JOURNAL OF MATHEMATICAL BEHAVIOR 14, 15-39(1995)

The Development of Algebra: Confronting

Historical and Psychological Perspectives
ANNA

SFARD

The Hebrew University of Jerusalem

COMBINING

HISTORY AND PSYCHOGENESIS

Why did the developmentof algebralag behind geometryfor so many centuries?

Why do todays pupils have difficulties with even the simplest word problems?
What preventedgenerationsof mathematiciansfrom acceptingthe idea of the
irrational and the negativenumbers?What arethe roots of thedifficulties experienced by studentsconfronted with the conceptof complex number for the first
time?
It is neitherby chance,nor by merecarelessness,thatmy list of questionsis a
mixture of psychological and historical puzzles. As differentas they seemat first
glance, thesetwo sets of problems may in fact have much in common. Indeed,
there are good reasons to expect that, when scrutinized, the phylogeny and
ontogeny of mathematicswill reveal more than marginal similarities. At least,
this is what follows from the constructivist view according to which learning
consists in the reconstructionof knowledge.
Piaget-one of the first and most outspokenprotagonistsof constructivism,
and thus of the thesis that the historical-critical and psychogenetic studies
[should] converge(Garcia & Piaget, 1989,p. 108)-grounds his position in the
claim that
theadvances
madein the courseof history of scientific thoughtfrom one period to
the next, do not, exceptin rareinstances,follow eachotherin randomfashion, but
can be seriated,as in psychogenesis,in the form of sequentialstages.(p. 28)

It is probably becauseof the inherentpropertiesof knowledge itself, becauseof

the nature of the relationship betweenits different levels, that similar recurrent
phenomenacan be tracedthroughoutits historical developmentand its individual
reconstruction. For the same reason, difficulties experiencedby an individual
learner at different stagesof knowledge formation may be quite close to those

I amindebted
to CarolynKieranfor herinspiration
andencouragement
andto IsraelKleiner
for numerous
helpfulremarks.
Correspondence
andrequests
for reprintsshouldbesentto AanaSfard,ScienceTeaching
Centre,
The HebrewUniversityof Jerusalem,GivatRam,Jerusalem91904,Israel.
15

16

SFARD

that once challenged generationsof mathematicians.The parallel terms, epistemic subject and collective epistemic subject usedby someresearchersto distinguish between the individual learner and the collective of the creators of
knowledge, bespeakthe widespreadbelief in thesesimilarities.
In this article I make a very quick journey through the history of algebra,
trying to detect certain recurrent phenomena in the developmentof abstract
ideas. The point of departurefor this patternfinding will be a theoreticalmodel
according to which the formation of mathematicalknowledge is more or less a
cyclic process,a processin which thetransitionsfrom one level to anotherfollow
some constant course. The particular scheme that will be used here pictures
mathematicsas a hierarchy in which what is conceivedoperationally (i.e., as a
computationalprocess)on one level is reified into an abstractobject and conceived structurally on a higher level. The idea of an operational-structuralduality of mathematicalconceptswith its numerousimplications was presentedin
detail in Sfard (1991, 1992).For the convenienceof thereader,a summaryof the
relevant elementsof this framework is given in the Appendix. While traveling
through the centuries, I confront, wheneverpossible, historical developments
with examplestaken from empirical studies on the ways in which todays students learn the subject.
During thehastyflight over history,our telescopewill be directedat what may
be consideredturning points in the developmentof algebraicthinking. Here I try
to fathom not only the mechanismsthat put such developmentsin motion, but
also the natureand the sourceof the cognitive difficulties which invariably pop
up whenever a crucial step forward is to be made. This topic deservesspecial
attentionbecausethe difficulties seem so ubiquitous both in history and in the
classroom that they ought to be regardedas a regular part of the process of
knowledge constructionratherthan as a madnesswith no methodin it. Indeed,
when history is considered,what seemsto be the most striking common characteristicof the many ways in which new ideasenteredthe sceneand then evolved,
is the greatdeal of distrustand reluctancewith which the candidatesfor citizenship in the kingdom of mathematicswere invariably greeted. (The turbulent
evolution of such conceptsas function or number may serve as good examples;
see, e.g., Hefendehl-Hebeker, 1991; Kleiner, 1988, 1989.)
According to a widespreadbelief often expressedby historians (e.g., Boyer,
1985),it was the lack of logical foundationsthatobstructedthe acceptanceof the
new typesof numbers. However, the schemeof conceptdevelopment,as well as
mathematiciansown utterances,suggest an additional explanation: In some
casesthe resistanceto a new abstractobject might havebeenof ontological rather
than of purely logical origins. It could stemfrom the inability to reify a process.
Reification is an act of turning computationaloperationsinto permanentobjectlike entities. For example,a complex number is born only when a personis able
to view the processof extractingthe squareroot of a negativenumber as a real

