REFLECTIONS ON MATHEMATICS EDUCATION

RESEARCH QUESTIONS IN ELEMENTARY NUMBER THEORY

Annie Selden
Tennessee Technological University
John Selden
Mathematics Education Resources Company

This monograph intentionally raises more questions than it answers. Indeed the editors'

aims included convincing readers that many interesting questions in the teaching and

learning of number theory await their attention, and that there are engaging number

theory tasks with which to investigate students' mathematical sense making.

Number theory appears to be a rather neglected area in the mathematics education

research literature. Whether one looks at the NCTM Handbook of Research on

Mathematics Teaching and Learning (Grouws, 1992) or the NCTM Yearbook,

Developing Mathematical Reasoning in Grades K-12 (Stiff, 1999), one finds almost

nothing on students' engagement with number theory topics. By number theory we mean,

as do the monograph editors, results concerning the structure of the integers that are not

primarily computational, for example, questions about factorization or the distribution of

primes. Whatever the reasons for this research lacuna, this monograph, devoted entirely

to various number theory investigations and reports, should help fill this void and

heighten the community's awareness of the potential for fruitful investigations. These

might concern individual students' cognitions on a variety of topics from divisibility and

prime factorization to number theoretic generic proofs and mathematical induction, or

they might be more socio-cultural studies.

 The authors of this monograph approach topics concerning the teaching/learning of

number theory from a broad range of perspectives and investigate university students

having various levels of mathematical sophistication from preservice elementary teachers

to second-year Cambridge University mathematics majors. Running across the

contributions, however, are several common threads upon which we will comment: the

potential of number theory for the teaching and learning of problem solving, reasoning

and proof; questions regarding the language and images of divisibility; philosophical

stances taken; theoretical frameworks used; and implications for teaching. In the spirit of

the monograph and drawing on it, we will focus primarily on questions and new

directions for investigation, some ranging well beyond number theory itself.

PROBLEM SOLVING, REASONING, AND PROOF

Elementary number theory seems an ideal place to learn about problem solving,

reasoning, generalization, abstraction, and proof because students from middle school

through university feel comfortable dealing with whole numbers. They are much more

familiar with the positive integers than with rational, real, or complex numbers.

Appropriately chosen elementary number theory problems can provide an excellent

opportunity for developing and evaluating mathematical arguments. The NCTM

Principles and Standards (2000) calls for such problem solving and argumentation in

both the Reasoning and Proof and the Number and Operations Standards for Grades 6-8.

Middle school students could, for example, explain why the sum of the digits of any

multiple of 3 is itself divisible by 3 (NCTM, 2000, p. 217). Also the Number and

Operations Standard for Grades 9-12 explicitly states that all students should "use

2
number-theory arguments to justify relationships involving the whole numbers." One

could ask: What can you say about the number that results when you subtract 1 from the

square of an odd integer? Students could easily verify that the number is divisible by 4

and conjecture that it is divisible by 8. Proving this would involve representing the

number as (2n  1) 2  1  4(n)(n  1) and recognizing that one of n or n  1 is even

(NCTM, 2000, pp. 290-292).

Teppo's vignette (this volume) shows how students can begin reasoning about the

number of divisors of small positive integers. Then, with scaffolding by the teacher

along with a suitable didactic object (a divisor table), students can conjecture rules and

explain why they hold. This can lead to subsequently considering the general question of

how to determine the number of divisors of any positive integer, given its prime

factorization. In general, however, care should be taken by teachers to ascertain whether

a proof is obtainable by elementary means. It is a well known "folk theorem" amongst

mathematicians that number theory questions and conjectures can be easy to state but

phenomenally hard to prove (e.g., about partitions or the distribution of primes). While

the ordered partition problem (Rowland, this volume) is a potentially rich classroom task,

finding regularity in numbers of (unordered) partitions is subtle and difficult and involves

an understanding of modular-forms. [See Peterson (2000) for a description of

mathematician Ken Ono's recent results on partition congruences.] A collection of

elementary number theory problems whose solutions/proofs are accessible by elementary

means, along with advice for teachers on potential student "pitfalls," would be a useful

pedagogical aid.

3
 Why might elementary number theory be a good place for university students to

learn about proofs? Apparently less abstraction is required to construct elementary

number theory proofs than say, abstract algebra proofs, and this can be an advantage for

students with a weak conception of proof. Currently in the U.S., university mathematics

majors, including preservice secondary teachers, can often avoid dealing with proofs until

a "bridge" course on proofs, often followed by an abstract algebra course. Such students'

entire previous experience may lead them to believe that mathematics deals mainly with

numbers and functions. Suddenly, in an abstract algebra course they need to treat

operations abstractly and see commonalties amongst radically different examples -- a

cause of considerable confusion.

Recently, while teaching abstract algebra to preservice secondary teachers and

mathematics majors, we observed some difficulties with abstraction. The students were

presented with various sets, together with operations. They were asked to determine

whether these were semigroups and contained an identity, a zero, or other idempotents.

The sets were subsets of the reals, 2  2 real matrices, or real functions. The students

"knew" that juxtaposition stood for the (appropriate) operation, yet several students

worked with juxtaposition as if it were ordinary multiplication. For example, the

condition for a zero element a of a semigroup S (for all x in S , ax  a and xa  a ) was

not readily rewritten in terms of other operations (e.g., as a  x  a and x  a  a for

addition, or as ( f oa )( x)  a( x) and (a o f )( x )  a ( x ) for function composition). One

student said, "When I see this [juxtaposition], I think it's multiplication." For these

students, juxtaposition should have been, but was not, functioning as a visual metaphor,

i.e., suggesting the operation worked a little, but not exactly, like ordinary multiplication.

