On The Dual Nature of Mathematical Conce

On the Dual Nature of Mathematical Conceptions: Reflections on Processes and Objects as
Different Sides of the Same Coin

Author(s): Anna Sfard
Source: Educational Studies in Mathematics, Vol. 22, No. 1 (Feb., 1991), pp. 1-36
Published by: Springer
Stable URL: http://www.jstor.org/stable/3482237 .
Accessed: 29/04/2013 02:13
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
Springer is collaborating with JSTOR to digitize, preserve and extend access to Educational Studies in
Mathematics.
http://www.jstor.org
This content downloaded from 132.74.40.78 on Mon, 29 Apr 2013 02:13:37 AM

All use subject to JSTOR Terms and Conditions
ANNA SFARD
ON THE DUAL NATURE OF MATHEMATICAL

CONCEPTIONS: REFLECTIONS ON PROCESSES AND
OBJECTS AS DIFFERENT SIDES OF THE SAME COIN
ABSTRACT. This paper presents a theoretical framework for investigating the role of
algorithms in mathematical thinking. In the study, a combined ontological-psychological
outlook is applied. An analysis of different mathematical definitions and representations
brings us to the conclusion that abstract notions, such as number or function, can be
conceived in two fundamentallydifferentways: structurally- as objects, and operationally-
as processes.These two approaches,although ostensiblyincompatible,are in fact complemen-
tary. It will be shown that the processes of learning and of problem-solvingconsist in an
intricate interplay between operational and structuralconceptions of the same notions.
On the grounds of historical examples and in the light of cognitive schema theory we
conjecturethat the operational conception is, for most people, the first step in the acquisition
of new mathematicalnotions. Thorough analysis of the stages in concept formation leads us
to the conclusion that transition from computational operations to abstract objects is a long
and inherentlydifficultprocess, accomplishedin three steps: interiorization,condensation,and
reification. In this paper, special attention is given to the complex phenomenonof reification,
which seems inherentlyso difficultthat at certain levels it may remainpracticallyout of reach
for certain students.
INTRODUCTION
It is more than eight decades now, since the well-knownFrench mathemati-

cian and philosopher Henri Poincare wrote in obvious despair:
One ... fact must astonish us, or rather would astonish us if we were not too much
accustomed to it. How does it happen that there are people who do not understand
mathematics?If the science invokes only the rules of logic, those accepted by all well-formed
minds ... how does it happen that there are so many people who are entirely imperviousto
it? (Poincare, 1952, p. 49; French original was published in 1908).
For all the knowledge accumulatedby psychologistsand educatorssince

then, this question seems today as challenging and teasing as ever. The
particular intricacy of mathematical thinking, the ubiquitous, sometimes
insurmountable difficulty experienced by those who learn it, and the
resultingpersistentlack of success in teaching the subject - all these facts
are not less puzzlingthan they are conspicuous.For the last severaldecades
ever growing resourceshave been invested in a search for an improvement
in mathematicsteaching. The results, however, are still far from satisfac-
tory - the sought-after solution seems to be as elusive as a cure for a
common cold.
EducationalStudies in Mathematics22: 1-36, 1991.
? 1991 KluwerAcademicPublishers.Printed in the Netherlands.

2 ANNA SFARD
There is probablymuch more to mathematicsthan just the rules of logic.

It seems that to put our finger on the source of its ostensibly surprising
difficulty,we must ask ourselves the most basic epistemologicalquestions
regardingthe natureof mathematicalknowledge.Indeed, since in its inacces-
sibility mathematics seems to surpass all the other scientific disciplines,
there must be something really special and unique in the kind of thinking
involved in constructing a mathematical universe. Saying what people
usually say, namely that mathematicsis the most abstractof sciences, does
not help very much. Being almost a cliche, this claim has little explanatory
power. The real question which should be asked here is qualitative rather
than quantitative: How does mathematical abstraction differ from other
kinds of abstraction in its nature, in the way it develops, in its functions
and applications?
The question itself is certainlynot new. The turn-of-the-centurycrisis in
mathematics forced mathematicians themselves into philosophical dis-
course on the most fundamentalquestions regardingthe nature of mathe-
matical thought. Within the framework of Piaget's genetic epistemology,
developed several decades later, it became possible to approach the same
problemspsychologically.But up to now, not enough has been done in the
directionof unifiedtheory which would addressphilosophy and psychology
of mathematics simultaneously,and would take an equal care of mathe-
matical thinking and of mathematicalthought - of both the process and
the product. In this context, the almost total neglect of advanced mathe-
matics is especiallyregrettable,since the advancedtopics are those in which
the difference between mathematics and other sciences becomes most
evident and the peculiaritiesof abstract thought can be observed in their
purest form.
It seems that the philosophical insight into the nature of mathematical
concepts is what we need in order to understandin depth the psychological
processes in which such concepts emerge. In the suggested kind of investi-
gation, epistemological and ontological analysis of "the stark, atemporal,
fornal universe of ideal [mathematical]knowledge"would hopefully shed
some light on the roots of this overwhelmingconfusion which only too
often seems to reign in "the organic, interior,processualuniverseof human
knowing" (Kaput, 1979).
1. THE DUAL NATURE OF MATHEMATICAL CONCEPTIONS
Peculiarityof mathematicalthinkinginvestigatedthroughreflectionson the

epistemological and ontological status of mathematicalconstructs is our

DUAL NATURE OF MATHEMATICALCONCEPTIONS 3
subject in this paper. Depending on the point of view assumed at a given

moment, two differentwords will be used to denote the building blocks of
mathematics(or any other science, for this matter): the word "concept"
(sometimes replaced by "notion") will be mentioned whenever a mathe-
matical idea is concernedin its "official"form - as a theoreticalconstruct
within "the formal universe of ideal knowledge"; the whole cluster of
internal representationsand associations evoked by the concept - the
concept's counterpartin the internal, subjective"universeof human know-
ing" - will be referredto as a "conception".
First, let us have a look at the world of mathematics,as it expressesitself
through formal descriptionsand representations.
As far as language is concerned, similaritiesbetween mathematicsand
other sciences seem more striking than the differences.Indeed, like physi-
cists or biologists, the mathematiciansuse to talk about a certain universe,
populated by certain objects. These objects have certain features and are
subjected to certain processes governed by well defined laws. The mathe-
matician describespropertiesof sets and numbersin much the same way as
the scientistpresentsthe structureof moleculesand crystals.Utteranceslike
"There exists a function such that.. ." are as commonplace in modern
mathematicsas the claims about the existence of certain subatomic parti-
cles are in physics.
Unlike materialobjects, however, advanced mathematicalconstructsare
totally inaccessibleto our senses - they can only be seen with our mind's
eyes. Indeed, even when we draw a function or write down a number, we
are very careful to emphasize that the sign on the paper is but one among
many possible representationsof some abstract entity, which by itself can
be neither seen nor touched. The mathematicianwould make claims about
existence and properties of this intangible object without giving much
thought to philosophical questions that his statements may evoke. Only
rarely would an author of a textbook make an apologetic remarksuch as
this: "We need not discuss how these abstract entities. .. may be catego-
rized from a philosophical point of view. For the mathematician .., it is
important merely to know the rules or laws by which they may be
combined" (Courant and John, 1962, p. 2). Being capable of somehow
"seeing" these invisible objects appears to be an essential component of
mathematicalability;lack of this capacity may be one of the majorreasons
because of which mathematicsappearspracticallyimpermeableto so many
"well-formedminds".
Even if this last claim is true (and I shall do my best in the sequel to
convince the reader that it really is), the careful analysis of textbook

4 ANNA SFARD
definitionswill show that treating mathematicalnotions as if they referred

to some abstractobjectsis often not the only possibility.Although this kind
of conception, which from now on will be called structural,seems to prevail
in the modem mathematics, there are accepted mathematicaldefinitions
which reveal quite a differentapproach. Function can be defined not only
as a set of orderedpairs, but also as a certain computationalprocess or as
a "method for getting from one system to another" (Skemp, 1971, p. 246).
Symmetrycan be conceived as a static propertyof geometricform, but also
as a kind of transformation.The latter type of description speaks about
processes, algorithmsand actions rather than about objects. We shall say
therefore, that it reflects an operationalconceptionof a notion.
Seeinga mathematicalentity as an objectmeans beingcapableof referring
to it as if it was a real thing - a static structure,existingsomewherein space
and time. It also means being able to recognizethe idea "at a glance" and
to manipulateit as a whole, without going into details. Using Hadamard's
metaphor(applied originallyin a slightlydifferentcontext), we can say that
structuralthinkingendows a concept with "a kind of physiognomy",which
allows a person to "think of it as a unique thing, however complicated it
may be,just as we see a face of a man" (Hadamard, 1949,p. 65). In contrast,
interpretinga notion as a process implies regardingit as a potential rather
than actual entity, which comes into existence upon request in a sequence
of actions. Thus, whereasthe structuralconception is static (or shall I say,
after Frege, 1970, "timeless"), instantaneous, and integrative, the opera-
tional is dynamic, sequential, and detailed.
It is practicallyimpossible to instantly pinpoint all the subtle aspects of
the above distinction, let alone to formulate exact definitions of the
structuraland operationalways of thinking.At this point, it should already
be quite evident that the former is more abstract,more integratedand less
detailed than the latter, but it should also be clear that such a comparison
leaves out at least as much as it comprises. Degrees of abstractionand of
integrationare but quantitativecharacteristics,whereasthe crucial,qualita-
tive, differencebetween the two modes of thinkinglies in the basic, usually
implicit, beliefs about the nature of mathematicalentities. In other words,
there is a deep ontologicalgap betweenoperationaland structuralconcep-
tions. It is the author's hope that as the discussiongoes on, this fundamen-
tal but elusive aspect of the distinction will become more and more clear.
All this being said, it is very importantto emphasizethat operationaland
structuralconceptions of the same mathematicalnotion are not mutually
exclusive. Although ostensibly incompatible(how can anything be a pro-
cess and an object at the same time?), they are in fact complementary.The

