J RES MATH EDUC v54 No 4 pp240-259

Journal for
Research in
Mathematics
Education
The Journal for Research in Mathematics Education is an official journal of the National Council
of Teachers of Mathematics (NCTM). JRME is the premier research journal in mathematics
education and is devoted to the interests of teachers and researchers at all levels—preschool
through college.
ARTICLE TITLE:
Unit Transformation Graphs: Modeling Students’ Mathematics in Meeting the Cognitive Demands of
Fractions Multiplication Tasks
AUTHOR NAMES:
Norton, Anderson; Ulrich, Catherine; and Kerrigan, Sarah
DIGITAL OBJECT IDENTIFIER: VOLUME: ISSUE NUMBER:
10.5951/jresematheduc-2021-0031 54 4
Mission Statement
The National Council of Teachers of Mathematics
advocates for high-quality mathematics teaching
and learning for each and every student.
CONTACT: jrme@nctm.org
Copyright © 2023 by The National Council of Teachers of Mathematics, Inc. www.nctm.org. All rights reserved. This material
may not be copied or distributed electronically or in any other format without written permission from NCTM.
Brought to you by Stanford Univ - Green Lib | Unauthenticated | Downloaded 02/08/24 04:43 PM UTC
Journal for Research in Mathematics Education
2023, Vol. 54, No. 4, 240–259
Unit Transformation Graphs: Modeling Students’ Mathematics in

Meeting the Cognitive Demands of Fractions Multiplication Tasks
Anderson Norton and Catherine Ulrich
Virginia Tech
Sarah Kerrigan
George Fox University
This article introduces unit transformation graphs (UTGs) as a tool for diagramming the ways students use sequences of mental
actions to solve mathematical tasks. We report findings from a study in which we identified patterns in the ways preservice elemen-
tary school teachers relied on working memory to coordinate mental actions when operating in fraction multiplication settings.
UTGs account for the constraint of working memory in sequencing mental actions to solve mathematical tasks. They also explain
the power of units coordinating structures in offloading demands on working memory. At the end of the article, we consider some
of the research implications for these findings—specifically, ways that UTGs can lend explanatory and illustrative power to analyses
of students’ mathematics.
In 2006, Educational Psychologist published a provocative article titled “Why Minimal Guidance During Instruction
Does Not Work” (Kirschner et al., 2006). The authors rebuked inquiry-oriented, problem-based, and constructivist
approaches to mathematics education, claiming that such approaches do not work because they do not account for the
limitations of working memory. Kirschner et al. argued, in particular, that the cognitive demands of problem solving
compete with the cognitive demand associated with structuring new mathematical knowledge.
The article provoked several published responses. For instance, Hmelo-Silver et al. (2007) defended problem-based and
inquiry-oriented learning by highlighting the use of scaffolding to moderate the cognitive demands of mathematical tasks
and by citing evidence of its effectiveness in supporting student learning. However, the broader critique of mathematics
education by Kirschner et al. (2006) still stands: Few models of mathematical teaching and learning explicitly account for
well-documented limitations in students’ working memory and, therefore, may overlook the cognitive demands of math-
ematical tasks intended to promote mathematical development. Conversely, little research from cognitive psychology
problematizes the nature of mathematics itself (Norton & Nurnberger-Haag, 2018).
Various frameworks used in mathematics education research characterize mathematical activity differently, but most
point to students’ actions as the basic atoms for mathematical knowledge (Alibali & DiRusso, 1999; Confrey & Kazak,
2006; Steffe et al., 1996). Sociocultural and emergent perspectives frame mathematics as a form of interaction in which
mathematicians participate, defining mathematics by the kinds of activities in which mathematicians engage (Cobb et al.,
1996). Researchers of embodied cognition go so far as to define mathematics—and all forms of knowledge—as sensorim-
otor activity (Nemirovsky & Ferrara, 2009; Núñez et al., 1999). Whether sensorimotor, interpersonal, or abstract, scientific
consensus now holds that mathematical learning is based in action.
Here, we adopt a Piagetian (e.g., Beth & Piaget, 1966) perspective and characterize mathematical actions as mental
actions that can be composed with other mental actions and reversed. Prior work from a Piagetian perspective has identi-
fied students’ construction and transformation of units as central to their development of number—extended to rational
numbers and even algebraic reasoning as generalized arithmetic (Hackenberg & Lee, 2015; Steffe & Olive, 2010). Our
work was motivated by a desire to explicitly identify mental actions that undergird the construction and transformation of
units in the context of fractions. For example, students may construct 1/7 as a unit that has a one-to-seven relationship with
a whole unit. This relationship may be established by the mental action of partitioning a continuous whole into seven equal
parts. Conversely, the relationship could be reversed by iterating a 1/7 part seven times to reproduce the whole. Thus,
partitioning and iterating constitute reversible and composable mental actions that can be used to construct a 1/7 unit and
transform it back into the whole unit (Wilkins & Norton, 2011).
The purpose of this article is to introduce unit transformation graphs (UTGs) as tools for modeling students’ mathematics.
UTGs are directed graphs (in the mathematical sense) that explicitly identify the sequences of actions and units coordinating
structures students use to solve mathematical tasks. We pose two questions related to the utility of UTGs in building models
of students’ mathematics:
This work was funded by a grant from the Adaptive Brain & Behavior Destination Area at Virginia Tech. Along with
the first and second authors, Martha Ann Bell served as a principal investigator on that grant and also provided use of her
Cognition, Affect, and Psychophysiology (CAP) Lab. Two undergraduate researchers, Katherine Wilde and Emily Wright,
contributed to initial data analysis.
Anderson Norton et al.

SA-NCTM-JRME230005.indd 240
The National Council of Teachers of Mathematics, Inc. www.nctm.org All
rights reserved.
5/2/2023 12:12:41 PM
Anderson Norton et al. 241
1. How can we use UTGs to predict and explain differing demands among fractions multiplication tasks? For example,
how can we use them to explain the asymmetric difficulty of taking a unit fraction of a non-unit fraction versus
taking a non-unit fraction of a unit fraction?
2. How can UTGs explain the affordances units coordination structures provide for students in managing the con-
straints of working memory and meeting the cognitive demands of fractions tasks?
The first question serves as a specific test of the utility of UTGs in illustrating sequences of mental actions students use
to solve mathematical tasks. The second question considers the broader utility of the framework in bringing together the
psychological construct of working memory with the construct of units coordination from mathematics education.
We present examples of UTGs in the context of 12 preservice teachers (PSTs) solving three fractions tasks. Results
explain the roles of working memory, mental actions, and units coordinating structures in meeting those demands. In
particular, students rely on working memory to sequence mental actions in service of a goal, and their available units
coordinating structures may offload demands on that limited capacity. We close the article by considering potential research
implications of our models and by returning to address related concerns raised by Kirschner et al. (2006).
Framework for Building Unit Transformation Graphs

In line with their various frameworks, mathematics educators have begun to account for students’ actions when building
models of their mathematical reasoning. Piaget’s schemes construct provided an early example (Beth & Piaget, 1966); more
recent examples included activity-effect relations (Tzur & Simon, 2004) and the construction of concepts as described within
Learning Through Activity framework (Simon, Placa, et al., 2018). Similar to those models, UTGs explicitly identify the
mental actions students use to construct and transform units, but have two distinguishing features. First, they account for
the role of working memory in sequencing those actions. Second, they account for the role of units coordinating structures
in reducing demands on working memory. Here, we describe how we operationalize existing theory to build UTGs.
Mental Actions for Constructing and Transforming Units

Although mathematics education researchers generally agree that mathematics is action based, they diverge in their
characterizations of mathematical actions—even among researchers who focus on individual cognition, as we do. Both
constructivists and researchers of embodied cognition take sensorimotor actions as the starting point. Some embodied
perspectives also account for internalized or imagined activity (e.g., Nemirovsky & Ferrara, 2009), as constructivists do.
The point of departure occurs when constructivists consider the emergence of structures that coordinate those internalized
actions as operations.
Beth and Piaget (1966) summarized, “An operation is an action which can be interiorised, is reversible, and always
dependent on other operations. . . . Logicomathematical experience is . . . only concerned with actions which will subse-
quently be transformed into operations” (p. 234). Following Piaget, we characterize mathematical actions (operations) as
mental actions that are potentially reversible and composable. We are particularly concerned with operations students use
to construct and transform units, where unit refers to a delimited instantiation of a quantity that can be used to compare
or measure other instantiations of the same quantity. For example, a student may construct a unit by isolating a collection
of items, treating them as identical, and taking them as a whole—an operation called unitizing (Steffe, 1991). Students may
also unitize a continuous span of attention, whether it be time, length, area, or volume. Once constructed, units can be
transformed into other units: students may iterate a unit, making copies of it and integrating the copies within a new
composite unit (a unit composed of units); or they may partition a unit into equal parts, forming smaller units.
Prior research on students’ mathematics has identified several operations used to construct and transform units that
satisfy Piaget’s criteria for mathematical actions. Namely, aforementioned operations of unitizing, iterating, and partitioning
are all potentially reversible and composable. In addition, disembedding enables a student to remove a unit (or collection
of units) from a composite unit, without destroying the composite unit (Steffe, 1992). A fifth operation—distributing—
involves inserting the units within one composite unit across the units of another composite unit (Steffe, 1992). For example,
in determining the value of 7 times 4, a student may distribute seven 1s across the four 1s that compose 4, making a sequence
of four units of seven units of 1. Table 1 summarizes these five operations—the operations most germane to our study.
These operations do not exist a priori for students but take years to develop through the internalization and coordination
of sensorimotor activity (Beth & Piaget, 1966). For example, the operation of partitioning develops through several levels
of fragmenting activity in which a student learns to coordinate three competing goals: to break a continuous whole into a
specified number of parts; to produce parts all of equal size; and to exhaust the whole (Steffe & Olive, 2010). Younger
students typically achieve two of these three goals, failing to simultaneously account for one of them (e.g., making five
equal parts within a whole but having some leftover part that is discarded; Hunt et al., 2016).
SA-NCTM-JRME230005.indd 241 5/2/2023 12:12:41 PM

242 Unit Transformation Graphs
Table 1
Operations for Constructing and Transforming Units

Operation Description
Unitizing (Un) Taking an item, a collection of n items, or a continuous span of attention as a whole unit
Iterating (In) Making n identical, connected copies of a unit
Partitioning (Pn) Creating n equal parts within a whole
Disembedding (Dn) Taking n parts out of a whole while maintaining their inclusion as part of the whole
Distributing (Tm:n) Inserting the m units of one composite unit into each of the n units in another composite unit to
produce a unit of units of units
In general, students construct operations first by internalizing their sensorimotor activity, so that they can perform it in
imagination. Then, they can begin to organize these internalized actions within structures for composing and reversing
them (Piaget, 1968/1970). For example, partitioning can become a mental action that students compose with itself (parti-
tioning a partition) or with other mental actions, like iterating. In composing partitioning with iterating, students may
iterate a unit fractional part, like 1/7, to make a non-unit fraction, like 5/7. They may also iterate to reproduce the whole
from a unit fractional part so that iterating and partitioning become organized as both reversible and composable.
We note that, in prior literature, researchers have used splitting to refer to a single operation that subsumes partitioning
and iterating as reversible operations (Hackenberg, 2007; Norton & Wilkins, 2012; Steffe, 2001; Wilkins & Norton, 2011).
We do not explicitly refer to splitting within our UTGs; instead, we indicate the mutual reversibility of the operations of
partitioning and iterating with double arrows. However, as noted by Confrey (1994), students often use the term “split” to
refer generically to various kinds of partitioning activity, as did the PSTs in our study.
Sequencing Operations in Working Memory

