MATHEMATICAL THINKING AND LEARNING, 7(1), 27–50

Copyright © 2005, Lawrence Erlbaum Associates, Inc.

Advanced Mathematical-Thinking
at Any Age: Its Nature
and Its Development
Guershon Harel
Department of Mathematics
University of California, San Diego

Larry Sowder
Department of Mathematics and Statistics
San Diego State University

This article argues that advanced mathematical thinking, usually conceived as think-
ing in advanced mathematics, might profitably be viewed as advanced thinking in
mathematics (advanced mathematical-thinking). Hence, advanced mathemati-
cal-thinking can properly be viewed as potentially starting in elementary school. The
definition of mathematical thinking entails considering the epistemological and
didactical obstacles to a particular way of thinking. The interplay between ways of
thinking and ways of understanding gives a contrast between the two, to make clearer
the broader view of mathematical thinking and to suggest implications for instruc-
tional practices. The latter are summarized with a description of the DNR system
(Duality, Necessity, and Repeated Reasoning). Certain common assumptions about
instruction are criticized (in an effort to be provocative) by suggesting that they can
interfere with growth in mathematical thinking.

The reader may have noticed the unusual location of the hyphen in the title of this
article. We relocated the hyphen in “advanced-mathematical thinking” (i.e., think-
ing in advanced mathematics) so that the phrase reads, “advanced mathemati-
cal-thinking” (i.e., mathematical thinking of an advanced nature). This change in
emphasis is to argue that a student’s growth in mathematical thinking is an evolv-
ing process, and that the nature of mathematical thinking should be studied so as to

Requests for reprints should be sent to Guershon Harel, Department of Mathematics, University of
California, San Diego, San Diego, CA 92093–0112. E-mail: [email protected]
28 HAREL AND SOWDER

lead to coherent instruction aimed toward advanced mathematical-thinking. These

arguments are embodied in our responses to four questions:

1. What is meant by “mathematical thinking”?1

2. What are the characteristics of advanced mathematical-thinking?
3. What are concrete reasoning practices by which advanced mathemati-
cal-thinking can be enhanced?
4. What are concrete reasoning practices by which advanced mathemati-
cal-thinking can be hindered?

We address these questions, in turn, in the four sections comprising this article.
Our earlier research necessitated these questions in the context of mathematical
proof (cf. Harel, 2001; Harel & Sowder, 1998). In this article, however, we do not
restrict our discussion to the process of proving. Rather, we demonstrate our
claims in a range of mathematical contexts across the grade-level spectrum, to
demonstrate that advanced mathematical-thinking is not bound by ad-
vanced-mathematical thinking.

ARTICLE’S ORGANIZATION

This article is organized in three sections (followed by a brief conclusion

section):

1. Our definition of advanced mathematical thinking is based on an impor-

tant distinction between two categories of knowledge: ways of understanding
and ways of thinking. In the first section, we define these notions and discuss
several responses and solutions by students to illustrate them. In particular, we
show how “proof schemes” (what constitutes truth for an individual), “prob-
lem-solving approaches,” and “beliefs about mathematics” are instances of one’s
ways of thinking.
2. Our definition of advanced mathematical thinking also utilizes Brousseau’s
(1997) notion of epistemological obstacle. Therefore, in the second section, we
discuss this important notion to argue for our relativistic view of the property “ad-
vanced” and to discuss examples of epistemological obstacles involved in the de-
velopment of advanced mathematical thinking (relative to our definition).
3. Finally, in the third section, we point to general reasoning practices by which
advanced mathematical thinking can be advanced or hindered.

1We use the terms mathematical thinking, a way of mathematical thinking, or just a way of thinking

interchangeably, although we are always referring to a mathematical context.

 ADVANCED MATHEMATICAL-THINKING AT ANY AGE 29

WHAT IS MEANT BY “MATHEMATICAL THINKING”?

Underlying the analysis presented in this article is the fundamental premise that
humans’ mental actions, observable or inferred, are induced and governed by their
general views of the world, and, conversely, humans’ general views of the world
are formed by these actions. Our probe into the above four questions through the
lenses of this duality led to a distinction between two categories of knowledge:
ways of thinking and ways of understanding.
The particular meaning students give to a term, sentence, or text, the solution
they provide to a problem, or the justification they use to validate or refute an asser-
tion—are ways of understanding, whereas students’ general theories—implicit or
explicit—underlying such actions are ways of thinking. This distinction, to be
elaborated upon shortly, has been both essential and valuable for our research and
for its instructional implications. We have observed that teachers often form, at
least implicitly, cognitive objectives in terms of ways of thinking, but their efforts
to these teach ways of thinking are often counterproductive because their efforts do
not build on ways of understanding. Conversely, teachers often focus on ways of
understanding but overlook the goal of helping students construct effective ways of
thinking from these ways of understanding. This observation is the basis for the
Duality Principle, one of the fundamental principles that underlie the instructional
treatment employed in our teaching experiments (see Harel, 1998, 2001). We re-
turn to the Duality Principle in the third section of this article.

WAYS OF THINKING VERSUS WAYS

OF UNDERSTANDING

We describe the distinction between ways of thinking and ways of understanding

in the context in which it initially arose. Consider the following three central, often
interrelated, mathematical activities:

1. Comprehension of mathematical content, as when reading texts or listen-

ing to others.
2. Carrying out an investigation, as when solving a problem.
3. Establishing truth, as when justifying or refuting.

Although it is pedagogically useful to distinguish among the three activities,

cognitively they can easily be subsumed under item (2), problem solving; for com-
prehension and communication, as well as justifying or proving, are all prob-
lem-solving processes.
Corresponding to these three types of mathematical activities, the phrase, ways
of understanding, refers to
30 HAREL AND SOWDER

1. The particular meaning/interpretation a person gives to a concept, relation-

ships between concepts, assertions, or problems.
2. The particular solution a person provides to a problem.
3. The particular evidence a person offers to establish or refute a mathemati-
cal assertion.

Examples of ways of understanding for (1) include the following: A student

may read or say the words, “derivative of a function,” understanding the phrase as
meaning the slope of a line tangent to the graph of a function, as the best linear ap-
proximation to a function near a point, as a rate of change, etc. On the other hand, a
student may understand this concept superficially (e.g., “the derivative is nxn-1 for
xn”) or incorrectly (e.g., “the derivative is the quotient (f(x + h) - f(x)) / h”). Simi-
larly, a student may understand the concept of a fraction in different ways. For ex-
ample, the student may understand the symbol a / b in terms of unit fraction (a / b is
a 1 / b units); in terms of part-whole (a / b is a units out of b units); in terms of par-
titive division (a / b is the quantity that results from a units being divided equally
into b parts); in terms of quotitive division (a / b is the measure of a in terms of
b-units). All of these would be ways of understanding derivatives or ways of under-
standing fractions.
Examples of ways of understanding for (2)—particular methods of solving a
problem—can be seen in the following. A ninth-grade class was assigned the fol-
lowing problem:

Problem 1: A pool is connected to 2 pipes. One pipe can fill the pool in 20
hours, and the other in 30 hours. Assuming the water is flowing at a constant
rate, how long will it take the 2 pipes together to fill the pool?