THE DEVELOPMENT

OF ALGEBRA

17

entity, as a permanentthing in its own right. To somepeople, all this may seemto
be conjuring up a new thing out of nothing.
Reification is a major change in the way of looking at things and as such is
inherently difficult to achieve.There are severaltypesof seriousobstaclesthatlie
in wait for those who dare to speak about new abstractobjects (see the Appendix). A revolutionary changein basic beliefs on the natureof mathematicsmust
sometimesoccur before the new idea is fully accepted.A natural resistanceto
upheavalsin tacit epistemologicaland ontological assumptions,which so often
obstructedthe historical growth of mathematics,can hardly be preventedfrom
appearingin the classroom.
In the next section I concentrateon algebra.Its evolution will be presentedas
a constant(but not necessarilyconscious) attemptat turning computationalprocedures into mathematicalobjects, accompaniedby a strenuous struggle for
reification. I hope that from this birds_eye view of history a lesson of some
practical importancewill be learnedregardingthe natureand sourcesof the traps
lying in wait for todaysstudentsthroughoutthe curriculum. To thosewho shrug
at my bold (some would say presumptuous)attemptto deal in one shortpresentation with the whole of thedevelopmentof algebra,and from a dual perspectiveat
that, let me say that history will be used here only to the extent which is
necessaryto substantiatethe claims abouthistorical and psychological parallels.
No more than a very general view of algebra will be presented.
STAGE 1: FROM ANTIQUITY
TO RENAISSANCE-TOWARD
THE SCIENCE OF GENERALIZED
NUMERICAL
COMPUTATIONS
What Is Algebra?
When and where did algebra begin? The literature provides more than one
opinion on this matter.There are many historiansof mathematicswho tracethe
origins of algebra to various nations of antiquity: the Assyrians, Babylonians,
Egyptians. Others, with more critical judgment, locatetheseorigins at the school
of Alexandria say Garcia and Piaget (1989, p. 143), immediately expressing
their disagreementand saying that for them algebra is a much more recent
invention. Were algebrareally known to the Greeks, they argue, pre-Euclidean
and Euclidean geometry,fairly well developedanyway,would haveopenedup to
becomeeven more impressiveachievements:It is clear that the difficulties the
Greeks encounteredin resolving their numerous geometricalproblems can be
explained only by the absenceof a science of algebra (p. 143). These words
indicate that the authorsdisagreementwith the others stemsnot so much from
different historical information as from the fact that they obviously have their
own interpretationof the term algebra. An answer to the question What is
algebra? must thereforeprecedeany historical account.