Students may need to perform a long sequence of operations to solve a mathematical task. With respect to working
memory—“the process of holding information in an active state and manipulating it until a goal is reached” (Pascual-Leone
et al., 2010, p. 62)—the cognitive demand of the task would increase with the length of this sequence. In our framework,
we adopt Pascual-Leone’s (1970) construct of M-demand, measured by the hypothetical number of action schemes required
to solve the task. This perspective largely aligns with cognitive load theory espoused by Kirschner et al. (2006) wherein
the operations required to move from an initial state to a goal state contribute to the overall demand of a task, competing
for the limited cognitive capacity of working memory (Sweller, 1988).
Cognitive load theory classifies the cognitive demand of (or cognitive load induced by) tasks into three categories:
intrinsic, extraneous, and germane (Sweller, 2010). Intrinsic demands are those inherent to the conceptual element being
learned. However, instructional approaches to teaching that element may introduce extraneous (unnecessary) demands.
To explain germane demand, Sweller (2010) introduced the idea of element interactivity, which describes how some
concepts (such as the elements of a mathematical equation) cannot be learned in isolation from one another. Thus, germane
load “refers to the working memory resources available to deal with the element interactivity associated with intrinsic
cognitive load” (p. 126).
Not coincidentally, many of the examples Sweller (2010) provided for high element interactivity come from mathematics.
We argue that element interactivity is essential to mathematics, as the coordination of actions. Our framework captures
these germane demands with respect to the mental actions involved in solving a mathematical task. It also accounts for
ways that students’ available structures may reduce those demands by enabling students to chunk elements into
single elements.
Several aspects of cognitive load theory align with Pascual-Leone’s (1970) characterization of working memory as
mental-attentional capacity or M-capacity. Pascual-Leone (1970) described M-capacity as “the number of separate schemes
(i.e., separate chunks of information) on which the subject can operate simultaneously using his mental structures” (p. 302).
Commensurate with other measures of working memory, Pascual-Leone (1970) found that adults can typically maintain
five to seven schemes at once. In numerical contexts, such as solving fractions tasks, schemes could refer to the operations
students use to construct and transform units. We note that, as a student of Jean Piaget, Pascual-Leone characterized
schemes and mental structures within a Piagetian perspective. Indeed, framing working memory from a scheme-theoretic
perspective was the motivation for his constructs of M-capacity and M-demand (Pascual-Leone, 1970).
A few studies have investigated ways that students may offload the cognitive demands of mathematical tasks (here
framed as M-demand) when those demands exceed the students’ working memory (M-capacity). For example, Alibali
and DiRusso (1999) found that, in counting, preschoolers rely on their fingers to reduce the cognitive demand of that

c oordinated activity. Elementary school students may similarly rely on figurative materials (e.g., drawings and manipula-
tives) to reduce M-demand of arithmetic tasks (Friso-van den Bos et al., 2013). Furthermore, studies have demonstrated
that students can reduce cognitive demands on working memory by organizing information into larger chunks (Thalmann
et al., 2019). In our framework, chunking would entail organizing units and unit transformations within units
coordinating structures.
Units Coordinating Structures and Stages

In the absence of structures for assimilating multiple levels of units, each unit or unit transformation (e.g., partitioning
a whole into n equal parts) would place separate demands on working memory. However, those units and the operations
that transform them can be organized within units coordinating structures (Boyce & Norton, 2016; Hackenberg, 2007;
Ulrich, 2016). Figure 1 illustrates two- and three-level structures for coordinating units.
The rectangle in Figure 1A represents a two-level structure for coordinating units—one that would organize the previ-
ously described one-to-seven relationship between a whole unit and the unit fraction 1/7. Note that the structure contains
two units (represented within circles) and a pair of reversible mental operations between them (represented by directional
arrows): The whole can be transformed into seven equal parts through the operation of partitioning, and this mental action
can be reversed by iterating one of those parts seven times, reproducing the whole. This units coordinating structure acts
as a single unit that can be used to assimilate two units and the action (in either direction) between them, thus reducing the
M-demand experienced from 3 to 1. In other words, with respect to cognitive demands placed on working memory, we
can think of the two-level structure as a single chunk (Pascual-Leone, 1970; Thalmann et al., 2019)
The triangle in Figure 1B represents a three-level structure that could reduce the M-demand experienced from 5 to 1.
As a larger chunk, it subsumes the roles of two overlapping two-level structures into a single units coordinating structure
that contains three units and the actions in between them. Note that the second two-level structure relates the unit fraction
1/7 to the non-unit fraction, 3/7, through a pair of reversible actions: iterating 1/7 three times to produce 3/7, and conversely,
partitioning 3/7 into three equal parts of 1/7 each. The three-level structure enables students to assimilate non-unit fractions,
such as 3/7, as “numbers in their own right” (Hackenberg 2007, p. 28).
The mathematical power of units coordinating structures is well documented. Student constructions of two- and three-
level unit structures unlock their potential for developing a host of new mathematical concepts, from number sequences,
to multiplicative reasoning, to fractions knowledge, and into algebraic reasoning and calculus (Boyce et al., 2021; Boyce
& Norton, 2016, 2019; Hackenberg, 2007; Hackenberg & Lee, 2015; Steffe & Olive, 2010; Tillema, 2013). In that literature,
students who reliably construct two-level unit structures are considered to be operating at units coordinating Stage 2, and
students who reliably construct three-level units structures are considered to be operating at units coordinating Stage 3.
Reversibility of an operation depends on a student’s ability to maintain the units being transformed. This principle is
evident in the definition of the disembedding operation, wherein the student maintains two levels of units: the composite
unit and its constituent units, when taken out of the whole (Steffe, 1992). With distributing, reversibility depends on main-
taining three levels of units: the two composite units, and the unit of units of units that results from distributing the units
of one composite unit over the units of the other composite unit. In our study, working with PSTs who all operated at units
Figure 1
(A) Two-Level and (B) Three-Level Units Coordinating Structures
Whole
P7
Whole 1/7
I7
P7 ,
I7
P3, I3
1/7 3/7
A B

coordination Stage 2 or 3, we could take for granted the first four operations listed in Table 1; the distributing operation
would be theoretically available to those students operating at Stage 3 (Steffe, 1992).
Fraction Composition
Our depiction of a three-level unit structure (see Figure 1B) resembles Davydov’s (1992) triangle model for multiplication,
wherein the multiplicand serves as an intermediary unit for measuring a quantity, as specified by the multiplier. In the case
of fraction multiplication, the situation becomes much more complex because each of the factors is already a quantity
measured by an intermediary unit. For example, Figure 1B depicts 3/7 as a quantity that is three measures of the interme-
diary unit, 1/7. Thus, fraction multiplication involves at least five levels of units: the whole, the initial unit fraction, the
first quantity measured in terms of that unit fraction, the second unit fraction (of the first fraction), and the measure of the
second fraction in terms of that unit. This complexity may help explain previous findings about the challenges inherent in
understanding fraction multiplication (e.g., Mack, 2001; Olive, 1999; Streefland, 1993).
Hackenberg and Tillema (2009) examined these challenges in a teaching experiment with four sixth-grade students.
Following Steffe (2003), they framed fraction multiplication as “fraction composition” to emphasize its dependence on
students’ ways of operating (composing fractions) rather than a formal operation, or procedure, to follow. Specifically,
they argued that fraction composition relies on recursive partitioning. This operation refers to partitioning the results of
an initial partition in service of another goal, such as determining the fractional quantity produced by taking a unit fraction
of a unit fraction. Hackenberg and Tillema (2009) determined that students needed to operate at Stage 2 to meaningfully
solve such tasks because they involve “anticipating the partitioning of a partition” (p. 14). In other words, students would
need to have available a two-level unit structure, like the one shown in Figure 1A, on which to operate through further
partitioning. The researchers also found that students needed to operate at Stage 3 to meaningfully solve tasks involving
the multiplication of two non-unit fractions, because of the extra units involved. Furthermore, they noted an asymmetry
in fractions multiplication tasks involving a unit fraction and a non-unit fraction.
From a formal perspective, fraction multiplication is commutative. However, students operating at Stage 2 seem to
struggle more in attempting to find a unit fraction of a non-unit fraction than they do in finding a non-unit fraction of a
unit fraction. Hackenberg and Tillema (2009) explained this finding as follows:
In situations involving taking a unit or proper fraction of a proper fraction, these students have difficulty keeping track of the work
on the fractional quantity in relation to the whole because they are “in” the activity of making three levels of units. (p. 16)
For example, when determining the fractional value of 1/4 of 5/6, Stage 2 students may become exhausted in the activity
of determining the value of 1/4 of 1/6 relative to the whole. In the process of operating on the unit fraction, 1/6, they may
lose track of the five units of 1/6 that make up 5/6. Our study provides a similar explanation, but one that explicitly accounts
for the role of working memory in keeping track of units.
Methods
We began the study with 10 fractions tasks, all borrowed from published research articles. We used the research articles
and their descriptions of students’ reasoning in response to the tasks to build a hypothetical UTG for each task. We then
conducted clinical interviews with 12 PSTs, posing subsets of those tasks. We posed tasks orally to reduce the availability
of figurative material on which PSTs may rely to offload demands on working memory. In this article, we focus on three
tasks that, with one exception, we posed to all 12 PSTs. In two phases of retrospective analysis, we assessed the relative
cognitive demand experienced by each PST (Low, High, or Over) and built student-dependent UTGs for each student and
each task. We share details of these methods in the following subsections.
The Tasks
The three tasks analyzed here were borrowed from Hackenberg and Tillema (2009). Table 2 describes each task, ranked
in accordance with its hypothetical M-demand. Note that the table includes two versions of Task 2—one contextualized
(Task 2a) and one decontextualized (Task 2b). All tasks involved fraction composition (ostensibly, fractions multiplication)
and were contextualized as cake problems. Task 1 involves finding a unit fraction of a unit fraction; Tasks 2a and 2b involve
finding a non-unit fraction of a unit fraction; and Task 3 involves finding a unit fraction of a non-unit fraction.
We determined the rank of each task by building a hypothetical UTG: an initial UTG that was independent of any particular
student’s response. We reviewed descriptions of students’ solutions shared in the Hackenberg and Tillema (2009) article to infer
the mental actions they may have used to solve the task without relying on units coordinating structures. In other words, our
hypothetical UTGs attempted to unpack and diagram every unit and every unit transformation a student may use within a
solution of the task. The task rank was the total number of units and unit transformations in this hypothetical UTG. This number
would be the hypothetical M-demand of the task experienced by a student if they had no units coordinating structures available.