4
It has been observed that in trying to cope with abstract algebra questions, university

students often attempt to reduce the level of abstraction by "folding back" to what they

know about numbers, thereby changing an unfamiliar problem into a more familiar one

(Hazzan, 1999). This kind of difficulty with abstraction often occurs while students are

also struggling to construct a conception of proof that is consistent with the standard

(mathematicians') concept. We suspect these two difficulties (concerning abstractions

and the nature of proof) interact, exacerbating both.

By contrast, in elementary number theory, students deal with objects (integers) and

operations (ordinary multiplication and addition) that are familiar to them. Hence, they

can concentrate on discovering and constructing proofs without being distracted by

simultaneously having to extend their conceptions of the operations and objects they are

studying. An example of an elementary number theory statement not involving excessive

abstraction is the following: A sequence of positive integers p, p  2, p  4 is called a

triple of primes if and only if all three of p, p  2, p  4 are primes. The only triple of

primes is 3, 5, 7. Our "bridge" course students have found this statement challenging, but

possible to prove. One student said, "The triple of primes problem kept me up until two

in the morning. I was in bed staring at the ceiling when inspiration struck around four in

the morning. It's not the kind of thing you do in calc."

Rowland (this volume) makes a convincing case for the use of generic proofs in

number theory but also notes (Footnote 12) that sometimes generic presentations seem

ruled out. One question to ask from a pedagogical point of view is whether a specific

generic proof is likely to be illuminating, even if one can easily find a suitable particular

case. For an example at the school level, consider the proof that subtracting 1 from the

5
square of an odd integer yields a number divisible by 8. It is hard to see that a generic

proof would be better understood beginning perhaps with 11 (instead of 2n  1 ) and

writing 112  1 = (2  5  1) 2  1 = 4  52  4  5  1  1 = 4  (52  5) = 4  5  (5  1) . Yet this seems

to follow Rowland's first four principles (this volume). The particular case, 11, is neither

trivial nor too complicated. After one has written 11= 2  5  1 , one can track 5 through the

stages of the argument. Invariant aspects of the proof transfer to values of n other than 5,

and the reasoning is constructive. By contrast, a generic proof of the digit-sum rule for

divisibility by 9 obtained by writing 864= 8 100  6 10  4 = (8  99  6  9) + (8  6  4) =

9(8 11  6)  (8  6  4) seems both illuminating and generalizable. At a more advanced

level, Rowland has developed an enlightening generic version of an elementary and

elegant proof that the Euler  -function is multiplicative. On the other hand, we rewrote

Bolker's (1970) textbook proof of the same result using m  3 and n  8 and concluded it

would not be illuminating. Thus, not every proof seems suitable, ex post facto, for

treatment via generic example, even following Rowland's principles, and we join him in

calling for a refinement of those principles.

On an individual level, the effectiveness of generic proofs depends on a student's

ability to see the features of the general argument in the particular case, i.e., on a student's

ability to make structural generalizations (Rowland, referring to Bills (1996)). Structural

generalization is a way of thinking similar to Harel's (this volume) process pattern

generalization for mathematical induction problems, in which students look for a pattern

relating a statement about one integer to the (same) statement about the next (or previous)

integer. Just one of Harel's student groups seems to have, on their own, come upon

6
process pattern reasoning for the sequence 2 , 2  2 , K ; a way of
2 2 2 ,

thinking subsequently encouraged by the instructor. Not all students seem to make such

generalizations naturally. Are there ways to help students become better at it? Harel

suggests the importance of repeated reasoning, the R in his DNR system. A carefully

structured sequence of problems together with suitable interventions by the teacher to

encourage internalization and interiorization (à la Harel) might well be helpful. Further

studies could help provide details which might help teachers engender such reasoning in

their students.

In summary, why is number theory ideal for introducing students to reasoning and

proof? (1) Students get to deal with familiar objects, thereby reducing the level of new

abstraction and the concomitant disequilbration. (2) When suitably selected, such proofs

are accessible, i.e., students need only reason from first principles together with a certain

amount of ingenuity. (3) Often number theory proofs have generic versions, allowing

students to see, or even to prove on their own, a general result after having seen a suitable

particular case.

QUESTIONS REGARDING THE LANGUAGE AND IMAGES OF DIVISIBILITY

Within mathematics, the concept of divisibility in the integers seems somewhat unusual

in having such a wide assortment of terms to describe different aspects of essentially the

same mathematical phenomenon: A is divisible by B, B divides A, B is a factor of A, B is

a divisor of A, and A is a multiple of B. In some textbooks for preservice teachers (e.g.,

Billstein, Libeskind, and Lott, 1984, p. 167 ) and for discrete mathematics (e.g., Epp,

1990, p. 164), it is simply stated that these five phrases are alternative ways of expressing

7
the fact that there is a unique integer C such that A=BC. But why are there so many

different terms? What aspects of the concept does each highlight? To what extent, does

the mere existence of so many essentially equivalent terms cause students confusion?

How do students cope with this variety of terms? What is the effect of students' own

alternative terms, e.g., B divides A evenly? We speculate on some of these issues, but

definitive answers need to come from research.