DUAL NATURE OF MATHEMATICAL CONCEPTIONS 5
term "complementarity"is used here in much the same sense as in physics,

where entities at subatomic level must be regardedboth as particlesand as
waves to enable full descriptionand explanation of the observed phenom-
ena (for a fuller discussion of complementarityin the context of education
see Otte, 1984, and Steiner, 1985). In the next section, I shall argue that in
a similarway, the ability of seeing a function or a numberboth as a process
and as an object is indispensablefor a deep understandingof mathematics,
whatever the definition of "understanding"is.
If we take a scrutinizinglook at any mathematicalconcept, more often
than not we shall find that it can be defined - thus conceived - both
structurallyand operationally.Some examples,chosen quite at random, are
presented in Figure 1.
The dual nature of mathematicalconstructs can be noticed not only in
verbal descriptions,but also through various kinds of symbolic representa-
tions. Although such property as structurality lies in the eyes of the
beholder rather than in the symbols themselves, some representations
appear to be more susceptibleof structuralinterpretationthan others. For
instance, differentapproachesto the concept of function can be detected in
Structural Operational
Function Set of ordered pairs Computationalprocess

(Bourbaki, 1934) or
Well defined method of
getting from one system
to another (Skemp, 1971)
Symmetry Property of a Transformationof

geometricalshape a geometricalshape
Natural Property of a set 0 or any number obtained

number or from another natural
The class of all sets number by adding one
of the same finite ([the result of]
cardinality counting)
Rational Pair of integers [the result ofl division

number (a member of a specially of integers
defined set of pairs)
Circle The locus of all points [a curve obtained by]

equidistant from rotating a compass
a given point around a fixed point
Fig. l. Structuraland operational descriptions of mathematicalnotions.