Among the different solutions provided by the students, there were the following
four—each represents a different way of understanding.

Solution 1.1: In 12 hours the first pipe would fill 3/5 of the pool and the sec-
ond pipe the remaining 2/5. (The student who provided this solution accom-
panied it with a sketch similar to Figure 1. We return to this solution later in
the article).

Solution 1.2: It will take the 2 pipes 50 hours to fill the pool.

Solution 1.3: It will take the 2 pipes 10 hours to fill the pool.

Solution 1.4: It would take x hours. In one hour the first pipe will fill 1/20 of
the pool, whereas the second will fill 1/30. In x hours the first pipe would fill
 ADVANCED MATHEMATICAL-THINKING AT ANY AGE 31

FIGURE 1 Student’s sketch for Solution 1.1.

x/20 and the second, x/30. Thus, x/20 + x/30 = 1. (The student then solved
this equation to obtain x = 12.)

Examples of ways of understanding for (3)—justifying or refuting—include

the following justifications by prospective secondary teachers to the problem:

Problem 2: Prove that log(a1 · a2 … an) = log a1 + log a2 + … + log an for all
positive integers n.

Solution 2.1:
log(4 · 3 · 7) = log 84 = 1.924
log 4 + log 3 + log 7 = 1.924
log(4 · 3 · 6) = log 72 = 1.857
log 4 + log 3 + log 6= 1.857
Because these work, then log(a1 · a2 … an) = log a1 + log a2 + … + log an.

Solution 2.2:
i. log (a1a2) = log a1 + log a2 by definition
log (a1a2a3) = log a1 + log a2a3. Similar to log (ax) as in step (i), where this
time x = a2a3.

Then

ii. log (a1a2a3) = log a1 + log a2 + log a3

We can see from step (ii) any log (a1a2a3 … an) can be repeatedly broken
down to log a1 + log a2 + log an.

In our usage, the phrase way of understanding, conveys the reasoning one ap-
plies in a local, particular mathematical situation. The phrase way of thinking, on
the other hand, refers to what governs one’s ways of understanding, and thus ex-
presses reasoning that is not specific to one particular situation but to a multitude
of situations. A person’s ways of thinking involve at least three interrelated catego-
ries: beliefs, problem-solving approaches, and proof schemes.
32 HAREL AND SOWDER

BELIEFS—VIEWS OF MATHEMATICS

“Formal mathematics has little or nothing to do with real thinking or problem solv-
ing,” and “The solution of a problem should not take more than five minutes” are
detrimental common beliefs among students (Schoenfeld, 1985, p. 43). On the
other hand, in our work with undergraduate mathematics students, we found that
enabling beliefs such as “A concept can have multiple interpretations” and “It is
advantageous to possess multiple interpretations of a concept,” although essential
in courses such as linear algebra, are often absent from the students’ repertoires of
reasoning. The development of these ways of thinking should not wait until stu-
dents take advanced-mathematics courses, such as linear algebra. Elementary
school mathematics and secondary school mathematics are rich with opportunities
for students to develop these ways of thinking. For example, the different ways of
understanding fractions we just presented should provide such an opportunity to
develop the above ways of thinking for all elementary-grade students; likewise the
(correct) different ways of understanding derivatives should provide such an op-
portunity for secondary-school students who take calculus.

PROBLEM-SOLVING APPROACHES2

“Look for a simpler problem,” “Consider alternative possibilities while attempting

to solve a problem,” “Look for a key word in the problem statement” are examples
of problem-solving approaches. The latter way of thinking might have governed
the way of understanding expressed in Solution 1.2. Which ways of thinking might
have governed the other three solutions to Problem 1? Of particular interest is So-
lution 1.1. Only one student, G, provided this solution, and she was briefly inter-
viewed. G indicated that she drew a diagram—a rectangle to represent the pool
(Figure 1)—and divided it into 5 equal parts. Then she noticed that 3(20/5) is the
same as 2(30/5). G was unwilling (or unable) to answer the question of how she
thought to divide the rectangle into 5 equal parts, so we can only conjecture that a
juxtaposition of ways of thinking had driven G’s solution. These may have in-
cluded “Draw a diagram,” “Guess and check,” and “Look for relevant relationships
among the given quantities.” It was shocking to learn that G’s score on this prob-
lem, as well as on three other problems she solved in a similar manner (i.e., without
any “algebraic representation”) was zero. Her teacher’s justification for this score
was something to the effect that G did not solve the problems algebraically, with
unknowns and equations, as she was expected to do.

2We chose not to use the term “heuristics” here because, although every heuristic is a general ap-
proach to solving problems, the converse is not true. Heuristics are defined as “rules of thumb for effec-
tive problem solving” (Schoenfeld, 1985, p. 23); students’ approaches to solving mathematical prob-
lems—needless to say—are not always heuristics in this sense.
 ADVANCED MATHEMATICAL-THINKING AT ANY AGE 33

PROOF SCHEMES

Proving is defined in Harel and Sowder (1998) as the process employed by a per-
son to remove or create doubts about the truth of an observation, and a distinction
is made there between ascertaining for oneself and persuading others. A person’s
proof scheme consists of what constitutes ascertaining and persuading for that per-
son. Thus, proof schemes include one’s methods of justification. In this sense,
“proving” and “justification” are used interchangeably in this article. One of the
most ubiquitous proof schemes held by students is the inductive proof scheme,
where students ascertain for themselves and persuade others about the truth of a
conjecture by direct measurements of quantities, numerical computations, substi-
tutions of specific numbers in algebraic expressions, etc. (Harel & Sowder, 1998).
We found that this way of thinking governed the way of understanding expressed
in Solution 2.1 (Harel, 2001). The way of understanding expressed in Solution 2.2,
on the other hand, was found to be a manifestation of a different way of thinking,
called transformational proof scheme. In Harel (2001) it is shown why Solution
2.1 contains the three essential elements that characterize the transformational
proof scheme: (a) consideration of the generality aspects of the conjecture, (b) ap-
plication of mental operations that are goal oriented and anticipatory—an attempt
to predict outcomes on the basis of general principles—and (c) transformations of
images that govern the deduction in the evidencing process.3

WHAT ARE THE CHARACTERISTICS OF ADVANCED

MATHEMATICAL-THINKING?