The words divisible, divides, and divisor emphasize those integer division situations

in which the remainder, R, in the division algorithm, A=BC+R, 0  R<B, is zero. In the

case of divisible, A's role is highlighted, whereas in the case of divides and divisor, B's

role is stressed. The terms factor and multiple emphasize two slightly different

multiplication situations. In the case of factor, it is the factorization, and to some extent

the prime factorization, of A that is being regarded, with B's role in that factorization

stressed. However with multiple, an Archimedean-like property (repeated addition) of

the integers may be being stressed, with emphasis on the fact that A=B+B+ L +B, where

there are C of the B's. Mathematically, the division algorithm, prime factorization, and

the Archimedean property stress somewhat different features of the positive integers, but

textbooks often only offer students a formal definition, i.e., the simple requirement that

there exist a unique C such that A=BC, together with illustrations and exercises. Yet in

selecting one term over another, it may sometimes appear that division is being stressed,

while at other times, multiplication. Also, sometimes A's role is emphasized, other times

B's. This, in itself, might cause students some confusion. This is not to say that

beginning university students should be introduced to all these nuances, but only that

there's probably more here than just five ways of saying exactly the same thing and that

8
the terminology used may influence what comes to mind at various times during

cognition.

In attempting clarification of tasks such as "Is M=33  52  7 divisible by 15?" or

"What is the remainder if you divide 6  147+1 by 6?", students seem to be trying to make

sense of the question in their own everyday language by asking the interviewers whether

"divides evenly" or "what's left over" is wanted (Campbell; Brown, Thomas, & Tolias;

Zazkis; this volume) . This suggests that some students may be bringing to mind various

visual images of division or multiplication. These may be partitive or quotitive as

suggested by Campbell (this volume), but might also be of other kinds. For example,

might the terms quotient and remainder bring to mind visual images such as

2 quotient
4 11 xxxx xxxx xxx 11  2  4  3
or   or   ?
8 2 groups of 4 3 left over
groups of 4 3 left over
3 remainder

In the first of these, it is possible to interpret the quotient as what sits on top and the

remainder as what's left at the bottom. One discrete textbook author uses these three

illustrations to introduce the division algorithm, apparently calling on students' everyday

conceptions of "left over" (Epp,1990, p. 173). It would be an interesting research

question to see what sorts of mental images of the division algorithm, or of prime

factorization, that university students bring with them and to what the degree they are

aware of using inner vision or inner speech.

9
 When one asks students such questions as "Is M=33  52  7 divisible by 15?" or "Is 34

 55  73  1318 a multiple of M?", not only is one asking them to discriminate between

multiplication and division and between integer and rational division, but one is also

introducing additional representational complexities. While the question "Is M=33  52  7

divisible by 5?" seemed relatively easy for many students like Alice because they could

see the 5 "in there" (Brown, et al., this volume), the question of divisibility by 15 often

found them resorting to their calculators to find M and divide by 15 to see whether "it"

would "go in evenly." Would asking whether M=32  5  7  15 is divisible by 15 have

elicited the same tendency to calculate M? We think not. This resort to an action

(procedure) seems to be an instance of "folding back" to the Primitive Knowing level

(Pirie & Kieren, 1994; Towers, Martin, and Pirie, 2000), i.e., to a portion of one's

mathematical knowledge with which one is comfortable and secure -- a way to deal with

the complexity of the situation. In addition, for preservice elementary teachers, it may be

that their view of doing mathematics consists largely of performing computations to get

answers, together with a propensity to view symbols such as + and  as commands to

calculate, further influencing their tendency toward a computational interpretation of

such tasks. Furthermore, to determine whether a specific number not given in prime

factored form, such as 344578, is divisible by and other number, say 457, it seems quite

natural to resort to long division or a calculator. While the formal definition of "B

divides A" is especially good for proving results, such as M|N and N|P together imply M|

P, it gives no means of deciding whether for specific integers there actually is a C such

that A=BC.

10
 We have noticed a slightly different image of divisibility used by some of our

"bridge" course students when attempting to construct proofs in elementary number

A
theory. They often want to convert "B divides A" into " is an integer." This leads to
B

contemplating properties of pairs of numbers, e.g., what can be said about a noninteger

times an integer? Such reflections do not seem as straightforward as working with the

equation A=BC, where A, B, C are integers. To see this, consider the following theorem:

If M divides N and M does not divide P, then N does not divide P. The essence of one

student's "proof" went as follows: "Since M divides N, there exists an integer K so that

N P
N=MK and so is an integer. Since M does not divide P, is not an integer. Then
M M

P P
 is not an integer. So N does not divide P." To complete the proof, the student
N MK

P
would have needed to argue that a noninteger such as divided by an integer such as K
M

A
is a noninteger. This image of divisibility results in the consideration of an object, ,
B

but one not as easily handled as the object A=BC, for which most students know

transformation rules. It would be interesting to investigate implications of this alternative

image of divisibility.

Sometimes in an attempt to answer questions conceptually, rather than procedurally

by resorting to their calculators, students in the interviews seemed to be using their

11
concept images, some which included an idea of "taking something out" (Campbell, this

volume). There seem to be several possible interpretations of this idea. One could "take

3  5 out" of M to produce 15 as Riva and Lanny did and then have difficulty interpreting

the role of 15 (Campbell, this volume). In the case of M, one "takes out" a factor of 15.