For example, Figure 2 illustrates the hypothetical UTG for Task 1. Units are represented by circles (vertices in the
graph), and unit transformations (operations) are represented by arrows (directed edges). Starting with a whole rectan-
gular cake (the unit represented by the top left circle), a student may apply the mental action of partitioning the cake
into 15 equal pieces (P15), resulting in a composite unit of 15. To work with one of those 15 pieces, a student may
disembed one of those pieces (D1). To share that piece fairly with another person, they may partition it into two equal
pieces (P2). Now, to determine how much of the original cake each of those two pieces is, the student may iterate one
of the two pieces (I30) within the original cake. That is, the student may try to figure out how many times the piece fits
into the original cake. In all, this UTG consists of five units and five unit transformations, rendering a hypothetical
M-demand of 10.
Participants
Participants consisted of PSTs at a large land-grant university in the mid-Atlantic United States. All participants were
enrolled in one of two sections of the same mathematics course (Mathematics for Elementary School Teachers) taught by
the same instructor. We visited the two sections of the course and invited all students in both sections to participate in the
study. A total of 12 students agreed to participate.
PSTs comprise a special population of participants for the study because they practice metacognitive skills in the context
of solving elementary school mathematics tasks. Specifically, they are encouraged to explain their reasoning when solving
tasks, and this aided our attempts to model their mathematical thinking. Also, fractions tasks remain problematic for many
PSTs (Lovin et al., 2018), so it was reasonable to assume that our fractions multiplication tasks would be appropriately
challenging for them, especially when required to reason through those tasks.
Data Collection
We conducted individual clinical interviews (Ginsburg, 1981) with all 12 PSTs. Each interview lasted about an hour and
occurred in three parts: an assessment of their units coordination stage using interview tasks from prior studies with middle
Table 2
Fractions Tasks
Task Rank Description
1 10 Imagine a rectangular cake that is cut into 15 equal pieces. You decide to share your piece of cake fairly
with one other person. So, how much of the whole cake would that person get?
2a 12 Imagine you are at a party and a cake is cut into nine equal pieces. Two people show up to the party late
and you decide to share your piece of cake with them. So, what fraction of the whole cake do the
latecomers get together?
2b 12 Imagine cutting off 2/5 of 1/3 of a cake. So, how much is that of the whole cake?
3 14 Imagine cutting off 1/4 of 5/6 of a cake. So, how much is that of the whole cake?
Note. From Hackenberg and Tillema (2009).
Figure 2
Hypothetical UTG for Task 1

school students (Norton et al., 2015); an assessment of working memory (M-capacity) using backward digit span with
digits read aloud (Morra, 1994); and a set of fractions tasks. All interviews were video recorded, and all writing (including
drawings) was captured using Livescribe pens.
We assessed units coordination and working memory during the first two parts of the clinical interview, relying on
methods described in the literature (Morra, 1994; Norton et al., 2015). On the basis of these assessments, we selected a
subset of the 10 ranked fractions tasks to be used in the third part of the interview for each PST. Tasks were ranked by
hypothetical M-demand as determined by the hypothetical UTGs. We intended for initial tasks to impose relatively low
cognitive demand on PSTs and for later tasks to impose high demand, so PSTs assessed with lower M-capacity began with
simpler (lower ranked) tasks.
The fractions tasks were administered one at a time. PSTs were asked to initially solve the tasks mentally with no figu-
rative material so that we may determine the number of units and unit transformations the PST could hold in working
memory without relying on drawings or other notations. We sometimes asked follow-up questions to solicit further behav-
ioral indications of their reasoning. In responding to follow-up questions, PSTs were sometimes encouraged to
make drawings.
Data Analysis
Data analysis included real-time and retrospective analysis of PSTs’ responses to the fractions tasks. The interview
required real-time analysis of the PSTs’ abilities to solve the tasks without using figurative material so that we could deter-
mine whether to continue to more challenging (higher ranked) tasks. We continued the interview with higher ranked tasks
until we inferred that the PST was unable to produce correct or confident solutions. In some cases, PSTs were explicit about
their own perceived limitations, saying, for example, “I have no idea,” “My head does not make sense,” or “My brain is
confused now.”
After all interviews were completed, the team began retrospective analysis of the video data. We conducted video anal-
ysis as a team, with at least two of the three team members present for each meeting (about 10 meetings total). We analyzed
all videos for each participant consecutively, moving from the lowest ranked to the highest ranked task. For each task, the
video analysis consisted of two phases.
In the first phase, we assigned each task a classification of relative cognitive demand using behavioral indicators during
the PST’s response to the task. We coded relative cognitive demand as Low, High, or Over, depending on how challenging
the task seemed to be for the PST in managing the units and unit transformations involved in solving the tasks. Evidence
for a Low code consisted of quick and confident responses, convinced run-through of the solution, and maintaining all the
units and relationships consistently throughout the task. We used the High code when the PST struggled, presumably
operating near the limits of their M-capacity, as indicated by expressed doubt, repeated rehearsal of the task’s solution,
requests for the task to be repeated during the solution process, or losing track of units and relationships during the task.
Use of the Over code indicates that the task was beyond the students’ ability to solve without figurative material or substan-
tial help from the interviewer. A few examples of relative cognitive demand coding follow.
In response to Task 1, PST 23’s behavior indicated that she started solving the task as it was being posed, and then she
immediately (less than one second after we asked the question) gave the answer. As she answered, she seemed relaxed and
confident. Moreover, she was able to readily explain her answer when we asked follow up questions. We coded her response
as displaying Low relative cognitive demand. In contrast, PST 9, who had displayed similar behavior to PST 23 on previous
tasks, did not give an answer to Task 1 immediately, but instead looked up and to the side as she was thinking. When she
did answer, she used a questioning intonation. We coded her as displaying High relative cognitive demand. Finally, PST
18 showed uncertainty in her answer—“one half of one fifteenth, one sixteenth?”—and gave an incorrect answer despite
clearly understanding that she was trying to find one half of one fifteenth. We assigned her response a relative cognitive
demand code of Over.
In the second phase, the research team went back and watched videos again to build a UTG for each task solved by each
PST. Building UTGs resembles conceptual analysis (Steffe et al., 1983; Thompson, 2008; Ulrich et al., 2014), because UTGs
serve as a kind of second-order model of students’ mathematics. However, UTGs apply to a single task and explicitly
include illustrations of the sequences of mental actions we infer PSTs to have used in reaching a solution to that task. In
building the models, the research team allowed relative cognitive demand codes, assessments of working memory, and
assessments of available units coordinating structures to inform graph design. These assessments helped us account for
the total number of units and transformations a PST may have used in solving a task. Thus, the graphs serve as explanatory
models for the PSTs’ observed behavior in solving tasks by drawing on available actions and their assessed M-capacity for
sequencing them.
Because UTGs constitute novel representations of students’ mathematical reasoning, we needed to pay special attention
to consistency across tasks and PSTs. For that reason, we employed a kind of constant comparative method (Grove, 1988)

in our analysis. Specifically, whenever we used a new representation in a UTG, we reviewed UTGs for all prior tasks and
PSTs to ensure that each representation was used consistently across all PSTs and all tasks. Adjustments were made to
prior graphs as new features emerged in newer graphs. For example, in building a UTG for one PST on one task, we intro-
duced the possibility of linking structures together (i.e., daisy-chaining), wherein two two-level structures may share a
common unit on which they overlap. This is illustrated by two overlapping rectangles, as shown in Figure 1B. Once we
introduced that feature for a PST, we needed to go back to determine whether our decision to include it applied to any prior
PST–task pairs.
Results
We present results in three parts. First, we present hypothetical UTGs for each of the three tasks, independent of PSTs
and their ability to structure units (as we began to do with Figure 2). These graphs, then, are theoretical models that demon-
strate an initial application of our framework. Next, we present descriptive statistics for the relative cognitive demand
experienced by each of the 12 PSTs, across the three tasks. Finally, we present UTGs for some of the PSTs’ task solutions,
to demonstrate the illustrative and explanatory power of the graphs in diagramming PSTs’ mathematical reasoning.
Collectively, these UTGs help us to identify patterns across PSTs’ solutions to the fractions tasks.
Hypothetical UTGs for the Three Fractions Tasks

As discussed in the Methods section, we developed hypothetical UTGs for each of the three tasks: Task 1, which involves
taking a unit fraction of a unit fraction; Task 2a (analogous to Task 2b), which involves taking a non-unit fraction of a unit
fraction; and Task 3, which involves taking a unit fraction of a non-unit fraction. We concluded that Task 1 involved
10 unit constructions or transformations, as shown in Figure 2, rendering it Rank 10. As shown in Figure 3, we similarly
determined that Tasks 2 and 3 were Rank 12 and Rank 14, respectively.
Like in all mathematical tasks, a student may take multiple paths in reasoning toward a solution, but in composing frac-
tions, all paths involve constructing and transforming units. In the case of Task 2a, starting from a whole unit (Whole1)—
the result of unitizing—a student may partition it into nine equal parts (P9). Next, the student may disembed one of those
parts (D1) by removing it from the whole while maintaining its status as one of the nine equal parts. Now the student has
a new whole on which to operate (Whole2). The student can operate on the new whole by partitioning it into three smaller
parts (P3), but then those parts would be smaller than the other eight parts making up the whole. This incongruence may
induce an action of iterating the part within the original whole 27 times (I27), or distributing the partition across those other
eight parts, producing 27 parts in total. Finally, the student could solve the task by iterating one of those parts twice (I2),
Figure 3
Hypothetical UTGs for (A) Task 2a and (B) Task 3
A Whole1 9 Whole2
P9 D1
I27 P3
I2 Whole3
D1
2/27 3
B Whole1 6 1/6 5/6