18

SFARD

The majority of authorsseemto be quite unanimous as to the early origins of

algebrabecausethey spot algebraicthinking whereveran attemptis madeto treat
computationalprocessesin a somehowgeneral way. Generality is one of these
salient characteristicsthat make algebradifferent from arithmetic.Boyer (1985)
explains his decision to call someproblemssolved in ancientEgypt algebraic by
saying that they do not concern specific concreteobjects, such as bread and
beer, nor do they call for operationson known numbers(p. 16).Novy (1973),in
the contextof somewhatlaterdevelopments,repeatedlystatesthatthe searchfor
a general [italics added] solution of . . . equationswas one of the two main
themesof algebra. The concept and the definition of realms of numbers was
the other (p. 25). Like Boyer, Kline (1972) agreesthat algebraicmethodswere
used as early as in ancient Mesopotamiaand Egypt, and like Novy he grounds
this claim in the fact that what was done theremay be interpretedin the modem
languageas solving equationsin a generalway: Though only concreteexamples
were given, many were intendedto illustratea generalprocedurefor quadratics
(P. 9).
On one point, therefore,there seemsto be perfect agreementamong all the
authors,including Garcia and Piaget: Algebra is a scienceof generalizedcomputations. Thus, the differencesmay only havetheir roots in the opinions aboutthe
means necessaryto implement algebraic methods.For Garcia and Piaget, the
symbolic notation, never heard of in Babylonia or Egypt, is clearly part and
parcel of the branch of mathematicscalled algebra. Similarly, Unguru (1975,
p. 77) names operational symbolism one of the main featuresof algebraic
way of thinking. According to otherauthors,the modem algebraicsymbols are
not the only possible vehicle of generality.
In this article, I join the latter school of thought.I use the term algebrawith
respectto any kind of mathematicalendeavorconcernedwith generalizedcomputational processes,whatever the tools used to convey this generality. This
definition brings into full relief the operationalorigins of algebraicthinking. In
the following sectionsthe history of algebrawill be presentedas a sequenceof
stepstoward ever greatergenerality and, at the sametime, toward structurality.
The three stagesthat will be listed-rhetorical and syncopatedalgebra, Vietan
symbolic algebra, and abstractalgebra-correspond roughly to what is taught
today at, respectively,the primary, secondary,and tertiary levels.
Rhetorical and Syncopated Algebras
For many people, even if well versedin mathematics,it often comesas a surprise
to learn that algebraic notation, which in our minds seems inseparablefrom
algebraitself, is quite a recentinvention. Until the 16thcentury, computational
processes,whether generalized or not, were presentedeither verbally or in a
mixture of words and symbols. Two examplesof theseearly ways of expressing
algebraic thought appearin Figure 1. The first, taken from Diophantus (c. 250
AD), includes somesymbols and thusrepresents
the so called syncopatedalgebra.

THE DEVELOPMENT

OF ALGEBRA

19

I, Syncopated algebra: from Diophantus, Arithmetica, c. 250 AD

To find two numbers such that their sum and productare given numbers:
Given sum 20, given product 96. 2x the difference of the required. Therefore
the numbers are 10 + x, 10 - x. Hence 100 - x2 is 96. Therefore,x is 2 and the
required numbers are 12, 8.
Remark: In fact, Diophantusused Greek lettersas symbols. We translatedthem
(after Fauvel and Grey, 1987,p. 218) into the modem signs for the sake of
clarity.
2. Rhetorical algebra: from Aryabhatia by Aryabhata, AD 499
To find the number of elementsin the arithmeticprogressionthe sum of which
is given: Multiply the sum of the progressionby eight times the common
difference, add the squareof the differencebetweentwice the first term, and the
common difference, take the squareroot of this, subtracttwice the first term,
divide by the common difference,add one, divide by two.
3. Geometric Algebra
a. Greek proof of identity
equivalentto
(a - b)(a + b) = a2 - b2

b. Greek solution to the problem

equivalentto the equation
x2 = ab

Figure 1. Rhetorical and syncopatedalgebra.

The second, coming from the Hindu Aryabhatia (AD 499), is purely verbal and a!
such belongs to the kind of algebraknown as rhetorical.
The presymbolic algebrathatbegan, as was alreadymentioned,almost4,00(
years ago in ancient Egypt and Mesopotamia developed-even if only mod,
erately-in the Greece of Pythagoras and Euclid, made a considerable steI