However, there are other possible interpretations of "taking out." Considering the case of

11 divided by 4, one might also think of "taking 4 out" of 11 twice, using a repeated

subtraction view of division, and having 3 "left over." Or, one might also "take 2 fours

out" of 11 and have "3 left over." Depending which image of "take out" comes to mind,

one might have a factor, a divisor, or a quotient, making it difficult to answer the question

correctly. In what ways do phrases like "take out" and "what's left over" influence

students' ways of looking at questions such as "Is M=33  52  7 divisible by 15?" While

students should learn to use mathematical terms precisely, they also need to unpack them

and call on relevant mental images and generic examples. If a term like "left over" calls

up an appropriate quotitive view of division, this would seem especially useful later for

preservice teachers.

Many other issues regarding divisibility have been raised by various chapter authors,

e.g., the necessity to distinguish the indivisible units of integer division from the

infinitely divisible units of rational division, as well as the importance of negotiating

successfully between the modular ( 2 10  1 ), fractional (10 1 2 ), and decimal (10.5)

representations of "21 divided by 2" (Campbell, this volume). All of these are ripe for

further investigation.

PHILOSOPHICAL POSITIONS

12
In the introduction, under Epistemological Orientation, Campbell and Zazkis remark on

the relevant philosophical views of the contributors to this monograph and, more

generally, of the mathematics education research community. The material in the

monograph, along with most research in mathematics education, reports on empirical

observations that can largely be understood without referring to the underlying

philosophical views of the authors. Nevertheless such views do play an important role in

research because they influence the kinds of questions that will be, or even can be, asked

as well as the kinds of studies regarded as providing acceptable warrants for answers to

these questions.

Campbell and Zazkis correctly point out that the research community in

mathematics education takes a constructivist view of learning. This entails two main

ideas. First, individuals construct their own knowledge from their experiences and what

they already know. Here "construct" is not synonymous with "invent" -- few people

would claim students invent all of mathematics themselves. "Construct" also refers to the

coordination and integration of new experiences with one's existing knowledge.

Constructing knowledge is often thought to depend on some form of reflection, i.e.,

thinking about one's actions (including mental actions or thoughts) and their effects. We

note that reflection requires focusing one's attention and suspect one's focus (James,

1910) plays a large role in learning. Perhaps the importance of focus in learning can be

most easily seen in situations involving its absence. Many people have had the

experience of looking up a phone number, dialing it, and moments later not being able to

remember it. No doubt their focus was on something other than the number, even though

they were sufficiently conscious of it to dial it correctly. Like dialing a phone number,

13
small mathematical tasks can be accomplished without much focus and it seems very

unlikely that such situations could lead to the construction of lasting conceptual

knowledge.

The role of focus in learning has not been well investigated in mathematics

education, although it could be. For example, one might study it by examining the effects

of tasks that incorporate distractions (i.e., that reduce focus) or by asking a number of

students to report on what they are thinking about at various times. It would be

interesting to know what students were focusing on during various kinds of lectures,

group problem-solving sessions, etc. It might also be interesting to know the effects of

focus on independent study (homework).

The idea that knowledge is individually constructed leads to studies of how this

occurs, that is, to studies of the mind. It provided legitimacy for research that had been

largely blocked as late as the mid twentieth century by behaviorism, the then dominant

view in psychology. The idea was that science should be very reliable and predictive,

rather like physics, and thus should depend only on directly observable phenomena. This

lead to concentrating on stimuli and responses, and eventually to the idea and study of

conditioning. Such work was very reliable, but of limited use in explaining mathematical

understanding, problem solving, learning, etc. More importantly, the logical positivist

views underlying behaviorism, which insisted on a level of reliability requiring direct

observations, prevented serious examination of mental phenomena because such

phenomena cannot be directly observed.

Returning to constructivism, the second major idea mentioned by Campbell and

Zazkis (this volume) is that "there is no way of knowing about any metaphysical or

14
mathematical realities independently of our own experience" or, more briefly, one cannot

objectively know about things in the external world. Constructivism that includes this

idea is called radical constructivism and only some of the mathematics education

community ascribe to it. It is often difficult to see the effect of radical constructivism in

individual research papers. Viewed as a whole however, much of the literature has little

to say about the relationship between an externally presented mathematical task and a

person's moment-to-moment cognition in executing it. This apparent neglect of the

boundary between "outside" and "inside" could be just a matter of chance, or it might

derive from a feeling that a person cannot really "know" about the "outside" task or that

all that really matters is the perception of the task -- basically radical constructivist points

of view. Seen in this way, radical constructivism may guide research to questions that do

not consider the boundary between "outside" and "inside" and be limiting in a way

similar to the way behaviorism was limiting some twenty or thirty years ago (although to

a lesser degree). For example, there is essentially no work on the role of mimicking in

constructing procedural knowledge even though most of what is assessed in K-14

mathematics teaching is procedural and the tendency to mimic is a major feature

distinguishing humans from other primates who favor emulation (Tomasello, 1997).

Brown, et al.'s interesting stage analysis of the construction of understanding from

procedural knowledge (this volume) illustrates this, in that nothing could be said about

the origins of that procedural knowledge.

The central idea of radical constructivism, i.e., that one cannot objectively know

about the external world, follows from the fact that one has a single set of senses and, as a

result, has nothing independent against which to test perceptions. Indeed the perceptions

15
themselves depend on one's own conceptions and knowledge, and thus inherently lack

objectivity. While these points seem irrefutable, there is more to say. The important

thing about human intelligence is not that people cannot objectively know about the

external world (in an absolute sense), but rather that they can know about it (in an

approximate way). Although people lack an automatic, biological way of objectively

knowing about the external world, they do gradually develop mental mechanisms that can

be used for reducing uncertainty about it. They are able to take several perspectives and

can call on previous experience. Observations may be revisited and the results of their

integration into a person's knowledge base can be questioned and adjusted (through

processes called disequilibration and reequilibration by Piaget). For these mechanisms to

yield a reliable, integrated, inner picture of (part of) the external world requires a way of

drawing inferences that is about as reliable as deductive reasoning. There appears to be

such a way, but it is not deductive reasoning itself because (two-valued) logic does not fit

uncertainty very well.

