensorimotorSensorimotor elements to warrant such a connection
and therefore have to construct it as a continuity outside our own field of experience. We
have to think of it as a link that is separate but, as it were, parallel to the succession of expe-
riences we have had in the interval between the two shirt perceptions. We remember the
succession of our actual experiences as continuous and having a sequential order, and we
can now project the patternPatterns of sequentiality on the imaginary line that preserved the
shirt’s individual identity. By this projection, we generate a sequentiality without events, an
abstracted flow (pp. 5–6).
hunky, Smooth, and Scaling Continuous

C
Variational Reasoning
Castillo-Garsow (2012) identified a distinction between two types of images of con-

tinuous reasoning about change, smooth, and chunky. Smooth reasoning is charac-
terized as “imagining a change in progress, of the physical experience of continuous
change based on real-world motion in space and/or time” (p. 69). Conceptions of
space and time are also constructions (Piaget, 1969) with a similar trajectory devel-
opmental trajectory to conceptions of number, but Castillo-Garsow highlights the
exclusion of reasoning about number to smooth continuous reasoning in his model.
He also states of smooth continuous reasoning that “in imagining changes in prog-
ress, there is no value that ever stays constant long enough to be calculated” (p. 65).
In contrast, chunky continuous reasoning is coordination of images of extents of
variation. “Chunky reasoning is first conceived at the end of a chunk, and only later
(or not at all) conceived of within the chunk, which disrupts the process of imagin-
ing change in real time” (p. 22). Thus, in Castillo-Garsow’s (2010, 2012) model of
chunky continuous reasoning, an image of a quantity’s variation and an image of a
quantity’s value are not simultaneous. Variation takes place within units, which can
be further partitioned or refined into sub-units.
The contrast between smooth and chunky reasoning about quantitative variation
can be described via the metaphor of motions of “extending a tape measure” (smooth
variation) versus “laying down rulers end-to-end” (chunky variation) (Castillo-
Garsow, 2012, p. 66). With chunky variation, there is an awareness of numerical
values between each endpoint of an interval, but the quantity’s magnitude actually
430 S. Boyce
taking on these intermediate values is not part of the image of the variation. The
“start” and “end” of the ruler (the interval) dominate the chunky yet connected
image; it is an incomplete integration of numerical and quantitative structure that is
completed with the construction of smooth variational reasoning. With a smooth
image of variation, the magnitude of a varying quantity takes on all of the values
between the endpoints of each interval, and there is nothing particularly special or
unique about the values of intervals’ endpoints, with regard to the quantity taking on
magnitudes.
At the onset, smooth and chunky variational reasoning were modeled by Castillo-
Garsow (2010) as independent ways of thinking about variation. Smooth variation
was modeled as a “qualitative” image of variation, devoid of arithmetical reasoning,
whereas “chunky” variation incorporated arithmetic schemes and operations, such
as partitioning (Castillo-Garsow, 2010). However, in Castillo-Garsow’s (2010)
paired-student teaching experiment, he found that one student could switch between
evoking “smooth” variational reasoning in some situations and “chunky” varia-
tional reasoning in others, whereas the other student never reasoned with smooth
variational reasoning. This led to a re-framing of smooth (co)variational reasoning
as developmentally beyond chunky (co)variational reasoning so that it incorporates
both smooth and chunky images of change (Thompson & Carlson, 2017). In
Thompson and Carlson’s (2017) variation framework, the previous use of the term
“smooth” (non-numerical) variational reasoning was recast as “gross” variational
reasoning: “The person envisions that the value of a variable increases or decreases,
but gives little or no thought that it might have values while changing” (Thompson
& Carlson, 2017, p. 440).
The role of imagined motion is inherent in the framing of smooth and chunky
continuous variation, and the distinguishing feature of smooth continuous variation
is the image of smooth, connected motion within an interval of change. Ellis et al.
(2020) identified an alternative image of continuous variation, which they termed
scaling continuous variation, incorporating a different motion: that of “zooming in”
on a chunk of variation. With regard to the previous metaphor of images of varia-
tion, scaling continuous variation is like laying out a sequence of rulers with antici-
pation that any single ruler could itself be partitioned into intervals, with smaller
rulers laid out end-to-end across that ruler. Thus, scaling continuous variation can
be considered recursive chunky variational reasoning. Table 12.2 summarizes my
conceptualization of the relationships between these three constructs.
Example: Contrasting Chunky, Smooth, and Scaling Continuous Variation
Imagine an analog clock’s hour hand moving with time. An image of chunky con-
tinuous variation could be an image of the hand rotating an amount of arc each
second. It is an image of continuous motion—the hand is not conceived as starting
and stopping—but within each second, the movement is conceived as a gross
change. A subsequent image might be of the clock hand rotating an amount of arc
each millisecond (instead of each second), but with chunky continuous variation,
the two images would not be coordinated as movements within movements. An
image of scaling continuous variation is of interiorized chunky variation, for which
Table 12.2 Relationships between within chunky, smooth, and scaling continuous variation
Chunky continuous Smooth continuous Scaling continuous
variation variation variation
Images of Composed of a Composed of Composed of a sequence
extensive sequence of arbitrarily incalculable extents; of arbitrarily small extents
variation small extents (chunks); simultaneous with (chunks); within each
within each chunk is an images of intensive chunk is an image of gross
image of gross motion (values of a variation. Images of
variation. Images of quantity continuously arbitrary sub-extents are
sub-extents of variation changing magnitude). anticipated as part of a
are not anticipatory and Independent of an image structure of variations
do not form a structure of chunky continuous within variations.
of variations within variation Recursive chunky
variations continuous variation
Coordination Unitizing of an extent Coordination of smooth, Unitizing of an extent of
of variations of chunky variation in intensive variation in one chunky variation in one
(images of one quantity and a quantity and quantity and a
covariation) corresponding corresponding corresponding
simultaneous chunky simultaneous smooth, simultaneous chunky
variation in another intensive variation in variation in another
quantity another quantity quantity, with anticipation
of invariance in the
correspondence
relationship under
sub-unitizing (scaling)
Relationship Image of variation in a Image of variation in a Image of variation in a
with extent of quantity is of an extent quantity is simultaneous quantity is of an extent of
variation of variation and a with the quantity’s variation and a
partitioning of the magnitude. There is partitioning of the
variation as a sequence coordination between an variation as a sequence of
of connected extent of previously connected sub-variations,
sub-variations. completed variation and with anticipation of
current variation, in the disembedding, re-framing,
same way that and partitioning any
continuous displacement sub-variation
is coordinated with
motion
there is potential coordination of chunks of variation within chunks of variation.

With scaling continuous variation, there is an anticipation of a potentially finer par-
titioning of any variation via an action of “zooming-in” and re-framing of the vary-
ing quantity. With this anticipation, there is a de-coupling of the simultaneity of the
motion of the hand and the variation in the amount it has rotated. The variation can
be “paused,” and a chunk of variation disembedded and re-framed without modify-
ing the extent of the variation. With smooth continuous variation, there is the simul-
taneity of previously completed change (e.g., the current magnitude of the arc) and
current variation (the hand’s movement) throughout the image of variation. Both
chunky continuous and scaling continuous variation lack this simultaneity because
they incorporate the unitizing of a “chunk” of variation.
432 S. Boyce
Quantitative Variational Reasoning
Johnson and McClintock (2018) distinguish quantitative variational reasoning as

“students’ reasoning about attributes that they can conceive of as capable of varying
and possible to measure” (p. 300). Note that the term “capable” in the definition
allows for both potential variation and lack of variation. This allows for constant
variation (e.g., increasing or decreasing at a constant rate) to be considered quantita-
tive. The second part of the distinction indicates how variational reasoning, in gen-
eral, need not be quantitative. For instance, reasoning involving arithmetic variations
involves reasoning about numerical values changing, in which numerical values are
dissociated from quantities.
An ambiguity in this definition of quantitative variational reasoning is what it
means for an attribute to be possible to measure. In Johnson and McClintock’s
(2018, p. 308) analysis of students’ reasoning filling the interior of polygonal
regions, they exemplify analytic differences by describing a student’s evoking
“smaller amounts of areas added” as indicating quantitative reasoning, but they
described another child’s evoking “the two sides [of the triangle] are getting closer”
as “physical” rather than quantitative. Inherent in their modeling was that the child’s
image of the triangle filling did not include anticipation of the possibility of measur-
ing the closeness. But with a broader view of quantity, the image of “getting closer”
may be of gross, intensive, or extensive quantity.
With a more inclusive framing of quantity, quantitative variational reasoning
involves reasoning about how quantities are changing. Are they increasing or
decreasing, discretely or continuously, arbitrarily or regularly, in bursts, or smoothly,
at a constant rate or a varying rate? Moreover, variation of a quantity can itself be
considered a varying quantity even if it is conceived without a unit of measure (e.g.,
increasing faster or increasing slower can be considered gross quantitative varia-
tional reasoning). This framing allows for smooth continuous variation without
anticipation of calculation to still be considered quantitative variational reasoning.
In addition to arithmetical and quantitative variation distinctions, researchers
have extended variational reasoning to model students’ reasoning about images of
variation when reasoning about variables within differential equations or to model
students’ reasoning about more than two variables (e.g., Jones, 2013, 2015; Jones &
Kuster, 2021). These models do not necessitate reasoning about quantities. Instead,
the distinguishing aspect is the structure of the relationships between variables. For
instance, Jones and Kuster (2021) identified an awareness of the dependence of a
variation within the value of a derivative (as determined by a differential equation)
on the variation with its independent variable as an important distinguishing feature
of students’ reasoning. Attention to how variables correspond, such as the roles of
dependent and independent variables, is of increased importance when modeling
covariational reasoning.
ovariational Reasoning as Correspondence

C
Between Variations
Saldanha and Thompson (1998) defined covariational reasoning as “holding in

mind a sustained image of two quantities’ values (magnitudes) simultaneously”
(p. 299). There are several aspects to unpack in this definition. The term “sustained”
within the definition refers to the simultaneity of the two quantities’ values rather
than the persistence of figurative imagery. The construction of such simultaneity is
a cognitive structure, which has been referred to as a “multiplicative object”
(Piaget, 1970b).
The development of covariational reasoning thus involves construction and coor-
dination of three cognitive structures: a structure for reasoning about variations
within one quantity, a structure for variations within another quantity, and a struc-
ture for reasoning about the correspondence between the two variations. An impor-
tant contrast with covariational reasoning is reasoning about a correspondence
between elements of two (or more) sets (Confrey & Smith, 1995). A correspondence
image includes identification of an element of one set, identification of an element
of a second set, and an experience of association, the result of the pairing action.
The processes leading to the identification of elements need not be reversible, and
there may be a sequential aspect to the association. That is, the resulting image may
or may not be a multiplicative object.
Generally, the elements of a correspondence need not refer to values of a quan-
tity’s magnitude, or even to numerical values. For example, one can form a multipli-
cative object by pairing two nominative or interval types, such as a color code, to a
hazard level. Covariational reasoning includes reasoning about correspondence but
with added specificity. For example, consider the problem, “What is the relationship
between the number of eyes that a group of dogs has and the number of dogs?”
(Blanton, 2008, p. 31). There is a correspondence relationship in this problem, in
that for every dog, there are two eyes. Suppose a child reasons, “If we add one more
dog to the group, the number of eyes in the group increases by two.” On the one
hand, the child is evoking reasoning about corresponding changes to the two dis-
crete quantities (adding one dog, increasing by two eyes). But this utterance does
not provide an indication that the two quantities are necessarily conceptualized as
varying; there is no indication that discrete variations in the number of dogs or in the
number of eyes are part of the images of those two quantities.
In order to describe images of covariational reasoning, it is helpful to first con-
sider structures for reasoning about variations within a single quantity. The image of
a quantity constrains but does not determine images of variations within that quan-
tity. For instance, it is possible to conceive of a variation as a signed difference for
a quantity whose image includes magnitude on an interval scale or a ratio scale. If
the image of quantity includes magnitude on but an ordinal scale, then an image of
variation will be limited to a nominal scale (e.g., gets larger / gets smaller / stays the
same) or an ordinal scale (e.g., goes up by 2 units / goes down by 3 units / stays the
same). If an image of quantity only includes magnitude on a nominal scale (i.e.,
434 S. Boyce
positive/negative), then an image of variation must be limited as well (same/

different).
Next, what are structures for reasoning about variation extensively? These are
structures for relating variations, including accumulating variations within a quan-
tity and for relationships between variations of different quantities. For instance, if
one reasoned about the accumulation of distance traveled on a walk from their
house to the park, they might conceive of the quantity as always increasing, and
thus, the variation in the distance traveled as always positive, or, perhaps, that the
total accumulated variation was a length of a particular magnitude. Either could be
considered gross variational reasoning. It is an image of the extent (the totality) of
variation without an image of intensive variations. Intensive variational reasoning is
a coordination of intensive and gross variations. With intensive variational reason-
ing, the extent of the variations is an accumulation of sub-variations, but the image
of sub-variations is of gross variation. Extensive variational reasoning includes
simultaneously reasoning intensively and extensively about accumulating varia-
tions. With extensive variational reasoning, the image of a sub-variation is interior-
ized as itself an extent of accumulated variation.
What are the possibilities for the correspondence between two (or more) varia-
tions? For instance, if one quantity is conceived as varying intensively, and another
quantity is conceived as varying extensively, is it possible for the correspondence
between the variations to be simultaneous? How does this relate to the construction
of instantaneous rate of change? Much of the earlier research regarding covaria-
tional reasoning has focused on students’ evocations and re-presentations of mental
imagery in their graphing activity when engaging in tasks designed to evoke covari-
ational reasoning about instantaneous rate of change (e.g., Carlson et al., 2002).
Progressively, research has focused on images of variation and coordinations of
variations prior to leading to the construction of instantaneous rate of change (e.g.,
Johnson, 2015).
Marilyn Carlson et al.’s (2002) seminal framework for covariational reasoning
identifies five mental actions (Table 12.3), along with behaviors of graphing activi-
ties indicative of those mental actions. Notice that each mental action is described
as a coordination consistent with forming a multiplicative object. First, the coordi-
nation is of values in variable one and changes in variable two (MA1), of direction-
ality of change in variable one and changes in variable two (MA2), of an amount of
change in variable one and changes in variable two (MA3), of average rate of change
of a function with uniform increments of change in the input variable (MA4), and
of instantaneous rate of change of the function with continuous changes in the inde-
pendent variable. The framework thus includes hallmarks of a progressive develop-
ment of images of extensive variations within quantities (from gross to intensive to
extensive) as well as a development of a correspondence between two variations
that were elaborated by Johnson (2015). We can interpret these frameworks of men-
tal actions as a trajectory for conceptualizing instantaneous rate of change as an
intensive quantity.
This is in concert with other ways of describing coordination when engaging in
tasks designed to engender covariational reasoning. In Kertil et al.’s (2019)
Table 12.3 Mental actions of the covariational reasoning framework (Carlson et al., 2002, p. 357)
Mental action Description of mental action
Mental Action Coordinating the value of one variable with changes in the other
1 (MA1)
Mental Action Coordinating the direction of change of one variable with changes in the other
2 (MA2) variable
Mental Action Coordinating the amount of change in one variable with changes in the other
3 (MA3) variable
Mental Action Coordinating the average rate-of-change of the function with uniform
4 (MA4) increments of change in the input variable
Mental Action Coordinating the instantaneous rate of change of the function with continuous
5 (MA5) changes in the independent variable for the entire domain of the function
elaboration on ways of coordinating variables, they frame four aspects: uncoordi-

nated, indirectly coordinated, directly coordinated, and direct and systematic coor-
dination (p. 211). Uncoordinated and indirectly coordinated variables are those that
lack simultaneity in reasoning about variations. With uncoordinated variables, two
quantities may each vary with respect to time, but the two variations are conceived
as separate and correspond to different variations of time. Indirectly coordinated
variables are those for which a conception of an invariant relationship between two
variables requires reasoning about a common parameter (most commonly time).
Directly coordinated variables are those for which variations in two variables are
conceived as simultaneous, but the relation between the variations is limited to
images of gross quantity (i.e., “the height decreases as the volume decreases”).
Direct and systematic coordination, the final level in Kertil et al.’s (2019) frame-
work, is the level at which one can also reason about the extent of variations, either
chunkily or smoothly. This level of coordination is required for the higher mental
actions (4 and 5) of Carlson et al.’s (2002) framework, which enable quantifying a
ratio of changes, and subsequently a rate of change, as a quantitative object.
ethodological Considerations for Investigating Quantitative

M
and Covariational Reasoning
For a researcher to investigate and model quantitative reasoning requires fostering

of mathematical norms for students to engage in non-routine yet accessible tasks,
with opportunities to present the structure of their reasoning (i.e., with diagrams)
and to listen to and respond to others’ questions about their thinking and their inten-
tions. Typically, researchers of quantitative variational reasoning introduce task
contexts which, to the researcher, have the potential to afford reasoning about quan-
tities extensively. This is because they provide more opportunities for students to
reason about relationships between varying quantities. Often, the tasks researchers
pose assess and perturb students’ abilities to inhibit reasoning about experiential
436 S. Boyce
time or to reason about aspects of space other than height in order to focus on other
quantities’ variations (e.g., Paoletti & Moore, 2017).
Models of reasoning quantitatively regard the development of strategies, rou-
tines, framings, and habits of mind that enable solving problems by thinking about
quantities and their relationships. Models of quantitative reasoning describe percep-
tions, actions, operations, schemes, conceptions, and images. Sometimes research-
ers are interested in the interplay between models of quantitative reasoning (ways of
understanding) and models of reasoning quantitatively (ways of thinking), but at
other times the interest is in one of the two.
One of the characteristics of the genetic epistemologist is concern that instruc-
tion that attempts to accelerate conceptual development may have negative reper-
cussions. As a way of understanding, quantitative reasoning is co-constructed with
the development of arithmetical reasoning (Steffe & Olive, 2010, p. 57). But as a
way of thinking, reasoning quantitatively may be hindered, for instance, by arith-
metical reasoning (Steffe & Olive, 2010, p. 72) or by graphical reasoning (Moore &
Thompson, 2015). In school mathematics, referral to quantities too commonly
serves as a backdrop or context to practice applying a numerical procedure absent
from quantitative structure. Accordingly, if students are presented with a task with
numbers, they tend to evoke arithmetical reasoning. If images of variation or quan-
tity focus on the presentation of others’ (e.g., teachers’) use of diagrams, utterances,
or procedural routines, this may lead to the establishment of purely figurative, pro-
cedural schemes that have limited potential for abstraction, generalizability, and
transfer (Lobato, 2012).
Another important contrast is reasoning about the shapes (or geometric proper-
ties) of the graphs of functions. For instance, Nemirovsky and Rubin (1992) distin-
guished between calculus students’ “resemblance thinking” and “variational
thinking” when relating graphs of functions. For example, if a student were reason-
ing that the graph of velocity would need to be above the x-axis, if the graph of
position were above the x-axis, they would be reasoning about a resemblance.
Moore and Thompson (2015) have elaborated on resemblance thinking in their
description of static shape-thinking. With static shape-thinking, the students’ rea-
soning focuses on building systems of relations between iconic imagery, treating the
graph of a function like a wire to be shaped rather than the result of an emergent
trace of covarying quantities.
A related aspect is the conceptualization of a framing for quantities (Joshua
et al., 2015). They describe this framing as including three commitments: (p. 32):
An individual conceives of measures as existing within a frame of reference if the act of
measuring entails: (1) committing to a unit so that all measures are multiplicative compari-
sons to it, (2) committing to a reference point that gives meaning to a zero measure and all
non-zero measures, and (3) committing to a directionality of measure comparison addi-
tively, multiplicatively, or both.
One of the methodological affordances of considering a student’s framing of

quantities is that it includes opportunities for students to indicate how they are rea-
soning about constituent quantities in a relationship before reasoning about how
they vary or covary, thus providing an indication of what images the student is refer-
ring to when they use words like “increasing at a steady rate.” Conversely, a poten-
tial flaw in research studies investigating covariational reasoning is that the frame of
reference for measuring the extents of quantities may not always be investigated and
instead assumed by the researcher to be normative with an absence of counterindi-
cation. The development of anticipation for curtailing part of an image of an object’s
attributes and that for re-framing quantities to reason about corresponding varia-
tions are important aspects of (co)variational reasoning (see Stevens et al., 2017 for
elaboration on task design principles).
Experiential time is (implicitly) part of reasoning about variation, even if time is
not explicitly evoked as a varying quantity (Thompson & Carlson, 2017). A concep-
tion of space (its framing, and its symmetries) will be implicitly (or explicitly) part
of the images of coordination as well (Lee et al., 2020). Committing to choices for
unit, reference point, and directionality requires awareness of such choices. For
instance, if a student is asked to imagine a ball bouncing up and down on a string
and to construct a graph of the displacement of its height versus the vertical differ-
ence traveled, passing timed “bouncing” will either be implicitly or explicitly part
of their images of the variations. Tasking students to reason about relationships
between quantities prescribed in unfamiliar (i.e., reversed or transposed) frames of
reference is one way of investigating what relationships between varying quantities
are simultaneous and necessary for students and what are “structured” as part of a
figurative image (Liang & Moore, 2020; Moore et al., 2014).
he Role of Technology in Quantitative

T
One strategy for investigating quantitative and covariational reasoning is to incorpo-

rate animations and applets by which dynamic imagery can be communicated.
Increasing familiarity with computer technology, in particular graphing technology,
makes physical actions upon static figurative materials, such as “zooming in,”
“zooming out,” or “sliding,” commonplace. The abilities to “pause” dynamic imag-
ery, or to “trace” a graph, are important as well. Task design decisions include what
tools to make available or constrain and what aspects to highlight. For instance, to
aid in evoking quantitative, rather than numerical, reasoning, researchers often pro-
hibit the use of measurement devices, such as rulers or number lines, and hide the
display of coordinates in a dynamic Cartesian graph in interviews and teaching
experiment sessions. The availability of measurement devices and the visibility of
coordinates should be described and explained in the methodology section of
research reports.
Often familiar approaches to representing quantities (e.g., a bar whose length
represents the magnitude of a quantity) are introduced in non-standard contexts
(e.g., a height being represented by a horizontal bar length). To evoke a student’s
438 S. Boyce
re-presentation of their perceived imagery of varying quantities depicted in an ani-

mation, a task may limit the number of times an animation can be viewed, or limit
the ability for the individual to pause or otherwise manipulate the animation. A
common task for students to assess their (co)variational and quantitative reasoning
is to be presented with an animation depicting some varying quantities, with a task
asking the student to verbalize relationships between quantities and subsequently
construct a graph depicting the relationships. Elaborations on this structure include
asking students to visualize a graph, or choose among existing graphs (e.g., Carlson
et al., 2010).
One lesson I learned when investigating students’ reasoning about a verbal
description of water evaporating from a vase is that not having an animation for the
student to reason about can contribute to a different range of foci in the students’
mental imagery of variation. One college student, when asked to reason about the
relationship between the volume of water in the vase and the height of the water in
the vase, evoked an image of non-monotonicity in the height of the water as evapo-
ration took place, describing an image of the formulation of hemispheric bubbles at
the surface that they had learned about in Chemistry. This student had an image of
the volume (and height) as varying discretely at a molecular level, which would not
have been apparent if an animation depicting an approximately constant change in
volume with respect to time had been shown.
On the other hand, when an animation is present, it may become difficult to dis-
tinguish between the students’ reasoning about an animation (and actions they
might imagine taking upon the animation, such as pausing or zooming in) and their
reasoning about the varying quantities that it represents. Ellis et al. (2020) depict
scaling continuous variational reasoning, which includes students’ reasoning about
varying areas of geometric figures that were depicted growing in animations. They
detail how one twelve-year-old student in their study reasoned about continuous
variation that was not vacuous within chunks and hence could not be considered
chunky continuous reasoning. However, the student did not describe imagining
motion within a frame of reference for the quantity, so in that sense, the reasoning
could not be considered smooth continuous variational reasoning either. The stu-
dent’s evoked reasoning about variation within chunks stemmed from an imagined
action of zooming in on a static geometric representation, which may not have been
available if he were imagining quantities distinct from their representations on com-
puter screens.
ole of Modeling Education in Promoting Quantitative

R
Throughout this chapter, the focus has been on describing second-order models of
students’ quantitative and (co)variational reasoning, with the discussion centering
on distinctions of students’ ways of understanding quantities and their relationships.
It is also important to consider the role of education in supporting quantitative and
covariational understandings and ways of thinking. This regards the design and
engineering of situations that can afford opportunities for students to assimilate and
accommodate their schemes, experiment with mental actions, and reflect on their
effects (Simon et al., 2004). In this section, I discuss how promoting quantitative
and covariational reasoning overlaps with the goals of applied mathematics and
mathematical modeling education.
Blum and Niss (1991, pp. 42–44) described five arguments for emphasizing the
application of mathematics, mathematical modeling, and problem-solving in math-
ematics education. The first argument regards general disposition and affect, with
its “fostering overall explorative, creative and problem-solving capacities (such as
attitudes, strategies, heuristics, techniques, etc.), as well as open-mindedness, self-
reliance, and confidence.” The second argument pertains to quantitative literacy and
social justice, enabling students to “recognize, understand, analyse and assess rep-
resentative examples of actual uses of mathematics, including (suggested) solutions
to socially significant problems.” The third argument pertains to the utility of math-
ematics, that in order to develop breadth in mathematical ways of thinking, students
should have experiences in which mathematical goals are not primary. The fourth
argument pertains to the mirroring of mathematics as a science in the classroom,
that applied mathematics should be represented as pure mathematics. The fifth argu-
ment is most germane to genetic epistemology: the “incorporation of problem solv-
ing, applications and modeling aspects and activities in mathematics instruction is
well suited to assist students in acquiring, learning and keeping mathematical con-
cepts, notions, methods and results, by providing motivation for and relevance of
mathematical studies.”
One of the main findings stemming from Piaget’s studies with children is that
coordinating structures develop in tandem across seemingly disparate quantitative
contexts, such as in the construction of number, motion, time, and space. Following
Piaget’s findings, researchers endeavored to replicate his experiments. Researchers
sometimes found that modest changes to Piaget’s tasks could lead to younger chil-
dren exhibiting more advanced reasoning and, conversely, that many older children
did not reach the levels of reasoning to be expected from generalizing Piaget’s stud-
ies. For instance, encouraging children to consider multiple quantities (not just the
most salient quantity) in a task situation seemed to “speed up” their conservation of
volume (Siegler, 2016). The same general principle applies to mathematical model-
ing instruction. Encouraging students to make their quantitative reasoning explicit—
to represent and discuss and critique and evaluate—provides opportunities for
developing quantitative ways of thinking. Engaging students in mathematical mod-
eling can thus explicitly encourage reflection on the relationships between their
quantitative and non-quantitative (e.g., algebraic, computational, and arithmetic)
ways of thinking (Lesh & Harel, 2003).
There is a reciprocal relationship between the long-term development of students’ algebraic
abilities and the long-term development of their reasoning from which these abilities
emerge. If algebra, meaning the use of representational practices that employ systematic
use of symbols to express quantitative and structural relationships, is to become students’
means of expressing and supporting their thinking, they must have experiences from
whence the thinking that those practices support emerges (Smith & Thompson, 2007, p. 7).
440 S. Boyce
Within genetic epistemology, it is the reflection upon actions on objects that

engender conceptions. Social interactions provide opportunities to reflect on com-
parisons with others’ reasoning. An important approach to engendering quantitative
covariational reasoning is to engage students in problem-solving and mathematical
modeling routines that provide opportunities for them to communicate with one
another (e.g., Carlson et al., 2003). For example, Kertil et al. (2019) designed a
modeling routine in which their students (prospective middle grades teachers)
engaged in a sequence of tasks, each ostensibly requiring more sophisticated coor-
dinations between quantities. In the initial task, there was only one quantity that
covaried with time so that it could be accessible to students who had yet to coordi-
nate quantities. In subsequent tasks, additional variations in attributes of dynamic
objects were introduced, such as the “filling tanks with water” task (Carlson et al.,
2003), and a reversal, “emptying” task, in which students were tasked to explain the
change of the height of the water with respect to the amount of water draining from
the tank. As part of the research design, students were asked to produce reflection
papers in addition to preparing group reports and participating in interviews.
In the modeling tasks described above, the launch began with dynamic imagery
without stipulating quantities or units of measure. In contrast, in the problem-
solving approach of Mkhatshwa (2020) and Mkhatshwa and Doerr (2018), they
provided students with few details about quantities at the onset of displaying written
tasks—their values and units of measure—and did not provide dynamic imagery.
They argue that “business calculus instructors and course materials such as business
calculus textbooks need to provide ample and meaningful opportunities for students
to reason about relationships among several quantities using numerical tables”
(p. 159). Given the growing importance of data science in mathematics education,
an important area of research in quantitative and covariational reasoning on the
horizon is investigating relationships between (big) tabular data, visualizations of
that data, and images of variation and covariation of quantities.
References
Barrett, J. E., Sarama, J., & Clements, D. H. (2017). Children’s measurement: A longitudinal study
of children’s knowledge and learning of length, area, and volume. Journal for Research in
Mathematics Education Monograph Series, 16.
Blanton, M. L. (2008). Algebra and the elementary classroom: Transforming thinking, transform-
ing practice. Heinemann.
Blum, W., & Niss, M. (1991). Applied mathematical problem solving, modelling, applications,
and links to other subjects – State, trends and issues in mathematics instruction. Educational
Byerley, C., Boyce, S., Grabhorn, J., & Tyburski, B. (2019). Investigating STEM students’ mea-
surement schemes with a units coordination lens. In Proceedings of the 22nd annual confer-
ence on research in undergraduate mathematics education.
Carlson, M., Larsen, S., & Lesh, R. (2003). Integrating a models and modeling perspective with
existing research and practice. In R. Lesh & H. Doer (Eds.), Beyond constructivism: Models and
modeling perspectives on mathematics problem solving, learning, and teaching (pp. 465–478).
Lawrence Erlbaum Associates.
Carlson, M., Oehrtman, M., & Engelke, N. (2010). The precalculus concept assessment: A tool for
assessing reasoning abilities and understandings. Cognition and Instruction, 28(2). https://doi.
org/10.1080/07370001003676587
Castillo-Garsow, C. C. (2010). Teaching the Verhulst model: A teaching experiment in covariational
reasoning and exponential growth. Unpublished Ph.D. dissertation, Arizona State University.
Castillo-Garsow, C. (2012). Continuous quantitative reasoning. In R. L Mayes & L. L. Hatfield
(Eds.), Quantitative reasoning and mathematical modeling: A driver for STEM integrated edu-
cation and teaching in context, 2 (pp. 55–73). Available at http://www.uwyo.edu/wisdome/_
files/documents/Castillo_Garsow.pdf
Confrey, J., & Smith, E. (1995). Splitting, covariation, and their role in the development of expo-
nential functions. Journal for Research in Mathematics Education, 26(1), 66–86.
Ellis, A. B. (2007). The influence of reasoning with emergent quantities on students’ generaliza-
tions. Cognition and Instruction, 25(4), 439–478. https://doi.org/10.1080/07370000701632397
Ellis, A., Ely, R., Singleton, B., & Tasova, H. (2020). Scaling-continuous variation: Supporting
students’ algebraic reasoning. Educational Studies in Mathematics, 104(1), 87–103. https://
doi.org/10.1007/s10649-020-09951-6
Euler, L. (1748). Introductio in Analysin Infinitorum. Annotated translation by Ian Bruce. Available
at http://www.17centurymaths.com/contents/introductiontoanalysisvol1.htm. Accessed 15
Nov 2021.
Glasersfeld, E. V. (1996). The conceptual construction of time. Paper presented at the conference
on mind and time. Available at http://www.oikos.org/Vonglasoct1.htm
Cognition and Instruction, 28(4), 383–432. https://doi.org/10.1080/07370008.2010.511565
Hackenberg, A. J., Aydeniz, F., & Jones, R. (2021). Middle school students’ construction of
quantitative unknowns. Journal of Mathematical Behavior, 61. https://doi.org/10.1016/j.
jmathb.2020.100832
Harel, G. (2008). DNR perspective on mathematics curriculum and instruction, Part I: focus on
proving. ZDM, 40, 487–500.
Jahnke, H. N. (2003). A history of analysis. American Mathematical Society.
Johnson, H. L. (2015). Together yet separate: Students’ associating amounts of change in quanti-
ties involved in rate of change. Educational Studies in Mathematics, 89, 89–110. https://doi.
org/10.1007/s10649-014-9590-y
Johnson, H. L., & McClintock, E. (2018). A link between students’ discernment of variation in
unidirectional change and their use of quantitative variational reasoning. Educational Studies
in Mathematics, 97, 299–316. https://doi.org/10.1007/s10649-017-9799-7
Jones, S. R. (2013). Understanding the integral: Students’ symbolic forms. The Journal of
Mathematical Behavior, 32(2), 122–141. https://doi.org/10.1016/j.jmathb.2012.12.004
Jones, S. R. (2015). Areas, anti-derivatives, and adding up pieces: Definite integrals in pure math-
ematics and applied science contexts. The Journal of Mathematical Behavior, 38, 9–28. https://
doi.org/10.1016/j.jmathb.2015.01.001
Jones, S. R., & Kuster, G. E. (2021). Examining students’ variational reasoning in differential equa-
tions. The Journal of Mathematical Behavior, 64. https://doi.org/10.1016/j.jmathb.2021.100899
Joshua, S., Musgrave, S., Hatfield, N., & Thompson, P. W. (2015). Conceptualizing and reasoning
with frames of reference. In Proceedings of the 18th meeting of the research in undergraduate
mathematics education conference (pp. 31–44).
Kertil, M., Erbas, A. K., & Cetinka, B. (2019). Developing prospective teachers’ covariational rea-
soning through a model development sequence. Mathematical Thinking and Learning, 21(3),
207–233. https://doi.org/10.1080/10986065.2019.1576001
442 S. Boyce
Kleiner, I. (1989). Evolution of the function concept: A brief survey. College Mathematics Journal,
20, 282–300.
Lee, H. Y., Hardison, H. L., & Paoletti, T. (2020). Foregrounding the background: Two uses of
coordinate systems. For the Learning of Mathematics, 40(2), 32–37.
Lesh, R., & Harel, G. (2003). Problem solving, modeling, and local conceptual development.
Mathematical Thinking and Learning, 5(2–3), 157–189. https://doi.org/10.1080/1098606
5.2003.9679998
Liang, B., & Moore, K. C. (2020). Figurative and operative partitioning activity: Students’ mean-
ings for amounts of change in covarying quantities. Mathematical Thinking and Learning,
23(4), 291–317. https://doi.org/10.1080/10986065.2020.1789930
research and practice. Educational Psychologist, 47(3), 232–247.
Mkhatshwa, T. P. (2020). Calculus students’ quantitative reasoning in the context of solving related
rates of change problems. Mathematical Thinking and Learning, 1–23. https://doi.org/10.108
0/10986065.2019.1658055
Mkhatshwa, T. P., & Doerr, H. M. (2018). Undergraduate students’ quantitative reasoning in eco-
nomic contexts. Mathematical Thinking and Learning, 20(2), 142–161. https://doi.org/10.108
0/10986065.2018.1442642
Moore, K. C., & Carlson, M. P. (2012). Students’ images of problem contexts when solving applied
problems. The Journal of Mathematical Behavior, 31(1), 48–59. https://doi.org/10.1016/j.
jmathb.2011.09.001
Moore, K. C., & Thompson, P. W. (2015, February). Shape thinking and students’ graphing activ-
ity. In Proceedings of the 18th meeting of the MAA special interest group on research in under-
graduate mathematics education (pp. 782–789). RUME.
Moore, K. C., Silverman, J., Paoletti, T., & LaForest, K. (2014). Breaking conventions to support
quantitative reasoning. Mathematics Teacher Educator, 2(2), 141–157. https://doi.org/10.5951/
mathteaceduc.2.2.0141
Nemirovsky, R., & Rubin, A. (1992). Students’ tendency to assume resemblances between a func-
tion and its derivative (TERC working paper 2–92).
Paoletti, T., & Moore, K. C. (2017). The parametric nature of two students’ covariational rea-
soning. The Journal of Mathematical Behavior, 48, 137–151. https://doi.org/10.1016/j.
jmathb.2017.08.003
Piaget, J. (1954). The construction of reality in the child (M. Cook, Trans.). Basic Books.
Piaget, J. (1965). The child’s conception of number. Norton.
Piaget, J. (1969). The child's conception of time (M. J. Pomerans, Trans.). Basic Books.
Piaget, J. (1970a). Genetic epistemology. Basic Books.
Piaget, J. (1970b). Structuralism. Basic Books.
Piaget, J., & Inhelder, B. (1969). The early growth of logic in the child. Norton.
Piaget, J., & Inhelder, B. (1976). The origin of chance in children. Norton.
Pimentel, J. L. (2019). Some biases in Likert scaling and usage and its correction. International
Journal of Sciences: Basic and Applied Research, 45(1), 183–191.
Saldanha, L. A., & Thompson, P. W. (1998). Re-thinking co-variation from a quantitative per-
spective: Simultaneous continuous variation. In S. B. Berensen et al. (Eds.), Proceedings of
the 20th annual meeting of the North American Chapter of the International Group for the
Psychology of Mathematics Education (Vol. 1, pp. 298–303). ERIC Clearinghouse for Science,
Mathematics, and Environmental Education.
Siegler, R. S. (2016). Continuity and change in the field of cognitive development and in the per-
spectives of one cognitive developmentalist. Child Development Perspectives, 10(2), 128–133.
Simon, M. A., & Placa, N. (2012). Reasoning about intensive quantities in whole-number
multiplication? A possible basis for ratio understanding. For the Learning of Mathematics,
32(2), 35–41.
Simon, M. A., Tzur, R., Heinz, K., & Kinzel, M. (2004). Explicating a mechanism for concep-
Mathematics Education, 305–329. https://doi.org/10.2307/30034818
Smith, J. P., & Barrett, J. E. (2017). The learning and teaching of measurement: Coordinating
quantity and number. In J. Cai (Ed.), Compendium for research in mathematics education
Smith, J., & Thompson, P. W. (2007). Quantitative reasoning and the development of algebraic
reasoning. In J. J. Kaput, D. W. Carraher, & M. L. Blanton (Eds.), Algebra in the early grades
(pp. 95–132). Erlbaum.
Stacey, K., & MacGregor, M. (1997). Ideas about symbolism that students bring to algebra. The
Mathematics Teacher, 90(2), 110–113.
Steffe, L. P., & Cobb, P. (1988). Arithmetical meanings and strategies. Springer.
Steffe, L. P., & Wiegel, H. G. (1996). On the nature of a model of mathematical learning. In
L. P. Steffe, P. Nesher, P. Cobb, G. A. Goldin, & B. Greer (Eds.), Theories of mathematical
learning (pp. 477–498). Erlbaum.
Steffe, L. P., Glasersfeld, E. von, Richards, J., & Cobb, P. (1983). Children’s counting types:
Steffe, L. P., Liss, D. R. I., & Lee, H. Y. (2014). On the operations that generate intensive quantity.
In K. C. Moore, L. P. Steffe, & L. L. Hatfield (Eds.), Epistemic algebraic students: Emerging
models of students’ algebraic knowing (WISDOMe monographs) (Vol. 4, pp. 49–79).
Stevens, I. E., Paoletti, T., Moore, K. C., Liang, B., & Hardison, H. (2017). Principles for design-
ing tasks that promote covariational reasoning. In A. Weinberg, C. Rasmussen, J. Rabin,
M. Wawro, & S. Brown (Eds.), Proceedings of the 20th annual conference on research in
undergraduate mathematics education (pp. 928–936).
teachers’ knowledge base. Journal of Mathematics Teacher Education, 23(1), 69–95. https://
doi.org/10.1007/s10857-018-9409-3
Studies in Mathematics, 25(3), 165–208. https://doi.org/10.1007/bf01273861
Thompson, P. W. (1994). The development of the concept of speed and its relationship to concepts
learning of mathematics (pp. 181–234). SUNY Press.
Thompson, P. W., & Carlson, M. (2017). Variation, covariation, and functions: Foundational ways
of thinking mathematically. In J. Cai (Ed.), Compendium for research in mathematics educa-
tion (pp. 421–456). National Council of Teachers of Mathematics.
Thompson, P. W., & Harel, G. (2021). Ideas foundational to calculus learning and their links
to student difficulties. ZDM Mathematics Education, 53, 507–519. https://doi.org/10.1007/
s11858-021-01270-1
Thompson, P. W., & Thompson, A. G. (1992, April). Images of rate. Paper presented at the annual
meeting of the american educational research association.
Tzur, R., & Simon, M. (2004). Distinguishing two stages of mathematics conceptual learning.
International Journal of Science and Mathematics Education, 2, 287–304.
Ulrich, C. (2012). The addition and subtraction of signed quantities. In R. L Mayes & L. L. Hatfield
(Eds.), Quantitative reasoning and mathematical modeling: A driver for STEM integrated edu-
cation and teaching in context, 2 (pp. 127–141). Available at: http://www.uwyo.edu/wisdome/_
files/documents/Ulrich.pdf
Vinner, S., & Hershkowitz, R. (1980). Concept images and common cognitive paths in the devel-
opment of some simple geometrical concepts. In L. Puig & A. Gutierrez (Eds.), Proceedings of
the fourth international conference for the psychology of mathematics education (pp. 177–184).
Voyat, G. (1982). Piaget systematized. Lawrence Erlbaum.
Part III
Commentaries on Genetic Epistemology
and Its Use in Ongoing Research
Chapter 13
Genetic Epistemology as a Complex
and Unified Theory of Knowing
Anderson Norton
Throughout this text, we have attempted to elucidate key constructs related to

Piaget’s genetic epistemology, especially as they apply to mathematics education.
Constructs like reflective abstraction provide critical instruments for understanding
students’ mathematics, i.e., in building models of that mathematics (Glasersfeld &
Steffe, 1991). This chapter represents an attempt to bring those constructs together
within a unified theory of mathematical knowledge. After all, an epistemology is a
theory of knowledge—one that explains its origins.
The chapter begins with a review of genetic epistemology, as framed by Piaget
and especially as it applies to mathematics. The review draws heavily from Piaget’s
“magnum opus” (Kitchener, 1985, p. 4) or “chef d’oeuvre” (Smith, 1993, p. 185) on
the topic, L’Epistemologie Genetique: La Pensee Mathematique (Piaget, 1950).
Having never been translated from French, the text contains at least two key insights
largely lost on the English-speaking community.1 First, the book describes how
genetic epistemology put social-historical methods on equal footing with psychoge-
netic methods, at least in the beginning. Second, it clearly articulates the mathemati-
cal motivation for Piaget’s entire research program.
Following a review of genetic epistemology, the remainder of this chapter draws
together the work Piaget’s program has inspired within the field of mathematics
education, as represented by the preceding chapters. Our field has greatly advanced
Piaget’s goal of explaining the human origins of mathematical knowledge. We have
identified unifying constructs, such as units coordination (Steffe, 1992; Hackenberg,
1
Chapman (1988, pp. 196–212) provides perhaps the most substantial English overview of
the book.
A. Norton (*)
https://doi.org/10.1007/978-3-031-47386-9_13
448 A. Norton
2007) and covariational reasoning (Carlson et al., 2002); we have operationalized

opaque constructs, such as reflective abstraction (Simon & Tzur, 2004); and we
have extended developmental models to include advanced mathematics (Dubinsky,
1991). The result is a complex web of related tools for understanding mathematical
development, from infancy through adulthood. This chapter concludes with some
open questions for motivating continued research in mathematics education.
Genetic Epistemology
Epistemologies answer the question, “How is knowledge possible?” Traditional

accounts of knowledge attempt to answer the question by judging and justifying a
presumed accord between human knowledge, in its static forms, and a reality that
exists independent of human experience. In contrast, genetic epistemology turns the
question toward the study of transformations of knowledge from one state of know-
ing to another. It thus transforms the fundamental epistemological question into,
“How does knowledge develop?” (Smith, 1993, p. xii). This is a question that psy-
chology can help answer: “The genetic method turns to the study of knowledge in
its active psychological construction” (Piaget, 1950, p. 19).
As the name implies (and unsurprisingly, given Piaget’s background as a biolo-
gist), genetic epistemology has its roots in biology: “My central aim has always
been the search for the mechanisms of biological adaptation and the analysis and
epistemological interpretation of that higher form of adaptation which manifests
itself as scientific thought” (Piaget, 1977, p. XI, as cited in Glasersfeld’s, 1997,
homage to Piaget). Piaget found the roots of human knowledge in our interactions
with the environments we experience. Beginning from reflexes, we learn to act in
the world in more or less productive ways. These actions fit the environment inas-
much as they accomplish the intended goals. Thus, we deem viable the ideas that
those actions comprise in much the same way we might deem a species viable by
virtue of its survival within an ecological system. In the case of generating ideas, or
human knowledge, we must pay special attention to interactions within sociocul-
tural environments.
It is true that the development of the child is always influenced by the social environment,
which plays not only the role of accelerator but also transmits a host of notions that are
themselves a collective history. Insofar as the individual in training thus receives social
heritage formed in the past by previous generations of adults, it is obvious that the historico-
critical method, extending to sociological methods, assumes control of the psychogenetic
method. But it is no less clear, even when receiving the notions already completely prepared
by the social environment, that children transform and assimilate these within their succes-
sive mental structures in the same way, they assimilate the environment formed by the
things that surround them: these forms of assimilation and their succession constitute, thus,
a given that sociology and history don’t suffice in explaining, and it is in their study that the
psychogenetic method, in turn, controls the historicocritical method.
13 Genetic Epistemology as a Complex and Unified Theory of Knowing 449
In sum, the complete method of genetic epistemology is constituted by an intimate collabo-

ration between historicocritical methods and genetic psychology, and these by virtue of the
following principle, without doubt, common to studies of all organic development: that the
nature of lived reality is not revealed by initial states alone nor by terminal states, but by the
actual processes of its transformations. (Piaget, 1950, pp. 22–23)
English readers of Piaget quickly become familiar with Piaget’s developmental

psychology. What is lost in translation (or lack thereof) is the balance between the
psychological (psychogenetic) and sociocultural (historicocritical) on which he
founded his genetic epistemology.
Historicocritical and Psychogenetic Methods
Among English-translated texts, and as its title would suggest, Mathematical

Epistemology and Psychology best encapsulates Piaget’s epistemology of mathe-
matics (Beth & Piaget, 1966). That text contains two parts: mathematical philoso-
pher Evert Beth wrote the first part, focusing on the history of mathematics, and
Piaget wrote the second part, focusing on the psychology of mathematics. As
Kitchener (1981) notes of Piaget’s later work, explanations of the development of
knowledge focused more and more on the psychogenetic side. Piaget seemed to use
work from prior epistemologists, who relied mostly on historicocritical methods, to
form a framework that he began to fill in through studies of child development.
Once again, the French text provides further insight.
We see, in summary, the dual task of genetic epistemology. At the start, it merges with a
certain aspect of the psychology of intellectual development: it seeks to explain the forma-
tion of particular knowledge and, thus, to resolve the problem of knowing how we accumu-
late delimited knowledge. So long as it stays on psychogenetic terrain, it requires, like
psychology itself, a system of reference constituted by scientific knowledge admitted at the
time of consideration. But, when psychogenetic analysis is extended to historical-critical
analysis, the system of reference, until then regarded as fixed, turns in motion… (Piaget,
1950, p. 49).
Thus, Piaget relied on the history of mathematics to frame his study of the ori-
gins of mathematical knowledge. However, we only have access to that system of
reference via its formalization, and ultimately, Piaget was interested in explaining
formalization itself.
There is no denying the influence of sociocultural history on mathematics. Take,
for example, Euclidean geometry—an example we will return to throughout this
chapter. We inherit this geometry from Euclid via its formalization in the form of the
following five axioms.
1. To draw a straight line from any point to any point
2. To produce a finite straight line continuously in a straight line
3. To describe a circle with any center and radius
4. All right angles are congruent to one another
450 A. Norton
5. If a line falls across a pair of lines so that the sum of the interior angles created
on one side of the line sum to less than π, then the pair of lines will intersect on
that side [equivalent to Playfair’s axiom regarding the existence and uniqueness
of parallel lines]
Much has been said about Euclid’s fifth postulate (axiom) and the alternative
geometries introduced by rejecting it. Here, we focus on the first three axioms,
which Euclid inherited from Plato: they are Plato’s rules for straightedge and com-
pass construction. Thus, all of Euclidean geometry depends on rules, invented by
Plato, which could have been otherwise.
Consider the question of constructible polygons: which regular polygons can we
construct with a straightedge and compass? It is a question inherently related to the
definition of trigonometric functions. The cosine and sine functions give the x- and
y-coordinates, respectively, of points on the unit circle, for a given central angle (see
Chap. 4). Whenever we construct a regular polygon inscribed in the unit circle, we
also construct lengths that correspond to those coordinates. Thus, we geometrically
derive exact values for sine and cosine.
In high school, we teach students about special triangles. We have just described
what makes them so special, but they are not the only ones. As it turns out, we can
construct any n-gon where n is a power of two times a product of distinct Fermat
primes. Fermat primes are prime numbers of the form 2^(2n) + 1. There are five
known Fermat primes (3, 5, 17, 257, and 65,541), so there are 31 (25, minus 1 unless
we want to count the 1-gon) constructible polygons with an odd number of sides.
We can then construct polygons with an even number of sides by recursive doubling
any of those numbers of sides (by bisecting the central angle). In a connection that
would have thrilled Plato and would seem to buttress his Platonist claims, we can
find the 31 constructible odd-gons hidden within the first 32 rows of Pingala’s tri-
angle (a.k.a. Pascal’s triangle). Ironically, Plato’s own rules belie his claims.
Just as Euclidean geometry is not the only possible geometry, Plato’s three rules
are not the only set of possible rules for construction. We can imagine a rule that
would allow us to wrap a string around the circumference of a circle and then to
straighten it as a length. Under such a rule, π would be constructible, the set of con-
structible polygons (and special triangles) would be entirely different, and the
uncanny connection—between constructible polygons, Fermat primes, and
Pingala’s triangle—would be broken. These connections exist only in the corre-
spondence between actions we are allowed to perform, as guided by the tools
(straightedge and compass) and formalizations that define Euclidean geometry, and
those that define the construction of Fermat primes and Pingala’s triangle.
We see then how fascinating connections within the body of knowledge that we
call “mathematics” are both psychogenetic (coordination of our own mental actions)
and historicocultural. The historicocultural contributions (e.g., tools, axioms, and
definitions) are intended to foster communication and build consensus about the
mental actions we employ and the rules for composing them. Figures represent and
communicate (even to ourselves) products of those coordinations. However, what
defines mathematics as a unique form of knowledge, what renders it certain, and
what enables us to extend it beyond past experience, are the actions themselves. We
can invent all sorts of geometries so long as the rules for engagement maintain the
reversibility and composability of the actions they describe. The rules then constrain
action and help us communicate that we are referring to the same kinds of action
(e.g., sweeping a line through two points, which are Euclid’s first axiom and Plato’s
first rule of construction).
Few researchers who focus on sociocultural dimensions of mathematics educa-
tion draw upon Piaget. Instead, they tend to rely on frameworks that characterize
mathematics as a socially constructed body of knowledge in which the role of the
individual is relegated to enculturation (e.g., Vygotsky, 1978). Geoffrey Saxe stands
out as a notable exception. He saw Piaget’s genetic epistemology as part a frame-
work for studying the individual within society. In one such study, he found that
college math students invented a linguistic register of specialized word forms to
communicate their mental actions and build upon the actions of others, as they
formed conjectures about the area and perimeter of Sierpinski’s triangle.
These schematizations are enabled and constrained, on the one hand, by the cognitive
resources that individuals have generated in their prior microgenetic acts, and, on the other,
in their incorporation of sociogenetic developments of forms and functions, such as the
emerging linguistic register that has been reproduced and altered over the conjectures dis-
cussions as participants work to establish common ground as they engage in communica-
tion and problem solving with one another. (Saxe & Farid, 2021, p. 9)
The microgenetic acts and sociogenetic developments that Saxe documents in

his studies mirror Piaget’s psychogenetic and historicocritical methods. If Piaget’s
primary interest lay in establishing an epistemology of mathematics (and we will
soon see that it did), Saxe’s primary interest is the same within the microcosms
formed by various mathematical cultures. However, unlike Piaget, Saxe consistently
maintained the equilibrium between the two methods.
How Is Mathematical Knowledge Possible?
In considering the possibility of knowledge, epistemologists hold mathematics in a

privileged position. This privilege owes to the immutability of mathematical knowl-
edge, whether seen from a psychogenetic or historicocritical perspective. As a
quintessential example, 2 + 2 = 4 across time, people, and cultures. We find nothing
so reliable in science, language, or art.
Any explanation of mathematical knowledge should also address its remarkable
applicability, known as the “unreasonable effectiveness of mathematics in the natu-
ral sciences” (Wigner, 1990, p. 291). This mathematical efficacy has marveled phi-
losophers and scientists for centuries: “How can it be that mathematics, being after
all a product of human thought which is independent of experience, is so admirably
appropriate to the objects of reality?” (Einstein, 1954, p. 233).
It sometimes happens that a mathematical model fails, as it has in the case of
physical Euclidean space, but when this happens, the mathematics never gets the
452 A. Norton
blame. Einstein (1954) elaborated on this point, beginning as follows: “For if con-
tradictions between theory and experience manifest themselves, we should rather
decide to change physical laws than to change axiomatic Euclidean geometry”
(p. 236).
Einstein explained that applying Euclidean geometry to models of the universe
involves the sum of two parts: the axioms that define the geometry (G) and the
physical laws to which they apply (P). “Using symbols, we may say that only the
sum of G + P is subject to experimental verification” (p. 236). This sum passed
myriad empirical tests when P represented the laws of Newtonian physics and
before Einstein’s theory of relativity. Then relativity opened our eyes to observa-
tions of a warped space-time and a new understanding of gravity. The old G + P
failed, and a new one replaced it, but Euclidean geometry itself remains valid.
Because Euclidean geometry is not the only possible G, we could indeed change
it, as well as physical laws, P. Even if we do develop a new G that better fits some P,
the validity of G never suffers the result of a failed experiment the way P does. This
distinction between G and P mirrors the distinction between mathematics and sci-
ence in general: as an internally consistent system for operating, G remains viable
(i.e., reliable) regardless of whether it describes any physical reality. No wonder we
marvel that it so often does fit and even anticipate the worlds we experience!
Consider Piaget’s (1950) remarks on the geometric example:
Mathematics is in accord with physical reality in the most detailed manner. It never happens
to the physicist—however many and diverse may be the structures or relations they discover
in the material world—to find one that may not be expressed with precision in the language
of mathematics, as if there is a preestablished harmony between all the aspects of the physi-
cal universe and the abstract framework of geometry and analysis. Moreover, this accord is
realized not only in the moment of discovery of a physical law, but mathematical schemes
often anticipate, many years before, the experiential contents that will come to fit them.
Geometric and analytic forms can, needless to say, be elaborated without any concern for
reality. However, to the degree that they are deductively coherent, one is assured, not only
that experience will never fault them, but also (and this is the paradoxical point) that experi-
ence will sooner or later partially fill them and adapt exactly to them. The most beautiful
example of this insertion of the real within the framework prepared by mathematical deduc-
tion is, without doubt, that of Riemannian geometry. Voila, a free and audacious construc-
tion, continuing along the lines of Euclidean geometry, but contradicting its famous fifth
postulates that, without demonstration, we considered as imposed by direct observation. So,
there is this type of free creation of the mathematical spirit unconcerned about the real.
More than a half-century after this defiance of physical reality, it turns out that physics itself
went on to consider Riemannian geometry as better adapted than Euclidean geometry in
rendering a complete account for the phenomenon of gravity: the theory of relativity used
this framework thus prepared, and experience gives reason for this stroke of genius.
(p. 57–58)
Not only is Euclidean geometry not the only possible geometry, it is not even the
one that fits best with modern physics. As with the pairing of Euclidean geometry
and Newtonian physics, Riemannian geometry seemed to anticipate the reality of
relativity. So, with its perfect reliability and prescient applicability, how is mathe-
matical knowledge possible? Translating into Piaget’s genetic epistemology, we
might ask instead, how does mathematical knowledge develop? Piaget (1950)
considered this question as central, and his genetic framework further translated it
as follows:
The grand problem of all epistemology, principally genetic epistemology, is in effect under-
standing how the mind succeeds in constructing necessary connections, appearing as “inde-
pendent of time,” if the instruments of thought are nothing more than the psychological
operations subject to evolution and constituted in time. (p. 28)
In describing necessary and timeless connections, Piaget refers to logicomathe-