P6 D1 I5
I27 P4
I5 D1
5/24 Whole3 4

or by disembedding two of them and naming them 2 out of 27. Because all the units are treated as whole units, the student
can interpret the final result using a part-whole scheme (2/27 as 2 out of 27), which treats fractions as a comparison of two
whole numbers (Pitkethly & Hunting, 1996; Saenz-Ludlow, 1994).
Note that the UTGs illustrated in Figure 3 contain no units structures beyond one level. Thus, the hypothetical M-demand
of Task 2a (Panel A) following this path would be 12: six unit constructions (circles) and six unit transformations (arrows).
This demand would exceed the M-capacity of virtually all adults (typically limited to an M-capacity of 7). Without the use
of figurative material or reliance on learned procedures, students may offload this excessive demand by assimilating some
of the units and actions within units coordinating structures they have constructed. For example, at the start of the task, a
student operating at Stage 2 may assimilate the whole, the action of partitioning, and the resulting nine parts within a
two-level structure (like the one shown in Figure 1A).
The main difference between Tasks 1 and 2a is that after the second partitioning, the student would need to iterate the
result (or disembed two of the final partitions). In the diagram (Figure 3A), this extra action and resulting unit is indicated
by an additional arrow (I2) and circle (2/27), respectively, in the lower left portion of the UTG. The extra arrow and circle
account for the extra cognitive demand of the task.
In response to Task 3, a student may start by constructing 5/6 of a cake. This action is represented by the top row of the
UTG (Figure 3B) and involves an iteration of the result of partitioning the whole (Whole1) into sixths (or disembedding
five of the sixths, which would yield the same M-demand). The extra iteration (I5) is represented by the additional arrow
and circle in the top right of the UTG. As in the other tasks, to find 1/4 of the 5/6 and relate the result back to the size of
the original cake, the student may construct a recursive partition (in this case, 1/4 of 1/6). However, during this process,
the student would need to remain aware of the mental action of iterating (I5) to then apply that mental action again to the
result of recursive partitioning (Whole3). The link between those actions (I5) is indicated by the two small tick marks on
the iterating arrows, representing the parallel awareness of carrying out five iterations to form the 5/6 and to form the final
quantity. This need to carry out the iteration twice instead of once results in a hypothetical M-demand that is 2 greater
than that of Task 2a (and analogously Task 2b).
That Task 1 should be less demanding than Tasks 2a and 3 is unsurprising, because Task 1 involves only unit frac-
tions. However, that Task 3 should be more demanding than Task 2a is somewhat surprising, because both involve a
unit fraction and a non-unit fraction. However, we know from past research (e.g., Hackenberg & Tillema, 2009) that
situations involving taking a unit fraction of a non-unit fraction appear more difficult for students than situations
involving taking a non-unit fraction of a unit fraction. The hypothetical UTGs in Figure 3 provide a clear illustration
of the difference.
Descriptive Statistics for Cognitive Demand

Table 3 summarizes the relative cognitive demand experienced by each PST in response to each task. We made assess-
ments of this relative demand, in qualitative descriptors of Low, High, and Over, as described in the Methods section. The
table also includes n and m codes for the assessed units coordinating stage and M-capacity of each PST, respectively. For
example, “n2m5” indicates that we assessed PSTs 7, 18, 20, and 21 to be operating at Stage 2 and with an M-capacity of 5.
Quantitative analysis of these data, across PSTs and all 10 tasks, can be found in Kerrigan et al. (2020).
Performance by all but one PST was consistent with an increasing hierarchy of task difficulty moving from Task 1
to Task 3. The one participant who did not fit this pattern, PST 10, seemed to find Task 2a less difficult than Task 1.
She had answered “Half of one fifteenth?” for Task 1, and then was able to say that the piece would fit into the whole
cake 30 times, after prompting by the interviewer. The task was clearly difficult for her, and her solution seemed to
Table 3
Summary of Relative Cognitive Demand by Task and PST

Code for each PST by units coordinating stage (n) and M-capacity (m)
n2m5 n2m6 n2m7 n3m5 n3m6 n3m7
Task Rank 7 18 20 21 9 10 22 8 23 11 17 13
3 14
2b 12
2a 12
1 10
Note. Relative cognitive demand codes: Low in green, High in yellow, and Over in red.

rely on the interviewer’s intervention (thus, the Over coding). The interviewer asked a follow-up question, similar to
the problem posed in Task 1, to see whether PST 10 could independently solve it. She did and then went on to solve
Task 2a, but still exhibited signs of relatively High cognitive demand. Therefore, we posit that she had made some
adaptations to her understanding of fractions on the basis of her response to Task 1 and its follow up that allowed her
to solve Task 2a independently.
Recall that we included two tasks that involved taking a non-unit fraction of a unit fraction: one contextualized task
(Task 2a) and one decontextualized task (Task 2b); both have rank 12 (see Table 2). All eight PSTs who responded to both
tasks experienced equal to less relative cognitive demand in response to Task 2b than to Task 2a. The mathematics educa-
tion literature provides no clear indication that contextualized tasks should be more or less cognitively demanding than
decontextualized tasks (cf. Stylianides & Stylianides, 2008), and it may be that these eight PSTs experienced less demand
in response to Task 2b because they could use the immediately prior experience with Task 2a to help solve Task 2b. Being
decontextualized, Task 2b might be more comparable to Task 3, so it is important to note that PSTs generally experienced
both Tasks 2a and 2b as less cognitively demanding than Task 3.
The hierarchy illustrated in Figure 3 confirms previous findings that working with non-unit fractions presents
cognitive challenges beyond unit fractions (Hunting, 1983; Norton & Wilkins, 2010; Pitkethly & Hunting, 1996;
Simon, Kara, & Placa, 2018). It also supports the idea that fraction multiplication is psychologically noncommutative;
that is, taking a unit fraction of a non-unit fraction is more demanding than taking a non-unit fraction of unit fraction.
Our hypothetical UTGs predicted this phenomenon of noncommutativity, previously noted by Hackenberg and
Tillema (2009). Considering UTGs for actual students provides further insight into the challenges inherent within
each task.
UTGs for the PSTs

We found specific ways of operating that strongly affected how PSTs deployed units coordinating structures, sometimes
resulting in solution types we had not anticipated. In this section, we use examples to illustrate three themes in students’
ways of operating. First, we found that the PSTs used their units coordinating structures to lower the M-demand of fractions
tasks that would have otherwise exceeded their M-capacities; we use an example from Task 1 to illustrate this finding.
Second, we found differences in PSTs’ ways of operating when reasoning with fractions as parts out of wholes versus as
measures; we share an example of each way of operating when solving Task 2a. Finally, we found that all the PSTs expe-
rienced limitations in their ways of operating that were based on available units coordination structures and working
memory, particularly in determining a fraction of a non-unit fraction (Task 3).
Meeting the Cognitive Demands of Task 1

We built UTGs for each of the 11 PSTs who attempted Task 1, a task that eight of the PSTs successfully solved (see
Table 3). The UTGs were identical for five of those eight PSTs (see Figure 4) and nearly identical for a sixth PST (PST 9,
who imagined partitioning the cake into 15 parts through a pair of partitionings: three horizontal and five vertical). Note
that the solution path fits the one we anticipated (see Figure 2), except that the PSTs applied two-level structures to two
pairs of units. Here, we discuss how we built the UTG in Figure 4, using our model of PST 13’s solution as representative.
We then discuss how our UTGs for other PSTs’ attempted solutions varied from it.
Figure 4
Common UTG for Task 1

Researcher: Imagine a rectangular cake that is cut into fifteen equal pieces. You decide to share your piece of cake fairly
with one other person. So, how much of the whole cake would that person get?
PST 13: [pauses for 6 s] So, you would have, like, one fifteen. So, they would have half of one fifteen, which would
be [begins to grimace and pauses for 11 s] Like, I don’t . . . like . . . would it be, like, one thirtieth? I don’t
know. I feel like I’m thinking too hard now.
Researcher: OK, so you’re, you’re sure it’s a half of one fifteenth? Or, or are you not sure about that part either?
PST 13: No, I’m sure it’s, like, half . . . it’s half of one fifteenth.
Researcher: OK, and you’re trying to figure out another name for half of one fifteenth?
PST 13: Yeah.
Researcher: And what do you think that name is?
PST 13: [pauses for 4 s] All I could think of is, like, point five over one … one fifteen, and I know that’s not
correct.
Researcher: OK, but you started to say one . . .
PST 13: Thirtieth.
Researcher: OK, so why, why do you think it might be one thirtieth?
PST 13: Because . . . if you, like, think about if everyone split their one piece in half then it would make thirty
pieces. So, then it would be one out of thirty. So, I’m not . . . I don’t know, but that’s, like, my thought process
of it.
Despite our assessment of PST 13 as a student operating at Stage 3—and with an M-capacity of 7—her solution to Task 1
demonstrated a High level of cognitive demand. This demand was particularly demonstrated by her hesitations and
expressed uncertainty (e.g., “I don’t know”). Recall that, without relying on units coordinating structures, Task 1 would
have a hypothetical M-demand of 10. As a student operating at Stage 3, PST 13 should have had available three-level
structures to reduce those demands. However, as we will see, none of the PSTs used three-level units structures germane
to fractions multiplication tasks.
As is evident in the UTG (Figure 4), we inferred that PST 13 did use a pair of two-level structures to reduce the cognitive
demands of the task. Specifically, she used a two-level units coordinating structure for the actions associated with each of
the two unit fractions, 1/15 and 1/2. This means, for example, she could partition the whole cake into 15 pieces (P15) and
remain simultaneously aware of the whole (cake), the 1/15, and the 1:15 relationship between them. Prior literature refers
to such a structure as a “partitive unit fraction scheme” (Steffe, 2001, p. 290). The partitioning action and the two quanti-
ties it connects are structured into one chunk (represented by the rectangle in the upper-left side of Figure 4), which would
reduce the experienced M-demand by 2; likewise for the two-level structure relating the units of 1/15 and 2 (as represented
by the rectangle on the right side of Figure 4).
PST 13 completed the task by distributing two halves across each of the 15 fifteenths (T2:15), which is equivalent to a
conceptual multiplication of the numbers 2 and 15 (Steffe, 1992). Note that this operation is available for students only
when they can maintain the 1:15 and the 1:2 structures simultaneously and distribute one across the other (Steffe, 1992).
Evidence of this coordination is apparent in the PST’s explanation of her thought process: “If everyone split their one piece
in half then it would make thirty pieces. So, then it would be one out of thirty.” Relative to the hypothetical UTG shared
in Figure 1, the distribution (T2:15) would further reduce the experienced M-demand of the task by 2, resulting in only four
units or transformations that had to be held in working memory.
Other PSTs’ reasoning followed the same pattern but seemed more certain and immediate. For example, PST 22
responded as follows:
You get one fifteenth of the cake and split that in half. My first thought was one thirtieth of the cake, because [makes partitioning
motion with her hands] splitting that in half, like if you were to split every piece of fifteen in half, then that would be like one
thirtieth of the entire cake.
Like PST 13, PST 22 seemed to imagine partitioning each one of the 15 original parts into two parts to produce 30 parts
in the whole. PSTs 9, 10, 11, 17, 18, 20, and 21 also reasoned along these lines. However, PSTs 10, 18, and 21 were not
successful in independently solving the task.
PSTs 18 and 21 were assessed as operating at Stage 2, with an M-capacity of 5, so theoretically they should have been
able to solve the task as PST 13 had. Instead, they seemed overwhelmed in attempting to carry out the distribution of units
discussed earlier. Both of them relied on a fractions multiplication algorithm in lieu of the distribution. As PST 21 put it,
“Like when you hear one thirtieth, it makes it seem like there are thirty slices [motions with hands] of cake, but knowing

that like when the fraction is on the bottom it’s actually smaller. Like the thirtieth is smaller.” The remaining two PSTs—
PSTs 8 and 23—seemed to engage in a pattern of reasoning that we refer to as daisy-chaining, discussed in the next section.
Two Distinct Ways of Operating in Task 2a