Radical constructivism does not take account of this perspective and implicitly

sees knowledge in an all or nothing way, i.e., as expressed by statements that have only

two truth values and that are combined using ordinary deductive reasoning and two-

valued logic. This view of knowledge and reasoning is very widely accepted, especially

in the West. However, in practice, people often appear to successfully use something like

logic and deductive reasoning with statements that are not quite certain, i.e., statements

describing a kind of approximate knowledge that radical constructivism appears to

neglect. We think approximate truth and reasoning can be precisely modeled. Doing so

16
might suggest a more moderate view of radical constructivism and thereby enlarge the

scope of research in mathematics education.

The uncertainty of statements can be viewed in two ways. One, that we will call

probabilistic uncertainty, views statements as two-valued, even though the value may not

be known. In the other, that we call synthetic uncertainty, the uncertainty is due to a

synthetic definition or concept (as opposed to the precise, analytic kind of definition

common in advanced mathematics). Synthetic definitions are at least partially

incomplete descriptions of existing things and introduce a kind of inherent uncertainty

into a statement. For example, if one says a country M is "fairly democratic," one does

not mean that out of ten copies of M, eight would be democratic and two would be

dictatorships. One is instead (partly) expressing uncertainty of the exact meaning of

"democratic." Indeed, in normal usage there might not be an exact meaning that is

capable in every case of sharply distinguishing democratic from nondemocratic, i.e.,

capable of supporting statements having only two truth values.

Often probabilistic uncertainty does not combine well with deductive reasoning

because valid arguments can yield conclusions with probabilities much smaller than the

probabilities of the premises. For example, in modus ponens, if the premises both have

probability at least x, 0  x  1 , then one can only be sure the conclusion has probability

at least 2x -1, e.g., premises with probability at least 0.8 only yield a conclusion with

probability at least 0.6. Thus, at this degree of probability, chains of several deductions

yield conclusions with probabilities too small to be informative. However, often a person

makes an observation of great (but not absolute) certainty. For example, a reader of this

monograph may be quite certain that he or she is actually reading it, as opposed to

17
dreaming, say with probability 0.99. This level of certainty works very well with modus

ponens because premises with probability at least 0.99 yield a conclusion with probability

at least 0.98. This is also true for modes tollens and a number of other valid arguments.

The effect of this is that modest chains of deductions yield "almost certain" conclusions

whenever the premises are "almost certain."

For synthetic uncertainty the situation is better. One might analyze this kind of

uncertainty by assigning statements a numerical valve between 0 (false) and 1 (true), just

as in the case of probabilistic uncertainty. But here it seems appropriate to take the value

of the statement "p and q" to be the minimum of the individual values of p and q, the

value of the statement "p or q" to be the maximum of the individual values, and the value

of "not p" to be the one minus the value of p. These values together with the operations

"and," "or," and "not" form a lattice with all of the properties of a Boolean algebra except

the existence of complements, i.e., "p or not p" might not have value 1, and "p and not p"

might not have value 0. Using these values and operations, modus ponens and modes

tollens and many other arguments valid in ordinary logic are valid in the sense that if the

premises are approximately true (to some degree), then so is the conclusion (to the same

degree). Indeed, any argument valid in the ordinary sense is also valid in this

approximate sense if all of the statements used in it are assumed to be either

approximately true or approximately false. In such situations, long chains of inferences

are reliable, i.e., do not lose truth value just as is the case with ordinary deductive

reasoning.

The above description of approximate reasoning under small degrees of

uncertainty suggests that a person can (tacitly) use it as part of the process of integrating

18
new information into his or her knowledge base and can very reliably know about the

external world, although perhaps not quite in the (two-valued, objective) sense referred to

in radical constructivism. Thus there appears to be no philosophical impediment to

investigations of the border between the external world and the inner mental world. In

particular, the relationship between a mathematical problem written on paper and the

corresponding inner speech version, that is normally produced when reading, can be

investigated. To what degree is this correspondence "accurate"? For example, one of our

college algebra students ignored the word "positive" in a test problem even after it was

pointed out to him. To what degree, and why, do such discrepancies occur? What

knowledge does a problem call to mind and by what mechanism? What part of solving a

problem is conscious? What part is mental and what part is external?

When solving a linear equation a person may report nothing conscious between

writing one line and writing the next. This suggests a significant part of the process is not

conscious. In addition, if immediately after a line of the solution is written, that line is

covered from the solver's view, then the solving process may become very difficult. This

suggests the actual written work is an integral part of the cognitive process of solving

certain problems. To understand the problem-solving process, it would be useful to

investigate the relationships between the nonconscious part, the conscious (mental) part,

the external physical part, and the way relevant information is brought to mind. Could

there be ways of writing when attempting the solution of a problem that enhance a

student's cognition?

These ideas extend far beyond the teaching and learning of elementary number

theory. However, number theory would be an ideal setting in which to investigate them.