matical operations (e.g., spatial rotations). He finds biological roots for these opera-
tions in sensorimotor action. Reframing the central epistemological question once
more, he referred to the “irreversibility of temporal environments” and equated the
necessity and timelessness of operations with their reversibility (p. 29). Piaget’s
research on children’s development of logic, number, and space, all focused on the
transformation from the sensorimotor actions that we perform or imagine in time, to
the logicomathematical operations that we coordinate all at once. In each case, the
transformation coincided with the reversibility of action.
Taking that habits and elementary perceptions are essentially one-directional, sensorimotor
(or preverbal) intelligence discovers already the conduits of detour and return, which intro-
duce, in part, the associativity and reversibility of operations. On the plan of actions inter-
nalized by intuitive representations, the child begins anew not knowing how to reverse the
composition of images, in the means by which they think, the progressive articulations of
intuition engendering, on the other hand, a growing reversibility that, toward 7–8 years old,
leads to the first concrete logical operations: these consist, in effect, of actions of reunion,
of seriations, etc., found thus reversible in course of a long evolution. But the latter will only
be achieved toward 11–12 years old, when the actions are rendered reversible, able them-
selves to be translated under the form of propositions, that is to say, purely symbolic opera-
tions. That’s so, and only so, thanks to operative reversibility finally generalized. (pp. 28–29)
Elsewhere, Piaget (1970) has explained how reversibility renders actions per-
fectly reliable by virtue of the ever-present possibility of returning to the starting
point, no matter how long the chain of composed or coordinated actions. And if the
reversibility of logicomathematical operations explains their reliability, their organic
origins in sensorimotor activity explain why they are so well suited for building
models of the environments we experience.
Actions begin from reflexes (no doubt part of our biological heritage, via human
evolution), such as the rooting reflex by which a baby reacts to a touch on the cheek
by turning its head in the direction of the perceived touch (Piaget, 1973). The baby
begins extending reflexes to more controlled sensorimotor actions through experi-
ence. All sensorimotor actions occur through immediate interaction with the envi-
ronment. Once the child has internalized an action to perform them in imagination,
it can be evoked without reference to any particular past experience. After all, from
the observer’s point of view, no two experiences are ever the same and are only
experienced as such by the subject in assimilating them within a scheme for acting/
operating. Moreover, internalization and coordination make possible imagined
actions that could never be carried out in sensorimotor activity, such as planar reflec-
tion or the projection of a line onto a point. We can extend our coordination by
coordinating internalized actions with one another, as we do in planning a sequence
454 A. Norton
of actions or in novel tool use. When this coordination forms a system of compos-
able and reversible actions, we refer to them as logicomathematical operations
(Piaget, 1970).
Thus, the central question of epistemology, when addressed from a genetic per-
spective, comes down to the interiorization of actions, meaning the coordination of
actions within reversible and composable systems. Through his empirical studies
with children, Piaget carefully documented the stage-wise progression from reflexes
and sensorimotor activity to operations across several domains of math and logic. In
explaining the final shift from internalized actions (still temporal and not fully
reversible) to interiorized (simultaneous and reversible) operations, Piaget relied on
the construct of reflective abstraction, and that is why the construct plays such a
central role in his work. Still, the construct remains opaque for many researchers,
and we hope prior chapters (as summarized in the next section) have contributed
clarity. Accepting that, we find several benefits to Piaget’s action-based approach in
answering epistemological questions:
• The basis of operations in sensorimotor interaction with the environment ulti-
mately renders them useful in building and extending models of the worlds we
experience, which helps explain the “unreasonable effectiveness of mathematics
in the natural sciences” (Wigner, 1990).
• Actions of the human body extend to the use of cultural tools that afford their
construction, which helps explain cultural and historical influences on mathe-
matical development.
• The internalization of actions as imagined actions allows them to be evoked
through social interaction, including language and other symbolic representa-
tions, which can suggest new coordination of actions (i.e., mathematical
learning).
• Their internalization also allows for coordination into new operations that would
be physically impossible to perform, such as planar reflections.
• Their organization within reversible and composable systems allows for their
reliability and extension, respectively.
Connections to Prior Chapters and Critical Constructs
The prior section highlighted two neglected aspects of Piaget’s genetic epistemol-
ogy: his use of historicocritical methods, as well as psychogenetic methods, and his
overarching focus on developing an epistemology of mathematics. Those two
aspects bring unity to the program by (respectively) including social-cultural influ-
ences (e.g., tools and symbols) as integral to studies of mathematical development,
and giving primacy to actions as the basis for mathematics itself. Both aspects
appear within each chapter of Part I, as the authors of those chapters elaborate on
particular constructs for carrying out the program of genetic epistemology within
Chapter 3: Schemes and Operations
From the outset of Chap. 3, Tillema and Gatza describe schemes as “core explana-
tory constructs for studying mathematical learning” (p. 1). In line with the preced-
ing section, the authors note the biological basis of schemes within reflexes and
other sensorimotor activity. However, they mark a critical distinction between
behaviorist accounts of reflexes and the constructivist construct of an action scheme.
Whereas reflexes involve only a stimulus and response, schemes include a sequence
of actions as their central component (see Chap. 3, Fig. 3.2). The authors draw upon
one of von Glasersfeld’s (1993) lucid elaborations of constructivism to describe
schemes as three-part constructs: a recognition template for perceiving relevant
situations, a way of acting in that situation, and an expected result of acting. As
noted in the chapter, Steffe (2010) later extended the three-part construct into a tet-
rahedral model, which introduced a fourth component/vertex to separately represent
goals (see Chap. 3, Fig. 3.4).
The scheme construct relates closely to a pair of complementary constructs at the
heart of genetic epistemology: assimilation and accommodation. Drawing again on
von Glasersfeld’s (1995) work, Tillema and colleagues describe assimilation as the
mechanism by which we make meaning of the worlds in which we act. These
actions include interaction with others and within (sociocultural) milieux, as
expressed through language, symbols, and tools. These cultural products may be
created or used to convey an intended meaning, but ultimately the success of this
intention depends upon the individual’s assimilation of those products within their
available schemes for acting/operating (as noted by Piaget, 1950).
On the one hand, the assimilation of experiences within schemes provides us
with a way of seeing and acting in society and the embodied world. On the other
hand, our expectations and goals do not always fit the perceived results of acting.
This perturbation can provoke us to modify, or make an accommodation in, one or
more of our schemes. In the previous section, we saw an example of this at the level
of scientific thought, through the history of gravity.
Euclidean geometry describes ways of operating in the world and, thus, ways of
building models of the worlds we experience. However, when observations do not
fit those models (light from a distant star bent around the Sun), we modify our
schemes. The actions, and even their coordinations with one another, remain valid,
but their application within a scientific theory (or scheme), such as Newtonian phys-
ics, requires accommodation—new ways of acting as described by a different
geometry.
Once more following von Glasersfeld (1995), the authors make a distinction
between sensorimotor schemes and conceptual schemes. Like the distinction
between reflexes and sensorimotor schemes, this new distinction follows Piaget’s
trajectory for the development of operations, as reversible and composable actions.
Specifically, when actions become internalized and coordinated with one another,
the child can anticipate results of activity without having to carry it out on figurative
material (we see a similar distinction, between figurative and operative thought, in
456 A. Norton
Chap. 4). The authors use the equipartitioning scheme as an example of a scheme
that may operate on figurative material but becomes operative through the interior-
ization of its actions and their coordination with one another (the process of reflec-
tive abstraction, as discussed in Chap. 8). From this perspective, and as the authors
note, we can think about a mathematical concept as an interiorized scheme.
Chapter 4: Figurative and Operative Thought
This next chapter picks up on an important theme raised in the previous chapter.
Specifically, although we might act on figurative material to carry out sensorimotor
activity, mathematical objects arise from the coordination of those actions them-
selves, ultimately structured within operational systems that do not rely on figura-
tive material. This distinction—between figurative and operative thought—is
especially important to Piaget’s epistemology of mathematics because it clarifies
the nature of mathematical objects.
As noted by the authors, and as recounted in the opening chapter of this book,
Steffe used the terms figurative and operative to mark distinctions in children’s
counting schemes. Whereas figurative counting is freed from perceptual material, it
still relies on carrying out sensorimotor activity (e.g., pointing at imagined locations
of counters hidden under a cloth). In contrast, operative counting operates on inter-
nalized actions themselves.
Following Thompson (1985), the authors extend the original characterizations of
operative and figurative thought to include distinctions made in advanced mathe-
matics. Specifically, they distinguish between the mathematical operations being
coordinated (the operative) and the mathematical objects on which they operate (the
figurative). Although students might be able to unpack those mathematical objects
as coordination of operations themselves, in operating at a higher level, they treat
those objects simply as material to operate upon.
The figurative-operative distinction foregrounds two key aspects of Piaget’s epis-
temology of mathematics. First, by virtue of being objectified, mathematical objects
at the figurative level can be symbolized so that operating upon or manipulating
those symbols serves as a proxy for operating on the mathematical object itself.
Second, the operative level introduces a level above the figurative, demonstrating
how mathematics builds upon itself, with new objects constructed at that higher
level: “What is operative at one level can become figurative at the next level as the
individual attempts to operate on prior abstracted mental actions” (Chapter 4, p. 10).
The chapter continues by elaborating on examples from different research pro-
grams, including research on shape thinking and quantitative reasoning. The authors
make a related distinction, between static shape thinking and emergent shape think-
ing, depending on whether the shape emerges from a coordination of operations.
This dichotomy recalls one Sfard (1992) marked between pseudoobjects and objects
(see also Zandieh, 2000). It also separates those aspects of shapes or graphs that are
operationally necessary from those that simply follow a convention (e.g., positive
numbers are on the right side of the graph).
Conventions, symbols, and axioms are historicocultural products. Within the
milieu of ancient Greece and its mathematical community, Euclid established a set
of five axioms to define Euclidean geometry, inheriting the first three axioms from
Plato’s laws of construction. Even outside of such products, the line between psy-
chogenetic and historicocultural contributions to mathematical development may
remain blurred because history and culture constrain or afford opportunities for
operational development. However, the figurative/operative distinction helps us
focus on the coordination of actions that define mathematical objects, even when
they include geometric objects from non-Euclidean geometries, and even if students
represent them in unconventional ways.
Chapter 5: Images
Thompson, Byerley, and O’Bryan’s chapter introduces the Piagetian construct of an

image, which they define as the representation of experience. They emphasize that
these experiences and the images they support include more than visual informa-
tion; images can include the totality of experience. The authors then relate imagery
to the schemes construct from Chap. 3, noting that images can serve as waypoints
in a scheme’s activity. They use the figurative/operative distinction from Chap. 4 to
explain two levels of imagery important to mathematical development: figurative
imagery and operative imagery.
Figurative imagery supports the sequencing of actions associated with past expe-
rience. As with figurative numbers, a student can use figurative imagery to stand in
place of perceptual material and imagine transforming it through activity. Because
figurative images involve imagined activity, they happen sequentially—in time—
and the student does not always anticipate the result of their imagined activity, espe-
cially if it involves long chains of transformations between images. Because
figurative images are contextualized within past experience, they also might not
generalize well to new mathematical experiences.
In contrast to the sequentiality of figurative images, operative images can repre-
sent results of acting, all at once. These images generalize better because the actions
and results are anticipated so that any of a number of images, across various con-
texts, can stand in their place. Again, we can draw an analogy to number, wherein
operative numbers contain records of counting, not simply figurative material for
counting.
We have seen the figurative/operative dichotomy plays out within a few different
constructs: number, shape, and now imagery in general. The contrast recalls a prin-
cipal distinction Piaget made within his explanation for mathematical development,
from sequential actions to the simultaneous coordination of operations. Images sup-
port that development.
458 A. Norton
Inhelder et al. (1971) dedicated their text, Mental Imagery in the Child, to studies
of the relationship between images and operations. Specifically, Piaget and Inhelder
asked, “To what extent does the anticipatory image depend on operations, and to
what extent does it further the functioning of operations” (p. 313). They determined
that, although images are subordinate to operations, images support their coordina-
tion: “after having structured and fashioned it in their own likeness, the operations
in fact come to depend upon the image” (p. 378). This is because images can mark
states between transformations and can symbolize the transformations themselves.
As stated in the closing paragraph of the text,
The image ensures final analysis of ‘states’, and even aids figurative anticipation of ‘trans-
formations’, in spite of the irreducibly static character of such a configuration. This makes
the image an indispensable auxiliary in the functioning of the very dynamism of thought—
but only as it remains consistently subordinate to such operational dynamism, which it
cannot replace, and which it can only express symbolically with degrees of distortion or
fidelity varying according to circumstances. (p. 390)
The symbolic function of images brings us back to the historicocritical method.

When images refer to states, they provide a “system of reference” in transforming
between states. Moreover, they provide ways for us to communicate transforma-
tions to others by way of initial and final states, as represented in images and
expressed in language.
Any representational cognition presupposes the bringing into play of some symbolic func-
tion. It would be preferable to describe this function as ‘semiotic,’ since it embraces arbi-
trary and social ‘signs’, motivated ‘symbols’, individual ‘symbols’, and social symbols.
Without this semiotic function, thought could not be formulated, and consequently could
not be expressed intelligibly to others or to oneself. (pp. 379–380)
Images serve a semiotic function for the individual, akin to language, but one
that can represent thought outside of language. Individuals can use images to com-
municate thought to themselves, similar to the way they might use language to com-
municate thought to others. Because these images symbolize states between
transformations, if the individual can describe these images in words or pictures (as
we often do in geometry), they can also express those transformations to others. We
have then a means for framing individual mathematical thought, as constituted by
transformations (i.e., operations), within shared formal systems (e.g., Euclidean
geometry, as described by its figures, axioms, and postulates).
Chapter 6: Empirical, Pseudoempirical,

and Reflective Abstraction
Chapter 6 introduces the most critical construct in Piaget’s genetic epistemology,

especially as it applies to mathematical development. If mathematics develops
through a coordination of actions, reorganized as operations, what is the mechanism
for such reorganization? Piaget introduced the novel and mysterious construct of
reflective abstraction to answer this question, which lies at the heart of his genetic
epistemology. The authors of Chap. 6 dissolve some of the mystery by first disen-
tangling it from closely related constructs, like pseudoempirical abstraction and
reflected abstraction.
Otherwise-divergent theories within mathematical education broadly agree that
students’ own activity is the basis for meaningful mathematical learning. Theories
of embodied cognition privilege sensorimotor activity, while sociocultural theories
foreground interaction between students within classroom-specific milieux. The
distinguishing feature of Piaget’s genetic epistemology is that it accounts for
abstraction from sensorimotor action, via the coordination of actions, and that is the
role of reflective abstraction as explicated by the authors of Chap. 6. Citing
Glasersfeld (1982), they explain, “reflective abstraction… occurs when the learner
abstracts mental operations from the sensorimotor context that may have given rise
to them” (p. 7).
Again citing Glasersfeld (1982), the authors describe reflective abstraction as a
two-phase process, consisting of projection and reorganization. Projection refers to
a transition to a higher level of operating, as in the distinction between the two levels
of figurative and operative thought (see Chap. 4): actions at one level become
objects of thought at the next level. Reorganization refers to the coordination of
these actions within a system for composing and reversing them (within group-like
structures discussed in Chap. 7).
Reflective abstraction not only describes the principal means of individual math-
ematical development, it also explains the historical development of mathematical
objects, like numbers and functions. When nineteenth-century French mathemati-
cian, Henri Poincare, claimed that “mathematicians do not study objects, but the
relations between objects” (Chapter 6, p. 20), he was highlighting the unique char-
acter of mathematical objects. They are not objects of perception but, instead, come
to life through relationships themselves. We see this in the history of number, for
example, especially in the sense that numbers only exist within systems for operat-
ing. Starting from the counting numbers, numbers continued to develop, to frac-
tions, negative numbers, and even complex numbers, by extending systems to
include new ways of operating (e.g., taking roots of numbers). Sfard (1991) outlined
a similar trajectory in the history of function.
Chapter 7: Groups and Groupings
If reflective abstraction results in the reorganization of actions, as operations, at a

higher level, groups and groupings describe the structure of that organization. Like
schemes, groups, and group-like structures are researcher constructs used to build
models of students’ ways of operating. Moreover, groups are also constructs in for-
mal mathematics, and this is no coincidence. We might say that the formal structure
of a group is the historical result of a reflected abstraction (the subject of Chap. 8)
460 A. Norton
on our ways of operating across multiple domains of mathematics, including num-

ber theory, geometry, and algebra.
Piaget (1970) specified that logicomathematical operations arise when they are
organized within a closed system in which they become reversible and composable.
He often included a criterion of associativity, which he described as an “indepen-
dence of path” in the transformations induced by operations. Here, then, we have
the four criteria of the formal structure from abstract algebra known as a group. And
here, then, we can see the psychological underpinnings of formal mathematics.
Chapter 7 uses the example of the splitting group to illustrate how children’s
mathematical development ultimately coincides with formal structures. Through
coordination of sensorimotor activity, children develop operations of partitioning
and iterating, which are organized as inverses of one another. Associativity follows
as students begin to coordinate the various units produced by transforming a whole
unit into parts and iterations of parts. Ultimately, we can model their coordinations
with a group structure, called the splitting group, that is isomorphic to the group of
positive rational numbers under the formal operation of multiplication.
Chapter 8: Reflected Abstraction
Chapter 8 complements Chap. 6 by situating reflected abstraction among the other

forms of abstraction. Like reflective abstractions, reflected abstractions operate on
actions at a higher level. Unlike reflective abstractions, they necessarily involve a
conscious awareness of that way of operating, including the organization of actions
at that higher level. “Reflecting abstractions constitute processes, and reflected
abstractions constitute becoming conscious of their results” (Piaget, 2001, p. 205, as
quoted in Chap. 8, p. 17).
Following psychological tradition, Piaget characterized consciousness as
control,2 so reflected abstractions introduce a greater degree of control over our
ways of operating. They do this (at least in part) by relying on the semiotic function,
allowing us to symbolize our ways of operating, and our coordination of actions. As
noted by the authors, this shift parallels the movement from sensorimotor thought to
representational thought. Representing our own ways of operating holds particular
personal power, as foreshadowed in Chap. 6: “On this higher cognitive level, the
subject can consciously manipulate these symbols independently of representing
the coordinated actions they signify, all the while being capable of doing so” (p. 4).
Such is the basis for algebraic reasoning, including abstract algebra, where we find
formalized fundamental structures for logicomathematical operations, such
as groups.
2
For example, Morgan (1894) proclaimed, “the primary aim, object, and purpose of consciousness
is control” (p. 182).
Because the semiotic function includes social signs, such as language, reflected
abstraction also enables us to communicate our ways of operating to others, assum-
ing they have compatible structures for assimilating our language. Drawing on the
example of “funky protractors” from Hardison and Lee (2020; as cited in Chap. 8),
consider the concept of angle measure.
How might you communicate to a friend, via text, a way of measuring angles so
that you could both measure any angle and always be in agreement? Assuming you
understand angle measure as the ratio between the arc length on a circle and the
radius of that circle, and that you have formed a reflected abstraction from that way
of operating, you might meaningfully communicate the concept of a protractor.
However, in order for the semiotic function to work through language, your friend
would need to have formed a similar reflected abstraction, so that your language
refers to a compatible way of operating. If your friend has not formed such an
abstraction, you might instead describe a procedure for determining an angle by
instructing them on how to build and use a protractor. You can do this thanks to
reflected abstraction from your own way of operating, whereas, for your friend,
everything would be reduced to figurative thought—a procedure carried out on figu-
rative material.
Chapter 8 includes explicit connections to figurative thought, operative thought,
and most of the constructs introduced in prior chapters. It also includes comparisons
to other frameworks inspired by Piaget’s genetic epistemology, such as Dubinsky’s
(1991) APOS theory and Harel’s (2007) DNR principles. What unifies these frame-
works and the various constructs introduced in this text is the central idea that all
mathematics—even formal mathematics—has as its source reflections on, and
abstractions from, our own mental actions.
Chapter 9: Decentering
Picking up where Chap. 8 left off, Chap. 9 describes a shift, called decentering,
from focusing on our own ways of operating, to focusing on another person’s ways
of operating. This shift in focus relies on reflected abstraction. Becoming aware of
our own ways of operating, via reflected abstraction, allows us to set it aside and
consider other ways of operating. “The process of comparing the characteristics of
one’s reasoning to the characteristics of another’s reasoning necessarily requires
that the individual has developed or is developing conscious awareness of their own
reasoning” (p. 8).
Piaget (1972) referred to decentering as a process that makes possible the repre-
sentation of thought. At early stages, it allows the subject to become aware of them-
self as an actor, differentiated from the objects on which they act; or even becoming
aware of their own body as “one object among others in a space containing them all”
462 A. Norton
(p. 22). Here, we focus on the decentering as the subject’s explicit awareness of
themself as one actor among others.3
Decentering enables us to communicate more effectively, especially in teaching,
where there is always a disparity between the knowledge of the teacher and their
students. Teachers regularly engage in communication with students wherein their
respective mathematical constructs seem incompatible. To teach effectively, partic-
ularly in such situations, the teacher needs to understand the student’s mathematical
ways of operating. The teacher needs to construct new ways of operating, which is
to say that, in teaching mathematics effectively, teachers are necessarily learning
mathematics from their students.
Chapter 9 describes decentering as a precursor for teachers’ continued learn-
ing of mathematics, specifically mathematical knowledge for teaching (Silverman
& Thompson, 2008). Figure 4 (p. 37) illustrates a cyclical process by which a
teacher can build models of their students’ mathematics, which then become the
basis for curriculum and instruction. First, the teacher becomes aware of their
own ways of operating, then they set them aside (decentering) in order to consider
how their students may be operating. These alternative ways of operating are
called second-order models, to distinguish them from the first-order models of the
teacher’s own mathematics, which result from reflected abstraction. The cycle
continues as these second-order models become part of the teacher’s own (first-
order) mathematics.
In addition to second-order models, the chapter also discusses the related con-
struct of epistemic students, first introduced in Chap. 1. Both constructs are revis-
ited in Chap. 11, but in Chap. 9, they are extended to research on teacher education.
Here, when the classroom is viewed as a microcosm for the historical development
of mathematics, we see some of the interplay between psychogenetic and historico-
critical themes in Piaget’s genetic epistemology. On the one hand, mathematics
arises through a coordination of actions, and even a teacher’s (second-order) models
of students’ mathematics involve such a coordination. On the other hand, we begin
to see how a teacher’s decisions, as influenced by both their mathematics and their
models of students’ mathematics, foster a classroom culture that can afford or con-
strain opportunities for students’ continued mathematical development.
3
Note that these two kinds of decentering are closely related for Piaget: “for the understanding of
other people as well as for the understanding of the outside world, two conditions are necessary:
(1) consciousness of oneself as a subject, and the ability to detach subject from object so as not to
attribute to the second the characteristics of the first; (2) to cease to look upon one’s own point of
view as the only possible one, and to coordinate it with that of others” (Piaget, 1959, p. 277, as
cited in Fox & Riconscente, 2008).
Chapter 10: Logic
Prior chapters have focused on mathematical development, but Piaget’s genetic

epistemology made little distinction between logic and mathematics, often referring
to operations as logicomathematical operations. Much of Piaget’s work specific to
children’s development of logic was conducted in partnership with his close col-
league, Barbel Inhelder, and its findings largely parallel those within various math-
ematical domains, including space and number. As with mathematics, they
understood the development of logic as a product of psychology, and just as formal
structures like groups were used to model later developments in mathematics, for-
mal logic was used to model later developments in the logic of proof and proving.
With regard to the two Piagetian themes highlighted in the present chapter, Chap.
10 brings them to logic by, first, explicating how logic arises through the coordina-
tion of mental actions and, second, explaining the relationship between historico-
cultural product of formal logic and children’s logical development. In making both
arguments, Dawkins homes in on children’s constructions of logical implications:
In contrast to formal logic that operates on a meta-language that is partitioned off from
actual states of affairs by truth-values, Piaget seeks to explicate the implications entailed in
actual processes of activity and reasoning. In particular, the anticipation that an effect will
result from an action (“if I do this, then that will occur”) is an implication for the reasoner
that, whether or not it is true to fact, tells the observing researcher about the child’s mean-
ings by which they draw such inferences. (p. 23)
Formally, we think about logical implications as structures independent of time

and context, and we sometimes assume that such structures are inherent in human
thought, as if from birth. However, just as Piaget demonstrated for time and space,
logic must be constructed on the basis of actions we perform within the worlds we
experience. How might we act in the world, and what effects might we expect as a
result of those actions? How might we coordinate mental actions—especially by
composing and reversing them—to anticipate those effects? Dawkins explains how
such intentions in action generate extensions, expressed as sets, like P and Q,
wherein we might formally say that proposition p implies proposition q because P
is contained in Q. Far from innate, this kind of formalism follows a period of devel-
opment that spans the life of the child (Inhelder & Piaget, 1958, 1964), similar to the
way that the group construct (see Chap. 7) emerges psychologically and historically
at the end of a series of reflective and reflected abstractions.
Chapter 11: Units Coordination
The final two chapters in Part I focus on constructs motivated by Piaget’s genetic
epistemology but contributed by the mathematics education research community
itself. Chapter 11 introduces the construct of units coordination, which describes
ways children construct and transform units, beginning from units of 1. As explained
464 A. Norton
by the authors, units coordination activity permeates mathematical domains from

counting to algebraic reasoning, and it provides the basis for multiplicative
reasoning.
We can find an early definition of unit in Book VII of Euclid’s Elements: “A unit
is that by virtue of which each of the things that exist is called one.” We can find
traces of children’s units coordinating activity throughout Piaget’s studies. However,
as a research construct, units coordination began within a research program carried
out by Les Steffe and colleagues (e.g., Steffe, 1992; Steffe et al., 1983).
As Steffe recounted in the opening chapter of this text, he had tried to apply
Piaget’s theory on the construction of number to teaching experiments but found the
framework too limiting.
The finding that counting was the children’s basic method of solving arithmetical tasks led
to abandoning attempts to apply Piaget’s idea that number is constructed as a synthesis of
classification and seriation operations and to investigating children’s construction of num-
ber sequences in the context of their spontaneous use of counting in solution of arithmetical
tasks in teaching experiments (Steffe et al., 1983; Steffe & Cobb, 1988; Steffe, 1991, 1994).
(Chapter 2, p. 4)
Informed by, but in place of, Piaget’s synthesis of classification and seriation,
Steffe and colleagues identified a progression of number sequences that children
construct through their counting activity.
In Chap. 11, Hackenberg and colleagues relate these successive number
sequences to stages of units coordination. The first of these number sequences, the
initial number sequence, consists of numbers that contain records of acts of count-
ing. As such, 5, for example, stands in place of the coordinated action of counting
“one, two, three, four, five,” including the one-to-one correspondence between those
number words and objects being counted. In subsequence number sequences, such
as the tacitly nested number sequences and the explicitly nested number sequence,
numbers become nested so that each number contains the numbers that precede it.
Ultimately number sequences are researcher constructs. We learn these ways of
operating from children as we build second-order models of their mathematics. We
generalize those models into epistemic students that replace the states that Piaget
had borrowed from the history of mathematics (e.g., class inclusion and ordering
relations) in following his historicocritical method. Such formal constructs pro-
vided Piaget with a framework for studying transformations from one state to the
next. Now the states themselves can be understood as ways of transforming units.
In sum, the units coordinating construct, as conceived by Steffe (1992) and
adopted by a growing number of mathematics education researchers, contributes to
a unified theory of mathematical knowing in at least two key ways. First, it explains
the development of large swaths of mathematical knowledge in terms of the ways
children construct and transform units. Second, it fulfills Piaget’s vision for a com-
plete genetic epistemology by replacing formal mathematical states, as inherited
from the history of mathematics, with stages of mathematical development, as
learned from children.
Chapter 12: Quantitative and Covariational Reasoning
Finally, Chap. 12 focuses on constructs that extend Piaget’s genetic epistemology

program to more advanced mathematical reasoning, especially with regard to con-
tinuous functions. These constructs relate centrally to the idea of quantities whose
values vary in relation to one another: covariational reasoning. They explain how
children coordinate quantities that covary, as in measuring rates of change. Much of
the foundational work in this area was conducted by Marilyn Carlson, Pat Thompson,
and their students.
Carlson and colleagues (2002) provided a framework for understanding stu-
dents’ construction of covariational reasoning. This framework describes levels of
development and specifies the kinds of mental actions involved at each level. In
Chap. 12, Boyce summarizes this framework (see Table 3), along with finer-grained
distinctions in ways students reason covariationally, such as “chunky” versus
“smooth” reasoning (Castillo-Garsow, 2012). Just as units coordination contributes
to our understanding of students’ mathematics across number and early algebra,
these new constructs support efforts to build models of students’ mathematics
related to advanced mathematical topics, such as integral Calculus, differential
equations, and analysis. In this sense, covariational reasoning is another unifying
construct informed by Piaget’s genetic epistemology.
With a common focus on measuring quantities, we should expect close interac-
tions between students’ construction of units coordination and their construction of
covariational reasoning. However, little work has been conducted to study potential
relationships between these domains. Indeed, such work presents one of many
future directions for continued research in genetic epistemology.
Future Directions for Genetic Research
In this chapter, we have focused on two neglected themes in Piaget’s genetic episte-
mology: his use of the historicocritical method and his central focus on building a
comprehensive epistemology of mathematics. Each of the chapters in Part I illus-
trates those unifying themes while introducing specific neo-Piagetian constructs.
Each chapter closes with suggestions for how those various constructs might be
applied to future research in mathematics education. In closing this chapter, we
highlight potential avenues for future research in mathematics education, as
informed by Piaget’s genetic epistemology. These avenues connect and extend con-
structs while interrogating their explanatory power in building models of students’
mathematics.
466 A. Norton
hat Are the Relationships Between Units Coordination

W
and Covariational Reasoning?
Covariational reasoning refers to the coordination of two covarying quantities, and

(extensive) quantities rely on a unit for measuring them. Whatever we take as a unit
becomes 1, and when 1 becomes an iterable unit, every other number in the number
sequence becomes some measure of 1 s (Steffe, 1992). So, theoretically, students’
number sequences and units coordinating activity should undergird their covaria-
tional reasoning. However, the constructs of units coordination and covariational
reasoning were developed independently, for modeling students’ mathematics in
complementary domains (see Chaps. 11 and 12). Research connecting them empiri-
cally has only recently begun to emerge.
In one quantitative study, Boyce et al. (2021) found that college students who
operated at the highest stage of units coordination performed significantly better on
assessments of calculus readiness. A separate study by Byerley (2019) achieved
similar findings and included a qualitative analysis that anticipated those findings by
explaining the role of units coordination in understanding rates of change. That
explanation aligns with results from previous studies suggesting a relationship
between students’ units coordination and their conceptualizations of rates of change
(e.g., Ellis et al., 2015). Future research might explicate those relationships by using
tasks informed by each framework—stages of units coordination and levels of
covariational reasoning—within quantitative studies using written instruments or
within teaching experiments.
ow Can We Support Students’ Stagewise Development

H
of Units Coordination?
Units coordination plays a critical role in students’ mathematical development,

from counting to algebra, and (via its emerging connections to covariational reason-
ing) into calculus. So, as mathematics educators, we should be concerned with stu-
dents’ development of units coordination from one stage to the next. So far, few
such studies exist.
Working with a first-grade student named Tyrone, Steffe (1991) supported
Tyrone’s construction of an initial number sequence by drawing on his ability to use
figurative number patterns to count on. Tyrone had counted two collections of chips
and was tasked with determining their total number after they were hidden under
two cloths. Over time, he began counting on from the number of chips under the
first cloth while pointing to five imagined items under the second cloth, by relying
on a figurative pattern for 5. Steffe’s questions about Tyrone’s activity prompted
reflection on that activity, and Tyrone began counting on in new situations, indicat-
ing the construction of an initial number sequence. Norton and Boyce (2015)
seemed to have similar success in supporting a sixth-grade students’ construction of
an explicitly nested number sequence by engaging him in sensorimotor activity that

involved working across levels of units in various ways, but that success was limited
to contexts involving discrete quantities.
The field needs more studies that investigate ways teachers and researchers can
provoke development from one number sequence to the next because that develop-
ment corresponds to development from one stage of units coordination to the next,
and units coordination pervades so much of students’ mathematics. Initial studies
might consist of teaching experiments with individual students, like the studies
summarized here. Later studies could consider the effects of whole-class instruction
on students’ stagewise development.
an We Specify the Coordination of Actions That Constitute

C
Reflective Abstraction Across Various Mathematical Domains?
The construction of logicomathematical operations, through reflective abstraction,

lies at the heart of Piaget’s epistemology of mathematics. Recall from Chap. 6, that,
at its simplest, reflective abstraction involves the abstraction of logical-mathematical
operations from sensorimotor activity. This abstraction occurs by projecting senso-
rimotor activity to the level of internalized (imagined) actions and then organizing
those actions within a structure for composing and reversing them. Note that reflec-
tive abstraction could also include cases where the abstracted actions are already
operations, but the simplest case may be the best place to start.
The simple case of reflective abstraction fits the example of Tyrone, above.
Tyrone’s construction of an initial number sequence constitutes a reflective abstrac-
tion because of its stagewise shift in his ways of operating. Furthermore, the way
this reflective abstraction seemed to occur fits descriptions from Chap. 6. Specifically,
he seemed to project the sensorimotor activity associated with counting-on to the
level of imagined activity, which he could perform on figurative material. By reflect-
ing on that imagined activity, he seemed to organize those imagined actions within
an initial number sequence, wherein each number contains a record of counting
from 1, including the possibility of continuing a count (counting on) from that num-
ber. Thus, the initial number sequence would allow Tyrone to additively compose
two numbers and even to begin considering the reversal of that composition, though
a complete reversal would involve disembedding one number from a larger number
in which it is contained, which falls in the purview of the explicitly nested number
sequence (see Chap. 11).
Tyrone’s construction of INS represents a powerful reorganization but one that
lacks formal reversibility (i.e., Tyrone could count on but not to count back). Steffe
(personal communication) abandoned much of the formalism contained within
Piaget’s genetic epistemology program, including the group structure discussed in
Chap. 7. Piaget’s epistemology relied on the group structure to describe the com-
posability and reversibility of operations, which in turn explain the fecundity and
468 A. Norton
reliability of mathematics. Steffe abandoned its use because it was too limiting in
understanding students’ construction of number sequences. Specifically, students
like Tyrone can engage in activity that Steffe deemed mathematical but not yet
reversible.
In modeling students’ development of logicomathematical operations, via reflec-
tive abstraction, we as researchers are obliged to account for the ways that actions
become organized at higher levels. Ultimately, this organization should specify the
reversibility of each operation, as group structures do (consider the example of the
splitting group, discussed in Chap. 7). However, along the way to that ultimate
development, we need ways to model children’s organizations of actions that might
not yet be reversible. Indeed, this was the purpose of Piaget’s group-like structures.
Can we use such structures, or other means, to model the organization of mental
actions that constitute children’s mathematics across various domains? If so, those
second-order models could be useful to teachers and researchers in simulating chil-
dren’s mathematics so that they can respond to it. As indicated in Chap. 10, such
models could even simulate children’s logic.
hat Is the Appropriate Role of Formalization

W
in Mathematics Education?
Chapters 4, 5, and 8 allude to the role of symbols in mathematics. Chapter 4 dis-

cusses the use of figures as material for operating; Chap. 5 discusses the role of
images as symbols; and Chap. 8 introduces the semiotic function, which becomes
available when we become aware of the products of our reflective abstractions. This
latter idea—reflected abstraction—is particularly powerful in the reasoning of pro-
fessional mathematicians because it allows them to formalize their ways of operat-
ing; that is, to represent their ways of operating symbolically and to manipulate
those symbols as a proxy for operating further, at higher and higher levels of abstrac-
tion. As mathematics educators, should be able to trace the development of this
operational power back to children’s early uses of figures and images; and intention-
ally support its development as it manifests itself in children’s meaningful use of
abstract symbols.
Algorithms constitute early uses of symbols. These algorithms include student-
invented algorithms, which students might find more meaningful because they
closely relate to students’ ways of operating. Beginning with the initial number
sequence, numerals can serve as proxies for acts of counting, and the addition of
two numerals might symbolize an act of counting on. With multidigit addition, stu-
dents can rely on algorithms to serve as proxies for such units coordination. We
should expect the meanings for symbols, including numerals, to evolve as students
construct more powerful ways of operating, and we should support developmentally
appropriate uses of those symbols to offload the cognitive demands of coordinating
long sequences of operations.
If the semiotic function becomes available upon reflected abstraction of our ways
of operating, we can use that framework to study developmentally appropriate intro-
ductions of abstract symbols. Such studies should especially include early algebra.
Indeed, some such studies have already occurred (e.g., Tillema & Hackenberg,
2011). What we are suggesting here is that Piaget’s genetic epistemology—espe-
cially in using constructs like reflected abstraction and in making distinctions
between figurative and operative thought—provides an overarching framework for
organizing those studies, from early childhood through college.
ow Might Teachers Assess Students’ Available Mental

H
Actions and Model Their Coordination as Reversible
and Composable Operations?
As with all scientific models, there is intellectual merit in building second-order

models of students’ mathematics. When built using a Piagetian framework, those
models hold additional, epistemic value, because such a framing suggests that stu-
dents’ mathematical development can help us better understand the historical devel-
opment of mathematics itself. After all, from a genetic epistemology perspective,
students’ mathematics is mathematics.
Still, mathematics education researchers ultimately concern themselves with
determining how knowledge of students’ mathematics can help us improve learning
opportunities in the classroom. For that, teachers will need to act as researchers,
building second-order models of their own students’ mathematics. Noting the
numerous intensive and longitudinal teaching experiments cited in this volume,
each conducted within a single mathematical domain (e.g., whole number, frac-
tions, and algebra), it does not seem feasible to call on classroom teachers to con-
duct similar work. However, teachers can benefit from constructs developed through
prior work, including reliance on epistemic students (see Chap. 9).
A few related professional development programs have begun to emerge. For
example, the Student-Adaptive Pedagogy Project (AdPed) led by Ron Tzur engaged
elementary school teachers in building models of students’ development of multi-
plicative reasoning (Tzur et al., 2018). To support those efforts, the AdPed team
introduced teachers to multiplicative reasoning schemes that the teachers might use
to better understand the mathematical ways of operating among students in their
own classrooms. Project results include reports on “manifestations of elementary
mathematics teachers’ shift toward second-order models” (Hodkowski, 2018, p. 1).
Relying on Carlson’s decentering framework (see Chap. 9), Bas-Ader and Carlson
conducted a similar study with college mathematics instructors, and Liang (2019)
relied on that same framework in a study of preservice teachers and the second-
order models they constructed of students’ mathematics.
This volume on genetic epistemology might challenge us, as mathematics teacher
educators, to push second-order models further, to specifying specific mental
470 A. Norton
actions that constitute students’ mathematics across various domains. In the domain
of number, we might rely on “unit transformation graphs” to illustrate graphically
the various units students construct and the various mental actions students use to
transform those units, while solving mathematics tasks, such as fractions multipli-
cation tasks (Norton et al., in press). Illustrating students’ mathematics graphically
could help teachers focus on students’ available mental actions and render the pro-
cess of model building more visual and tangible.
Second-order models of students’ mathematics could inform teachers’ instruc-
tional decisions, such as choosing tasks and making decisions about the appropriate
use of technology. Specifically, teachers and researchers might ask, what actions
(and coordinations thereof) are afforded by, and constrained by, various instruc-
tional technologies (Biddlecomb, 1994; Olive, 2000)? Additionally, mathematics
educators who research teacher education might be interested in building models of
teachers’ second-order models of students’ mathematics—something we might call
a third-order model (Wilson et al., 2011).
Conclusion
Piaget left the mathematics education community with the legacy of his genetic
epistemology. We build upon his legacy as we build models of students’ mathemat-
ics, from childhood through advanced mathematical study. The constructs presented
in this book and reviewed in this chapter provide us with tools for conducting this
work. What unifies those constructs and our collective work is the fundamental
principle of Piaget’s epistemology of mathematics, which is the idea that mathemat-
ics itself arises from the coordination of reversible and composable mental actions,
which Piaget termed logicomathematical operations. This principle is embedded in
each of the constructs and underlies even formal mathematical constructs passed on
to us through the history of mathematics (e.g., formal logic and algebraic groups).
By explicitly identifying the mental actions students rely upon to construct mathe-
matics in various domains, from counting through Calculus, we are reclaiming all
of mathematics as students’ mathematics.
References
Beth & Piaget. (1966). Mathematical epistemology and psychology. (Trans. W. Mays). Gordon
and Breach.
Biddlecomb, B. D. (1994). Theory-based development of computer microworlds. Journal of
Research in Childhood Education, 8(2), 87–98.
Boyce, S., Grabhorn, J. A., & Byerley, C. (2021). Relating students’ units coordinating and calcu-
lus readiness. Mathematical Thinking and Learning, 23(3), 187–208.
Byerley, C. (2019). Calculus students’ fraction and measure schemes and implications for teach-
ing rate of change functions conceptually. Journal of Mathematical Behavior, 55. https://doi.
org/10.1016/j.jmathb.2019.03.001
Castillo-Garsow, C. (2012). Continuous quantitative reasoning. In R. L. Mayes & L. L. Hatfield
(Eds.), Quantitative reasoning and mathematical modeling: A driver for STEM integrated
education and teaching in context (Vol. 2, pp. 55–73). Available at http://www.uwyo.edu/wis-
dome/_files/documents/Castillo_Garsow.pdf
Chapman, M. (1988). Constructive evolution: Origins and development of Piaget’s thought.
Cambridge University Press.
Dubinsky, E. (1991). Reflective abstraction in advanced mathematical thinking. In D. Tall (Ed.),
Advanced mathematical thinking (pp. 95–123). Kluwer.
Einstein, A. (1954). Ideas and opinions (S. Bargmann, Trans.). Bonanza Books (Original work
published 1921).
Ellis, A. B., Özgür, Z., Kulow, T., Williams, C. C., & Amidon, J. (2015). Quantifying exponential
growth: Three conceptual shifts in coordinating multiplicative and additive growth. Journal of
Fox, E., & Riconscente, M. (2008). Metacognition and self-regulation in James, Piaget, and
Vygotsky. Educational Psychology Review, 20(4), 373–389.
sion of the splitting hypothesis. Journal of Mathematical Behavior, 26, 27–47.
Hardison, H. L., & Lee, H. Y. (2020). Funky protractors for exploring angle measure. Mathematics
Teacher: Learning and Teaching PK-12, 113(3), 229–232.
Harel, G. (2007). The DNR system as a conceptual framework for curriculum development and
instruction. In R. Lesh, J. J. Kaput, & E. Hamilton (Eds.), Foundations for the future in math-
ematics education. Routledge.
Hodkowski, N. M. (2018). Manifestations of elementary mathematics teachers’ shift towards
second-order models (Doctoral dissertation, University of Colorado at Denver).
essay on the construction of formal operational structures. Basic Books.
Inhelder, B., & Piaget, J. (1964). The early growth of logic in the child. W. W. Norton.
Inhelder, B., Piaget, J., Bovet, M., Frank, F., & Étienne, A. (1971). Mental imagery in the child: A
study of the development of imaginal representation. Routledge & Kegan Paul.
Kitchener, R. F. (1981). The nature and scope of genetic epistemology. Philosophy of Science,
48(3), 400–415.
Kitchener, R. F. (1985). Genetic epistemology, history of science and genetic psychology. Synthese,
65(1), 3–31.
Liang, B. (2019). A radical constructivist model of teachers’ mathematical learning through
student-teacher interaction. In Proceedings of the 41st annual meeting of the North American
Chapter of the International Group for the Psychology of Mathematics Education (Vol. 11,
pp. 1814–1819).
Morgan, C. L. (1894). Introduction to comparative psychology. Walter Scott.
Norton, A., & Boyce, S. (2015). Provoking the construction of a structure for coordinating n+1
levels of units. Journal of Mathematical Behavior, 40, 211–242. https://doi.org/10.1016/j.
jmathb.2015.10.006
Norton, A., Ulrich, C., & Kerrigan, S. (in press). Unit transformation graphs: Modeling the cogni-
tive demands of mathematical tasks. Journal for Research in Mathematics Education.
Olive, J. (2000). Computer tools for interactive mathematical activity in the elementary school.
International Journal of Computers for Mathematical Learning, 5(3), 241–262.
Piaget, J. (1950). Introduction à l'épistémologie génétique Vol. I: La pensée mathématique. Presses
Universitaires de France.
472 A. Norton
Piaget, J. (1959). The language and thought of the child (3rd ed.) (M. Gabain, Trans.). Routledge
& Kegan Paul.
Piaget, J. (1970). Structuralism (C. Maschler, Trans.). Basic Books (Original work published 1968).
Piaget, J. (1972). Principles of genetic epistemology (W. Mays, Trans.). Routledge & Kegan Paul
Piaget, J. (1973). The child and reality: Problems of genetic psychology (A. Rosin, Trans.).
Grossman Publishers.
Piaget, J. (1977). Foreword to H. In E. Gruber & J. Voné che (Eds.), The essential Piaget. Routledge
and Kegan Paul.
Saxe, G. B., & Farid, A. M. (2021). The interplay between individual and collective activity:
An analysis of classroom discussions about the Sierpinski triangle. International Journal of
Research in Undergraduate Mathematics Education, 1–34.
Sfard, A. (1991). On the dual nature of mathematical conceptions: Reflections on process and
objects on different sides of the same coin. Educational Studies in Mathematics, 22(1), 1–36.
Sfard, A. (1992). Operational origins of mathematical objects and the quandary of reification—
The case of function. In G. Harel & E. Dubinsky (Eds.), The concept of function: Aspects of
epistemology and pedagogy (pp. 59–84). Mathematical Association of America.
Silverman, J., & Thompson, P. W. (2008). Toward a framework for the development of math-
ematical knowledge for teaching. Journal of Mathematics Teacher Education, 11(6), 499–511.
https://doi.org/10.1007/s10857-008-9089-5
Simon, M. A., & Tzur, R. (2004). Explicating the role of mathematical tasks in conceptual learning:
An elaboration of the hypothetical learning trajectory. Mathematical Thinking and Learning,
6(2), 91–104.
Smith, L. (1993). Necessary knowledge: Piagetian perspectives on constructivism. Lawrence
Erlbaum Associates.
Steffe, L. P. (1991). Stages in the construction of the number sequence. In C. Meljac & J. B. Bideaud
(Eds.), Les Chemins du nombre. Editions des Presses Universitaires de Little. (Book published
by Lawrence Erlbaum in English, 1992).
Steffe, L. P. (1994). Children's construction of meaning for arithmetical words: A curricu-
lum problem. In D. Tirosh (Ed.), Implicit and explicit knowledge: An educational approach
(pp. 131–168). Ablex.
Steffe, L. P. (2010). Perspectives on children’s fraction knowledge. In L. P. Steffe & J. Olive (Eds.),
Springer-Verlag.
in developing mathematics curricula. In E. Silver (Ed.), Teaching and learning mathematical
problem solving: Multiple research perspectives (pp. 189–243). Erlbaum.
Tillema, E. S., & Hackenberg, A. J. (2011). Developing systems of notation as a trace of reasoning.
For the Learning of Mathematics, 31, 3.
Tzur, R., Johnson, H. L., Hodkowski, N. M., Jorgensen, C., Nathenson-Mejia, S., Wei, B., Smith,
A., & Davis, A. (2018). Impact of a student-adaptive pedagogy PD program on students’ mul-
tiplicative reasoning. North American Chapter of the International Group for the Psychology
of Mathematics Education.
von Glasersfeld, E. (1982). Subitizing: The role of figural patterns in the development of numerical
concepts. Archives de Psychologie, 50, 191–218.
von Glasersfeld, E. (1993). Learning and adaptation in the theory of constructivism. Communication
and Cognition, 26(3/4), 393–402.
von Glasersfeld, E. (1997). Homage to Jean Piaget (1896–1982). The Irish Journal of Psychology,
18(3), 293–306.
von Glasersfeld, E., & Steffe, L. P. (1991). Conceptual models in educational research and prac-
tice. The Journal of Educational Thought, 25(2), 91–102.
Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes.
Harvard University Press.
Wigner, E. P. (1990). The unreasonable effectiveness of mathematics in the natural sciences. In
Mathematics and science (pp. 291–306).
Wilson, P. H., Lee, H. S., & Hollebrands, K. F. (2011). Understanding prospective mathematics
teachers’ processes for making sense of students’ work with technology. Journal for Research
in Mathematics Education, 42(1), 39–64. https://doi.org/10.5951/jresematheduc.42.1.0039
Zandieh, M. (2000). A theoretical framework for analyzing student understanding of the concept
of derivative. CMBS Issues in Mathematics Education, 8, 103–127.
Chapter 14
Second-Order Models as Acts of Equity
Amy J. Hackenberg, Erik S. Tillema, and Andrew M. Gatza
In 2017, the Research Committee of the National Council of Teachers of Mathematics

called for all mathematics education researchers to engage in the intentional, collec-
tive, and professional responsibility of becoming equity researchers (Aguirre et al.,
2017). To respond to this call, we consider two conceptions of equity in relation to
doing research in mathematics education. In the first, researchers articulate how
their research addresses one or more equity-related issues. In the second, research-
ers articulate how equity is foundational to their research. Researchers have
expressed concerns that following the first conception may only superficially
address issues of equity because these issues are added on to a research program
without considering the principles, assumptions, and aims of the program (Barajas-
López & Larnell, 2019). In this chapter, we make an account grounded in the second
conception of equity. That is, we view equity as foundational to the approach to
research in mathematics education that is discussed in this book. In this chapter, we
explain why.
A main feature of this book is to describe and demonstrate theoretical constructs
that support researchers to make second-order models of students’ mathematics
(Steffe et al., 1983). These models are explanatory accounts of particular students’
mathematical reasoning and learning that can be used to orchestrate future mathe-
matical interactions with other students. Although we view the making and using of
second-order models to be acts of equity in research settings, we do not know of any
place where that position has been articulated (cf. Ellis, 2022). Indeed, we think it
A. J. Hackenberg (*) · E. S. Tillema

e-mail: ahackenb@iu.edu; etillema@iu.edu
A. M. Gatza
Department of Mathematical Sciences, Ball State University, Muncie, IN, USA
e-mail: amgatza@bsu.edu
https://doi.org/10.1007/978-3-031-47386-9_14
476 A. J. Hackenberg et al.
is important for researchers who make these models to define what they mean by an
act of equity (Gutiérrez, 2002), make explicit how making and using these models
can support acts of equity in mathematics education, and identify how these models
could be enhanced to better support acts of equity. We take on these three tasks in
this chapter. Then, we turn to recommendations and directions for future research
that need to be undertaken so that second-order models may better support acts
of equity.
Before moving on, we point out two issues. First, there are many outcomes one
could work for with regard to equity. We are focused on the outcomes of supporting
learning and opportunities for learning because that is our primary research domain.
Second, we view the call to become equity researchers as an active, incomplete
process of becoming. This stance means that there is always room to grow, learn,
and articulate more. We invite the reader to think in these terms because it suggests
that there is always more to learn about how equity is or can be foundational to one’s
research in mathematics education.
Defining Acts of Equity
We now take on our first task, to define what we mean by an act of equity. In order
to do so, we organize themes in research on equity issues in the field, highlighting
those relevant for learning. These themes include identifying structural and institu-
tional inequities in the field (e.g., Battey & Franke, 2015; Berry, 2015), disrupting
deficit narratives in order to focus on asset-based framings of knowing (e.g.,
Adiredja, 2019; Gholson & Martin, 2014; Maloney & Matthews, 2020; Stinson,
2013), understanding the impact of people’s identities on mathematical learning
(e.g., Walshaw, 2013; Shah, 2017), supporting students to learn mathematics as a
tool for inquiry into issues of social justice (e.g., Gutstein, 2016; Kokka, 2022), and
reconsidering power and access by broadening what counts as mathematical knowl-
edge and who generates it (e.g., Gutiérrez, 2018).
Given this background, we conceive of equity to be about how power in society
affects people (Ball, 2013; Foucault, 1980; Hayward, 1998; Walshaw, 2013) com-
bined with respect for people in all of their varied ways of being and identifying in
the world (Gutiérrez, 2018; Walshaw, 2013). We conceive of power as the “social
boundaries that, together, define fields of action for all actors” (Hayward, 1998,
p. 12).1 Social boundaries include “laws, rules, norms, institutional arrangements,
social identities, and exclusions” (p. 2). Our definition of power leads to a particular
view of freedom: “the capacity to act upon the boundaries that constrain and enable
social action, for example, by changing their shape or direction” (p. 12).
1
From the perspective of radical constructivism, a teacher or researcher does not have access to
what fields of actions for all actors are but to the teacher’s or researcher’s interpretation of
those fields.
14 Second-Order Models as Acts of Equity 477
We define respect as “an act of giving particular attention, including high or spe-
cial regard” (Merriam-Webster dictionary online). We view respect for people as an
important component of what the literature on rehumanizing mathematics educa-
tion (Gutiérrez, 2018) calls for: respect for students in mathematics classrooms. As
an example, in a dialogue about race and racism in mathematics education, Martin
repeatedly called for mathematics education researchers to get to know, deeply, the
Black children and teachers with whom they are engaged in research (D’Ambrosio
et al., 2013a). When pressed on whether by this call he meant respect, he said,
“Exactly. To be able to understand the social realities that individuals and collec-
tives from these groups experience” (p. 27). Similarly, Gutiérrez (2018) has called
for students to be regarded as doers of mathematics and as whole people (p. 1). She
stated, “Not until we seek to stand in the shoes of our students, to understand their
conceptions, will we be on the path toward recognizing and embracing their human-
ity” (p. 2). We concur with this view and seek to show in this chapter what we
mean by it.
The focus on respect for all people means taking an interest in how people iden-
tify themselves and are identified by others in society. A person’s identity refers to
the person’s ways of associating themselves with or against mathematics (Cobb
et al., 2009; Martin, 2000), as well as with (or outside of) a racial group, culture,
gender, etc. A person’s identities are changeable and partial, and what a person
identifies with, as well as how others perceive and conceive of the person, is not
fixed (Walshaw, 2013). Furthermore, a person’s identities are partly shaped by the
expectations of others and norms of the institution or environment or, as Hayward
(1998) puts it, by social boundaries. People have partial agency over their identities
within these social boundaries (Cobb et al., 2009). In addition, one’s individual
sense of agency may be directly linked to the extent to which a person experiences
freedom, as defined above (Hayward, 1998).
We note here that power, freedom, respect, and identity all have multiple layers:
We can think of power and freedom in a classroom, as well as in larger systems like
communities, states, and countries. We can think of respect for people and their
identities individually and in terms of how people have been identified, grouped,
and treated in schools, communities, cities, and countries over time (historically).
With these definitions in mind, we define an act of equity to be acting on social
boundaries with the intent of changing their shape or direction in order to address
known inequities. We note that this definition is very close to our definition of free-
dom, but we view freedom as a capacity or potential, whereas an act of equity is an
action. To complete an act of equity requires identifying how the changes produce
more equitable outcomes (i.e., intent alone is insufficient). For example, in mathe-
matics classrooms, students are often trained to learn the thinking of their teacher,
as we will explain in this chapter. The first part of an act of equity (i.e., the intent)
would be changing that norm so that students’ mathematical ways of thinking, and
especially the thinking of students at the margins, are a central and sustained subject
matter of the class. We would consider these changes to be theoretical until an actor
(e.g., a teacher-researcher) establishes markers that the changes produce more
equitable outcomes. Because we are focused on mathematical learning, this

documentation would focus on how the change in social boundaries positively

affected the learning of the students or the opportunities for student learning.
One way to act upon social boundaries is to demonstrate respect for people (stu-
dents) at a personal or individual level, as well as at a historical level. For example,
in the classroom, an act of equity would be listening to students to hear their think-
ing, not assuming that it is the same as one’s own thinking, and yet viewing it as a
set of ideas that has coherence for the students. This act demonstrates respect pri-
marily for students as individuals. An act of respect at a historical level would be
learning the history of racial groups in one’s community in order to understand the
families and students outside of the school setting (D’Ambrosio et al., 2013b). One
outgrowth of that learning might be recognizing that Black or Latinx students in a
class likely have experienced individual and structural discrimination in schools and
that those experiences may exacerbate frustrations they may feel in math class
(Maloney & Matthews, 2020; Mendoza et al., 2021). Again, we would consider
these changes to be theoretical until a teacher-researcher establishes markers that
the changes produce more equitable outcomes, positively affecting student learning
or opportunities for learning.
To be able to engage in acts of equity as a researcher requires, at the very least,
knowledge of historical inequities, knowledge of how one has participated in and
benefited from the institutions of one’s society, and a willingness to question and
continue to learn about both. Within the US, we consider these institutions to have
been infused with racial, gender, linguistic, and other biases, and so, we offer brief
information about our positionalities (D’Ambrosio et al., 2013c). We are former
public middle and high school teachers who are white, cisgender, and monolingual.
We have all participated in groups working for equity in schools that have been led
by both scholars of color and white scholars. All of us have taught racially, socio-
economically, linguistically, geographically, gender-, and cognitively diverse stu-
dents, both as former classroom teachers and in our research. All of us studied issues
of equity during our academic training, and one of us has a Ph.D. in Urban Education
Studies, with an explicit focus on racial equity, as well as a Ph.D. in mathematics
education. All of us are makers and users of second-order models of students, two
of us experienced. All of us consider it our professional responsibility to inquire into
and disrupt inequities, to shape our research and teaching to support equity and to
continue to learn. Although this paragraph only gives a few details about our posi-
tionalities, we shared what we think will help readers see why we are compelled by
the issues in this chapter.
In the rest of this chapter, we work on our second and third tasks. In the next sec-
tion, we describe how we see making and using second-order models to be acts of
equity based on how they have been presented in the research literature to date. In
the following section, we adapt van Es et al.’s (2022) notions of stretch and expanse
to demonstrate how research on second-order models could be enhanced to better
support acts of equity, with examples from recent research.
ow Are Making and Using Second-Order Models Acts