It is clear that some ways of thinking are flawed (e.g., relying solely on empirical
observations to justify mathematical arguments, as we have seen in Solution 2.1;
over-generalizing mathematical ideas, as in the common inference students make:
“since 2(a + b) = 2a +2b is valid, then (a + b)2 = a2 + b2 must also be valid” [Matz,
1980]), although others are sound (e.g., looking for elegant solutions to problems;
generalizing mathematical ideas). But in what sense is “mathematical thinking”
advanced? Does “advanced” imply “effective,” “efficient,” or “elegant”? Is
nonadvanced mathematical thinking necessarily lacking or faulty? “Advanced”
implies there is also an “elementary.” If so, in what sense is “mathematical think-
ing” elementary? It is extremely difficult to characterize these properties, even if
we share an intuitive understanding of their meaning, and it is even more difficult
to build a taxonomy that differentiates among properties of mathematical thinking.
Yet it is of paramount importance to characterize qualities of mathematical think-
ing to

3For the full taxonomy of proof schemes, see Harel and Sowder (1998) and Harel (in press).
34 HAREL AND SOWDER

translate them into essential cognitive objectives—objectives that would position el-
ementary mathematics content for the successful subsequent learning of advanced
mathematical content. But what is the complete set of such ways of thinking? Is the
set a mere list, or does it have an underlying structure and is it guided by a small num-
ber of principles? Advanced mathematical-thinking research can and should take the
lead in answering these critical questions. (Heid, Harel, Ferrini-Mundy, & Graham,
2000, p. 35)

One goal of this article is to contribute to our understanding of these issues.

The term “advanced” implies that a developmental process is involved. “Ad-
vanced” is, therefore, not an absolute but a relative term, both in relation to a single
way of thinking and in relation to different ways of thinking. The attainment of a cer-
tain way of thinking is not all or nothing but gradual, and likewise, one might demon-
strate a high level of mastery of one way of thinking and little or none of another.
In addition to this relativistic view of the property “advanced,” we consider the
kind of obstacles one encounters in developing a way of thinking. We adopt
Brousseau’s distinction between didactical obstacles and epistemological obsta-
cles. The former are the result of narrow or faulty instruction, whereas the latter are
unavoidable due to the nature of the development of human knowledge
(Brousseau, 1997). But what are the criteria for determining whether the develop-
ment of a particular type of mathematical thinking necessarily involves
epistemological obstacles? Although this question itself requires serious re-
search—cognitive, historical, and epistemological—there already exist some cri-
teria with which to begin a debate on this question. Duroux (1982, cited in
Brousseau, 1997) lists necessary conditions for a piece of knowledge to be consid-
ered an epistemological obstacle. The first of Duroux’s conditions is that
epistemological obstacles have traces in the history of mathematics. The second
condition is that an epistemological obstacle is not a missing conception, or a lack
of knowledge; rather, it is a piece of knowledge or a conception that produces re-
sponses that are valid within a particular context, and it generates invalid responses
outside this context. To overcome the epistemological obstacle, one must construct
a notably different point of view. The third and last condition is that an
epistemological obstacle “withstands both occasional contradictions and the es-
tablishment of a better piece of knowledge. Possession of a better piece of knowl-
edge is not sufficient for the preceding one to disappear” (Brousseau, 1997, pp.
99–100).
These considerations—the relativistic view of the property “advanced,” and the
obstacles involved in the developmental process—led us to the following defini-
tion, which suggests a research agenda for determining ways of thinking that are
advanced, as well as the level of their development:

Mathematical thinking is advanced, if its development involves at least one

of the above three conditions for an obstacle to be epistemological. The level
 ADVANCED MATHEMATICAL-THINKING AT ANY AGE 35

of acquisition of a way of thinking by an individual is determined by the ex-

tent to which the individual has overcome these obstacles.

It should be noted that the first condition—that an epistemological obstacle must

have traces in the history of mathematics—is particularly problematic. It is difficult,
and in many cases it may not be possible, to establish whether an obstacle has mani-
fested itself in the history of mathematics. Many obstacles have likely occurred in the
historical development of mathematics but have never been observed by historians.
A ready example of an obstacle that satisfies one of Duroux’s conditions is the
transition from solely additive reasoning to proportional reasoning, a commonly
observed difficulty (see discussion following). Also, the notion of epistemological
obstacle applies to the construction of both ways of understanding and ways of
thinking. For example, the understanding of negative integers and imaginary num-
bers meets some, if not all three, criteria. The mathematical community of the time
(17th century) had to reconstruct—even revolutionize—its ways of thinking about
the concepts of number and quantity to accept these new constructs (Klein, 1968;
see also Kline, 1972, p. 252).
We conclude this section with two episodes to illustrate the above consider-
ations—not the definition per se.

THE RELATIVISTIC VIEW OF THE

PROPERTY “ADVANCED”

This consideration is discussed in the context of Problems 3–5 below, about a stu-
dent who can do proportional reasoning but is not yet able to reason in terms of
functional representation, and hence does not work in a mathematically efficient
fashion.
In a secondary mathematics lesson on exponential decay, the homework in-
cluded the problem:

Problem 3: The annual rate of inflation in a certain year is 8%. How much
will the dollar lose of its purchasing power during this year?

Student H’s solution was the following

Solution 3:
H: What costs $1 at the beginning of the year will cost $1.08 at the end of the
year. If a product costs $1 at the beginning of the year, that product would
cost $1.08 at the end of the year. We want to know how much of the product
we can buy for $1 at the end of the year. We are not going to be able to buy
the whole product for one dollar, only a portion of it. Let’s say we can buy x
of it for $1. Then [reasoning proportionally] 1/1.08 = x/1. x = 1/1.08 = 1/(1
36 HAREL AND SOWDER

+ 8/100) = 100/108. We can buy only 100/108 = 92.6% of the product. The
dollar lost about 7.4% of its purchasing power.

Following H’s presentation of her solution, the teacher introduced the following
generalization (without labeling it so):

Problem 4: The annual rate of inflation in a certain year is a%. How much
will the dollar lose of its purchasing power during this year?

The teacher went on to present the following solution:

Solution 4:
Teacher: As H said, a product that costs $1 at the beginning of the year
would cost $(1 = a/100) at the end of the year. Our goal is to find out how
much of the product we can buy for $1 at the end of the year. If x is the frac-
tion of the product we can buy for $1, then, as H did, x can be obtained from
the equation: 1/(1 + a/100) = x/1. Solving for x, we get: x = 1/(1 + a/100) =
100/(100 + a), or 100(100/(100 + a))%. Thus, if the annual inflation rate is
a%, then the dollar loses (100 – 10000/(100 + a))% of its purchasing power.