19
 THEORETICAL FRAMEWORKS
Several monograph authors refer to interesting theoretical frameworks, i.e., collections of

specialized, interrelated concepts that help express their point of view. One of these,

APOS theory, is used by Ferrari (this volume) and is central in the analysis of Brown, et

al. (this volume). Campbell and Zazkis have previously relied on APOS theory and here

it becomes part of their view of number theory as a conceptual field, in the sense of

Vergnaud. While we agree this can be a fruitful point of view, we also think APOS

theory might be more widely applied, and to that end, we will sketch an extended view of

it.

APOS is an acronym for action, process, object, and schema, which together form

a classification system for ways an individual might conceive of a particular concept.

Here, in discussing an idea such as number or function, we will reserve the term concept

for a building block of "public knowledge" -- knowledge in a form that can be

communicated, defined, described, and written down. In contrast, we will speak of a

conception as a (roughly) corresponding building block of an individual's "private"

mental knowledge -- knowledge that can become conscious, but cannot be precisely

communicated or directly observed, only inferred. Because APOS has been described by

Brown, et al. (this volume), we will only recall it through an example. A student, on first

hearing about functions, might look at f ( x )  3 x  2 and, guided in a procedural way by

the (external) written expression, be able to say f ( x)  5 when x  1 . If he or she could

only work with specific examples and could not think about or discuss functions in

general, he or she would have an action conception. Repeating this kind of action a

20
number of times and reflecting on it can lead to the ability to think about functions

without having a specific function in mind, i. e., can lead to interiorizing the experience

into a process conception of function. With such a conception the student could think in

a general way about carrying out the required operations without having a specific

formula such as f ( x )  3 x  2 or might think of input and output, but still might not see

this process as a whole or single unit that could be acted upon. On (probably

considerable) reflection he or she might encapsulate his or her process conception of

function into an object conception, i.e., be able to think of it as a single unit that could be

acted upon.

While the idea of an object conception of function seems unproblematic --

functions can be regarded metaphorically as objects rather like physical objects -- Brown,

et al.'s "object conception of divisibility" seems counterintuitive because in the everyday

sense "divisible" is a property (of two numbers), rather than an object. In the context of

APOS theory, there seem to be two main ideas associated with "object." One of these,

perhaps intended by those referring to "cognitive object," is the idea that an object

conception is a mental structure that can be experienced in an undivided way, as a whole

without parts. The other is that an object conception is like a physical object -- something

that not only can be seen as an undivided whole, but can also be acted upon, can be an

element of a collection, etc. Separating these two ideas, wholeness and objectness,

provides an extended categorization that we see as potentially useful in studying the

development of conceptions.

We propose that concepts, and the associated conceptions, can be separated into

objects (numbers, groups, topological spaces), properties (prime, commutative, compact),

21
and activities (factoring, solving, dividing, integrating). Relations (less than, divides) can

be regarded as properties of two or more objects. In this categorization, objects can have

properties and be the subject of activities; properties can be associated with objects and

activities; and activities can transform objects to yield other objects. Thus "object" as

used in APOS theory might better be called entity (in this proposed categorization) with

its key feature being a unitary or gestalt nature that allows it to take on the role of object,

property, or activity. One might, for example, have an action, process, or entity

conception of the activity, factoring, or of the relation, divisible. This extended

framework allows distinctions between kinds of entities, however it has no effect on the

APOS view of schema and does not alter the central ideas behind APOS theory itself.

There appears to be an additional feature that a conception (either action, process,

or entity) may or may not have. This might be called efficacy and it permits fluent speech

and cognition. That is, an efficacious conception can be used in cognition (and speech)

without consciously "unpacking," or calling to mind, its defining aspects or the

experiences that went into its formation. This may appear to resemble the unitary nature

of entity conceptions, but is meant only to refer to the way a conception is used in speech

or cognition. The distinction between an efficacious conception and an entity conception

can be seen in the first transcript of Brown, et al. (this volume). There Adam appears to

be fluent with his conception of divisor because he begins calculating at once without

having to think about the meaning, despite only having an action or process, not an entity,

conception.

Among mathematicians, instances of both this efficacy feature of conceptions and

the lack of it are common. For example, most mathematicians have an efficacious, entity

22
conception of the concept "group;" they can understand or speak complex sentences

about groups without bringing the definition to mind. In contrast consider a somewhat

obscure mathematical definition from the 1950's: a clan is a compact connected

topological semigroup with identity. We suspect that many mathematicians would have

to consciously review the meaning of "clan" to fully comprehend, "In any clan there is a

minimal subclan connecting the identity to the minimal ideal." That is, they would not

have an efficacious conception of "clan" even though they would see it as an object.

On being given an unfamiliar abstract definition, mathematicians seem to treat it

logically as an entity, and to suspend or ignore the confusion associated with not having

either an efficacious or an entity conception, while they work procedurally at the action

or process level, accumulating the experiences needed to yield an efficacious entity

conception. Many university students appear unable to apply this "bootstrap" process and

may even be unaware that such a thing is possible. They may also avoid use of

conceptions which are not efficacious. We suspect that beginning number theory students

who are not fluent with one of their conceptions -- divides, multiple, or factor -- may

avoid using it and the symbolic representation it calls for, perhaps substituting more

familiar terminology of their own, e.g., divides evenly. It would be good to know the

degree to which fluency (efficacy) affects use.