20

SFARD

forward thanksto the work of Diophantus,and for severalcenturiesflourished in

the medieval Hindu, Arab, and Italian writings. The salient trait of this algebra
was its predominantly operational character.Garcia and Piaget would add the
prefix intra to the operational to emphasizethe fact that the computational
processeswere observedfrom inside ratherthan from a higher level perspective. Its sole concern was finding generalprescriptionsfor computingunknown
values out of concretenumerical data. Thus, algorithmswere soughtthat could
be used for entire families of problems.
For all their pursuit of generality,however,the ancientand eventhe medieval
mathematiciansusually explainedtheir computationalmethodsthroughconcrete
numerical examplesratherthan by universal prescriptions.Diophantussolution
in Figure 1 is a representativeexampleof this. Although the problem was stated
in general terms, concretenumbers were chosen to explain the solution; other
questions of the samekind could from then on be worked out by analogy-by
substitutionof new numbersinsteadof thosechosenin the example.Sometimes,
a special name was given to the unknown quantity: Length and width evidently
servedthis purposein ancientMesopotamia,a letterplayedthis role in Diophantus Arithmetica, root appearedin Al-Khowarizmis (9th century) Al-jabr, and
tunto in thewritings of such Italian algebraistsas Bombelli (c. 1526-1573).Only
rarely, however, and relatively late, could an author be found who, like Aryabhatain Example 2, addressedboth the given and the sought-fornumbersby
generalnames, thus paving the way for the later idea of a variable. In theserare
cases,the languageusedby the writer might be quite imaginativeas exemplified
by the following instruction for finding the fourth item in a proportion:Multiply
the fruit by the desire and divide by the measure.The result will be the fruit of
the desire (Aryubhutiu, from Boyer, 1985, p. 233).
Thus, until the 16th century the developmentof algebra was marked not
by changes in either the general characterof the endeavoror in the methods
employed-these remainedbasically the same for more than 2 millennia-but
by the gradual increase in the complexity of the investigated computational
processes.What beganwith an equivalentof solution of linear equationsof the
form x + ax = b or x + ax + bu = c (Boyer, 1985,p. 16)in theAhmes Papyrus
(c. 1650 BC), continued with quadratics everywhere from Mesopotamia to
Greece to the medieval East and Europe, and ended with truly complicated
prescriptionsfor solving equationsof the third and fourth degreein CardansArs
Magna (1545; see Cardan, 1968).
It is quite obvious thatin comparisonto the modem notation, whereformulas
such as 12.3 + 2V% or u2 + 23b concisely representboth a computational
processand its product and thus facilitate a transition to structuralthinking, the
rhetoricaland syncopatedexpressions,with their prolixity and tediously sequential character,impose an operationaloutlook. The operationalmodeof thinking,
however, puts a substantialburden on the working memory and thus is more
strenuousand less effectivethan the structuralapproachinduced by the modern

THE DEVELOPMENT OF ALGEBRA

21

notation. This may be clearly seen when the Aryabhatas prescription is confronted with its symbolic counterpart.It would be quite reasonable,therefore,to
expectthat the moment todays studentgains accessto the algebraicsymbolism
he or she becomeswilling to use it in every possible context.We can only guess
that this is exactly what Cardan would have done had algebraic symbols been
available to him at the time he toiled to explain the solutions of the cubic and
quartic equations.
More than one empirical studyhas shown, however,thatin reality things look
quite different. What was noticed for the first time in a series of experiments
performed by Clement and his colleagues (Clement, Lochhead, & Soloway,
1979; Soloway, Lochhead, & Clement, 1982)found its further confirmation in
the systematicstudiesby myself (Sfard, 1987)and by Harper (1987).All thedata
unanimously showed that even pupils with several years of (symbolic) algebra
behind them may do better with verbal than with symbolic methods.In 1979,
Clement and his colleaguesdiscoveredthata largeproportionof college students
could not translate such simple sentencesas There are six times as many
studentsas professorsinto equationsand in 1982they found thatstudentswere
much more successful when required to write appropriatecomputerprograms.
These observations led me to the conjecture that it may be the operational
characterof the computerencodingthatmakesthis seeminglyawkwardrepresentation somehoweasierfor pupils than the structuralalgebraicsymbolism. To test
this supposition, an experimentwas carried out in which two groups of highschool students,14 to 17 yearsof age, werepresentedwith questionslike theone
presentedin Figure 2. The participantswere askedto choose, from among three
possibilities, a formula that matchedthe situation described in a problem. In
anotherquestionnaire,they were requiredto find a verbal prescriptionfor solving
a similar problem. Both groups succeededin the latterkind of task significantly
better than in the former. In light of these findings, it no longer comes as a
surprise that, as was shown by Harper, studentsoften choose the rhetorical
methodif not obliged to use algebraicsymbols. In his experiment,Harper asked
a group of pupils of different agesto solve one of Diophantusproblems(similar