H
of Equity?
In this section, we show that making and using second-order models of students’
mathematics are acts of equity that demonstrate respect for students and that can
influence the social boundaries that define fields of action for students and teachers
in a mathematics classroom. However, we are aware that we should make clear the
nature of the interactions we are addressing. We write about ourselves as teacher-
researchers in research settings (e.g., design experiments, teaching experiments)
and classrooms (e.g., design experiments, university teaching). We address three
domains of activity: making second-order models, establishing epistemic students
or generalized models, and using second-order models and epistemic students in
our research and teaching.
To accomplish the task of this section requires that we address how we conceive
of mathematical knowing. We follow von Glasersfeld (1995) and others in this book
who view rational human knowing as an active process of perceiving and conceiv-
ing. Humans can never get outside of their ways of perceiving and conceiving in
order to check whether their ways of perceiving and conceiving are the way things
“really” are in the world. Humans can check with sources outside themselves, cer-
tainly! But the check a person makes is performed via their ways of perceiving and
conceiving. Therefore, the function of knowing is to organize one’s experiential
world rather than to discover an objective ontological reality. Yet, this position on
knowing is not solipsistic because humans cannot just organize the world however
they want—humans repeatedly encounter constraints through interaction that help
them shape their constructions. These constraints include social boundaries that
define fields of action for all actors, as well as actions upon these boundaries.
Within this framing of knowing, we differentiate between first-order and second-
order knowledge. First-order knowledge is the knowledge that a person constructs
“to order, comprehend, and control his or her experience” (Steffe et al., 1983, p.
xvi). In contrast, second-order knowledge is the knowledge that a person constructs
about another person’s knowledge “in order to explain their observations (i.e., their
experience) of the subject’s [another person’s] states and activities (p. xvi).” An
example of first-order knowledge is a calculus teacher’s knowledge of integrals.
Included in this knowledge may be multiple ways of thinking about an integral—for
example, as a way to find the area under a curve or as an accumulation function cre-
ated from the rate of change of the function to be integrated. The teacher may or
may not be able to link these two meanings for integral. The same teacher’s second-
order knowledge would be their knowledge about how their students might reason
about integrals, which the teacher has developed from interacting with their stu-
dents over time. Steffe and D’Ambrosio (1995) call this kind of second-order
knowledge a working (or living) model because it consists of descriptions formed in
the context of interactions.
In this section, we demonstrate that making second-order models is an act of
respect; establishing epistemic students is an augmented act of respect, and using
second-order models and epistemic students with particular students is what we call
“respect in action.” Furthermore, using second-order models and epistemic students
in research and teaching can support a teacher-researcher to disrupt the norm of
students being tasked with learning the mathematical knowledge of their teacher, a
norm we view as fundamentally inequitable (as we will explain). That is, using
second-order models and epistemic students in our research and teaching can
address how power affects people in research settings and classrooms. We begin
with an explanation of making second-order models.
Making Second-Order Models
What Is a Second-Order Model?
When first interacting with students in a research setting, teacher-researchers can

make working second-order models of students’ mathematical reasoning. Repeated
interactions over time are essential for these models, which are framed by the
teacher-researcher’s prior experience with students in the specific mathematical
domain, as well as by the teacher-researcher’s understanding of prior second-order
models in the domain.
At the beginning, a teacher-researcher often interacts intuitively with students in
order to harmonize themselves with the students’ current reasoning (Hackenberg,
2010; Steffe & Thompson, 2000). Interactions include posing questions to the stu-
dents, listening to the responses, asking more questions, observing what the stu-
dents write or draw, observing the students’ body language, etc. Based on the
teacher-researcher’s interpretations of these interactions, the teacher-researcher
might conjecture that the students understand in a particular way or show an under-
standing that seems quite different from the teacher-researcher’s understanding.
To test that conjecture, the teacher-researcher designs problems that allow stu-
dents’ reasoning to surface and that can open possibilities for students to learn.
Indeed, one aim of a teacher-researcher is to maximize opportunities for learning—
but the teacher-researcher does not assume at the outset that they know how to
orchestrate that learning. Toward this end, the teacher-researcher engages in ongo-
ing analysis (Steffe & Thompson, 2000), which involves watching video recordings
of interactions, taking notes, and discussing interpretations with a research team in
order to continue to revise conjectures, make new conjectures, and test them.
As the teacher-researcher develops language to describe the student’s reasoning
and learning, they are beginning to make a second-order model—if they set their
own ways of thinking to the side in order to conceive of the student’s ways of think-
ing as distinct from their own (Steffe & Thompson, 2000). This is a big “if.” This
activity is the heart of decentering (Bas-Ader & Carlson, 2021; Carlson et al., Chap.
9, this volume; Teuscher et al., 2016), which means “taking actions that adopt the
perspective of another” (Bas-Ader & Carlson, 2021, p. 2). It often requires new
language to describe a particular student’s understanding in order to not conflate the
teacher-researcher’s ideas with the student’s ideas. It requires making space for stu-
dents’ ideas (cf. Gutiérrez, 2018; Kokka, 2022) as well as articulating them pre-
cisely. Please see Carlson, et al. (Chap. 9, this volume) for more.
Why would a teacher-researcher need to make such a model? A main reason is
that if a teacher-researcher wants to understand a student’s mathematical thinking,
they can only get to know it through their own ways of perceiving and conceiving.
Thus, any second-order model of a student’s knowledge is not the same as the stu-
dent’s first-order knowledge (Steffe & Thompson, 2000).
When interactions have concluded, the teacher-researcher goes through a formal
analytic process that includes iteratively watching the video recordings to make a
coherent account of the student’s reasoning. This process transforms a working
second-order model into an analytic second-order model that includes explanatory
accounts. When we use the term second-order model, we mean this analytic model.
A second-order model, then, is a constellation of constructs that a teacher-researcher
makes to describe and account for another person’s mathematical reasoning, includ-
ing changes in the person’s reasoning (i.e., learning) (Steffe et al., 1983; Ulrich
et al., 2014). So, a second-order model is a scientific model of someone else’s first-
order knowledge.
Why is what we have described above a model? Remember, the teacher-
researcher has no direct access to what the student perceives and thinks. The con-
structs are a model in the sense that they provide “a hypothetical system of thinking
that would result in the same externalized results of the student’s thinking that are
observed” (Ulrich et al., 2014, pp. 329–30). In other words, if someone were to fol-
low the thinking described in the model, that person would act as the student does.
What language is used in making a second-order model? A more fundamental
question is what a second-order model consists of. Second-order models of stu-
dents’ mathematics are founded on the idea that mathematics consists of mental
actions, or operations, and complex organizations of those operations into schemes
and concepts (Steffe, 2010). For example, we invite the reader to imagine a cube.
What can you do with the cube you are imagining? You might be able to turn it.
Turning is an operation. In fact, the creation of a mental image of a cube is a com-
plex organization of mental actions, so you are right now applying an operation
(turning) to other operations that you have previously constructed. A second-order
model consists of language to describe a person’s operations, schemes, and con-
cepts and make accounts of changes in them. Please see Tillema and Gatza
(Chap. 3, this volume) for more.
Why Is Making a Second-Order Model an Act of Equity?
Making a second-order model is an act of equity because it is an act of respect for

the thinking of the students whose mathematical thinking is being modeled. The
message of making a second-order model is “I see your way of thinking, and I am
going to hold it out as important and unique.” Making second-order models is a
time-intensive endeavor, on the order of months and years, not days or weeks. So,
making a second-order model is not something to engage in lightly. In our view, the
time-intensive nature of model-making deepens it as an act of respect for the stu-
dents whose mathematical thinking is being modeled. Furthermore, making a
second-order model is never fully complete due to the complexity of each individual
student’s thinking. However, teacher-researchers can get to the point where they
have a model that holds up across all of their interactions with the student (Steffe &
Thompson, 2000).
In addition, the orientation to making second-order models is an act of equity
because it can disrupt disciplinary norms. A teacher-researcher who makes a second-
order model is also saying “I will put my own first-order mathematical knowledge
to the side and not assume that that knowledge is what you (the student) are to
learn.” A teacher-researcher cannot separate themselves from their own first-order
knowledge, but the intentional and repeated practice of putting their knowledge to
the side has two consequences with respect to mathematical knowledge: First, it
means that the student’s mathematical knowledge is actually being privileged over
anything seen as “standard” (first-order) mathematical knowledge. Second, it means
that new mathematical knowledge can be co-constructed between the students and
the teacher-researcher. A good example of this kind of extensive co-construction
that can result in rethinking an entire mathematical domain can be found in Steffe
and Olive’s monograph on fractions, Children’s Fractional Knowledge (2010).
And yet, if the making of second-order models was only about highlighting the
individual thinking of certain students, it would be an honorable pursuit with lim-
ited impact. A central reason that making second-order models is an act of equity is
because making them can contribute to the construction of epistemic subjects (Beth
& Piaget, 1966), although we will use the term epistemic students.
Establishing Epistemic Students

What Is an Epistemic Student?
An epistemic student is “that which is common to all subjects [students] at the same
level of development, whose cognitive structures derive from the most general
mechanisms of the co-ordination of actions” (Beth & Piaget, 1966, p. 308). Our
interpretation of an epistemic student is as follows. In a very broad sense, people
have similarities biologically, and all children go through a process of organizing
their experiences as they grow. We are speaking here at the level of the species, not
at the level of the numerous social, cultural, and individual differences we can
observe with a view that is more detailed. Due to these very broad similarities, we
can expect some similarity in the operations and schemes that students construct.
So, although there is always a myriad of differences in thinking that are unique to
individual students, there are also similarities, and these similarities can be identi-
fied and organized.
For example, in a first-grade classroom with 25 students, a teacher has 25 indi-

vidual thinkers. Yet, in terms of how the students think about number, second-order
model building has shown that there are similarities among students that can help
teacher-researchers organize interactions with them (Steffe & Cobb, 1988; Steffe
et al., 1983). Teacher-researchers have found that children construct four different
number sequences as they construct whole number knowledge. These number
sequences are epistemic students, and they are one support in managing the com-
plexities of interacting with a multitude of students.
For example, when children first become numerical (meaning, construct and
count abstract unit items), they construct the initial number sequence (INS) (Steffe
& Cobb, 1988; Steffe et al., 1983). Prior to this point, children count to solve prob-
lems, but numbers do not signify an amount to them. For example, if they have just
counted 13 beads and are asked how many beads there are, they will engage in
counting again rather than say “thirteen.” When children construct the INS, they
spontaneously count on to solve problems. For example, if a child who has con-
structed the INS encounters eight beads on the desk and five more in a cup and is
asked how many beads in all, the child will start from the eight and count on “nine,
ten, eleven, twelve, thirteen,” usually keeping track of the additional counting on
fingers. The child counts on from eight because, for them, “eight” means the result
they would get if they counted from 1 to 8. So, the child does not have to count
that—they can start at 8 and continue. Thus, the INS signifies the beginning of an
“amountness” meaning for whole numbers. In addition, the INS is an example of an
epistemic student because it indicates key cognitive characteristics—such as the
ability to count on to solve problems—that are common to students who have con-
structed it. In other words, although INS students are certainly not all identical in
how they think, they have this similarity that marks a significant change in how they
understand number.
The INS as a construct grew out of specific second-order models of individual
students constructed by Steffe and colleagues (Steffe & Cobb, 1988; Steffe et al.,
1983) that they then generalized in relation to second-order models of students who
had not constructed amountness meanings for numbers and of students who demon-
strated more sophisticated ways of operating with numbers. Initially, two class-
rooms of first-grade children were involved in exploratory teaching to allow the
researchers to get to know how these children thought about number (Steffe &
Thompson, 2000). These classrooms were located in a school near the University of
Georgia in Athens, where the student population in schools are about 50% African
American and 50% non-African American.2 Then, the researchers conducted a
2-year teaching experiment with six first-grade children to intensively test a large
hypothesis about the learning of the children (Steffe & Thompson, 2000). That
study was pivotal in the construction of children’s number sequences (Steffe &
Cobb, 1988; Steffe et al., 1983). These models have subsequently been used in a
2
Today, the schools in Athens, Georgia, are 48.5% African American, 25.2% Latinx, 20.2% White,
4.3% two or more races, and 1.4% Asian or Asian Pacific Islander (https://www.usnews.com/edu-
cation/k12/georgia/districts/clarke-county-111273).
number of projects and studies by a range of researchers to provide elaboration and

confirmation of them (e.g., Clements & Sarama, 2009, MacDonald & Wilkins,
2019; Ulrich, 2016; Wright et al., 2006, 2012).
However, we hasten to point out that a teacher-researcher does not automatically
establish epistemic students upon the making of specific second-order models. In
our view, it is essential to make detailed second-order models before trying to
abstract or generalize from them. The importance of this recommendation cannot
really be understated: Building specific models can help teacher-researchers develop
facility with analytic constructs in relation to their student-participants. It can be
helpful to build second-order models in the context of other epistemic students from
prior research. For example, when the first author first began to build second-order
models of middle school students’ fractions knowledge and algebraic reasoning, she
did so in the context of the number sequences research. In addition, it is especially
useful to engage with students whose mathematical activity and ways of thinking
are challenging to understand or explain. Faithful work in this domain can support
a teacher-researcher to become a good second-order model builder. It can also sup-
port a teacher-researcher to imagine what they would do in future similar interac-
tions with students—for example, if they were to conduct the teaching experiment
again with similar students or with new populations.
When a teacher-researcher does conduct these other studies on similar or related
topics, they might have the experience that they have met a new student before. In
other words, a teacher-researcher can have the feeling that a new student’s ways of
thinking have similarities to prior students because the new student’s ways of think-
ing demonstrate key cognitive characteristics that were part of the second-order
models of the prior students. There will be differences between the students, cer-
tainly: Each student’s mind is unique. Yet, recognizing key cognitive characteristics
in new students helps a teacher-researcher start to establish epistemic students.
Why Is Establishing an Epistemic Student an Act of Equity?
Establishing epistemic students is a welcoming, respectful position that augments

the respect of making a second-order model. It says to individual students, “Your
ways of thinking are worth knowing; they have features that are unique to you, but
also they have features in common with other students.” So, establishing epistemic
students is an act of equity because it is a deepened act of respect for the thinking of
the students whose mathematical thinking is being modeled.
Before moving on to the implications of this position, we want to emphasize two
issues. First, every student’s mathematics can be modeled, even those who have
been historically marginalized in mathematics education research (e.g., Hackenberg,
2013; Hackenberg et al., 2021; Tillema, 2018, 2020; Tillema & Hackenberg, 2017).
Here, by historically marginalized, we are pointing to students who are marginal-
ized because of their mathematical thinking, not due to other identifiers that are
more typically associated with marginalization, such as being a student of a particu-
lar race or gender identity or from a lower socioeconomic background. By
marginalization of mathematical thinking, we mean that a student’s mathematical

thinking is frequently not present in curricular materials available to teachers and,
therefore, it is unlikely that teachers will be able to make the thinking of these stu-
dents central to instruction. Moreover, students remain marginalized when mathe-
matics education researchers do not take on the goal of modeling the thinking of
these students and including them in research reports.
One example of historically marginalized students in mathematics education
research are middle school students who are not yet multiplicative thinkers. We
view students in middle school who are not yet multiplicative thinkers as histori-
cally marginalized because (1) middle school curricular materials focus primarily
on topics that require multiplicative reasoning, and (2) little attention has been paid
by researchers to this issue, to our knowledge. Zwanch and Wilkins (2021) charac-
terize multiplicative thinkers as students who have constructed composite units
where the units of 1 are iterable. Students who have constructed composite units
think of a pack of six yogurt cups as one 6 and six 1 s. Students who have con-
structed iterable units of 1 anticipate the operation of iteration on 1 to create larger
numbers. So, for example, a priori, these students think of 6 as 1 iterated 6 times or
102 as 1 iterated 102 times. Prior to this construction of composite units with iter-
able units of 1, students feel a need to have 102 ones to conceive of a number like
102. Students who have not yet constructed iterable units of 1 are quite challenged
by the multiplicative nature of typical middle school learning goals and curricular
materials (Tillema & Hackenberg, 2017). And yet, their mathematics can be identi-
fied and modeled (e.g., Hackenberg, 2013; Hackenberg et al., 2021; Tillema, 2018,
2020) rather than left entirely out of research. Please see Hackenberg and Sevinc
(Chap. 11, this volume) for more information on these and related epistemic
students.
Second, readers might get the idea that epistemic students are exclusionary. That
is, one might ask: What about students who do not fit into the epistemic students
that have been created so far in research? Our response to this question is twofold.
First, radical constructivism is a self-reflexive stance (Steffe & Wiegel, 1996): What
we apply to students’ ways of thinking and knowing, we (endeavor to) apply to
ourselves. “Endeavor to” in parentheses means self-reflexivity is a practice, always
in the process of becoming. This self-reflexivity means, at minimum, that teacher-
researchers need to engage in productive doubt of their models. Productive doubt
means that we hold the models out for scrutiny and do not view them as the best or
only way to construct a model or to generalize a model. When published, models are
our most current interpretations that have been subjected to rounds of scrutiny, but
they could certainly be revised in the future with further analysis. Second, if a
teacher-researcher encounters students who are unusual with respect to prior mod-
els and prior epistemic students, then the teacher-researcher has two important
actions to take: (1) create a second-order model for interaction with that student
(e.g., Steffe & Ulrich, 2010; Ulrich, 2016) and (2) engage in a search over time,
creating second-order models with more students to determine warrants for the cre-
ation of a new epistemic student. Researchers in recent years have abstracted new
epistemic students (Hackenberg & Sevinc, 2021; Ulrich, 2016).
Using Second-Order Models and Epistemic Students
What Does It Mean to Use Second-Order Models and Epistemic Students?
The implications of making epistemic students are “If we knew more about your
ways of thinking and how they fit into the ways of thinking of other students, that
would help us organize learning experiences for you—and for other students who
have cognitive similarities to you.” In other words, for teacher-researchers, the mak-
ing of epistemic students has an immediate implication of use: They are a conduit
for using second-order models in research and teaching. That is why teacher-
researchers write about second-order models as tools for interaction (e.g., Steffe &
Thompson, 2000; Ulrich et al., 2014).
For example, the students described above who have constructed composite units
but not iterable units of 1 are often referred to as “students operating at stage 1 of
units coordination” or “students at stage 1.” These students can solve equal groups
of multiplication problems by tracking composite units with another composite
unit. For example, they can determine the number of yogurt cups in eight 6-packs
by using the units of eight to keep track of amassing 6 s (Steffe, 1992). They make
this coordination in their problem-solving activity, and so they do not anticipate it
when solving problems. These students can learn to solve division problems, but the
results of division do not feed back into the coordination described above. For
example, suppose a stage 1 student was asked to determine how many packages of
yogurt cups could be made from 42 yogurt cups. Students at stage 1 can learn to
solve this problem by tracking the number of 6s they count until they reach 42. Let
us further suppose that directly after this solution, a stage 1 student were asked how
many cups were in 7 packs. They would solve this second problem anew—they
would not consider that in solving the first problem, they had already solved this
second problem (Steffe, 1992). This lack of reversibility between multiplication and
division is one key cognitive characteristic of the epistemic student, a “student at
stage 1.”
Considering this knowledge about students at stage 1 can influence the learning
goals a teacher-researcher makes for these students when working on multiplication
and division with them. For example, if a teacher-researcher knows that reversible
thinking is problematic, they would pose problems to explore and support students
to engage in it but not expect that they will solve problems similarly to students who
have constructed multiplication and division as reversible (e.g., Steffe, 1992).
Taking this approach allows the thinking of these students to be more visible in
interactions and means they would be less likely to be subjected to expectations that
are wildly out of step with their thinking (e.g., Steffe, 1994). For older students, this
knowledge about students at stage 1 can influence learning goals for topics that
require multiplication and division, such as proportional reasoning (e.g., Hackenberg
et al., 2023) and exponentiation. It can also influence other aspects of teaching, such
as task design, questioning, and interpretations of students’ thinking.
And yet, there are possible negative consequences to organizing the understand-
ing of students in this way. One possible negative consequence is that epistemic
students might become reified, and deficit narratives could grow around them (cf.
Adiredja & Louie, 2020). For example, we could imagine deficit narratives growing
around middle school students who are operating at stage 1: These students could
simply be labeled as students who do not understand typical topics in middle school
mathematics curricula and slotted into remedial classes. We cannot state strongly
enough that, in our view, this would be a misuse of epistemic students. Epistemic
students can be used to assess and categorize students. However, if the use of them
ends there, and they are not used as tools for designing interactions with students,
then they have been misused.3
Furthermore, because students are always changing, as is our knowledge of epis-
temic students, epistemic students should be orienting for a teacher-researcher, not
deterministic. In other words, a teacher-researcher has to hold lightly to the guid-
ance that the epistemic student provides. Holding lightly means repeatedly testing
to see if a student has made progress unexpected for students at stage 1, for exam-
ple. This evidence could mean that the student is progressing to a new stage of units
coordination, or it could mean that the knowledge of students at stage 1 has to be
further refined (e.g., Ulrich, 2016). The use of epistemic students in interaction does
not imply a list of forbidden topics for certain students. For example, students at
stage 1 are likely to work on some topics or problems in different ways from stu-
dents at other stages, but they are not barred from working on topics or problems.
As Steffe (1992) has explained, he lets the students be the guide regarding what
topics to work on, as he adapts to their ways of thinking. In short, epistemic students
should be used in a way that allows the teacher-researcher to reveal and understand
more about the student, to craft tasks and interactions that are sensible to the student
and that spur operative activity (Tillema & Gatza, Chap. 3, this volume), and to take
seriously the constraints that the student may demonstrate while always holding
open possibilities for growth.
Taking constraints that students demonstrate seriously is one aspect of using
epistemic students that does not get a lot of attention. We want to say carefully what
we mean by it. For us, a constraint is a way of thinking that we as teacher-researchers
view as a conflation, but that seems to persist for a student despite our best efforts to
pose problems or ask questions that would allow the student to eliminate it. The
example above, where the stage 1 student correctly solved how many 6-packs of
yogurt cups could be made with 42 cups and then did not know how many yogurt
cups were in 7 packages, is a good example of a constraint. Even persistent probing
and questioning, or trying other problems or smaller numbers, would be unlikely to
have an immediate influence on this student’s way of thinking because changing it
requires a significant reorganization in how the student conceives of units. When we
as teacher-researchers encounter a constraint like this, we do our best to explore its
extent—is it really a constraint? Can a set of problems or questions support the
3
Note that this does not mean all studies have to be design studies; some research does need to
assess students so that we know about particular populations, e.g., Zwanch & Wilkins, (2021). But
research of this kind has to conduct this assessment work extremely carefully, with an eye toward
using the results to support students.
student to eliminate it? And, we show respect for it—and for the student—by not
wishing it away or hiding it. We support it to surface. We include and account for it
in our second-order model.
We note that as students get older, these constraints may change. For example, a
middle school stage 1 student may be repeatedly told that division and multiplica-
tion are inverses, and so they may be able to respond that 7 packages are 42 cups.
However, being able to respond to this particular question does not necessarily indi-
cate an underlying change in the unit structure the student establishes when solving
problems involving multiplication and division. In our experience, aspects of the
student’s unit structure will emerge in other problem contexts, such as proportional
reasoning, as constraints (e.g., Hackenberg et al., 2023).
How Is Using Epistemic Students an Act of Equity?
Now, we turn to how using second-order models and epistemic students in interac-
tion with particular students is an act of equity. In our view, doing so is a higher level
of respect for particular students than simply making second-order models. It is a
higher level of respect because it involves interacting supportively with particular
students in the moment and with students who have similar ways of thinking in the
future. That is, a teacher-researcher can say that they respect the thinking of students
at stage 1 in middle school—students who have not constructed composite units
with iterable units of 1. But until the teacher-researcher uses their particular and
generalized models (their epistemic students) to support interaction with these stu-
dents, their respect is a bit “empty.” It is respect mostly in name, not in deed. When
a teacher-researcher uses their epistemic students to inform their learning goals,
task design, and plans for interaction (e.g., questioning), a teacher-researcher is try-
ing to act in a way that allows these students to engage in operative activity. A
teacher-researcher is also simultaneously building their knowledge of how to inter-
act with students who think similarly in the future. The teacher-researcher is learn-
ing how to interact. We view using epistemic students in interactions with particular
students as “respect for students in action.”
In addition, using epistemic students in interaction with particular students can
address how students and the teacher-researcher are affected by and act upon power
in the research setting or classroom. Recall that we define power as the “social
boundaries that, together, define fields of action for all actors” (Hayward, 1998,
p. 12), where social boundaries include “laws, rules, norms, institutional arrange-
ments, social identities and exclusions” (Hayward, 1998, p. 2). Using epistemic
students in interactions in research settings and classrooms can allow teacher-
researchers and students to act upon social boundaries to change their shape or
direction in order to rectify past inequities.
For example, a norm that is typical in middle school and high school mathemat-
ics classrooms is that students are to learn the first-order mathematical knowledge
of the teacher (Teuscher et al., 2016). In other words, in typical mathematics class-
rooms, teachers present mathematical ideas neutrally, as “the mathematics” to be
learned. From our perspective, the mathematics to be learned in any given class-
room is determined by the teacher’s first-order knowledge. Most of the time, teach-
ers attempt to convey their first-order knowledge to students. And students often
tacitly agree to this norm and try to learn these meanings (Liljedahl, 2021;
Thompson, 2016). Yet, students cannot learn these meanings directly; they can only
interpret the teacher’s activity with the meanings (operations, schemes) that they
have constructed at that point—with their own first-order knowledge. Nevertheless,
many of them try. And thus, the norm of students being tasked with learning the
teacher’s knowledge is generated and reinforced. This norm becomes one of the
central boundaries that shapes the fields of action for teachers and students.
We view this norm as a fundamental inequity because of the heavy burden it
places on students and because of the way in which it keeps student thinking hidden.
First, trying to learn the first-order knowledge of the teacher means that, at mini-
mum, the students are tasked with trying to learn what they often do not have the
ways of thinking (operations, schemes) to learn. That is, without certain operations
and schemes, making interpretations that are compatible with the teacher’s mean-
ings is impossible. Therefore, this task can be a heavy burden for many students.
Furthermore, this burden does not fall equally. For example, students at stage 1 are
at a greater disadvantage to make interpretations of a typical middle school teach-
er’s first-order knowledge than students at stage 3, who have constructed reversibil-
ity between multiplication and division. So, students who need the most support are
having to do the most work. And yet, all students are burdened.
Second, student thinking stays hidden. It may surface at times, but when it does,
it usually is as a curiosity or anomaly, something unusual. There is no sense that
students’ thinking could be a body of knowledge that is different from the teacher’s
first-order knowledge and yet is still knowledge, with coherence and depth, and
with ways that it could be modified and advanced under supportive pedagogical
environments. When student thinking stays hidden, students are essentially kept
submerged. Many students quickly learn that math class is not about them or how
they think, and many of them, therefore, view the subject as boring, mystifying, or
even detrimental to their well-being. The pain experienced by some students in
math class (e.g., Gholson & Martin, 2019) may be related to the norm of deeply
hiding student thinking in favor of the teacher’s knowledge. That is, Gholson and
Martin document the “suffocating, headache-inducing, and painful” (p. 402) experi-
ence of a middle school Black girl in mathematics class and focus on how she expe-
rienced microaggressions related to her race and gender. We conjecture that an
additional factor in this kind of pain in math class is the deep hiding of students’
mathematical thinking.
In contrast, a teacher-researcher who is using second-order models and epistemic
students in interaction with particular students can subvert the larger norm of stu-
dents being tasked with learning the teacher’s first-order knowledge. Whether the
teacher-researcher is working outside the classroom (e.g., in a teaching experi-
ment or outside-of-school design experiment) or inside classrooms (e.g., in a whole
classroom experiment, in university teaching), the typical norm may surface. For
example, students may wait for the teacher-researcher to tell them what to do, or
they may not share their ideas. A teacher-researcher who is using epistemic students
in interaction with particular students must learn the mathematical ways of thinking
of the particular students. So, this teacher-researcher will have goals of bringing
student thinking to the forefront of discussion and using that thinking to support the
progress of all students involved. This orientation and these goals, if enacted, can
disrupt the typical norm of students being tasked with learning the first-order knowl-
edge of the teacher. This orientation and these goals can instead work to establish
the norm that the teacher-researcher must build (working) second-order knowledge
of the students, and that the students’ mathematics should be the subject matter of
the research setting—even in whole classrooms. Thus, using epistemic students in
interaction with particular students can influence how power affects students and
the teacher-researcher in research settings or classrooms by shifting whose mathe-
matical knowledge is valued and centered.
Documenting this phenomenon has not typically been done by researchers who
make and use second-order models, to our knowledge (cf., McClain & Cobb, 2001).
This chapter is a call to make this type of documentation, which would include care-
ful articulation about how this disruption of an oppressive norm facilitated student
learning and opportunities for learning.
How Can We Enhance Current Second-Order Models?
We now turn to our third task in the chapter, to identify how second-order models
could be enhanced to better support acts of equity. To do so, we focus on the inclu-
sion of social identity categories and social identities in the creation of second-order
models. We take this focus because researchers have identified the impact of iden-
tity both on opportunities for mathematical learning and on mathematical learning
itself (e.g., Aguirre et al., 2013; Martin, 2009; Martin et al., 2017; Shah, 2017). We
begin by defining social identity categories and social identity.
Social Identity Categories and Social Identities
We use the term social identity categories to refer to the range of identifiers com-
monly used to categorize people, including in the reporting of research data, such as
race, gender, socioeconomic status, etc. (Langer-Osuna & Esmonde, 2017, p. 637,
membership identity; Martin et al., 2017, p. 612, theme 1). We define social identi-
ties in a way that is consistent with our definition of identity at the beginning of the
chapter. Here, we use the word social to signal that we are talking about non-
mathematical identities, such as an individual’s ways of identifying themselves with
(or outside of) a racial group, culture, class, gender, linguistic group, etc. (Langer-
Osuna & Esmonde, 2017, p. 637 individual identity; Martin et al., p. 613, theme 3).
We consider this identification to be a dynamic process that individuals engage in

via interactions with others.
We make the distinction between social identity categories and social identities so
that we can differentiate between the reporting of social identity categories in
research and the use of social identities to support analyzing data. As Martin et al.
(2017, theme 1) have pointed out, when researchers report on social identity catego-
ries, they often do so without theorizing about them—the categories are treated
simply as static identifiers for participants or researchers. We consider reporting
social identity categories for both researchers and participants to be an important
way to contextualize research studies. We hasten to add that it is also important to
theorize about the relationship between social identity categories and social identi-
ties so that researchers can use these constructs analytically in studies (Martin,
et al., 2017, theme 3), rather than only to contextualize a study. We discuss reasons
to include social identity categories in the reporting of research, and then, we theo-
rize further about the relationship between social identity categories and social
identities.
hose Reasoning Is Represented in Our Models? Participants’

W
Social Identity Categories
As part of making accounts of mathematical learning, we consider it important to

report on participants’ social identity categories (e.g., gender, racial, socioeco-
nomic, linguistic). To accomplish this goal, we recommend allowing participants to
identify the social identity categories to which they see themselves as belonging.
Doing so ensures that a researcher captures how the participants identify themselves.
We make this point because this information is not always reported in studies
aimed at creating second-order models (Zahner, 2019).4 When this information is
not reported, it can lead readers to assume that the participants in studies are White,
middle class, and well educated. This assumption is reasonable given that, in gen-
eral, social science research has tended to lack inclusion of participants from diverse
social identity categories (Henrich et al., 2010). This lack of representation has led
to researchers making claims about human cognition that are largely based on
research that was conducted with White middle-class participants (Henrich et al.,
2010). When this phenomenon occurs, then researchers are essentially taking White
middle-class children’s reasoning as representative of all children’s reasoning
(Martin et al., 2017). We do not want to perpetuate this problem either by working
with only White middle-class participants or by giving the perception that we may
be by not clearly reporting this information. Both can support a range of biases in
research (Zahner, 2019), including, for example, structural manifestations of
4
Zahner (2019) did a relatively small meta-analysis of studies, 21 peer-reviewed published studies,
but found an inconsistent reporting of this data.
Whiteness—taking White people’s culture, customs, and beliefs as the standard to

which all people are compared (National Museum of African-American History and
Culture, 2023).
By reporting participants’ social identity categories, researchers can look across
studies to determine the extent to which we have included a diversity of social iden-
tity categories in them. Looking across studies is particularly important to achieve a
diverse representation of social identity categories because most studies involve
only a small number of participants. Therefore, to make claims about who is repre-
sented in studies aiming to make second-order models necessarily requires looking
across multiple studies. We also think that this reporting increases the trustworthi-
ness of our models, in that reporting social identity categories allows other research-
ers to consider the relevance and possible use of a second-order model in their own
contexts (Mertens, 2010).
That said, we acknowledge that identifying this information requires care because
the identification of racial, gender, socioeconomic, linguistic, or other social iden-
tity categories can open the door to making comparisons that erroneously support
broader deficit-based social narratives about the relationship between particular
social identity categories and mathematical ability (Gutiérrez, 2008; Martin, 2009).
We suggest that a researcher attend to this potential and frame the reporting of
results with this issue in mind (see, e.g., Valencia’s (2010) discussion of “at-risk
schools” rather than “at-risk students”).
ho Are the Model Builders? Researchers’ Social

W
Identity Categories
We also consider it important to include the social identity categories of the research-
ers who are making the second-order models. Making explicit a researchers’ social
identity categories can contextualize the questions they ask and the interpretations
they give, and it can orient a reader to possible affordances and limitations that may
be a part of the study (Bartell & Johnson, 2013; D’Ambrosio et al., 2013c). It can
also help mathematics educators to identify what research is being taken up and
by whom.
We see contextualizing the second-order models that we produce with our own and
participants’ social identity categories as an important starting place for taking into
consideration how our and our participants’ social identities may impact research
interactions. That is, the methods of research we use to develop second-order mod-
els rely heavily on researchers using their own teaching as a basis for scientific
inquiry (Steffe & Thompson, 2000). However, little work that emanates from this
research tradition has been done to examine how our social identities and those of
our participants impact these interactions. To address this issue, we first theorize
about the relationship between social identity categories and social identities.
heorizing About Social Identity Categories

T
and Social Identities
Researchers have argued that social identity categories are deeply ingrained in soci-
ety in that they are encoded in law (e.g., Feagin, 2010; Lipsitz, 1995) and used to
shape institutions like schools (Battey & Leyva, 2016; Ladson-Billings, 2006;
Tyack, 1974). Therefore, they have material consequences for all individuals, and
these material consequences are differential, depending on how an individual is
categorized by others and how the individual categorizes themselves within the
broader system (Battey & Franke, 2015; Leonardo, 2009). Moreover, individual
members of a society have highly differential access to creating or making changes
to these categories or to the consequences of the system of categorization (Gimenez,
2014). And the categories can be used by those who have more access as one way
to define the social boundaries of action for those who have less access (Freire,
1970). Thus, the extent to which individuals experience a sense of agency to change
the shape or direction of these social boundaries depends on the access they have to
mechanisms that can effect change, often requires collective action over time, and is
related to the material consequences an individual may experience for working to
make such changes.
With these observations in place and commensurate with radical constructivist
epistemology, we position social identity categories as constraints, which function
similarly to physical or conceptual constraints (von Glasersfeld, 1984, 1995), in that
they are the reality that “kicks back” as individuals’ construct their social identities
in interactions with each other. We note that social identity categories are them-
selves constructions, but because they have a legal basis and history, the categories
have a different level of durability and social sanctioning than, for example, indi-
viduals’ construction of their social identities in a mathematical interaction.
Moreover, the durability of these categories, and the way that they constrain indi-
viduals’ interactions, depends on the context in which they are invoked and on how
they have been encoded and used within that context. We further note that the con-
straints that individuals experience in constructing their social identities are differ-
ential across social identity categories. That is, laws that result in differential benefits
across different social identity categories (e.g., access to well-funded schools)5 form
the basis for individuals to experience differential constraints in their construction
of their social identities within those contexts.6
5
We note that laws leave historical legacies, so they need not be currently in place to shape current
social institutions.
6
Other researchers have framed differential constraints as involving privilege (e.g., White privilege
or cis-gender privilege). We consider privilege to be an observer making a comparison of con-
straints between people who are members of two different social identity categories and positing
that individuals belonging to one social identity category do not experience a constraint, whereas
those belonging to another social identity category do experience a constraint. Members of the
social identity category who do not experience a constraint are considered to have a privilege.
With social identity categories positioned as constraints that individuals experi-

ence in their construction of their social identity, we return to an epistemological
point that von Glasersfeld (1983, 1995) frequently made: An individual’s first-order
knowledge is constructed in relation to, but not determined by, their experience of
constraints. That is, we propose that social identity categories form one set of con-
straints in individuals’ constructions of their social identities but that there are mul-
tiple viable (von Glasersfeld, 1984, 1995) responses to these constraints. We see this
theoretical point as an important way to understand differences in the way members
of the same social identity categories construct their social identities—there are
multiple viable responses to any set of constraints. We note, too, that the intersec-
tion of multiple of these identity categories is often at play as individuals construct
and enact their social identities within a particular context (Collins & Bilge, 2020).
These considerations allow us to differentiate between first- and second-order
knowledge within the arena of teacher-researchers’ and students’ social identities.
We remind the reader that first-order knowledge is the knowledge a person con-
structs “to order, comprehend, and control his or her experience” (Steffe et al., 1983,
p. xvi). In contrast, second-order knowledge is the knowledge that a person con-
structs about another person’s knowledge “in order to explain their observations
(i.e., their experience) of the subject’s [another person’s] states and activities (Steffe
et al., p. xvi).” We have described how social identity categories can form con-
straints in a person’s construction of their first-order knowledge of their social iden-
tities. Second-order knowledge of social identities, then, entails a teacher-researcher
making working second-order models of the ways that their own and their partici-
pants’ social identities impact interactions and, in particular, impact interactions
aimed at mathematical learning. We address important considerations in making
second-order models of social identities next.
onsiderations for Making Second-Order Models That Account

C
for Social Identities
To discuss how second-order models could be enhanced by accounting for social

identities, we follow van Es et al. (2022), who have defined two terms, expanse and
stretch, that we apply to including social identities in the creation of second-order
models. van Es et al. define expanse as “the breadth and range of what teachers
[-researchers] identify as noteworthy” (p. 115) in a research interaction. They define
stretch as the ways in which a teacher-researcher “reaches back historically and
forward futuristically” (p. 117) to consider their own and their participants’ pasts
and futures.
With these definitions in mind, we note that teacher-researchers have typically
situated their current second-order models in relation to prior models of students’
mathematical reasoning. Obviously, mathematical reasoning is not the only factor
that contributes to our interactions with others and, in particular, that contributes to
others’ opportunities for mathematical learning in interaction with us. Thus, one
way we can expand current second-order models is by attending to the ways that
researchers’ and participants’ social identities contribute to our interactions aimed
at learning. We see doing so as one way to broaden, in the sense of van Es et al.
(2022), the analytic lens of a researcher making second-order models. To expand
one’s analytic lens can involve van Es et al.’s notion of stretch by making explicit
our understanding of our own and our participants’ pasts and futures relative to their
social identity categories. In this section, we make some recommendations about
what is necessary to expand and stretch current second-order models in this way.
We begin with the observation that there is well-documented research evidence
of the inequitable distribution of resources in mathematics education, for example,
along racial and socioeconomic lines (e.g., experienced teachers, quality curricu-
lum, technology, etc.) (e.g., Oakes et al., 2004). Therefore, we recommend that
researchers interested in making second-order models that attend to social identities
aim to document successful avenues for working with students that belong to his-
torically marginalized social identity categories (Langer-Osuna & Esmonde, 2017).
That is, rather than have a study’s findings potentially reflect the broader inequitable
distribution of resources in mathematics education, we think it is more productive to
explicitly design studies aimed to disrupt these inequities (e.g., Gatza, 2021; Kokka,
2022; Stephan et al., 2021). With that goal in mind, it may be important for a
teacher-researcher to read research bases that extend outside of mathematics educa-
tion (Aguirre et al., 2017; Gutiérrez, 2017) that can inform the specific equity issue
they hope to address.
Beyond knowing research as a basis for intentionally designing studies, it is
important for a teacher-researcher who has different social identities than the stu-
dents with whom they work to consider specific ways to cultivate respect and free-
dom (i.e., the capacity to engage in acts of equity) in the interactions with their
study participants (D’Ambrosio et al., 2013b). That is, when a teacher-researcher
and participants share fewer social identities, a sense of respect and power may be
more fragile in these interactions than in ones in which the teacher-researcher and
participants have more shared social identities (Solomon, et al., 2011). For this rea-
son, we recommend that teacher-researchers plan for prolonged and persistent
engagement aimed at building relationships with students in which the teacher-
researcher and students can build trust (Mertens, 2010). Doing so can include inter-
actions that are outside of the immediate research setting, including those that reveal
a researcher’s commitment to working productively with students.
Another consideration for teacher-researchers is to be explicit with participants
about how they think about their social identities, in relation to themselves and in
relation to a broader social, cultural, and historical context (Clark et al., 2013; de
Freitas et al., 2012), and offer their research participants opportunities to do the
same (cf. Langer-Osuna & Esmonde, 2017). Doing so allows a teacher-researcher
not to assume that the way they have organized their social identities within a
broader context is necessarily like the way that their research participants have. It
also provides space for nuance so that the teacher-researchers’ and participants’
social identities are not reified or discussed univocally across a particular social
identity category (Martin et al., 2017).
Like our discussion of mathematical knowing, we consider integrating these fea-
tures into the design of a study as conveying that a teacher-researcher is aiming “to
see” participants in a way that promotes respect for them. Moreover, it can help to
cultivate freedom among participants, in that it can open the way for them to act on
social boundaries with the intent of changing their shape or direction (i.e., engaging
in acts of equity) within the research interactions. We now turn to two examples of
stretching and expanding current second-order models.
Example 1: Addressing gender equity in interactions
Acts of Equity in an Interaction
During a recent teaching experiment that the second author conducted with two
preservice secondary mathematics teachers, a graduate student interested in gen-
dered patterns of interaction pointed out that the two participants were engaged in
some gender-normative patterns of interaction related to their expressions of confi-
dence in their mathematical thinking (Lubienski & Ganley, 2017; Bench et al.,
2015) and how they each attended to the other person’s thinking. Specifically, the
female participant was quite attentive to her male partner’s mathematical language
and reasoning. She regularly formed a working second-order model of her partner’s
mathematical thinking. Although she was a powerful reasoner herself, she also
often doubted her mathematical thinking in these interactions. Her male partner, on
the other hand, rarely doubted his mathematical thinking and, at the same time, had
difficulty making sense of his female partner’s thinking (Ippolito et al., 2021). These
gender-normative patterns of interaction, which were linked to the social identities
of the participants and researcher, were one component of the mathematical interac-
tions in the teaching experiment.
As part of the teaching experiment, the teacher-researcher formed the goal of
disrupting these gender-normative patterns of interaction by framing each partici-
pant as possessing a strength—the female participant being a powerful mathemati-
cal reasoner who could make strong interpretations of her partner’s mathematical
thinking and the male participant having a high level of confidence in his mathemat-
ical thinking. He asked each participant to work on the other person’s strength rela-
tive to themselves—the female participant to work on being more confident in her
mathematical thinking and the male participant to work on understanding his part-
ners’ thinking. The preservice teachers appeared to appreciate the teacher-
researcher’s suggestion at the time he made it, and they willingly took it up in future
teaching episodes.
The teacher-researcher intentionally framed each participant as having a strength.
That is, the teacher-researcher’s intent was to focus on the assets of each participant
rather than framing either as having a deficit. Moreover, it would have been possible
for the teacher-researcher to either ignore the gender-normative patterns of interac-

tion or to simply frame the male participant as being the source of the problem. We
do not see either of these options as likely to bring about changes in the pattern of
interaction. In fact, we see the teacher-researcher as having a greater responsibility
to ask for and to support these changes because he was the facilitator of the interac-
tions in the study. For the remainder of the experiment, the teacher-researcher
brought the participants back to this conversation so that they had the opportunity to
work on disrupting the gender-normative patterns of interaction.
We see the teacher-researcher and participants’ actions as an example of an act
of equity in that the researcher and participants acted on a social boundary with the
intent of changing its shape or direction in order to address known inequities in pat-
terns of participation related to the gender identities of the participants and research-
ers. Moreover, we consider the participants’ willingness to work on disrupting
gender-normative patterns of interaction (i.e., to work on the other person’s strength)
as evidence that, at least at the level of a working model, the teaching action pro-
duced a more equitable outcome. Here, by a more equitable outcome, we mean that
the teacher-researcher and male participant acknowledged in new ways the impor-
tance of the female participant’s mathematical contributions to the interaction, a
form of respect. It also highlighted for the male participant the importance of work-
ing to understand the thinking of others as a teacher, which we see as a basis for
supporting him to engage in potential acts of respect for his future students.
This observation, however, does not specifically tie the teaching actions to the
study of mathematical learning. To do so would require an analysis of how the
actions produced greater opportunities for learning or actual occasions of learning
itself. That is, it would involve creating a second-order model that attends to link-
ages between the teaching action and participants’ learning. We turn to outlining the
theoretical tools that we think would be useful for doing so.
Making Second-Order Models That Include Acts of Equity in Interactions
We do not frame teaching actions as causally related to the construction of specific

schemes and operations. Steffe (1996) has differentiated between two domains of
interaction: interaction among the constructs within a person and individual-
environment interaction, which includes interactions with the physical, social, and
cultural environment. Interaction in one domain is not causally related to interaction
in the other domain. The noncausal relationship, however, does not preclude estab-
lishing linkages between the two domains in the creation of second-order models.
We see Thompson’s (2013) framing of communication among two or more people
that involves reciprocal assimilation as one theoretical tool that could be used to
establish such an analysis.
We return to the specific example to suggest what such an analysis could look
like. Although not causally related, we see the teaching action as having the poten-
tial to open opportunities for learning for both participants. For the male participant,
making a working second-order model of another person’s mathematical reasoning
has the potential to be a source of new learning because it could provide new inputs
against which to compare his mathematical reasoning. For the female participant,
developing confidence in her own reasoning has the potential to support reflective
abstraction (see Ellis et al., Chap. 6, this volume, and Tallman & O’Bryan, Chap. 8,
this volume) because it could promote heightened attention to her own schemes and
operations as a source for reflection.
Thus, creating a second-order model that attends to this feature of relationships
between the two domains of interaction could entail analyzing the data for linkages
between the two domains. For example, it could involve analyzing changes in the
schemes of the male participant as he listened to the female participant’s reasoning
and changes in the female participant’s schemes in moments when she prioritized
reflection on her own reasoning over attending to her partner’s reasoning.
This type of analysis would support an expansion of second-order models in that
the models would seek to draw connections between the schemes and operations of
the participants (Steffe’s intraindividual domain of interaction) and components of
the individual-environment interaction for the participants. We see this type of anal-
ysis as an expansion for researchers working within this tradition because current
analyses have tended to focus more heavily on the intraindividual domain of inter-
action, with less attention given to how the two domains of interaction are related to
each other. We see this work as also stretching prior second-order models in that the
genesis of the teaching action was based on the researchers’ knowledge of historical
inequities related to gendered patterns of interaction, and changing the patterns of
interaction could contribute to changes in each preservice teacher’s future.7
We also note that we see this type of analysis as augmenting the analysis that we
proposed in the prior section. That is, this type of analysis supports tying the more
equitable outcomes to the participants’ mathematical learning not just to changes
related to gender-normative patterns of interaction.
We turn now to a second example. Our second example is different from our first
because the design of the second study was specifically tied to an equity goal.
Example 2: Designing to address equity
Designing Interactions to Address an Equity Issue
A second way to expand second-order models is by designing mathematical work

on an issue of social justice into a study in order to make models of students’ math-
ematical reasoning and of students’ understanding of the equity-related issue. For
example, the third author of this chapter designed a mathematical intervention in
which he had middle grade students explore racial bias in jury selection (Gatza,
2021). There were five participants in his study, most of whom were students in his
7
We think that such teaching moves help research participants see new possibilities for themselves
and others that they may carry forward if they have other similar experiences.
eighth-grade Algebra class. Of the five participants, three self-identified as Black/