6 ANNA SFARD
Graph Algebraic expression Computer program
10 INPUT X
20 Y= I
\ | y= 3X4 30 FOR I=I TO 4
1 Y=Y.X
~~~~~~40
50 NEXT I
60 Y=3*Y
x>
Fig. 2. Different representationsof a function.
the three distinct ways in which the mappingy = 3x4 has been presentedin
Figure 2. The computer program seems to correspond to an operational
conception rather than to a structural,since it presents the function as a
computational process, not as a unified entity. In the graphic representa-
tion, on the other hand, the infinitelymany components of the function are
combined into a smooth line, so they can be grasped simultaneouslyas an
integrated whole; the graph, therefore, encourages a structuralapproach.
The algebraicrepresentationcan easily be interpretedboth ways: it may be
explained operationally,as a concise description of some computation, or
structurally,as a static relation between two magnitudes (this duality of
interpretationcorrespondsto the alreadywidely noticed and discusseddual
meaningof the equality sign: " = ' can be regardedas a symbol of identity,
or as a 'command" for executingthe operations appearingat its right side;
see e.g. Behr et al., 1976; Kaput, 1979; Kieran, 1981).
Different kinds of conception - structural and operational - manifest
themselvesalso in the special representationsof which people avail them-
selves while processing knowledge mentally. According to what is already
known on the internal encoding (see e.g. Paivio, 1971; Clements, 1981,
Bishop, 1988; Eisenberg and Dreyfus, 1989), mathematical concepts are
sometimes envisioned by help of "mental pictures", whereas on other
occasions the same ideas are handled mainly through verbal representa-
tions. Mental images, being compact and integrative, seem to support the
structuralconception. Hadamard's introspective observations on the role
of visualizationreinforcesthis supposition: "I need [an image] in order to
have a simultaneous view of all elements. . . to hold them together, to
make a whole of them .. .; to achieve synthesis. .; and give the concept its
physiognomy" (Hadamard, 1949, p. 77). Visualization, therefore, makes
abstract ideas more tangible, and encourages treating them almost as if
they were material entities. Indeed, mental images can be manipulated

almost like real objects. Like in face recognition, the pictures would
preservetheir identity and meaning when "observed"from differentpoints
of view and in different contexts. Visual representationis holistic in its
nature and various aspects of the mathematicalconstruct may be extracted
from it by "random access". In contrast, verbal encoding cannot be
grasped "at one glance" and must be processed sequentially, so it seems
more appropriate for representingcomputational procedures. Thus, the
non-pictorial inner representation is more pertinent to the operational
mode of thinking. [Please note: the above claims should not be interpreted
as an attempt to say that there is a one-to-one correspondencebetween
operational/structuralconceptions and verbal/visualinner representations.
The only suggestion I have been trying to make here is that some kinds of
inner representationsfit one type of conception better than the other.]
Before closing this introductory section it would be in point to notice

that mathematical, psychological and philosophical literature teems with
allusions to various dichotomies in mathematical universe. Distinctions
between "two types of mathematical knowledge/thought/understanding"
go like a thread of scarlet throughoutall kinds of recent writings,and some
of them may have certain bearings on the operational-structuralduality
suggested in this article.
Out of the long list of dichotomies proposed by differentwriters, let me
mention only a few (for a fuller catalogue see Hiebert, 1985, pp. 1-2).
According to some researchers,mathematicscan be divided into abstract
and algorithmic(see e.g. Halmos, 1985) or into declarativeand procedural
(Anderson, 1976). The names are almost self-explanatory,so even without
any formal definitions the connection between these distinctions and the
ideas presented in this paper should be quite obvious. The already men-
tioned observations on process/productduality of mathematicalsymbolism
(Kaput, 1979; Davis, 1975), although much more restrictedin scope, seem
to go hand in hand with this kind of divisions. Another categorization,
which has perhaps even more in common with our suggestions, is the one
which splits mathematics into dialectic and algorithmic (Henrici, 1974).
While algorithmic mathematics deals mainly with all kinds of computa-
tional processes, "dialectic mathematics is a rigorously logical science,
where statements are either true or false, and where objects with specified
propertieseither do or do not exist".
A certain kind of dichotomy has been observed also when psychological
rather than philosophical aspects of mathematics were concerned. Two
differentmodes of mathematicalthinkinghave been distinguishedby Piaget
(1970, p. 14):figurative, which refers to seeing "states as momentaryand

8 ANNA SFARD
static" and thus corresponds to our structuralconception; and operative,

which "deals.. . with transformations.. .", so has much to do with our
operational approach. Incidentally, this distinction is deeply rooted in
Piaget's theory of reflective abstraction which, especially in its further
elaborations (Thompson, 1985; Dubinsky and Lewin, 1976), has touched
explicitly upon the role of processes and objects in mathematicalthinking.
Even the widely accepted categorizations of mathematical understanding
(or knowledge) into conceptualand procedural(see e.g. Lesh and Landau,
1983;Hiebert, 1985) or into instrumentaland relational(Skemp, 1976) seem
to be somehow in point here. Indeed, in the next section an attempt will be
made at showing how our capability for developing operational and
structuralconceptions bears upon the type of understandingwe achieve.
The classificationsuggested in this article puts us, therefore, in a good
company. Moreover, even if our distinction still looks somewhat fuzzy, so
do all the others. For instance, while talking about conceptual and proce-
dural knowledge, Hiebert and Lefevre complained: "the relationship be-
tween these forms of knowledgeis not yet well understood",and "the types
of knowledge themselvesare difficultto define; the core of each is easy to
describe, but outside edges are hard to pin down" (ibid., p. 3). "Pinning
down the edges", at least in the case of operational-structuraldistinction,
is my objective in the remainderof this article.
To begin with, let me confront our division with those listed above. On
the face of it, the idea of operationaland structuralconceptions may seem
not much different from some of the dichotomies which have just been
mentioned. For all the similarities,however, two fundamentalcharacteris-
tics of our distinction - its combined ontological-psychologicalnature and
its complementarity- put it apart from the majority of other classifica-
tions. Firstly, most of those who suggested some kind of dichotomy rarely
gave much attention to the question of tacit philosophical assumptions
underlyingany mathematicalactivity;rather,they referredeither to certain
more obvious aspects of the subject-matter(such as its structureor the role
of its different components in problem-solving),or to the cognitive pro-
cesses involved in handlingthe knowledge.In our classification,we tried to
address the first and the last of these issues simultaneouslyby focusing on
the nature of mathematicalentities (ontological issue) as perceived by a
thinker (psychological perspective). Secondly, whereas other distinctions
lead to decompositionof mathematicalknowledgeinto two separatecompo-
nents (e.g., concepts vs. procedures), our complementarian approach
stresses its unity. True, recently the former position seems to be gradually
abandoned. While referring to the issue of concepts and procedures,

Hiebert and Lefevre observed: "Historically,the two kinds of knowledge

have been viewed as separate entities, . . . coexisting as disjoint neigh-
bours ... In contrast, there is a growing interesttoday in how concepts and
proceduresare related" (ibid., p. 2). Nevertheless, this new approach still
cannot be regardedas complementarian:"currentdiscussionstreat the two
forms of knowledge as distinct", even though "linked in critical, mutually
beneficialway". Let me stress once more: unlike "conceptual"and "proce-
dural", or "algorithmic" and "abstract", the terms "operational" and
"structural"refer to inseparable, though dramaticallydifferent, facets of
the same thing. Thus, we are dealing here with duality rather than
dichotomy.
To complete the picture, let me make yet another remarkregardingthe
normative approach some writers assume while talking about "different
kinds of mathematics". "Algorithmic" and "abstract" mathematics, for
example, are sometimes assessed and contrasted against each other as if a
contest was held between them. While there seems to be a consensus that
the "abstract" mathematics deserves the highest esteem, the algorithmic,
proceduralaspects are rather controversial.Recently, the dispute over the
topic has become more heated than ever: provocativelyexaggerateddecla-
ration that "algorithmicway of life is best" (Maurer, 1985) evoked angry
reactions from those who feel that "algorithmdrives out thought" (Stein,
1988). Even though everybody admits that "algorithmic"mathematics is
important, the opinion seems to prevail that it is somehow second-rate.
Our complementarianapproach strips this kind of discussion of any
meaning. Whether the issue of applications or of education is concerned,
the operational and structural elements cannot be separated from each
other. Therefore,we may only repeat after Halmos (1985) that to "try to
decide which component is more important is not much more meaningful
than to debate whether for walking you need your right foot more than
your left". This statementwill be furthersubstantiatedin the next sections,
where the mutual dependence and the necessity of both operational and
structural conceptions will be carefully explained and illustrated. The
thorough discussion of the role played by them in all kinds of cognitive
processes will help us to grasp the reason for which abstract mathematics
- the one more heavily based on a structuralapproach- is so highly-rated.
Indeed, it will be shown that the real insight necessary for mathematical
creation can hardly be achieved without the ability to "see" abstract
objects, and that, on the other hand, the structural conception is very
difficultto attain (that is probablywhy some people may feel, intuitivelyof
course, that the special ability to develop a structuralconception is what

10 ANNA SFARD
distinguishesmathematiciansfrom "mere mortals"). For all the attention

and respect given to the structural approach, the operational mode of
thinking will also get its due share:we shall argue that a profound insight
into the processesunderlyingmathematicalconcepts, maybe even a certain
degree of mastery in performing these processes, should sometimes be
viewed as a basis for understanding such concepts rather than as its
outcome. Consequently, the "technical skills" will be rehabilitated, after
they were unduly demoted in a somewhat exaggeratedreaction to behav-
iorism. At last, we may even find ourselves in the position to offer a
tentative answer to the vexing question so often asked by educators:
Why is it that so many intelligent,well-trained,well-intentionedteachersput such a premium
on developing students' skill in the routines of arithmetic and algebra despite decades of
advice to the contrary from so-called experts?What is it that teachers know that others do
not? (Kilpatrick, 1988).
2. THE ROLE OF OPERATIONAL AND STRUCTURAL CONCEPTIONS

IN THE FORMATION OF MATHEMATICAL
CONCEPTS - HISTORICAL OUTLOOK
Of the two kinds of mathematicaldefinitions, the structuraldescriptions

seem to be more abstract. Indeed, in order to speak about mathematical
objects, we must be able to deal with productsof some processes without
bothering about the processesthemselves.In the case of functions and sets
(in their modem sense) we are even compelled to ignore the very question
of their constructivity. It seems, therefore, that the structural approach
should be regardedas the more advancedstage of concept development.In
other words, we have good reasons to expect that in the process of concept
formation, operationalconceptionswouldprecede the structural.Different
kinds of evidencewill be brought in this article to show that this statement
is basically true whether historical development or individual learning is
concerned.
Before any examplewill be given, however, it should be pointed out that
the proposed model, although believed to be very broad in its scope, might
be inadequatein certaincases. Geometricideas, for example, for which the
unifying, static graphical representationsappear to be more natural than
any other, can probably be conceived structurallyeven before full aware-
ness of the alternative procedural descriptions has been achieved. The