STRUCTURAL SOLUTION

PROBLEM
In a classtheboys
outnumber
thegirls

To find thenumberof

x =numberof girls

girlswe haveto:

y = numberof boys

by four

a. add4 to thenumberof boys

OPERATIONAL SOLUTION

b. subtract
4 from thenumberof

ax+$=y

boys
c. noneof theabove

b.x=y+rl
c.y=-x+4

Figure 2. A Problemusedin theexperiment

by Sfard(1987)

22

SFARD

to the one presentedin Figure 1). He notedthatnot only in theyoungeststudents,

but also amongthe older therewas a distinct preferencefor verbal prescriptions.
In all theseexperimentsthe authorsemphasizethat the discoveredphenomenon cannot be regardedas a mereoutcomeof classroomexperiencebecausethe
studentswere never trained in constructingverbal solutions to word problems.
Thus, the rhetorical methodwas used spontaneously,independentlyof instruction. It seems,therefore,thatthe precedenceof operationalover structuralthinking must be, at least in this case, one of thosedevelopmentalinvariants we are
looking for in this article-it was observed in the historical developmentof
mathematicsas well as in the processof individual learning. All thesefindings
speakwith force for the thesis of the inherentdifficulty of the transition from an
operationalto a structuralapproach.In thenext chapter,I look moredeeplyat the
impedimentsto progressat thejunction betweenrhetorical/syncopatedand symbolic algebra.
Geometric Algebra
Our account of presymbolic algebra would be seriously incomplete without a
mention of the so called geometric aZgebru, a very specialbreedof mathematics
which developed in ancient Greece. Its story enlightens anotheraspectof the
difficulty with reification.
As its name suggests,geometricalgebrawas a result of a mergerbetweenthe
two central componentsof ancient mathematics.Basically, it consistedin interpreting quantities expressedwith letters as lengths of line segmentsand the
operationson thesequantitiesas finding lengths, areas,or volumesof the figures
built from these segments. Thus, solving equations could be translatedinto
finding the geometricalconstructionthatwould producea segmentof the length
equal to the sought-forquantity (see Example 3b in Figure 1). Algebraic identities could also be proved in this way (seeExample 3a). To be sure, someauthors
opposethe view thatgeometricalalgebrawas algebradressedup and claim that
the problemsdealtwith were essentiallygeometricand were tackledonly for this
reason (Unguru, 1975, p. 69). This discussion, however, is irrelevant to the
presentsubjectbecauseit was agreedthatall thoseeventsthatcontributedto the
science of generalizedcomputations,even if only indirectly, qualify to be included in the historical account.
The desire to marry the science of computationswith geometryis not much
younger than the two disciplines themselves,and geometricalgebrais one of its
first results. The main reason usually brought by historians (see, e.g., Boyer,
1985; Kline, 1972, 1980)to explain this ancient urge for unification is the fact
that, unlike algebra, geometry was considered in Greece to be a paragon of
consistency and mathematicalrigor. Those who tackled concretequantitiesand
unknowns throughlengthsand areasevidently hopedthat in this way the advantagesof geometrywould be transmittedto the scienceof computations.Another
explanation brings the problematic idea of irrational number as the trigger for
separatingthe concept of number from that of continuous magnitude. Because