African-American, one self-identified as Hispanic, and one self-identified as multi-
racial, Black and White.
Gatza conducted the design study after school. His analysis focused on two stu-
dents, one African-American male and one Hispanic male, who participated in 18
teaching episodes. Gatza began with an initial interview in which he identified key
features of students’ multiplicative reasoning coupled with key features of their
understanding of race, racial identity, racism, and racial bias. This initial interview
was designed so that the researcher could share information about how he thought
about his own racial identity and to allow the students to share their understandings
of their racial identity.8
The broad intent of the study was to understand how students’ mathematical
reasoning impacted their understanding of racial bias and how their understanding
of racial bias impacted their mathematical reasoning. Gatza (2021) outlined how
students initially interpreted quadratic situations, situations that were mathemati-
cally similar to a jury selection process of two jurors, by using schemes that involved
inferences based on linear meanings of multiplication. Later in his study, the stu-
dents developed schemes that allowed them to interpret the jury selection process as
involving a quadratic relationship, which in turn supported them to make mathemat-
ically correct inferences about the likelihood of racial bias in a jury selection pro-
cess involving two jurors. Thus, the mathematical schemes that they established
informed their reasoning about racial bias.
As part of the study, Gatza (2021) also worked with students on differentiating
conscious from unconscious bias as one way to support explanations of how an
actual outcome could differ significantly from an expected outcome. He started on
ideas about conscious and unconscious bias by having students examine a flyer
from the American Red Cross about pool safety that showed children phenotypi-
cally presenting as White labeled as acting good and children phenotypically pre-
senting as children of color as acting bad. The students initially interpreted the flyer
as not involving bias but simply as happenstance. Through multiple discussions of
the meaning of both conscious and unconscious bias that occurred over time, the
students were able to interpret what each form of bias would “look like” when they
analyzed artifacts like the flyer. They were, then, able to apply these ideas to the jury
selection process to reason about factors that could produce differences between
actual and expected outcomes. Thus, their development of schemes related to racial
bias supported the development of mathematical understandings.
Within the interactions, Gatza’s (2021) design study aimed to cultivate respect
for the individual participants by basing the teaching episodes on the students’
evolving mathematical and racial bias schemes. Moreover, the design was tied to a
social and historical inequity, racial bias. Researchers have found that this particular
inequity impacts many facets of Black and Latinx students’ lives, including their
experiences in schools (Gutiérrez, 2018; Kozol, 2005; Ladson-Billings, 2006;
8
The students did not initially differentiate between racial and cultural identity.
Martin, 2009) and that Black and Latinx students who have strong understandings
of race, racism, and racial bias are more likely to thrive in schools (Carter, 2008;
Oyserman et al., 2007). Thus, there was an explicit intent on the part of the teacher-
researcher to design contexts that would open opportunities for students to act upon
the social boundaries that constrain and enable fields of action by changing their
shape or direction. Moreover, throughout the study, there was an indication that the
participants willingly engaged with this intent, working to develop their ideas about
race, racism, and racial bias.
We note that a common feature of studies similar to this one is that they involve
a component of social action that is related to, but broader than, the study itself—for
example, contacting local leaders about an issue (e.g., Gutstein, 2006). We agree
that social action can be one outcome of such a study, but here we emphasize that
an important outcome that we consider transformative is the learning that occurs for
the teacher-researcher and participants.
Making Second-Order Models from the Study
The summary above does not capture the nuance in the development of the partici-
pants’ reasoning (see Gatza, 2021, for a full report). It does, however, indicate both
how second-order models can be stretched and expanded. We see the stretch in
Gatza’s work as his engagement with research related to race, racism, and racial
bias that he used as a tool to design the study. That is, the content of the design was
informed by his understanding of the role that race has played in shaping facets of
the US legal system over time.
Gatza’s (2021) work expands on current second-order models by attending to
two kinds of schemes and how they informed one another: racial bias schemes and
mathematical schemes. We see this kind of expansion as different from the first, in
that the analysis falls into Steffe’s intra-domain of interaction; that is, it examined
the ways in which participants’ schemes changed over time and how these schemes
were related to each other. However, the lens in Gatza’s study was broader than in
many studies in the tradition of research explored in this book, in that many studies
attend only to mathematical schemes, whereas Gatza needed to include additional
constructs to analyze the development of students’ schemes for racial bias.
Looking Ahead
We now summarize our points and consider ways in which teacher-researchers who
make second-order models can further develop this work.
First, we have argued that making and using second-order models and epistemic
students, as we have portrayed them in this chapter, are acts of equity in research.
They are acts of respect for students, and when using second-order models and
epistemic students in interaction with particular students, they are an active form of
respect. Furthermore, using epistemic students in interaction can influence how

power affects people in a research setting by allowing a teacher-researcher to influ-
ence the social boundaries that define fields of action and disrupt the inequitable
norm of students being tasked with learning the first-order mathematical knowledge
of their teacher.
Second, we have demonstrated with two examples how to augment second-order
models and epistemic students to better address equity. In the first example, we
demonstrated how the teacher-researcher addressed the local dynamics in the inter-
action by relating it to broader gender norms and patterns of interaction. So, this
example was an instance of stretching into social-historical knowledge about
gender-normative patterns of behavior and also expanding what the teacher-
researcher attended to in the interactions, namely, the impacts of the social identities
of the participants. The stretching and expanding then influenced how he interacted
with the two participants. In the second example, we demonstrated how the teacher-
researcher designed a study to address an issue of equity in conjunction with math-
ematical reasoning. This example was also an instance of both stretching into
social-historical knowledge about racial bias in jury selection over time in the US
and expanding what the teacher-researcher attended to in the interactions, namely,
social identity categories in relation to jury selection and his participants’ construc-
tion of their own racial identities. This instance of stretching and expanding then
allowed him to study and support participants’ construction of mathematical and
racial bias schemes.
Based on our writing of this chapter, we have recommendations for those build-
ing second-order models. First, write a positionality statement and allow your par-
ticipants to characterize their social identity categories and social identities. Do
research to understand the histories of the groups your participants identify with.
Consider the ways that the nonoverlapping social identity categories and social
identities of you and your participants may affect the interactions of you and your
participants and, therefore, the models you make.
Be vigilant about this issue: The misuse of epistemic students could reify deficit
narratives. Anticipate how readers might interpret claims from a study in relation to
deficit narratives, and provide explicit guidance to avoid this kind of reification.
Document acts of equity and how they influenced learning or opportunities for
learning. Design a study to make a model of participants’ mathematical reasoning
and their construction of an issue related to equity.
That said, what we have discussed in this chapter is just a starting point. We now
ask three questions to frame further work that is needed to continue to articulate
how equity is foundational to this tradition of research and to augment second-
order models.
The first question is: How can making and using second-order models sup-
port advocating for equity in standards and curricula? A central purpose of
research on second-order models and epistemic students has been to base school
mathematics, or mathematics for students, on second-order models of students’
mathematics, which Steffe (2010b) calls the mathematics of students. The proposal
to base mathematics for students on the mathematics of students differs from the
current basis of many common curricular materials and standards, including the
Common Core State Standards for Mathematics (CCSS-M) (National Governors
Association Center for Best Practices and Chief Council of State School Officers
[NGACBP & CCSSO], 2010). For example, consider this third-grade standard:
“Understand a fraction 1/b as the quantity formed by 1 part when a whole is parti-
tioned into b equal parts; understand a fraction a/b as the quantity formed by a parts
of size 1/b.” This standard communicates one way to think about fractions. However,
as we have explained, the thematization of second-order models can lead to three or
four epistemic students at a grade level (Steffe, 2017). These different epistemic
students structure numbers and quantities differently, with implications for the con-
struction of fractions (Steffe & Olive, 2010). So, there are (at least) three or four
ways that students in third grade are likely to conceive of fractions, with this stan-
dard representing a relatively advanced conception (Steffe, 2010a).
Yet, other ways of conceiving of fractions are simply absent from the CCSS-M
(NGACBP & CCSSO, 2010). Instead, a “common sense,” first-order adult under-
standing of a fraction is presented. We see this issue as a pervasive problem with
current standards because they communicate to curriculum designers and teachers
what should be taught; thus, we see the standards as implicitly taking the stance that
adults’ first-order knowledge is what should be taught. A great deal more could be
explored here, in the quest to conceptualize what standards that respect student
knowledge and are based on second-order models and epistemic students could
look like and how that conceptualization could provide a basis for the design of
more robust curricular resources that better meet the needs of students.
We also think that the ways we have proposed to expand and stretch second-
order models can be a site for the development of novel curricular resources. That
is, Gatza’s (2021) design study produced an array of curricular materials that could
be used by others. When possible, we suggest that teacher-researchers make these
resources available, along with guidelines for their use. These guidelines might
include descriptions of the social identity categories and social identities of the stu-
dents with whom they were used as well as factors that were necessary and contin-
gent for the learning that was documented as part of the study.
The second question is: How can making and using second-order models sup-
port equity in work with teachers? Since we have argued that making and using
second-order models is, at its foundation, an equitable practice, a natural question is
how this practice might be useful to and taken up by classroom teachers. Research
to support teachers in constructing working second-order models of their students is
underway (Bas-Ader & Carlson, 2021; Carlson et al., Chap. 9, this volume; Liang,
2021; Teuscher et al., 2016). So far, most of this research has focused on supporting
teachers to decenter, a critical step in making a working second-order model of
students. Bas-Ader and Carlson developed their decentering framework based on
teachers of college students, so much more needs to be done to expand to K-12
teachers. There are multiple aspects that require research, including how teachers
make working models, what resources teachers draw upon to make their models,
how teachers use these models in instruction, whether using these models disrupts
the norm of students being tasked with learning the first-order knowledge of their
teacher or other inequitable norms, and what professional learning experiences sup-
port teachers in this entire process. The ways we have proposed to expand and
stretch second-order models are an important component in the work with teachers.
We also think that the work to link second-order models to instructional moves
like those aimed at disrupting gender-normative patterns of interaction has the
potential to provide linked accounts of how such moves afford and constrain learn-
ing and opportunities for learning. We see that as important because, for example, it
is still an open question as to what kinds of instructional practices might support
acts of equity and learning, particularly among teacher-researchers who give
detailed accounts of learning. Doing so can help to link accounts of learning to acts
of equity in an interaction, providing an avenue to more robustly theorize about the
relationship between the two.
The third question is: What conceptual foundations need to be examined and
refined in order to develop ideas about power, privilege, and society from a
radical constructivist perspective? Our basic stance in writing this chapter is that
radical constructivism, as a theory of knowing, has to be paired with first- and
second-order models in many domains in order to develop the theory and its useful-
ness. For example, radical constructivism has already been augmented by a model
of communication (Thompson, 1999, 2013), as well as second-order models of spe-
cific mathematical domains like students’ construction of whole numbers (Steffe &
Cobb, 1988; Steffe et al., 1983). Another domain that has been addressed is ethics,
where authors have discussed in what ways models of ethics are related to radical
constructivism (e.g., Gash, 2000; Larochelle, 2000; Lewin, 2000). An overarching
conclusion is that radical constructivism does not point to a particular stance on eth-
ics, but it “predicts that ethical considerations will emerge as humans interact and
form expectations about others’ actions” (Thompson, 2000, p. 299).
We consider our chapter to be a (first) step in contributing to a model of equity
related to radical constructivism. For example, in our definitions of power and iden-
tity, we have borrowed from research traditions with theoretical assumptions that
differ from those of radical constructivism. We have endeavored to use and adapt
definitions to fit with radical constructivism. However, we know that these con-
structs need further refinement to solidly fit with radical constructivism and to dis-
tinguish them from other constructs rooted in other traditions. We see our chapter as
opening a conversation. What refinements to our definitions, or alternative defini-
tions, would be useful? Other questions include the following: How do teacher-
researchers build up models of power in society? How do they build up models of
historical power relations? How do they build up a model of society (cf. von
Glasersfeld, 2008)? How do these models inform the making of second-order mod-
els and epistemic students? What are the implications for working toward a more
equitable mathematics education for students? We ask these questions to support
continued elaboration of how to stretch and expand second-order models in order to
advance equity and disrupt inequities.
References
Adiredja, A. P. (2019). Anti-deficit narratives: Engaging the politics of research on mathematical

sense-making. Journal for Research in Mathematics Education, 50(4), 401–435.
Adiredja, A. P., & Louie, N. (2020). Untangling the web of deficit discourses in mathematics
education. For the Learning of Mathematics, 40(1), 42–46. https://proxyiub.uits.iu.edu/
login?url=https://search.ebscohost.com/login.aspx?direct=true&db=edsjsr&AN=edsjs
r.27091140&site=eds-live&scope=site
Aguirre, J., Mayfield-Ingram, K., & Martin, D. (2013). The impact of identity in K-8 mathematics:
Rethinking equity-based practices. National Council of Teachers of Mathematics (NCTM).
Aguirre, J., Herbel-Eisenmann, B., Celedón-Pattichis, S., Civil, M., Wilkerson, T., Stephan, M.,
Pape, S., & Clements, D. H. (2017). Equity within mathematics education research as a politi-
cal act: Moving from choice to intentional collective professional responsibility. Journal for
Ball, S. J. (2013). Foucault, power, and education. Routledge.
Barajas-López, F., & Larnell, G. V. (2019). Unpacking the links between equitable teaching prac-
tices and standards for mathematical practice: Equity for whom and under what conditions?
Journal for Research in Mathematics Education, 50(4), 349–361. https://doi.org/10.5951/
jresematheduc.50.4.0349
Bartell, T. G., & Johnson, K. R. (2013). Making unseen privilege visible in mathematics education
research. Journal of Urban Mathematics Education, 6(1), 35–44.
Bas-Ader, S., & Carlson, M. P. (2021). Decentering framework: A characterization of graduate
and Learning. https://doi.org/10.1080/10986065.2020.1844608
Battey, D., & Franke, M. (2015). Integrating professional development on mathematics and equity:
Countering deficit views of students of color. Education and Urban Society, 47(4), 433–462.
https://doi.org/10.1177/0013124513497788
Battey, D., & Leyva, L. A. (2016). A framework for understanding whiteness in mathematics edu-
cation. Journal of Urban Mathematics Education, 9(2), 49–80.
Bench, S. W., Lench, H. C., Liew, J., Miner, K., & Flores, S. A. (2015). Gender gaps in overestima-
tion of math performance. Sex Roles, 72(11), 536–546.
Berry, R. Q. (2015). Addressing the needs of the marginalized students in school mathematics:
A review of policies and reforms. In T. G. Bartell, K. N. Bieda, R. T. Putnam, K. Bradfield,
& H. Dominguez (Eds.), Proceedings of the 37th annual meeting of the North American
Chapter of the International Group for the Psychology of Mathematics Education (pp. 19–32).
Michigan State University.
D. Riedel.
Carter, D. J. (2008). Cultivating a critical race consciousness for African American school success.
The Journal of Educational Foundations, 22, 11–28.
Clark, L., Badertscher, E., & Napp, C. (2013). African American mathematics teachers as agents
in their African American students’ mathematics identity formation. Teachers College Record,
115(2), 1–36.
Clements, D. H., & Sarama, J. (2009). Learning trajectories in early mathematics–sequences of
acquisition and teaching. Encyclopedia of Language and Literacy Development, 7, 1–6.
Cobb, P., Gresalfi, M., & Hodge, L. L. (2009). An interpretive scheme for analyzing the identi-
ties that students develop in mathematics classrooms. Journal for Research in Mathematics
Education, 40(1), 40–68.
Collins, P. H., & Bilge, S. (2020). Intersectionality. Wiley.
D’Ambrosio, B., Martin, D. B., Frankenstein, M., Moschkovich, J., Gutierrez, R., Taylor, E.,
Kastberg, S., & Barnes, D. (2013a). Addressing racism. Journal for Research in Mathematics
Education, 44(1), 23–36.
Kastberg, S., & Barnes, D. (2013b). Introduction to the JRME equity special issue. Journal for
Kastberg, S., & Barnes, D. (2013c). Positioning oneself in mathematics education research.
de Freitas, E., Wagner, D., Esmonde, I., Knipping, C., Lunney Borden, L., & Reid, D. (2012).
Discursive authority and sociocultural positioning in the mathematics classroom: New direc-
tions for teacher professional development. Canadian Journal of Science, Mathematics and
Technology Education, 12(2), 137–159.
Ellis, A. B. (2022). Decentering to build asset-based learning trajectories. In A. E. Lischka,
E. B. Dyer, R. S. Jones, J. N. Lovett, J. Strayer, & S. Drown (Eds.), Proceedings of the forty-
fourth annual meeting of the North American Chapter of the International Group for the
Psychology of Mathematics Education (pp. 15–29). Middle Tennessee State University.
Feagin, J. R. (2010). Racist American: Roots, current realities, and future reparations (2nd ed.).
Routledge.
Foucault, M. (1980). Power/knowledge: Selected interviews and other writings, 1972–1977
(C. Gordon, Trans.). Pantheon Books.
Freire, P. (1970). Pedagogy of the oppressed (M. B. Ramos, Trans.). Continuum.
Gash, H. (2000). Epistemological origins of ethics. In L. P. Steffe & P. W. Thompson (Eds.),
Radical constructivism in action (Vol. 15, pp. 80–90). RoutledgeFalmer
Gatza, A. M. (2021). Not just mathematics, “just” mathematics: Investigating mathematical learn-
ing and critical race consciousness. Doctoral dissertation, Indiana University, Bloomington.
ProQuest.
Gholson, M., & Martin, D. B. (2014). Smart girls, Black girls, mean girls, and bullies: At the
intersection of identities and the mediating role of young girls’ social network in mathematical
communities of practice. Journal of Education, 194(1), 19–33.
Gholson, M., & Martin, D. B. (2019). Blackgirl face: Racialized and gendered performativity
in mathematical contexts. Zentralblatt für Didaktik der Mathematik, 51, 391–404. https://doi.
org/10.1007/s11858-019-01051-x
Gimenez, M. E. (2014). Latino/“Hispanic”—Who needs a name? The case against a standard-
ized terminology. In A. Darder & R. Torres (Eds.), Latinos and education: A critical reader
(pp. 93–104). Routledge.
Gutiérrez, R. (2002). Enabling the practice of mathematics teachers in context: Toward a new
equity research agenda. Mathematical Thinking & Learning, 4(2/3), 145–187. https://doi.
org/10.1207/S15327833MTL04023_4
Gutiérrez, R. (2008). Research commentary: A gap-gazing fetish in mathematics education?
Problematizing research on the achievement gap. Journal for Research in Mathematics
Education, 39(4), 357–364.
Gutiérrez, R. (2017). Why mathematics (education) was late to the backlash party: The need for a
revolution. Journal of Urban Mathematics Education, 10(2), 8–24.
Gutiérrez, R. (2018). The need to rehumanize mathematics. In I. Goffney & R. Gutierrez (Eds.),
Rehumanizing mathematics for Black, Indigenous, and Latinx students (pp. 1–10). NCTM.
Gutstein, E. (2006). Reading and writing the world with mathematics: Toward a pedagogy for
social justice. Routledge.
Gutstein, E. (2016). “Our issues, our people—Math as our weapon”: Critical mathematics in a
Chicago neighborhood high school. Journal for Research in Mathematics Education, 47(5),
454–504.
Hackenberg, A. J. (2010). Mathematical caring relations in action. Journal for Research in
Mathematics Education, 41(3), 236–273. https://doi.org/10.5951/jresematheduc.41.3.0236
Hackenberg, A. J., & Sevinc, S. (2021). A boundary of the second multiplicative concept: The case
of Milo. Educational Studies in Mathematics. https://doi.org/10.1007/s10649-021-10083-8
Hackenberg, A. J., Walsh, P. A., & Valero, J. (2021). A case of units coordination stage change in
middle school. In D. Olanoff, K. Johnson, & S. M. Spitzer (Eds.), Proceedings of the forty-third
annual meeting of the North American chapter of the International Group for the Psychology
of mathematics education (pp. 1281–1286). Towson University.
Hackenberg, A. J., Aydeniz, F., & Borowksi, R. (2023). Middle school students at three stages of
units coordination learn to make same speeds. Journal of Mathematical Behavior. https://doi.
org/10.1016/j.jmathb.2023.101085
Hayward, C. R. (1998). De-facing power. Polity, 31(1), 1–22. https://doi.org/10.2307/3235365
Henrich, J., Heine, S. J., & Norenzayan, A. (2010). Most people are not WEIRD. Nature,
466(7302), 29–29.
Ippolito, D., Ataide Pinheiro, W., & Liu, J. (2021, July 11–18). Gender differences in student-student
interaction (Conference presentation). The 14th International Congress on Mathematical
Education, Shanghai, China.
Kokka, K. (2022). Toward a theory of affective pedagogical goals for social justice mathemat-
ics. Journal for Research in Mathematics Education, 53(2), 133–153. https://doi.org/10.5951/
jresematheduc-2020-0270
Kozol, J. (2005). The shame of a nation: The restoration of the apartheid schooling in America.
Three Rivers Press.
Ladson-Billings, G. (2006). From the achievement gap to the education debt: Understanding
achievement in U.S. schools. Educational Researcher, 35(7), 3–12.
Langer-Osuna, J. M., & Esmonde, I. (2017). Identity in research on mathematics education. In
J. Cai (Ed.), First compendium for research in mathematics education (pp. 637–648). NCTM.
Larochelle, M. (2000). Radical constructivism: Notes on viability, ethics, and other educational
issues. In L. P. Steffe & P. W. Thompson (Eds.), Radical constructivism in action (Vol. 15,
pp. 55–68). RoutledgeFalmer
Leonardo, Z. (2009). Reading whiteness: Anti-racist pedagogy against white racial knowledge. In
W. Ayers, T. Quinn, & D. Stovall (Eds.), Handbook of social justice in education (pp. 231–248).
Routledge.
Lewin, P. (2000). Constructivism and paideia. In L. P. Steffe & P. W. Thompson (Eds.), Radical
constructivism in action (Vol. 15, pp. 37–54). RoutledgeFalmer
Liang, B. (2021). Learning about and learning from students: Two teachers’ construction of stu-
dents’ mathematical meanings through student-teacher interactions. Doctoral dissertation, The
Liljedahl, P. (2021). Building thinking classrooms in mathematics, grades K-12: 14 teaching prac-
tices for enhancing learning. Corwin Press, Inc.
Lipsitz, G. (1995). The possessive investment in whiteness: Racialized social democracy and the
‘white’ problem in American studies. American Quarterly, 47(3), 369–387.
Lubienski, S. T., & Ganley, C. M. (2017). Research on gender and mathematics. In J. Cai (Ed.),
First compendium for research in mathematics education (pp. 649–666). NCTM.
MacDonald, B. L., & Wilkins, J. L. (2019). Subitising activity relative to units construction: A case
study. Research in Mathematics Education, 21(1), 77–95.
Maloney, T., & Matthews, J. S. (2020). Teacher care and students’ sense of connectedness in
the urban mathematics classroom. Journal for Research in Mathematics Education, 51(4),
399–432.
Martin, D. B. (2000). Mathematics success and failure among African- American youth. Erlbaum.
Martin, D. B. (2009). Researching race in mathematics education. The Teachers College Record,
111(2), 295–338.
Martin, D. B., Anderson, C. R., & Shah, N. (2017). Race and mathematics education. In J. Cai
(Ed.), First compendium for research in mathematics education (pp. 607–636). NCTM.
McClain, K., & Cobb, P. (2001). An analysis of development of sociomathematical norms in one
first-grade classroom. Journal for Research in Mathematics Education, 32(3), 236–266.
Mendoza, E., Hand, V., van Es, E. A., Hoos, S., & Frierson, M. (2021). ‘The ability to lay your-
self bare’: Centering rupture, inherited conversations, and vulnerability in professional devel-
opment. Professional Development in Education, 47(2–3), 243–256. https://doi.org/10.108
0/19415257.2021.1891955
Mertens, D. (2010). Research and evaluation in education and psychology: Integrating diversity
with quantitative, qualitative, and mixed methods (3rd ed.). Sage.
National Governors Association Center for Best Practices & Council of Chief State School Officers.
(2010). Common core state standards for mathematics. Author. http://corestandards.org/
National Museum of African-American History and Culture [NMAAHC]. (2023, January 19).
Whiteness. NMAAHC. https://nmaahc.si.edu/learn/talking-about-race/topics/whiteness
Oakes, J., Joseph, R., & Muir, K. (2004). Access and achievement in mathematics and science:
Inequalities that endure and change. In J.A. Banks & C.A.M. Banks (Eds.), Handbook of
research on multicultural education (Vol. 2, pp. 69–90). Routledge Falmer.
Oyserman, D., Brickman, D., & Rhodes, M. (2007). Racial-ethnic identity in adolescence: Content
and consequences for African American and Latino youth. In A. Fuligni (Ed.), Social catego-
ries, identities and educational participation (pp. 91–114). Russell-Sage.
Shah, N. (2017). Race, ideology, and academic ability: A relational analysis of racial narratives in
mathematics. Teachers College Record, 119(7), 1–42.
Solomon, Y., Lawson, D., & Croft, T. (2011). Dealing with “fragile identities”: Resistance and
refiguring in women mathematics students. Gender and Education, 23(5), 565–583.
Steffe, L. P. (1994). Children’s multiplying schemes. In G. Harel & J. Confrey (Eds.), The develop-
ment of multiplicative reasoning in the learning of mathematics (pp. 3–39). State University
of New York Press.
Steffe, L. P. (1996). Social-cultural approaches in early childhood mathematics education: A dis-
cussion. In H. Mansfield, N. A. Pateman, & N. Bednarz (Eds.), Mathematics for tomorrow’s
young children (pp. 79–99). Kluwer. https://doi.org/10.1007/978-94-017-2211-7
Steffe, L. P.(2010a). The partitive and part-whole schemes. In L. P. Steffe & J. Olive (Eds.),
Steffe, L. P. (2010b). Perspectives on children’s fraction knowledge. In L. P. Steffe & J. Olive
Steffe, L. P. (2017). Psychology in mathematics education: Past, present, and future. In E. Galindo
& J. Newton (Eds.), Proceedings of the 39th annual meeting of the North American Chapter
of the International Group for the Psychology of Mathematics Education (pp. 27–56). Hoosier
Association of Mathematics Teacher Educators.
Springer-Verlag.
reaction to Simon. Journal for Research in Mathematics Education, 26(2), 146–159.
Steffe, L. P., & Olive, J. (2010). Children’s fractional knowledge. Springer. https://doi.
org/10.1007/978-1-4419-0591-8
Steffe, L. P., & Ulrich, C. (2010). Equipartitioning operations for connected numbers: Their
use and interiorization. In L. P. Steffe & J. Olive (Eds.), Children’s fractional knowledge
Steffe, L. P., & Wiegel, H. G. (1996). On the nature of a model of mathematical learning. In
L. P. Steffe, P. Nesher, P. Cobb, G. A. Goldin, & B. Greer (Eds.), Theories of mathematical
learning (pp. 477–498). Erlbaum.
Stephan, M., Register, J., Reinke, L., Robinson, C., Pugalenthi, P., & Pugalee, D. (2021). People use
math as a weapon: Critical mathematics consciousness in the time of COVID-19. Educational
Stinson, D. (2013). Negotiating the “White male math myth”: African American male students and
success in school mathematics. Journal of Research in Mathematics Education, 44(1), 69–99.
Teuscher, D., Moore, K., & Carlson, M. (2016). Decentering: A construct to analyze and explain
teacher actions as they relate to student thinking. Journal of Mathematics Teacher Education,
19(5), 433–456. https://doi.org/10.1007/s10857-015-9304-0
Thompson, P. W. (1999). Some remarks on conventions and representations. In F. Hitt &
M. Santos (Eds.), Proceedings of the 21st annual meeting of the north American chapter of the
International Group for the Psychology of mathematics education. ERIC Clearinghouse for
Science, Mathematics, and Environmental Education.
Thompson, P. W. (2000). Radical constructivism: Reflections and directions. In L. P. Steffe
& P. W. Thompson (Eds.), Radical constructivism in action (Vol. 15, pp. 291–315).
RoutledgeFalmer
Thompson, P. W. (2013). In the absence of meaning…. In K. Leatham (Ed.), Vital directions for
Thompson, P. W. (2016). Researching mathematical meanings for teaching. In L. D. English
& D. Kirshner (Eds.), Handbook of international research in mathematics education
(pp. 435–461). Taylor & Francis.
representation of these problems using arrays. Journal of Mathematical Behavior, 52, 1–20.
Tillema, E. S. (2020). Students’ solution of arrangement problems and their connection to
Cartesian product problems. Mathematical Thinking and Learning, 22, 23–55. https://doi.org/1
0.1080/10986065.2019.1608618
Tillema, E. S., & Hackenberg, A. J. (2017). Three facets of equity in Steffe’s research programs. In
E. Galindo & J. Newton (Eds.), Proceedings of the 39th annual meeting of the North American
Chapter of the International Group for the Psychology of Mathematics Education (pp. 57–67).
Hoosier Association of Mathematics Teacher Educators.
Tyack, D. (1974). The one best system: A history of American urban education. Harvard
University Press.
Ulrich, C. (2016). The tacitly nested number sequence in sixth grade: The case of Adam. Journal
Valencia, R. R. (2010). Dismantling contemporary deficit thinking. Routledge.
van Es, E. A., Hand, V., Agarwal, P., & Sandoval, C. (2022). Multidimensional noticing for equity:
Theorizing mathematics teachers’ systems of noticing to disrupt inequities. Journal for Research
in Mathematics Education, 53(2), 114–132. https://doi.org/10.5951/jresematheduc-2019-0018
von Glasersfeld, E. (1983). Learning as constructive activity. In J. C. Bergeron & N. Herscovics
(Eds.), Proceedings of the 5th annual meeting of the north American Group of Psychology in
mathematics education (Vol. 1, pp. 41–101). Montreal.
von Glasersfeld, E. (1984). An introduction to radical constructivism. In P. Watzlawick (Ed.), The
invented reality (pp. 17–40). Norton.
von Glasersfeld, E. (2008). Who conceives of society? Constructivist Foundations, 3(2), 59–64.
Walshaw, M. (2013). Post-structuralism and ethical practical action: Issues of identity and power.
Wright, R. J., Martland, J., & Stafford, A. (2006). Early numeracy: Assessment for teaching and
intervention (2nd ed.). Sage.
Wright, R. J., Ellemor-Collins, D., & Tabor, P. (2012). Developing number knowledge: Assessment,
teaching and intervention with 7- to 11-year-olds. Sage.
Zahner, W. (2019). Does participant selection skew mathematics education research find-
ings? Considering quantitative reasoning research. In S. Otten, A. G. Candela, Z. de Araujo,
C. Haines, & C. Munter (Eds.), Proceedings of the forty-first annual meeting of the north
American chapter of the International Group for the Psychology of mathematics education
(pp. 1466–1471). University of Missouri.
Zwanch, K., & Wilkins, J. L. (2021). Releasing the conceptual spring to construct multiplicative
reasoning. Educational Studies in Mathematics, 106(1), 151–170.
Chapter 15
Reflections on the Power of Genetic
Epistemology by the Modern Cognitive
Psychologist
Percival Matthews and Alexandria Viegut
Reflections of Respectful Tourists
As cognitive developmental psychologists who study mathematical cognition, we

would like, on the front end, to express our deep appreciation for this volume and
what it has done for our perspectives. We both are products of a world inhabited
primarily by psychologists, cognitive scientists, and neuroscientists, but we have
worked hard over the years to continue to try to keep a toehold in mathematics edu-
cation research – to explore the space as frequent and respectful tourists.1 We do this
because, despite our affinities for our chosen field, we also appreciate the nuances
and limits of the methods that typically characterize psychological approaches to
mathematical cognition. We strongly believe in methodological pluralism – or
minimally, an appreciation for others’ uses of alternative methods – as a path for
1
We acknowledge that there are real difficulties in delimiting psychology and mathematics educa-
tion research at the boundaries, as things become fuzzy at the margins. That said, we find it uncon-
troversial to suggest that they constitute different disciplines, with different degrees, different
flagship journals, and somewhat divergent methods. If we are to grant that there is an epistemic
subject in genetic epistemology, as suggested in some chapters in this volume, we would hope to
be granted the grace to posit a sort of epistemic psychologist who can be differentiated from the
epistemic mathematics education researcher.
P. Matthews (*)
Department of Educational Psychology, University of Wisconsin–Madison,
Madison, WI, USA
e-mail: pmatthews@wisc.edu
A. Viegut
Department of Psychology, University of Wisconsin-Eau Claire, Eau Claire, WI, USA
e-mail: viegutaa@uwec.edu
https://doi.org/10.1007/978-3-031-47386-9_15
512 P. Matthews and A. Viegut
accounting for our own blind spots. As such, we have an abiding respect for what
mathematics education research brings to the table to complement, supplement, and
even to contest psychological approaches.
In fact, our appreciation of the diversity of methods for studying mathematical
cognition frequently borders on something like awe. The sheer diversity of
approaches employed by people who purport to study the same topic is impressive.
When considering the contents of the current volume, we find the contrast between
fields particularly striking, as the work centers Piaget, an absolute giant in the field
of developmental psychology, historically speaking. To be sure, the “historically
speaking” caveat qualification is an important one. Piaget remains widely discussed
in introductory psychology courses, but the narrative is substantially different from
that presented in this collection. In the modern discipline of psychology, Piaget is
simultaneously (a) revered for helping usher in the cognitive revolution which intro-
duced human reasoning back into an equation whence the behaviorists had banished
it long ago and (b) marginalized somewhat as a giant who didn’t quite get it right.
Piaget is typically presented to Introductory Developmental Psychology students
as something like that old stage theorist. Moreover, his constructivist account of
learning and cognition is often characterized as one that is not adequately concerned
with language or social interaction (Beilin, 1992). Perhaps, most problematic from
our perspective is that Piaget’s position on thinking and learning is often presented
as painting a picture of cognitive development that is limited by biology in a manner
parallel to that of Arnold Gesell. In fact, perhaps, the lion’s share of what most
graduate psychology students read in class about Piaget remains dedicated to listing
all of the ways in which he may have underestimated children’s abilities. Thus, a
very large number of psychological testing paradigms (and associated illustrious
careers) were born of a cottage industry dedicated to attempting to prove what
Piaget got wrong while paying scant attention to what he got right. Several psy-
chologists have penned compelling defenses of Piaget, detailing the ways in which
this industry was unfair to him (e.g., Beilin, 1992; Flavell, 1996; Lourenço &
Machado, 1996). These accounts have typically proceeded by highlighting the
extent to which critics change the goalposts, illustrating young children’s compe-
tence for tasks that are very-near-but-not-quite-the-same as those Piaget employed.
However, what is rarely discussed in psychological circles, even among his apol-
ogists, is the depth of his thinking about mathematics and how these thoughts were
integrally tied up in the development of his genetic epistemology. We believe that
the fruits of Piaget’s genetic epistemology (GE), as typified in these chapters, have
the potential to be wildly impactful for psychological approaches to studying math-
ematical cognition. As respectful tourists, we would like to add to the broader
appreciation for Piagetian analysis in our native land of psychology and hopefully
to show how reinterpreting and interrogating the GE approach through the gaze of
modern psychology might lead to a productive synthesis that can make a substantial
contribution to research on children’s mathematical thinking.
In what follows, we will first strive to make a clear case that there is much that
the GE approach to mathematical cognition stands to offer for modern psychologi-
cal approaches. Second, we will speculate about how a few specific insights from
15 Reflections on the Power of Genetic Epistemology by the Modern Cognitive… 513
cognitive psychology might be used to update the GE analytic in ways that might
make GE more powerful – and that subsequently may pay dividends back to psy-
chology. Third, we will share one big “what if” question that lurked in the back-
ground for us as we read these chapters – one that we submit has the potential to
help drive a productive evolution of the GE project. Fourth and finally, we will offer
some closing thoughts regarding paths to expand the broader impacts of the GE
project in the field of psychology.
hat a GE Approach to Mathematical Cognition Offers

W
for Psychology
A Paean to Construct Validity
In our reading, perhaps the most striking aspect of the GE approach is the extent to
which it is deeply attentive to the way mathematical thinking unfolds over time. To
be sure, psychology offers its share of longitudinal analyses, but there are precious
few psychological studies that follow individuals closely using fine-grained mea-
sures with substantial temporal resolution for any real length of time. There are
certainly some notable exceptions to this rule in modern psychology, such as the
microdevelopmental or microgenetic approaches that began to gain more traction in
developmental psychology in the early aughts (e.g., Granott & Parziale, 2002).
These theorists argued that ostensibly discontinuous jumps in competence could
often be resolved into much more gradual steps if development was followed with
repeated, dense, within-subject measures on short timescales. In the realm of math-
ematics, the approach was perhaps best exemplified by a strand of Bob Siegler’s
work that charted changes in children’s strategy use (e.g., Siegler, 1995; Siegler &
Crowley, 1991; see also Chi et al., 1989; Van der Ven et al., 2012). Still, in the realm
of mathematics, this work often focused primarily on the evolution of strategy
choice and lacked the rich dives into children’s subjective reasoning experiences
that characterize the clinical interviews pioneered by Piaget and adopted and
adapted by modern researchers in the GE tradition (but see Mix, 2009 for a notable
counterexample).
There are compelling and obvious reasons for psychologists to use such methods
sparingly – or even to avoid them altogether. One’s choice of method inevitably
involves tradeoffs, and deciding to employ clinical interviews is tantamount to mak-
ing two decisions that are somewhat difficult to square with the mode of inquiry
valued in modern psychology departments. First, clinical interviews cost a lot in
terms of interviewer and participant time, and these resource demands almost nec-
essarily require dramatic reductions to potential sample size. This in turn compro-
mises statistical power, making it difficult to detect effects even when they exist.
Moreover, small samples complicate the researcher’s ability to secure a sample that
represents the broader population of interest, which threatens the generalizability of
any findings that do emerge. Viewed all together, the consequences of choosing to
employ clinical interviews therefore present a fundamental threat to the statistical
inference that is central to the quantitative bent of modern psychology. Second,
clinical interviews are far from neat and tidy. Their very nature requires a high
degree of sophistication on the part of any given interviewer, and the hermeneutic
concerns with such qualitative approaches are fundamentally more complicated
than those involving simpler measures such as percent correct answers or reaction
time. This complexity regarding coding and interpretation poses validity and reli-
ability problems that can make even the stoutest of psychologists shudder – even
those who majored in anthropology! These costs noted, we believe that the costs of
GE in terms of sample size and interpretive complexity are offset by what it offers
in terms of rich models of children’s thinking and in terms of task analyses.
Somewhat counterintuitively, we view the benefits that accrue to the use of quali-
tative clinical interviews partially in terms of their potential impacts on improving
quantitative assessments, chiefly via improving construct validity. Consider our
appreciation for the concept of second-order models. The concept is one that has
much to offer psychology, not because the epistemology behind it is so foreign to
psychologists but because it offers a mirror for psychologists to reflect upon the
extent to which we practice the principles of good measurement that we know we
should. The idea that an observer can never simply look inside a learner’s head and
directly observe that learner’s understanding is not the least bit controversial to
cognitive psychologists. One of the first things budding cognitive psychologists
learn in our introductory measurement and statistics courses is that most of our
constructs of interest are not directly observable. Rather, we learn that they are
latent constructs and that we must conceive of a theoretically based way of choosing
a set of manifest observables that we can interpret as indicators of the latent con-
struct of interest. This interpretive turn is at the heart of modern conceptions of
validity (APA, AERA, NCME, 2014) that view validity less as a quality of a given
test or measure and view validity more as tied up in the way one interprets a given
test or measure. The map from this fundamentally interpretive turn that psycholo-
gists learn early on in our training to the second-order model is relatively
straightforward.
One might reasonably ask, how could the concept of the second-order model
possibly make a substantial contribution to psychology if the line from introductory
measurement is indeed so straightforward? Our answer is simple: it is one thing to
pay lip service to a set of principles and another thing entirely to impose account-
ability and to honor those principles. We identify quite strongly as psychologists,
and we routinely use measurements such as accuracy on problem-solving tasks or
reaction times for speeded comparison tasks in our work (e.g., Matthews & Chesney,
2015; Park et al., 2021; Viegut & Matthews, in press) – measures that are not typi-
cally the province of GE researchers. Moreover, we do so without shame and have
some well-elaborated theories about what those tasks measure and a large store of
information regarding how reliably they do so. Accordingly, on the one hand, we
take the measurement bent of our field as a point of pride. On the other hand, how-
ever, we believe that psychological approaches to studying mathematical cognition
broadly suffer from a bit of a construct validity problem.
Take for example, psychological approaches to studying something like algebra.
Algebraic competence is frequently treated as an important outcome in psychology
(e.g., Barbieri & Booth, 2016; Booth et al., 2014; Byrd et al., 2015; Hornburg et al.,
2022; Matthews & Fuchs, 2020; Powell et al., 2019; Siegler et al., 2012), but the
nature of algebraic thinking is arguably relegated to the background. That is, our
field seldom begins by problematizing algebra as a construct. We suspect this is in
part because our motivations often enough are not math first; one experience that
makes this perfectly clear is alternating between attending mathematical cognition
presentations at math education research conferences and at psychology confer-
ences. As tourists at math education research conferences, we get the impression
that those around us hold math to be nigh autotelic – an end in itself. By contrast, at
psychology conferences, there is often a much more utilitarian ethos in the air. We
trumpet the importance of algebra in terms of what it can do for us, reporting time
and again what algebra completion predicts, be it high school graduation (Gamoran
& Mare, 1989), college admission and graduation (Adelman, 2006), or STEM
degree attainment (Chen, 2009). Similarly, we might frame our studies of mathe-
matical equivalence in terms of procedural or conceptual knowledge or the impact
of self-explanation prompts (e.g., Matthews & Rittle-Johnson, 2009), arguably rel-
egating the mathematics itself to second place.
As a result, for much of psychology, math learning has been conceived of,
whether implicitly or explicitly, as success on mathematics tests or as success on
some subset of items culled from a collection of tests. To those who might think this
an unfair account, consider this: the first author, upon reading an article by a giant
in our field, an article that has since been cited hundreds of times, contacted the
author to ask how the items that tested the key mathematical construct were gener-
ated. After checking around with their team, the author replied that items were
culled from preexisting standardized tests that tested general mathematical content
based on whether the items seemed like they measured the construct of interest. In
this classic example of putting the cart before the horse, the construct was defined
by the available measures, reifying a measure without following a thorough inter-
rogation of construct validity.
This question of validity is critical for the study of mathematical cognition – is
there a solid basis for interpreting a given measure as indicating what we think it indi-
cates? Considering a few simple examples from our own work on fractions can help
make it clear that this question of validity is not at all simple. We frequently use an
item asking students to circle the larger fraction from the pair 4/9 and 2/3. Our group
has previously taken this item to be one that indexes children’s conceptual knowledge
of fractions, as many others in our field have (e.g., Bailey et al., 2012, 2015;
Fuchs et al., 2013; Gunderson et al., 2019; Hallett et al., 2010; Hecht & Vagi, 2010;
Siegler & Pyke, 2013). The problem, however, is that the item can be approached in a
number of different ways. A nonexhaustive list includes:2
• One could go with a conceptually oriented measurement model of fractions and
think about the distance from 1 to 4/9, focusing on its complement 5/9. Similarly,
one could see that the complement of 2/3 is 1/3, meaning that 2/3, with a smaller
complement, is closer to 1 and is therefore larger.
• One could procedurally multiply 2/3 by 3/3 to yield a common denominator,
transforming the item to a comparison between 4/9 and 6/9. One could then pick
the fraction with the larger numerator. This would represent a much more proce-
dural approach.
• One could reflect on the fact that 2/3 is larger than ½ whereas 4/9 is smaller than
one-half and decide that 2/3 must therefore be larger by logical necessity.
• One could subtract the numerator from the denominator of each and pick the
fraction with the smallest difference. This heuristic will actually yield a correct
result fairly frequently.
Our point is that a forced-choice test wherein students choose a response without
explaining their thinking cannot discriminate between these five widely divergent
modes of thinking about fractions nor among a larger set of alternative methods that
would lead to an accurate answer (e.g., cross multiplication, dumb luck, or explic-
itly reasoning about how three-ninths are nested in each third to facilitate compari-
son in ninths). Without an adequate task analysis to populate the space of possible
moves a student can make and without adequate opportunities to observe and inter-
act with students to observe their reasoning, forced-choice items are fundamentally
limited. When confronted with the explicit task of constructing a second-order
model of the participants’ thinking, it is painfully clear that we cannot build one
from forced choices due to a lack of information, and this poses a threat to validity.
In contrast, GE researchers’ detailed task analyses combined with intimate observa-
tions of the clinical interviews have yielded useful constructs that stand to add sub-
stantial leverage to psychological research. We describe two such constructs below.
E’s Potential Contribution to Psychology, Construct #1:

G
Fraction Schemes
One such construct is fraction schemes, an indicator of the sophistication of fraction

knowledge that makes repeated appearances throughout this volume (e.g., Chaps. 5,
7, 11, and 12). There is a long and voluminous tradition of investigating fraction
knowledge in mathematics education research, and fractions are also currently
among the hottest topics in psychological research on mathematical cognition.
2
For a much more detailed list of possible approaches to the task, see Table 2 from Fazio
et al. (2016).
However, these two branches of research on fraction knowledge have remained rela-
tively isolated from each other, typically employing divergent research designs and
measures for investigating children’s developing knowledge of fractions. As a
result, the construct of fraction schemes remains mostly absent from psychological
studies.
In recent years, psychological studies have come to privilege learners’ under-
standings of fraction magnitude as a chief measure of fraction concepts. In this
work, fraction magnitude knowledge is typically assessed either by forced-choice
comparisons between fractions pairs as described above or by number line estima-
tion tasks, in which students estimate the position of fractions on unmarked number
lines from 0 to 1 (or, less commonly, from 0–2, 0–3, or 0–5). The dominance of
magnitude in psychological studies of fraction cognition is largely due to the work
of Siegler and colleagues, who both (a) elaborated an influential theory that numeri-
cal development is at its core a process of understanding that numbers possess mag-
nitudes that can be represented as positions on a number line (e.g., Siegler et al.,
2011; Siegler & Lortie-Forgues, 2014) and (b) collected an impressive body of
empirical data demonstrating the measure’s strong predictive validity. Even after
controlling for other math and cognitive skills, children’s accuracy on fraction num-
ber line estimation and comparison tasks predicts concurrent and later fraction
arithmetic skill (e.g., Bailey et al., 2017; Siegler et al., 2011), overall math achieve-
ment (e.g., Bailey et al., 2012; Hansen et al., 2017a, b; Torbeyns et al., 2015), and
algebra knowledge (Booth et al., 2014; Siegler et al., 2012; but see Barbieri et al.,
2021). These consistent associations suggest that fraction number line estimation
and comparison capture something important about children’s fraction
understanding.
Separately, math education researchers have spent decades documenting the
complex, multifaceted nature of fraction knowledge (e.g., Behr et al., 1983;
Carpenter et al., 2012; Kieren, 1976; Lamon, 1996; Ohlsson, 1988; Steffe & Olive,
2010). For example, many in the GE tradition have carefully traced the learning
trajectories of children’s fraction schemes as well as their tight connection to chil-
dren’s unit coordination abilities (e.g., Boyce & Norton, 2016; Hackenberg, 2007;
Hackenberg & Tillema, 2009; Norton & Hackenberg, 2010; Steffe, 2002, 2003).
However, the constructs of fraction schemes and unit coordination have never to our
knowledge been assessed in psychological research on fraction cognition. Their
absence from psychological studies investigating the connections between fractions
and algebra knowledge is a conspicuous gap, given that GE researchers have offered
compelling accounts of how students’ fraction schemes are closely related to their
algebraic reasoning (Hackenberg, 2013; Hackenberg & Lee, 2015; Nabors, 2003;
Thompson & Saldanha, 2003).
We recently sought to address this gap by examining the relations between mag-
nitude knowledge and fraction schemes as well as by comparing the extent to which
each was independently related to algebraic competence (Viegut et al., 2023; OSF
project). Our study was the first, to our knowledge, to measure students’ fraction
knowledge using measures of both fraction magnitude and fraction schemes with
the same students. We summarize the design and results of this study here, because
we believe they provide a convincing empirical example of what is to be gained

from combining the strengths of GE and quantitative psychological research.
We investigated relations among different aspects of fraction knowledge and
algebra knowledge in an online study of 59 eighth-grade students recruited from the
American Midwest through social media posts and online flyers. In three one-hour
sessions on Zoom, participants completed various measures of fraction knowledge
and algebraic competence, along with multiple domain-general cognitive measures.
Fractions measures included those typically privileged by psychological researchers
(i.e., fraction number line estimation, fraction comparison, and fraction arithmetic)
along with a written assessment of children’s fraction schemes adapted from Norton
and Wilkins (2012). The algebra measure included 25 items assessing conceptual
knowledge, procedural knowledge, and flexibility in algebra (Star & Rittle-Johnson,
2008) and five items designed by experts in early algebraic thinking to measure
students’ proportional reasoning and functional thinking (Blanton et al., 2015;
Kaput & West, 1994; Rivera & Becker, 2011; Stephens et al., 2021). We also mea-
sured unit coordination, whole number magnitude knowledge and arithmetic flu-
ency, non-symbolic visuospatial ratio processing, non-verbal abstract reasoning,
and working memory. Then, using regression analyses we investigated (1) which
aspects of fraction knowledge were most closely related to fraction schemes and (2)
which aspects of fraction knowledge uniquely predicted overall algebra knowledge.
After accounting for whole number knowledge and general cognitive skills, we
found that fraction schemes were associated with fraction arithmetic and number
line estimation, but not with fraction comparison. This was striking given that many
studies in psychological research consider fraction number line estimation and com-
parison as tapping the same construct – fraction magnitude knowledge (Barbieri
et al., 2021; Hamdan & Gunderson, 2017; Siegler, 2016). Even more strikingly, we
found that when we included all fraction measures and unit coordination in the same
models, only unit coordination and fraction schemes emerged as significant predic-
tors of algebra knowledge. That is, once we included the constructs of fraction
schemes and unit coordination, the measures of fraction magnitude that dominate
psychological research failed to explain significant additional variance in algebra
knowledge. Our results suggest that the fraction schemes-algebra relation is stron-
ger than the relation of either fraction magnitude knowledge or fraction arithmetic
skill with algebra knowledge.
Ours is but a single study and has yet to pass peer review, much less to be repli-
cated, yet it offers tantalizing prospects. First, our results suggest that introducing a
construct new to psychological studies of mathematical cognition may explain con-
siderably more variation in an important outcome of popular interest. Second, our
results open the door to a more complete articulation of the fraction magnitude
construct and its mechanisms of action. The fact that variance explained by fraction
magnitude knowledge is more powerfully explained by fraction schemes and unit
coordination invites us to reexamine existing theory about the nature of fraction
magnitude knowledge and to consider whether it might be reframed as a subset or a
result of fraction schemes. Finally, in addition to offering striking evidence that
psychological researchers should more deeply engage with fraction schemes and
unit coordination research, we should consider that our results may be but the tip of
the iceberg for a potentially generative project. Perhaps, there are many more mea-
sures from mathematics education research that, when employed alongside psycho-
logical methods and measures, can explain more outcomes of interest and enrich
theories. Perhaps, populating a larger space of parallel measures across the disci-
plines can allow us to sift, winnow, and combine via methods like factor analysis to
develop ever more powerful measures and theories – both about mathematical cog-
nition and about the nature of cognition more generally.
E’s Potential Contribution to Psychology, Construct #2:

G
The Figurative/Operative Distinction
A second potentially useful construct for psychology is the figurative/operative dis-

tinction. Specifically, the figurative/operative analytic may stand to contribute to the
way psychologists who study educational domains theorize about the differences
between procedural and conceptual knowledge. To be sure, the procedural vs. con-
ceptual knowledge distinction looms large in psychological studies of mathematics
thinking and learning. Indeed, both the first and second authors are the academic
progeny of a long line of researchers whose work are standard bearers in psychol-
ogy for conceptualizing and measuring procedural vs. conceptual knowledge of
mathematics (e.g., Crooks & Alibali, 2014; Rittle-Johnson & Alibali, 1999; Rittle-
Johnson et al., 2001, 2015; Siegler, 1988).
As important and illuminating as this work has proven to be, there are at least
two deep unresolved tensions that remain regarding procedural vs. conceptual
knowledge. The first tension is that the procedural vs. conceptual knowledge dis-
tinction is very different from the procedural vs. declarative knowledge dichotomy
that is prevalent throughout a much broader swath of cognitive psychology (e.g.,
Anderson, 1996; Beilock & Carr, 2001; Logan, 1985; Ullman, 2004). Throughout
much of cognitive psychology, the construct of knowledge is divided into two kinds,
based largely on the extent to which the knowledge is accessible to top-down pro-
cesses. In this system, declarative knowledge can be roughly described as con-
sciously accessible knowledge that a knower can describe or convey to others.
Examples include knowledge of one’s social security number, any recipes one may
have memorized, and the vast majority of the things learned in school, such as the
ability to recite the Pythagorean theorem, the ability to use the Pythagorean to solve
a given problem, or the ability to prove the theorem. The reader can see from this
description that both conceptual and procedural knowledge as typically conceived
in circles concerned with mathematics education typically qualify as declarative
knowledge.
By contrast, procedural knowledge for many cognitive psychologists is of a dif-
ferent kind; it is a knowledge that is read directly from short-term memory and is
frequently thought of as automated. Moreover, both the learning of the knowledge
and the knowledge itself are generally not available to conscious access (e.g.,
Ullman, 2004). Examples include the ability to touch one’s ear or to look to the left.
Most people can do it, but no one can really tell you how they do it. Other examples
include motor skills such as walking or dribbling a basketball, and research on these
examples have underscored the difference in accessibility for the two types of
knowledge. It is simple enough for most typically developing adults to walk, but
actively thinking about walking while trying do so rapidly results in a clumsy failure
to walk properly. As for dribbling a basketball or a soccer ball, it seems that when
initially learning a skill, some portion of the learning process involves declarative
knowledge, as thinking through step-by-step instructions on how to complete the
task is helpful. Yet once learned, the action sequence is read directly from long-term
memory, and thinking too hard about it has the same result as thinking about walk-
ing – it rapidly degrades competence (Beilock et al., 2002).
Behavioral theories such as Anderson et al.’s (1997) ACT-R very clearly lay out
the distinction between the procedural and declarative knowledge, with conscious
access as one major distinguishing factor. Moreover, a vast body of cognitive and
neuropsychological research has attributed the two to different memory structures
in the brain (e.g., Fogel et al., 2007; Ullman, 2020). In sum, the procedural vs.
declarative knowledge dichotomy is based on robust theory and supported by a
rigorous body of empirical results. On this theory of knowledge, the procedural vs.
conceptual distinction that is current among researchers in education is not a dis-
tinction of kinds. Instead, the procedural vs. conceptual taxonomy is so difficult to
articulate that on our reading, it at times seems to be a somewhat arbitrary system
that really just breaks declarative knowledge out into different bins.
Herein lies the second tension: the procedural vs. conceptual distinction is argu-
ably not much of a distinction at all. Hiebert and Lefevre’s (1986) distinction is
among the most widely cited in studies of mathematical cognition, defining concep-
tual knowledge as something that “can be thought of as a connected web of knowl-
edge, a network in which the linking relationships are as prominent as the discrete
pieces of information” (pp. 3–4) and defining procedural knowledge alternatively as
“familiarity with the individual symbols of the system and with the syntactic con-
ventions for acceptable configurations of symbols” or as knowledge of rules or pro-
cedures for solving mathematical problems” (p. 7). The distinction has intuitive
appeal in its many different iterations from author to author, yet there enough vague-
ness baked in that Rittle-Johnson et al.’s (2001) qualifying statement is often
invoked:
These two types of knowledge lie on a continuum and cannot always be separated; however,
the two ends of the continuum represent distinct types of knowledge. (p. 346)
Perhaps, the first thing to note here is that the claim about realness of the distinction
at the extremes remains an assertion without a deep theoretical argument. In fact,
the literature features several examples whereby researchers disagree on how to
conceptualize procedural and conceptual knowledge. For instance, Art Baroody
et al. (2007) and Jon Star (2005) sparred about the distinction. Star argued that
knowledge type was often conflated with quality and that there is such a thing as
deep procedural knowledge “that is associated with comprehension, flexibility, and

critical judgment and that is distinct from (but possibly related to) knowledge of
concepts” (p. 408). Baroody and colleagues countered that the distinction need not
refer to connectedness or depth and that the real key is in tracking the extent to
which a particular learner’s procedural and conceptual knowledge are connected to
each other. Table 1 of Baroody et al. presents a summary of various attempts to
define procedural and conceptual knowledge that remains interesting today, more
than 15 years later.
We suggest that none of these accounts offer much in the way of clarity, and a
quick example can help provide important clues as to why. In psychological studies
of children’s understanding of equivalence, children’s ability to accept or evaluate
nonstandard forms has long been considered an index of their conceptual knowl-
edge (e.g., Behr et al., 1980; McNeil et al., 2015; Rittle-Johnson & Alibali, 1999;
see also Baroody & Ginsburg, 1983; Carpenter et al., 2003). For instance, Matthews
and Rittle-Johnson (2009) used items asking children whether expressions like
8 = 2 + 6 or 3 + 2 = 7–2 “make sense” to measure conceptual knowledge. The prob-
lem, however, is that one really efficient way to answer the question is simply to use
procedures to evaluate the expressions on each side of the equals sign. This raises
an important question: might this “conceptual” knowledge item instead be getting
at procedures? If so, then our measures of the two may be confounded, in which
case the frequently cited iterative relations between procedural and conceptual
knowledge (Rittle-Johnson et al., 2001, 2015) may spring less from bidirectional
effects of two different kinds of knowledge and may in part be the result of a jangle
fallacy: calling two ostensibly different constructs by separate names when they are
not different in kind.
Of course, to conduct a rigorous treatment of this issue requires that conceptual
knowledge in mathematics has a functional definition in the first place. Unfortunately,
it may not. In a wide-ranging review of the construct that cites liberally from psy-
chology, math education research, and the learning sciences more generally, Crooks
and Alibali (2014) found that out of 265 sources that purported to be about concep-
tual knowledge, only 74 (35%) even offered an explicit definition. Among the
sources that did offer explicit definitions, there was wide variation, both in the con-
ceptualization of the construct and in the tasks used to measure it. Considering this
state of affairs, it may be time to question the usefulness of the procedural vs. con-
ceptual knowledge divide in mathematics cognition. The figurative/operative dis-
tinction, featured in multiple chapters of this volume, (most prominently in Chaps.
4 and 5), may provide an analytic that can help clarify or supplement the constructs
that the procedural vs. conceptual distinction attempts to index.
To illustrate, we return to an example we considered earlier, asking participants
to judge which is larger between 4/9 and 2/3. This is typically taken to be a measure
of conceptual knowledge in psychological research on fractions, but as described
above, many routes to answering the question are explicitly algorithmic in nature
(e.g., finding a common denominator, converting to decimals, or choosing the frac-
tion with the smallest difference between the numerator and the denominator). If we
simply evaluate correctness, it is unclear whether the person completing the task is
using an explicitly algorithmic (i.e., procedural) route or if they are doing some-
thing more conceptual. To complicate matters further, even when participants
manipulate symbols, it is difficult to know whether their manipulations are focused
directly on the symbols at a purely syntactic level or if the same symbolic manipula-
tions reflect a focus that looks through the symbols to the referents behind them
(Kaput et al., 2008; Sfard, 1991; Chap. 5). There is little in the procedural vs. con-
ceptual distinction current in psychological approaches that helps clarify the type of
thought behind symbolic manipulations or how it might be measured. By way of
contrast, the figurative/operative distinction puts this issue front and center.
We were particularly struck by Thompson’s (1985) account, as described in
Moore et al.’s chapter. On this conception, notions of figurative and operative
thought seek to account for the developmental nature of mathematical meanings and
the fact that a system of meanings at one developmental level can become the source
material (i.e., the figurative ground) for a system of meanings at a subsequent devel-
opmental level (Chapter 4, p. 16).
To put the distinction plainly, Thompson writes that
the relationship between figurative and operative thought is one of figure to ground. Any set
of schemata can be characterized as figurative or operative, depending upon whether one is
portraying it as background for its controlling schemata or as foreground for the schemata
that it controls (p. 195).
On our reading, this figurative/operative conception stands to add to the psycholo-

gist’s toolbox, because it offers some important advantages over the procedural vs.
conceptual distinction. To briefly describe a few:
1. It points less to an item as a static measure and more to differences in a problem-
solver’s perspectives. On this account, it is not the item itself, such as 4/9 vs. 2/3,
which decides whether procedural or conceptual knowledge is involved but
rather a thinker’s approach to that item. This can help orient psychological the-
ory and measurement by explicitly introducing the solver’s perspective as a vari-
able to be considered and measured.
2. The figurative/operative conception never aspires to constitute the same differ-
ence in kind that the procedural vs. conceptual dichotomy does. Instead, it antici-
pates that what is taken to be figurative in early learning is likely to be more
operative with development. At the same time, it allows that a given thinker
might flexibly approach a particular item (or number or graph, etc.) more figura-
tively or more operatively depending on context. Rejecting this difference in
item kind seems important from a validity perspective.
3. Explicitly introducing the figure/ground gaze consideration to theory regarding
test construction opens the door to leveraging a large existing body of psycho-
logical research on figure-ground perception in psychology (e.g., Lamme, 1995;
Vecera et al., 2002; Wagemans et al., 2012). On its own, this has the potential to
enrich the theory behind psychological research on mathematical cognition.
This is not to say that the figurative/operative conception is wholly unproblem-
atic. Its complexity introduces a raft of measurement issues. Most immediately, it
first suggests that the procedural/conceptual taxonomies used to classify items in

many psychological studies of mathematical cognition almost certainly miss the
mark by failing to measure the participant’s perspective. Moreover, it is one thing to
suggest that measuring a participant’s figurative/operative perspective is important
to do, but it is another thing entirely to specify how to do it. Thus, GE offers up the
construct but does not provide psychologists with methods for operationalizing the
construct reliably or for measuring it at scale given limited resources and time.
These measurement issues are daunting indeed and limit the practical usefulness for
psychological approaches.
However, as proud psychologists, we assert that we should fear not these mea-
surement issues! Our field has proven remarkably agile and creative at wringing
ever more leverage out of simple measures like accuracy and response time. For
instance, drift diffusion models have allowed us to take simple forced choice com-
parisons and to estimate differences in the speed at which individuals accumulate
information (drift rate), differences in hesitancy to make decisions (decision bound-
ary), and various biases (boundary distance metrics) (e.g., Binzak & Hubbard, 2020;
Ratcliff et al., 2012, 2016). Moreover, combining methods like drift diffusion with
creative experimental paradigms and computational models allows us to model ever
more complexity with some simple measures. All of this is before considering the
ways that self-reports of figurative/operative focus might potentially be yoked to
neuroimaging, eye-gaze analysis, and pupillometry to train computer classifiers that
may eventually be able to develop nonintrusive, yet valid and reliable, measures of
perspective taking for a given item in real time. We are quite bullish on what our
tools can accomplish given the proper seeds and feel that the limiting factor is often
simply in securing the seeds. The figurative/operative distinction has the potential to
be such a seed that GE has to offer that psychologists can nourish and grow, and for
that, we are grateful.
What Can Psychology Offer to the Genetic Epistemologist?
We hope that what we have written to this point has gone some way toward express-
ing the profound respect we have for the GE approach to researching mathematical
cognition. To summarize a thread that runs the above, much of our respect for GE
flows from its deep commitment to construct validity. In one of the most highly
cited psychological works of all time, Cronbach and Meehl (1955) famously
wrote that:
Scientifically speaking, to “make clear what something is” means to set forth the laws in
which it occurs. We shall refer to the interlocking system of laws which constitute a theory
as a nomological network. (p. 290)
In our estimation, the products of the GE approach can help modern cognitive psy-
chological approaches more reliably meet Cronbach and Meehl’s bar with respect
to mathematical thinking in at least two important respects: first, GE directly
attempts to deal with questions such as “what is mathematics” and “what does it
mean to think mathematically?” Second, the methods of GE, involving teaching
experiments and clinical interviews, offer dense longitudinal data and attending
theories that attempt to build the sorts of nomological networks that constitute deep
understanding of mathematical constructs. Minimally, we offer that there is very
little modern work in cognitive psychology that asks the same theoretical questions
or offers the same sorts of data. We find this to be noteworthy and think that to dis-
miss GE or to give it short shrift as incompatible with modern cognitive psychology
is to commit a colossal error. Indeed, as our empirical example integrating fraction
schemes and unit coordination with measures currently privileged by psychologists
demonstrated, some GE constructs stand up very well to scrutiny by psychological
methods.
However, just as we believe there is much that psychologists can learn from GE,
we also believe that modern cognitive psychological approaches have much to offer
genetic epistemologists. Perhaps, the most obvious sphere in which cognitive psy-
chology stands to contribute is by helping to update GE’s conception of human
thinker qua organism. The neurophysiologist Warren McCulloch (1961) once asked,
“What is a number, that a man may know it, and a man, that he may know a num-
ber?” If GE more frequently engages with the part of the question regarding the
nature of number when compared to cognitive psychology, it is much less likely to
engage with the second part of the question – at least in any operational sense.
Although Neo-Piagetians, with Robbie Case as perhaps the most notable exemplar
(e.g., Case, 1993; Case et al., 1996; Demetriou et al., 2016; Kalchman et al., 2001;
Morra et al., 2012; Norton et al., 2023), attempted to account for the human cogni-
tive architecture that might instantiate mathematical thinking, this approach has
been largely absent from the bulk of genetic epistemology. Recalling again our
intuitive impressions as tourists at math education research conferences, it seems to
us as though many GE researchers dismiss mechanistic information processing
accounts of mathematical cognition out of hand as reductive and dehumanizing. We
hope to convince some to soften that position and to consider the productive poten-
tial of a synthetic approach.
As a point of departure, Dehaene and Cohen’s (2007) neuronal recycling hypoth-
esis offers a compelling rationale for why those who would study mathematical
thinking should pay heed to the specifics of human cognitive architecture – that is,
the hardware with which we do our thinking. To summarize this hypothesis briefly,
Dehaene and Cohen make four big points:
1. The advent of human cultural achievements such as mathematical thinking
occurred far too recently to have influenced the evolution of our species.
2. “Human brain organization is subject to strong anatomical and connectional
constraints inherited from evolution” (p. 384).
3. Cultural acquisitions like mathematics must operate using a set of preexisting
circuits that are both (a) sufficiently close to the newly desired function and (b)
sufficiently plastic that they can reorient a significant fraction of their neural
resources to the new use.
4. The prior organization of these circuits is never really erased. Thus, prior neural
constraints exert a powerful influence on the culturally determined new usage –
in this case, mathematical thinking.
To these, we add a fifth point for the sake of clarity: that the same arguments apply
not only to the brain but to the human body more generally. This point extends the
applicability of the argument beyond any simple mind vs. body distinction such that
it also applies to embodied accounts of cognition (e.g., Abrahamson et al., 2020;
Nathan & Walkington, 2017; Radford, 2014; Wilson, 2002).
On this logic, there are at least two major advantages to updating GE theories to
account for the characteristic capabilities and constraints of the human cognitive
architecture. First, theories that exclude consideration of these constraints may
seem reasonable in some narratives but may prove to be flatly implausible given
what we know about how human bodies work. Second, updating theories with
information about our cognitive architecture’s characteristic capabilities and func-
tioning may help integrate additional wrinkles into theories that enrich their nomo-
logical networks. This can both make individual theories stronger and enhance the
potential for bridging between theories that currently operate in separate silos.
Given the demonstrated power of GE approaches to date, we find it exciting to
speculate about the additional heights to which GE might climb when equipped
with theory updated to account for our hardware. Below, we discuss two specific
examples of how psychological perspectives might help update GE for the purposes
of illustration.
he GE Approach Would Be Much more Powerful If Updated