Following this work, the teacher discussed with the students the graph of the
function f(a) = 100 – 10000/(100 + a), and its physical (economic) meaning. Spe-
cifically, he discussed these questions: What are the roots of the function? Where is
it defined? What is the behavior of the graph of the function, and what is the eco-
nomic meaning of these behaviors (e.g., the economic meaning of a = –100, or an
annual rate of 100% deflation)?
The next set of homework included the following problem.

Problem 5: During one year, the dollar lost 12.7% of its value. What was the
annual rate of inflation during that year?

H applied a similar reasoning to that which she used to solve Problem 3:

Solution 5: At the end of the year, with $1 I can buy only (100 – 12.7)% =
87.3/100 of the product. The whole product would cost $y. y/1 =
1/(87.3/100). y = 100/87.3 ≈ 1.145. The annual rate of inflation is about
14.5%.

H’s solution involves an application of proportional reasoning—a sophisticated

way of thinking that warrants the label “advanced,” by our definition. First, addi-
tive reasoning—an antecedent to proportional reasoning—produces responses that
are valid within a particular context but generates invalid responses outside this
 ADVANCED MATHEMATICAL-THINKING AT ANY AGE 37

context. Indeed, the transition from additive reasoning to proportional reasoning

requires one to construct a different way of understanding relationships between
quantities. Second, research has shown that additive reasoning withstands “occa-
sional contradictions,” in that students continue to reason additively after they are
shown its inapplicability in certain situations. Finally, the “establishment of a
better piece of knowledge,” that of proportional reasoning, does not completely re-
move its application in multiplicative situations—students continue to use it after
they have been exposed to the concept of proportionality.
Going back to Solution 5, note that H did not realize that she could obtain
the solution by substituting 12.7 for f(a) and solve the equation, 12.7 =
100–10000/(100 + a) to obtain the annual rate of inflation a. When she was shown
this approach, she had difficulty comprehending it. The latter approach exempli-
fies a way of mathematical thinking that manifests, among other things, economy
of thought. “Economy of thought,” in this case, has to do with one’s ability to reify
Solution 4 into a “solution method.” It has been shown that reification is one of the
most complex processes in the conceptual development of mathematics—with the
individual (e.g., Dubinsky, 1991; Greeno, 1983; Harel & Kaput, 1991) and in the
history of mathematics (Sfard, 1992).
It is critical to emphasize that one cannot and would not appreciate the effi-
ciency of the latter solution if he or she has not gone, in various problematic situa-
tions, through an elaborated solution, such as that offered by H. Hence, although
we desire to label the functional solution as more advanced than the elaborated so-
lution, it may be unlikely that the former could be constructed without the latter. Of
course, the student’s background plays a critical role. For example, if a student un-
derstood functions before studying inflation, the function solution would likely be
easier for her. This raises a question that is important to curriculum development
and instruction: What possible instructional treatments can help H construct this
and other ways of advanced mathematical-thinking?
“Proportional reasoning” and “reification of a solution into a solution
method”—the two ways of mathematical thinking that emerged in the analysis of
this last problem—are examples of what we, as mathematics educators, feel
should be labeled “advanced.” This is so because we recognize that these develop
during a long period of intellectual effort and have proved essential and effective in
doing and creating mathematics. Proportional reasoning, for example, is indis-
pensable in many areas of mathematics, and it demands a reconceptualization of
mathematical reality—from a world that is organized solely according to additive
principles to a world that is organized according to a differentiation of additive
phenomena from multiplicative ones. Noelting (1980a, 1980b) found that even
among students who had had the usual instruction dealing with proportions, it was
quite common for the students to instead use a unit-rate thinking in working pro-
portion problems, a practice also observed among practicing teachers in the inter-
mediate grades (Harel & Behr, 1995). Lamon (1999) has identified several steps in
38 HAREL AND SOWDER

a possible development of proportional thinking, and Cai and Sun (2002) have de-
scribed the carefully planned development of proportion in a Chinese curriculum.
Arriving at a level of thinking that might be called genuine proportional thinking is
not just a matter of telling students about cross-multiplication.

OBSTACLES INVOLVED IN THE

DEVELOPMENTAL PROCESS

This consideration is discussed in the context of Problem 6 below. It shows a major

obstacle—to our knowledge little discussed in the literature—that students en-
counter in building the way of thinking of representing word problems algebra-
ically. The obstacle is not in forming a propositional representation of the problem;
rather, its roots seem to lie in the subtle distinction between “variable” and “un-
known”—a difficulty that might be appreciated through historical considerations.

Problem 6. Find a point on the number line whose distance from 1 is half its
distance from –4.

Solution 6. L, a prospective elementary school teacher, drew a number line

and marked on it the points, 1 and –4. After a long pause, L indicated that he
did not know what to do next. His teacher proceeded by asking him to de-
scribe the problem. In the process of doing so, L indicated—erroneously—
that the unknown point couldn’t be to the left of 1. It was clear from his de-
scription that he understood the problem. L’s argument about the location of
the unknown point—despite being erroneous—supports this claim.

Teacher: Very good. What is the distance between x and –4?

L: x plus 4
Teacher: Write that down, please.

L writes x + 4.

Teacher: And what is the distance between x and 1?

L: Half of x + 4

L writes (x + 4)/2.

Teacher: How else can you express the distance between x and 1?

L reads the problem again.

 ADVANCED MATHEMATICAL-THINKING AT ANY AGE 39

L: It says it is half the distance from –4.

At this point L was unable to express the distance between x and 1 in a differ-
ent way from (x + 4)/2.

In our experience, the difficulty of forming equations, as in this case, is common

among students. A possible conceptual basis for this difficulty is the following. For
an expert, a value x representing an unknown in a word problem would involve two
ways of understanding. One is expressed in the condition of the problem; the other
in the variability of the quantities involved. In our case the condition is “The dis-
tance of the unknown point x from 1 is half its distance from –4,” and the variability
is that of the functional expressions x + 4 and x – 1. These two ways of understand-
ing are independent of each other. In the former x is an unknown whereas in the lat-
ter it is a variable. There might be different explanations for L’s difficulty. L may
not have constructed these two ways of understanding, may have had difficulty co-
ordinating them, or once he constructed one way of understanding had difficulty
attending to the other.
The distinction between “variable” and “unknown” is likely to be more
epistemological than didactical—a claim that can be supported by the historical
development of the notion of “variable” in the 17th century. As we have discussed
earlier, for an obstacle to be epistemological it is necessary that it has occurred in
the historical development of mathematics.

WHAT ARE REASONING PRACTICES BY

WHICH ADVANCED MATHEMATICAL-THINKING
CAN BE ENHANCED?