Of the 11 PSTs who attempted Task 2a, only five were successful, and all five experienced relatively High cognitive
demand. Our hypothetical UTG (Figure 3A) accounts for a hypothetical cognitive demand of 12, 2 more than Task 1.
However, as with Task 1, PSTs could reduce the M-demand they would experience by applying two-level structures. In
particular, they may assimilate the two unit fractions, 1/9 and 1/3, within two-level structures, and determine the value of
1/3 of 1/9 by distributing. Still, they would need to account for an additional action and an additional unit in producing
2/27 from the 1/27 that results from the distribution, inducing an experienced M-demand of 6 (2 more than Task 1).
Among the six PSTs coded Over (red in Table 3), four (PSTs 9, 11, 13, and 21) produced 1/27 in much the same way they
had produced 1/30 in response to Task 1, but they did not independently proceed further to produce 2/27. For example,
after partitioning each ninth into three parts, PST 11 responded, “They’d be getting one [pause] twenty-sevenths each.”
When prompted to state the amount the latecomers would get together, she hesitated and ultimately responded as follows:
“Six ninths. I don’t know . . . two thirds of one ninth, but I don’t know what that is.” We take this response as evidence that
her working memory had been exhausted in building 1/27 or that 1/27 was not an iterable unit for her; this despite the fact
that we had supposed she would be operating at Stage 3, with an M-capacity of 6.
Among the five PSTs who were successful in independently solving the task, we identified two distinct ways of operating.
One involved beginning as described previously, and then using part-whole reasoning and the mental action of disembed-
ding to produce 2/27 as “two out of 27.” The other involved treating 1/27 as an iterable unit, made possible by daisy-chaining.
We next discuss each strategy in turn.
Determining 2/27 as a Part-Whole Relationship

PSTs 10, 17, and 22 were successful in solving the task using part-whole relations. We use PST 10’s response as
an example.
PST10: OK, so this is my cake [sweeps hand across the table], and it’s nine equal pieces so three rows of three. So I
take one out [gestures taking one out], so there would be eight-ninths of the cake left [gestures as if to shove
the other eight ninths away], but I’m taking one ninth, and I’m sharing that with three equal parts [gestures as
if to cut three times] because there are two other people. So that’d be . . . well, it’d be two thirds out of my one
ninth, and then I’d multiply that. So it would be [starts drawing with her finger on the desk, as if to compute
the product] . . . one over nine times three over two. So it would be . . .
Researcher: Instead of doing it that way could you figure it out thinking about the—what are we doing?—cake, like you
were doing before where you keep dividing up the cake?
PST 10: So then [pauses]. So then I have one ninth divided by three. So then I guess I do nine times three [looks at
researcher], which is 27. So they get, they get two twenty-sevenths, cause you said they’re putting them
together, right?
PST 10’s gestures and verbalizations indicate use of the operations illustrated in the UTG depicted in Figure 5. She described
a relationship between the whole and its nine parts (“Three rows of three”), as illustrated by the left-hand rectangle in the
UTG. She explicitly referred to disembedding (D1), in terms of “taking one out,” and then she gestured as if to partition
that piece into three parts (P3). She also demonstrated the increased demand of the task (relative to Task 1) in reconciling
the “two thirds out of one ninth” (emphasis added). Note in particular the part-whole language: “out of” (Watanabe, 2006).
At that point, PST 10 began to rely on a multiplication algorithm, until the researcher encouraged her to reason through
the task. She responded by determining the number of parts in the whole, as she had in Task 1, by distributing three units
into each of the nine parts (T3:9). All the while, she maintained the two parts she would need to disembed from that whole
(D2), ultimately naming them “two twenty-sevenths.”
By illustrating the unit constructions and transformations PST 10 performed in response to Task 2a, as well as her use of
available units coordinating structures, we can better understand why the task is more demanding than Task 1 and why PSTs
18, 20, and 21—assessed with M-capacities of 5—did not complete it. Conversely, PSTs 9, 11, and 13 were assessed with
M-capacities of 6 or higher—theoretically sufficient to meet the M-demand of Task 2a when solved in the manner described here.
Working From 1/27 as a Measure

The triangle in Figure 1B includes two overlapping rectangles. Whereas the triangle represents a three-level unit struc-
ture, the rectangles represent two-level structures that share a common unit. The overlap represents a pattern of reasoning

Figure 5
UTG for PST 10 in Response to Task 2a
we refer to as daisy-chaining. PSTs 8 and 23 exhibited this pattern in response to Task 2a, wherein they seemed to shift
seamlessly from one two-level structure to the next, through the overlapping unit, but without coordinating all three units
at once. Operating at Stage 3, both PSTs may have had three-level structures available—structures that would have allowed
them to coordinate 1/3 of 1/9 as a new unit of measure—but neither PST exhibited the use of such a structure. Instead, we
argue that daisy-chaining enabled them to treat 1/27 as an emergent quantity (cf. Ellis, 2007): a unit of measure available
to them in the context of operating on the task. They could then iterate this emergent unit to produce 2/27. We use the case
of PST 23 to demonstrate our argument.
PST 23: [looks up and to the left] One twenty-seventh each. One twenty-seventh piece of cake each. I guess, is that
what you’re asking?
Researcher: OK. So how much do they get together? It was the question. You said one twenty-seventh each, and I asked,
“What fraction of the whole cake do the latecomers get together?”
PST 23: Two twenty-sevenths. Well are you splitting the one piece for you, just for those two people, or all
three?
Researcher: Let me read it again and then ask the question. So this was the original task, I’m just trying to clarify; you’ve
already answered; I’m just trying to clarify. Imagine you are at a party and a cake is cut into nine equal
pieces. Two people show up to the party late and you decide to share your piece of cake with them. So, what
fraction of the whole cake do the latecomers get together?
PST 23: Well they get two thirds of one ninth of the cake.
Researcher: That’s a nice way to say it. And then could you describe that as a single fraction?
PST 23: Ah, I guess I will just stay with the two twenty-sevenths of a cake [nervous laugh].
In comparison with PST 10 (and the other participants who used part-whole reasoning), PST 23 readily alighted on 1/27
(as she had for Task 1). She gave no indication that she disembedded one of the first nine parts as PST 10 had (“So I take
out one”), and she referred to 1/27 as a single unit of measure: “One twenty-seventh piece of cake.” However, she did not
immediately generate 2/27 from this new unit and, when reminded of the question, returned to her activity of producing
1/27 by splitting (partitioning: P3) 1/9 into three parts (see Figure 6). Because she consistently referred to the fraction, 1/27,
rather than a whole number of parts, we infer that her final action in producing 2/27 involved iterating the emergent unit
of 1/27 twice (I2), instead of disembedding two parts from 27 parts.
Note that PSTs 8 and 23 were the only participants assessed with an M-capacity of 5 to independently solve
Task 2a, albeit while experiencing High cognitive demand. Figure 6 illustrates how this was theoretically possible,
just as Figure 5 illustrates why it was theoretically impossible for the other three PSTs with M-capacity of 5 (PSTs
18, 20, and 21) to solve Task 2a using part-whole reasoning. Specifically, the UTG in Figure 5 represents an experi-
enced M-demand of 6 (reduced from 10 by assimilation of units or actions into two two-level structures, as
represented by the rectangles), whereas the UTG in Figure 6 represents an experienced M-demand of 5. Thus, the
strategy illustrated in Figure 6 was accessible to PSTs with an M-capacity of 5 but the strategy illustrated in
Figure 5 was not.

Figure 6
UTG for PST 23 in Response to Task 2a
Whole 1/9
P9, I9
P3 I3
I2
1/3 2/27
T3:9
An Attempt to Solve Task 3

None of the PSTs successfully solved Task 3. Two PSTs (PSTs 9 and 17) gave the correct answer of 5/24 through the use
of multiplication algorithms, but neither could relate that algorithm to the cake context. In general, responses were so
curtailed that we were able to create UTGs for only six of the 12 participants. In all six cases, the UTGs began with the
use of a two-level structure for understanding 1/6. This structure would decrease the experienced M-demand by 2 from
our hypothetical UTG (Figure 3B). However, all six students then disembedded 5/6 and attempted to operate on it through
partitioning, as opposed to disembedding 1/6 and operating on it without losing track of the five iterations, as in our
hypothetical UTG.
PST 22 made the most progress on this task. This finding is not surprising given that PST 22 was one of two PSTs assessed
with an M-capacity of 7, and none of the PSTs used three-level structures (or even daisy-chaining) in response to this task.
PST 22: You’re cutting off one fourth of five sixths of a cake?
Researcher: Yes.
PST 22: [uses hands to show number of pieces on the desk and begins talking to herself] So, you’d have, so you’d
have six pieces . . . and out of those five . . . you want to cut off one fourth of that. Um . . . I guess you
would . . . I mean I guess you could split those five pieces into four and get one of those, but I’m trying
to think like numbers-wise what that would . . . I . . . well . . . [pauses for 7 s] I guess of those five pieces
you could . . . split them into . . . like you could get a . . . split them into twenty pieces because five times
four is twenty and then, um, you would take one fourth of that . . . I guess it would be five pieces. Yeah,
it would be five pieces of that twenty to find the one fourth of the five sixth. Is that, do I need to explain
it more?
Researcher: OK, uh, let’s . . .
PST 22: Which would be, do you want me to draw it? [reaches toward paper]
Researcher: Well, tell me the final answer and then we can draw it.
PST 22: Um, oh gosh it would be . . . [pauses for 4 s]. Splitting twenty, it would be five . . . well it would be five
twentieths, which would equal one fourth, so like five of those, but then I don’t know how to figure that out
into sixths. I think that’s my . . .
Researcher: Yeah, that’s cool. I like the way you’re reasoning. Let’s draw it, and I think you will figure it out.
PST 22 seemed to conceptualize the initial fraction (5/6) as a part-whole relation, partitioning the whole into six equal parts
(P6) and disembedding five of them (D5; see Figure 7). This inference is supported by her verbalization, “So you’d have
six pieces . . . and out of those five.” As she tried to find 1/4 of 5/6, she operated only on the five disembedded parts and
lost track of the sixth part making up the whole. Thus, instead of taking 1/4 of 5/6, as originally intended, she ended up
distributing four parts into each of the five parts (T4:5), producing 20 parts, and taking 1/4 of those 20 parts (P4) instead,
which brought her back to five parts. Because she lost track of the whole and the six parts that constitute it, she had begun
treating five—with its 20 parts—as the whole, and renamed the five parts (out of 20) as 1/4. This partial solution would