T
to Feature a Theory of Memory
Arguably, one of the goals of school learning is to find a path to instantiate to-be-
learned content and skills in long-term memory. This is not to be mistaken with a
position that takes math learning to be about rote memory, but memory IS incontro-
vertibly an important part of the story. After all, what good would it be to construct
a scheme if it could not be stored in memory and recalled for later use? And how
could the concept of a scheme be sensible without the ability to recall operations in
some combination? Even the briefest reflection reveals that memory necessarily
delimits important constraints on the thinker at the heart of GE approaches, yet
memory receives almost no explicit theoretical treatment. We hold that foreground-
ing and problematizing memory stands to substantially inform our conceptions of
some of the key constructs upon which GE is built.
What is long-term memory? How stable is it? What kind of fidelity can it con-
ceivably have as a recording device? Consideration of these questions is key to
problematizing central constructs such as operations and schemes. For instance, if
long-term memories are veridical and detailed, then this would have important
implications for the nature of the schemes we construct; they could be quite robust.
However, one could also imagine that extremely stable, well-defined schemes might
be more limited for the purposes of abstraction, requiring the storage of many dif-
ferent specific peculiarities and conditions to populate some specific set of instances
to which each scheme should apply. In contrast, suppose that long-term memory is
instead characterized predominantly by fuzzy, decay-laden impressions of core con-
tent that the thinker essentially fills out in different moments and that even then
remain susceptible to suggestion. This state of affairs would have very different
implications for how schemes might work. Schemes might then be much more
unstable, sometimes decaying to the point that they are inaccessible altogether. But
they might also be much more flexible and adaptive, as one fuzzy memory of an
exemplar might be suitable for helping support a wider range of compatible associa-
tions, thus boosting the human potential for abstraction. We do not pretend to be
experts on long-term memory, and this is not the space for a deep review, but we
note the empirical record leans much more heavily in favor of the fuzzier, more
unstable view of long-term memory (Brainerd & Reyna, 1990; Conway & Pleydell-
Pearce, 2000; Sekeres et al., 2018).
The subject of short-term or working memory is different and represents a transi-
tory gateway into long-term memory. Take, for example, the fact that to participate
in mathematical thinking, the thinker has to be able to hold relevant information in
memory long enough to operate on it. We typically accomplish this in two ways.
One is by off-loading the information to some sort of inscription to which we might
attend. Another is by holding the information in what cognitive psychologists call
working memory. Critically, there is a relatively constrained number of items that
humans can maintain in working memory for any length of time; hence, Miller
(1956) titled his groundbreaking piece The Magical Number Seven, Plus or Minus
Two. Miller also clarified that although the number of units in memory was finite,
those units can be composed of chunks – units that are themselves composed of
other units of information. As a result, chunking can bundle things that would need
to be held in memory separately, effectively making working memory much more
powerful. Later in the 1970s, Baddeley (1976) provided us with a more elaborate
model of working memory that featured three major components:
1. A central executive that controls the flow of information.
2. A phonological store for verbal content, including verbal mathematical content
such as the fact that the internal angles of a triangle add to 180. Notably, this
phonological store decays rapidly, so an internal auditory loop repeats important
verbal information in order to maintain the information and prevent decay from
working memory.
3. A visuospatial sketchpad that stores visuospatial data.
There have been myriad advances to Baddeley’s model and debates about whether
the magical number really is seven. Additionally, modern neuroscience aided by
computational methods offers much more elaborate mechanistic theories. However,
GE need not dig down too far into those technical weeds to benefit from an explicit
treatment of working memory.
Because of its gateway status, working memory and its management constitute
important limiting factors for learning. This consideration is at the heart of cognitive
load theory (Plass et al., 2010; Sweller, 2011), and parallel considerations may help
GE accounts a) develop more advanced measures of student abilities and b) develop
design principles that lead to more efficient instruction. One recent example that
attempts such an update is Norton et al.’s (2023) work with unit transformation
graphs (UTGs). Briefly, UTGs attempt to explicitly identify the sequences of actions
and unit coordinating structures that individuals use to complete a particular math-
ematical task. The work draws on the idea that different coordinating structures can
chunk together different elements of a more exhaustive sequence of actions for a
given task. This means two thinkers with different structures for knowledge that
essentially contain the same elements can have different UTGs, which means the
same task may have very different memory loads for different people.
Norton et al. (2023) studied a sample of preservice teachers, attempting to build
second-order models of their UTGs – essentially individualized models of the work-
ing memory load for a given task. The researchers also measured the preservice
teachers’ working memory capacity using a backward digit span task. Although
UTGs were far from perfect predictors, combining UTGs with individual working
memory capacity was broadly predictive of the ease or difficulty that PSTs encoun-
tered with different classes of problems. We hope that this work is but an opening
salvo in such efforts to update GE with explicit concerns for memory. Moreover, we
look forward to watching partnerships bloom whereby computational researchers
with noninvasive measures can help automate the construction of UTGs, making the
measure more accessible for practical use.
he GE Approach Would Be Much more Powerful If Updated

T
with a Probabilistic, Emergentist Conception of Cognition
In addition to memory decay, a second notion should also give us pause when con-
sidering the extent to which schemes are likely to be stable. Namely, it is highly
likely that the genesis of schemes is less unitary in nature than commonly conceived
and may instead emerge from a dynamic system characterized both by marked mod-
ularity and a probabilistic nature. Cognitive theorists from Elizabeth Bates and
Esther Thelen to David Rumelhart and James McClelland have pushed the bounds
of our thinking about how nonlinear dynamics between interacting components of
systems can lead to emergent aspects of cognition (Bates et al., 1998; Rumelhart
et al., 1986; Thelen, 2005). These theories allow for intensely mechanistic accounts
of thinking without being susceptible to charges of reductionism; because nonlinear
dynamics in a combinatorial space of billions of neurons are so complex and so
sensitive to minor changes, be they changes in initial conditions or minor differ-
ences in connectivity between neurons, such dynamics are not fully predictable. Yet,
the models work. Advances in these theories have led to rapid advances both in
artificial intelligence and in theories about the nature of human thought.
Again, genetic epistemologists need not become experts on the ins and outs of
neural networks or computational models to reap many benefits from emergentist
perspectives, as many implications follow, even when zooming out to a coarser level
of analysis. For instance, theories of learning like Siegler’s overlapping waves the-
ory (e.g., Siegler, 2016; Siegler & Shipley, 1995) posit that learners’ approaches to
different math problems are characterized not by deep stability but by repertoires of
approaches that coexist at different levels of activation. At first glance, this might
seem to suggest that one conception or strategy may be active in response to a given
item 80% of the time, another 15% of the time, and another 5%. However, a more
accurate reading would be to think in terms of simultaneous activations that may be
pulled to the fore in an instant. Importantly, many of these ways of approaching a
given mathematical task may not even be accessible to conscious thought; they may
not spring into consciousness until they meet a particular level of activation. These
possibilities may have very substantial implications for the ways that we think about
schemes, their emergence, their stability and accessibility, and the ways that we
assess students’ constructed schemes.
We are not sure what a GE updated to incorporate such a perspective might yield.
On the one hand, we were frequently struck by the extent to which the chapters
described schemes as relatively stable. For instance, Thompson et al. (Chap. 5)
wrote “An understanding is stable if it is the result of an assimilation to a scheme. A
scheme, being stable, then constitutes the space of implications resulting from the
person’s assimilation of anything to it” (p. xx). Even if the authors allow that
schemes to develop with time and experience, they are often treated as if they are
stable from moment to moment. On the other hand, there is clearly room in GE for
a probabilistic, emergentist perspective on schemes. Here, it is perhaps instructive
to reflect on Thompson et al.’s (Chap. 5) suggestion that “Piaget’s use of schemes is
quite utilitarian” (p. 132) or Tillema and Gatza’s (Chap. 3) reminder that schemes
are living constructs. Similarly, second-order models are explicitly designed as pro-
visional accounts based on behavior. These models are always there for the updat-
ing, and this presumably includes the situation in which behavior is reconceived on
some level as epiphenomenal and amenable to some sort of probabilistic formal
modeling. This is all to say that emergentist accounts seem to us to be compatible
with the GE project.
A Major “What If”
When we began this chapter, we promised to end with a big “what if?” that we
thought might have the potential to be transformative for the field of GE. The big
question we have in mind stems from our own ongoing interests in symbols. Part of
what motivated the first author’s initial foray into mathematical cognition was the
conviction that our symbolic abilities are what make us quintessentially human and
that mathematics is a domain that offers special cases for considering the ways
humans understand arbitrary symbols. It is perhaps fitting then that our interest was
piqued by the volume’s treatment of the construct of semiotic function – the repre-
sentational ability to use something (a signifier) to stand for something else (the
signified), often when the signified is not present. The idea of the distance imposed
between the signifier and the signified seems to play a particularly important role in
Piagetian thought. According to Tallman and O’Bryan (Chap. 8), Piaget considers
the development of the semiotic function to be ““the most decisive turning point in
the mental development of the child” because of its potential to create objects of
thought detached from concrete supports and ultimately to facilitate an individual’s
dissociation of the form of their thought from its content – an essential feature of
reflected abstraction” (pp. X).
In this view, the semiotic function allows humans to engage in truly representa-
tional thought.
Whether it is the authors’ intent or not, it is easy to read many of the arguments
in this volume as resting in part upon an assumption that signifiers (henceforth sym-
bols) are very different from sensorimotor percepts. A reader might reasonably con-
clude that the abstract nature of symbols and the concomitant psychological distance
they offer from the sensorimotor allow for learners to free their thoughts from the
particularities of context; this is the doorway to reflected abstraction and the possi-
bility of generalization. But what if symbols aren’t quite so distant from the senso-
rimotor after all? What if the symbol/percept dichotomy fails to hold up? What if
the very thing that allows symbols to be abstract is that they are permanently yoked
to sensorimotor representations?
Although the symbol/percept distinction is compatible with many approaches in
developmental psychology, there have long been strands of thinking that posit much
less daylight between the two. For example, in writing of representations and sym-
bols, Thomas Hobbes wrote in his Leviathan (1994) that:
The original of them all is that which we call sense. (For there is no conception in a man’s
mind which hath not at first, totally or by parts, been begotten upon the organs of sense.)
The rest are derived from that original. (p. 6)
Hobbes thought that there was very little to distinguish between sensation and rep-
resentations, including symbolic representations, and he was not alone. As reviewed
in Barsalou (1999), it appears that prior to the twentieth century, most theories of
knowledge were inherently perceptual in nature. Multiple modern theorists offer
current accounts that at root share a common thrust: that symbolic thinking and
perceptual thinking may not be so different after all (e.g., Casasanto, 2009;
Goldstone et al., 2010; Lupyan et al., 2010; Wilson, 2002).
To be sure, it seems eminently reasonable to expect that abstract symbolic think-
ing should be dissociated from perception at some point, especially if symbolic
thinking is to be freed from the yoke of particular images and contexts. As the first
author has written before, a vulgar view of perception might cast it as primarily
concerned with lower-order thinking as opposed to higher-order operations and
schemes (Matthews & Ziols, 2019). In this sense, perception is the province of
general animal cognition, whereas the schemes of the semiotic function appear to be
what separates homo sapiens from the others. However, the theorized distance
between perception and higher-order thought is potentially overstated. In truth, a
sizable body of empirical work suggests that perception involves some powerful
forms of abstraction that can support deep engagement in mathematical thinking.
To take the sensorimotor and abstract thinking to be different classes of thought
may be to commit a grave error. Researchers from the Gibsons (Gibson, 1950,
1969) to Goldstone and Barsalou (1998), to Gauthier et al. (2010) have demon-
strated that perception is highly selective and can be a source of abstract understand-
ing. Indeed, from a Gibsonian perspective, perception is at root about selectively
attending to relevant information and relations from among the vast amount of
information taken in by the senses (e.g., Gibson & Gibson, 1955).
As much becomes clear when considering the Rubin vase in the left panel of
Fig. 15.1, below. Instead of merely registering a series of lines and shades, human
perception attempts to make sense of the mass of information presented, yielding a
face or a vase. As for the right panel, most will rapidly perceive an image of a
Dalmatian, as opposed to seeing some random mass of black shapes. These exam-
ples illustrate that the perceptual system is remarkably adept at sifting through mas-
sive amounts of sensory inputs and focusing on features.
The power of this perceptual system to guide complex thinking should not be
underestimated. Writing back in 2013, Kellman and Massey pointed out that the
best human grandmaster could still beat the world’s best chess-playing computer,
despite the fact that the computer could examine up to 200 million possible moves
per second. To the question of how humans, who at best might tend to examine four
moves per turn, could manage to compete, Kellman and Massey offered that:
Fig. 15.1 Left: A Rubin vase, alternately depicting faces or a vase, depending on what is taken as
a figure or as ground. Right: Emergent depiction of a Dalmatian sniffing tree from Sadil et al. (2019)
Whatever the human is doing, it is, at its best, roughly equivalent to 2 billion moves per
second of raw search. It would not be overstating to describe such abilities as “magical.”
(p. 118)
Those magical abilities emerge from perceptual learning or experience-induced

changes in how we perceive and acquire information (Goldstone et al., 2010;
Kellman et al., 2010). Here, it seems instructive to invoke Moravec’s paradox:
Encoded in the large, highly evolved sensory and motor portions of the human brain is a
billion years of experience about the nature of the world and how to survive in it. The delib-
erate process we call reasoning is, I believe, the thinnest veneer of human thought, effective
only because it is supported by this much older and much more powerful, though usually
unconscious, sensorimotor knowledge. (Moravec, 1988, p. 15)
We have previously used these reflections as a point of departure for arguing that
those interested in optimizing teaching should consider how we might leverage per-
ceptual learning to give a powerful boost to mathematical cognition (Matthews &
Ellis, 2018; Matthews & Ziols, 2019). However, even these arguments still allow for
a strong distinction between symbols and perception. What if symbols themselves
at root similarly rely on sensorimotor knowledge? What if the semiotic function
operates on perceptual symbols?
At first blush, the concept of a perceptual symbol may seem oxymoronic.
However, revisiting Dehaene and Cohen’s (2007) neuronal recycling hypothesis can
help illuminate the plausibility of perceptual symbols. The crux of the neuronal
recycling account is that evolutionarily ancient brains come to operate in new ways
by co-opting old circuits to perform new functions. In considering the emergence of
symbolic thinking, one option is to posit the relatively sudden emergence, sui
generis, of an independent capacity to process this novel thing – the symbol. Another
option is to posit a relatively gentle shift in neural functional connectivity that
repackaged preexisting perceptual capacities that already represented different phe-
nomena neurally (e.g., a dog and verbally emitted sound patterns /d/ /ɔ/ /g/) and
complexed those capacities to form a symbol-referent relationship (i.e., united the
visual representation of a dog to the sound pattern that matches the word “dog”).
Goldstone and Barsalou (1998) offer a compelling and accessible perceptual sys-
tems account arguing that the second option is the more parsimonious one.
At the risk of oversimplification, below we briefly adumbrate three key tenets of
Goldstone and Barsalou’s (1998) perceptual symbol systems account:
1. “Perceptual representations are not necessarily conscious images but are often
unconscious states of perceptual systems specified neurally” (p. 235). For exam-
ple, instead of defining a representation of a chair in terms of a conscious mental
image, one could define the representation of a chair as the pattern of neural
activations stimulated when the image of a chair falls on the retina. We might
similarly define a memory of a chair as the reactivation of this pattern of neural
activations. Goldstone and Barsalou further argue that these perceptual represen-
tations can be abstract:
For example, selective attention might focus on the form of an object, storing only its shape
in memory and not its color, texture, position, size and so forth. This schematic extraction
process not only operates on sensory states, it also operates on internal mental events,
extracting aspects of representational states, cognitive operations, motivational states and
emotions. Once these schematic perceptual representations become established in memory,
they can function as symbols. (p. 235)
We note that the meaning of “schematic” in this case is more in line with “outline”
or “framework” than it is with the GE sense that invokes complex operations.
2. “Mechanisms used to represent information in perception perform double duty,
also representing information in concepts” (p. 232). Just as recalling a memory
of a chair involves reactivating portions of the same neural patterns that were
active when directly perceiving the chair, the perceptual symbol systems account
suggests that the word “chair” and associated concepts also activate portions of
the same perceptually based neural pattern. To demonstrate that this can also
extend to more abstract concepts, Goldstone and Barsalou point out that:
In problem-solving, one of the most effective ways of deriving an abstract schema, such as
‘overcoming an object by converging weak intensity forces from several pathways onto the
object’ is to use several concrete examples. (p. 244; see
Fig. 15.2)
Duncker’s radiation problem describes a situation in which using a direct radia-
tion blast powerful enough to destroy a cancerous tumor would also do irreparable
damage to the patient’s surrounding tissue. One solution is to provide multiple
smaller blasts from different directions. For example, Gick and Holyoak (1983) suc-
cessfully used the diagram above to help participants develop an abstract conver-
gence solution that could be applied to various analogous problems such as
Duncker’s radiation problem.
3. Because symbols and related conceptual structures develop from perceptual
processes, they continue to bear vestiges of their perceptual legacy. Two
brief examples can help bring this point home. First, when asked to indicate the
larger of two Hindu-Arabic digits, such as comparing 1–9 or 8–9, both children
Fig. 15.2 Diagram used in

Duncker’s radiation
problem
and adults exhibit distance effects, whereby they are faster and more accurate at
making these symbolic comparisons as the distance between the compared digits
increases (Moyer & Landauer, 1967; Sekuler & Mierkiewicz, 1977). That is,
participants are faster and more accurate when comparing 1–9 than when com-
paring 8–9. There is no theoretical reason to expect such distance effects if digits
were truly arbitrary symbols, wholly divorced from perception. However, Moyer
and Landauer (1967) noted that the observed pattern of results “strongly suggest
that the process used in judgements of differences in magnitude between numer-
als is the same as, or analogous to, the process involved in judgements of inequal-
ity for physical continua.” (p. 1520)
All told, in such a system, aspects of a given perceptual state that outline core
aspects of its essence or structure can be stored in long-term memory. Such repre-
sentations can then function as symbols. Moreover, they can combine via combina-
torial and recursive mechanisms, yielding ever more productive possibilities.
As a result of these characteristics, perceptual symbols are both modal and ana-
logical. They are modal in that they are represented in the same systems as the
perceptual states from whence they arise. This is especially powerful, because our
perceptual system allows simple features to be registered in parallel, effectively
without a capacity limit. Moreover, to the extent that symbols are rendered modally,
it is simple to see how perceptual operations can act upon these symbols, without
the need to posit the creation of some new higher-order capacity to operate. Finally,
because they are modal, there is a clear path to leverage the power of analogy to map
modal structure to other analogous modal structures.
What might an update to GE that integrates perceptual symbol systems look
like? Here, we remind you that we are but tourists in this realm and that we lack the
deep familiarity with GE required to point toward definitive solutions. We can, how-
ever, detail some of the questions for GE theorists that arise for us when we suppose
that even the most abstract of symbols may be largely perceptual at base:
1. How might we reconceive of the base material that the semiotic function
operates on?
2. How might theory on the figurative/operative distinction evolve if even operative
thought is conceived of as riding the coattails of perception?
3. How might the distinctions between types of abstraction need to be updated?
4. Can knowing and abstraction rest on what is often perceived as a “lower level”
form of cognition? If so, is there a role for perceptual learning modules in GE?
We anticipate that some might consider our question not to be so big after all. Due
to the volume, richness, and evolving character of Piaget’s writings, there are many
ways to read his thought, and there are certainly arguments to be made that the sen-
sorimotor and abstract symbols are not so far apart in Piaget. For instance, Thompson
et al. (Chap. 5) point out that imagery is about re-represented experience and that
such experiences are never re-represented veridically, an account that is in the ball-
park with Goldstone and Barsolou’s (1998). However, we recall Tallman and
O’Bryan’s (Chap. 8) description of Dubinsky’s account:
Consistent with Piaget’s distinction between figurative and operative modes of thought,
reflective abstraction is the means by which form or process is divorced from sensorimotor
content and converted into objects of thought that the subject is then able to act upon to
form new constructions. (p. 265)
On our reading, the treatment of reflective abstraction in this volume seems much
more aligned with this conception that separates sensorimotor from abstract think-
ing and, presumably, from the symbols that support it. To the extent that this con-
ception pervades some theorists’ approach to GE, we eagerly await the answers to
the questions we posed above and the theoretical and methodological sequelae to
which those answers lead.
References
Abrahamson, D., Nathan, M. J., Williams-Pierce, C., Walkington, C., Ottmar, E. R., Soto, H., &
Alibali, M. W. (2020). The future of embodied design for mathematics teaching and learning.
In Frontiers in education (Vol. 5, p. 147). Frontiers Media SA.
Adelman, C. (2006). The toolbox revisited: Paths to degree completion from high school through
college. US Department of Education.
American Educational Research Association (AERA), American Psychological Association
(APA), National Council on Measurement in Education (NCME), & Joint Committee on
Standards for Educational and Psychological Testing (U.S.). (2014). Standards for educational
and psychological testing. AERA.
Anderson, J. R. (1996). ACT: A simple theory of complex cognition. American Psychologist,
51(4), 355.
Anderson, J. R., Matessa, M., & Lebiere, C. (1997). ACT-R: A theory of higher level cognition and
its relation to visual attention. Human–Computer Interaction, 12(4), 439–462.
Baddeley, A. D. (1976). The psychology of memory. Basic Books.
Bailey, D. H., Hansen, N., & Jordan, N. C. (2017). The codevelopment of children’s fraction
arithmetic skill and fraction magnitude understanding. Journal of Educational Psychology,
109(4), 509.
Bailey, D. H., Hoard, M. K., Nugent, L., & Geary, D. C. (2012). Competence with fractions pre-
dicts gains in mathematics achievement. Journal of Experimental Child Psychology, 113(3),
447–455.
Bailey, D. H., Zhou, X., Zhang, Y., Cui, J., Fuchs, L. S., Jordan, N. C., et al. (2015). Development
of fraction concepts and procedures in US and Chinese children. Journal of Experimental Child
Psychology, 129, 68–83.
Barbieri, C., & Booth, J. L. (2016). Support for struggling students in algebra: Contributions of
incorrect worked examples. Learning and Individual Differences, 48, 36–44.
Barbieri, C. A., Young, L. K., Newton, K. J., & Booth, J. L. (2021). Predicting middle school pro-
files of algebra performance using fraction knowledge. Child Development, 92(5), 1984–2005.
Baroody, A. J., Feil, Y., & Johnson, A. R. (2007). Research commentary: An alternative recon-
ceptualization of procedural and conceptual knowledge. Journal for Research in Mathematics
Education, 38(2), 115–131.
Baroody, A. J., & Ginsburg, H. P. (1983). The effects of instruction on children’s understanding of
the “equals” sign. The Elementary School Journal, 84(2), 199–212.
Barsalou, L. W. (1999). Perceptual symbol systems. Behavioral and Brain Sciences, 22(4),
577–660.
Bates, E., Elman, J. L., Johnson, M., Karmiloff-Smith, A., Parisi, D., & Plunkett, K. (1998).
Innateness and emergentism. In W. Bechtel & G. Graham (Eds.), A companion to cognitive
science (pp. 590–601). Basil Blackwell.
Behr, M., Erlwanger, S., & Nichols, E. (1980). How children view the equals sign. Mathematics
Teaching, 92(1), 13–15.
Behr, M., Lesh, R., Post, T., & Silver, E. (1983). Rational number concepts. In R. Lesh & M. Landau
(Eds.), Acquisition of mathematics concepts and processes (pp. 91–125). Academic Press.
Beilin, H. (1992). Piaget’s enduring contribution to developmental psychology. Developmental
Psychology, 28(2), 191.
Beilock, S. L., & Carr, T. H. (2001). On the fragility of skilled performance: What governs choking
under pressure? Journal of Experimental Psychology: General, 130(4), 701.
Beilock, S. L., Carr, T. H., MacMahon, C., & Starkes, J. L. (2002). When paying attention becomes
counterproductive: Impact of divided versus skill-focused attention on novice and experienced
performance of sensorimotor skills. Journal of Experimental Psychology: Applied, 8(1), 6.
Binzak, J. V., & Hubbard, E. M. (2020). No calculation necessary: Accessing magnitude through
decimals and fractions. Cognition, 199, 104219.
Blanton, M., Stephens, A., Knuth, E., Gardiner, A. M., Isler, I., & Kim, J. S. (2015). The develop-
ment of children’s algebraic thinking: The impact of a comprehensive early algebra interven-
tion in third grade. Journal for Research in Mathematics Education, 46(1), 39–87. https://doi.
Booth, J. L., Newton, K. J., & Twiss-Garrity, L. K. (2014). The impact of fraction magnitude
knowledge on algebra performance and learning. Journal of Experimental Child Psychology,
118, 110–118.
Boyce, S., & Norton, A. (2016). Co-construction of fractions schemes and units coordinat-
ing structures. Journal of Mathematical Behavior, 41, 10–25. https://doi.org/10.1016/j.
jmathb.2015.11.003
Brainerd, C. J., & Reyna, V. F. (1990). Gist is the grist: Fuzzy-trace theory and the new intuition-
ism. Developmental Review, 10(1), 3–47.
Byrd, C. E., McNeil, N. M., Chesney, D. L., & Matthews, P. G. (2015). A specific misconception of
the equal sign acts as a barrier to children’s learning of early algebra. Learning and Individual
Differences, 38, 61–67.
Carpenter, T. P., Fennema, E., & Romberg, T. A. (Eds.). (2012). Rational numbers: An integration
of research. Routledge.
Carpenter, T. P., Franke, M. L., & Levi, L. (2003). Thinking mathematically. Heinemann.
Casasanto, D. (2009). Embodiment of abstract concepts: Good and bad in right-and left-handers.
Journal of Experimental Psychology: General, 138(3), 351.
Case, R. (1993). Theories of learning and theories of development. Educational Psychologist,
28(3), 219–233.
Case, R., Okamoto, Y., Griffin, S., McKeough, A., Bleiker, C., Henderson, B., Stephenson, K. M.,
Siegler, R. S., & Keating, D. P. (1996). The role of central conceptual structures in the devel-
opment of children’s thought. Monographs of the Society for Research in Child Development,
61(1/2), i–295. https://doi.org/10.2307/1166077
Chen, X. (2009). Students who study science, technology, engineering, and mathematics (STEM)
in postsecondary education. In Stats in brief (NCES 2009-161). National Center for Education
Statistics.
Chi, M. T., Bassok, M., Lewis, M. W., Reimann, P., & Glaser, R. (1989). Self-explanations: How
students study and use examples in learning to solve problems. Cognitive Science, 13(2),
145–182.
Conway, M. A., & Pleydell-Pearce, C. W. (2000). The construction of autobiographical memories
in the self-memory system. Psychological Review, 107(2), 261.
Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological
Bulletin, 52(4), 281.
Crooks, N. M., & Alibali, M. W. (2014). Defining and measuring conceptual knowledge in math-
ematics. Developmental Review, 34(4), 344–377.
Dehaene, S., & Cohen, L. (2007). Cultural recycling of cortical maps. Neuron, 56(2), 384–398.
Demetriou, A., Shayer, M., & Efklides, A. (Eds.). (2016). Neo-Piagetian theories of cognitive
development: Implications and applications for education. Routledge.
Fazio, L. K., DeWolf, M., & Siegler, R. S. (2016). Strategy use and strategy choice in fraction mag-
nitude comparison. Journal of Experimental Psychology: Learning, Memory, and Cognition,
42(1), 1–16.
Flavell, J. H. (1996). Piaget’s legacy. Psychological Science, 7(4), 200–203.
Fogel, S. M., Smith, C. T., & Cote, K. A. (2007). Dissociable learning-dependent changes in
REM and non-REM sleep in declarative and procedural memory systems. Behavioural Brain
Research, 180(1), 48–61.
Fuchs, L. S., Schumacher, R. F., Long, J., Namkung, J., Hamlett, C. L., Cirino, P. T., et al. (2013).
Improving at-risk learners’ understanding of fractions. Journal of Educational Psychology,
105(3), 683.
Gamoran, A., & Mare, R. D. (1989). Secondary school tracking and educational inequality:
Compensation, reinforcement, or neutrality? American Journal of Sociology, 94(5), 1146–1183.
Gauthier, I., Tarr, M., & Bub, D. (Eds.). (2010). Perceptual expertise: Bridging brain and behav-
ior. OUP USA.
Gibson, E. J. (1969). Principles of perceptual learning and development. Appleton-Century-Crofts.
Gibson, J. J. (1950). The perception of the visual world. Houghton Mifflin.
Gibson, J. J., & Gibson, E. J. (1955). Perceptual learning: Differentiation or enrichment?
Psychological Review, 62(1), 32.
Gick, M. L., & Holyoak, K. J. (1983). Schema induction and analogical transfer. Cognitive
Psychology, 15, 1–39.
Goldstone, R. L., & Barsalou, L. W. (1998). Reuniting perception and conception. Cognition,
65(2–3), 231–262.
Goldstone, R. L., Landy, D. H., & Son, J. Y. (2010). The education of perception. Topics in
Cognitive Science, 2(2), 265–284.
Granott, N., & Parziale, J. (Eds.). (2002). Microdevelopment: Transition processes in development
and learning (Vol. 7). Cambridge University Press.
Gunderson, E. A., Hamdan, N., Hildebrand, L., & Bartek, V. (2019). Number line unidimension-
ality is a critical feature for promoting fraction magnitude concepts. Journal of Experimental
Child Psychology, 187, 104657.
org/10.1016/j.jmathb.2007.03.002
first multiplicative concept. Journal of Mathematical Behavior, 32(3), 538–563. https://doi.
org/10.1016/j.jmathb.2013.06.007
Hackenberg, A. J., & Lee, M. Y. (2015). Relationships between students’ fractional knowledge and
equation writing. Journal for Research in Mathematics Education, 46(2), 196–243.
Hackenberg, A. J., & Tillema, E. S. (2009). Students’ whole number multiplicative concepts: A
critical constructive resource for fraction composition schemes. The Journal of Mathematical
Behavior, 28(1), 1–18.
Hallett, D., Nunes, T., & Bryant, P. (2010). Individual differences in conceptual and procedural
knowledge when learning fractions. Journal of Educational Psychology, 102(2), 395.
Hamdan, N., & Gunderson, E. A. (2017). The number line is a critical spatial-numerical represen-
tation: Evidence from a fraction intervention. Developmental Psychology, 53(3), 587.
Hansen, N., Jordan, N. C., & Rodrigues, J. (2017a). Identifying learning difficulties with frac-
tions: A longitudinal study of student growth from third through sixth grade. Contemporary
Educational Psychology, 50, 45–59.
Hansen, N., Rinne, L., Jordan, N. C., Ye, A., Resnick, I., & Rodrigues, J. (2017b). Co-development
of fraction magnitude knowledge and mathematics achievement from fourth through sixth
grade. Learning and Individual Differences, 60, 18–32.
Hecht, S. A., & Vagi, K. J. (2010). Sources of group and individual differences in emerging frac-
tion skills. Journal of Educational Psychology, 102(4), 843.
Hiebert, J., & Lefevre, P. (1986). Conceptual and procedural knowledge in mathematics: An intro-
ductory analysis. In J. Hiebert (Ed.), Conceptual and procedural knowledge: The case of math-
ematics (pp. 1–27). Erlbaum.
Hobbes, T. (1994). Leviathan: With selected variants from the Latin edition of 1668. Hackett
Publishing.
Hornburg, C. B., Devlin, B. L., & McNeil, N. M. (2022). Earlier understanding of mathematical
equivalence in elementary school predicts greater algebra readiness in middle school. Journal
of Educational Psychology, 114(3), 540.
Kalchman, M., Moss, J., & Case, R. (2001). Psychological models for the development of math-
ematical understanding: Rational numbers and functions. In S. M. Carver & D. Klahr (Eds.),
Cognition and instruction: Twenty-five years of progress (pp. 1–38). Lawrence Erlbaum
Associates.
Kaput, J., Blanton, M., & Moreno-Armella, L. (2008). Algebra from a symbolization point of
view. In J. J. Kaput, D. Carraher, & M. Blanton (Eds.), Algebra in the early grades (pp. 19–56).
Lawrence Erlbaum Associates.
Kaput, J. J., & West, M. M. (1994). Missing-value proportional reasoning problems: Factors
affecting informal reasoning patterns. In G. Harel & J. Confrey (Eds.), The development of
multiplicative reasoning in the learning of mathematics (pp. 235–287). State University of
New York Press.
Kellman, P. J., & Massey, C. M. (2013). Perceptual learning, cognition, and expertise. In Psychology
of learning and motivation (Vol. 58, pp. 117–165). Academic Press.
Kellman, P. J., Massey, C. M., & Son, J. Y. (2010). Perceptual learning modules in mathematics:
Enhancing students’ pattern recognition, structure extraction, and fluency. Topics in Cognitive
Science, 2(2), 285–305.
Kieren, T. E. (1976). On the mathematical, cognitive and instructional. In Number and measure-
ment. Papers from a research workshop (Vol. 7418491, p. 101).
Lamme, V. A. (1995). The neurophysiology of figure-ground segregation in primary visual cortex.
Journal of Neuroscience, 15(2), 1605–1615.
Lamon, S. J. (1996). The development of unitizing: Its role in children’s partitioning strategies.
Logan, G. D. (1985). Skill and automaticity: Relations, implications, and future directions.
Canadian Journal of Psychology/Revue canadienne de psychologie, 39(2), 367.
Lourenço, O., & Machado, A. (1996). In defense of Piaget’s theory: A reply to 10 common criti-
cisms. Psychological Review, 103(1), 143.
Lupyan, G., Thompson-Schill, S. L., & Swingley, D. (2010). Conceptual penetration of visual
processing. Psychological Science, 21(5), 682–691.
Matthews, P. G., & Chesney, D. L. (2015). Fractions as percepts? Exploring cross-format distance
effects for fractional magnitudes. Cognitive Psychology, 78, 28–56.
Matthews, P. G., & Ellis, A. B. (2018). Natural alternatives to natural number: The case of ratio.
Journal of Numerical Cognition, 4(1), 19.
Matthews, P. G., & Fuchs, L. S. (2020). Keys to the gate? Equal sign knowledge at second grade
predicts fourth-grade algebra competence. Child Development, 91(1), e14–e28.
Matthews, P. G., & Rittle-Johnson, B. (2009). In pursuit of knowledge: Comparing self-
explanations, concepts, and procedures as pedagogical tools. Journal of Experimental Child
Psychology, 104(1), 1–21.
Matthews, P. G., & Ziols, R. (2019). What’s perception got to do with it? Re-framing founda-
tions for rational number concepts. In A. Norton & M. W. Alibali (Eds.), Constructing number
McCulloch, W. S. (1961). What is a number, that a man may know it, and a man, that he may know
a number. General Semantics Bulletin, 26(27), 7–18.
McNeil, N. M., Fyfe, E. R., & Dunwiddie, A. E. (2015). Arithmetic practice can be modified
to promote understanding of mathematical equivalence. Journal of Educational Psychology,
107(2), 423.
Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity
for processing information. Psychological Review, 63(2), 81.
Mix, K. S. (2009). How Spencer made number: First uses of the number words. Journal of
Experimental Child Psychology, 102(4), 427–444.
Moravec, H. (1988). Mind children: The future of robot and human intelligence. Harvard
University Press.
Morra, S., Gobbo, C., Marini, Z., & Sheese, R. (2012). Cognitive development: Neo-Piagetian
perspectives. Psychology Press.
Moyer, R. S., & Landauer, T. K. (1967). Time required for judgements of numerical inequality.
Nature, 215(5109), 1519–1520.
Nabors, W. K. (2003). From fractions to proportional reasoning: A cognitive schemes of opera-
tion approach. Journal of Mathematical Behavior, 22(2), 133–179. https://doi.org/10.1016/
S0732-3123(03)00018-X
Nathan, M. J., & Walkington, C. (2017). Grounded and embodied mathematical cognition:
Promoting mathematical insight and proof using action and language. Cognitive Research:
Principles and Implications, 2, 1–20.
Norton, A., & Hackenberg, A. J. (2010). Continuing research on students’ fraction schemes. In
L. P. Steffe & J. Olive (Eds.), Children’s fractional knowledge (pp. 341–352). Springer. https://
doi.org/10.1007/978-1-4419-0591-8
Norton, A., Ulrich, C., & Kerrigan, S. (2023). Unit transformation graphs: Modeling students’
mathematics in meeting the cognitive demands of fractions multiplication tasks. Journal for
Ohlsson, S. (1988). Mathematical meaning and applicational meaning in the semantics of frac-
tions and related concepts. In Number concepts and operations in the middle grades (Vol. 2,
pp. 53–92).
Park, Y., Viegut, A. A., & Matthews, P. G. (2021). More than the sum of its parts: Exploring the
development of ratio magnitude versus simple magnitude perception. Developmental Science,
24(3), e13043.
Plass, J. L., Moreno, R., & Brünken, R. (Eds.). (2010). Cognitive load theory. Cambridge
University Press.
Powell, S. R., Gilbert, J. K., & Fuchs, L. S. (2019). Variables influencing algebra performance:
Understanding rational numbers is essential. Learning and Individual Differences, 74, 101758.
Radford, L. (2014). Towards an embodied, cultural, and material conception of mathematics cog-
nition. ZDM, 46, 349–361.
Ratcliff, R., Love, J., Thompson, C. A., & Opfer, J. E. (2012). Children are not like older adults: A
diffusion model analysis of developmental changes in speeded responses. Child Development,
83(1), 367–363.
Ratcliff, R., Smith, P. L., Brown, S. D., & McKoon, G. (2016). Diffusion decision model: Current
issues and history. Trends in Cognitive Sciences, 20(4), 260–281.
Rittle-Johnson, B., & Alibali, M. W. (1999). Conceptual and procedural knowledge of mathemat-
ics: Does one lead to the other? Journal of Educational Psychology, 91(1), 175.
Rittle-Johnson, B., Schneider, M., & Star, J. R. (2015). Not a one-way street: Bidirectional rela-
tions between procedural and conceptual knowledge of mathematics. Educational Psychology
Review, 27, 587–597.
Rittle-Johnson, B., Siegler, R. S., & Alibali, M. W. (2001). Developing conceptual understanding
and procedural skill in mathematics: An iterative process. Journal of Educational Psychology,
93(2), 346.
Rivera, F. D., & Becker, J. R. (2011). Formation of pattern generalization involving linear figural
patterns among middle school students: Results of a three-year study. In Early algebraization
Rumelhart, D. E., McClelland, J. L., & PDP Research Group, C (Eds.). (1986). Parallel distrib-
uted processing: Explorations in the microstructure of cognition, Vol. 1: Foundations. The
MIT Press.
Sadil, P., Potter, K. W., Huber, D. E., & Cowell, R. A. (2019). Connecting the dots without top-
down knowledge: Evidence for rapidly-learned low-level associations that are independent of
object identity. Journal of Experimental Psychology: General, 148(6), 1058.
Sekeres, M. J., Winocur, G., & Moscovitch, M. (2018). The hippocampus and related neocortical
structures in memory transformation. Neuroscience Letters, 680, 39–53.
Sekuler, R., & Mierkiewicz, D. (1977). Children’s judgments of numerical inequality. Child
Development, 48(2), 630–633. https://doi.org/10.2307/1128664
Sfard, A. (1991). On the dual nature of mathematical conceptions: Reflections on processes and
objects as different sides of the same coin. Educational Studies in Mathematics, 22(1), 1–36.
Siegler, R. S. (1995). How does change occur: A microgenetic study of number conservation.
Cognitive Psychology, 28(3), 225–273.
Siegler, R. S. (1988). Strategy choice procedures and the development of multiplication skill.
Journal of Experimental Psychology: General, 117(3), 258.
Siegler, R. S. (2016). Continuity and change in the field of cognitive development and in the per-
spectives of one cognitive developmentalist. Child Development Perspectives, 10(2), 128–133.
Siegler, R. S., & Crowley, K. (1991). The microgenetic method: A direct means for studying cogni-
tive development. American Psychologist, 46(6), 606–620. https://doi-org.udel.idm.oclc.org/1
0.1037/0003-066X.46.6.606
Siegler, R. S., Duncan, G. J., Davis-Kean, P. E., Duckworth, K., Claessens, A., Engel, M., et al.
(2012). Early predictors of high school mathematics achievement. Psychological Science,
23(7), 691–697.
Siegler, R. S., & Lortie-Forgues, H. (2014). An integrative theory of numerical development. Child
Development Perspectives, 8(3), 144–150.
Siegler, R. S., & Shipley, C. (1995). Variation, selection, and cognitive change. In T. Simon &
G. Halford (Eds.), Developing cognitive competence: New approaches to process modeling.
Erlbaum.
Siegler, R. S., Thompson, C. A., & Schneider, M. (2011). An integrated theory of whole number
and fractions development. Cognitive Psychology, 62(4), 273–296.
Siegler, R. S., & Pyke, A. A. (2013). Developmental and individual differences in understanding
of fractions. Developmental Psychology, 49(10), 1994.
Star, J. R. (2005). Reconceptualizing procedural knowledge. Journal for Research in Mathematics
Education, 36(5), 404–411.
Star, J. R., & Rittle-Johnson, B. (2008). Flexibility in problem solving: The case of equation solv-
ing. Learning and Instruction, 18(6), 565–579.
Steffe, L. P. (2002). A new hypothesis concerning children’s fractional knowledge. Journal of
Mathematical Behavior, 20, 267–307. https://doi.org/10.1016/S0732-3123(02)00075-5
Steffe, L. P. (2003). Fractional commensurate, composition, and adding schemes: Learning trajec-
tories of Jason and Laura: Grade 5. The Journal of Mathematical Behavior, 22(3), 237–295.
Stephens, A., Stroud, R., Strachota, S., Stylianou, D., Blanton, M., Knuth, E., & Gardiner,
A. (2021). What early algebra knowledge persists 1 year after an elementary grades interven-
tion? Journal for Research in Mathematics Education, 52(3), 332–348.
Sweller, J. (2011). Cognitive load theory. In Psychology of learning and motivation (Vol. 55,
pp. 37–76). Academic.
Thelen, E. (2005). Dynamic systems theory and the complexity of change. Psychoanalytic
Dialogues, 15(2), 255–283.
Torbeyns, J., Schneider, M., Xin, Z., & Siegler, R. S. (2015). Bridging the gap: Fraction under-
standing is central to mathematics achievement in students from three different continents.
Learning and Instruction, 37, 5–13.
Ullman, M. T. (2004). Contributions of memory circuits to language: The declarative/procedural
model. Cognition, 92(1–2), 231–270.
Ullman, M. T. (2020). The declarative/procedural model: A neurobiologically motivated theory
of first and second language 1. In Theories in second language acquisition (pp. 128–161).
Routledge.
Van der Ven, S. H., Boom, J., Kroesbergen, E. H., & Leseman, P. P. (2012). Microgenetic pat-
terns of children’s multiplication learning: Confirming the overlapping waves model by latent
growth modeling. Journal of Experimental Child Psychology, 113(1), 1–19.
Vecera, S. P., Vogel, E. K., & Woodman, G. F. (2002). Lower region: A new cue for figure-ground
assignment. Journal of Experimental Psychology: General, 131(2), 194.
Viegut, A. A., & Matthews, P. G. (in press). Building fraction magnitude knowledge with number
lines: Partitioning versus analogy. Developmental Psychology.
Viegut, A. A., Stephens, A., & Matthews, P. G. (2023). Exploring connections between fractions
knowledge and algebra knowledge. Retrieved from osf.io/h89vf
Wagemans, J., Elder, J. H., Kubovy, M., Palmer, S. E., Peterson, M. A., Singh, M., & Von der
Heydt, R. (2012). A century of Gestalt psychology in visual perception: I. Perceptual grouping
and figure–ground organization. Psychological Bulletin, 138(6), 1172.
Wilson, M. (2002). Six views of embodied cognition. Psychonomic Bulletin & Review, 9, 625–636.
Chapter 16
Skepticism and Constructivism
I initiated this book project as a way of saying “thank you.” When speaking at a
conference with one of the contributing authors, they asked me (paraphrasing),
“Why are you the one to do this?” It is a very reasonable question. I am not leading
this project because I am the most qualified to speak on matters of Piagetian theory.
Unlike me, I think my coeditors can claim incredibly deep and profound insights
into genetic epistemology and its use, which is why I recruited them. I am leading
this project because I saw the need and recognized the potential value this volume
could bring. I anticipated the value of such a book because I know how I have ben-
efitted from this body of theory and from the community of scholars represented in
the author list. In my professional career as a mathematics education researcher, I
have been deeply shaped and guided by my colleagues who work in the Piagetian
tradition. Every bit of my scholarship is infused with the sensibilities that I gain
from Piagetian constructivism. I thus wanted to say “thank you” to this group of
researchers by attempting, as best I knew how, to offer some of those learning
opportunities to a broader audience.
In this essay, I want to accomplish three goals. First, I want to describe what I see
as some of the key lessons I have learned from constructivism that guide my work.
Second, I want to connect these lessons to the theme of skepticism that runs through-
out mathematics education research in the Piagetian tradition, often under the label
of radical constructivism (Glasersfeld, 1995). I think this skepticism is a key com-
ponent in the lessons I have learned, which I think are particularly valuable. Third,
I offer a critique of the way this skepticism is often discussed by exploring the dif-
ference between (what I shall call) philosophical agnosticism or skepticism and
scientific skepticism. By drawing this distinction, I hope to endorse what I value in
P. C. Dawkins (*)
https://doi.org/10.1007/978-3-031-47386-9_16
542 P. C. Dawkins
constructivist skepticism and to argue against other forms that I see as philosophi-
cally problematic and ultimately unproductive for mathematics education research
and communication.
Key Lessons I Glean from Piagetian Constructivism
I think one of the most fundamental sensibilities we gain from Piaget’s work is the
sense that someone else’s experiential world (especially children) is qualitatively
different from my own. Piaget and his colleagues were especially good at designing
seemingly simple tasks that children would solve in manners that felt surprising as
compared to adult expectations. The tasks were often so simple in design (related to
physics and movement) that the observed responses suggested the child literally
saw something different than the adult did. The inanimate objects that children
interacted with obeyed different rules to them or seemingly no rules at all. By seeing
task performance that seems contrary to expectation, adult observers are driven to
seek an explanation and insight into the research subject’s experience and reason-
ing. This has become such a powerful feature in some forms of psychological study
that one cognitive psychologist noted how you almost cannot publish in that field if
your results do not appear somehow counterintuitive. Recognizing differences
between adult experience and child experience, though, is not enough. What Piaget
offered that was so powerful were models of (1) the experiential world of the child
and (2) how children progressively shaped that experiential world (primarily
through abstraction, see Chaps. 6 and 8). These types of models provided opportu-
nities for explanation, prediction, and insight into learning that were incredibly
fruitful for psychologists.
A Useful Example of a Piagetian Experiment
I will share one experiment of Piaget’s that has been particularly insightful to me
(Piaget & Garcia, 1989). In the experiment, children were given a ball on a string.
This was placed on a table at which the children sat. A box was placed on the table
a little way in front of them (see Fig. 16.1). Children were to spin the ball on the
string and then release it so it went into the box. To get the ball into the box, the
child had to release the string when it was perpendicular to a linear path that led the
ball into the box. Like many of Piaget’s tasks, this task connected something chil-
dren could easily be asked to do with a key bit of expert knowledge: when the string
is released, the ball will travel along a path tangent to the circular path of rotation.
What I found especially illuminating was that some children could accurately get
the ball into the box and yet subsequently claimed that they released the string when
it was pointed at the box (extended directly in front of the child). There are a few
paradigmatic insights available from this. First, the child is able to do more in
16 Skepticism and Constructivism 543
Fig. 16.1 Ball on string task layout with two relevant points of release
activity (when they spin the ball) than they can in reflective analysis of the task
(verbally describing what they did). They “know” when to release the string in
action, but the child’s mental model of the physics of the situation still tells them
that they should release it at another spot. I have found this focus on knowledge in
activity deeply fruitful in my own thinking about mathematics education research.
When I want to understand student activity, I try to focus on the tasks that students
are asked to carry out in the relevant learning context. The meaning of something
arises from what we do with it.
Second, the ball on the string finding portrays the sense in which the child is
constructing their experience from what they understand of the world rather than
from direct “recall.” I often try to explain this point to future teachers by saying,
“Your students do not remember what you said; they remember what they under-
stood.” Students construct meaning for what we say rooted in their current under-
standing. There is nothing wrong with giving a clear and thoughtful explanation, but
constructivism teaches us to remain skeptical about the efficacy of communication.
We should be especially skeptical when we are trying to help someone construct
new knowledge, which ostensibly covers most of our interactions in the classroom.
Constructivism leads us to consider instead how to understand what alternative
meaning students might have constructed for our explanations and what abstrac-
tions might be necessary for them to interpret and approach the task more like we
do. I shall return to this later on. The third point I draw from this experiment is
important enough I shall give it a separate section.
The Principle of Subjective Rationality
The ball on the string gives us a rather clear picture of the child’s model of the phys-
ics of the situation. As with most things we throw, we release them along a path that
is directed at the goal we intend them to reach. The string introduces a systematic
deviation from this common sense idea, and it is not surprising that students try to
544 P. C. Dawkins
construe the ball on the string situation within this linear path model. What Piaget’s
work especially teaches us is the need to construct a descriptive model (when pos-
sible) that helps explain: in what experiential world would that child’s answer make
sense? This need is related to the pervasive constructivist principle of subjective
rationality. Unless we have evidence that someone has simply made a mistake (i.e.,
they would quickly change their answer if invited to think on it again), we should
assume that people answer tasks in ways that make sense given their current under-
standing. This imposes upon the researcher the task of trying to describe the nature
of that understanding that helps explain how that other person behaved.
One lesson I learned by applying the principle of subjective rationality regarded
the meaning of the word or. In studying how college students might learn mathe-
matical logic, I presented students with lists of statements that (to me) had the same
logical (i.e., grammatical) form. The first two statements I presented to students
were as follows: “given any integer x, x is even or x is odd” and “15 is even or 15 is
odd.” Multiple students in my studies declared the first statement true and the sec-
ond false. To a mathematician, the second statement is an instance of the first (set-
ting x = 15), and thus, this answer seems counterintuitive. Applying the principle of
subjective rationality, I sought to understand why this answer made sense to these
college students. As I talked with students, I came to understand that some of them
rejected “15 is even or 15 is odd” because it did not convey alternatives or possibili-
ties. We generally use or to convey that different things are possible or that we are
uncertain about which alternative is (or will be) the case. This applies nicely to the
statement with the variable x. The two alternatives are for x to be even or to be odd,
but we cannot be certain until we choose an x. However, once we choose an x, we
know what is the case. Accordingly, students were giving the answer “false” not so
much to say “untrue,” but rather to say that they do not affirm the claim “15 is even
or 15 is odd.” At the very least, they would say something different. At most, it is an
inappropriate way to speak since it suggests we are unsure whether 15 is even or odd
and thus are stating an untruth.
What I love about this example is that I now see how much more natural (or
rational?) the students’ interpretation of or is than the mathematical interpretation.
Before I talked with students, I could not easily go back and hear the statement
without the influence of my mathematical training. I was thus blind to the conflict
between the (rather artificial) mathematical usage and common language. Much
like the experiential world of physical objects, the way we hear and understand
language is often not part of our conscious processing. Somehow, these statements
sounded different to students; they said different things to them than to me.
Rationality and Normativity
To portray the value and entailments of the principle of subjective rationality, I want
to compare it to some analysis set forth by the cognitive psychologist Stanovich
(1999) in the book Who is Rational? The book takes a critical stance toward ideas
like subjective rationality, instead of arguing that researchers should view certain
kinds of responses to tasks as more rational than others. This is a consideration (i.e.,
implicit debate) that is quite pertinent in mathematics education. Stanovich consid-
ers the gap between the ways people typically reason on certain types of tasks (the
common response) and the ways experts think they ought to reason (the normative
response).1 Stanovich considers different ways of interpreting the widely docu-
mented gaps between these two. The Panglossian view assumes that people are
basically rational, so the deviations from normative performance are due to simple
errors, bad normative models, or alternative ways of construing the task (i.e., the
solver is rationally solving a different, unintended task). For instance, many econo-
mists assume that people spend their money rationally, given their goals and subjec-
tive value functions. Researchers need to engineer value functions that make
seemingly irrational spending habits appear to maximize the value function or
achieve some goal. This sounds very similar to many kinds of constructivist analy-
ses in which we seek to model the child’s experiential world that makes their
responses to tasks seem rational to them. Stanovich objects to this view because it
suggests that people would not benefit from expert knowledge because they are
already acting rationally.
Another view, favored by Stanovich (1999), is the Meliorist view that claims that
people can reason in manners compatible with expert rationality and that common
responses indicate irrational behavior. One of Stanovich’s arguments for this view
is that this is more amenable to education. This view recognizes the gap between
novice and expert knowledge such that instructors have something to teach and also
asserts that learners are capable of learning. Thus, he concludes that we must recog-
nize the distinction between expert rationality and student rationality and favor the
former (and possibly label the latter “irrational”).
I appreciate this line of argument because I think it captures some key tensions
felt in mathematics education. Stanovich seems to acknowledge the nature of the
Panglossian move to recognize alternative rationalities (what I called subjective
rationality), especially in the construal of the task. This matches Thompson’s (1982)
observation of a key difference between environmentalist (nonconstructivist)
researchers and constructivist researchers. “To an environmentalist, a problem is an
objective entity that exists independently of any one problem solver. To a
1
There are a range of tasks that might be cited in this kind of research, but they tend to match
Piaget’s tasks in terms of being relatively easy to state and respond to while operationalizing a bit
of expert knowledge. They also tend to invite an initially plausible but inaccurate answer. Frederick
(2005), for instance, used the task “In a lake, there is a patch of lily pads. Every day, the patch
doubles in size. If it takes 48 days for the patch to cover the entire lake, how long would it take for
the patch to cover half of the lake?” Kahneman and Tversky (1972) introduce the task “All families
of six children in a city were surveyed. In 72 families, the exact order of births of boys and girls
was GBGBBG. What is your estimate of the number of families surveyed in which the exact order
of births was BGBBBB? In the same survey set, which, if any, of the following two sequences
would be more likely: BBBGGG or GBBGBG?” The “expert” answers are 47 days and 72 fami-
lies, and neither are more likely, respectively.
546 P. C. Dawkins
constructivist, a problem is of necessity idiosyncratic to the person solving it and is