Our answer to this question is an instructional treatment guided by a system of

learning-teaching principles, called the DNR system. The three chief principles of
the system are Duality, Necessity, and Repeated Reasoning. In this section we
briefly describe the first and last; the middle will be mentioned in the next section.
(For the complete description of the system, see Harel, 1998, 2001.)

The Duality Principle.

This principle asserts that
Students’ ways of thinking impact their ways of understanding mathematical con-
cepts. Conversely, how students come to understand mathematical content influences
their ways of thinking. (Harel, 1998, in press)

Clearly, one’s ways of thinking, both good and bad, influence one’s further
ways of understanding. A student whose way of thinking involves believing that a
40 HAREL AND SOWDER

mathematics story problem should be solved quickly by looking for a key word
and then waiting for a teacher’s reaction to the answer will certainly derive a differ-
ent way of understanding for story problems (i.e., will solve them differently) than
a student willing to spend several minutes making a drawing, looking for relation-
ships, and then striving for some sort of self-verification. The Duality Principle
asserts that the converse is also true, and so teachers and curriculum developers in
all grade levels should structure their instruction in a way that provides students
with opportunities to construct advanced mathematical-thinking from ways of
understanding.
There are powerful examples of the relationship of advanced mathemati-
cal-thinking in school mathematics to advanced-mathematical thinking. Consider
again the “multiple ways of understanding” we mentioned earlier.
Most students’ repertoires of reasoning do not include the way of thinking that
“A concept can be understood in different ways,” and that “It is often advantageous
to change ways of understanding of a concept when attempting to solve a prob-
lem.” The learning of linear algebra, an advanced-mathematical thinking topic, re-
quires multiple ways of understanding, for one must realize, for example, that
problems about systems of linear equations are equivalent to problems about ma-
trices, which, in turn, are equivalent to problems about linear transformations. Stu-
dents who are not equipped with these ways of thinking are doomed to encounter
difficulties. At the precollege level, there are various opportunities to help students
think in these ways. The list of ways of understanding fractions mentioned earlier
provides one such opportunity. Students should learn, for example, that the frac-
tion 3/4 can be understood in different ways: 3 individual objects, each of quantity
1/4; the result when 3 objects of the same size are shared among 4 individuals; the
portion of the quantity 4 that equals the quantity 3; and 3/4 as a mathematical ob-
ject, a conceptual entity, a number. Similarly, students should become comfortable
with the different ways in which many functions can be represented—table, graph,
equation, for example—and translations among these representations. Students
should also learn that depending on the nature of the problem, some interpretations
or representations are more advantageous than others. We believe that it is from
these kinds of ways of understanding that students construct the aforementioned
ways of thinking.

THE REPEATED REASONING PRINCIPLE

Research has shown that repeated experience, or practice, is a critical factor in en-
hancing, organizing, and abstracting knowledge (Cooper, 1991). The question is
not whether students need to remember facts and master procedures but how they
should come to know facts and procedures and how they should practice them.
This is the basis for the Repeated Reasoning Principle: “Students must practice
 ADVANCED MATHEMATICAL-THINKING AT ANY AGE 41

reasoning internalize and interiorize specific ways of thinking and ways of under-
standing” (Harel, 2001).
Consider again two important ways of thinking we mentioned earlier: “mathe-
matical efficiency” and “transformational proof scheme.”
Two elementary school children, S and T, were taught division of fractions. S
was taught in a typical method, where he was presented with the rule (a/b) ÷ (c/d) =
(a/b) ⋅ (d/c). The rule was introduced to him in a meaningful context and with a
mathematically correct justification that he understood, but was asked to repeat. T,
on the other hand, was presented with no rule but consistent with the duality princi-
ple and the repeated reasoning principle, she was always encouraged to justify her
mathematical actions. Each time she encountered a division of fractions problem,
she explained its meaning using her understanding of division of whole numbers as
the rationale for her solution. S and T were assigned homework problems to com-
pute divisions of fractions. S solved all the problems correctly, and gained, as a re-
sult, a good mastery of the division rule. It took T a much longer time to do her
homework. Here is what T—a real third-grader—said when she worked on (4/5) ÷
(2/3):

How many 2/3s in 4/5? I need to find what goes into both [meaning: a
unit-fraction that divides 4/5 and 2/3 with no remainders]. 1/15 goes into
both. It goes 3 times into 1/5 and 5 times into 1/3, so it would go 12 times into
4/5 and 10 times into 2/3. [She writes: 4/5 = 12/15; 2/3 = 10/15; (4/5) ÷ (2/3)
= (12/15) ÷ (10/15)]. How many times does 10/15 go into 12/15? How many
times do 10 things go into 12 things? One time and 2/10 of a time, which is 1
and 1/5.

T had opportunities for reasoning of which S was deprived. T practiced reasoning

and computation, S practiced only computation. Further, T eventually discovered
the division rule and learned an important lesson about mathematical efficiency—
a way of thinking S had little chance to acquire.
In Harel and Sowder (1998) we argued that a key to the concept of mathematical
proof is the transformational proof scheme—a scheme characterized by consider-
ation of the aspects of the conjecture, application of mental operations that are goal
oriented and anticipatory, and transformations of images as part of a deduction pro-
cess. The education of students toward transformational reasoning must not start in
college. Otherwise, years of instruction that focus on the results in mathematics,
rather than the reasons behind those results, can leave the impression that only the re-
sults are important in mathematics, an opinion sometimes voiced even by university
mathematics majors. We argued that instructional activities that educate students to
reason transformationally about situations are crucial to students’mathematical de-
velopment, and that these activities must begin at an early age.
42 HAREL AND SOWDER

The building of environments in which students regard the giving of reasons as

a natural part of mathematics is one of the more exciting aspects of some studies
with children in the primary grades (Carpenter, Franke, Jacobs, Fennema, &
Empson, 1998; Fuson et al., 2000; Maher & Martino, 1996; Yackel, Cobb, Wood,
Wheatley, & Merkel, 1990). Having discussions about which of 2.12 and 2.113 is
larger can reveal something important about the children’s ways of understanding,
and hence, have implications for their ways of thinking. Some may rely errone-
ously on the number of digits, a way of understanding that naturally develops with
whole number work. Or, in comparing 4.21 and 4.238, it may come out that some
students focus on the right-most place value and decide that 4.21 is larger because
hundredths are larger than thousandths (Resnick et al., 1989). Such discussions
would seem to be more valuable in the long run than practicing a teacher-given rule
about annexing zeros until each number has the same number of decimal places,
especially if the discussions led naturally to the rule.
Similarly, ready-made theorems, formulas, and algorithms, even when moti-
vated and completely proved, are often hastily introduced in undergraduate mathe-
matics courses. An interesting phenomenon was observed in our teaching experi-
ments (Harel, 2001; Harel & Sowder, 1998). It illustrates the importance of
practicing mathematical reasoning. Until a mathematical relationship was de-
clared a theorem, the students continued—either voluntarily when they needed to
use the relationship or upon request—to justify it. Once the relationship was stated
as a theorem, there seemed to be a reduced effort, willingness, and even the ability
of some of the students to justify it. This phenomenon was explained in terms of
the students’ authoritarian view of mathematics (another example of an undesir-
able, yet common, way of thinking): For them, the label “theorem” renders the re-
lationship into something to obey rather than to reason about. Or, possibly, in the
teaching experiment context these students had not practiced enough the reasoning
behind the theorem.