Figure 7
UTG for PST 22 in Response to Task 3
exhaust her M-capacity of 7 before she could return to the six parts in the whole that had been lost when she disembedded
five parts from it.
Working with an M-capacity of 7 or less, successful solutions would need to rely on daisy-chaining or a three-level
structure for coordinating fractional units, such as 1/6 and 1/4. Operating on a non-unit fraction, like 5/6, seemed to preclude
the use of daisy-chaining that PSTs 8 and 23 had successfully employed in solving Task 2a. For the other PSTs, who did
not employ daisy-chaining or three-level structures and relied on part-whole reasoning instead, the M-demand of the task
would be too great. Even PST 22’s response to Task 3 (as the most successful response to that task) demonstrates the
increased demands of taking a unit fraction of a non-unit fraction (versus vice versa) and the psychological noncommuta-
tivity of fraction multiplication. The UTG (Figure 7) illustrates the phenomenon because PST 22 operated on only one of
the two units making up 5/6 as a part-whole relation and, in the process, lost track of the other unit.
Discussion
UTGs explicitly recognize students’ mental actions as the basis for mathematics itself (Piaget, 1968/1970). Specifically,
they diagram the unit constructions and unit transformations that undergird number and numerical operations. They also
account for the role of two cognitive constructs in meeting the cognitive demands of mathematical tasks: the psychological
construct of working memory, framed here with the concept of M-capacity (Agostino et al., 2010; Pascual-Leone, 1970);
and units coordinating structures (see Figure 1), a theoretical construct that arose through research on students’ mathematics
(Steffe, 1992). We have introduced UTGs as a tool for building models of students’ mathematics in the context of multi-
plication of fractions. In demonstrating the utility of this tool, we have addressed two related questions.
We began by considering the psychological noncommutativity of fraction multiplication: Why are tasks involving unit
fractions of non-unit fractions more cognitively demanding than tasks involving non-unit fractions of unit fractions? In
addressing this question, we drew heavily on the work of Hackenberg and Tillema (2009) who conducted a teaching exper-
iment in which they posed a sequence of related tasks (see Table 2). Our hypothetical UTGs accounted for the increased
cognitive demand from Task 1 to Tasks 2a and 2b, and from Tasks 2a and 2b to Task 3, with respect to additional actions
and units involved in their solutions (see Figures 2 and 3). In particular, when taking a unit fraction of a non-unit fraction,
students may need to maintain an additional unit transformation, related to the numerator in the non-unit fraction, as they
operate on its unit fraction with the other (unit) fraction (Figure 3); or, as seen in the case of PST 22 on Task 3, operate on
the numerator and risk losing track of the original unit fraction. Empirical results regarding the relative cognitive demand
exhibited by PSTs upheld these models (see Table 3; also see Kerrigan et al., 2020). UTGs from actual PSTs provide for
nuance when addressing our second question regarding the affordances of units coordination structures in meeting cogni-
tive demands of fractions tasks.
Units Coordinating Structures

We had assessed the 12 PSTs who participated in our study as operating at units coordinating Stage 2 or 3 (see Table 3),
which means they would have two- or three-level structures available for solving numerical tasks (Steffe & Olive, 2010).
Indeed, each of the PSTs employed two-level structures to solve each of the fractions tasks they attempted. Specifically,
they regularly used two-level structures to coordinate a whole unit with a unit fraction of it; this two-level structure char-
acterizes a scheme identified in prior literature as the partitive unit fraction scheme (Steffe, 2001). They also engaged in
the activity of distributing one composite unit over the units in another composite unit—an activity Steffe (1992) refers to
as a units coordination. However, none of the PSTs used a three-level structure to coordinate the three levels of units (the
whole, the fraction, and its unit fraction) within a non-unit fraction as a single unit (“A number in its own right”; Hackenberg,

2007). This finding aligns with prior research indicating that Stage 3 students operate with three levels of whole-number
units before learning to coordinate fractional units in an analogous manner (Boyce & Norton, 2016).
The UTGs also illustrated different patterns of reasoning when students conceptualize fractions as part-whole relation-
ships versus fractions as measures. These distinct patterns were especially evident in PSTs’ bifurcated responses to Task 2a
(see Figures 5 and 6). In part-whole reasoning, disembedding is the key action (Steffe & Olive, 2010). PSTs used that action
to take one part out of the nine-part whole and then to take two parts out of the 27 parts, in naming 2/27. PSTs who engaged
in part-whole reasoning to produce 1/27 first (e.g., “They’d be getting one twenty-sevenths each”) struggled to name the
total amount (2/27). This finding fits prior literature suggesting that iterating a unit fraction, like 1/27, relies on producing
the unit fraction as a unit of measure, rather than a part-whole relationship between two whole numbers (Kieren, 1980;
Simon, Placa, et al., 2018; Tzur, 1999).
In employing measurement reasoning, PSTs did use iterating to produce 2/27 from 1/27, and 1/27 did seem emergent for
them, in the truest sense of the word (cf. Ellis, 2007). That is, 1/27 seems to have emerged as a unit from their activity of
taking a unit fraction of a unit fraction (cf. Hackenberg & Tillema, 2009). These PSTs showed no signs of disembedding
a part from the first partitioned whole before finding a unit fraction of it. Rather, they seemed to continue operating on the
unit fraction contained within the first two-level structure, within a second two-level structure. We refer to this pattern of
reasoning as daisy-chaining. Although this way of operating is not as advanced as working within a single three-level
structure, it may prove productive in constructing such structures.
As students operating at Stage 3, PSTs 8 and 23 could have had available a three-level structure for composing unit
fractions (Hackenberg & Tillema, 2009), but we know that students do not necessarily construct all possible fractions
structures by virtue of operating at Stage 3 (Izsák, 2008). Still, Stage 3 students may learn to “experientially bound” (Ulrich,
2015, p. 5) their activity of daisy-chaining, as indicated by the dashed lines in Figure 6. Reflecting on that experience may
support the construction of a single three-level structure. After all, the two overlapping two-level structures contain all
the units and actions contained within the three-level structure (see Figure 1). Constructing such a structure would render
the students’ reasoning more operationally powerful (Hackenberg, 2007) and potentially render fractions tasks less
cognitively demanding.
Working Memory
Prior research in cognitive psychology has demonstrated a strong relationship between working memory and students’
problem-solving ability (Bull & Lee, 2014; Swanson & Beebe-Frankenberger, 2004), including a predictive relationship
between assessed working memory and future mathematical achievement (Blankenship et al., 2018; De Smedt et al., 2009).
Our models explain this relationship by (a) illustrating ways that students can use working memory to sequence their mental
actions in the service of solving mathematical tasks, and (b) suggesting ways that sequences of actions may become struc-
tured so that future tasks become less demanding.
As with units coordinating structures, it appears that PSTs did not always use their available working memory optimally.
For example, we assessed PSTs 18 and 21 with M-capacities of 5, which should have been sufficient to continue their ways
of operating toward a successful solution of Task 1, but both PSTs became overwhelmed as they attempted to coordinate
1/2 and 1/15. It may be that the units coordination itself was not available to the PSTs, as has been found in prior literature
among some students operating at Stage 2 (Hackenberg & Tillema, 2009). However, we find it conspicuous that these two
particular students were assessed with lower M-capacities, and Pascual-Leone’s (1970) model of working memory accounts
for related challenges they may have experienced.
In his model of working memory, Pascual-Leone (1970) noted that M-capacities indicate “the maximum number of discrete
‘chunks’ of information or schemes that M [the central executive] can control or integrate in a single act” (p. 301). In our
framework, these chunks comprise actions and units or unit structures, and none of the PSTs in our study overperformed
the theoretical limit prescribed by this maximum number. Rather, PSTs’ available units coordinating structures explain how
they chunked units and actions to lower the experienced M-demand of mathematical tasks while working within this limit.
Precisely to the point, Pascual-Leone’s framework for working memory was particularly powerful for us because it allowed
us to draw on knowledge of units coordinating structures (and related schemes) as “chunks.” However, students will not
always function at maximal capacity, and Pascual-Leone gave a number of reasons for this, including motivation and
fatigue—factors upheld in several other studies of working memory and task performance (e.g., Eysenck & Calvo, 1992).
Implications for Future Research

Thanks to a growing body of research specifying the mental actions students use to construct and transform units in
other domains, we see potential to expand the use of UTGs in building models of students’ mathematics. Multiplicative
reasoning offers particularly fertile ground because it is essentially characterized as a transformation of units (Davydov,
1992) and involves many of the same operations identified in fractions reasoning (Hackenberg & Tillema, 2009;

Steffe, 2001). Specifically, in reasoning through whole number multiplication problems, students construct composite units
through iteration of a unit of 1 and distribute the units within one composite unit across the units within another composite
unit (Steffe, 1992). Moreover, they often rely on operations of disembedding and partitioning to break composite units into
smaller units for ease of computation (e.g., 7 as 5 + 2). Thus, we can imagine UTGs similar to the ones shared here that
illustrate students’ various forms of multiplicative reasoning. When compared with one another, these UTGs may provide
further insight into the co-construction of fractions knowledge and multiplicative reasoning (Boyce & Norton, 2016).
We see similar opportunities for building explanatory models for students’ solutions of algebra tasks, when algebraic
reasoning is characterized as operating on quantitative unknowns (Hackenberg & Lee, 2015). By illustrating units and
unit transformations, UTGs could elucidate students’ ways of operating with unknowns compared with their ways of
operating with other kinds of units, such as composite units and unit fractions. Furthermore, Reynolds and Wheatley (1996;
also see Wheatley & Reynolds, 1996) have demonstrated a psychological basis for geometric reasoning in unit construction
and coordination, so we could extend UTGs to geometry as well. Across each context, when units are assimilated into
existing structures, working memory is freed to focus on ever more complex tasks; a special case of the phenomenon
cognitive psychologists call chunking (e.g., Pascual-Leone, 1970; Sweller, 2010). This structuring and offloading may
explain the sense in which mathematics builds on itself.
In the present study, we built UTGs of PSTs’ mathematical activity. We noted that this special population served the
study’s purpose in developing UTGs as a methodology for modeling students’ mathematics. However, our primary interest
is in building models of K–12 students’ mathematical activity, whether solving multiplicative, fractions, or algebraic
reasoning tasks. We would also like to engage PSTs in building UTGs of K–12 students’ mathematics, particularly in
support of efforts to draw these future teachers’ attention to students’ mathematics and the coordinations of mental actions
that it comprises (Liang, 2021).
We have noted that UTGs are directed graphs, in the mathematical sense of the term: Vertices represent units (or unit
structures), and directed edges represent actions between units. This formal mathematical connection suggests that, once
developed as models of students’ ways of operating, UTGs become available for further mathematical analysis. For example,
as a directed graph, we could represent a UTG with an adjacency matrix, which would indicate direct connections between
pairs of units in a student’s solution to a task. The nth power of that matrix would represent the number of n walks between
pairs of vertices, which we could interpret as distances between units, as measured in the number of actions the student
took to transform one unit into the other.
Limitations
The research team developed each UTG through consensus; the codes of relative cognitive demand of tasks were devel-
oped likewise. This collaboration precluded cross-validation of the UTGs and codes. The choice to collaborate was largely
a consequence of our purpose. Because UTGs had not been used before, we developed them as a team of researchers.
Consequently, we cannot make strong claims about the validity or reliability of our assessments. Such claims would require
multiple assessments of M-capacity and relative cognitive demand, as well as independent researcher development of
UTGs. Other threats to validity include the use of case studies that might not generalize to other populations, especially
from PSTs to elementary and middle school students.
Conclusions
We have framed our article in response to the critique by Kirschner et al. (2006) of problem-based approaches to math-
ematics education. We have accepted the authors’ main critique that instructional theories in mathematics education rarely
give explicit consideration to limitations in students’ working memory when solving novel tasks and constructing new
schemes. In contrast to existing theory, UTGs do account for the role of working memory during problem solving so that
teachers and researchers may predict and attend to task demands while supporting students’ productive struggle (Lynch
et al., 2018). For example, teachers could help students offload excessive demands on working memory by allowing them
to use figurative materials, such as manipulatives, drawings, and notation. Indeed, the very purpose of algebraic notation
is to offload some of the demands of mathematical tasks that could overwhelm any mathematician. However, the key
distinction in using those symbols meaningfully versus the kinds of symbolic manipulation that may be taught through
direct instruction lies in understanding them as proxies for mental actions we could perform (Philipp & Schappelle, 1999).
The argument by Kirschner et al. (2006) is ultimately grounded in Sweller’s (1988) cognitive load theory, which asserts
that “solving a problem and acquiring schemas may require largely unrelated cognitive processes” (p. 261). We respond
to the argument by Kirschner et al. (2006) by noting that sequencing actions in the service of solving a problem is inherent
to mathematical schemas themselves (Beth & Piaget, 1966). No mathematical structure exists without the operations it
comprises. Thus, students’ problem-solving activity does not place an extraneous load on working memory but constitutes
the most germane elements of meaningful mathematical learning.