concept of a circle, for instance, could develop in the steps envisioned by
M. Boole (Tahta, 1972): "The elementarygeometricianwho first conceived
the idea of the circle caught his suggestion from looking at things whose
forms were approximatelyround; but as soon as he discovered the law of

roundness [the algorithm for obtaining a circle] within his own mind, he
was able to express roundness in a new material". Even though this
developmental scenario looks plausible, the ensuing historical analysis of
other concepts will hopefully convince the reader that in computational
mathematics, the majority of ideas originated in processes rather than in
objects. Indeed, a close look at the history of such notions as number or
function will show that they had been conceived operationallylong before
their structuraldefinitions and representationswere invented.
What follows now is a very brief and by no means exhaustivepresenta-
tion of a long and turbulenthistory of some of the most centralmathemat-
ical concepts. In this article I shall deal with only those historicalfacts and
events which highlight the point I would like to make here (for a more
detailed historical account see e.g. Cajori, 1985; Kleiner, 1989).
Let us begin our analysis with the notion of number. For a very long
time the meaning of this term was restrictedto what is known today under
the name "naturalnumber".This kind of numberoriginatesin the process
of counting, so even before turning to the history, let us consider a certain
well-known phenomenon observed in young children. It was noticed by
researchers(see e.g. Piaget, 1952, p. 62) that when a child learns counting,
there is a stage when he or she can already build a one-to-one mapping
between the words "one", "two", "three",.. and the objects in a given
set, but would not use the last number-nameused in this process as an
answerto a question "How many objects are there?".Wheneverasked, the
child would just repeat the procedure of counting. This phenomenon
clearly shows the operational roots of natural numbers:for the child, the
process of counting itself, not its abstract product, is what is meant
whenever the term "number"is mentioned.
The meaningof the term "number"has been generalizedseveraltimes in
the course of the last three thousand years. For long periods did mathe-
maticians perform some special manipulationswith already acknowledged
kinds of numbersbefore they were able to sever an abstractproduct from
these new processes and to accept the resulting entities as a new kind of
mathematical objects. For instance, a ratio of two integers was initially
regarded as a short description of a measuring process rather than as
a number. Incidentally, some traces of purely operational approach to
rationals were noticed by researchers (Carpenter et al., 1980) also in
today's 13-year-oldstudents, 50 percent of whom were found unable "to
representa division problem like 7 divided by 4 as a fraction". In these
students, the division of integers was still only a process which could not
yet be seen at will as a static entity.

12 ANNA SFARD
For a long time the term "number"appearedmainly in the context of

measuring processes. The Pythagoreandiscovery that in certain squares,
the usual procedure for finding the length of the diagonal cannot be
describedin terms of integersand their ratios (because the diagonal and the
sides have no common measure), was greeted with astonishment and
bewilderment(Hippasus of Metapontum,the unfortunatediscovererof the
incommensurability,was reportedlythrown to the sea for his "heresies").
Much time elapsed before mathematicianswere able to separatethe notion
of number from measuringprocessesand to acknowledgethe fact that the
length of any segment representsa numbereven if it cannot be found in the
"usual"way. Eventually,the set of numbershas been broadenedagain, to
include positive irrationalsalong with integers and fractions.
This enlarged set, in its turn, gave birth to new kinds of computational
processes, and then to new kinds of numbers. Cardan's prescriptionsfor
solving equations of the third and fourth order, publishedin 1545, involved
subtractingpositive rationals from smaller ones and even finding roots of
what is today called negative numbers.Despite the widespreaduse of these
algorithms, however, mathematiciansrefused to accept their by-products
and for some centuries referredto them as "absurd"or "imaginary".The
term "negative number" and the symbol /- 1 were initially considered
nothing more than abbreviations for certain "meaningless" numerical
operations. They came to designate a fully-fledged mathematical object
only after mathematiciansgot accustomedto these strangebut useful kinds
of computation.
As to the operational origins of the negative and complex numbers, it
would be most illuminating to look into the historical writings by the
logician and philosopher P. E. Jourdain (1879-1919). In the following
passage, Jourdain (1956, p. 27) says explicitly that negative number is
nothing but a type of process:
Let a - b be c. To get c from a we carry out the operation of taking away b. This operation,
whichis thefulfillmentof the order:"Subtractb", is a "negativenumber".Mathematicianscall
it a "number" and denote it by "-b" simply because of analogy: the same rules for
calculation hold for "negative numbers"and "positive numbers".
Jourdain'sexplanation highlights the developmentalgap between his con-

ceptions of the "unsigned"and the negative numbers:while the formerare
alreadyregardedby the author as "real",genuine objects, their operational
origins totally forgotten, the latter are still identified with processes and
allowed to be treated as static entities only by force of convention. Since
the idea of complex numbersstems from certain manipulationsperformed
on negatives, it should surprise nobody that Jourdain's interpretationof

the "imaginary"numbersis also purely operational:while referringto the

number "i", he states that "it representsan operation,just as the negative
numbers do, but of a different kind" (p. 30).
The most important thing to be learned from this short historical
account is that the development of the notion of number was a cyclic
process, in which approximately the same sequence of events could be
observed time and again, whenevera new kind of numberwas being born.
These iterations have been summarized schematically in Figure 3. Each
recurrentsegment of the schema representsa lengthy process, consisting of
three phases:
(1) the preconceptualstage, at which mathematicianswere getting used to
certain operationson the already known numbers(or, as in the case of
counting - on concreteobjects);at this point, the routinemanipulations
Complex
InumbersI
Real ex. root

numbers - Imaginary
Positive subtr. Negative

real real
Positive cx. root Positive

rational measuring irrational
Natural division Positive

numbers imeasunng fractions
Sets of
concrete
objects counting
Fig. 3. Development of the concept of number.

14 ANNA SFARD
were treated as they were: as processes, and nothing else (there was no
need for new objects, since all the computations were still restrictedto
those procedureswhich produce the previously accepted numbers).
(2) a long period of predominantlyoperational approach, during which a
new kind of number begun to emerge out of the familiar processes
(what triggered this shift were certain uncommon operations, previ-
ously regarded as totally forbidden, but now accepted as useful, if
strange); at this stage, the just introduced name of th- new number
served as a cryptonym for certain operations rather than as a signifier
of any "real" object; the idea of a new abstract construct, although
already in wide use, would still evoke strong objections and heated
philosophical discussions;
(3) the structuralphase, when the number in question has eventually been
recognized as a fully-fledged mathematical object. From now on,
different processes would be performed on this new number, thus
giving birth to even more advanced kinds of numbers.
To sum up, the history of numbers has been presented here as a long
chain of transitions from operational to structuralconceptions: again and
again, processes performedon already accepted abstract objects have been
converted into compact wholes, or reified (from the Latin word res - a
thing), to become a new kind of self-contained static constructs. Our
conjectureis, that this model can be generalizedto fit many other mathe-
matical ideas.
For instance, the just presented pattern repeats itself in the history of
function. This important idea, born in the end of seventeenthcentury (at
least officially),was the result of a long search after a mathematicalmodel
for physical phenomena involving variable quantities. When the term
"function" appearedfor the first time (in a work by Leibniz, in 1692), the
recently invented algebraic symbolism was gaining popularity and gradu-
ally entering every branch of mathematics. No wonder then, that the
notion of function was initially tightly connected to algebraic processes.
The new term was first used to denote "a quantity composed in any
mannerwhateverof [a] variableand constant" (by Jean Bernoulliin 1718),
or the so called "analyticexpression"(by Euler in 1747). Thus, in a sense,
the concept of function was for algebraicmanipulationson variableswhat
the idea of a negative numberwas for subtraction:something between the
product and the process itself.
The main problem with the early definitions of function was that they
leaned heavily on the concept of variable,which by itself was rather fuzzy

and escaped every attempt at reification. That is probably one of the

reasons why in 1755, after a long discussion with d'Alembert, Euler
suggested another definition, in which the term "variable"was not explic-
itly mentioned:"a quantity"should be called function only if it dependson
another quantity "in such a way that if the latter is changed the former
undergoes change itself". The operational flavor emanates from this de-
scription even more clearly than from the earlier versions.
All the furtherhistory of the notion may be seen as a long sequence of
strenuous, if mostly failed, attempts at reification. Euler himself tried to
arrive at a fully-fledged structural version by endowing his "changing
quantities" with a "solidifying" graphic representation.Euler's idea was
not very helpful, however, since neither he, nor any of his contemporaries
were able to build a truly satisfactory bridge between the algebraic and
graphic approach:each time a definition had been proposed which would
fit the algebraic-operationalintuition, after a while somebody would find
an example showing that the new description fell short of the structural-
graphic version; and vice versa (for a detailed account see Kleiner, 1989).
It should be pointed out that at a certain stage, mathematiciansand
philosophersbecame fully aware of what for some time had probably been
done only intuitively - of their striving for reification, of their need of
definitionwhich would justify the common practiceof referringto function
as if it was a real "thing". For example, let us look at the following remark
from the beginning of our century:
In recent times the word 'variable' is predominant in the definitions [of function]. Conse-
quently Analysis would have to deal with a process in time, since it takes variables into
consideration. But in fact it has nothing to do with time; its applicabilityto occurrencesin
time is irrelevant .... as soon as we try to mention a variable, we shall hit upon something
that varies in time and thus does not belong to pure Analysis. And yet it must be possible to
point to a variablethat does not involve something alien to arithmetic,if variablesare objects
of Analysis at all. (Frege, 1970, p. 107; German original: 1904.)
Frege's call for elimination of time is an explicit request for reification.

Also, this remarkmakes it clear how difficultthe struggle for a structural
version of the notion was.
The numerous failed attempts at translating operational intuition into
structural definition led to Dirichlet's rebellion against the algorithmic
approach, and eventually to the now widely accepted, purely structural
Bourbaki'sdefinition.This simple descriptionpresentedfunction as a set of
orderedpairsand madeno referencewhatsoeverto any kind of computational
process. Bourbaki'sgroup solved the time-reveredproblem by eliminating
the "unreificable"notion of variableand substitutingit with purelystructural
set-theoreticconcepts. Not surpnsingly,this new definition,which had very

16 ANNA SFARD
little in common with its intuitive operationalorigin, evoked much criticism

when first proposed. But when at a long, long last function - initially only
a computationalprocess - was converted into a mathematicalobject, our
schema of concept development could repeat itself once more: on the new
objects new operations could now be performed.These new operations are
known today as functionals.
Let us summarizeonce again what has been observedthus far. In all our