WHAT ARE REASONING PRACTICES BY

WHICH ADVANCED MATHEMATICAL-THINKING
CAN BE HINDERED?

Epistemological obstacles are perhaps more fascinating as objects of scholarly

study than didactical obstacles, but we must attend to the latter, for if narrow or
faulty instruction leads to problems in thinking or understanding, it should be
easier to correct such instruction than it may be to overcome an epistemological
obstacle.
Certain teaching practices are still in existence, and even widely used, despite
the consensus among mathematics education researchers that they lead to
didactical obstacles that are difficult to eradicate. The emphasis on “key” words in
 ADVANCED MATHEMATICAL-THINKING AT ANY AGE 43

instructing students on how to decide what operation to do in solving story prob-

lems is an example. Students learn that the phrase “all together” in a problem state-
ment should signal addition; “left” should signal subtraction; “per” should sum-
mon multiplication or division, etc. Such instruction, although well intentioned,
will give at best short-lived success, and will fail completely if problems are not al-
ways written to follow such guides (as in “Thirty rows, with 42 seats in each row,
will seat how many, all together?”). More important, these ways of understanding
would reinforce faulty ways of thinking—that in doing mathematics what counts is
the result, not the reasoning process.
It is fair to say that most instructional planning is a mix of art and science, with
art playing the major role. In an effort to be provocative, we challenge some of the
usual principles—in our view they are myths—that might guide one’s instruction.
Like the “key words” approach above, these principles may be helpful in the short
run, but may prove to be unhelpful or even counterproductive in the long run. Each
of them certainly merits research attention.

Myth 1: In sequencing instruction, start with what is easy. For example,

it is common to introduce methods of solving equations with examples like x + 2 = 7
and 3x = 15. Because these can be solved virtually by inspection, the students may
see no need for the usual canons for solving equations, and thus the Necessity Princi-
ple (Harel, 1998)—students are more likely to learn when they see a genuine need
(intellectual, not necessarily social or economic)—is violated. Much better first ex-
amples might be x + 75.6 = 211.3 and 1.7x = 27.2 or even 2.4x + 9.6 = 17.28, examples
not likely to be easily solved by inspection or guessing. In the same vein, perhaps a
treatment of congruent figures should start with complicated figures rather than the
usual congruence of segments, angles, and triangles. Dienes and Golding long ago
suggested that such a “deep-end” approach might be appropriate in many cases:

At first it is not always wise or useful to present a new mathematical concept in its
simplest form …. It has been found that, at least in some cases, it is far better to intro-
duce the new structure at a more difficult level, relying upon the child to discover the
less complex sections within the whole structure. (1971, p. 57)

Hence, a building-blocks metaphor in designing curriculum may not be the

most useful one, especially if the learner has no idea of the building that will even-
tually be finished. A more apt metaphor for designing curriculum might be based
on some sort of deep-end metaphor, perhaps starting with a picture of the building
and the question, “How would you build this?”

Myth 2. The best mental model is a simple one, preferably one quite fa-
miliar to the students. For example, instruction in linear algebra often uses co-
ordinate 2-D and 3-D geometry as the first examples of a vector space. Harel
44 HAREL AND SOWDER

(1999) argued that these examples constrain students’ understanding, so that they
think vector space ideas are just ideas about geometry: Linear algebra “=” geome-
try. Consequently they have difficulty dealing with nongeometric vector spaces.
He suggests that using systems of linear equations as a first way of understanding
vectors at least keeps the students’ thinking in an algebraic domain.
Here is another instance in which starting with the simplest situations may cre-
ate a didactical obstacle. Multiplication is always introduced as repeated addition;
this natural but confining approach seems to lead almost inexorably to the errone-
ous “multiplication makes bigger” idea (e.g., Fischbein, Deri, Nello, & Marino,
1985; Greer, 1987). Perhaps introducing multiplication as meaning “copies of”
would serve the students better (Thompson & Saldanha, 2003): 2 × 4 tells you how
many are in 2 copies of 4, and 2/3 × 6 tells you how many are in 2/3 of a copy of 6—
thus enveloping repeated-addition and fractional-part-of-an-amount interpreta-
tions into one way of thinking about multiplication. We do recognize that other
ways of understanding multiplication should also, and usually do, come up in the
mathematics curriculum. Such an instructional approach is needed to advance the
ways of thinking “a concept can have multiple interpretations” and “it is advanta-
geous to have multiple ways of understanding.”
In general, instruction that uses examples limited in some irrelevant or confin-
ing way runs the risk of over-generalization, with the irrelevant characteristic per-
haps becoming a part of the concept—everyone knows what to draw when asked to
draw an “upside-down” trapezoid (cf. Sowder, 1980). The first choices of exam-
ples may be crucial, as Marshall’s work (1995) with schemas for story problems
suggests.

Myth 3. In advanced undergraduate mathematics, begin with the axi-

oms. Starting with the basic rules of the game might seem sensible, but we feel
that the typical undergraduate student is not yet ready to play the game that way.
Our argument builds on our notion of “proof scheme” mentioned earlier—a proof
scheme guides what one does to convince oneself and to convince others (Harel &
Sowder, 1998). Our studies of the proof schemes of undergraduate mathematics
majors suggest that extensive earlier work entailing deductions by the student,
putting two or more results together to get a new result (deductive proof schemes)
must precede any meaningful work with axiomatic developments (axiomatic proof
schemes). Otherwise the student may just go through the motions, often rotely,
without any genuine appreciation of the development from axioms.