The unit structures identified in UTGs specify the atoms of mathematical knowledge—students’ own mental actions—
while aligning with the overarching framework of cognitive load theory. Consider the following example, about mastering
the game of chess, cited by Sweller (1988) and repeated in Kirschner et al (2006):
Chase and Simon (1973a, 1973b) found that while both novices and masters remembered both realistic chess configurations and
sequences of moves in chunks, the number of chunks did not differ appreciably. Differences did occur in chunk size with masters’
chunks being far larger than those of novices. (Sweller, 1988, p. 258)
This example also fits within Pascual-Leone’s (1970) theory of M-capacity: “M operates upon the units or behavioral
segments available in the subject’s repertoire, as the repertoire changes with learning so will the level of performance even
if the subject’s M value remains constant” (p. 304). Our study, too, affirms the power of constructing new schemes, such
as units coordinating structures, so we are in agreement with Kirschner et al. (2006) that scheme construction should be
the major focus of mathematics education, even if we disagree on what a scheme comprises.
Even if problem solving and scheme construction did compete for the same limited capacity, a readily available
instructional solution presents itself: prompting students to reflect. Beyond problem-solving activity itself, reflection
on activity plays a critical role in the structuring of actions (Wheatley, 1992; Tzur & Simon, 2004). From a construc-
tivist perspective, reflection (or reflective abstraction; Beth & Piaget, 1966) is the cognitive mechanism by which
actions become structured. Once students have solved a problem, they can reflect on their activity, even using drawings
to explicitly trace their actions. Indeed, the PSTs in our study often reflected on the actions they had just enacted,
especially when the task was cognitively demanding. The importance of reflection in problem solving has been recog-
nized in influential publications as far back as 1945, when Pólya included it as the fourth and final principle of How
to Solve It (1945/2004).
References
Agostino, A., Johnson, J., & Pascual-Leone, J. (2010). Executive functions underlying multiplicative reasoning: Problem type matters. Journal of
Experimental Child Psychology, 105(4), 286–305. https://doi.org/10.1016/j.jecp.2009.09.006
Alibali, M. W., & DiRusso, A. A. (1999). The function of gesture in learning to count: More than keeping track. Cognitive Development, 14(1), 37–56.
https://doi.org/10.1016/S0885-2014(99)80017-3
Beth, E. W., & Piaget, J. (1966). Mathematical epistemology and psychology (W. Mays, Trans.). Gordon and Breach.
Blankenship, T. L., Keith, K., Calkins, S. D., & Bell, M. A. (2018). Behavioral performance and neural areas associated with memory processes
contribute to math and reading achievement in 6-year-old children. Cognitive Development, 45, 141–151. https://doi.org/10.1016/j.cogdev.2017.07.002
Boyce, S., Grabhorn, J. A., & Byerley, C. (2021). Relating students’ units coordinating and calculus readiness. Mathematical Thinking and Learning,
23(3), 187–208. https://doi.org/10.1080/10986065.2020.1771651
Boyce, S., & Norton, A. (2016). Co-construction of fractions schemes and units coordinating structures. The Journal of Mathematical Behavior, 41,
10–25. https://doi.org/10.1016/j.jmathb.2015.11.003
Boyce, S., & Norton, A. (2019). Maddie’s units coordinating across contexts. The Journal of Mathematical Behavior, 55, Article 100696. https://doi
.org/10.1016/j.jmathb.2019.03.003
Bull, R., & Lee, K. (2014). Executive functioning and mathematics achievement. Child Development Perspectives, 8(1), 36–41. https://doi.org/10.1111
/cdep.12059
Cobb, P., Jaworski, B., & Presmeg, N. (1996). Emergent and sociocultural views of mathematical activity. In L. P. Steffe, P. Nesher, P. Cobb, G. A.
Goldin, & B. Greer (Eds.), Theories of mathematical learning (pp. 3–19). Erlbaum. https://doi.org/10.4324/9780203053126-6
Confrey, J. (1994). Splitting, similarity, and rate of change: A new approach to multiplication and exponential functions. In G. Harel & J. Confrey
(Eds.), The development of multiplicative reasoning in the learning of mathematics (pp. 293–331). SUNY Press.
Confrey, J., & Kazak, S. (2006). A thirty-year reflection on constructivism in mathematics education in PME. In Á. Gutiérrez & P. Boero (Eds.),
Handbook of research on the psychology of mathematics education: Past, present and future (pp. 305–345). Sense. https://doi
.org/10.1163/9789087901127_012
Davydov, V. V. (1992). The psychological analysis of multiplication procedures. Focus on Learning Problems in Mathematics, 14(1), 3–67.
De Smedt, B., Janssen, R., Bouwens, K., Verschaffel, L., Boets, B., & Ghesquière, P. (2009). Working memory and individual differences in
mathematics achievement: A longitudinal study from first grade to second grade. Journal of Experimental Child Psychology, 103(2), 186–201.
https://doi.org/10.1016/j.jecp.2009.01.004
Ellis, A. B. (2007). The influence of reasoning with emergent quantities on students’ generalizations. Cognition and Instruction, 25(4), 439–478.
https://doi.org/10.1080/07370000701632397
Eysenck, M. W., & Calvo, M. G. (1992). Anxiety and performance: The processing efficiency theory. Cognition and Emotion, 6(6), 409–434.
https://doi.org/10.1080/02699939208409696
Friso-van den Bos, I., van der Ven, S. H. G., Kroesbergen, E. H., & van Luit, J. E. H. (2013). Working memory and mathematics in primary school
children: A meta-analysis. Educational Research Review, 10, 29–44. https://doi.org/10.1016/j.edurev.2013.05.003
Ginsburg, H. (1981). The clinical interview in psychological research on mathematical thinking: Aims, rationales, techniques. For the Learning of
Mathematics, 1(3), 4–11. https://flm-journal.org/Articles/7744A5621728F82314F0485A33BFF3.pdf
Grove, R. W. (1988). An analysis of the constant comparative method. International Journal of Qualitative Studies in Education, 1(3), 273–279.
https://doi.org/10.1080/0951839900030105a
Hackenberg, A. J. (2007). Units coordination and the construction of improper fractions: A revision of the splitting hypothesis. The Journal of
Mathematical Behavior, 26(1), 27–47. https://doi.org/10.1016/j.jmathb.2007.03.002

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.org/10.5951/jresematheduc.46.2.0196
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. https://doi.org/10.1016/j.jmathb.2009.04.004
Hmelo-Silver, C. E., Duncan, R. G., & Chinn, C. A. (2007). Scaffolding and achievement in problem-based and inquiry learning: A response to
Kirschner, Sweller, and Clark (2006). Educational Psychologist, 42(2), 99–107. https://doi.org/10.1080/00461520701263368
Hunt, J. H., Tzur, R., & Westenskow, A. (2016). Evolution of unit fraction conceptions in two fifth-graders with a learning disability: An exploratory
study. Mathematical Thinking and Learning, 18(3), 182–208. https://doi.org/10.1080/10986065.2016.1183089
Hunting, R. P. (1983). Alan: A case study of knowledge of units and performance with fractions. Journal for Research in Mathematics Education,
14(3), 182–197. https://doi.org/10.2307/748381
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. https://doi.org/10.5951/jresematheduc.43.4.0391
Kerrigan, S., Norton, A., & Ulrich, C. (2020). Ranking the cognitive demand of fractions tasks. 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 (pp. 2253–2261). Cinvestav/AMIUTEM/PME-NA. https://doi.org/10.51272/pmena.42.2020-383
Kieren, T. E. (1980). The rational number construct: Its elements and mechanisms. In T. E. Kieren (Ed.), Recent research on number learning
(pp. 125–150). ERIC/SMEAC.
Kirschner, P. A., Sweller, J., & Clark, R. E. (2006). Why minimal guidance during instruction does not work: An analysis of the failure of constructivist,
discovery, problem-based, experiential, and inquiry-based teaching. Educational Psychologist, 41(2), 75–86. https://doi.org/10.1207
/s15326985ep4102_1
Liang, B. (2021). Learning about and learning from students: Two teachers’ constructions of students’ mathematical meanings through student-
teacher interactions [Doctoral dissertation, University of Georgia]. ScholarWorksUGA. https://esploro.libs.uga.edu/esploro/outputs/994937
5152802959
Lovin, L. H., Stevens, A. L., Siegfried, J., Wilkins, J. L. M., & Norton, A. (2018). Pre-K–8 prospective teachers’ understanding of fractions: An
extension of fractions schemes and operations research. Journal of Mathematics Teacher Education, 21(3), 207–235. https://doi.org/10.1007
/s10857-016-9357-8
Lynch, S. D., Hunt, J. H., & Lewis, K. E. (2018). Productive struggle for all: Differentiated instruction. Mathematics Teaching in the Middle School,
23(4), 194–201. https://doi.org/10.5951/mathteacmiddscho.23.4.0194
Mack, N. K. (2001). Building on informal knowledge through instruction in a complex content domain: Partitioning, units, and understanding
multiplication of fractions. Journal for Research in Mathematics Education, 32(3), 267–295. https://doi.org/10.2307/749828
Morra, S. (1994). Issues in working memory measurement: Testing for M capacity. International Journal of Behavioral Development, 17(1), 143–159.
https://doi.org/10.1177/016502549401700109
Nemirovsky, R., & Ferrara, F. (2009). Mathematical imagination and embodied cognition. Educational Studies in Mathematics, 70(2), 159–174.
https://doi.org/10.1007/s10649-008-9150-4
Norton, A., Boyce, S., Phillips, N., Anwyll, T., Ulrich, C., & Wilkins, J. L. M. (2015). A written instrument for assessing students’ units coordination
structures. International Electronic Journal of Mathematics Education, 10(2), 111–136. https://doi.org/10.29333/iejme/295
Norton, A., & Nurnberger-Haag, J. (2018). Bridging frameworks for understanding numerical cognition. Journal of Numerical Cognition, 4(1), 1–8.
https://doi.org/10.5964/jnc.v4i1.160
Norton, A., & Wilkins, J. L. M. (2010). Students’ partitive reasoning. The Journal of Mathematical Behavior, 29(4), 181–194. https://doi.org/10.1016
/j.jmathb.2010.10.001
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
Núñez, R. E., Edwards, L. D., & Matos, J. F. (1999). Embodied cognition as grounding for situatedness and context in mathematics education.
Educational Studies in Mathematics, 39(1–3), 45–65. https://doi.org/10.1023/A:1003759711966
Olive, J. (1999). From fractions to rational numbers of arithmetic: A reorganization hypothesis. Mathematical Thinking and Learning, 1(4), 279–314.
https://doi.org/10.1207/s15327833mtl0104_2
Pascual-Leone, J. (1970). A mathematical model for the transition rule in Piaget’s developmental stages. Acta Psychologica, 32, 301–345. https://doi
.org/10.1016/0001-6918(70)90108-3
Pascual-Leone, J., Johnson, J., & Agostino, A. (2010). Mental attention, multiplicative structures, and the causal problems of cognitive development.
In M. Ferrari & L. Vuletic (Eds.), Developmental relations among mind, brain and education: Essays in honor of Robbie Case (pp. 49–82).
Springer. https://doi.org/10.1007/978-90-481-3666-7_3
Philipp, R. A., & Schappelle, B. P. (1999). Algebra as generalized arithmetic: Starting with the known for a change. The Mathematics Teacher, 92(4),
310–316. https://doi.org/10.5951/MT.92.4.0310
Piaget, J. (1970). Structuralism (C. Maschler, Trans.). Basic Books. (Original work published 1968)
Pitkethly, A., & Hunting, R. (1996). A review of recent research in the area of initial fraction concepts. Educational Studies in Mathematics, 30(1),
5–38. https://doi.org/10.1007/BF00163751
Pólya, G. (2004). How to solve it: A new aspect of mathematical method (Vol. 85). Princeton University Press. (Original work published 1945)
Reynolds, A., & Wheatley, G. H. (1996). Elementary students’ construction and coordination of units in an area setting. Journal for Research in
Mathematics Education, 27(5), 564–581. https://doi.org/10.2307/749848
Saenz-Ludlow, A. (1994). Michael’s fraction schemes. Journal for Research in Mathematics Education, 25(1), 50–85. https://doi.org/10.2307/749292
Simon, M. A., Kara, M., & Placa, N. (2018). Promoting reinvention of a multiplication-of-fractions algorithm: A study of the Learning Through
Activity research program. The Journal of Mathematical Behavior, 52, 174–187. https://doi.org/10.1016/j.jmathb.2018.03.007
Simon, M. A., Placa, N., Avitzur, A., & Kara, M. (2018). Promoting a concept of fraction-as-measure: A study of the Learning Through Activity
research program. The Journal of Mathematical Behavior, 52, 122–133. https://doi.org/10.1016/j.jmathb.2018.03.004
Steffe, L. P. (1991). Operations that generate quantity. Learning and Individual Differences, 3(1), 61–82. https://doi.org/10.1016/1041-6080(91)
90004-K