entirely of his own making” (p. 153).
I recount this argument here because I think Piagetian constructivism gives rich
tools for formulating another view that Stanovich does not describe, one that cap-
tures what is useful in the Panglossian view while avoiding Stanovich’s critiques.
First, Piaget’s models of students almost always expressed stages in the progression
of children’s thinking, which corresponded to different “rationalities” relative to the
task. This may be compared to states of local equilibrium in a dynamical system that
can be perturbed until they move to some new local equilibrium. Thus, acknowledg-
ing the subjective rationality of a descriptive model of human reasoning does not
mean giving up learning goals for students to develop more expert ways of reason-
ing. Indeed, recognizing the rationality of both the student view and the expert view
may allow us to recognize the differences and hypothesize how we might perturb a
learner to problematize their current way of reasoning and move toward construct-
ing a more expert-like approach to the task.
Piaget’s work goes much further to describe the mechanisms by which people
learn: abstraction (see Chaps. 6 and 8). What is harder to explain in general is what
is meant by the subjective rationality of a particular way of reasoning and on what
grounds we prefer the expert way(s) of reasoning. There is an uneasy tension con-
structivist researchers must balance between trying to respect the constructive inde-
pendence of student thinking (from expert models) while maintaining learning
goals that almost necessarily correspond to expert knowledge (both knowledge of
content and knowledge of student learning) or the particular tasks that expert knowl-
edge supposedly addresses. What is most often done in mathematics education is to
explain particular (nonnormative) ways of reasoning in light of what tasks they help
the child to complete (or from which activities they arise), how prior instruction
might (unintentionally) foster them, what tasks perturb the child to sense a limita-
tion in their current understanding so as to extend it (tasks the expert way of reason-
ing can address), and how we might build upon the students’ current schemes to
support new abstractions. In addition, the Piagetian respect for the rationality of
student mathematics opens new doors for experts to reformulate mathematics cur-
ricula of knowledge in light of students’ mathematics. In my own work, my stu-
dents’ reasoning has guided me to develop a logic curriculum organized around
quantification and sets rather than truth tables and nonsense statements. When I
analyzed some of the gaps between student reasoning and the standard curriculum,
I found student ideas more coherent and useful.
Knowledge Assumes a Knower
One of the key lessons I learn from Piagetian constructivism is to consistently attri-
bute knowledge to some knower. This is especially pertinent in mathematics because
many people take mathematical facts to be indubitably true and conceptualize that
the body of mathematical knowledge exists apart from particular knowers (as
embedded in the universe, textbooks, axiom systems, or whatever). A mathemati-

cian might say that such and such is “what this concept means.” A constructivist
would ask, “To whom?” It is common in mathematical discourse to say “this theo-
rem says” or “this theorem implies,” as though the statement itself has agency and
can reason. Constructivism teaches us to recognize this as a way of hiding the
human agents who reason about the theorem and the interpretive/reasoning process
by which it “speaks.”
A very productive habit fostered among mathematics education researchers in
the Piagetian tradition is to consistently identify the knower of any particular claim.
They tend to very carefully distinguish when they are making a claim about what a
child knows or what the researcher knows. However, since their claim about what a
child knows is a piece of knowledge, they usually give a more careful formulation:
their model of the child’s knowledge is such and such. As we shall discuss later on,
this is a key point at which skepticism is usually expressed. The actual knowledge
of the child is taken to be inaccessible to the researcher; thus, we construct models
of it based on inferences from the child’s observable activity and interactions.
Before we discuss skepticism, I want to convey what I find so deeply productive
about consistently attributing knowledge to a knower. I see it linked to one of the
most important lessons I have learned in analyzing others’ reasoning: that my
knowledge of mathematics cannot play a causal role in structuring someone else’s
reasoning. Here is what I mean. When students perform computations of fractions,
for example, we may make claims about impossibility, necessity, or license, such as
“you must have common denominators” or “you cannot add that way” when some-
1 3 4
one writes . A student might defend their solution by saying, “I just did.”
3 5 8
The key point is to distinguish between a child finding a common denominator
because their teacher told them to and doing so because it follows from the meaning
of adding fractions (e.g., combining numbers of same-size pieces). My understand-
ing of adding fractions can only influence the way children operate with fractions
inasmuch as I have supported them to construct similar meanings for fractional
quantities (and encouraged them to think with such meaning as they compute).
Constructivism guides me to pursue such supports. Otherwise, the question “Why
do we need common denominators” has no better an answer than “Why does the
horse-shaped piece move in an L-shape in chess?”
Let me give another instance of this issue from my own work to portray the prin-
ciple. I do research on logic, and one of the important principles of conditional
statements is that converse claims are treated as independent (i.e., they cannot imply
one another). Consider the two converse claims:
“Given any integer x, if x is a multiple of 2 and x a multiple of 7, then x is a multi-
ple of 14.”
“Given any integer x, if x is a multiple of 14, then x is a multiple of 2 and x a mul-
tiple of 7.”
These claims are both true. What is different about what they assert? If we prove
one of them true, have we then proven the other one as well? I recall vividly the first
548 P. C. Dawkins
time that I was discussing this issue with a pair of undergraduates, and one of them
claimed that a proof of one such claim actually proved both. What would convince
this student that this was untrue? This claim runs contrary to formal logic (I should
say “my understanding of formal logic”), in which the converse claims are logically
independent and cannot imply each other. However, as a constructivist researcher, I
must recognize that my understanding cannot play a causal role in the students’
sensemaking unless I perturb them to think in a new way or I simply invoke author-
ity, which would undercut the subjective rationality of their response (and my com-
mitment to the students’ sensemaking). In other words, invoking authority may
influence my students’ task behavior and their affective experience of learning, but
it is unlikely to influence their logical sensemaking.
As I pondered that question, I looked in textbooks at the justifications given for
this. One justification is that the truth tables are different. Another was that some
true statements have false converses; thus, no conditional statement could prove its
converse. What is implicit in both of these justifications is that we need a definition
of the truth of a conditional statement that applies to all conditional statements in
mathematics. We needed to view any particular conditional as an instance of the
class of all such statements, and our rules for such statements should be uniform.
This is implicit in what it means to “do logic” to the textbook authors (and to me).
This was precisely what I understood the undergraduate student to be denying. She
did not see true conditional statements with true converses as being the same sort of
thing as true conditional statements with false converses.
I tell this story to point out that my understanding of the constraints on the task
has no causal influence on how this student reasoned. I could invoke (or claim)
authority to tell her, “We want uniform rules for how any conditional statement is
related to its converse,” but this injunction will have a different status in the stu-
dent’s understanding. What is more, I conjecture that for the student to experience
an intrinsic sense of necessity that all mathematical conditional statements should
operate in some uniform way requires her to abstract her understanding of mathe-
matical statements in a powerful way that had not yet happened. By abstracting
logical form, she might come to see all statements as instances of the same thing
(and thus exhibiting uniform relationships between proofs and claims to be proven).
This is by no means a foregone conclusion. She indeed might instead abstract that
(1) conditional situations, (2) biconditional situations, and (3) situations in which
both converses are false are all distinct types of conditional statements. She may
decide that the way proofs work in each situation may have no bearing on how they
work in the other situations.
By carefully distinguishing my model of the student’s understanding from my
own understanding of the logic situation, I am able to really feel the tension created
between the subjective rationality of the students’ construal of the task and my
expectations rooted in my own expert understanding of the logic of conditionals. I
recall Les Steffe saying (paraphrased) “The student is my constraint.” I am tasked
with considering how I might guide the student to construct the situation so they
might be perturbed to adjust their understanding. How might that construction be
rooted in their ongoing mathematical activity of reading, articulating, arguing, and
proving? I think this exemplifies the ways that mathematics researchers and instruc-
tors are pushed to learn with and from students by carefully attributing knowledge
to particular knowers (especially when we are the knower). This is closely related to
the practice of decentering (see Chap. 9).
Implicit in what I said above, mathematics educators are largely unsatisfied with
relying too heavily on instructor authority to impose constraints on students’ math-
ematical activity. Rather, we would like them to be guided by what makes sense to
them, what provides powerful ways of approaching tasks, or what avoids internal
contradictions. How do we really apply this as a principle across various arenas of
mathematics education teaching, research, and curriculum design? Indeed, one of
my favorite inheritances from Pat Thompson is the freedom to call out common
curricula as conceptually incoherent and not offering “a mathematics worth know-
ing.” I am compelled by his vision of a conceptually coherent curriculum. I must
admit, though, that I struggle to really articulate what I mean by it other than “I
know it when I see it” or “I recognize when it is missing.”2 I think that this notion of
conceptual coherence is largely under-theorized in constructivist work, though I
think Norton’s work on reversible and composable mental actions provides a rich
tool for trying to make sense of it (see Chap. 7). Indeed, I am curious what construc-
tivist mathematics educators would claim as the reason why children do or should
construct understandings in mathematics that, to whatever degree, approximate
expert ways of knowing. In other words, given that expert ways of knowing only
indirectly influence the constructions of students, why do at least a modest number
of students end up constructing understandings that are reasonably compatible with
their teachers’ understandings? Furthermore, why do we expect that our models
developed from studying one student hold any value for teaching or researching
another student? Why do our epistemic student models work at all? Given that, as I
mentioned above, constructivism tends to adopt skepticism about the efficacy of
communication (and skepticism about our ability to know students’ understanding),
I think there is insufficient theory within constructivism to really explain why and
how communication can be largely effective at least a reasonable portion of the time.
Skepticism in Modeling
Part of the great challenge in doing the work described above is to reason about the
mathematical understanding of students as distinct from our own mathematical
understandings. Naturally, the primary tools we have to model their understandings
are our own. It is thus very easy for our knowledge to flavor our models of students’
mathematics. To fight against this tendency, constructivist mathematics educators
have developed a rich tradition of skepticism toward their models of students’ math-
ematical understanding. This skepticism expresses itself in careful attention to
2
Moving beyond, “I know it when I see it” involves reflected abstraction (see Chap. 8).
550 P. C. Dawkins
student activity, in detailed attention to standards of evidence, in careful methodol-

ogy for eliciting students’ mathematical thinking, and in seeking alternative expla-
nations that are testable via subsequent tasks. This iterative model building is at the
heart of teaching experiment methodology (Steffe & Thompson, 2000). This kind of
analytical skepticism is among the most valuable things I have inherited from the
Piagetian tradition in mathematics education. For me, this skepticism is a localiza-
tion of scientific skepticism to mathematics education. Here, I am focusing on the
work of mathematics education as modeling student mathematical reasoning and
learning as it happens.3 Much science advances by forming and testing models of
whatever phenomenon is being studied. I appreciate the tools Piagetian constructiv-
ism has given me for forming and skeptically testing my own models of students’
reasoning and learning.
In Chap. 5, Thompson makes this exact analogy to talk about how “schemes” are
researcher constructs meant to model students’ ways of reasoning. Thompson notes
the interplay between the way a researcher reasons with the model and the way the
researcher must critique and challenge the model. He explains:
The prior paragraph points to a methodological aspect of scheme as a theoretical
construct. On one hand, we say schemes are organizations of a person’s mental
activity that express themselves in what an observer sees as behavior. From this
perspective, schemes reside in individuals. On another hand, we say scheme is a
theoretical construct that researchers impute to individuals to explain their behavior.
They are a researcher’s construct. This is much like stances taken by natural scien-
tists. They realize anything they say is based on models built from theory-laden
observations, but in doing their science, they act as if their models describe real-
ity—until observations force them to step back and question their assumptions and
their models. Likewise, we infer schemes from students’ behavior in response to
carefully defined probes. We impute schemes to students to form explanations of
their behavior and to design supports we think will advance their thinking. We step
back and question ourselves when our explanations become inconsistent or inade-
quate or our designed supports do not have their intended effects.
This analogy to modeling in the natural sciences is worth exploring a bit further.
First, let me note some seemingly obvious and yet essential background assump-
tions to the activity of building models, as mathematics educators do:
1. The phenomenon being studied is able to be modeled by the researcher (statisti-
cally, mathematically, qualitatively, etc.).
2. The models the researcher creates say something about the phenomenon being
studied, even if not a precise ontological account of it (i.e., the model does not
precisely correspond to the reality of the phenomenon in some strong sense).
3. Elaborating on the previous, the researcher’s model holds some value for describ-
ing, explaining, predicting, or predictably influencing the phenomenon of study.
4. The researcher’s model is communicable to other researchers.
3
Naturally, mathematics education research may focus on many other phenomena, but this is the
“bread and butter” of the kind of research done in the genetic epistemology tradition. It is also
central to much of the work I do.
If we do not hold these basic assumptions, it makes little sense why we engage
in the modeling activity in the first place and/or why we attempt to share these mod-
els within our research community. We dedicate huge numbers of hours to con-
structing these models, and we work hard to share them in writing and other forms
of communication. I find it hard to explain the rationality of researcher activity if we
do not make these assumptions.
Scientific Model Building and Ontological Correspondence
The second assumption above is quite important since different kinds of models
reflect what is being modeled in very different ways. I use the term “ontological”
because I will have more to say about this in the rest of the chapter. Ontology is the
study of what is, meaning the existence and nature of things. I recall years ago, a
philosopher friend of mine told me that I do not understand ontology because I am
so focused on epistemology. Epistemology is the study of knowledge (how we know
and what we can know). I think my friend was quite right that, as a psychologist, I
find it hard to separate the question of what something is (ostensibly apart from my
knowing) from the question of how I know or experience it. However, I have come
to better appreciate the difference, and it is crucial for this chapter.
In our science classes, we learn that many aspects of our experience of the world
are not completely accurate as we naively experience them. Take, for instance, our
experience of color. We learn that colors are not materially different as we might
perceive but rather are mere differences of wavelength among light waves. There
are wavelengths too short or too long for our eyes to perceive, and thus, other kinds
of light constantly reach us that we are simply unequipped to perceive with the
naked eye. This example portrays how science may teach us that our experiential
world, by which we “know” the world, is not ontologically accurate. We may look
to the history of science then to teach us about the skepticism toward our naïve
experience of color that led scientists to propose the wave accounts. The point I
want to draw from this example is that this skepticism toward naïve experience was
productive for advancing new (and often surprising) claims about the world. A key
point here (parallel to Piaget’s view on learning) is that the shift came through a
constructive process of developing a richer and more comprehensive model of light
and color.
We may then naturally ask, “Is the wave account of light and color correct, as in
ontologically correct?” This is raising what I shall refer to as the question of corre-
spondence: To what extent does my knowledge correspond to ontological reality?
Many scientists would rightly demure at this point. The question of correspondence
is the bridge between epistemology and ontology since it considers whether the
knowledge we construct (epistemology) reflects reality (ontology). We can now
describe a few key stances regarding the question of correspondence:
552 P. C. Dawkins
• Realism is the strong stance that our best expert knowledge largely corresponds
to ontological reality.
• Scientific skepticism is the stance that we apply skepticism toward our under-
standing to produce more robust and explanatory models of reality. The precise
extent of correspondence is not fully known, and we consistently apply skepti-
cism even to our best models, but our models say something about ontological
reality.
• Philosophical agnosticism is the stance that we cannot be certain that we know
ontological reality, and thus, we must abandon any claims of correspondence
between our knowledge and ontological reality.
• Philosophical skepticism is the stance that due to our limited epistemic resources,
we do not know (much of) anything about ontological reality.
Notice that the distinction between the third and fourth stances is whether we
merely abandon questions of correspondence to reality (and thus all ontology) or if
we commit to a denial of correspondence. Stated another way, philosophical skepti-
cism posits that we cannot know reality, while philosophical agnosticism only
claims that we cannot know that we know reality. However, for many who do not
adopt either of these stances, the term “know” implicitly connotes “knowledge of
reality.” For those in the first two categories, to make a claim of knowledge is to
make a claim about correspondence. Scholars who are philosophical agnostics or
skeptics do not cease talking about knowledge but rather reframe what we mean by
the term trying to separate this implicit link to correspondence. Philosophical agnos-
ticism and philosophical skepticism instead adopt a functional view of knowledge.
Our knowledge merely helps us navigate our experiential worlds to pursue our goals
and, in that sense, operates as adequate (inasmuch as we achieve our goals). Such
functionality of understanding is divorced from any claims of correspondence to
reality.
Let me clarify upfront that I am not going to advocate for any of these four
stances in this chapter. Rather, my intent is to examine the relationship between
these stances and radical constructivism. To be forthright, I personally hold to the
second stance but do not think I have means by which to refute philosophical agnos-
ticism or skepticism. The two questions I want to consider are (1) do many radical
constructivists hold to philosophical agnosticism or skepticism, and (2) are those
views necessary to be a radical constructivist (or someone who works in the
Piagetian tradition more broadly)?
In my experience, the answer to the first question is yes. To listen to much of the
discourse around mathematics education research in this tradition, one would infer
that radical constructivism entails abandoning claims of correspondence. Certainly,
as it pertains to student knowledge, the limitations to researcher knowledge are
consistently and strongly asserted. I am not sure whether such scholars apply these
same epistemic limitations to our knowledge of physical reality as well, but I sus-
pect they often do. From what I understand of the history, this commitment to philo-
sophical agnosticism/skepticism (or at least the discourse that sounds similar to it)
arose less from Piaget and more from Ernst von Glasersfeld (1995). He is the thinker
associated with the adjective radical as a particular flavor of constructivism, and
this term is generally used to invoke abandonment of correspondence.4 Glasersfeld
(1995) championed the notion that we should define “knowledge” apart from cor-
respondence to reality but rather define “knowledge” as functional to meet our
goals. He explained:
[Constructivism] deliberately and consequentially avoids saying anything about ontology,
let alone making any ontological commitments. It intends to be no more and no less than
one viable model for thinking about the cognitive operations and results which, collectively,
we call ‘knowledge.’ (Glasersfeld, 1985, p. 100)
One of my key goals in this chapter is to argue that the skepticism of Piagetian con-
structivism can be employed as a form of scientific skepticism, which is to say that
this body of constructivist theory does not entail a commitment to philosophical
skepticism or to philosophical agnosticism. I mean this in two ways: the theory of
radical constructivism cannot justify philosophical agnosticism or skepticism (even
if it is often asserted), and it should not be required for those working in this tradi-
tion to adopt those stances. I will develop this argument in two ways. First, I will
reflect on the goal and value of scientific skepticism and suggest that radical con-
structivists desire this value in their modeling. In other words, philosophical skepti-
cism is not really conducive to our goals as researchers. Philosophical agnosticism
also imposes difficulties in this regard; it does not add anything essential, and I am
not convinced that it is truly free of ontological commitments (specifically regard-
ing the claims of constructivism itself). Second, I will argue that there is a contradic-
tion in claiming that radical constructivism entails philosophical skepticism since
this would be making a (negative) correspondence claim, which is precisely what
the stance seeks to deny as possible. I do not think this contradiction exists for philo-
sophical agnosticism, which I think is why it is the stance many radical constructiv-
ists adopt. Finally, I will argue that many are turned off from radical constructivism
because of the ways philosophical agnosticism is communicated, often leaving the
impression that philosophical skepticism is being endorsed. By identifying these
issues of communication, I hope I can contribute to the removal of unnecessary
confusion and philosophical tension. My hope is that drawing this distinction will
help clarify the entailments of these claims and the means by which we can dis-
cuss them.
4
This is the reason I have shifted from speaking of Piagetian constructivism or genetic epistemol-
ogy to speaking of radical constructivism, by which I mean Glasersfeld’s interpretation of Piagetian
constructivism with the abandonment of correspondence brought to the foreground.
554 P. C. Dawkins
Wittgenstein’s on Certainty (1969)
One of the last things Wittgenstein wrote before he died was a series of reflections
on skepticism and certainty, which were later compiled into the book On Certainty
(1969). I happened to read Wittgenstein’s text as I was preparing this chapter, and it
struck me that he puts forth some central insights relevant to this chapter’s consid-
eration of philosophical skepticism (he uses the word doubt). Since I think his mus-
ings are much better than my own, I shall use them in this section to explore why I
think philosophical skepticism is ill-fitting to mathematics education research and
model building.
One of the keen insights Wittgenstein offers is that doubt of one claim or other
only makes sense if we consider some other things as known. He says:
114.5 If you are not certain of any fact, you cannot be certain of the meaning of your
words either. (p. 17)
115. If you tried to doubt everything you would not get as far as doubting anything.
The game of doubting itself presupposes certainty. (p. 18)
341. That is to say, the questions that we raise and our doubts depend on the fact that
some propositions are exempt from doubt, are as it were like hinges on which
those turn. (p. 44)
342. That is to say, it belongs to the logic of our scientific investigations that certain
things are in deed not doubted. (p. 44)
343. But it isn’t that the situation is like this: We just can’t investigate everything,
and for that reason we are forced to rest content with assumption. If I want the
door to turn, the hinges must stay put. (p. 44)
625. But these rules of caution only make sense if they come to an end somewhere.
A doubt without an end is not even a doubt. (p. 83)
The point is that we may doubt particular claims, but we need a broader back-
ground of knowledge to explore such doubt. Doubting everything ceases to have
meaning. What is further, the one who claims to doubt everything may often hide
from themselves how much they are taking for granted as known (such as the mean-
ing of their own discourse and the ability to communicate their doubt).
Another point is that we do not come to know initially by doubting, but rather,
this is a much later issue that arises as people consider the question of their own
knowledge:
94. But I did not get my picture of the world by satisfying myself of its correctness;
nor do I have it because I am satisfied of its correctness. No: it is the inherited
background against which I distinguished between true and false. (p. 15)
5
Note that Wittgenstein often wrote in small sections that were numbered. I maintain the numbers
but do not always quote the entire section.
More significantly, Wittgenstein explores how doubt ought to be motivated and,

in some sense, productive. This is not to say that we have access to certainty in the
face of doubt, but rather doubt needs to be justified (as does certainty):
117. Why is it not possible for me to doubt that I have never been on the moon? And
how could I try to doubt it?
First and foremost, the supposition that perhaps I have been there would strike me
as idle. Nothing would follow from it, nothing be explained by it. It would not tie
in with anything in my life.
When I say “Nothing speaks for, everything against it,” this presupposes a principle
of speaking for and against. That is, I must be able to say what would speak for
it. (p. 18).
392. What I need to shew is that a doubt is not necessary even when it is possible.
That the possibility of the language-game doesn’t depend on everything being
doubted that can be doubted. (This is connected with the role of contradiction in
mathematics.) (p. 50).
In sum, doubt/skepticism needs to be targeted, not universal, since it requires
many other background assumptions to pursue such doubt. Doubt requires some
justification for its pursuit, as it is not its own justification. As Wittgenstein noted of
idle doubt: “Nothing would follow from it, nothing be explained by it. It would not
tie in with anything in my life” (p. 18). Finally, to employ doubt about our knowl-
edge assumes we have some standards by which we imagine being able to resolve it
(even if not met).
Considering Constructivist Skepticism
As I have attempted to convey in this chapter, I am deeply committed to scientific

skepticism as a means of advancing the science of teaching and learning. Piagetian
constructivism has taught me how to do this well when trying to study another per-
son’s mathematical reasoning and learning. However, for me, the goal of scientific
skepticism is to create better models of the phenomenon being studied. Here, I use
the term “better” both in the functionalist sense of explaining and predicting experi-
ence and the correspondence sense of likely saying something about ontological
reality.6 This is to say, the doubt has a purpose and operates against a background of
many other claims taken to be true (even if not in the strongest sense of certain).
I then wonder two things: (1) what is the purpose of radical constructivist skepti-
cism and (2) what else is being taken for granted in the work of modeling students’
mathematical cognition? One way to consider radical constructivist skepticism is as
6
I take these to be deeply interrelated. The best way to predict and explain would be to approximate
reality, and we can gain some sense of certainty of correspondence the more a model allows us to
explain and predict. I maintain the fallibility of such models as discussed above but do not abandon
the question as does the philosophical agnostic.
556 P. C. Dawkins
a motivated response to many research findings. As I recounted early in the chapter,

Piaget’s studies showed how different children’s experiential worlds are from
adults’ experiential worlds. The physics we learn in school offers us models of the
world that seem greatly at odds with our naïve experience (e.g., light and color). Is
not the rational response to continue to doubt even our current knowledge of the
world? I think this argument might speak against a thoroughgoing form of realism.7
I think that scientific skepticism allows us a middle ground in which we think our
best models provide us with better and better approximations of ontological reality.
We are not certain of knowing ontological reality (which would be beyond our abil-
ity to truly justify our knowledge), but we maintain that our knowledge corresponds
in at least some ways to reality.
Nevertheless, I observe that for many radical constructivists, their skepticism is
tied to a philosophical agnostic view (as espoused by Glasersfeld), if not philo-
sophical skepticism. I think the real tension in this position is that we have no
resources for explaining why we are able to model students’ mathematical reason-
ing with any success, why we can communicate these models to one another and
find them useful, or why our models can have any predictive power to study and
teach future students. I think this is a really essential point tied to the assumptions I
gave above about engaging in scientific model building (which I repeat here for
convenience):
1. The phenomenon being studied is able to be modeled by the researcher (statisti-
cally, mathematically, qualitatively, etc.).
2. The models the researcher creates say something about the phenomenon being
studied, even if not a precise ontological account of it (i.e., the model does not
precisely correspond to the reality of the phenomenon in some strong sense).
3. Elaborating on the previous, the researcher’s model holds some value for describ-
ing, explaining, predicting, or predictably influencing the phenomenon of study.
4. The researcher’s model is communicable to other researchers.
By their behavior, I infer that radical constructivist researchers affirm these
assumptions. They are part of the background for their doubt about their knowledge
of student thinking. The fact that there are (or at least appear to be) recurrent pat-
terns in how children reason about particular mathematical topics is not only useful
for our work, but it is also essential. However, if we remain utterly ontologically
agnostic, there seem to be no resources for explaining why this should be the case.8
I think many such researchers would explain it in terms of biology and evolution
(humans’ common origins, as well as their learning through social interactions in
highly uniform learning environments, explain their common emergent patterns of
reasoning), but this would reveal that they are not really philosophically agnostic
7
My personal comfort with realism stems from my position as a theist.
8
I think another reason that radical constructivists attack correspondence is to convince those
involved in teaching that mathematics is neither monolithic nor is it given in the universe or human
psychology. I shall not explore this aspect of the argument.
but rather hold to scientific skepticism believing that their knowledge holds at least
some kind of correspondence to reality (however tenuous).
he Contradiction of Radical Constructivism

T
and Philosophical Skepticism
I sense that some radical constructivists take the theory itself to be a justification for
philosophical skepticism. If we adopt this epistemology, it assumes that our knowl-
edge is (merely) our construction of the world around us and requires no ties to
ontological reality. We cannot know one another’s experiential worlds in any strong
sense (skepticism toward communication), and we have no means of testing whether
our knowledge of the world corresponds. Thus, we are committed to agnosticism at
best (abandoning questions of correspondence) and possibly philosophical skepti-
cism (knowing that our knowledge does not correspond).
The key problem with this whole line of argument is that it assumes that our
theory of radical constructivism actually corresponds to the way people think and
learn. There is an assumption of correspondence that underlies the whole line of
reasoning, which is precisely what the argument concludes we cannot do. This is an
internal contradiction. Notice that this is only a contradiction if we adopt philo-
sophical skepticism, which asserts a lack of correspondence. Philosophical agnosti-
cism avoids this contradiction, but I think the tension still remains that we are using
our knowledge (specifically our knowledge of radical constructivism) to draw onto-
logical claims about the limitations of our knowledge. Furthermore, this view risks
taking doubt as evidence against the claim being doubted, which is a fallacy. I affirm
the claim that we cannot verify (in a strong sense) the correspondence between our
knowledge and ontological reality. This claim speaks against strong forms of cer-
tainty, but it does not itself provide evidence against correspondence either. Indeed,
the assertion that we cannot verify our knowledge in a strong sense is not incompat-
ible with being a realist; it merely means one is not epistemically justified in such
realism. Similarly, one may choose to be a philosophical skeptic (asserting that
knowledge does not correspond), but you cannot justify that stance via radical con-
structivism. It is an axiom, not a corollary, in that case. What the philosophical
agnostic cannot do is tell the realist or scientific skeptic that they know they are
wrong since this entails an ontological claim (and moves into philosophical
skepticism).
I think radical constructivists doubt our certainty of knowledge in a principled
way. Specifically, modern science tells us of many ways that our experiential world
does not correspond, so we extrapolate and infer that there are many other such
instances. The history of science suggests that even our best theories have been
“wrong” (according to modern theories) so many times before. How can we not
anticipate that we are similarly wrong now? However, the same stories can be seen
either as evidence of how scientists were wrong or how scientists were able to
558 P. C. Dawkins
improve their theories to learn. We developed the wave model of light to explain
apparent discrepancies between our naïve experience and further observations. We,
as adults, are able to observe children’s behavior and make inferences about their
experiential worlds that help explain and predict their behavior. This is not to say
that we are epistemically empowered to certainty, but we do not know that we are
epistemically helpless either.
I imagine the following interchange, to sum up the relative impasse:
Philosophical agnostic: You lack the view of an outside observer to in any way
verify whether your constructed understanding of the world corresponds to real-
ity or not, so you should abandon the question of ontological correspondence or
else you are unjustified in this belief.
Scientific skeptic: Why then do you spend so much time studying people’s cognition
if it is truly unknowable?
Philosophical agnostic: It is not that it is unknowable in every sense. I do not know
in the sense of my understanding corresponding to ontological reality, but I can
pursue knowing in terms of functional understanding useful enough to meet my
goals as a teacher and researcher.
Scientific skeptic: But by your same argument, you cannot be sure that your knowl-
edge does not correspond to ontological reality. This would require an outside
observer’s view. If you can simply redefine knowledge to match this limitation,
why do we need to emphasize our ontological limitations so much? Can we not
act in a state of mutual intersubjectivity in discussing our models even if I think
they hold some ontological correspondence and you remain dubious on that
front? Can I still engage in constructivist modeling from a stance of consistent
self-critique even if I do not adopt philosophical agnosticism?
I cannot form a final response here since I am really in the position of the scien-
tific skeptic. It is clear to me that most radical constructivists are strongly committed
to philosophical agnosticism. My unwillingness to adopt this commitment makes
my sense of belonging in this community somewhat tenuous. I think it would largely
benefit the community of researchers in this area (and those who might want to join
them) to downplay claims against correspondence, which I do not see as really
essential to the power and utility of the theory overall.
Improving Communication
As I said above, my goal is not to dissuade scholars who are committed to philo-
sophical agnosticism. My goals are more modest. I claim (1) that this commitment
is not necessary for engaging in research in this tradition and (2) that scholars in this
tradition can emphasize philosophical agnosticism more clearly and less often. I
argue the first claim because, in practice, I think scientific skeptics and philosophi-
cal agnostics can remain committed to continual efforts to challenge, test, refine,
and at times overturn our models without much difficulty.
Indeed, what are the differences between the qualitative model building of a
scientific skeptic and a philosophical agnostic? One of the main differences I see in
practice regards the frequency with which many radical constructivists assert the
limitations of our knowledge. Both in research reports and in presentations, many in
this community consistently call attention to the fact that what they present is only
their model of what their subjects are reasoning about (both taking ownership of the
knowledge and distinguishing it from the unknowable knowledge of the research
subject). I think this is productive inasmuch as it calls us to attend to the distinction
between “x is my account of what I think the child thought or perceived” and “the
child’s reasoning was of type x.” I frequently find myself frustrated when I read
research reports that make claims of the latter type. This is because failure to distin-
guish between the researcher’s model and the students’ mathematics being modeled
often leads researchers to provide less evidence linking the two.
However, I claim that there is a different kind of confusion that naturally arises
when radical constructivists make their frequent reminders about the nature of their
models. In particular, it is easy for such a scholar to sound as though they are ques-
tioning whether they know rather than whether they know that they know. For exam-
ple, a close colleague of mine wrote the following explanation about a leading
radical constructivist scholar: “[Professor X] defines these constructs in a manner
commensurate with the radical constructivist premise that individuals lack the
capacity to experience an objective reality.” This claim can be easily taken in one of
two ways: we lack the capacity to experience reality, or we lack the capacity to
know the objectivity of our experience. The first sounds more like philosophical
skepticism, while the latter like philosophical agnosticism. If we do not clarify
which we mean, then we run the risk of trying to express one and easily being inter-
preted as endorsing the other.
Indeed, radical constructivist scholars at times bring up examples from the his-
tory of science to portray how even science’s best theories may be wrong (e.g.,
Newtonian physics of gravitation).9 However, the point of such examples must be
made clear. Do we mean to assert that what we know is (probably) wrong?
Alternatively, do we mean to say that we simply cannot be sure of our knowledge –
we do not know that we know – and thus must maintain that knowledge is only
functional? As before, the first sounds much more like philosophical skepticism and
the latter like philosophical agnosticism.
I sense that the goal of such examples is often primarily to remind the hearer that
“know” is intended to carry the functional meaning rather than the correspondence
meaning. In that case, we might go back and repair the claim that Newton was
wrong (did not know physics). The claim that Newtonian physics was wrong (did
not correspond) is based on the claim that we now have better evidence for Einstein’s
9
I should note that there is an epistemic asymmetry between verifying and falsifying, similar to
Popper’s notion of falsifiability in science. Radical constructivists generally will not claim that we
can verify our knowledge of reality to be sure that it corresponds, but we can have sufficient evi-
dence to be relatively sure that some knowledge is incorrect. We cannot know that we know (some-
thing true), but we can know with greater certainty that something is false.
560 P. C. Dawkins
theories and ideas of where to find counterexamples of Newton’s theories. However,

if we are saying that knowledge only means functional for meeting our goals and
answering our questions, Newton was not wrong – his physics was functional for
objects that are medium in size and speed.
I do not make this point idly. Rather, I mean to draw two implications from it.
First, in communicating their philosophical agnosticism, radical constructivists
should be careful to draw the distinction between when are they arguing against
knowledge of reality and when are they arguing against certainty of knowledge
about reality. As I tried to portray in the two examples above, it is very easy for the
message of the example to shift from saying, “You cannot be certain of your knowl-
edge” to “Your knowledge is (probably) wrong.”
Second, once we shift to a functional meaning for knowledge, it is quite difficult
to really distinguish what a scientific skeptic means by a model and what a philo-
sophical agnostic means by knowledge. Consider talking to a biologist about their
differential equation models of the populations of antelope or a chemist about their
models of subatomic particles. If you pressed them about whether they are sure that
their models correspond to ontological reality, I think many of them would be rela-
tively noncommittal. For many scientists, the whole idea of calling something a
model is to acknowledge that it is not perfectly true but is still useful for reasoning
about some phenomenon and answering questions about it. I once heard someone
say that the first few chemistry classes in college each explain why everything
taught in the previous class was false. Each new class introduced more complex
models that, in some respects, disagreed with the previous models, but students
benefitted from learning the models in this sequence as their complexity (and coun-
terintuitiveness) increased. This aspect of scientific modeling is why I described
scientific skepticism as corresponding in some sense to ontological reality. If indeed
the scientific skeptic’s and the philosophical agnostic’s meanings of knowledge can
operate in quite compatible ways, then I argue for the following:
Scientific skeptics can take part in Piagetian constructivist research and radical constructiv-
ists do not need to so consistently reassert their philosophical agnosticism if the primary
force of the argument is merely to support a functionalist meaning of knowledge.
Summary and Conclusions
I have attempted in this chapter to convey the many ways I have benefitted from the
Piagetian tradition in my work as a mathematics educator. I consistently work to
focus on student activity, to attend to their subjective rationality, to avoid the fallacy
that my mathematical understandings have any causal force in student reasoning, to
attend to the particular knower behind any knowledge claim, and to apply consistent
skepticism to my models of others’ reasoning so that I can challenge and improve
them. As has probably become clear, my skepticism is more of a pragmatic stance
that helps me improve my models. I do not attempt to study students’ mathematics
because I think it is unknowable, but rather precisely because I think I can model it.
I do not believe my models correspond to their understanding in a strong sense, but
I have found that my models are useful for explaining student reasoning, anticipat-
ing productive teaching moves, and helping students to learn (as best I can discern
such changes in their understanding). I think these aspects of Piagetian constructiv-
ism are all compatible with either scientific skepticism or philosophical agnosti-
cism. There are some ways that philosophical agnosticism is a very pragmatic
stance since it allows us to get on with the work of teaching and learning without
delving unnecessarily into ontological quandaries on the nature of mathematics
itself. It is clear that many mathematics educators are committed to claiming that
mathematical ideas are merely human constructions and cannot have any ontologi-
cal status (e.g., Godino et al., 2005). As someone who is not committed to such a
negative claim, I appreciate the comparative openness of philosophical agnosticism,
provided it is not really philosophical skepticism.
As I have argued, I think philosophical skepticism is untenable and certainly
does not follow from radical constructivism. I think many of my radical constructiv-
ist readers may sense that my discussion of philosophical skepticism set up a straw
man since they only ever argued for philosophical agnosticism. I included it in this
chapter in part because it allowed me to hone my arguments, but I also claim that it
represents a common understanding of radical constructivism by those who are
learning this body theory for the first time. I hope that by drawing out these distinc-
tions more carefully, I can aid in clarifying the discourse around this theory and
contribute in a practical way to its use. As part of my way of saying “thank you” to
this community, who has taught me so much, I want to advise that scholars who
teach Piagetian constructivism leave open the possibility that researchers can use
this theory as an expression of scientific skepticism. Philosophical agnosticism is
not necessary for using this body of theory in research, and I think this commitment
is unhelpful for attracting and training new mathematics education researchers. I
offer the lessons I have learned from this body of theory as a functional core of what
I think the theory has to offer, which (1) does not essentially depend upon question-
ing correspondence and (2) can thus be agreed upon by scientific skeptics and philo-
sophical agnostics.
To younger scholars who are reading this text to learn this body of theory, I want
to say that there is much to be gained from Piagetian constructivism, even if you are
a scientific skeptic or a realist. I hope this chapter has helped you understand the
intent of the radical constructivist skepticism (no matter how that is philosophically
framed). Even if you find such questions challenging, I encourage you to try to learn
from them. Let them help you to challenge your own assumptions about how much
you know about someone else’s reasoning. This body of theory has so much to offer
in terms of learning about learning and reasoning about reasoning, and I hope you
will benefit as I have in my research work (and my life in society). It is my sincere
hope that this book will help many scholars understand and make productive use of
these powerful tools. As in many parts of my work, learning has been my goal: both
researcher learning and student learning. I hope I have helped realize that goal,
however, modestly.
562 P. C. Dawkins
References
Frederick, S. (2005). Cognitive reflection and decision making. Journal of Economic Perspectives,
19, 25–42.
Glasersfeld, E. (1985). Reconstructing the concept of knowledge. Archives de Psychologie,
53(204), 91–101.
Glasersfeld, E. (1995). Radical constructivism: A way of knowing and learning. Falmer Press.
Godino, J. D., Batanero, C., & Roa, R. (2005). An onto-semiotic analysis of combinatorial prob-
lems and the solving processes by university students. Educational Studies in Mathematics,
60, 3–36.
Kahneman, D., & Tversky, A. (1972). Subjective probability: A judgment of representativeness.
Cognitive Psychology, 3, 430–454.
Piaget, J., & Garcia, R. (1989). Psychogenesis and the history of science. Columbia University Press.
Stanovich, K. (1999). Who is rational? Studies of individual differences in reasoning. Lawrence
Erlbaum Associates.
and essential elements. In R. Lesh & A. E. Kelly (Eds.), Research design in mathematics and
science education (pp. 267–307). Lawrence Erlbaum Associates.
Thompson, P. W. (1982). Were lions to speak, we wouldn’t understand. Journal of Mathematical
Behavior, 3(2), 147–165.
Wittgenstein, L. (1969). On certainty. Harper.
Part IV
Using Constructs from Genetic
Epistemology to Develop Agendas
of Research
Chapter 17
Researching Special Education: Using
and Expanding Upon Genetic
Epistemology Constructs
Jessica H. Hunt
Genetic epistemology, or the study of the origins of “knowing” (Piaget, 1972; Piaget
& Duckworth, 1970), has been interpreted and used to explain learning for decades.
Yet, its application in special education intervention research is rare (Lambert &
Tan, 2016). In this chapter, I describe how I have leveraged Piaget’s seminal ideas
about knowing and learning as one pathway to access, equity, and inclusion for
students with disabilities. I begin by reviewing traditional depictions of mathemat-
ics interventions for students with disabilities. Next, I discuss how I leveraged
Piaget’s ideas to “turn around” or reframe my program of research, to center the
voices of students with disabilities in mathematics. Finally, I touch on how and why
using tenets of genetic epistemology together with other theoretical frameworks can
build new evidence bases that present students with disabilities as thinkers and
doers of mathematics. Specifically, I argue that Piaget’s ideas can be used with other
theoretical frameworks to center students’ lived experiences and reasoning within
complex systems.
Mathematics Interventions and Students with Disabilities
Mathematics interventions for students in special education are historically struc-

tured to improve student performance through the use of explicit instruction (Fuchs
et al., 2021; Gersten et al., 2008), and the majority of the research base is centered
here. For example, there are a plethora of studies that document how interventions
J. H. Hunt (*)
Department of Teacher Education and Learning Sciences, North Carolina State University,
Raleigh, NC, USA
e-mail: jhunt5@ncsu.edu
https://doi.org/10.1007/978-3-031-47386-9_17
566 J. H. Hunt
that privilege explicit instruction can bolster students’ procedural knowledge of

fractions (e.g., Butler et al., 2003; Flores et al., 2014). These researchers report
using components of the method, such as multiple sequenced representations and
teacher think aloud of procedural steps, to improve students’ performance.
Contemporary research focuses on improving students’ conceptual understanding
of fractions through the use of models, such as number lines, to depict fractions as
measures from the perspectives of the researchers (e.g., Fuchs et al., 2017; Jordan
et al., 2017). Although the focus of this work is on conceptual understanding, the
privileging of explicit instruction remains. In fact, much, if not all, of this research
recommends teaching students to model problem structures through replicating
teachers’ think-aloud problem-solving strategies, teachers’ modeled representations
of fractions using a number line or length model, and teachers’ explanations of
thinking (see Fuchs et al., 2017).
Turning Around
Throughout much of my line of research (e.g., Hunt et al., 2016a, b, 2019, 2020;
Hunt & Empson, 2015; Hunt & Silva, 2020), I have argued that privileging the sole
use of explicit instruction (Flores et al., 2014; Gersten et al., 2008) positions inter-
vention as deficit focused spaces of remediation. In addition, it may place limita-
tions on the mathematics that students with disabilities can access and construct.
For example, teacher-modeled instruction rarely considers students’ own reasoning,
places notions of responsiveness on the individual student (Hunt et al., 2016a), and
arguably blocks students from activating their own viable ways of reasoning (Silva,
2021). Positioning the teacher’s knowledge as the authority also works to normalize
some ways of reasoning yet pathologizes others (Silva, 2021): arguably, only stu-
dents who can relate to the teacher’s thinking will make sense of it. Thus, although
interventions structured by explicit instruction do show consistent, positive effects
on students’ procedural and conceptual performance with mathematics, its domi-
nance raises ethical questions regarding not only what but who we are designing
intervention research for.
The origins of this critique rest in my lived experiences, which by and large
changed the way I conceive of interventions, what and who they are structured for,
and what kinds of outcomes should be used as documentation of their “success.” As
a doctoral student in special education, I read deficit narratives of students with dis-
abilities and the mathematics they “need” as papers outlining their deficiencies built
up. When designing my dissertation intervention pedagogy, I took in this research
and used it as a reason to employ explicit, systematic instruction (see Hunt, 2014)
that fronted my modeled strategies and think aloud of my reasoning to students. It
served as a predictable structure for me to create and enact intervention lessons and
measure the success of the intervention program.
17 Researching Special Education: Using and Expanding Upon Genetic Epistemology… 567
In doing so, I found that most students in the study were able to deploy the strate-
gies that I modeled for them within the lessons and evidenced improved perfor-
mance before and after the intervention; three focus students also seemed to improve
their strategy use in varying degrees as uncovered by task-based interviews in vary-
ing degrees. However, one of the focus students who had disabilities continued to
use alternative strategies and reasoning even after the intervention was completed
(Hunt et al., 2016b). At the time, I argued that the continued use of these strategies
was an affirmation of the research base that seemed to show that students with dis-
abilities held atypical understandings of mathematics, despite intervention.
However, upon closer examination, the qualitative data across the intervention
seemed to tell a different story. In fact, these students were using strategies that were
not atypical but based on informal ideas related to partitioning. In this way, the mis-
conceptions and seemingly “atypical” strategy use were, in reality, a reflection of
the misalignment of the intervention pedagogy and tasks with the viable notions of
fraction equivalence the student did possess (Hunt, 2022). Namely, the student con-
ceived of the fraction unit informally through equal sharing interpretations. This
work taught me the pitfalls of designing interventions without the knowledge of
how students currently conceive of, or know, mathematics concepts and taught me
that I needed to better understand the process of how students with LD interpreted
fractions through equal sharing contexts in their mathematical worlds such that I
could build on these understandings through responsive instruction.
enetic Epistemology as Part of a Framework for Equity

G
and Inclusion
When working with children, there is often a tendency to use a binary to view
“knowing” as performance; children’s learning and subsequent conceptions seem to
be labeled as either right or wrong. Based on this assessment, children with sus-
tained low achievement are often times given labels of “deficient,” “not ready,” and
“unable” (Tan & Kastberg, 2017). Mathematical knowledge, then, also takes on an
altered form and is interpreted for these children as something that needs to be
poured in rather than something that already exists in the form of physical and men-
tal actions that students are already engaging in and that can be the basis for growth
(Piaget, 1972), that can grow and change, and that can be supported and extended in
the midst of instruction. In other words, the more children are placed into instruc-
tional situations that remove the responsibility for reasoning from the child and
place it onto the teacher, the more these children experience an altered means of
knowing and learning mathematics in school (Hunt et al., 2017). Arguably, such an
experience might work to further marginalize and separate these children.
From my own early experiences with students with disabilities, I began to docu-
ment alternatives to the dominant methods of instructional intervention. I draw upon
568 J. H. Hunt
Piaget’s ideas to reframe intervention as an adaptive form of instruction based on a

complex model of knowing and learning that I call Small Environments (see Hunt,
2018; Hunt & Silva, 2020). Small Environments grounds learning within children’s
adaptive activity rather than in remediation. Students with disabilities possess pow-
erful and, often, diverse ways of reasoning. Even if they engage in learning situa-
tions in unexpected ways, their knowing and reasoning cannot be conceptualized as
deficient. It must be conceptualized as their knowledge: unique, complex organisms
composed of strengths and challenges that every person utilizes to make sense of
their world (Rose & Fischer, 2009). For example, students with dyspraxia experi-
ence differences in fine motor movements that may impact how they are able to act
on objects as they solve problems (American Psychiatric Association, 2013). These
students may rely upon strengths in whole number relationships and aspects such as
visual symmetry to partition (as opposed to physical partitions of an object) to make
connections to iteration (Hunt et al., 2019).
In this way, I draw from Piaget (1972, 1980) when he argued that the ways that
any student makes sense are, by their nature, dissimilar from that of adults and can-
not be divorced from human interaction (Piaget, 1972). Instead, and as shown in my
own story, cognition is intricately intertwined with one’s environment and experi-
ences (Flavell, 1996). The idea that students’ knowledge belongs to them and is
influenced by their lived experiences requires that intervention settings center stu-
dents’ ways of reasoning, as opposed to adults. This shift in perspective supports a
move away from asking “Does the student know it?” and toward far more powerful
ones - “How does the student understand it?” and “What knowledge and expere-
inces are studnets drawing upon and bringing forward to make sense?” Furthermore,
the question, “How can I center and build upon students’ experiences?” becomes
critical.
Equity and inclusion can begin from this “centering” or, when we elevate the
diversity and power of students’ experiences and ways of reasoning, celebrating
them and positioning them as competent as opposed to reinforcing ability hierar-
chies or sustaining deficit language about who they are as mathematics thinkers and
doers (Rodríguez et al., 2022). Returning to my previous example concerning a
student with dyspraxia, one way that we can center their reasoning by getting to
know them and their mathematics, being responsive to the context and ways they
make sense, and altering instruction to provide the student access to their actions (as
opposed to expecting students to respond to a task or an environment that is inacces-
sible to them) (Hunt et al., 2019). In these ways, exploring and documenting stu-
dents’ ways of reasoning over the course of time becomes valuable. Yet,
understanding student thinking void of the complexity (e.g., education systems,
teacher perspectives and beliefs, learner identity, language, race and culture, cogni-
tion) that may influence it may come up short in promoting rich educational experi-
ence for students with disabilities in schools and improving their outcomes in
society (Rodríguez et al., 2022; Silva et al., 2023).
Expanding Theory and Building a New Evidence Base
Interventions can be structured (and researched) as spaces that listen for (and figure
out how to use) the powerful ways of reasoning and knowledge that students already
possess and understand how these ways may advance (Hunt et al., 2016a, b, 2019;
Hunt & Silva, 2020; Hunt & Tzur, 2017; Martin & Hunt, 2022) and how to use these
ways of reasoning with students. The impact on outcomes includes but is not limited
to mathematical reasoning, such as participation, agency, identity, self-efficacy, and
self-regulation. This is no small task; students’ knowing and learning belong to
them yet exist and grow within larger systems (Davis & Simmt, 2003) that, argu-
ably, cannot be explained by genetic epistemology alone. In this way, there exist
opportunities for new insights and collaborations when supported by multiple theo-
retical frameworks. In 1970, Piaget and Duckworth wrote, “A great deal of work
remains to be done in order to clarify th[e] fundamental process of intellectual cre-
ation, which is found at all the levels of cognition…. (pg. 77–78).” I argue that one
interpretation of this quote is that we connect and expand theoretical frameworks as
a foundation from which to build new evidence bases in mathematics intervention
research for students with disabilities.
For example, researchers who use radical constructivism, information process-
ing theory, sociocultural theory, and sociopolitical theory each bring rich knowl-
edge and experiences related to what it means to effectively know, learn, and teach
mathematics within complex systems, such as tiered intervention systems.
Researchers across these theoretical disciplines agree about the importance of math-
ematics learning yet have different ideas that drive how they depict instruction that
often do not inform each other. Yet, synergy and new avenues for research and prac-
tice can result when multiple perspectives, such as those from radical constructiv-
ism, cognitive psychology, special education, and critical scholars, communicate
and interact. Without a way to learn from and with each other, our disagreements
become a primary focus, and opportunities for new knowledge are stymied or lost.
Nonetheless, finding ways to work together could lead toward transformative edu-
cational experiences for students with disabilities, creating and sustaining views of
students as mathematically able and removing the “problem” from students, and
placing it, as a challenge, on the design of instructional environments, schools, and
systems.
References
American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders
(DSM-5). American Psychiatric Publishing.
Butler, F. M., Miller, S. P., Crehan, K., Babbitt, B., & Pierce, T. (2003). Fraction instruction for stu-
dents with mathematics disabilities: Comparing two teaching sequences. Learning Disabilities
Research & Practice, 18(2), 99–111.
570 J. H. Hunt
Davis, B., & Simmt, E. (2003). Understanding learning systems: Mathematics education and com-
plexity science. Journal for Research in Mathematics Education, 34(2), 137–167.
Flavell, J. H. (1996). Piaget’s legacy. Psychological Science, 7(4), 200–203.
Flores, M. M., Hinton, V. M., & Schweck, K. B. (2014). Teaching multiplication with regroup-
ing to students with learning disabilities. Learning Disabilities Research & Practice, 29(4),
171–183.
Fuchs, L. S., Malone, A. S., Schumacher, R. F., Namkung, J., & Wang, A. (2017). Fraction inter-
vention for students with mathematics difficulties: Lessons learned from five randomized con-
trolled trials. Journal of Learning Disabilities, 50(6), 631–639.
Fuchs, L. S., Newman-Gonchar, R., Schumacher, R., Dougherty, B., Bucka, N., Karp, K. S.,
Woodward, J., Clarke, B., Jordan, N. C., Gersten, R., Jayanthi, M., Keating, B., & Morgan,
S. (2021). Assisting students struggling with mathematics: Intervention in the elementary
grades (WWC 2021006). National Center for Education Evaluation and Regional Assistance
(NCEE), Institute of Education Sciences, U.S. Department of Education. Retrieved from http://
whatworks.ed.gov/
Hunt, J. H. (2014). Effects of a supplemental intervention focused on equivalency concepts for
students with varying abilities. Remedial and Special Education, 35(3), 135–144.
Hunt, J. (2022). Keynote: Reframing intervention as asset-based learning environments:
Considerations for structuring time and space. Math Learning Center Intervention Conference.
Hunt, J. H., Welch-Ptak, J. J., & Silva, J. M. (2016a). Initial understandings of fraction concepts
evidenced by students with mathematics learning disabilities and difficulties: A framework.
Learning Disability Quarterly, 39(4), 213–225.
Hunt, J. H., Westenskow, A., Silva, J., & Welch-Ptak, J. (2016b). Levels of participatory concep-
tion of fractional quantity along a purposefully sequenced series of equal sharing tasks: Stu’s
trajectory. The Journal of Mathematical Behavior, 41, 45–67.
Hunt, J., Westenskow, A., & Moyer-Packenham, P. S. (2017). Variations of reasoning in equal
sharing of children who experience low achievement in mathematics: Competence in context.
Education Sciences, 7(1), 37.
Hunt, J. H. (2018). Navigating “disability” in “intensive instruction”: Learner complexity and
small environments. International Society of the Learning Sciences, Inc.[ISLS].
Hunt, J. H., Silva, J., & Lambert, R. (2019). Empowering students with specific learning disabili-
ties: Jim’s concept of unit fraction. The Journal of Mathematical Behavior, 56, 100738.
Hunt, J. H., Martin, K., Khounmeuang, A., Silva, J., Patterson, B., & Welch-Ptak, J. (2020).
Design, development, and initial testing of asset-based intervention grounded in trajectories of
student fraction learning. Learning Disability Quarterly, 46, 0731948720963589.
Hunt, J. H., & Empson, S. (2015). Strategies and representations used in equal partitioning prob-
lems by students with mathematics learning disability. Learning Disabilities Quarterly, 38(4),
208–220.
Hunt, J., & Silva, J. (2020). Emma’s negotiation of number: Implicit intensive intervention.
Hunt, J., & Tzur, R. (2017). Where is difference? Processes of mathematical remediation through
a constructivist lens. The Journal of Mathematical Behavior, 48, 62–76.
Gersten, R., Chard, D. J., Jayanthi, M., Baker, S. K., Morphy, P., & Flojo, J. (2008). Mathematics
instruction for students with learning disabilities or difficulty learning mathematics: A synthe-
sis of the intervention research. Center on Instruction.
Jordan, N. C., Resnick, I., Rodrigues, J., Hansen, N., & Dyson, N. (2017). Delaware longitudi-
nal study of fraction learning: Implications for helping children with mathematics difficulties.
Journal of Learning Disabilities, 50(6), 621–630.
Martin, K., & Hunt, J. H. (2022). Learning trajectory based fraction intervention: Building a math-
ematics education evidence base. Investigations in Mathematics Learning, 14(3), 235–249.
Lambert, R., & Tan, P. (2016). Dis/ability and mathematics: Theorizing the research divide between
special education and mathematics. North American Chapter of the International Group for the
Psychology of Mathematics Education..
Piaget, J. (1972). The principles of genetic epistemology. Routledge & Kegan Paul.
Piaget, J. (1980). Schemes of action and language learning. In M. Piattelli-Palmarini (Ed.),
Language learning: The debate between Jean Piaget and Noam Chomsky (pp. 163–167).
Harvard University Press.
Piaget, J., & Duckworth, E. (1970). Genetic epistemology. American Behavioral Scientist, 13(3),
459–480.
Rose, L. T., & Fischer, K. W. (2009). Dynamic development: A neo-piagetian approach. In The
Cambridge companion to Piaget (p. 400). Cambridge University Press.
Rodríguez, L. A. M., Jessup, N., Myers, M., Louie, N., & Chao, T. (2022). A critical lens on
cognitively guided instruction: Perspectives from mathematics teacher educators of color.
Mathematics Teacher Educator, 10(3), 191–203.
Silva, J. M. (2021). Through an equity lens: Teaching practices for children who are bilingual with
learning dis/abilities during mathematics discussions. Insights on Learning Disabilities, 18(2),
187–209.
Silva, J. M., Hunt, J. H., & Welch-Ptak, J. (2023). From (and for) the Invisible 10%: Including
Students With Learning Disabilities in Problem-Based Instruction. Journal for Research in
Tan, P., & Kastberg, S. (2017). Calling for research collaborations and the use of dis/ability studies
in mathematics education. Journal of Urban Mathematics Education, 10(2), 25–38.
Chapter 18
Research in Subitizing to Examine Early
Number Construction
Beth L. MacDonald
In the spring of 2012, I enrolled in a psychological structures of mathematics course

at Virginia Tech. In this course, the class and I were charged to read and discuss
Piaget’s (1968a) Genetic Epistemology book. Guided by Andy Norton, this book
and these class discussions gave me enough questions to last a career examining
how young children can construct number through their subitizing activity (a quick
apprehension toward the numerosity of a small group of items; Clements, 1999). In
this chapter, I will discuss (1) my questions regarding young children’s number
construction, (2) how Piaget’s Genetic Epistemology framework has led me to focus
on young children’s units construction and subitizing activity, (3) my next steps in
early childhood mathematics education, and (4) equitable access to opportunities
for number construction.
Questions Regarding Children’s Number Construction
In 1917, Warren McCulloch’s academic advisor asked him to describe his research
interests. He replied, “What is number, that a man [sic] may know it, and a man that
he may know number?” His advisor responded, “Friend, thee will be busy as long
as thee lives” (McCulloch, 1963, p. 1).
I have found this quote to be quite true. When I first began examining young
children’s number development, I was intrigued by the idea that mathematics was
the result of an individual’s psychological interpretation of their perceptual environ-
ment. Often, we witness children count objects and talk about aspects of their
B. L. MacDonald (*)
School of Teaching and Learning, Illinois State University, Normal, IL, USA
e-mail: blmacd1@ilstu.edu
https://doi.org/10.1007/978-3-031-47386-9_18
574 B. L. MacDonald
Fig. 18.1 To conserve number, eight black counters in the top row can be understood as the same
number of gray counters in the bottom row
number development (i.e., number words, pointing) as if they had the same goal
when counting as we did. We count to determine the numerosity of a set, but young
children count with very different goals in mind. For instance, young children may
begin mimicking their parent when finishing a count to three or sing- their counting
with the words no different than the words of any other song. Children’s goals may
be to say the number words in a particular order and possibly point at items while
saying the words. These activities are rooted in perception and often have very little
to do with their construction of number.
Piaget (1941) explained that young children first construct number when they
conserve number. To conserve number, Piaget explained that children would simul-
taneously coordinate their “structure for ordering” and “algebraic thinking struc-
tures” (Piaget, 1968a, p. 3). The structure for ordering deals with relationships. For
instance, when children place items in a row and count these items, they are relying
on their structure for ordering because they may relate the growth of their number
sequence (i.e., 1, 2, 3, 4, …) to the growth of the row of items. The algebraic think-
ing structure focuses on children’s development of groups and numbers (Piaget,
1968a, p. 25). When children develop groups of items (i.e., “cups” indicates when
several items, each identified as a cup, can be grouped based on the boundness of
the set of cups), they understand the group of items as representing a bounded set.
By coordinating two grouping structures (i.e., twos and threes with regard to five),
children can understand number as a quantity (i.e., five is composed of three and
two, four and one, and five and zero) and can be constructed through their counting
activity. This simultaneous coordination allows children to have what Piaget (1941)
described as “lasting equivalence” (p. 37), meaning they are able to conserve num-
ber across a set of items, regardless of the arrangement of the items. By conserving
number, children understand that regardless of the space between perceptual items,
the items can be understood as representing the same number in each set (see
Fig. 18.1). Piaget found children were constructing number through physical activ-
ity, which could serve as material for their construction of number through mental
activity later.
Units Construction and Subitizing Activity
What intrigued me most, as a scholar and a mathematics teacher, was how this view
could be both limiting and insightful when describing children’s construction of
number. For instance, when constructing number, subitizing was not explicitly
18 Research in Subitizing to Examine Early Number Construction 575
examined by Piaget (1941), limiting some of these theoretical perspectives.