Myth 4. In school practice, use mathematical proofs to convince the stu-

dents that a mathematical result is certain. We know that an argument of
“but how can you be sure, without a proof” is often used, and that of course mathe-
maticians do look for arguments to assure themselves (and their referees) that the
result is indeed established. But mathematicians often look for more than certainty
 ADVANCED MATHEMATICAL-THINKING AT ANY AGE 45

in their proofs—What is the key to the result? Or, does a slight modification in the
proof suggest another result? Rav (1999) even claimed that mathematical knowl-
edge is embedded in the proofs, with the theorem only a “headline” (p. 20). But, to
repeat an earlier point, we have noticed that a proof for many students is either
something to ignore in favor of studying the result, or something only to be duti-
fully memorized for purposes of repetition on an examination. Indeed, labeling a
result with “theorem”—and that labeling alone—often means that the result is cer-
tain and requires nothing more, as we noted earlier.
We hypothesize that it is better to emphasize the reasoning, perhaps in several
examples, that a proof generalizes. The earlier example in which the child continu-
ally utilized a meaning-based argument for calculating divisions by fractions illus-
trates our point. Brownell (1956) emphasized that the quality of practice, rather
than just practice itself, was most important. Carefully planned practice could
guide the student’s thinking to a higher level. For example, the exercises in Figure
2 could precede, indeed could generate, the result about the relation between the
measures of vertical angles, at the same time they are providing practice with the
angle sum for a linear pair.
Here is another example of practice paving the way to a result. Suppose the tar-
get is one version of the fundamental theorem of calculus: Under certain condi-
b
tions on f, with F an antiderivative of F, ∫ f ( x) dx = F( b) − F( a). A common start-
a

ing point for this version is another version of the fundamental theorem:
d x
dx ∫a
f (t) dt = f ( x), again with conditions on f. Paraphrasing the latter gives that
x
the integral is an antiderivative of f(x). Hence, for example, ∫ (cos t) dt is the
2

antiderivative of cos x, or cin x + C. (Then the practice begins.) Therefore,

3 2
∫ (cos t) dt = sin 3 + C , but ∫ (cos t) dt = sin 2 + C = 0, so C = –sin2, and
2 2
3
∫
2
(cos t) dt = sin 3 + C = sin 3 − sin 2. Repetitions of the argument with other
integrals sets the stage for the general argument (= proof) that
b
∫ f ( x) dx = F( b) − F( a), with F(x) an antiderivative of f(x).
a

FIGURE 2 Practice leading to a general result.

46 HAREL AND SOWDER

Opportunity to Learn
The most serious didactical obstacle is a lack of opportunity to learn. In particular,
we have in mind the (good) ways of thinking and understanding mentioned earlier,
and the “habits of mind” of Cuoco, Goldenberg, and Marks (1996). Instruction (or
a curriculum) that ignores sense-making, for example, can scarcely be expected to
produce sense-making students. Computational shortcuts like “move the decimal
point” or “cross-multiply” or “invert and multiply” given as rules without any at-
tention as to why these work turns elementary school mathematics into what is de-
servedly called a bag of tricks. Also, students who never have a chance to make
conjectures cannot become more skilled at conjecturing—and it may be hypothe-
sized that students who have never conjectured do not see any need for mathemati-
cal proof. And so on.

Not an Exclusive Or
Whether a particular obstacle to learning is didactical or epistemological, in an
exclusive-or sense, is, we believe, too limiting. Harel (in press) offers the view
that an obstacle may be partly didactical and partly epistemological. Consider,
for example, “multiplication makes bigger” (MMB), the well-documented mis-
conception mentioned earlier that is an obstacle for many students (through col-
lege) in choosing an operation for solving a story problem (Greer, 1987). MMB
clearly meets Duroux’s partially-valid and obstinacy criteria, and one might ar-
gue that it also has historical roots, with multiplication probably first formalized
with whole numbers. Yet, MMB could perhaps have its influence allayed, if not
nullified, by some instructional modification like some more-inclusive view, say
the “copies of” interpretation mentioned earlier, or perhaps by exploring “what it
would be” via a calculator calculation of something like 0.2 × 15 or (1/2)x24 at
an age before extensive experience with whole numbers leads to MMB. Hence,
MMB might be positioned on a didactical versus epistemological set of axes as
in Figure 3.
In a similar way, one can conjecture difficulties with proportional reasoning,
with understanding (–1)(–1) = +1, with linear independence, or with some nota-
tional conventions like sin–1 x, as being both didactical and epistemological in na-
ture, as we have speculated in Figure 4.

SUMMARY

Our view is that the roots of mathematical thinking for advanced mathematics
must be fostered during the study of elementary mathematics. General ways of
thinking, built on rich ways of understanding in elementary mathematics, can then
symbiotically support further ways of understanding and of thinking in advanced
 ADVANCED MATHEMATICAL-THINKING AT ANY AGE 47

FIGURE 3 Multiplication makes bigger (MMB) as a mix of obstacles.

FIGURE 4 Hypothesized mixes of types of obstacles.

mathematics. Obstacles to ways of thinking and ways of understanding may be

epistemological and/or didactical, with didactical obstacles more easily identified
and perhaps more easily overcome than epistemological obstacles. We propose
that a way of mathematical thinking be called “advanced” if its development nec-
essarily involves at least one of the three necessary conditions for epistemological
obstacles identified by Duroux (1982, cited in Brousseau, 1997). An important
next step will be to identify ways of thinking that meet this criterion.
We endorse the DNR-based instruction for furthering ways of thinking and ways
of understanding: (Duality Principle) make the dually supportive roles of ways of
thinking and ways of understanding a conscious, carefully planned part of the cogni-
tive objectives for coursework in mathematics; (Necessity Principle) build instruc-
tion via problems that contain intellectual appeal to the students; and (Repeated Rea-
soning Principle) involve repeated reasoning to give a firm foundation for ways of
48 HAREL AND SOWDER

thinking and ways of understanding. Finally, by labeling them “myths,” we offer a

critique of some teaching “axioms” that have face validity but might actually hinder
the development of fruitful ways of thinking and ways of understanding.

ACKNOWLEDGMENTS

Some of the ideas presented in this article are the result of National Science Foun-
dation Project 9355861; opinions expressed here, of course, are those of the au-
thors and are not necessarily those of the Foundation. We also acknowledge with
gratitude the clarifying conversations with Alfred Manaster, and the suggestions
from the Advanced Mathematical Thinking Group of PME–NA, particularly those
at the Tucson meeting: Stacy Brown, Walter Houston, Kathy Ivey, Barbara Loud,
Denise Mewborn, and Sharon Walen.

REFERENCES

Brousseau, G. (1997). Theory of didactical situations in mathematics (N. Balacheff, M. Cooper, R.