examples, the same phenomenon could be distinguished over and over
again: various processes had to be converted into compact static wholes to
become the basic units of a new, higher level theory. When we broaden our
view and look at mathematics(or at least at its big portions) as a whole,
we come to realizethat it is a kind of hierarchy,in which what is conceived
purely operationally at one level should be conceived structurally at a
higherlevel. Such hierarchyemergesin a long sequenceof reifications,each
one of them startingwhere the formerends, each one of them adding a new
layer to the complex system of abstractnotions. In certaincases, of course,
this picturewould seem a little simplistic.The process of concept formation
would look more intricate than implied by our unidirectionalmodel. This
model, however, is to be regardedas not more than a first approximation,
indicating only the prevalent tendency.
3. THE ROLE OF OPERATIONAL AND STRUCTURAL CONCEPTIONS

IN THE FORMATION OF MATHEMATICAL
CONCEPTS - PSYCHOLOGICAL OUTLOOK
What really strikes the eyes in the already given examples is that the
fonnation of a structuralconception is a lengthy, often painfully difficult
process. The question which now cries out to be asked concernsthe sources
of this difficulty.Naturally, this issue should be tackled from the psycho-
logical point of view, and this is exactly what will be done in the next
section. In the remainderof the present part I shall restrict myself to the
preliminaryproblem:is the proposed model of concept formation in force
also when individual learning is concerned?Or, in other words, is it true
that when a person gets acquaintedwith a new mathematicalnotion, the
operationalconception is usually the first to develop?The odds are that the
answer to this question should be yes. Let me put it even more clearly:it
seems that the scheme which was constructed on the basis of historical
examples can be used also to describe learning processes.
At this point, some objectionsmay be raised by a careful reader.Firstly,
the above statementsimply that there is some "natural"course of events in

processes, which can hardly be regardedas spontaneous. Indeed, mathe-

matical learning,expeciallyat more advancedlevels, cannot be expected to
take place without external intervention(of a teacher, of a textbook), and
may therefore be highly dependent on a kind of stimulus (of teaching
method) which has been used. Moreover, given a specificteaching strategy,
it is sometimes almost impossible to know to what extent the observed
learning process has been influenced by this particularmethod, and how
differentit could be in other circumstances.The simplestway to deal with
this kind of doubt would be to say that in the psychological context, the
statement "operationalbefore structural"should be understood merely as
a prescriptionfor teaching. Yet, although such interpretationshould not be
dismissed,it would not make full justice to the suggestedmodel. Our whole
argumentationis based on the assumption (which, incidentally, seems to
underlie most of cognitive research since Piaget) that in the process of
learning - any kind of learning! - certain constant characteristicscan be
identifiedwhich appear to be quite immune to changes in external stimuli.
The precedenceof the operationalconceptions over structuralis presented
here as one of such invariants.
Secondly, it must be stressed right away that the suggested model of
concept acquisition should by no means be viewed as a result of a
light-minded, automatic projection from history to psychology. Claims
about operational origins of mathematical notions were made by many
writers,often without any referenceto history. The pioneeringwork in this
field has been done by Piaget, who wrote in his book on genetic epistemol-
ogy (1970, p.16): "the [mathematical]abstraction is drawn not from the
object that is acted upon, but from the action itself. It seems to me that this
is the basis of logical and mathematicalabstraction".For the last twenty
years, this supposition has guided both theoretical and empirical research
on mathematicalthinking. Recent studies elaboratedPiaget's original ideas
and put into them new contents (see e.g. Thompson, 1985; Sinclair and
Sinclair, 1986; Dubinsky and Lewin, 1986; Dorfler, 1987, 1989). Our
somewhat broader conjectureabout the duality of mathematicalthinking
and the developmentalprecedence of the operational conceptions can be
substantiatedin many differentways. Some empiricalevidence has already
been woven into our historical account, and many other recent findings in
'the field of learning mathematics may serve as additional reinforcement
(see Sfard, 1987, 1988, 1989). But first and foremost, there is a strong
theoreticalargument speaking for our thesis. If the structuralapproach is
more abstractthan the operational,if from the philosophicalpoint of view
numbersand functions are basically nothing but processes, if doing things

18 ANNA SFARD
is the only way to somehow "get in touch" with abstractconstructs - if all

this is true,then to expectthat a personwould arriveat a structuralconception
without previous operational understanding seems as unreasonable, as
hoping that he or she would comprehendthe two-dimensionalscheme of a
cube without being acquaintedwith its "real-life"three-dimensionalmodel.
According to our scheme of historical development, three steps can be
distinguished in the process of concept formation. These three stages
correspondto three "degreesof structuralization"which may be named on
the grounds of purely theoretical analysis of the relationship between
processes and objects. In the light of the same analysis, our model of
learning can now be refined along similar lines: if the conjecture on
operationalorigins of mathematicalobjects is true, then first there must be
a process performed on the already familiar objects, then the idea of
turning this process into an autonomous entity should emerge, and finally
the ability to see this new entity as an integrated,object-likewhole must be
acquired. We shall call these three stages in concept development interi-
orization,condensationand reification,respectively.
At this point, a detaileddescriptionof each stage should be given. Before
this is done, however, we must be aware of a methodological difficulty
stemming from the fact that we are dealing here with student's implicit
beliefs about the nature of mathematicalobjects. Unable to investigatethe
problem in a direct way, how shall we diagnose the differentstages in the
conceptual developmentof a learner?It seems that we have no choice but
to describeeach phase in the formation of abstractobjects in terms of such
external characteristicsas student's behaviours,attitudes, and skills. The
resultingspecificationwill hopefully be clear enough to serve as a tool for
diagnosing, maybe even measuring, student's ability to think structurally
about a concept at hand.
At the stage of interiorizationa learnergets acquaintedwith the processes
which will eventuallygive rise to a new concept (like counting which leads
to naturalnumbers,subtractingwhich yields negatives,or algebraicmanip-
ulations which turn into functions). These processes are operations per-
formed on lower-level mathematical objects. Gradually, the learner
becomes skilled at performingthese processes.The tenn "interiorization"is
used here in much the same sense which was given to it by Piaget (1970, p.
14): we would say that a process has been interiorizedif it "can be carried
out through [mental] representations",and in order to be considered,
analyzed and compared it needs no longer to be actually performed.
In the case of negative number, it is the stage when a person becomes
skillful in performingsubtractions. In the case of complex number, it is

when the learneracquireshigh proficiencyin using squareroots. In the case

of function, it is when the idea of variableis learnedand the ability of using
a formula to find values of the "dependent"variableis acquired.
The phase of condensationis a period of "squeezing"lengthy sequences
of operations into more manageableunits. At this stage a person becomes
more and more capable of thinking about a given process as a whole,
without feeling an urge to go into details. It is like turninga recurrentpart
of a computer program into an autonomous procedure:from now on the
learnerwould refer to the process in terms of input-outputrelations rather
than by indicating any operations. As in the case of computer procedures,
a name might be given to this condensed whole. This is the point at which
a new concept is "officially"born. Any difficultywith indicatingthe output
of the underlyingprocess (like in the case of subtractinga number from a
smaller one while only unsigned numbers are known) will serve as an
additional trigger for the idea of a new mathematicalentity. Thanks to
condensation, combining the process with other processes, making com-
parisons, and generalizingbecome much easier. A progressin condensation
would manifestitself also in growing easiness to alternatebetweendifferent
representationsof the concept.
In the case of the negative numbers, condensation may be assessed
through student's proficiency in combining the underlyingprocesses with
other computationaloperations;or, in other words, in his or her ability to
perform such arithmetic manipulationsas adding or multiplying negative
and positive numbers. In the case of complex numbers, condensation is
what helps the learner to realize that reversingthe operation of squaring
may be useful as a part of lengthy calculations even if it would not, by
itself, yield a legitimate mathematical object. The student may still treat
such symbol as 5 + 2i as nothing but a shorthandfor a certain procedure,
but at this stage it would not prevent him from skillfuly using it as a part
of a complex algorithm.When function is considered,the more capable the
person becomes of playing with a mapping as a whole, without actually
looking into its specific values, the more advanced in the process of
condensation he or she should be regarded. Eventually, the learner can
investigatefunctions, draw their graphs,combine couples of functions (e.g.
by composition), even to find the inverse of a given function.
The condensation phase lasts as long as a new entity remains tightly
connected to a certain process. Only when a person becomes capable of
conceivingthe notion as a fully-fledgedobject, we shall say that the concept
has been reified. Reification,therefore,is defined as an ontological shift -
a sudden ability to see something familiar in a totally new light. Thus,

20 ANNA SFARD
whereas interiorizationand condensation are gradual, quantitative rather

than qualitative changes, reification is an instantaneous quantum leap: a
process solidifiesinto object, into a static structure.Various representations
of the concept become semanticallyunified by this abstract, purely imagi-
nary construct. The new entity is soon detached from the process which
produced it and begins to draw its meaning from the fact of its being a
memberof a certaincategory. At some point, this category ratherthan any
kind of concrete construction becomes the ultimate base for claims on the
new object's existence. A person can investigate general propertiesof such
category and various relations between its representatives.He or she can
solve problems involving finding all the instances of the category which
fulfill a given condition. Processescan be performedin which the new-born
object is an input. New mathematicalobjects may now be constructedout
of the present one. Penrose's (1989, p. 67) statement referringto Church's
lambda-calculusmay serve as a telling example: "one is concerned [here]
with a 'universe' of objects, denoted by say a, b, c, . . ., each of which
stands for a mathematicaloperation or function . . . The things on which
these functions act - are other things of the same kind, i.e. also functions"
(incidentally, this statement has another interesting aspect: by referring
to function first as 'object'or 'thing', and then as 'mathematicaloperation',
it clearly indicates the operational-structuralduality of the author's
approach).
The stage of reificationis the point where an interiorizationof higher-
level concepts (those which originate in processes performedon the object
in question) begins.
In the case of negative numbers,it is learner'sability to treat them as a
subset of the ring of integers(without necessarilybeing aware of the formal
definition of ring) which can be viewed as a sign of reification. Complex
numbermay be regardedas reifiedwhen the symbol 5 + 2i is interpretedas
a name of a legitimate object - as an element in a certain well-definedset
- and not only (or even not at all) as a prescriptionfor certain manipula-
tions. In the case of function, reificationmay be evidenced by proficiency
in solving equations in which "unknowns" are functions (differentialand
functional equations, equations with parameters),by ability to talk about
general propertiesof differentprocesses performed on functions (such as