Subitizing was first named and defined by Kaufman et al. (1949) as significantly
faster than counting and drawing from significantly higher degrees of confidence in
the numerosity of the set than estimating. Often, this is witnessed by educators
when children roll a die and determine “how many” pips are on the face of the die.
Distinctions in the literature (Sarama & Clements, 2009) between perceptual and
conceptual subitizing provided broad differences between children’s description of
the “shape of the items” (perceptual subitizing) and the groups used to subitize the
entire group (conceptual subitizing). Broadly, I wondered with which thinking
structure (algebraic or ordering) subitizing aligned. Broadly, subitizing activity is
described with several different models in both the psychology literature and the
mathematics education literature (Clements et al., 2019), but for the purposes of my
work, I took up Pylyshyn’s (2001) model for subitizing describing individual’s per-
ceptual parallel preattentive behavior when subitizing. What this means is that when
subitizing, children develop a visual awareness of small sets of items at the same
time. Pylyshyn describes this awareness as individuation where children may attend
to, “that [which] refers to something we have picked out in our field of view without
reference to what category it falls under or what properties it may have” (p. 129). In
short, the items children individuate are brought into preattentive mechanisms as a
group (making this activity parallel) and considered without reference.
By considering subitizing in this way, I wondered how it might connect later with
children’s conceptual number understanding. Piaget’s (1968a) work has impacted
my work since then, as I could study the variations of subitizing as it relates to alge-
braic thinking activity and number conservation. Moreover, in addition to the order-
ing and algebraic thinking structures, Piaget (1968a) described a third thinking
structure, the topological thinking structure (i.e., proximity). I wondered if children
developed early forms of topological logic when subitizing.
Essentially, Piaget (as cited by Flavell, 1963) describes two areas of children’s
topological thinking structures, space, and geometry. Due to the nature of young
children’s number construction through subitizing, I focused on children’s concep-
tions of space. Often, people assume our understanding of space between objects is
given and not developed as concepts. However, Piaget and Inhelder (1948) found
quite the opposite, suggesting that five areas of space were developed through child-
hood: proximity, order, enclosure, separation, and continuity. These five areas were
described as fundamental in developing conceptual understandings of Euclidean
features (i.e., lines, closed figures). In my dissertation, I examined how children’s
conceptions of proximity and the separation between items informed their forma-
tion of sets when subitizing and constructing number. Essentially, I found that when
children perceived a large space between sets of items, they often subitized these
items and developed perceptual units for these items (see Fig. 18.2). For instance,
when children were shown the set of dots in Fig. 18.2, they often described this as
two, two, and one. The children either subitized subgroups and then the larger group
or the larger group and then the subgroups.
This finding more closely represented Piaget’s discussion of children’s develop-
ment of classification systems wherein they would ascend toward the larger group
576 B. L. MacDonald
Fig. 18.2 Dot image

shown to students in
MacDonald’s (2013)
dissertation study
(understanding sets of items distinctively before relating the sets together in a larger
class) or descend from the larger group (understanding, the larger class as related
before distinguishing between the sets as unique groups) (Inhelder & Piaget, 1958).
Given that their creation of groups seemed to be at the root of their number develop-
ment, I focused on their development of classification systems when subitizing and
constructing number.
When children in my dissertation study engaged in early forms of perceptual
subitizing, I found they developed similar activity wherein they subitized two small
sets of items and then composed these to determine the numerosity of the entire set
(i.e., “I saw two and three. Did you know two and three is five?”). Children also
descended from a set when subitizing. Often, this was evidenced when children
would subitize the entire set of items and then describe the smaller sets (i.e., “I saw
five. Two and three make five.”). This activity was important for early childhood
educators to recognize because it provided insight into the algebraic groups young
children construct and provided insight into children’s early segmenting activity
when creating a variety of number combinations (i.e., four is composed of four and
zero, three and one, and two and two). I also found that children’s development of
numbers in subitizing the relatively smaller groups of items had the potential to
engender children’s development of more number combinations. In other words,
when a child constructed six with three and three, they were able to use three in later
activities to construct five as two and three and four as three and one. This increased
the number of combinations for five and four, as children often used groups of two
and one more in previous settings. Given some of these critical contributions to the
literature, I still had questions as to how children’s subitizing might also be lever-
aged to support particular counting development.
enetic Epistemology as a Pathway to Early Childhood

G
Mathematics Education Scholarship
Once I completed my dissertation, individuals reached out to ask about further

implications of this work. For instance, I was hired by the Activity Integrating Math
and Science (AIMS) nonprofit group to serve as a research consultant. I worked
with scholars, students, and educators at AIMS to help them better understand how
I designed my tasks, how 4-year-old children engaged with these tasks, and how this
could inform our future task designs, professional development, and our under-
standing of children’s number construction. Through this work, several questions
became relevant. Some examples include:
• When children represent the numerosity of a small set of items, what perceptual,
figurative, motor, and verbal materials do they rely on?
• When subitizing, how might children’s representations and physical activity
relate to the same type of material used when counting?
• How do children’s subitizing vary across children’s activity in a classroom?
I left these conversations with much to do.
After discussing my next research initiatives with scholars and mentors, I also
became intrigued with children’s early development of reversibility. Piaget (1968a)
describes reversibility as an action that “can take place in one direction or in the
opposite direction” (p. 21). Reversibility is described more fully in Chap. 7 of this
book. The premise for my work began when I compared children’s activity required
for units construction through their counting versus through their subitizing.
Theoretically, I reflected on young children’s subitizing and units construction as
children may first individuate and then perceptually carry in a small set of items in
parallel to their working memory. Throughout this activity, children may develop
pre-numerical units (units developed with perceptual resources and through physi-
cal activity) (Steffe & Cobb, 1988). Pre-numerical units are similar to Hackenberg
and Sevinc’s (Chap. 11) description of a unit, but this unit is characterized by the
different types of children’s physical activity used when constructing these units.
For instance, Steffe and Cobb explain that children creating units with the percep-
tual materials provided them are said to have constructed perceptual units. Moreover,
children may also represent perceptual items with other material (e.g., finger pat-
terns to stand in for counters), described as figurative units. When children use
physical movement or verbal gestures, they are constructing motor units and verbal
units, respectively (Steffe & Cobb, 1988). In fact, when I reexamined one student’s
activity from my dissertation, I found her to conceptually subitize but not yet con-
struct and use abstract units (internalized or interiorized units, operating at least at
an INS stage) in her solutions (see Chap. 11). In other words, just because children
are constructing a conceptual unit in their subitizing activity does not mean this unit
is abstract (MacDonald & Wilkins, 2019). In fact, through this preliminary analysis,
I found that this one child often represented their subitizing activity with a figurative
unit, motor unit, or verbal unit, all particular types of pre-numerical units. However,
I was intrigued with the activity children used when they subitized compared to
when they counted. When children subitized more than one unit, they sometimes
segmented entire figures depicting dot arrangements for children to construct pre-
numerical units (MacDonald et al., 2015; MacDonald & Wilkins, 2019; MacDonald,
2015). This segmenting activity and figurative unit development may support chil-
dren’s construction of pre-numerical and abstract units beyond the perceptual items
shown to them (Wilkins et al., 2022).
578 B. L. MacDonald
Comparatively, when children count, Steffe and Cobb (1988) found they first
construct a unit with sequential activity, associating a number word with a pointing
action. This counting allows children to construct pre-numerical units in activity
through an iterative process. Given this, if children’s segmenting and iteration were
(at times) being developed through both types of activity and were reversible of
each other, would both activities better promote children’s development of an
abstract unit, instead of only one of these activities?
In other words, by simultaneously coordinating iterating and partitioning (seg-
menting), children have the potential to develop reversibility, which characterizes an
operation and can be used to develop reciprocal relationships and inverse under-
standings. By leveraging both subitizing and counting, I developed questions
such as:
• Would children, who perform relatively lower than their peers, be more capable
of developing abstract units when subitizing and counting?
• How does both subitizing and counting serve children’s ability to develop revers-
ible actions?
Given these questions, I reexamined some data from my dissertation (MacDonald
& Wilkins, 2019) to consider, when did one of my participants create units when
subitizing, and when did she create units when counting? What activity was associ-
ated with these units, and did she draw on similar activity in both? I found that this
participant did create figurative units when solving a counting task, and she seemed
to represent these figurative units with grouping activity (i.e., flashing fingers in
small groups). In fact, she created the same figurative units in her counting activity
(two and three) that she had been creating in her subitizing activity. These findings
were interesting because it seemed this participant was drawing from her subitizing
activity to construct some counting activity, and the units she represented were
organized with the most efficient finger patterns (three fingers representing “three”
and two fingers on each hand representing “four”). Moreover, given this partici-
pant’s use of groups, it seemed she was relying primarily on her algebraic thinking
structure. This finding was also insightful because this participant was in the midst
of constructing her pre-numerical units, allowing her physical activity to possibly
reflect on, which may engender her later mental activity development.
Next Steps in Early Childhood Mathematics Education
These findings, evidencing possible relationships between children’s counting and

subitizing, struck me as meaningful because it seemed that by subitizing more than
one group, this participant may have created pre-numerical units. It was unclear if
her segmenting when subitizing was related to her iterating when counting. I wanted
to examine reversible activity children constructed when subitizing and counting.
I began a small pilot study (MacDonald et al., 2016), developing tasks that drew
from inverse and compensation activity associated with number. These tasks were
designed to engender K-2 students’ +1/−1 activity across the equal sign. For
instance, for inverse tasks, children were often given the “bears in the cave” task,
where all counters (representing bears) were first shown to the students, and then
some counters were placed under a piece of paper (hiding in a cave). The students
were asked how many bears went into the cave. This task was designed to engender
inverse activity because they would “undo” what they had observed to determine
what items were taken from the entire group.
The compensation tasks included two sheets of paper, each representing a tray
where counters (a.k.a cupcakes) were placed in patterned arrangements (similar to
those of the face of a die). One paper had two sections, each with two sets of coun-
ters, and a second paper with two sections, but only one section had a set of coun-
ters. The counters were arranged so that there was relatively one more or one less
set of counters on one section of the first paper compared to one section of the sec-
ond paper (see Fig. 18.3). The students were told they needed the same number of
cupcakes on each tray and were asked, “How many cupcakes should we add to this
tray (second paper) so they are the same as this tray (first tray)?” This second task
was designed to develop early childhood students’ understanding of the relationship
of an action and its effect when operating on number (Piaget, 1968b).
Findings from this pilot study suggested that children were able to engage in
meaningful reversibility when they were capable of counting on (MacDonald et al.,
2016). When children count on, Steffe and Cobb (1988) explain that they are said to
have constructed internalized units by partitioning their number sequence. The pat-
terned items (see Fig. 18.3) children subitized gave them the opportunity to develop
reversibility. For instance, when children compensated for differences among pat-
terned items on either tray of cupcakes (see Fig. 18.3), they often engaged in a
“match and move” (MacDonald et al., 2016). Essentially, when children developed
a “match and move” between two different sets of items on either cupcake tray, the
patterns allowed them to rely on the visual patterns of each set of items before com-
paring patterns across sets of items among the trays (see Fig. 18.4). By adding or
taking away “one” from a perceptual unit, we found that students might construct
the entire unit and then segment this to create a new perceptual unit.
MacDonald et al. (2016) also found that these “match and move” physical actions
became mental activity wherein students relied on imagined matching the dots and
imagined moving one dot to keep the same number of cupcakes the same but have
Fig. 18.3 Compensation tasks for young children to examine reversibility development
580 B. L. MacDonald
Fig. 18.4 “Match and move” strategy used during the compensation tasks for young children to
examine reversibility development
them change by +1 or −1. Echoing Steffe and Cobb’s (1988) findings, we also
found that once students were able to develop figurative units, motor units, and
verbal units, with this “match and move” activity, they were better positioned to
develop abstract units and count on. Once students were able to count on, they were
able to transition to problems that distanced them from perceptual material (i.e.,
larger sets of items, covered items, and symbolic representations).
It was still unclear as to how students’ subitizing and counting might relate to
their early addition and subtraction development. In fact, I worked with my col-
leagues at Virginia Tech – Jay Wilkins and Andy Norton – on some data analysis,
examining relations between kindergarten children’s subitizing and their counting
in first grade (Wilkins et al., 2022). In this study, we hypothesized that when con-
trolling for children’s abstract unit development in kindergarten, children capable of
subitizing in kindergarten are more likely to count on in first grade (Wilkins et al.,
2022, p. 143). Findings indicated that when kindergarten children were unable to
accurately subitize both items, they had a 40.13% chance of counting on in first
grade. When kindergarten children subitized five items or three items, they had a
61.77% chance of counting on in first grade. Finally, when children subitized both
sets of items correctly, they were 78.92% more likely to count on in first grade. We
posited this likelihood developed because children were asked to show in some way
“how many” without the perceptual items present. This meant that children were no
longer constrained to only rely on perceptual items to enumerate, which encouraged
them to develop figurative units. These findings echoed the work I had done with
young children’s subitizing and counting (MacDonald & Wilkins, 2019) and
explained some children’s activity from Steffe’s (1991, 1992, 2010) findings where
children used an available pattern for three when counting on.
Based on some observations, I still wondered how students subitized differently
and how subitizing activity could be coupled with counting activity to assess young
children’s number construction. I worked with my colleague, Jessica Shumway, to
explain how to use my subitizing framework (MacDonald & Wilkins, 2019) and the
subitizing games. We wrote a professional journal article for early childhood educa-
tors on subitizing and games as a form of assessment (MacDonald & Shumway,
2016). After presenting some of this work, and introducing these games to early
childhood preservice teachers, I realized there was a lot about students’ number
understanding that we still did not yet know. For instance, we wondered: What are
the varying degrees and nuances of subitizing activity, and would students from dif-
ferent home experiences or with different learning abilities subitize and construct
number in the same way? From these questions, I began to wonder how I might
begin to use my expertise around young children’s number construction with subi-
tizing and counting to examine wider variability in students’ number development
to help educators provide equitable support to students as they construct number in
the early childhood grade levels.
Equitable Access to Opportunities for Number Construction
Following some discussions about the questions that we still had in mathematics
education, I began talking with my colleague, Jessica Hunt. We wondered how chil-
dren with learning disabilities constructed number when subitizing and counting
and how their number understanding was reorganized to develop fraction
conceptions.
To dive into how children with learning disabilities vary in their number develop-
ment, many questions were formed. Verschaffel et al. (2018) synthesized findings
from the neuroscience field on the development of number among children’s who
are identified as having a learning disability. This work posits that young children
developing number in activity, similar to subitizing, may have varying cognitive
factors, such as in their attentional mechanisms, constraining or differing their subi-
tizing development. This additional literature suggests that children with a learning
disability have varying types of counting development. From this, we began to won-
der if there were strengths or assets that we could build upon in the early grade
levels when considering ways to support these varying forms of development. By
reaching this goal, we seek to help all children access opportunities to develop con-
ceptual number as well as begin transitioning from preschool classrooms to kinder-
garten classrooms in more coherent ways.
Given that we don’t yet have the ability to identify children who have significant
developmental delays in mathematics in the younger grade levels, we realized that
we needed to first work with children identified with a learning disability in upper
elementary school (i.e., third grade, fourth grade). By working with children from
this population, we might be able to better characterize and delineate children’s
strengths instead of their deficits in their number development. Currently, it is often
rare to have asset-based performance measures in the special education intervention
design. By examining a child’s mathematical reality, we are more readily able to
help children develop new mathematical activity from their current activity.
In addition to the asset-based frame we aimed to carry into our work, we also
hoped to take up a reorganization hypothesis (Steffe, 2001). Steffe (2001) posits
children reorganize their whole number understandings to construct fraction under-
standings. By examining children’s reorganization, we began examining the broad
structures of children’s units coordination (see Chap. 11). To develop fraction
582 B. L. MacDonald
understandings, children reorganize their activity and units to develop a “units

within units” mental structure. When reorganizing three levels of units with whole
number understandings to make sense of fraction understandings, children revisit
some of their earliest whole number activity and transition to two levels of frac-
tional units.
By examining this reorganization that upper elementary children with a learning
disability engage in, we can better understand the activity they develop when first
coordinating units with whole numbers. Throughout such a study, there are particu-
lar unique features of this population of children’s activity we need to be aware of.
For instance, often, a child is determined as having a learning disability because
their working memory is developing at a different rate than their peers (e.g.,
Compton et al., 2012). Moreover, many children with a learning disability are not
yet capable in the third or fourth grade of coordinating three levels of units in their
whole number activity. Thus, such children’s reorganization of their units coordina-
tion would look different and possibly not provide enough mental activity or struc-
ture to support a part-whole understanding of fractions. Given this, we wondered if
we need to begin with older children or if we develop tasks that engender children’s
development of varying levels of units coordination (drawing on both mental and
physical activity) with whole number at the same time as they are developing part-
whole understandings and measurement understandings for fractions (see Chap. 17).
Finally, given my questions about how young children develop whole number
understandings and develop their subitizing and counting activity, I wondered if
these differences among children with varying abilities may explain a simple mis-
alignment with our children’s mathematical realities with our mathematics instruc-
tion. Could it be possible that all of our children are counting and subitizing in
different ways and need varying degrees of physical activity so that they can develop
meaningful mental activity when constructing whole number units coordination?
Given this, I also wondered how tasks could be better developed to allow meaning-
ful entry points, exit points, and moments of perturbation.
Conclusion and Final Thoughts
By examining children’s number development in early childhood, I have so many

more questions. How does gameplay at home and in school support children’s
counting, subitizing, and whole number strategy development? How can home
mathematics inform school mathematics?, and What cultural differences between
urban areas and rural areas, between varying ethnicities, and between varying fam-
ily structures uniquely contribute toward young children’s counting and subitizing
development? These questions (and more) continue to swim in my mind as I develop
and work in collaboration with colleagues on different research studies. I am also
collaborating with colleagues to consider how preservice elementary and early
childhood teachers’ specialized content knowledge with fractions develops in math-
ematics methods courses. More questions surface around this work, considering the
degree of impact this may have on educators’ instruction with whole number,
geometry, measurement, and fractions. I also wonder how children’s units con
struction and coordination relate to their Science, Technology, Engineering, and
Mathematics (STEM) learning. How might children’s computational thinking
evolve in relation to their units coordination?
To examine this work, I have had the honor of leading a working group for the
past several years at the North American Chapter of the International Group for the
Psychology of Mathematics Education conference. Through this working group,
multiple scholars have joined, and we have begun creating a space to discuss, reflect
on, develop, and refine our tasks, assessments, and studies. These efforts have been
carefully organized on a website: https://unitscoordination.wordpress.com/.
By examining young children’s subitizing development within a Neo-Piagetian
paradigm, I have had the pleasure of developing more questions than a research
career in mathematics education requires. The best part of working within this para-
digm is that children are often at the center of the work. By attempting to determine
a child’s mathematical reality, we are often surprised at the novel ideas and the
activity they carry into their mathematics solutions. This makes for a rich career
filled with new contributions, meaningful work, and always more questions.
References
Clements, D. H. (1999). Subitizing: What is it? Why teach it? Teaching Children Mathematics,
5(7), 400–405.
Clements, D. H., Sarama, J., & MacDonald, B. L. (2019). Subitizing: The neglected quantifier. In
A. Norton & M. W. Alibali (Eds.), Constructing number: Merging perspectives from psychol-
ogy and mathematics education (pp. 13–45). Springer.
Compton, D. L., Fuchs, L. S., Fuchs, D., Lambert, W., & Hamlett, C. (2012). The cognitive
and academic profiles of reading and mathematics learning disabilities. Journal of Learning
Disabilities, 45(1), 79–95.
Flavell, J. H. (1963). The developmental psychology of Jean Piaget. Litton Educational Publishing.
Inhelder, B., & Piaget, J. (1958). The growth of logical thinking: From childhood to adolescence
(A. Parsons & S. Milgram, Trans. 1959). Basic Books.
Kaufman, E. L., Lord, M. W., Reese, T. W., & Volkmann, J. (1949). The discrimination of visual
number. The American Journal of Psychology, 62(4), 498–525.
MacDonald, B. L. (2013). Subitizing activity: Item orientation with regard to number abstraction
(Doctoral dissertation, Virginia Polytechnic Institute and State University).
MacDonald, B. L. (2015). Ben’s perception of space and subitizing activity: A constructivist
teaching experiment. Mathematics Education Research Journal, 27(4), 563–584. https://doi.
org/10.1007/s13394-015-0152-0
MacDonald, B. L., & Shumway, J. F. (2016). Subitizing games: Assessing preschool children’s
number understanding. Teaching Children Mathematics, 22(6), 340–348.
MacDonald, B. L., & Wilkins, J. L. M. (2019). Subitising activity relative to units construc-
tion: A case study. Research in Mathematics Education, 21(1), 77–95. https://doi.org/10.108
0/14794802.2019.1579667
MacDonald, B. L., Boyce, S., Xu, C. Z., & Wilkins, J. L. M. (2015). Frank’s perceptual subi-
tizing activity relative to number understanding and orientation: A teaching experiment. In
T. G. Bartell, K. N. Bieda, R. T. Putnam, K. Bradfield, & H. Dominguez (Eds.), Proceedings of
the 37th annual Psychology of Mathematics Education conference, North American Chapter
(pp. 149–156). Michigan State University.
584 B. L. MacDonald
MacDonald, B. L., Ashby, M. J., & Litster, K. (2016). Preliminary findings of first grade stu-
dents’ development of reversibility. In M. B. Wood, E. E. Turner, M. Civil, & J. A. Eli (Eds.),
Proceedings of the 38th annual meeting of the North American Chapter of the International
Group for the Psychology of Mathematics Education (pp. 205–205). The University of Arizona.
McCulloch, W. S. (1963). Embodiments of mind. MIT Press.
Piaget, J. (1941). The child’s conception of number (B. Inhelder, Trans. 1965). Norton Library.
Piaget, J. (1968a). Genetic epistemology (E. Duckworth, Trans. 1970). Columbia University Press.
Piaget, J. (1968b). Structuralism (C. Maschler, Trans. 1970). Basic Books.
Piaget, J., & Inhelder, B. (1948). The child’s conception of space (F. J. Langdon & J. L. Lunzer,
Trans. 1967). Norton Library.
Pylyshyn, Z. W. (2001). Visual indexes, preconceptual objects, and situated vision. Cognition,
80(1–2), 127–158.
Sarama, J., & Clements, D. H. (2009). Early childhood mathematics education research: Learning
trajectories for young children. Routledge.
Steffe, L. P. (1991). The learning paradox: A plausible counterexample. In L. P. Steffe (Ed.),
Steffe, L. P. (2001). A new hypothesis concerning children’s fractional knowledge. The Journal of
Mathematical Behavior, 20(3), 267–307.
Steffe, L. P. (2010). Operations that produce numerical counting schemes. In L. P. Steffe & J. Olive
Steffe, L. P., & Cobb, P. (1988). Construction of arithmetical meanings and strategies. Springer.
Verschaffel, L., Baccaglini-Frank, A., Mulligan, J., Heuvel-Panhuizen, M. V. D., Xin, Y. P., &
Butterworth, B. (2018). Special needs in research and instruction in whole number arithmetic.
In Building the foundation: Whole numbers in the primary grades (pp. 375–397). Springer.
Wilkins, J. L. M., MacDonald, B. L., & Norton, A. (2022). Construction of subitizing units is
related to the construction of arithmetic units. Educational Studies, 109, 137–154.
Chapter 19
Researching Coordinate Systems Using
Genetic Epistemology Constructs
Hwa Young Lee
[T]he perception of space involves a gradual construction and certainly does not exist ready
made at the outset of mental development. (Piaget & Inhelder, 1967, p. 6)
The ‘intuition’ of space is not a ‘reading’ or apprehension of the properties of objects, but
from the very beginning, an action performed on them. (Piaget & Inhelder, 1967, p. 449)
My ongoing research agenda involves studying how students deliberately select,

establish, and coordinate frames of reference and construct coordinate systems to
organize phenomena via graphical representations. An overarching assumption I
make in my work is that mathematical objects or “representations” do not exist as
things in themselves; instead, they are constructed by individuals in goal-directed
activity (von Glasersfeld, 1987). Therefore, phenomena or graphical representa-
tions of such phenomena do not contain frames of reference, coordinate systems,
and graphs until an individual imputes them onto a situation. Taking such a perspec-
tive has led me to focus my research efforts on three interrelated goals:
1. Making theoretical distinctions that could help better understand why students
might engage in graphing activities the way they do.
2. Designing mathematical tasks that could be used to leverage students’ spontane-
ous and intuitive thinking to support their graphing activities.
3. Building conceptual models that contribute to a better understanding of students’
constructive resources they develop when engaging in graphing activities.
When building conceptual models of students’ thinking, I adopt Piaget’s genetic
epistemology constructs, such as scheme, operation, and assimilation, via von
Glasersfeld’s (1995) interpretations and elaborations.1 Specifically, I attend to
students’ constructive resources—both cognitive and assimilatory resources in their
constructive activities. By cognitive resource, I refer to schemes and operations
that constitute students’ mental activity. By assimilatory resource, I refer to the
1
Please refer to Tillema and Gatza’s (Chap. 4) discussion on these constructs.
H. Y. Lee (*)
e-mail: hylee@txstate.edu
https://doi.org/10.1007/978-3-031-47386-9_19
586 H. Y. Lee
experiences and intuitions in a context (e.g., map reading, navigating, relaying

explanations of locations to others, viewing objects from various vantage points,
etc.) that may contribute to the recognition and enactment of cognitive schemes and
operations based on what I learn from students.2 Using these constructs has helped
me make sense of the patterns of thinking and actions that I observed in students’
mathematical activity.
When it comes to studying students’ coordinate systems, I found Piaget and col-
leagues’ work on children’s conception of space and geometry to be insightful. In
this chapter, I elaborate on some of the main constructs that have guided my
research. Specifically, I discuss children’s organization of space and logical multi-
plication of measurements coupled with units coordination.
Piaget’s Distinctions in Children’s Organizations of Space
As illustrated in the opening quotes, the active role of a cognizing subject in their
organization of space is essential in Piaget’s genetic epistemology. Relatedly, Piaget
and colleagues studied the process by which children construct, abstract, and opera-
tionalize conceptual space from perceptual space (Piaget & Inhelder, 1967; Piaget
et al., 1960). To illustrate, consider one of the tasks from Piaget et al.’s (1960) clini-
cal interviews with children. Given two congruent sheets of rectangular paper (S
and S′ in Fig. 19.1), children were asked to mark a point on the blank sheet of paper
(rectangle S′) exactly where a mark (point P in Fig. 19.1) was made on the other
sheet of paper (rectangle S) so when the two sheets of paper were superimposed on
top of each other, the marks would line up. The children were given a ruler, stick,
strips of paper, and thread as tools they could use. I invite the reader to pause and
think about how they might copy the location of point P onto Rectangle S′.
From their observations of children’s activities, Piaget and colleagues abstracted
three stages in which children developed a reference system to mark the point’s
Fig. 19.1 An illustration of rectangles S and S′ and point P in Piaget et al.’s (1960) task
2
See Lee (under review) for illustrations of these resources.
19 Researching Coordinate Systems Using Genetic Epistemology Constructs 587
location. In the first stage, children made a simple estimate by looking at the point
on the first sheet of paper. Without using any of the tools available, the children in
this stage made visual judgments about the location of the point based on qualitative
features, such as proximity, in the rectangular space. Children in the second stage
started using the ruler but attended to only one measurement. Some children
attempted to preserve the inclination of the ruler, but this preservation was a visual
one. For example, children connected one corner of the paper and point P and tried
to move the ruler to the other sheet of paper, keeping the same inclination. The
preservation of the inclination was made visually and not based on specific mea-
surements carried out. Finally, children in the third stage operationalized how to
measure and maintain this inclination: “Gradually, they decompose [the ruler’s]
inclination and express it in terms of two separate measurements along different
axes” (Piaget et al., 1960, p. 155). In the end, the children took measurements along
both vertical and horizontal dimensions into account and “dissociate them and coor-
dinate them operationally” (Piaget et al., 1960, p. 159).
These three stages in which children developed a reference system to represent a
point’s location illustrate Piaget and colleagues’ distinctions between different
organizations of space, which have guided my work: (a) qualitative and quantitative
organization of space, and (b) perceptual, sensorimotor, and representational space.
Qualitative and Quantitative Organization of Space: Developing My Notion of
Frames of Reference and Coordinate Systems Piaget et al. (1960) distinguished
between qualitative and quantitative organization of space. Organizing space in a
qualitative sense means to think about objects within the space by spatial features
without consideration of the scale or measurement of the intervals between the
objects (e.g., proximity, separation, enclosure, or order). Children in the second
stage above had a qualitative conception of where the point should be—a certain
inclination from a corner; however, they were yet to quantify the inclination. When
one accounts for the measurements of intervals between objects (in the metric
sense), then the space is organized quantitatively. Children in the third stage estab-
lished some quality of an object (e.g., inclination of the ruler from one corner of the
paper to the point) but also accounted for the extent of the quality via associated
quantities measured (e.g., horizontal and vertical distances tied with the inclination
of the ruler). Piaget et al. (1960) explained that children’s construction of space
develops from qualitative to quantitative space and quantitative space presupposes a
qualitative organization of space.
This distinction has influenced my conceptualization of frame of reference and
coordinate system, which developed initially in spatial contexts. I note that Piaget
and colleagues used reference frames, reference systems, and coordinate systems
interchangeably. In sum, for Piaget et al. (1960), a reference frame or coordinate
system is like a container, which entails the relations of order (qualitative relations)
and distance (quantitative relations) between objects within the space and a simul-
taneous organization of all possible positions of any but no particular point within
the space. However, I purposefully distinguish between frames of reference and
coordinate systems to attend to students’ qualitative/quantitative organization of
588 H. Y. Lee
space. This distinction has also helped me further examine students’ logical multi-
plication of measurements, which I discuss later in this chapter.
My initial conceptualization of a frame of reference involved an individual’s
qualitative conception of relative locations of objects within a space, and a coordi-
nate system involved establishing a quantitative organization of where a point
should be. I further extended these initial conceptions into other contexts beyond
spatial. More generally, by frames of reference, I refer to mental structures that
individuals construct and use to gauge the relative extent of various attributes in the
phenomenon being organized (Joshua et al., 2015; Lee, 2017; Levinson, 2003).
Thinking within frames of reference involves establishing a qualitative idea of
quantities involved in the phenomenon, which entails attending to and establishing
reference points, directionality, and thinking about how to measure the quantities
being organized (Joshua et al., 2015; Lee et al., 2019). By coordinate system, I mean
a space in which quantities (Thompson, 2011) are coordinated to systematically
organize some phenomenon. The axes constituting a coordinate system are geomet-
ric embodiments of the frames of reference an individual established with specified
units of measure. In this way, a coordinate system allows an individual to coordinate
multiple frames of reference. For example, an individual may overlay two quantities
each onto a number line and arrange each number line perpendicularly with their
reference points coinciding to form a Cartesian coordinate system. This geometric
object allows the individual to coordinate both quantities. Guided by these concep-
tions, I have focused my attention on investigating ways students deliberately select,
establish, and coordinate frames of reference to construct coordinate systems as
representational space.
Studying Students’ Coordinate Systems as Constructions of Representational
Space The characterizations of children’s activities in each stage, as described
above, have also helped me make sense of Piaget and Inhelder’s (1967) distinction
between perceptual, sensorimotor, and representational space. Perceptual space is
the space perceived through perceptual activity on elements of raw material. For the
children in the first stage, as described above, the rectangular sheet of paper space
was perceptual—they relied on visually estimating the point on the piece of paper.
Both sensorimotor and representational space are considered conceptual space,
space abstracted from perceptual space. Sensorimotor space is an organized and
experienced space; however, it lacks a symbolic function in that the subject is unable
to imagine it or mentally reconstruct it. The children in the second stage, as described
above, are likely to have established the rectangular sheet of paper as sensorimotor
space—they tried to move the ruler while preserving its inclination; however, this
preservation relied on their sensorimotor actions. Representational space entails a
symbolic function, which leads one “to regulate his spatial behavior through a sys-
tem of total representation of his displacements rather than according to simple
motor expectations” (Laurendeau & Pinard, 1970, p. 12). For children in the third
stage, as described above, the rectangular sheet of paper space must have become a
representational3 space. Note perceptual space and conceptual space are not sepa-
rate, as Laurendeau and Pinard (1970) explained:
Between these two types of structures a reciprocal influence or functional interaction must
operate; at all levels of development, the information provided by perception (or the mental
image) serves as raw material for the intellectual action or operation, and, reciprocally,
these intellectual activities exert an influence (direct or indirect) on perception, enriching
and increasing the flexibility of its functioning with development. (p. 10)
As Piaget and Inhelder (1967) stated, “[K]nowing nothing of the stages which led
up to this transformation, the adult assumes that perception involves co-ordinate
systems or vertical-horizontal relations right from the outset, when in fact such
systems are extremely complicated” (p. 4). In that regard, I study how students’
coordinate systems develop into representational space from their activities with
perceptual/sensorimotor space. Relatedly, I have designed mathematical tasks that
involve students engaging in spontaneous locating activity in various contexts (e.g.,
locating a missing person in the North Pole region; locating a fish in a fish tank).
These locating activities are built on the premise that determining an object’s loca-
tion is a complex, coordinated performance that involves various strategies to see
where things lie in respect to one another. I have been inspired by Piaget et al.’s
(1960) tasks involving locating a point on a rectangular sheet of paper (as men-
tioned above) or locating a point in a wired box (three-dimensional analog of the
rectangular paper task) in their study of children’s construction of reference systems.
In designing my own tasks, my first guiding principle has been to create situa-
tions in which students will engage in goal-directed locating activity. In other words,
I try to come up with tasks where students feel the need to (systematically) locate a
point in two- or three-dimensional space. For example, I ask students to describe the
location of fish in a fish tank to someone in another room who is trying to make a
replica of the given fish tank. Second, and relatedly, I embed tasks in experientially
real contexts (Gravemeijer & Doorman, 1999) that build on students’ experiences
with everyday activity (e.g., navigating) such that students can be motivated to use
their intuitions of and operations on perceptual/sensorimotor space and quantities.
Third, different from approaches taken by Piaget et al., I do not necessarily limit
contexts to those related to the Cartesian coordinate system. Because my goal is to
observe students’ spontaneous constructions of coordinate systems, I vary not only
contexts but also the contour of the spatial contexts (e.g., providing cubic, cylindri-
cal, and spherical fish tanks). These principles have helped me attend to different
ways students deliberately select, establish, and coordinate frames of reference and
construct coordinate systems to organize phenomena. An elaboration of these design
principles in relation to two tasks I designed and models of students’ coordinate
systems can be found in Lee (2020, under review).
3
Refer to Thompson and Byerley (Chap. 5) for a further elaboration on the distinction between
representation and representation in relation to imagery.
590 H. Y. Lee
iaget’s Logical Multiplication of Measurements

P
and Units Coordination
Several studies have documented difficulties and challenges students experience

when creating or interpreting coordinate systems (Herscovics, 1989; Mevarech &
Kramarsky, 1997; Sarama et al., 2003). Collectively, these studies show that two
elements are critical but also challenging for students when engaging in graphing
activity: (1) establishing axes that entail both spatial positioning and numerical val-
ues and (2) coordinating two axes in a specific way to allow coordination of two
quantities. I view these two elements as captured in Piaget et al.’s (1960) notion of
establishing a logical multiplication of measurements. Drawing from the notion of
logical multiplication of measurement, I hypothesized that students’ units coordina-
tion serves as a cognitive resource in coordinating multiple frames of reference, and
hence, in constructing coordinate systems.
Logical Multiplication of Measurements Piaget et al. (1960) referred to chil-
dren’s coordination of lengths along both vertical and horizontal dimensions in the
final stage (as described in the task in Fig. 19.1) as a logical multiplication of mea-
surements, which is a key mental process needed in establishing reference systems.
A logical multiplication of measurements involves a twofold one-to-one correspon-
dence of points on the horizontal and vertical axis. Specifically, it involves holding
a point’s specific qualitative position (order and direction) along one axis with that
along another axis and also holding a point’s quantitative location (length) from a
reference point along one axis with that along another axis (see detailed illustration
in Lee, under review).
Guided by this notion of logical multiplication of measurements, my work has
focused on modeling how students might hold a point’s qualitative position via a
frame of reference with that along another frame of reference and also hold a point’s
quantitative location from a reference point along a frame of reference with that
along another frame of reference via a coordinate system. Relatedly, in my work
with ninth graders and undergraduate preservice teachers, I observed a variety of
ways (not restricted to the Cartesian coordination) students established and coordi-
nated their frames of reference and coordinated measurements guided by their
frames of reference to quantitatively describe the location of points in various con-
texts. From students’ activities, I abstracted the frames-of-reference (FR)-
coordinating scheme and have argued that the FR-coordinating scheme is necessary
for an individual to construct a coordinate system (Lee, 2017, under review).
Units Coordination Researchers have identified units coordination4 as an impor-
tant operation involved in students’ constructions of various mathematical concepts
or ways of reasoning, such as fractions (e.g., Norton & Boyce, 2013; Steffe & Olive,
2010), multiplicative relationships (e.g., Hackenberg, 2010; Tillema, 2012),
proportional reasoning (e.g., Steffe et al., 2014), and quantifying angularity (e.g.,
4
Please see Hackenberg (Chap. 14) for a detailed discussion on this construct.
Hardison, 2018). In my work, drawing from the aforementioned notion of logical

multiplication of measurements (Piaget et al., 1960), I hypothesized that students’
units coordinations serves as a cognitive resource in constructing coordinate
systems.
Piaget et al. (1960) explained, “A coordinate system, even though a qualitative
one, requires the imposition of spatial order in two or more dimensions between the
several elements, and also demands a systematic nesting of partial intervals between
such elements” (p. 160). Recall that a logical multiplication of measurements
involves an individual holding a point’s specific qualitative position (order and
direction) along one axis with that along another axis and also holding a point’s
quantitative location (length) from a reference point along one axis with that along
another axis. What this means is that even at the qualitative level, holding in mind a
specific position in relation to order and direction involves a systematic nesting of
partial intervals; it seems needless to say that a systematic nesting of partial inter-
vals along with subdivision is involved in holding a point’s quantitative location.
Therefore, taking frames of reference as spatio-structural units, I hypothesized that
students’ units coordinations serves as a cognitive resource in constructing coordi-
nate systems.
Through teaching experiments (Steffe & Thompson, 2000) with ninth graders
and undergraduate preservice teachers, I worked with students ranging from those
operating with two levels of units in activity (Hackenberg, 2007) to those operating
with three levels of units as given. From their locating activities, I have found
nuanced distinctions in the activities in their FR-coordinating schemes—namely,
simultaneous or sequential coordination of one, two, or three frames of reference in
two- or three-dimensional space, consistent with what I expected from their levels
of units coordinations (Lee, 2017). Although my findings supported my hypothesis,
future work is needed to further explore and elaborate on this hypothesis.
Moving forward, I intend to continue testing my research hypotheses inspired by
Piaget’s genetic epistemology and building second-order models of students’ graph-
ing activities. In closing, I emphasize that the aforementioned elaborations are my
interpretations of genetic epistemology constructs and that my understanding of
these constructs, hypotheses, and conceptual models are evolving as I continue to
work with students.
Acknowledgments I thank Les Steffe for his influence on sparking my interest in and my under-
standings of Piaget’s genetic epistemology. I also thank Andy Norton and Jessica Hunt for their
thoughtful comments on an earlier version of this chapter.
References
Gravemeijer, K., & Doorman, M. (1999). Context problems in realistic mathematics education: A
calculus course as an example. Educational Studies in Mathematics, 39(1–3), 111–129.
sion of the splitting hypothesis. The Journal of Mathematical Behavior, 26(1), 27–47.
592 H. Y. Lee

Cognition & Instruction, 28(4), 383–432. https://doi.org/10.1080/07370008.2010.511565
Hardison, H. L. (2018). Investigating high school students’ understandings of angle measure.
Unpublished doctoral dissertation, University of Georgia, Athens, GA.
Herscovics, N. (1989). Cognitive obstacles encountered in the learning of algebra. In S. Wagner &
C. Kieran (Eds.), Research issues in the learning and teaching of algebra (Vol. 4, pp. 60–86).
Lawrence Erlbaum Associates, Inc.
Joshua, S., Musgrave, S., Hatfield, N., & Thompson, P. W. (2015). Conceptualizing and reason-
ing with frames of reference. In T. Fukawa-Connelly, N. Infante, E. K. Keene, & M. Zandieh
(Eds.), Proceedings of the 18th meeting of the MAA special interest group on research in under-
graduate mathematics education (pp. 31–44). RUME.
Laurendeau, M., & Pinard, A. (1970). The development of the concept of space in the child.
International Universities Press.
Lee, H. Y. (2017). Students’ construction of spatial coordinate systems. Unpublished doctoral dis-
sertation, University of Georgia, Athens, GA.
Lee, H. Y. (2020). Tell me where they are. Mathematics Teacher: Learning and Teaching PK-12,
113(11), e78–e84.
Lee, H. Y. (under review). Inventing coordinate systems: Modeling students’ spontaneous coordi-
nation of frames of reference. Cognition and Instruction
Lee, H. Y., Moore, K. C., & Tasova, H. I. (2019). Reasoning within quantitative frames of refer-
ence: The case of Lydia. The Journal of Mathematical Behavior, 53, 81–95.
Levinson, S. C. (2003). Space in language and cognition: Explorations in cognitive diversity (Vol.
5). Cambridge University Press.
Mevarech, Z. R., & Kramarsky, B. (1997). From verbal descriptions to graphic representations:
Stability and change in students’ alternative conceptions. Educational Studies in Mathematics,
3, 229c263. https://doi.org/10.2307/3482634
Norton, A., & Boyce, S. (2013). A cognitive core for common state standards. The Journal of
Mathematical Behavior, 32(2), 266v279.
Trans.). The Norton Library.
Piaget, J., Inhelder, B., & Szeminska, A. (1960). The child’s conception of geometry. Basic
Books.
Sarama, J., Clements, D. H., Swaminathan, S., & McMillen, S. (2003). Development of mathemat-
ical concepts of two–dimensional space in grid environments: An exploratory study. Cognition
and Instruction, 3, 285–324. https://doi.org/10.2307/3233812
Steffe, L. P., Liss, D. R., II, & Lee, H. Y. (2014). On the operations that generate intensive quantity.
In K. C. Moore, L. P. Steffe, & L. L. Hatfield (Eds.), Epistemic algebraic students (Vol. 4,
pp. 49–79). University of Wyoming Press.
Thompson, P. W. (2011). Quantitative reasoning and mathematical modeling. In S. Chamberlin,
in mathematics education: Papers from a planning conference for WISDOMê (pp. 33–57).
University of Wyoming.
Tillema, E. S. (2012). Relating one and two-dimensional quantities: Students’ multiplicative
reasoning in combinatorial and spatial contexts. In R. Mayes, R. Bonillia, L. L. Hatfield, &
S. Belbase (Eds.), Quantitative reasoning: Current state of understanding WISDOMe mono-
graphs (Vol. 2, pp. 143–148). University of Wyoming Press.
von Glasersfeld, E. (1987). Preliminaries to any theory of representation. In C. Janvier (Ed.),

Problems of representation in the teaching and learning of mathematics (pp. 215–225).
Lawrence Erlbaum.
Chapter 20
Researching Quantifications of Angularity
Using Genetic Epistemology Constructs
Hamilton L. Hardison
In this chapter, I discuss the interplay between constructs with Piagetian roots and
the planning of my dissertation study (Hardison, 2018), which involved a teaching
experiment focused on ninth-grade students’ quantifications of angularity. I reflect
on how exposure to such constructs supported my articulation of a dissertation topic
and my organization of prior literature relative to this topic. Following this, I con-
sider how reflection on Neo-Piagetian work impacted the initial design of my dis-
sertation study. Specifically, I discuss how Steffe and colleagues’ approach to
studying children’s fractional knowledge impacted the design of my dissertation
study in at least two regards: (a) articulating a guiding hypothesis for my program
of research and (b) selecting and designing initial interview tasks. I close the chapter
by mentioning a persistent challenge I encounter when presenting my work, as well
as a promising indication regarding progress in research on quantifying angularity.
istinguishing Quantifications of Angularity Using

D
Piagetian Constructs
My interest in how individuals quantify angularity was occasioned, in part, by my

participation in two courses during my third year as a doctoral student. The first was
a content course with a precalculus emphasis, which was designed for undergradu-
ates who were prospective secondary mathematics teachers; the second was a doc-
toral seminar on quantitative reasoning.
H. L. Hardison (*)
e-mail: HHardison@txstate.edu
https://doi.org/10.1007/978-3-031-47386-9_20
596 H. L. Hardison
Thinking About One-Degree Angles in the Content Course
Early in the content course, the instructor’s goals included activating and enriching
the prospective teachers’ ways of reasoning about angular measure. To some extent,
my dissertation research stemmed from extended reflection on a single prompt,
which the instructor posed during an initial course meeting: What does it mean for
an angle to have a measure of one degree?
When considering how I would respond to this prompt, I realized I could con-
ceive at least two different, coherent ways an individual might reason about the
meaning of a one-degree angle. First, if an individual drew an arbitrary circle cen-
tered at the vertex of a one-degree angle, then the minor arc intercepted by the angle
would have an arc length that was 1/360 of the circle’s circumference; therefore, an
individual could produce a one-degree angle by partitioning an arbitrary circle into
360 equal-length arcs and selecting a central angle subtending any one of these arcs.
For the moment, I refer to this first way of reasoning as circle reasoning. Second, if
an individual were to iterate a one-degree angle (along with its interior) 360 times,
then the result of the iteration would exhaust the plane; therefore, an individual
could produce a one-degree angle by radially partitioning a plane into 360 equian-
gular parts and selecting any one of those parts. For now, I refer to this second way
of reasoning as non-circle reasoning. Models of one-degree angles (shown in red)
produced via reasoning with circles and reasoning without them are depicted in
Figs. 20.1 and 20.2, respectively, though these figures push the boundaries of my
current image production capabilities.
These two ways of reasoning about degrees as a unit of angular measure are
distinct in that different geometrical objects are foregrounded. Circle reasoning
foregrounds circles and arcs, and the measure of a central angle can be determined
Fig. 20.1 A model of a

one-degree angle produced
by selecting a central angle
subtending one arc of a
circle equipartitioned into
360 congruent arcs
20 Researching Quantifications of Angularity Using Genetic Epistemology Constructs 597
Fig. 20.2 A model of a one-degree angle produced by equipartitioning a plane into 360 equiangu-
lar parts and selecting one of those parts
via a composite structure involving circumference and arc length.1 In contrast, nei-
ther circles nor arc lengths are involved in non-circle reasoning. Instead, such rea-
soning can entail conceiving of the measure of an angle in degrees via a composite
structure between an angular section (i.e., an angle along with its interior) and the
plane, which can also be considered an angular section.
stablishing Orienting Constructs in the Doctoral Seminar

E
and Elsewhere
Fortunately, I was enrolled in a doctoral seminar on quantitative reasoning (e.g.,