Steffe, L. P. (1992). Schemes of action and operation involving composite units. Learning and Individual Differences, 4(3), 259–309. https://doi
.org/10.1016/1041-6080(92)90005-Y
Steffe, L. P. (2001). A new hypothesis concerning children’s fractional knowledge. The Journal of Mathematical Behavior, 20(3), 267–307. https://doi
.org/10.1016/S0732-3123(02)00075-5
Steffe, L. P. (2003). Fractional commensurate, composition, and adding schemes: Learning trajectories of Jason and Laura: Grade 5. The Journal of
Mathematical Behavior, 22(3), 237–295. https://doi.org/10.1016/S0732-3123(03)00022-1
Steffe, L. P., Nesher, P., Cobb, P., Goldin, G. A., & Greer, B. (Eds.). (1996). Theories of mathematical learning. Erlbaum. https://doi
.org/10.4324/9780203053126
Steffe, L. P., & Olive, J. (2010). Children’s fractional knowledge. Springer. https://doi.org/10.1007/978-1-4419-0591-8
Steffe, L. P., von Glasersfeld, E., Richards, J., & Cobb, P. (1983). Children’s counting types: Philosophy, theory, and application. Praeger.
Streefland, L. (1993). Fractions: A realistic approach. In T. P. Carpenter, E. Fennema, & T. A. Romberg (Eds.), Rational numbers: An integration of
research (pp. 289–326). Erlbaum.
Stylianides, A. J., & Stylianides, G. J. (2008). Studying the classroom implementation of tasks: High-level mathematical tasks embedded in “real-life”
contexts. Teaching and Teacher Education, 24(4), 859–875. https://doi.org/10.1016/j.tate.2007.11.015
Swanson, H. L., & Beebe-Frankenberger, M. (2004). The relationship between working memory and mathematical problem solving in children at risk
and not at risk for serious math difficulties. Journal of Educational Psychology, 96(3), 471–491. https://doi.org/10.1037/0022-0663.96.3.471
Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257–285. https://doi.org/10.1207
/s15516709cog1202_4
Sweller, J. (2010). Element interactivity and intrinsic, extraneous, and germane cognitive load. Educational Psychology Review, 22(2), 123–138.
https://doi.org/10.1007/s10648-010-9128-5
Thalmann, M., Souza, A. S., & Oberauer, K. (2019). How does chunking help working memory? Journal of Experimental Psychology: Learning,
Memory, and Cognition, 45(1), 37–55. https://doi.org/10.1037/xlm0000578
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. Sepúlveda (Eds.), Proceedings of the joint meeting PME 32 and PME-NA XXX (Vol. 1, pp. 31–49).
Cinvestav-UMSNH.
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. https://doi.org/10.1016/j.jmathb.2013.03.006
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(4), 390–416. https://doi.org/10.2307/749707
Tzur, R., & Simon, M. (2004). Distinguishing two stages of mathematics conceptual learning. International Journal of Science and Mathematics
Education, 2(2), 287–304. https://doi.org/10.1007/s10763-004-7479-4
Ulrich, C. (2015). Stages in constructing and coordinating units additively and multiplicatively (Part 1). For the Learning of Mathematics, 35(3), 2–7.
https://flm-journal.org/Articles/6B439227432674A49489C6917444E.pdf
Ulrich, C. (2016). Stages in constructing and coordinating units additively and multiplicatively (Part 2). For the Learning of Mathematics, 36(1),
34–39. https://flm-journal.org/Articles/435C6B29393207D2E4957C48F3156.pdf
Ulrich, C., Tillema, E. S., Hackenberg, A. J., & Norton, A. (2014). Constructivist model building: Empirical examples from mathematics education.
Constructivist Foundations, 9(3), 328–339. https://constructivist.info/9/3/328.ulrich
Watanabe, T. (2006). The teaching and learning of fractions: A Japanese perspective. Teaching Children Mathematics, 12(7), 368–374. https://doi
.org/10.5951/TCM.12.7.0368
Wheatley, G. H. (1992). The role of reflection in mathematics learning. Educational Studies in Mathematics, 23(5), 529–541. https://doi.org/10.1007
/BF00571471
Wheatley, G. H., & Reynolds, A. (1996). The construction of abstract units in geometric and numeric settings. Educational Studies in Mathematics,
30(1), 67–83. https://doi.org/10.1007/BF00163753
Wilkins, J. L. M., & Norton, A. (2011). The splitting loope. Journal for Research in Mathematics Education, 42(4), 386–416. https://doi.org/10.5951
/jresematheduc.42.4.0386
Authors
Anderson Norton, Department of Mathematics, Virginia Tech, Blacksburg, VA 24060; norton3@vt.edu
Catherine Ulrich, School of Education, Virginia Tech, Blacksburg, VA 24061; culrich@vt.edu
Sarah Kerrigan, Mathematics Department, George Fox University, Newberg, OR 97132; skerrigan@georgefox.edu
Submitted September 8, 2021

Accepted January 18, 2022
doi:10.5951/jresematheduc-2021-0031

J RES MATH EDUC v54 No 4 pp240-259

Uploaded by

Document Information

Original Description:

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

J RES MATH EDUC v54 No 4 pp240-259

Uploaded by

Copyright:

Available Formats

Journal for

Norton, Anderson; Ulrich, Catherine; and Kerrigan, Sarah

DIGITAL OBJECT IDENTIFIER: VOLUME: ISSUE NUMBER:

Unit Transformation Graphs: Modeling Students’ Mathematics in

Anderson Norton et al.

Framework for Building Unit Transformation Graphs

Mental Actions for Constructing and Transforming Units

SA-NCTM-JRME230005.indd 241 5/2/2023 12:12:41 PM

Operations for Constructing and Transforming Units

Sequencing Operations in Working Memory

SA-NCTM-JRME230005.indd 242 5/2/2023 12:12:41 PM

Units Coordinating Structures and Stages

(A) Two-Level and (B) Three-Level Units Coordinating Structures

SA-NCTM-JRME230005.indd 243 5/2/2023 12:12:41 PM

SA-NCTM-JRME230005.indd 244 5/2/2023 12:12:41 PM

Hypothetical UTG for Task 1

SA-NCTM-JRME230005.indd 245 5/2/2023 12:12:41 PM

SA-NCTM-JRME230005.indd 246 5/2/2023 12:12:41 PM

Hypothetical UTGs for the Three Fractions Tasks

Hypothetical UTGs for (A) Task 2a and (B) Task 3

B Whole1 6 1/6 5/6

SA-NCTM-JRME230005.indd 247 5/2/2023 12:12:42 PM

Descriptive Statistics for Cognitive Demand

Summary of Relative Cognitive Demand by Task and PST

SA-NCTM-JRME230005.indd 248 5/2/2023 12:12:42 PM

UTGs for the PSTs

Meeting the Cognitive Demands of Task 1

Common UTG for Task 1

SA-NCTM-JRME230005.indd 249 5/2/2023 12:12:42 PM

SA-NCTM-JRME230005.indd 250 5/2/2023 12:12:42 PM

Two Distinct Ways of Operating in Task 2a

Determining 2/27 as a Part-Whole Relationship

Working From 1/27 as a Measure

SA-NCTM-JRME230005.indd 251 5/2/2023 12:12:42 PM

UTG for PST 10 in Response to Task 2a

SA-NCTM-JRME230005.indd 252 5/2/2023 12:12:42 PM

UTG for PST 23 in Response to Task 2a

An Attempt to Solve Task 3

SA-NCTM-JRME230005.indd 253 5/2/2023 12:12:42 PM

UTG for PST 22 in Response to Task 3

Units Coordinating Structures

SA-NCTM-JRME230005.indd 254 5/2/2023 12:12:43 PM

Implications for Future Research

SA-NCTM-JRME230005.indd 255 5/2/2023 12:12:43 PM

SA-NCTM-JRME230005.indd 256 5/2/2023 12:12:43 PM

SA-NCTM-JRME230005.indd 257 5/2/2023 12:12:43 PM

SA-NCTM-JRME230005.indd 258 5/2/2023 12:12:43 PM

Submitted September 8, 2021

SA-NCTM-JRME230005.indd 259 5/2/2023 12:12:43 PM

You might also like