composition or inversion), and by ultimate recognition that computability
is not a necessarycharacteristicof the sets of orderedpairs which are to be
regardedas functions.
How and to what extent the just presenteddevelopmentalschememay be
influenced by deliberate instructional actions is an important question

which must be dealt with in a separatearticle. There is one thing, however,

which is much too essential to be passed over in silence. It is the potential
role of names, symbols, graphs, and other representationsin condensation
and reification.Judging from the history, the importanceof this factor can
hardly be overestimated.For example, the introductionof the number line
may be viewed as a final triggerfor the reificationof negativenumbers,and
the invention of what is known today as Argand plane can be regardedas
a decisive step in turning complex numbers into legitimate mathematical
objects. It seems reasonable to expect that representationswould play a
similar role in individual learning. This topic, although already studied
in a similar context (see e.g. Dorfler, 1987), asks for more empirical
investigation.
At this point it should already be clear that our three-phaseschema is to
be understood as a hierarchy, which implies that one stage cannot be
reached before all the former steps are taken (see Figure 4). One word
should be added, however, regardingcertain side routes which can be taken
by a learner. The student may manipulate a concept through a certain
prototype (for example, the data collected by Markovitzet al., 1985, show
that beginners tend to imagine linear mappings whenever the notion of
function is mentioned) or, unable to come to terms with the invisible
'objects', he or she can develop a debased, quasi-structuralapproach,
namely a tendency to identify the notion at hand with one of its represen-
tations (in the case of function: formula or graph). This stage, which is
clearly a deviation from our scheme, may be transitory or permanent.
Let me conclude this part of the study with two remarksaddressedto a
critical reader.First, those who feel that at some points the author sounded
somewhat too assertiveshould be remindedthat the majorityof statements
made in this section were analytic rather than synthetic. For example, the
hierarchicalnature of the scheme is implicitin the definitionsof interioriza-
tion, condensation and reification. More generally, the model of concept
acquisition presented in this section has been deduced from onc basic
conjecture- from the thesis about the operationalorigins of mathematical
objects. Like so many other ideas contrivedby those who still believe in the
possibility of a theoretical framework for cognitive research, our three-
phase schema has a highly speculativecharacter- and this is only natural.
Hypothetical and simplified as it is, it has already started to prove itself
useful as a tool for planning, integrating, and interpretingempirical re-
search (see Sfard, 1987, 1989). Ultimately, it may lead to some important
didactic implications (Sfard, 1988).
Secondly, the claim about the developmental priority of operational

22 ANNA SFARD
Concept C
Object C
I
reification
I
condensation
Concept B interioization
Object B ocesses
on
I ~~~~B
reification
condensation
Concept A interionzation
on concrete ~ o
inteiorization
4 |Concrete | | j Projcss s l o
|objects rC objects l
Fig. 4. Gcneral model of concept formation.
conceptions over structural, although historically sound and empirically

demonstrable,may seem somehow at variancewith the common practiceof
introducingnew concepts by help of structuraldefinitions,usually without
an explicit referenceto any kind of underlyingprocesses. More often than
not, a modemnmathematical textbook would start its presentation of the
concept of complex number with a simple statement:"Let us consider the
set of all the pairs of real numberswith the following properties.. .". The
contemporarymathematicianwould offer an entirelynew idea in a forn of
a ready-madeobject, clearly believing that the abstract construct may be
brought into being just by force of an appropriatedefinition. Thus, the
possibility must be considered that, after all, structural conception may
sometimesbe the first. This can certainlybe true in the case of professional

mathematicians - their well-trained minds can indeed be capable of

manipulatingabstractobjects right away, without the mediation of compu-
tational processes. Even so, in the light of both theoreticalargumentsand
experimentalfindings, our model does seem to present a prevailing ten-
dency. In fact, this tendency may be so strong, that even if a new concept
is introducedstructurally,the studentwould initiallyinterpretthe definition
in an operational way (much empirical evidence for this may be found in
recent studies on algebra and on the concept of function; see Vinner and
Dreyfus, 1989; Sfard, 1987, 1989).
4. THE ROLE OF OPERATIONAL AND STRUCTURAL CONCEPTIONS IN

COGNITIVE PROCESSES
Before we try to find an explanation for the inherent difficultyof reifica-

tion, let us tackle yet another more basic issue. The just suggestedmodel of
concept formation implies that certain mathematical notions should be
regardedas fully developed only if they can be conceived both operationlly
and structurally.The question arises, what is it that necessitatesthis dual
outlook; or, in other words, what can be achievedwith the ability of seeing
a concept both operationallyand structurally,which would not be attained
if only one approach was always assumed.
4.1. OperationalApproach:CertainlyNecessary, Sometimesalso Sufficient

Apparently,a purely operational way of looking at mathematicswould be
quite appropriate. Indeed, to an unprejudicedand insightful person, the
very notion of "mathematicalobject" may appear superfluous:since pro-
cesses seem to be the only real concern of mathematics,why bother about
these elusive, philosophicallyproblematic "things", such as infinite sets or
"aggregatesof ordered pairs"? Theoretically it would be possible to do
almost all the mathematics purely operationally:we could proceed from
elementary processes to higher-level processes and then to even more
complex processes without ever referringto any kind of abstract objects.
As a matter of fact, a careful look at history would reveal that for a very
long time big portions of mathematicswere done alnost exactly this way.
As Davis and Hersh (1983, p. 182) notice, "The mathematicsof Egypt, of
Babylon, and of ancient Orientwas all of algorithmictype .... It is only in
modern times that we find mathematics with little or no algorithmic
content, which we could call purely dialectical or existential". Indeed,
the science of computation, known today under its relatively new name

24 ANNA SFARD
"algebra",has retained a distinctly operational characterfor thousands of

years.The so called "rhetorical"algebra,whichprecededthe syncopatedand
symbolicalgebras(the last one developednot before the 16thcentury!)dealt
with computationalprocessesas such, while the only kind of abstractobjects
permittedin the discoursewere numbers.Even most complex sequencesof
numericaloperations were presentedby help of verbal prescriptions,which
bore distinctly sequentialcharacterand did not stimulatecondensationand
reification.As long as the computationalprocesses have been presentedin
the purely operationalway, they could not be squeezedinto static abstract
entities, thus were not susceptibleof being treated as objects.
4.2. The Necessity of StructuralConceptions

Twentieth-centurymathematics,however, seems to be so deeply permeated
with the structural outlook, that a modem mathematician had to be
exceptionally open-minded - indeed, not himself at all - to realize that
from a philosophical (not psychological!) point of view he could do
without "mathematicalobjects". In his eyes, this notion is probably so
inherent to mathematicsas the idea of matter is to physics. (Can anybody
imagine motion without physical bodies? Can anybody talk about compu-
tational processes performedon ... nothing?).Why do we have this strong
propensityfor making abstractionsin the image of the materialword?This
important question may be addressedat several levels. The most obvious,
top-level answer is that our imaginationis shaped by our senses. Probably
that is why we have this overpoweringfeeling that we cannot perform a
process, unless there is an object on which this process is carried out, and
there is another object, which this process produces.
For a more profound explanation we shall turn in a moment to the
theory of cognitive schemata. Before we do this, however, the reader is
invited to perform an experiment, which will hopefully illuminate our
subsequentclaims in the most persuasiveway.
Our exercisewill be performedin two steps. Let us begin with mathemat-
ical definitions, presentedin Figure 5. The first step, which should not be
taken until the definition of stroll is learned, is to perform the three tasks
listed in the box. The reader is invited to tackle the problems one by one,
without changing the order.
Fulfilling the first two requirements,and especiallythe second, can be a
rathertough job. Also, those who managedto solve the first problemmight
notice that respondingto the now seeminglyobvious third question is not
at all as straightforwardas it looked.

Definition: Promenadeis the set P of all natural numbers from 1 to 25 together with the
following four functions:
S(x)= x + 5, allowed only for x e P, x 6 20
N(x) =x-5, allowed only for x e P, x> 5
E(x)= x + 1, allowed only for x e P, xmod 5#O
W(x) = x-1, allowed only for x E P, x mod 5 # I
Any composition of the above functions is called a stroll. We say that stroll s
leadsfrom a to b iff s(a) = b.
Example:the stroll S o W3 o S2 leads from 5 to 17:
I
(S o S2)(5) = (W3 S2X 10) = (W2o S2X9) = ... S2(7) = S(12) = 17
Tasks:
1. Give an example of a stroll which would lead from 11 to 3.
2. Find all the numbers which can be reached by strolls from 9 without using
the steps N and W.
3. Without looking into the answer you gave to the question 1 above, give an
example of a stroll from 11 to 3 once again.
Fig. 5. The experiment- Part 1.
Even if not all three tasks have been successfullyaccomplished,we shall

turn now to the second part of our experiment.The reader is invited to
study a new description of the concept of stroll, as presentedin Figure 6,
and to solve the three problems once again.
If everythinggoes well, there should be a considerabledifferencebetween
the first and the second trial at solving the problems. By introspection,the
readercould probably find out what the author has observedwith particu-
lar claritywheneverthe experimentwas carriedout in a classroom:with the
second kind of definition (the one presented in Figure 6) all three tasks
become much easier. What is it about this new representationthat makes
such difference?It is probablyits distinctlystructuralcharacter:the "prom-
enade", which was first regardedas nothing but a bunch of algorithms,has
now been combined into an easily manipulable object-like structure;the
"stroll", which according to the first definition was a computational
procedure, can now be viewed as a path of a constant shape - as a
polygonal line. It is important to notice that no information has been
added when the shift from the operational to the structuralapproach was
made: the computationalprocesseswere caught into a static constructjust
like water is frozen in a piece of ice.

26 ANNA SFARD
In a quest for a more insightful explanation of the role of structural

conceptions, it is time now to make use of the theory of cognitive
schemata.Our example shows in a convincing,if somewhat simplifiedway,
what could be figured out also on the grounds of purely theoretical
considerations: the operationally conceived information, although abso-
lutely indispensableand seeminglysufficientfor problem-solving,cannot be
easily processed. This kind of information can only be stored in unstruc-
tured, sequential cognitive schemata, which are inadequate for the rather
modest dimensions of human working memory. Consequently,the purely
operational ideas must be processed in a piecemeal,cumbersomemanner,
which may lead to a great cognitive strain and to a disturbingfeeling of
only local - thus insufficient- understanding.Naturally, such strainwould
be totally counter-productive for anybody trying to solve a complex
problem. For instance, think about the difficulty you must have experi-
enced when still endowed with only operational definitions and trying to
perform the first two tasks in our experiment;or imagine how hard it
would be to solve an advanced word problem in a "rhetorical"manner,
without using algebraicsymbols. It should also be pointed out that in the
sequentialcognitive schema there is hardly a place for assimilationof new
knowledge, or for what is usually called meaningful learning. That is
probably why even our third task, which requiredrecallingthe answer to
the first question, seemed quite difficult as long as only the operational
representationwas available.
It is the static object-like representationwhich squeezes the operational