Thompson, 1990, 2011) during the same semester that I began thinking more deeply
about angular measure.2 Delving into the literature on quantitative reasoning (and
related topics, many of which are contained in previous chapters) provided me with
analytical footholds for elaborating the distinctions between the circle reasoning
and non-circle reasoning described above. For example, these ways of reasoning
implied different imagined measurement processes for (what an observer could
call) the same attribute—in the verbiage of Thompson’s theory of quantitative rea-
soning, these indicated different quantifications. Thus, at some point, rather than
1
If the circle is conceived as a unit composed of congruent unit arcs, then I consider these units to
have a composite structure (cf. Hackenberg and Sevinc, Chap. 11).
2
Boyce (Chap. 12) has provided a discussion of quantitative reasoning, its constituent constructs,
and those who have contributed to its development.
598 H. L. Hardison
describing my research interest as “angle measure,” I began to describe my interest

as “how individuals quantify angularity.”3
Furthermore, other Piagetian-rooted constructs gradually allowed me to more
precisely specify the nature of the distinctions I was considering. For example,
although both quantifications suggested above leveraged extensive quantitative
operations like iteration and partitioning, the figurative material subjected to these
operations differed across the quantifications. Put succinctly, non-circular quantifi-
cations of angularity entail operations enacted on angular sections; circular quanti-
fications entail operations enacted on portions of circles (e.g., arcs).
Consulting and Organizing Prior Literature
Although being able to better articulate distinctions was progress toward conduct-
ing my dissertation research, purposeful consultation of literature on angle and
angular measure was critical in several regards. Since I wanted to make a novel
contribution to the field, discovering whether other researchers had attended to
angular measure similarly was of primary importance. To my pleasant and genuine
astonishment, I found that, although some literature on angle and angular measure
existed, remarkably few works focused on individuals’ quantifications of angular-
ity. These few works prioritized what I have called circular quantifications of angu-
larity and were authored by Thompson or his academic descendants (Moore, 2009,
2010, 2012, 2013, 2014; Tallman, 2015; Thompson, 2008).4,5 Although I found
some practitioner literature directed toward activities that might be used to foster
non-circular quantifications of angularity (Browning et al., 2007; Host et al., 2014;
Millsaps, 2012; Sherman, 2017; Wilson, 1990; Wilson & Adams, 1992), I found no
empirical research investigating non-circular quantifications of angularity.
Thompson’s characterization of a quantity in terms of three interrelated compo-
nents—object, attribute, and quantification—yielded a convenient organizational
structure for critiquing reports of extant empirical studies in my dissertation’s litera-
ture review. By this, I mean the characterization afforded categorizing studies about
angle and angular measure. Some studies (Keiser, 2000, 2004; Keiser et al., 2003;
Mitchelmore, 1997, 1998; Mitchelmore & White, 1995, 1998, 2000; Prescott et al.,
2002; Silfverberg & Joutsenlahti, 2014) focused primarily on individuals’ concep-
tions of angles as objects without sufficiently attending to an attribute of interest or
3
The latter phrasing clearly positions my research as part of “a core mission of mathematics educa-
tion research” (Thompson and Byerley, Chap. 5) in that it implies attending to viable mathematics
of students; the former phrasing fails to incorporate human actors.
4
See also Tallman and O’Bryan’s discussion of angle measure in Chap. 8.
5
Piaget et al. (1960, 1967) attended to children’s constructions of angles as objects as well as their
quantifications of angularity in the sense of establishing operations constitutive of angular congru-
ence; however, I am unaware of any Piagetian work explicitly addressing the quantifications of
angularity alluded to here.
the process through which this attribute might be measured. Other studies (Devichi
& Munier, 2013; Lehrer et al., 1998; Noss, 1987) focused primarily on attributes
individuals find salient when attending to angle models but did not foreground
enacted or imagined measurement processes. Finally, some studies did focus on
angular measurement (Clements et al., 1996; Clements & Burns, 2000; Crompton,
2013, 2017; Simmons & Cope, 1990; Smith et al., 2014); however, the participants
in this final group of studies were provided with virtual protractors, which, I argued,
obscured reliable insights regarding the participants’ quantifications of angularity.
Formulating Hypotheses
Consulting and organizing extant literature indicated that a dissertation investigat-

ing quantifications of angularity would constitute a novel contribution to the field.
To bolster the chance that this novel contribution might also be substantive, I sought
to emulate Les Steffe, specifically with respect to his collaborative long-term
endeavor to elucidate how children can construct fractional knowledge (Steffe &
Olive, 2010).6 Here, I focus on just one distinguishing characteristic of this program
of research, which was the articulation (and yearslong empirical investigation) of
the reorganization hypothesis: “The basic hypothesis that guides our work is that
children’s fraction schemes can emerge as accommodations in their numerical
counting schemes” (Steffe & Olive, 2010, p. 1). Put in terms of quantitative reason-
ing, the reorganization hypothesis states that children’s quantifications of (rational)
length can emerge as accommodations in their quantifications of discrete numerosity.
Paralleling Steffe’s work, I situated my research program about a basic reorgani-
zation hypothesis focused on angularity rather than fractions. In particular, the
angular reorganization hypothesis I began to test in the dissertation was as follows:
individuals’ quantifications of angularity can emerge as modifications in their quan-
tifications of length.7 To argue the viability of this reorganization hypothesis, I pre-
sented, early in the dissertation, an initial first-order conceptual analysis considering
what early quantifications of angularity might entail and how modifications to such
quantifications might yield superseding quantifications (e.g., circular quantifica-
tions of angularity, which have been advocated by Moore (2013) and Thompson
(2008)). The following paragraphs briefly illustrate that a reorganization hypothesis
of this sort is potentially viable.
From a Piagetian perspective, an extensive quantity arises when an individual
introduces units into a gross quantity (Piaget, 1965; Steffe, 1991; cf. Boyce, Chap.
12). The operations Steffe and colleagues abstracted to build second-order models
6
See also Chap. 2.
7
Steffe’s reorganization hypothesis stands in opposition to the interference hypothesis, which
states that “whole number knowledge interferes with the learning of fractions” (Steffe & Olive,
2010, p. 2). The analogous angular interference hypothesis is that quantifications of length inter-
fere with quantifying angularity.
600 H. L. Hardison
Fig. 20.3 A model

illustrating an angle four
times as open as a given
angle produced via
iteration
of fractional knowledge (e.g., iterating, partitioning, and splitting) are examples of

extensive quantitative operations because these operations produce units. From my
perspective, if an individual’s quantification of length entails particular extensive
quantitative operations, then so too might their eventual quantifications of angular-
ity. For instance, if an individual can mentally enact iteration to produce a segment
four times as long as a given segment, then I anticipated conceiving an angle four
times as open as a given angle via iteration (Fig. 20.3) would be part of their quan-
tification of angularity or in their zone of potential construction. In this way, apply-
ing extensive quantitative operations to arbitrary angular units permits quantifying
angularity without attention to circles, arcs, or standard units of angular measure
(e.g., degrees and radians).
As suggested by my earlier discussion of non-circular quantifications of angular-
ity (see Fig. 20.2), I further hypothesized that degrees could emerge as a standard
unit of angular measure as a consequence of non-circular quantifications of angular-
ity. Prior to my teaching experiment, I conjectured that one-degree angles could be
constructed via the application of the splitting operation to a full angle.8 In other
words, one way to produce a one-degree angle is to solve the following splitting
task: A full angle is 360 times as open as some other angle; how would you make
this other angle?9
Designing Some Initial Interview Tasks
A methodological consequence of the angular reorganization hypothesis I adopted

was the need to account for participants’ quantifications of length—and, therefore,
their fractional and counting schemes—at the onset of my dissertation study.
However, an in-depth investigation of these schemes was not feasible, given my
focus on angularity and my temporal constraints as a doctoral student. Therefore, to
indirectly account for these schemes, I elected to assess whether participants had
constructed selected mental operations for linear contexts, as well as their stages of
8
The splitting operation involves simultaneous implementation of partitioning and iterating opera-
tions. See Norton’s discussion of splitting in Chap. 7.
9
My subsequent dissertation results and later work (Hardison, 2018, 2019, 2020; Hardison & Lee,
2019; Hardison et al., 2022) indicated that other angular templates, like right angles and straight
angles, are more viable starting points than a full angle for constructing the degree as a standard
angular unit.
units coordination.10 Ultimately, a subset of my initial interview protocol comprised

tasks about length, which were essentially borrowed directly from Steffe or others
who had extended his work (e.g., Hackenberg, 2005). For example, I used the fol-
lowing linear splitting task: “This is my piece of string. My piece of string is five
times as long as your piece of string. Can you make your string?” (cf. Norton’s
splitting task, Chap. 7). I emphasize that there was little, if any, need for me to be
creative when designing these tasks involving linear material precisely because
other researchers had already established the predictive utility of these tasks, pro-
duced descriptions of different ways students approached them, and considered the
implications of these differences. However, a modicum of creativity was required in
designing tasks for assessing the operations, including units coordination, to which
students could assimilate angular material at the onset of the study. So, I developed
parallel tasks involving hinged wooden chopsticks to investigate the angular ways
of reasoning students brought into the teaching experiment. For example, I used the
following angular splitting task in initial interviews: “This is my pair of chopsticks.
Can you set your pair of chopsticks so that my chopsticks are five times as open as
your chopsticks?”
Concluding Remarks
Emerging researchers may find two more points helpful should they investigate
notions of quantification or angularity. The first point is a persistent challenge I have
encountered in presenting my work—mathematics educators sometimes conflate
quantification and measurement, which impacts their responses to my research. For
example, after sharing my work at a conference, an attendee wondered why I was
worried about angular measure at all when dynamic geometry software (e.g.,
GeoGebra) could quickly report accurate measurements. Such comments serve as
an important reminder that mathematics educators’ conceptions of quantification
(as a general construct) require our support in the form of intentional interactions,
which is reminiscent of Thompson’s (2011) remark: “We must take quantification
seriously if we want to develop instructional and curricular strategies for having
students’ mathematics be of use to them in dealing with their world” (p. 38).
The second point is an acknowledgment that progress has been made since I
conducted my dissertation research. Here, I highlight the dissertations of Hanan
Alyami (2022), Brooke Mullins (2020), Erell Germia (2022), and Grace Vishner
(2020) so that emerging researchers have additional models for the forms research
on quantifying angularity might take, as well as additional persons to contact if they
are interested in talking more about this work. Considering the scarcity of research
on quantifying angularity when I was planning my dissertation study, the comple-
tion of so many studies within such a short time frame suggests an awareness of the
10
Hackenberg and Sevinc discuss stages of units coordination in Chap. 11.
602 H. L. Hardison
importance of attending to quantifications of angularity has gained some promi-

nence in the field. Moreover, these collective works suggest that significant progress
in understanding and supporting individuals’ quantifications of angularity is on the
horizon.
Acknowledgments I thank the editors for organizing this book and inviting me to contribute. To
this volume’s authors, I express my gratitude for furthering my understanding of genetic episte-
mology constructs and the roles they have and might play in mathematics education. I appreciate
those who offered feedback on an earlier version of this chapter: Hanan Alyami, Suby Kandasamy
Kularajan, Beth MacDonald, and Andy Norton. To Les Steffe, I extend special thanks, both for
introducing me to the extension of Piaget’s epistemology that became known as radical construc-
tivism and also for teaching me to conduct independent research, which he accomplished while
treating me as though I was capable of it throughout my doctoral studies.
References
Alyami, H. (2022). Preservice mathematics teachers’ conceptions of radian angle measure

[Unpublished doctoral dissertation]. Purdue University.
Browning, C. A., Garza-Kling, G., & Sundling, E. H. (2007). What’s your angle on angles?
Teaching Children Mathematics, 14(5), 283–287.
Clements, D. H., & Burns, B. A. (2000). Students’ development of strategies for turn and angle
measure. Educational Studies in Mathematics, 41(1), 31–45.
Clements, D. H., Battista, M. T., Sarama, S., & Swaminathan, S. (1996). Development of turn
and turn measurement concepts in a computer-based instructional unit. Educational Studies in
Mathematics, 30(4), 313–337.
Crompton, H. C. (2013). Understanding angle and angle measure: A design-based research study
using context aware ubiquitous learning. International Journal of Technology in Mathematics
Education, 22(1), 19–30.
Crompton, H. C. (2017). Using mobile learning to support students’ understanding in geometry: A
design-based research study. Educational Technology & Society, 20(3), 207–219.
Devichi, C., & Munier, V. (2013). About the concept of angle in elementary school: Misconceptions
and teaching sequences. Journal of Mathematical Behavior, 32, 1–19.
Germia, E. (2022). Investigating elementary school students’ reasoning about dynamic
angles [Doctoral dissertation, Montclair State University]. Montclair State University
Digital Commons. https://digitalcommons.montclair.edu/cgi/viewcontent.cgi?article=2123&c
ontext=etd
Hackenberg, A. J. (2005). Construction of algebraic reasoning and mathematical caring relations
[Doctoral dissertation, University of Georgia]. UGA Theses & Dissertations. https://getd.libs.
uga.edu/pdfs/hackenberg_amy_j_200508_phd.pdf
Hardison, H. L. (2018). Investigating high school students’ understandings of angle measure
[Doctoral dissertation, University of Georgia]. UGA Theses & Dissertations. http://getd.libs.
uga.edu/pdfs/hardison_hamilton_l_201805_phd.pdf
Hardison, H. L. (2019). Four attentional motions involved in the construction of angularity. In
S. Otten, A. G. Candela, Z. de Araujo, C. Haines, & C. Munter (Eds.), Proceedings of the 41st
annual meeting of the North American Chapter of the International Group for the Psychology
of Mathematics Education (pp. 360–369). University of Missouri.
Hardison, H. L. (2020). Acknowledging non-circular quantifications of angularity. In A. I. Sacristán,
J. C. Cortés-Zavala, & P. M. Ruiz-Arias (Eds.), Proceedings of the 42nd annual meeting of
the North American Chapter of the International Group for the Psychology of Mathematics
Education (pp. 671–675). https://doi.org/10.51272/pmena.42.2020-98
Hardison, H. L., & Lee, H. Y. (2019). Supporting prospective elementary teachers’ non-circular
quantifications of angularity. In S. Otten, A. G. Candela, Z. de Araujo, C. Haines, & C. Munter
(Eds.), Proceedings of the 41st annual meeting of the North American Chapter of the
International Group for the Psychology of Mathematics Education (pp. 360–369). University
of Missouri.
Hardison, H. L., Lee, H. Y., Guajardo, L. R., & Bui, M. T. T. (2022). How many angles do you
see? Prospective teachers’ assimilatory domains for angularity. In A. E. Lischka, E. B. Dyer,
R. S. Jones, J. N. Lovett, J. Strayer, & S. Drown (Eds.), Proceedings of the 44th annual meeting
of the North American Chapter of the International Group for the Psychology of Mathematics
Education (pp. 583–591). Middle Tennessee State University.
Host, E., Baynham, E., & McMaster, H. (2014). Using digital technology to see angles from dif-
ferent angles. Australian Primary Mathematics Classroom, 19(2), 18–22.
Keiser, J. M. (2000). The role of definition. Mathematics Teaching in the Middle School, 5(8),
506–511.
Keiser, J. M. (2004). Struggles with developing the concept of angle: Comparing sixth-grade stu-
dents’ discourse to the history of the angle concept. Mathematical Thinking and Learning,
6(3), 285–306.
Keiser, J. M., Klee, A., & Fitch, K. (2003). Time for action: An assessment of students’ under-
standing of angle. Mathematics Teaching in the Middle School, 9(2), 116–119.
Lehrer, R., Jenkins, M., & Osana, H. (1998). Longitudinal study of children’s reasoning about
space and geometry. In R. Lehrer & D. Chazan (Eds.), Designing learning environments for
developing understanding of geometry and space (pp. 137–167). Erlbaum.
Millsaps, G. M. (2012). How wedge you teach the unit-angle concept? Teaching Children
Mathematics, 18(6), 362–369.
Mitchelmore, M. C. (1997). Children’s informal knowledge of physical angle situations. Learning
and Instruction, 7(1), 1–19.
Mitchelmore, M. C. (1998). Young students’ concepts of turning and angle. Cognition and
Instruction, 16(3), 265–284.
Mitchelmore, M. C., & White, P. (1995). Development of the angle concept by abstraction from
situated knowledge. Paper presented at the 1995 annual meeting of the American Educational
Research Association, San Francisco.
Mitchelmore, M. C., & White, P. (1998). Development of angle concepts: A framework for
research. Mathematics Education Research Journal, 10(3), 4–27.
Mitchelmore, M. C., & White, P. (2000). Development of angle concepts by progressive abstrac-
tion and generalization. Educational Studies in Mathematics, 41(3), 209–238.
Moore, K. C. (2009). An investigation into precalculus students’ conceptions of angle measure.
In Twelfth annual Special Interest Group of the Mathematical Association of America on
Research in Undergraduate Mathematics Education (SIGMAA on RUME) conference. North
Carolina State University.
Moore, K. C. (2010). The role of quantitative reasoning in precalculus students’ learning central
concepts of trigonometry [Unpublished doctoral dissertation]. Arizona State University, Tempe.
Moore, K. C. (2012). Coherence, quantitative reasoning, and the trigonometry of students. In
R. Mayes & L. L. Hatfield (Eds.), Quantitative reasoning and mathematical modeling: A
driver for STEM integrated education and teaching in content (Vol. 2, pp. 75–92). University
of Wyoming.
Moore, K. C. (2013). Making sense by measuring arcs: A teaching experiment in angle measure.
Educational Studies in Mathematics, 83(2), 225–245.
Moore, K. C. (2014). Quantitative reasoning and the sine function: The case of Zac. Journal for
Mullins, S. B. (2020). Examining the relationship between students’ measurement schemes for
fractions and their quantifications of angularity [Doctoral dissertation, Virginia Tech]. Virginia
Tech Electronic Theses and Dissertations. https://vtechworks.lib.vt.edu/handle/10919/107124
604 H. L. Hardison
Noss, R. (1987). Children’s learning of geometrical concepts through Logo. Journal for Research
in Mathematics Education, 18(5), 343–362.
Piaget, J. (1965). The child’s conception of number. Norton.
Trans.). Norton.
Piaget, J., Inhelder, B., & Szeminska, A. (1960). The child’s conception of geometry. Routledge.
Prescott, A., Mitchelmore, M., & White, P. (2002). Student difficulties in abstracting angle con-
cepts from physical activities with concrete materials. In B. Barton, K. C. Irwin, M. Pfannkuch,
& M. O. J. Thomas (Eds.), Mathematics education in the South Pacific (Proceedings of the 25th
annual conference of the Mathematics Education Research Group of Australasia, Auckland)
(pp. 583–591). MERGA.
Sherman, L. (2017). Problem solvers: Problem: Angle detectives. Teaching Children Mathematics,
24(3), 154–157.
Silfverberg, H., & Joutsenlahti, J. (2014). Prospective teachers’ conceptions about a plane
angle and the context dependency of the conceptions. In C. Nichol, S. Oesterle, P. Liljedahl,
& D. Allan (Eds.), Proceedings of the joint meeting of PME-38 and PME-NA 36 (Vol. 5,
pp. 185–192). PME.
Simmons, M., & Cope, P. (1990). Fragile knowledge of angle in turtle geometry. Educational
Smith, C. P., King, B., & Hoyte, J. (2014). Learning angles through movement: Critical actions
for developing understandings in an embodied activity. The Journal of Mathematical Behavior,
36, 95–108.
Steffe, L. P. (1991). Operations that generate quantity. Journal of Learning and Individual
Differences, 3(1), 61–82.
Tallman, M. (2015). An examination of the effect of a secondary teacher’s image of instructional
constraints on his enacted subject matter knowledge [Unpublished doctoral dissertation].
Arizona State University.
bra. Center for Research in Mathematics & Science Education, San Diego State University.
dations of mathematics education. In O. Figueras, J. L. Cortina, S. Alatorre, T. Rojano, &
A. Sépulveda (Eds.), Proceedings of the annual meeting of the International Group for the
Psychology of Mathematics Education (Vol. 1, pp. 45–64). PME.
Thompson, P. W. (2011). Quantitative reasoning and mathematical modeling. In S. Chamberlin,
in mathematics education: Papers from a planning conference for WISDOMê (pp. 35–57).
University of Wyoming.
Vishner, G. N. (2020). An investigation of sixth-grade students’ conceptualization of angle and
angle measure: A retrospective analysis of design research study of a real-world context
[Doctoral dissertation, Syracuse University]. Syracuse University Libraries. https://surface.syr.
edu/cgi/viewcontent.cgi?article=2179&context=etd
Wilson, P. S. (1990). Understanding angles: Wedges to degrees. Arithmetic Teacher, 39(5),
294–300.
Wilson, P. S., & Adams, V. M. (1992). A dynamic way to teach angle and angle measure. Arithmetic
Teacher, 39(1), 6–13.
Chapter 21
Using Constructivism to Develop
an Agenda of Research in Stochastics
Education Research
Neil Hatfield
Stochastics education research (the combination of statistics, probability, and data

science education research) is at an intriguing developmental place. Simultaneously,
the discipline is old and young, established and emerging. As such, there are excit-
ing opportunities for researchers of all backgrounds to find a place for their work.
As a researcher in stochastics education who comes from a tradition of mathematics
education, I have found that radical constructivism provides me a foundation upon
which to set my work. In what follows, I hope to share how the tools of constructiv-
ism assist me in my research, both qualitative and quantitative.
Researching People’s Meanings
Suppose for a moment that we have a collection of 36 Oreo cookies (one regular
package in the USA) and have recorded the filling mass (in grams) for each cookie.
How would you interpret the following statement? A 95% confidence interval for
the mean filling mass of an Oreo® is (3.05, 3.31). Many people who have had some
introduction to confidence intervals will say something along the lines of “We’re
95% confident that the true mean filling mass is between 3.05 and 3.31.” However,
when asked to unpack this statement further, I observed a myriad of responses. For
example, some people say that “95% of the sample (i.e., the data) are between 3.05
and 3.31 grams.” Others might say, “The true value must be between 3.05 and 3.31.”
A few individuals will state that “If we were to repeat the entire process of making
the interval, 95% of the time, we’ll be between 3.05 and 3.31.” None of these
N. Hatfield (*)
Department of Statistics, Pennsylvania State University, University Park, PA, USA
e-mail: neil.hatfield@psu.edu
https://doi.org/10.1007/978-3-031-47386-9_21
606 N. Hatfield
interpretations are consistent with the underlying concept of confidence intervals

(see Hoekstra et al., 2014, 2018; Morey et al., 2016).
What mental operations and imagery would a person need to have in order to
have a meaning consistent with the theory that underpins frequentist confidence
intervals? How might individuals develop these meanings? These are the kinds of
questions that lie at the heart of my research. Fortunately, constructivism positions
me to not only pose such questions but to investigate them. Von Glasersfeld (1995)
highlighted that in doing conceptual analysis, a researcher is seeking to answer the
question, “What mental operations must be carried out to see the presented situation
in the particular way one is seeing it?” (p. 78). By interrogating my own thinking
and stepping beyond my own thinking into models of others (i.e., decentering, see
Chap. 9), I can build my own understanding of a particular concept and the ways in
which other individuals may understand the idea.
What thinking would an individual need to go through to come up with a produc-
tive meaning for a 95% confidence interval for the mean filling mass for Oreos® (or
for any confidence interval)? The first thing that anyone needs to do is to step back
and ask what it is, in general, they are trying to do. In this situation, I am attempting
to estimate how much filling the manufacturing process puts into Oreo cookies.
Articulating what I am trying to do gives me an underlying process that I can imag-
ine constantly running to build the population of all Oreos. I can now imagine a
process by which we get a subset of the cookies. For example, I can go to a store,
select and purchase a package of Oreos, return home, and then carefully weigh each
cookie twice in grams—once as a whole cookie and once again after carefully
scraping the filling off the wafers. At this point in time, I have had to imagine and
coordinate two processes (the generative and the sampling) and two collections (the
population and a sample). In coordinating the two collections, I need to think of
collections as objects in their own right, possessing their own attributes, such as size
or group performance, which I can measure. Further, I must imagine that both col-
lections are frozen at some moment in time so that the sample collection is a proxy
for the population collection. By positioning the sample as a proxy for the popula-
tion at that moment in time, I can further connect one of the sample’s attributes, say
group performance, to the corresponding attribute of the population. The popula-
tion’s attributes are direct consequences of the underlying generative process we
want to learn something about. Therefore, exploring how well the sample does at
accumulating filling mass (i.e., group performance) is a powerful tool at my dis-
posal. In applying the sample arithmetic mean to our sample collection, I observe a
value of 3.18 grams/cookie.
How accurate is this estimate? I can imagine that there is some measurement
error as removing the filling from each wafer is not an easy task. Further, there
might be some variability in the package I selected. In this last thought, I find the
roots of an important coordination I must make—I must coordinate this estimate
with the sampling process. If I carried out the sampling process again, I could get a
different package of Oreos whose value of the sample arithmetic mean could be
different (e.g., 2.85 g/cookie or 3.27 g/cookie). Scientists rarely report a single
value for a measurement; rather, they tend to provide an estimate and some
21 Using Constructivism to Develop an Agenda of Research in Stochastics Education… 607
measurement error. Thus, I need to revise the goal: How do I estimate the target
attribute with some measure of measurement error? This is the essential question
that Neyman (1937) sought to answer when he developed the theory behind confi-
dence intervals.
Answering this question requires focusing on a construction method, not the end
result. For example, I could plan to use two statistics (i.e., functions of data), say the
40th percentile and the 60th percentile, to act as lower and upper estimation bounds.
As I imagine carrying this estimation method out across the many samples—each
formed via a sampling process—I imagine getting a variety of different interval
estimates. Each of these intervals either successfully captures the population’s value
or doesn’t. Coordinating this thought with my repeated sampling and construction,
I can conceptualize my method’s success rate: The multiplicative comparison of
how often my method successfully captures the value I am looking for relative to
how many times I carry out the method. I can use this success rate to compare dif-
ferent methods to each other; the higher the rate, the more successful the method is.
When I imagine this rate over infinitely many repetitions of the sampling and con-
struction processes, the rate turns into a probability traditionally referred to as con-
fidence. Bringing all this imagery together allows me to interpret a 95% confidence
interval for the mean Oreo filling mass of (3.05, 3.31) as “My specific sample of
Oreos suggests that when used in conjunction with my chosen 95% confidence
method, the filling performance might be between 3.05 and 3.31.”
There are multiple key aspects that I believe an individual needs to hold in mind
to reach such a productive understanding of confidence intervals. First, they need to
be able to imagine multiple kinds of processes where the results are stochastic, not
deterministic. In other words, they need to be able to envision carrying out the pro-
cess in an essentially identical way and anticipate getting different results (see
Chap. 5). Second, they need to be able to coordinate multiple of these processes
together (see Chap. 11). For the productive way of thinking about interval estima-
tion, the individual needs to be able to coordinate (1) the underlying generative
process we’re trying to learn something about, (2) the sampling process by which
we get a set of data to act as a proxy for the population and thereby the generative
process, and (3) our construction method being applied to the results of each run of
our sampling process. Further, the individual needs to conceptualize these processes
as entities in their own right, separate from their products, but keep a coordination
between the process and products. The construction method has a success rate, but
each constructed interval does not. By focusing on the mental operations, the imag-
ery a person needs to construct, and the various coordinations necessary, I am better
positioned to think about developing tasks that can elicit responses from individuals
as they work toward constructing the above-described meaning or some other mean-
ing. From those responses, I can then work on building a model for how they might
be thinking about interval estimation. Through conceptual analysis, constructivism
better positions me to explore individuals’ meanings for stochastic concepts. While
my example here focuses on interval estimation, there are a host of stochastic con-
cepts that need investigation.
608 N. Hatfield
Researching Teaching
Like many education researchers, my work does not end with me generating models
for how individuals might think about some concept. Rather, we must also think
about how we can leverage what we’ve learned from our participants in classroom
settings. This can be challenging, as we might not be able to go from one or two
individuals to half a dozen, then to 20. We might have to scale to 45, 60, or, in some
cases, 1000+ students. This is another place where constructivism helps to shape my
research. The second tenant of radical constructivism—“the function of cognition is
adaptive and serves the organization of the experiential world”—is key for me (von
Glasersfeld, 1995, p. 18). As I think about translating research into the classroom,
my belief in constructivism predisposes me to think about creating experiences and
conversations which I believe will support students in developing meanings consis-
tent with my goals for their learning. Whether we converse with ourselves or with
others, the act of conversing is where we begin the journey to abstraction and
meaning-making.
Why is it that so many people will state, “We’re 95% confident that the true mean
filling mass is between 3.05 and 3.31”? First and foremost, they have been trained
to give such a response. The way in which interval estimation has been and contin-
ues to be taught is procedurally focused, with instructors (and AP scorers) taking
this phrasing to mean that the individual has an appropriate meaning for interval
estimation (Hoekstra et al., 2018). While there are some lessons out there that
attempt to highlight the sample-to-sample variability of interval estimates, they still
miss many of the key elements I identified earlier. Thus, what would a sequence of
activities look like that could help individuals build a productive meaning for inter-
val estimation? This was a challenge that my colleagues and I undertook in our
forthcoming entry in the Interweaving Mathematics Pedagogy and Content for
Teaching series (Saldanha et al., 2023). As part of our activity sequence, I developed
a Shiny app (https://psu-eberly.shinyapps.io/Building_Intervals/) that interested
readers may explore. The goal of the sequence is to problematize relying upon point
estimation and then to get students to engage in developing and evaluating their own
methods for interval construction. We do not intend for the sequence to result in
students re-creating the generally accepted central limit theorem-based methods
they might have previously encountered. Rather, we strive to get them to think about
the construction process and make judgments about that process rather than the end
result (i.e., the interval estimate). I contend that we could not have conceptualized
this activity (nor the app) without drawing upon our backgrounds as constructivists.
Around this activity sequence, I have two current lines of inquiry. The first is to
explore how this activity sequence impacts the meanings that undergraduate stu-
dents develop for confidence intervals in an upper-division statistics course. The
second revolves around the sense that PhD statistics graduate students make of the
activity sequence as they engage with it as a teaching tool. Another possible research
area is to support high school teachers who teach statistics to examine their mean-
ings for confidence intervals.
A Radical Constructivist Statistician
The title of this section may seem like an oxymoron as quantitative research often
tends to come from a position of [post] positivism. There can be utility in thinking
about the model we generate in quantitative work as reflecting some ontological
reality. That is to say, we can find a certain power in saying that, since we’ve mea-
sured this thing, then that is the measurement of that thing. After all, our experiences
in measuring objects (e.g., ourselves) support this way of thinking more often than
not. As I think back upon my many experiences as a student in statistics courses
(undergraduate, Master’s, and PhD level), this positivist bent lurked throughout.
The de facto reign of positivism has an effect of painting quantitative work as being
objective. I am reminded of one of Tukey’s “badmandments”: “the one and only
proper use of statistics is for sanctification” (Tukey, 1986, p. 199).1 This particular
badmandment (among others) reflects an attitude or belief that Tukey saw as being
taught explicitly/implicitly and that would lead to bad science. Sadly, I believe that
Tukey’s sarcastic/satirical approach to communicating the dangers of such beliefs
worked against him.
Quantitative researchers need to critically examine the methods that they use and
how they go about using them. Researchers drawing upon critical theories have
made progress on this front with the development of QuantCrit (see Gillborn et al.,
2018). I would like to share how constructivism has helped to shape my work as a
statistician. I believe that constructivism, while long held as a foundational theory
in qualitative work, also has a place in quantitative work.
The reflexive nature of constructivism prompts us to apply its tenets to ourselves
and our understandings just as much as we apply it to build our models of others’
thinking. I take von Glasersfeld’s (1995) second tenet as a reminder that I’m using
quantitative methods to help me make sense of what I am experiencing. Part of my
job as a statistician is to critically examine my own work and then to present that
work in such a way that supports my consumers in constructing their own under-
standings of the data in a way that is, hopefully, compatible with mine (i.e., the
notion of intersubjectivity). Constructivism helps to keep me grounded and centered
in my quantitative work. Data do not collect themselves, do not clean themselves,
do not speak for themselves, and do not place themselves into a particular model.
Throughout every step of quantitative work, there is a statistician/data analyst/data
scientist (i.e., a person) who is making thousands of decisions that can (and will)
ripple throughout the analysis. Quantitative methods are not objective—each
reflects the decisions that we have made as we actively build our understanding of
the data (including how we assimilated that data to our existing schemes) and our
understanding of those methods.
A second place that constructivism has impacted my quantitative research is
through Thompson’s theory of quantitative reasoning (Thompson, 1990, 2011; also
see Chap. 12). Thompson’s work into the ways in which a person might reason
1
For those unfamiliar with Tukey, a badmandment was like a negative commandment.
610 N. Hatfield
about measurable attributes is truly at the heart of statistics. Lurking throughout my

previous conceptual analysis for interval estimation is quantitative reasoning.
Confidence was my measure of how successful my chosen construction method was
when I applied that method infinitely many times. Further, I measured the perfor-
mance of the filling process with the sample arithmetic mean and reported the value
as 3.18 grams/cookie rather than 3.18 grams. This choice of unit reflects what I’m
taking as the object whose attribute I’m attempting to measure. I’m not attempting
to measure some cookies, but the underlying process, complete with natural fluctua-
tions. Thus, my data collection is the entity with an attribute called group perfor-
mance that I can measure and use as a proxy for the generative process’s performance.
Attending to what I’m conceptualizing as the object I want to measure—and how I
conceptualize this object—is a critical and often overlooked aspect of quantitative
research. I strive in my quantitative work to bring quantitative reasoning to bear and
work from a constructivist position.
Final Thoughts
Stochastics education inherits all the challenges from mathematics education and
adds on additional challenges, such as computational thinking and, importantly,
stochasticity. While I do not believe that constructivism holds all the answers, it is a
tool kit we can and should make use of. Constructivism has had a profound impact
on my research and how I mentor students (undergraduate and graduate) and faculty
in doing research (within and beyond education). I have found great utility in theory
not only for investigating how individuals might understand ideas in statistics and
the development of teaching materials but also as a critical component for how I
approach quantitative work as a statistician.
References
Gillborn, D., Warmington, P., & Demack, S. (2018). QuantCrit: Education, policy, ‘Big Data’ and
principles for a critical race theory of statistics. Race Ethnicity and Education, 21(2), 158–179.
https://doi.org/10.1080/13613324.2017.1377417
Hoekstra, R., Morey, R. D., Rouder, J. N., & Wagenmakers, E.-J. (2014). Robust misinterpreta-
tion of confidence intervals. Psychonomic Bulletin & Review, 21(5), 1157–1164. https://doi.
org/10.3758/s13423-013-0572-3
Hoekstra, R., Morey, R. D., & Wagenmakers, E.-J. (2018). Improving the interpretation of confi-
dence and credible intervals. In Proceedings of the tenth international conference on teaching
statistics (p. 6).
Morey, R. D., Hoekstra, R., Rouder, J. N., Lee, M. D., & Wagenmakers, E.-J. (2016). The fallacy
of placing confidence in confidence intervals. Psychonomic Bulletin & Review, 23(1), 103–123.
https://doi.org/10.3758/s13423-015-0947-8
Neyman, J. (1937). Outline of a theory of statistical estimation based on the classical theory of
probability. Philosophical Transactions of the Royal Society A: Mathematical, Physical and
Engineering Sciences, 236(767), 333–380. https://doi.org/10.1098/rsta.1937.0005
Saldanha, L., Hatfield, N. J., Chernoff, E. J., & Primi, C. (2023). The learning and teaching of sta-
tistics and probability: A perspective rooted in quantitative reasoning and conceptual coher-
ence. Routledge.
bra. http://patthompson.net/PDFversions/1990TheoryQuant.pdf
Thompson, P. W. (2011). Quantitative reasoning and mathematical modeling. In L. L. Hatfield,
S. Chamberlain, & S. Belbase (Eds.), New perspectives and directions for collaborative
research in mathematics education (Vol. 1, pp. 33–57). University of Wyoming. http://bit.
ly/15II27f
Tukey, J. W. (1986). Data analysis and behavioral science or learning to bear the quantitative man’s
burden by shunning badmandments. In L. V. Jones (Ed.), The collected works of John W. Tukey
(Vol. 3, pp. 187–389). Wadsworth & Brooks/Cole.
von Glasersfeld, E. (1995). Radical constructivism: A way of knowing and learning. Routledge.
Index
A C
Abstraction, 31, 53, 91, 169, 241, 293, 369, Carlson, M., 264, 271, 430, 434, 435, 437,
374, 426, 459, 526, 542, 608 438, 440, 448, 465
Accommodation Cartesian plane, 278
functional accommodation, 30, 31, 33–36, Chunky change, 429
48, 62, 67–68, 72, 78, 83–86, 134, 150 Circumference, 245, 450, 596, 597
metamorphic accommodation, 25, 30, 33, Clinical interviews, 17, 421, 513, 514, 516,
37, 83, 85, 86, 145, 150, 155 524, 586
Additive reasoning, 226 Commutativity, 213, 226
Algebra Complex numbers, 213, 459
abstract algebra, 210, 213, 219, 230, Composite unit, 30–34, 36, 37, 55–57, 60, 61,
361, 460 63, 67, 70–72, 74, 75, 77, 78, 81, 84,
linear algebra, 262 215, 372–382, 385, 386, 394, 397,
Angles 399–401, 404, 407, 485, 486, 488
angle measure, 140, 242, 243, 245, 252, Concept, 5, 12, 49, 116, 135, 169, 211, 239,
259, 260, 272, 300, 301, 317, 461, 598 341, 373, 415, 456, 481, 514, 547, 567,
subtended angles, 247, 317 575, 590, 606
APOS theory, 264, 266, 267, 461 Conceptual analysis, 17–22, 26, 30, 32, 79,
Area, 13, 14, 18, 19, 90, 96, 98, 108–110, 123, 164, 165, 267, 281, 387, 599, 606,
124, 145, 170–174, 181, 182, 205, 244, 607, 610
260, 283, 340, 363, 405, 414, 416, 432, Concrete operational stage, 234
438, 440, 451, 465, 479, 558, 575, Continuous functions, 465
582, 608 Continuum, 427, 428, 520
Assimilation Coordination, 5, 52, 93, 156, 176, 210, 246,
generalizing assimilation, 256, 257 293, 361, 372, 428, 450, 486, 517, 574,
Associativity, 213, 217–221, 225–227, 235, 590, 606
453, 460 Cosine function, 450
Axioms, 347, 351, 449–452, 457, 458, Counting, 6, 15, 52, 91, 192, 210, 273, 371,
547, 557 416, 456, 483, 512, 574, 599
Covariation, 90, 102, 103, 184, 188, 205, 263,
414, 415, 428, 431, 440
B Covariational reasoning, 98–100, 102, 106,
Binary operations, 213–215, 219, 220, 115, 116, 121–123, 136, 187, 264,
225, 232 270–272, 413–440, 448, 465, 466
© The Editor(s) (if applicable) and The Author(s), under exclusive license to 613
Springer Nature Switzerland AG 2024
https://doi.org/10.1007/978-3-031-47386-9
614 Index
D Formal operational stage, 235

Decentering, 13, 279, 461, 462, 469, 480, 502, Fractions, 7, 18, 54, 133, 211, 268, 317, 373,
549, 606 422, 459, 482, 515, 547, 566, 581,
Decimals, 422, 521 590, 599
Derivatives, 179, 432 Functions, see Continuous functions; Cosine
Dimension, 91, 221, 273, 274, 451, 587, function; Logarithmic function; Sine
590, 591 function; Tangent function
Discrete change, discrete variable, 415
Disembedding, 379, 381–385, 390, 401, 406,
417, 431, 467 G
Displacements, 139, 140, 163, 210, 217, 222, Generalization, 90, 159, 162, 164, 172,
228, 273, 274, 278, 417, 426, 431, 174–176, 178, 198, 202, 204, 205, 250,
437, 588 261, 265–267, 281, 341, 343, 349, 350,
Distributing, 372, 381 353, 356–358, 364, 368, 389, 421, 529
Division, 133, 138, 154, 155, 157, 165, 219, Generalized number sequence (GNS), 374,
268, 356, 372, 373, 395, 397, 404, 405, 378, 381, 405
422, 486, 488, 489 Genetic decomposition, 265, 267
DNR-based instruction, 260, 267, 268 Geometry
Double counting, 32, 376 Euclidean geometry, 5, 228, 449, 450, 452,
455, 457, 458
projective geometry, 228
E transformational, 227, 228
Egocentrism, 291, 325 Glasersfeld, E. von, 4, 14, 47, 170, 210, 239,
Embodied cognition, 284, 459 428, 429, 447, 448, 455, 459, 541, 553,
Empirical abstraction 556, 585, 606
pseudo-empirical abstraction, 31, 143, 153, Graphs, 100–103, 105–108, 111, 113, 114,
164, 179–182, 186, 188, 195, 199, 120–122, 124, 136, 138, 139,
202–205, 244, 246, 248, 251, 257, 259, 159–163, 166, 181, 186–192,
274, 278, 459 203–205, 262, 263, 278, 436–438,
Encapsulation, 265–267 456, 457, 470, 527, 585
Epistemic subject, epistemic student, 102, 374, Gross quantity, 416, 419, 435, 599
382, 462, 464, 469, 479, 480, 482–490, Group, 14, 110, 130, 174, 210, 273, 340, 374,
500–503, 549 422, 459, 477, 515, 541, 573, 599, 606
Equilibrium, 22, 240, 256, 257, 355, 451, 546 Grouping, 18, 202, 210, 217–223, 225, 226,
Erlangen program, 227–229, 233 230–232, 234, 356, 362, 459, 578
Euclid, 19, 449–451, 457, 464
Explicitly nested number sequence (ENS), 22,
33–36, 38, 39, 372, 374, 377, 381, 405, H
407, 464, 467 Hindu-Arabic numerals, 532
Exponents, 307, 308 Historicocultural, see Sociocultural
Extension, 95, 111, 123, 124, 181, 210, 213,
229, 233, 245, 267, 347–351, 357, 359,
366–367, 380, 381, 454, 463, 602 I
Extensive quantity, 416, 417, 419, 422, 424, i, see Imaginary numbers
432, 599 Identity element, 213, 216, 217, 219, 221, 225
Image, 21, 48, 92, 129, 173, 251, 290, 347,
377, 417, 453, 481, 529, 576, 589, 596
F Imaginary numbers, see Complex numbers
Figurative thought, 94, 97, 98, 108, 110, 113, Infinity, 37
261–263, 456, 461 Inhelder, B., 5, 15, 16, 21, 90, 91, 132, 158,
Formal logic, 219, 236, 340, 342, 344, 348, 209, 210, 215, 228, 232, 241, 256, 340,
351, 357, 358, 363–365, 367, 463, 356, 357, 359, 362–365, 368, 417, 427,
470, 548 428, 458, 463, 575, 585, 586, 588, 589
Index 615
Initial number sequence (INS), 22, 24–26, N

30–33, 36, 38, 39, 372, 401, 404, 405, Negative numbers, see Integers
407, 464, 466–468, 483, 577 Neuroscience, 229, 526, 581
Integers, 139, 140, 214, 215, 219, 222, 226, Newton, I., 559
262, 347, 366, 373, 422, 424, 425, Noether, E., 227, 229, 236
544, 547
Integrals, 259, 454, 465, 479
Intensive quantity, 271, 416, 417, 419, O
422–424, 434 Operative thought, 90–115, 121, 123, 163,
Intention, 14, 31, 90, 116, 132, 133, 246, 261, 277, 455–457, 459, 461, 469,
203, 259, 267, 347, 366, 435, 522, 533
455, 463 Orthogonal, see Perpendicular
Interaction, 6–8, 16, 17, 20, 21, 23, 26, 135,
150, 158, 185, 187, 191, 240, 241,
257, 267, 282, 361, 372, 396–399, P
403, 440, 448, 453–455, 459, 465, Parallel
475, 479–501, 503, 512, 543, 547, parallelogram, 108
556, 568, 589, 601 Pascal, B., 209, 227, 450
Interiorization (action/concept), 25, 70, 71, Pascal’s triangle, see Pingala’s triangle
265, 267, 375, 454, 456 Patterns, 19, 22, 24, 25, 27, 28, 31, 135, 143,
Internalization, 453, 454 153, 172, 174, 177, 179, 181, 182, 199,
Inverses, 212, 227, 231, 405, 460, 488 202, 203, 205, 209, 218, 254, 255, 265,
Isometries, 212, 214, 215, 228, 229, 233 270, 271, 278, 341, 345, 348, 351, 355,
Iterating, 99, 212, 218, 221, 223–227, 229, 358, 375, 376, 404, 425, 466, 496–498,
234, 235, 250, 379–383, 385, 388, 501, 503, 531–533, 556, 577–580, 586
423–425, 460, 578, 600 Pedagogical content knowledge (PCK), 241,
275, 279–282
Perception
K perceptual material, 70, 95, 97, 179–182,
Kant, I., 4, 5, 169, 170, 175, 209, 186, 188, 262, 263, 456, 457, 577, 580
228, 234 Perpendicular, 542
Key developmental understanding (KDU), Perturbation, 22, 25, 100, 122, 124, 241, 256,
297–298, 300, 315 421, 425, 455, 582
Klein, F., 211, 215, 227–231, 233, 235, 236 Pi, 225
Piaget, J., 3, 11, 47, 90, 130, 170, 209, 239,
291, 339, 413, 448, 482, 512, 542, 568,
L 573, 585, 586, 599
Limits, 12, 152–154, 230, 268, 344, 438, 511, Pingala’s triangle, 450
523, 533, 589, 608 Plato
Linear algebra, see Algebra Platonism, 5
Logarithmic function, 306–308, 311 Plato’s rules for construction, 450
Logic, 3, 12, 124, 144, 183, 209, 239, 339, Polar coordinates, 262–264
453, 525, 544, 575 Postulates, see Axioms
Logical implication, 343, 463 Pre-operational stage, 23, 254, 291
Prime numbers
Fermat primes, 450
M Projecting, 25, 150, 156, 162, 182, 269, 345,
Mathematical knowledge for teaching (MKT), 429, 467
293, 297–298, 315, 462 Proportion
Mathematical meaning for teaching (MMT), proportional reasoning, 18, 359, 373, 398,
317, 319, 332 406, 407, 417, 486, 488, 518, 590
Matrices, 200, 217, 230 proportional relationship, 263, 276, 304,
Multiplicative reasoning, 55, 57, 70–72, 226, 361, 365
303, 406, 464, 469, 485, 499 Psychogenetic, 180, 447–451, 454, 457, 462
616 Index
Q Sensori-motor activity (sensorimotor, sensory

Quantitative reasoning motor, sensory-motor), 4, 5, 19, 24, 52,
quantitative variational reasoning, 432, 435 53, 93–98, 100, 175–179, 181, 185,
Quantity 187, 188, 195, 199, 202, 209, 211,
covarying quantities, 97, 103, 186, 188, 254–258, 262, 263, 265, 283, 291,
190, 192, 204, 205, 271, 272, 275, 276, 453–456, 459, 460, 467, 529–531, 533,
415, 436 534, 587–589
directed quantities, 215, 222, 226 Shape thinking
emergent, 95, 102, 103, 106, 136, 456
static, 102–104, 106, 108, 436, 456
R Sine function, 165, 166, 450
Radian, 117, 243, 244, 263, 600 Slope, 100–102, 136, 139, 262, 273, 274,
Rate of change, 97, 101, 102, 132–133, 139, 278, 354
144, 145, 205, 262–264, 275, 276, 278, Smooth change, 303, 429–431
434, 435, 479 Sociocultural, 284, 448, 449, 451, 455, 459, 569
Rational numbers, see Fractions Splitting
Real numbers, 213 splitting group, 211, 226, 229, 235,
Rectangular coordinates, 588, 589 460, 468
Reflected abstraction, 143–145, 179, 180, 183, splitting loope, 211, 212, 223–227, 235
184, 189, 191, 192, 198, 200–205, Stage (of development), 254, 291
239–285, 459–463, 468, 469, 529, 549 Steffe, L.P., 5, 13, 93, 131, 175, 210, 277, 413,
Reflective abstraction, reflecting abstraction, 455, 479, 517, 548, 577, 590, 599
7, 19, 23–27, 33, 37, 53, 83, 86, 142, Subitizing, 573–583
143, 145, 155, 159, 161, 166, 169–205, Symbols, 14, 37, 94, 95, 103, 106, 108–110,
211, 213, 242, 244, 246–252, 255, 257, 135, 141, 165, 183, 184, 244, 251, 254,
261, 264–267, 269, 273–275, 277, 278, 255, 257, 340, 343, 345, 362, 389, 439,
284, 375, 418, 427, 447, 448, 454, 456, 452, 454–458, 460, 468, 469, 520, 522,
459, 460, 467, 468, 498, 534 528, 529, 531–534
Reflex, 209, 257, 448, 453–455 Symmetry, see Reflecting
Re-presentation (representation), 16, 21, 23,
24, 27, 73, 91, 94, 95, 98, 101,
106–112, 115, 117, 119, 130, 138, 146, T
156, 166, 169, 170, 174, 181–184, 186, Tacitly nested number sequence (TNS), 22,
204, 205, 246–251, 253–255, 257, 258, 32, 33, 36, 39, 372–377, 381, 399, 401,
261, 262, 265, 270, 271, 291–293, 306, 405, 407
319, 331, 341, 354, 360, 398, 491, 492, Teaching experiments, 7, 15–18, 20–27, 30,
529, 531, 533, 566, 577, 580, 585, 588 32, 34–39, 54, 55, 62, 108, 184, 202,
Reversal, 230, 232, 265–267, 440, 467 275–277, 372, 373, 382, 387, 396, 430,
Rotating, 93, 113, 212, 215, 247, 252, 437, 464, 466, 467, 469, 479, 483, 484,
259, 430 496, 524, 550, 591, 595, 600, 601
Thompson, P., 15, 17, 19, 34, 160, 285, 367,
465, 549
S Translating, see Displacements
Scaling continuous variation, 430, 431
Scheme, 7, 20, 47, 93, 129, 175, 210, 241,
292, 342, 372, 452, 481, 517, 546, 585, U
599, 609 Unit circle, 90, 450
Second-order model, 8, 13, 20, 24, 26, 28, 34, Unitizing, 19–22, 24, 99, 218, 371, 375,
84, 86, 115, 204, 281, 373, 381, 395, 381, 431
396, 398, 408, 413, 438, 462, 464, Units
468–470, 475–503, 514, 516, 527, 528, composite units, 30–34, 36, 37, 55–57, 60,
591, 599 61, 67, 70–72, 74, 75, 77, 81, 84, 215,
Semiotics, 183, 184, 244, 251, 253–255, 458, 372–382, 386, 394, 399–401, 404, 407,
460, 461, 468, 469, 529–531, 533 485, 486, 488
Index 617
unit fractions, 7, 37, 211, 224, 225, 381, W

385–389, 394, 395, 425 Ways of thinking, 6, 8, 102, 108, 134, 135,
unknown units, 60, 390, 393–395 256, 260, 264, 265, 268–271, 278, 280,
Units coordination, 31, 33, 48, 56, 69, 70, 78, 282, 407, 408, 420, 421, 430, 436, 439,
79, 84, 85, 371–408, 447, 463–468, 477, 479, 480, 484–490
486, 487, 581–583, 586, 590, 591, 601 Ways of understanding, 19, 165, 264,
268–270, 272, 280–282, 413, 418, 420,
436, 438
V Working memory, 518, 526, 527, 577, 582
Variable, 99, 109, 197, 229, 254, 262, 272,
307, 323, 341, 342, 353, 354, 356–358,
362–364, 389, 415, 428, 430, 432, 434, Z
435, 522, 544 Zero, 112, 139, 145, 152, 156, 172, 173, 195,
Variational reasoning, see Covariational 196, 436, 574, 576
reasoning Zone of potential construction (ZPC), 233,
Vectors, 192, 193, 214 381, 382, 600

Paul Christian Dawkins, Amy J. Hackenberg, Anderson Norton - Piaget's Genetic Epistemology For Mathematics Education Research-Springer (2024)

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

Paul Christian Dawkins, Amy J. Hackenberg, Anderson Norton - Piaget's Genetic Epistemology For Mathematics Education Research-Springer (2024)

Uploaded by

Copyright:

Available Formats

Research in Mathematics Education

Series Editors: Jinfa Cai · James A. Middleton

Paul Christian Dawkins

Piaget’s Genetic Epistemology

ISSN 2570-4729 ISSN 2570-4737 (electronic)

Paper in this product is recyclable.

Part I Introduction to Piaget’s Genetic Epistemology and the

Part II Key Constructs from Genetic Epistemology Being Used

9 The Construct of Decentering in Research on

Part III Commentaries on Genetic Epistemology and Its Use

Part IV Using Constructs from Genetic Epistemology to Develop

21 Using Constructivism to Develop an Agenda of Research

Sinem Bas-Ader is an assistant professor in the Department of Mathematics and

Steven Boyce is an associate professor of Mathematics Education at Portland

Cameron Byerley is an assistant professor in the Department of Mathematics,

Marilyn P. Carlson is a professor of Mathematics Education studying the teach-

Paul Christian Dawkins is a professor of Mathematics Education in the

Amy Ellis is a professor of Mathematics Education at the University of Georgia.

Andrew M. Gatza is a former middle school mathematics teacher and completed

Hamilton L. Hardison is an assistant professor of Mathematics at Texas State

Neil Hatfield is an assistant research professor in the Department of Statistics at

Jessica H. Hunt began her career in education as a middle school mathematics

Elise Lockwood is a professor in the Department of Mathematics at Oregon State

programming can be leveraged to support students’ combinatorial thinking and

Beth L. MacDonald is an associate professor of Early Childhood Mathematics

Percival Matthews is an associate professor of Educational Psychology at the

Anderson Norton is a professor of Mathematics Education in the Department of

Alan O’Bryan has worked for almost 20 years as a mathematics education

Teo Paoletti is an associate professor specializing in Mathematics Education in

Abby Rocha completed her PhD in Mathematics Education at Arizona State

Serife Sevinc (Şerife Sevinç; in native language) is an associate professor at the

Leslie P. Steffe is a research professor emeritus of Mathematics Education,

Interdisciplinary Research on Number (IRON) with the epistemologist, Ernst von

Irma E. Stevens is an assistant professor in the Department of Mathematics and

Michael A. Tallman received his BSc and MA in Mathematics from the

Halil I. Tasova is an assistant professor of Mathematics Education at California

Patrick W. Thompson is professor emeritus of Mathematics Education in the

Alexandria Viegut is an assistant professor of Psychology at University of

Paul Christian Dawkins, Amy J. Hackenberg, and Andy Norton

Piaget began to study children’s psychological development as a means of investi-

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 3

Consider Kant’s (1781) famous statement: “The concept of Euclidean space is by

Why Is Piaget’s Genetic Epistemology Useful?

For Glasersfeld (1995), a way of knowing is valuable to the extent it is useful.

Fig. 1.1 3/5 as three of 1/5

Organization of the Book

We mention these models of students’ mathematics because they portray how

Piaget and Modern Mathematics

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 11

stages of cognitive development as a rationale for the elementary programs because

Development vs. Learning

Although he could have oriented his papers as elaborations of genetic epistemology,

In contrast, he explained learning as presenting the opposite case.

The curriculum developers considered Piaget to be an observer rather than a teacher,

The lack of appreciation for genetic epistemology was addressed by Eleanor

Developmental vs. Mathematical Structure

Sinclair used a metaphor to explain what she regarded as difficulties in trying to

Preludes to IRON (Interdisciplinary Research on Number)

A Change in Research Program

I was working as a realist and an empiricist in my attempts to apply Piaget’s devel-

Interdisciplinary Research on Number (IRON)

Original Members of IRON

During the time we engaged in IRON, I knew Glasersfeld best as a coinvestigator,

Part I Introduction to Piaget’s Genetic Epistemology and the

Part II Key Constructs from Genetic Epistemology Being Used

9 The Construct of Decentering in Research on

Part III Commentaries on Genetic Epistemology and Its Use

Part IV Using Constructs from Genetic Epistemology to Develop

21 Using Constructivism to Develop an Agenda of Research

Why Is Piaget’s Genetic Epistemology Useful?

Organization of the Book

Piaget and Modern Mathematics

Development vs. Learning

Developmental vs. Mathematical Structure

Preludes to IRON (Interdisciplinary Research on Number)

A Change in Research Program

Interdisciplinary Research on Number (IRON)

Original Members of IRON

Piaget’s Epistemic Child

The First Type

Attentional Moments and the Unitizing Operation

I recall Glasersfeld commenting that extracting attentional patterns from sensory-

The Second Type

The Initial Number Sequence as a Scheme

Operationalizing Reflective Abstraction

Constructing the Initial Number Sequence

Further Operationalizing Reflective Abstraction

The Perceptual Stage

The Figurative Stage

Percent of First-Grade Children in the Perceptual Stage

Further Criteria of Learning Stages

Modifications in the Initial Number Sequence

Modifications in the Explicitly Nested Number Sequence

The Fractions Teaching Experiment

The Equipartitioning Scheme: A Functional Accommodation

The Partitive Fraction Scheme

The Iterative Fraction Scheme

rief Overview of Glasersfeld’s Radical

iaget’s Development and Glasersfeld’s Three-Part Definition

Schemes from Reflexes

Sensory Motor and Conceptual Schemes

mpirical Example of Glasersfeld’s Three-Part Definition

uilding on Glasersfeld’s Definition of Schemes: Steffe’s

mpirical Example of Nuances in Steffe’s Definition of Scheme:

Using Schemes to Investigate Learning

Assimilation, Perturbation, and Accommodation

mpirical Example of Assimilation: Carlos’s Solution

mpirical Example of Perturbation: Carlos’s Solution