Sutherland, & V. Warfield, Eds. & Trans.). Dordrecht, The Netherlands: Kluwer Academic.
Brownell, W. A. (1956). Meaning and skill—Maintaining the balance. Arithmetic Teacher, 3, 129–136.
Cai, J., & Sun, W. (2002). Developing students’ proportional reasoning: A Chinese perspective. In B.
Litwiller & G. Bright (Eds.), Making sense of fractions, ratios, and proportions (pp. 195–205).
Reston, VA: National Council of Teachers of Mathematics.
Carpenter, T. C., Franke, M. L., Jacobs, V. R., Fennema, E., & Empson, S. (1998). A longitudinal study
of invention and understanding in children’s multidigit addition and subtraction. Journal for Re-
search in Mathematics Education, 29, 3–20.
Cooper, R. (1991). The role of mathematics transformation and practice in mathematical development.
In L. P. Steffe (Ed.), Epistemological foundations in mathematical experience (pp. 102–123). New
York: Springer-Verlag.
Cuoco, A., Goldenberg, E. P, & Marks, J. (1996). Habits of mind: An ongoing principle for mathemat-
ics curricula. Journal of Mathematical Behavior, 15, 375–402.
Dienes, Z. P., & Golding, G. W. (1971). Approach to modern mathematics. New York: Herder and
Herder.
Dubinsky, E. (1991). Reflective abstraction in advanced mathematical thinking. In D. Tall (Ed.), Ad-
vanced mathematical thinking (pp. 95–102). Dordrecht, The Netherlands: Kluwer Academic.
Fischbein, E., Deri, M., Nello, M. S., & Marino, M. S. (1985). The role of implicit models in solving
verbal problems in multiplication and division. Journal for Research in Mathematics Education, 16,
3–17.
Fuson, K. C., De La Cruz, Y., Smith, S. T., Cicero, A. M. L., Hudson, K., Pilar, R., et al. (2000).
Blending the best of the twentieth century to achieve a mathematics equity pedagogy in the
twenty-first century. In M. J. Burke & F. R. Curcio (Eds.), Learning mathematics for a new century
(pp. 197–212). Reston, VA: National Council of Teachers of Mathematics.
Greeno, J. (1983). Conceptual entities. In D. Genter & A. L. Stevens (Eds.), Mental models (pp.
227–252). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.
 ADVANCED MATHEMATICAL-THINKING AT ANY AGE 49

Greer, B. (1987). Nonconservation of multiplication and division involving decimals. Journal for Re-
search in Mathematics Education, 18, 37–45.
Harel, G. (1998). Two dual assertions: The first on learning and the second on teaching (or vice versa).
The American Mathematical Monthly, 105, 497–507.
Harel, G. (1999). Students’ understanding of proofs: A historical analysis and implications for the
teaching of geometry and linear algebra. Linear Algebra and Its Applications, 302–303,
601–613.
Harel, G. (2001). The development of mathematical induction as a proof scheme: A model for
DNR-based instruction. In S. Campbell & R. Zazkis (Eds.), Learning and teaching number theory
(pp. 185–212). Westport, CT: Ablex.
Harel, G. (in press). Students’ proof schemes revisited: Historical and epistemological considerations.
In P. Boero (Ed.), Theorems in school. Dordrecht, The Netherlands: Kluwer Academic.
Harel, G., & Behr, M. (1995). Teachers’ solutions for multiplicative problems. Hiroshima Journal for
Research in Mathematics Education, 3, 31–51.
Harel, G., & Kaput, J. (1991). The role of conceptual entities and their symbols in building advanced
mathematical concepts. In D. Tall (Ed.), Advanced mathematical thinking (pp. 65–79). Dordrecht,
The Netherlands: Kluwer Academic.
Harel, G., & Sowder, L. (1998). Student’s proof schemes: Results from exploratory studies. In A. H.
Schoenfeld, J. Kaput, & E. Dubinsky (Eds.), Research in collegiate mathematics education. III (pp.
234–283). Providence, RI: American Mathematical Society.
Heid, M. K., Harel, G., Ferrini-Mundy, J., & Graham, K. (2000). Advanced mathematical thinking: Im-
plications of various perspectives on advanced mathematical thinking for mathematics education re-
form. In M. L. Fernandez (Ed.), Proceedings of the 22nd Annual Meeting of the North American
Chapter of the International Group for the Psychology of Mathematics Education (Vol. 1, pp.
33–37). Columbus, OH: ERIC Clearinghouse for Science, Mathematics, and Environmental Educa-
tion.
Klein, J. (1968). Greek mathematical thought and the origin of algebra. New York: Dover.
Kline, M. (1972). Mathematical thought from ancient to modern times. New York: Oxford University
Press.
Lamon, S. J. (1999). Ratio and proportion: Elementary didactical phenomenology. In J. Sowder & B.
Schappelle (Eds.), Providing a foundation for teaching mathematics in the middle grades (pp.
167–198). Albany: State University of New York Press.
Maher, C. A., & Martino, A. M. (1996). The development of the idea of mathematical proof: A 5-year
case study. Journal for Research in Mathematics Education, 27, 194–214.
Marshall, S. P. (1995). Schemas in problem solving. New York: Cambridge University Press.
Matz, M. (1980). Toward a computational theory of algebraic competence. Journal of Mathematical
Behavior, 3, 93–166.
Noelting, G. (1980a). The development of proportional reasoning and the ratio concept, Part I—Differ-
entiation of stages. Educational Studies in Mathematics, 11, 217–253.
Noelting, G. (1980b). The development of proportional reasoning and the ratio concept, Part II—Prob-
lem-Structure at successive stages; problem-solving strategies and the mechanism of adaptive re-
structuring. Educational Studies in Mathematics, 11, 331–363.
Rav, Y. (1999). Why do we prove theorems? Philosophia Mathematica, 7, 5–41.
Resnick, L. B., Nesher, P., Leonard, F., Magone, M., Omanson, S., & Peled, I. (1989). Conceptual bases
of arithmetic errors: The case of decimal fractions. Journal for Research in Mathematics Education,
20, 8–27.
Schoenfeld, A. H. (1985). Mathematical problem solving. Orlando, FL: Academic Press.
Sfard, A. (1992). Operational origins of mathematical objects and the quandary of reification—The case
of function. In E. Dubinsky & G. Harel (Eds.), The concept of function: Aspects of epistemology and
pedagogy (pp. 59–84). MAA Notes No. 25. Washington, DC: Mathematical Association of America.
50 HAREL AND SOWDER

Sowder, L. (1980). Concept and principle learning. In R. J. Shumway (Ed.), Research in mathematics
education (pp. 244–285). Reston, VA: National Council of Teachers of Mathematics.
Thompson, P. W., & Saldanha, L. (2003). Fractions and multiplicative reasoning. In J. Kilpatrick, W. G.
Martin, & D. Schifter (Eds.), A research companion to “Principles and Standards for School Mathe-
matics” (pp. 95–113). Reston, VA: National Council of Teachers of Mathematics.
Yackel, E., Cobb, P., Wood, T., Wheatley, G., & Merkel, G. (1990). The importance of social interaction
in children’s construction of mathematical knowledge. In T. J. Cooney & C. R. Hirsch (Eds.),
Teaching and learning mathematics in the 1990s (pp. 12–21). Reston, VA: National Council of
Teachers of Mathematics.