infornation into a compact whole and turns the cognitive schema into a
more convenient structure. To grasp the nature and the implications of
such change, compare schemata A and B in Figure 7 (naturally, I am not
trying to imply that these schemata are anything like faithful images of
"real" mental structures in which our knowledge is stored; I use the
picturesjust as a convenient means for elucidatingthe technicalaspects of
our claims). From a purely technical point of view, the mathematical
objects are the upper "nodes" in the hierarchicalschema resulting from
reification. Each of them serves as a single item in the catalogue of our
mind. For a cognizing person, they function like simplified pictures or
symbols, which can be seized at one glance and may be used instead of "the
real thing" (the correspondingprocess) at certain stages of problem-solv-
ing. Naturally, more often than not, these abstract constructscan only be
seen with our mind's eyes; most regrettably,very rarelyare we as fortunate
as in the case of our promenade and strolls, where it was possible to
actually draw the "reifying"pictures and symbols on the paper.

Definition: PromenadeP is the graph presented below.
I-2 -3-4 -5 If xis anode in P

I I I I I then S(x), N(x), W(x) and E(x)
6- 7- 8 - 9-10 are the adjacent nodes,
I I I I I placed south, north, west,
11-12-13-14-15 and east to x, respectively.
I I I I l
16-17-18-19-20 Stroll is defined like in
l I I I I box 5.
21-22-23-24-25
Fig. 6. The experiment- Part 2.
Schema A Schema B
Any informationcan be stored in many differentschemata.For example, the two schemata

pictured in this figurecontain the same information (representedby (1, 2, 3, 4, 5, 6, 7, 8)).
Schema A is sequential,shallow and wide. As a result of reificationit can be reorganized
into a deeper and narrowerstructure,such as Schema B. With the new organization, all
cognitive processes (retrieval and storing) become much faster.
Fig. 7. Different organizationsof a hierarchicalschema.
Both problem-solvingand learning processes may be effectively "navi-

gated" by the help of these compact, if not detailed, overall representations,
just as movements of a ship are controlled and directed with maps.
While tackling a genuinelycomplex problem, we do not always get far if
we start with concrete operations;more often than not it would be better
to turn first to the structuralversion of our concepts. These upper-level
representationsprovide us with a "generalview", so we can use our system
of abstract objects just like a person looking for information uses a
catalogue; or like anybody trying to get to a certain street consults a map
before actually going there. In other words, in problem-solvingprocesses

28 ANNA SFARD
the compact abstract entities serve as pointers to more detailed informa-

tion. Thus, almost any mathematicalactivity may be seen as an intricate
interplay between the operational and the structuralversions of the same
mathematicalideas: when a complex problem is being tackled, the solver
would repeatedlyswitch from one approachto the other in order to use his
knowledge as proficiently as possible. ParaphrasingHenrici (1974), who
spoke about dialectic and algorithmicmathematics,we may say that "the
structuralapproachinvites contemplation;the operationalapproachinvites
action; the structuralapproach generatesinsight; the operationalapproach
generates result". An excellent illustration for the object-navigatedprob-
lem-solving process can be found in Hadamard's account of his own
thinking (Hadamard, 1949, pp. 76-7): in a most persuasive way this
mathematiciantells the reader how he recapitulatesan argument for the
existence of infinitely many prime numbers, through switching back and
forth from objects (numbers and sets of numbers) to the underlying
computations.
Let us list now the beneficialeffects reificationcan have on learning.As
explainedbefore, formation of a structuralconception means reorganizing
the cognitive schema by adding new layers - by turning sequential aggre-
gates into hierarchicalstructures.Obviously, the deeper and narrowerthe
hierarchy,the greater the capacity of the schema. To have a better idea
about the change resulting from the restructuring,let us compare the two
schemata presented in Figure 7. In schema A, new items can hardly be
added because the number of "sons" of the upper-level node is already
quite large. In contrast, no node in schema B has yet been "saturated",
since the average quantity of "sons" does not surpass the "magic number
7 + 2" (accordingto Miller, 1956, this is the maximalnumberof chunks of
information that can be kept simultaneously in our working memory).
Thus, within the structuralapproach more room is available for inserting
new information. As a result, learning becomes more effective, more
meaningful.Also, the retrievalprocesses become faster when the necessary
information is stored in hierarchical tree-shaped structures. The reader
might be able to observe the differencewhile tackling again the third task
in Figure 5.
We are now in a position to give a tentativeanswerto the question about
the necessityof a structuralapproachasked at the beginningof this section:
in the light of the explanations and examples given so far it seems that
without the abstractobjects all our mental activity would be more difficult.
Since we are not super-computers,we just could not get along with very
complex processes without breaking them into small pieces and without

squeezing each part into a more manageable whole. In other words, the
distance between advancedcomputationalprocessesand the concrete mate-
rial entities which are the objects of the most elementaryprocesses (such as
counting) is much too large to be grasped by us in its totality. We
overcome this difficultyby creatinginterveningabstractobjects which serve
us as a kind of way-stations in our intellectualjourneys. These compact
cognitive entities effectively shield our working memory against overflow.
Abstract objects, neatly organized into a well-structuredcognitive schema,
were probably what allowed Poincare (1952, p. 51) to make the following
declaration: ". . . I can perceive the whole of the [lengthy mathematical]
argument at a glance. [Thus] I need no longer be afraid of forgetting one
of the elements; each of them will place itself naturally in the position
preparedfor it, without my having to make any effort of memory."
At certain stages of knowledge formation (or acquisition)the absence of
a structuralconception may hinderfurtherdevelopment.As the amount of
information grows, the old schema may become saturated and practically
imperviousto any enrichment.It was certainlynot by pure chance that the
transition from rhetorical to symbolic algebra - a transition from opera-
tional to structuralapproach in computationalmathematics- occurredin
the sixteenth century. And it was not just a historicalaccident that several
different systems of symbols were invented almost simultaneously by
independentlyworking mathematicians.By that time, too great a complex-
ity of computationalprocessesbrought the rhetoricalalgebrato a stalemate
and practicallyput an end to its development.Looking back even further,
we can venture a conjecturethat the absence of structuralrepresentations
(thus structuralconceptions) was one of the factors that slowed down the
development of computational science in Ancient Greece and caused
algebra's falling behind geometry for centuries.
I have just presented the structuralthinking as a very powerful weapon
against the limitations of our working memory. At a less technical, more
philosophical level, we can say that in mathematics,transition from pro-
cesses to abstract objects enhances our sense of understandingmathemat-
ics. After all, reificationincreasesproblem-solvingand learningabilities, so
the more structuralour approach,the deeperour confidencein what we are
doing. At least some of the readers may be able to convince themselves
about the accuracyof this claim by recallingthe flash of enlightenmentthey
probably experiencedwhen presentedwith the structuraldefinition of the
concept of stroll. To sum up, structural conception is probably what
underliesthe relationalunderstanding,defined by Skemp (1976) as "know-
ing both what and why to do", or having both rules and reasons. Purely

30 ANNA SFARD
operationalapproachwould usually give no more than instrumentalunder-

standing,once presentedby Skempas having ruleswithoutreasons.It should
immediately be stressed that this somewhat disdainful description (which
apparentlywas revisedby Skemphimselfafter a deeperthought was given to
the subject)does not do fulljustice to the kind of understandingwhichis both
valuableand inevitableat certainstagesof learning.This claimwill be clarified
and substantiatedin the last part of this section. Before I do this, however,I
shall push Skemp'soriginalidea a little further,and talk about a thirdkind of
understanding:reasonswithoutrules.What I mean by that is a kind of purely
intuitiveunderstanding,attainedin those rarecaseswhen the vaguestructural
conceptionis achievedbefore the operationalbasis has fully developed.This
is probablythe kind of understandingmathematicianshad when the earliest
versionsof the concept of function were introduced.As can be learnedfrom
the history,having "reasonswithout rules"may be not enough for creatinga
fully-fledgedmathematicaltheory,but it is certainlymost helpfulin discover-
ing theoremsand in decidingabout directionsof future development.
The roles and the featuresof operationaland structuralconceptionshave
been schematicallysummarizedin Figure 9.
4.3. TheInherentDifficultyof Reification

After showing the importance of structuralconceptions, we should now
return to the question asked in the previous section: why is reification
obviously so very difficult?Why mathematiciansthemselvesneeded several
centuriesto arriveat fully structuralversionsof the most basicconcepts,such
as numberor function?
The problemwill seem less puzzlingif we remindourselvesthat reification
is an ontologicalshift, a qualitativejump. Suchconceptualupheavalis always
a rathercomplex phenomenon, especiallywhen it is accompaniedby subtle
alternations of meaning and applications (which is usually the case; for
example, the structural set-theoretic definition of function considerably
extendedthe scope of the notion). The ability to see somethingfamiliarin a
totally new way is nevereasyto achieve.The difficultiesarisingwhena process
is converted into an object are, in a sense, like those experiencedduring
transition from one scientific paradigm to another; or - to make a less
ambitious but perhaps more persuasive comparison - they are like the
obstaclesencounteredby a person looking at a cube picturedon a paperand
tryingto perceiveit as if it was presentedfrom a differentangle(see Figure8).
Our three-phaseschema of concept formation would shed more light on
these difficulties.Accordingto the model, reificationof a givenprocessoccurs

If you feel that the cube presented

here is viewed from above -
try to convince yourself that
it is seen from below; and
vice versa.
Fig. 8. The cube.
simultaneouslywith the interiorizationof higher-levelprocesses. For exam-

ple in the case of negativenumbers,the reificationbecomesmost likelywhen
the algebraic operations on this kind of numbers are, at least partially,
interiorized.Indeed, what leads to recognizingsuch operationslike 2-5 and
0-6 as numbersis the similarityof the algorithmsinvolvingthese operations
to those performedon more familiarnumbers(namely the fact-that adding
and multiplyingtwo negative or miscellaneousquantitiesis very much like
adding or multiplying positive numbers). In order to notice the likeness,
however, one has to achieve some mastery in the operations on negative
quantities. Similarly,in order to see a function as an object, one must try
to manipulateit as a whole: there is no reason to turn process into object
unless we have some higher-level processes performed on this simpler
process. But here is a vicious circle:on one hand, without an attempt at the
higher-levelinteriorization,the reificationwill not occur;on the other hand,
existenceof objects on which the higher-levelprocessesare performedseems
indispensablefor the interiorization - without such objects the processes
must appearquite meaningless.In other words:the lower-levelreificationand
the higher-levelinteriorizationare prerequisitefor each other!
It seems that from the psychologicalpoint of view, what has just been said
can have quite importantimplications.The "viciouscircle"of reificationwill
help us now explain why for so many people "mathematicsat school [is] a
collection of unintelligiblerules which, if memorizedand applied correctly,
[lead] to 'the right answer'"(Skemp, 1971, p. 3).
The question of the order in which different mathematical abilities
should be developed has always drawn the attention of psychologists and
educators. As Kilpatrick (1988) put it, "One of the most venerable and
vexing issues in mathematics education concerns the trade-off between