Zazkis (this volume) remarks in her conclusion that many preservice elementary

teachers are reluctant to adopt formal mathematical terminology, e.g., persistently using

terms like "goes into" instead of "divides." In explanation, following Sfard, she suggests

that mathematical objects emerge in an interaction between metaphor and rigor. The

terminology we have introduced here allows us to suggests another explanation.

23
 We suspect that, as children, many of the students Zazkis is referring to

constructed a relation conception of divides based on an activity, namely that of long

division with remainder zero. For many, this conception would be efficacious and have

the sign (name, symbol) "divides evenly" (or some synonym), with "divides" treated as a

contraction of "divides evenly." While for young children such a conception might be

metaphorical because numbers might be seen as attached to specific sets, for adults

numbers have become objects in their own right and this conception seems neither

especially metaphorical nor unrigorous. Later, when the standard mathematical concept

of divides is introduced, the students may construct another relation conception, but one

based on an object, i.e., the defining equation A=BC, and be encouraged to use what to

them is approximately the same sign, "divides." It is hard to see how two different

conceptions with (essentially) the same sign could both be efficacious. Furthermore, it

seems unlikely that the first conception ("divides evenly"), or any conception, could be

eradicated -- short of waiting a very long time. To take an everyday example, can people

stop having a conception of car, or for that matter of unicorn? In this situation, students

might best be helped to construct a schema containing both conceptions so that when one

comes to mind, the other is also readily available. Fortunately, the two conceptions are

complementary. The one based on an activity can yield numerical results such as 15|

1080, while the object conception can be used to support reasoning and proofs.

In addition to APOS theory, Brown, et al. (this volume), introduce a second

theoretical framework inspired by the three stages Piaget (1978) employed to describe

children's progress from successfully executing a physical task without understanding to

doing so with understanding. Brown, et al., substitute a procedural mathematical task for

24
the physical task. They follow Piaget in taking indicators of understanding to include the

ability to explain why parts of the procedure work and the ability to predict the results of

carrying out parts of the procedure before actually doing so. This rather broad

perspective on understanding complements that of APOS theory, which has focused

mainly on the development of particular conceptions. Furthermore substituting a

procedural mathematical task for a physical task may turn out not to be a very radical

change. Much procedural work in mathematics appears to be partly physical in nature,

i.e., to depend on the partly physical acts of writing and reading in a way inextricably

integrated with the mental activity. For example, in solving an equation one of our

calculus students wrote "+ 4" so that the horizontal bar of the "+" was half complete

causing the inscription to resemble "- 14" after which the student erroneously used "- 14."

Such errors depend specifically on interpreting the most recent inscription and could not

happen if the written work was merely a (partial) transcription of otherwise wholly

mental activity. Writing and seeing what she had written seems to have been an essential

part of this student's cognition.

Using the three stage framework to examine the way procedural knowledge

supports understanding in elementary number theory allows Brown, et al., to give an

interesting picture of their preservice teachers' knowledge. But it also suggests a much

broader application. The teaching and learning of procedural knowledge has been

neglected in recent years by the mathematics education research community (Star, 2000).

Yet procedural knowledge is often most of what is assessed at the lower division

undergraduate level and some mathematicians believe it supports, and its mastery must

precede, the teaching of the corresponding conceptual understanding. While there is

25
evidence this last belief is inaccurate, that the mastery of procedures need not precede the

teaching of conceptual understanding (Heid, 1988), procedural knowledge does appear to

support the construction of conceptual understanding in two main ways. It is needed to

provide situations, patterns, and experiences upon which reflection can lead to the

construction of new conceptions and understandings. By way of illustration, one can

learn to execute the rules of the game of checkers (procedurally) after which one has an

opportunity to reflect and discern its strategies (Simon, Tzur, Heinz, & Kinzel, 2000). In

addition, some understanding appears to be constructed out of reflection on procedures

themselves -- the basis of the three stage framework. Perhaps one might think of a

procedure as providing the central core of one kind of explanatory schema.

Because of the nature of their data -- single interviews with preservice teachers --

Brown, et al., mainly describe the current state of their subjects' knowledge, not lasting

changes in knowledge that would indicate learning. However, in a number of transcripts

one can see students actively constructing understanding, e. g., Alice progresses from

Stage I to Stage II and Karyl moves through Stage II to Stage III. This suggests single

student teaching experiments consisting of sequences of probing interviews might be

designed specifically to encourage the construction and retention of new understandings.

By focusing as much on the interviewer as the student, it might be possible to identify

situations that facilitate students' progress from Stage I to Stage III, i.e., that facilitate the

construction of conceptual understanding from procedural knowledge. Elementary

number theory would be a good topic for such a study because it can be taught early,

minimizing the effects of prior knowledge and habits of mind so as to isolate situations

26
that facilitate conceptual learning. Any results, however, might be of interest far beyond

the teaching of elementary number theory.

IMPLICATIONS FOR TEACHING

This monograph contains a number of mathematically rich number-theoretic tasks. With

them, teachers could try to probe students' understanding or help them with

generalization and abstraction. It seems that through the use of an unexpected

representation, students can be disequilibrated and sometimes brought to reconsider their

knowledge. Asking what would be the remainder when M=33  52  7 is divided by 15

(Campbell, this volume) or whether M= 34  53  76 198 is divisible by 63 (Ferrari, this

volume) seem sufficiently novel that students first resort to their calculators, and when

pushed (or finding the calculator of no use), try other approaches.