proficiencyand comprehension,betweenpromoting the smooth performance
of mathematicalprocedureand developing understandingof how and why
that procedureworks and what it means .. ..". Differentanswershave been

32 ANNA SFARD
given to this question by differentpsychologicalschools. The generalstance

taken by behaviouristsimplied that skills should be learned, whereas the
more recent theories have seemed to prefer the view that "drill [should be]
recommended[only] when ideas and processes already understood are to
be practicedto increaseproficiency"(Brownell, 1935, p. 19). According to
our model of concept development,however, no clear order of abilities can
be established. The thesis of the "vicious circle" implies that one ability
cannot be fully developed without the other: on one hand, a person must
be quite skillful at performingalgorithmsin order to attain a good idea of
the "objects"involved in these algorithms;on the other hand, to gain full
technical mastery, one must already have these objects, since without them
the processes would seem meaninglessand thus difficultto performand to
remember.For example, the concept of complex number cannot be reified
until a person is able to make computationsinvolving these numbers;at the
same time, however, conceiving such constructs like i or 3 + 2i as fully-
fledged numbers (and not just symbols for operations "without result") is
a prerequisitefor being proficientin manipulatingthem. This statement is
in line with the results of a large-scale study carried out among 13- and
17-year olds in United States (Carpenter et al., 1980). According to the
findings, "the development of a skill is closely tied to understandingthe
concept underlyingthe skill".
In the light of these claims it should not surpriseus that ever so often,
"[s]tudents appear to be learning many mathematical skills at a rote
manipulation level and do not understand the concepts underlying the
computation"(ibid.) For instance,pupils can be quite successfulin compu-
tations involving fractions in spite of being unable to treat fractions as
numbers. Because of the complex nature of their mutual dependence, it
seems inevitable that in the process of learning, student's understanding-
this feeling of competence and mastery which accompanies the ability of
"seeing" abstract structures - will sometimes drop behind the technical
proficiency.This implies that in some cases the learnermust put up with a
certain amount of "mechanical"drill accompaniedby doubts about mean-
ing and by a feeling of insufficient(instrumentalonly) understanding.Even
professional mathematicianscannot escape this fate, and they sometimes
complain about the necessity of strugglinghard for meaning of ostensibly
simple ideas. Halmos (1985a) recalls the times when he was a university
student:in spite of "workingfuriously"on the concept of lambda-matrices,
he "didn't really begin to understandwhat the subject was about till four
or five years later" (pp. 40-1). We may conjecture that it was first
condensation, and then reificationthat took so long.

More often than not, both students and teachersfail to acknowledgethe

fact which is one of the most important implications of our three-phase
schema: insight cannot always be expected as an immediate reward for a
person'sdirectattempts to fathom a new idea. The reification,which brings
relationalunderstanding,is difficultto achieve, it requiresmuch effort, and
it may come when least expected, sometimes in a sudden flash. In his
pioneering book on the psychology of mathematics, Hadamard (1949)
mentions an "illumination effect" which may occur after a period of
intensive work followed by days of rest ("incubation period").
From the educational point of view, the main problemwith this delay in
reificationand with the resulting periods of doubts about meaning is that
they may bring a permanentharm - a life-long apprehensionof mathemat-
ics and a conviction that it cannot be learned. Some people may be unable
to recover from the shock caused already by the first encounter with the
problematicsituation. Those who are not preparedto actively struggle for
meaning (for reification) would soon resign themselves to never under-
standing mathematics. The ability of orchestratinglower-level reification
with higher-levelinteriorizationin a subtle, painless mannermay be one of
the most important features which make a person capable of coping with
mathematics."Mathematicallyfittest", even if they feel at times somewhat
shaky in their understanding,seem to have enough motivation, patience,
and intellectual discipline to put up with this situation in a trustful
anticipation of a salutary insight. It is certainly what can be learned from
Operationalconception Structuralconception
General a mathematicalentity a mathematicalentity is

characteristic is conceived as a product conceived as a static
of a certain process or is structure- as if it was
identifiedwith the a real object
process itself
Internal is supported by verbal is supported by visual

representations representations imagery
Its place develops at the first evolves from the
in concept stages of concept operational conception
development formation
Its role is necessary, but not facilitates all the cognitive
in cognitive sufficient,for effective processes (learning,
processes problem-solving& learning problem-solving)
Fig. 9. Operationaland structuralconceptions - summary.

34 ANNA SFARD
history. As Jourdain (1956) put it, "when logically-mindedmen objected"

to the "absurd" notions of negative and imaginary (complex) numbers,
"mathematicianssimply ignored them and said 'Go on; faith will come to
you'". Those who could see the inner beauty of the idea thought that the
new numbers,"though apparentlyuninterpretableand even self-contradic-
tory, musthave logic. So they [the numbers]were used with a faith that was
almost firm and was justified much later" (pp. 29-30).
In the light of the "vicious circle" thesis, it seems that in the search for
an improvementin mathematicseducation we should focus on the question
what, and how much, can be done to unravel the harmful tangle and to
stimulatereification.Also, we ought to ask ourselveswhat means should be
used to ensure that the students go safely through these doubts-about-
meaning periods, when they still feel certain uneasiness about the object
they have to manipulateand sense, as a result, that their understandingis
far from satisfactory. Another article will be devoted to these issues.
REFERENCES
Anderson, J. R.: 1976, Language, Memory,and Thought,Erlbaum, Hillsdale, NJ.
Behr, M., Erlwanger,S., and Nichols, E.: 1976, How childrenview equalitysentences(PMDC
Technical Report No. 3), Florida State University (ERIC Document ReproductionService
No. ED144802).
Bishop, A. J.: 1988, 'A review of research on visualisation in mathematics education', in
Proceedings of The Twelfth InternationalConferencefor the Psychology of Mathematics
Education,Hungary, Vol. 1, pp. 170-176.
Brownell, W. A.: 1935, 'Psychologicalconsiderationsin learning and teaching arithmetic',in
The Teachingof Arithmetic: Tenth Yearbookof the NCTM, Columbia University Press,
New York.
Cajori, F. A.: 1985, History of Mathematics,Fourth Edition, Chelsea Publishing Company,
New York.
Carpenter, T. P., Corbitt, M. K., Kepner, H. S., Lindquist, M. M., and Reys, R.: 1980,
'Results of the second NAEP mathematicsassessment:Secondaryschool', The Mathematics
Teacher73(5), 329-338.
Clements, K.: 1981, 'Visual imageryand school mathematics',For the Learningof Mathemat-
ics 2(2), 33-38.
Courant, R. and John, F.: 1965, Introductionto Calcuhlsand Analysis, Vol. I, Interscience
Publishers,New York.
Davis, R. B.: 1975, 'Cognitive processes involved in solving simple algebraic equations',
Journalof Children'sMathematicalBehavior1(3), 7-35.
Davis, P. J. and Hersh, R.: 1983, The MathematicalExperience,Penguin Books, London.
Dorfier, W.: 1987, 'Empirical investigation of the construction of cognitive schemata from
actions', in Proceedingsof the EleventhInternationalConferenceof PME, Vol. III, pp. 3-9.
Dorfier, W.: 1989, 'Protocols of actions as a cognitive tool for knowledge construction', in
Proceedingsof the ThirteenthInternationalConferenceof PME, Paris, Vol. 1, pp. 212-9.
Dubinsky, E. and Lewin, P.: 1986, 'Reflective abstraction and mathematicseducation: The
genetic decomposition of induction and compactness',Journalof MathematicalBehavior5,
55-92.

Eisenberg,T. and Dreyfus, T. (eds.): 1989, Visualizationin the mathematicscurriculum,Special

issue of Focus on LearningProblemsin Mathematics11(1&2).
Frege, G.: 1970, 'What is function', in Geach, P. and Black, M. (eds.), Translationsfrom the
PhilosophicalWritingsof GottlobFrege, Blackwell, Oxford.
Hadamard, J. S.: 1949, The Psychology of Invention in the Mathematics Field, Princeton
University Press, NJ.
Halmos, P. R.: 1985, 'Pure thought is better yet.. .', The College MathematicsJournal 16,
14-16.
Halmos, P. R.; 1985a, I Wantto be a Mathematician,An Autobiography,Springer,New York.
Henrici, P.: 1974, 'The influence of computing on mathematicalresearchand education', in
Proceedingsof Symposiain AppliedMathematics,Vol. 20, American MathematicalSociety,
Providence.
Hiebert, J. and LefevreP.: 1986, 'Conceptualand proceduralknowledge in mathematics:An
introductoryanalysis',in Hiebert, J. (ed.), Conceptualand ProceduralKnowledge:The Case
of Mathematics,Eribaum, Hillsdale, NJ.
Jourdain, P. E. B.: 1956, 'The nature of mathematics',in Newman, J. R. (ed.), The Worldof
Mathematics,Simon and Schuster, New York.
Kaput, J. J.: 1979, 'Mathematicsand learning:Roots of epistemologicalstatus', in Lochhead,
J. and Clement, J. (eds.), CognitiveProcess Instruction,Franklin Institute Press.
Kieran, C.: 1981, 'Concepts associated with the equality symbol', EducationalStudies in
Mathematics12(3), 317-326.
Kilpatrick,J.: 1988, 'Editorial',Journalfor Researchin MathematicsEducation19(4).
Kleiner, I.: 1988, 'Evolution of the function concept: A brief survey', College Mathematics
Journal20(4), 282-300.
Lesh, R. and Landau, M. (eds.): 1983, Acquisitionof MathematicsConceptsand Processes,
Academic Press, New York.
Markovits, Z., Eylon, B., and Bruckheimer,M.: 1986, 'Functions today and yesterday',For
the Learningof Mathematics6(2), 18-24.
Maurer, S. B.: 1985, 'The algorithmic way of life is best', College MathematicsJournal 16,
2-5.
Miller, G. A.: 1956, 'The magic number seven plus minus two: Some limits on our capacity
for processing information', PsychologicalReview63, 81-96.
Otte, M.: 1984, 'Komplementaritat',Dialektik 8, 60-75.
Paivio, A.: 1971, Imagery and VerbalProcesses, Holt, Rinehart, and Winston, New York.
Penrose, R.: 1989, The Emperor'sNew Mind, Oxford University Press, Oxford.
Piaget, J.: 1952, The Child's Conceptionof Number,Routledge and Kegan, London.
Piaget, J.: 1970, Genetic Epistemology,W. W. Norton, New York.
Poincare, H.: 1952, Science and Method, Dover Publications, New York.
Sfard, A.: 1987, 'Two conceptions of mathematicalnotions: operational and structural',in
Proceedingsof the EleventhInternationalConferenceof PME, Montreal, Vol. 3, pp. 162-9.
Sfard, A.: 1988, 'Operationalvs structuralmethod of teaching mathematics:A case study', in
Proceedingsof the TwelfthInternationalConferenceof PME, Hungary, pp. 560-7.
Sfard, A.: 1989, 'Transitionfrom operationalto structuralconception:the notion of function
revisited',in Proceedingsof the ThirteenthInternationalConferenceof PME, Paris, Vol. 3,
pp. 151-8.
Sinclair,H. and Sinclair A.: 1986, 'Children'smastery of written numeralsand the construc-
tion of basic numberconcepts', in Hiebert, J. (ed.), Conceptualand ProceduralKnowledge:
The Case of Mathematics,Eribaum,Hillsdale, N.J.
Skemp, R. R.: 1971, The Psychology of Learning Mathematics, Penguin Books, Har-
mondsworth, England.

36 ANNA SFARD
Skemp, R. R.: 1976, 'Relational understandingand instrumentalunderstanding',Mathematics

Teacher77, 20-26.
Stein, S. K.: 1988, 'Gersham'slaw: Algorithm drives out thought', Journal of Mathematical
Behavior7, 79-84.
Steiner,H.-G.: 1985, 'Theoryof mathematicseducation:An introduction',For the Learningof
Mathematics5(2), 11-17.
Tahta, D. (ed.): 1972, A Boolean Anthology,Derby, ATM.
Thompson, P. W.: 1985, 'Experience,problem solving, and learning mathematics:consider-
ations in developing mathematicalcurricula',in E. A. Silver (ed.), Teachingand Learning
MathematicalProblemSolving: Multiple ResearchPerspectives,Erlbaum, Hillsdale, N.J.
Vinner, S. and Dreyfus, T.: 1989, 'Images and definitionsfor the concept of function', Journal
for Researchin MathematicsEducation20(5), 356-66.
The Science TeachingCentre

The Hebrew Universityof Jerusalem
Givat-Ram,Jerusalem91904
Israel


On The Dual Nature of Mathematical Conce

Uploaded by

Copyright:

Available Formats

On The Dual Nature of Mathematical Conce

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

On The Dual Nature of Mathematical Conce

Uploaded by

Copyright:

Available Formats

On the Dual Nature of Mathematical Conceptions: Reflections on Processes and Objects as

Different Sides of the Same Coin

This content downloaded from 132.74.40.78 on Mon, 29 Apr 2013 02:13:37 AM