The two problems: (1) Can you find a rule for determining the number of different

divisors of any number, given its prime factorization? (Teppo, this volume). (2) Given

an N  K grid, how many squares are crossed by the diagonal? (Edwards & Zazkis, this

volume), are requests for a generalization, with the accompanying need to consider

examples and make and test conjectures. Both problems can be used with preservice

elementary teachers as a way of introducing them to these important aspects of

mathematical reasoning. In a somewhat similar way, the six process pattern

generalization problems, e.g., find an upper bound for the sequence 2 , 2 2 ,

K (Harel, this volume), can prod at least some students to find

2 2 2 ,

27
relationships between consecutive terms and explain their reasoning (in the form of

quasi-induction) as a precursor to formal mathematical induction.

In addition to such tasks, ideas/principles for teaching mathematics can be drawn

from the chapters, although some of these can also be found elsewhere. Teppo's

classroom vignette can serve as a prototype for guiding students to think for themselves,

using appropriate teacher scaffolding and a suitable didactic object (divisor table), while

also leading to a more general question about numbers of divisors.

The interviews of Brown, et al. (this volume) were not meant as teaching sessions,

but seemed to help students develop insights. Perhaps with some modest adjustments

they could serve as a model for teaching the understanding of procedures on an individual

or small group level. Providing individual assistance is a small, but important, role of

many university mathematics teachers. The semi-structured nature of the interviews,

with lists of problems and questions, might prove helpful to teachers. Indeed, it might be

useful to circulate such interview material, together with annotations on student reactions,

amongst teachers as a form of pedagogical content knowledge.

The value of appropriate generic proofs for exposition in number theory was argued

convincingly, along with some general principles for selecting a particular case. There is

also the admonition that writing up a more formal proof can be a nontrivial task for

students (Rowland, this volume).

The idea that the type and sequence of problems is of "paramount pedagogical

importance" is not new, nor is the idea that such sequencing of tasks should follow

students' conceptual development (Harel, this volume). What is new is the somewhat

counterintuitive notion that teaching, at least for mathematical induction, should begin

28
with suitably chosen "harder" problems, rather than the more usual progression from

"easy" to "hard," as preconceived by textbook authors. Other factors that probably

influenced students' success in the mathematical induction teaching experiment were the

opportunity for repeated reasoning of an appropriate type (process pattern

generalization), the opportunity to clarify and rewrite one's solutions, and the request that

students summarize the most important ideas of the mathematical induction unit in their

own language, followed by the instructor's incorporation of these into a presentation in

precise mathematical language.

Elementary number theory tasks are a rich and accessible area with which to enhance

and probe students' mathematical sense making. Many fascinating research

directions/questions have been considered in this monograph. Some are specific to

elementary number theory, but others suggest using number theory tasks to investigate

more general aspects of mathematical learning and cognition.

REFERENCES

Bills, L. (1996). The use of examples in the teaching and learning of mathematics. In L.
Puig and A. Gutierrez (Eds.), Proceedings of the 20th Conference of the
International Group for the Psychology of Mathematics Education (Vol. 2, pp. 81-
88), University of Valencia, Spain.

Billstein, R., Libeskind, S., and Lott, J. (1984). A problem solving approach to
mathematics for elementary school teachers, 2nd edition. Menlo Park, California:
Benjamin Cummings Publishing Company.

Bolker, E. D. (1970). Elementary number theory: An algebraic approach. New York:

W. A. Benjamin, Inc.

Epp, S. S. (1990). Discrete mathematics with applications. Belmont, California:

Wadsworth Publishing Company.

Grouws, D. A. (Ed.). (1992). NCTM Handbook of research in mathematics teaching

29
 and learning, New York: Macmillan Publishing Company.

Hazzan, O. (1999). Reducing abstraction level when learning abstract algebra concepts.
Educational Studies in Mathematics, 40(1), 71-90.

Heid, M. K. (1988). Resequencing skills and concepts in applied calculus using the
computer as a tool. Journal for Research in Mathematics Education, 19, 3-25.

James, W. (1910). The principles of psychology. New York, NY: Holt.

National Council of Teachers of Mathematics (2000). Principles and standards for

school mathematics. Reston, VA: Author.

Peterson, I. (2000). The power of partitions: Writing a whole number as the sum of
smaller numbers springs a mathematical surprise. Science News, 157(25), 396-397.

Piaget, J. (1978). Success and understanding. Cambridge, Massachusetts: Harvard

University Press.

Pirie, S. E. B. & Kieren, T. E. (1994). Growth in mathematical understanding: How can

we characterise and how can we represent it? Educational Studies in Mathematics,
26, 165-190.

Simon, M. A., Tzur, R., Heinz, K. & Kinzel, M. (2000). Articulating theoretical
constructs for mathematical reasoning. In M. L. Fernández (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. 213-218), Columbus, OH:
ERIC Clearinghouse for Science, Mathematics, and Environmental Education.

Star, J. R. (2000). Re-"conceptualizing" procedural knowledge in mathematics. In M. L.

Fernández (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. 219-223), Columbus, OH: ERIC Clearinghouse for Science,
Mathematics, and Environmental Education.

Stiff, L. V. (1999). Developing mathematical reasoning in grades K-12: 1999 yearbook.

Reston, Virginia: National Council of Teachers of Mathematics.

Tomasello, M. (1997). Human see, human do. Natural History, 106(8), 45-47.

Towers, J., Martin, L., & Pirie, S. (2000). Growing mathematical understanding:
Layered observations. In M. L. Fernández (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. 225-230), Columbus, OH: ERIC
Clearinghouse for Science, Mathematics, and Environmental Education.

30
31