THE DEVELOPMENT

OF ALGEBRA

23

this distinction first arose in the context of incommensurability, it was only

natural to view the continuous quantities as tightly connectedto the realm of
geometry.
Our model of concept developmentsuggestsan additional explanationof the
phenomenonof geometricalgebra.Greek geometry,with its thinking embodied
in, fused with graphic, diagrammaticrepresentation(Unguru, 1975,p. 76) was
clearly at its structuralstage,whereasalgebra,preoccupiedwith verbally representedcomputationalprocesses,could be conceivedin no otherway than operationally. The structurality of the geometry facilitated thinking and enabled
effectiveinvestigations. The operationalrhetoricof algebramadeit cumbersome
and unyielding. No deeply penetrating,generalizing insight was possible. What
algebraneededfor further developmentwas reification of its basic concepts.At
this time, no bettermeanswere available to help in reification of the growingly
complex computationsthan the palpable geometricobjects. Geometric figures
renderedsome kind of tangible existencenot only to the idea of irrational quantity, but also to the elusive concept of variable magnitude. However, although
initially helpful, these tools soon proved restrictive. They createda systemof
prohibitions that greatly limited the range of problems qualifying for algebraic
treatment.For instance, becausethe unknown was usually referredto as length,
its squareas area, and its cube as volume, adding differentpowersremainedfor
some time entirely out of the question. Also, no power greaterthan 3 was
admittedin calculations.
Once again an important lesson can be learned from history by teachersand
psychologists. The current studies on visualization (e.g., Dreyfus, 1991) leave
little doubt as to the effectivenessof graphical representationseven in learning
such abstract subjects as algebra. No wonder, then, that the Greeks found it
useful to give numerical computationsa geometricalinterpretation.For the same
reason, graphical meansare offeredtoday to thosewho teachalgebra.However,
while employing geometry to support the science of computation we should
rememberthat, if used without precautionsand treatedtoo literally, the models
may become restrictive rather than helpful. The following declarationby Bell
(1951) is pertinent here: Real mischief is done when the credulous pupils
acquire an ineradicable belief that their purely metaphoricallanguagedescribes
an existentspaceor an objective reality (p. 140).This statementmay sound
too emphatic but, stripped of its exaggeratedrhetoric, what it really says is
probably this: Algebra is an inherently abstractdiscipline and one cannotescape
teaching it as such.
STAGE 2: FROM VIETE TO PEACOCK-ALGEBRA
A SCIENCE OF UNIVERSAL
COMPUTATIONS

AS

Vi&es Invention: Variable as a Given

To find the right tools for the reification of generalizedcomputationalprocedures
took many centuries,which shows once morehow genuinely difficult theprocess

24

SFARD

was. How the structuralstagein algebrawas eventually attainedis told in this

section. It is the story of modem algebraic symbolism.
Today historians seem united in the opinion that althoughletterswere often
used by mathematiciansbefore the 16thcentury, it was the way FranGois Vi&e
(1540- 1603)employedthemwhich madethereal.difference.The French mathematician was the first to replace numerical givens with symbols. To put it in
modem language,Vi&e was theinventor of parametricequations,equationswith
literal coefficients. Until this point, letterswere used in algebrato symbolize the
sought-forunknown quantities. Viete decidedto denotethemby vowels. Given
numbers, namely thosewhich were assumedto be known and provided as data,
had to be representedby consonants.Thanks to this convention, entire families
of problems (equations)could now be dealt with by means of concisely stated
algorithms. Thus, the introduction of parameterwas a great step toward the
generality so intensely pursued in mathematicsin general and in algebra in
particular. To use Boyers (1985) imaginative expression, Vibtes givens
helped convert algebrafrom Diophantus bag of tricks into a genuine science
of generalcomputations(p. 334). V&e himself was awareof the fact thathe had
added a floor to the hierarchical edifice of mathematicalgeneralization and
abstraction.According to his own description, whereasarithmeticis the science
of concretenumbers(logica numerosa), his type of algebrais a scienceof species
(logica speciosa) or of typesof things ratherthanof the things themselves.Thus,
this is probably where the concept of variable was born.
To fully appreciateVibtes achievementone has to consider its impact on
mathematicsin general. Employing letters as givens, togetherwith the subsequentsymbolism for operationsand relations,condensedand reified thewhole of
existing algebraicknowledge in a way that made it possible to handle it almost
effortlessly, and thus to use it as a convenient basis for entirely new layers of
mathematics.In algebraitself, symbolically representedequationssoon turned
into objectsof investigationin their own right and the purely operationalmethod
of solving problems (by reversecalculations) was replacedby formal manipulations on propositional formulas. These manipulationsare addressedin the sequel
as secondaryoperations, as opposedto the underlying arithmeticprocessesreferred to as primary.
The advent of symbolic algebra was soon followed by the emergenceof an
entirely new kind of natural science. For the first time in history, mathematics
had the meansfor dealing with changing magnitudesand not just constantquantities. It was only natural that scientistswould seize the new invention to represent all kinds of natural processes.Along with the intense investigation of the
mathematicalstructureof physical movement,the conceptof function, innately
and indissolubly linked with the idea of variable, began to arise. It is this
developmentthat made it possible for physics to be translatedinto the precise
languageof mathematics.
The impact of symbolic algebrawas also felt in geometry.As emphasizedin

THE DEVELOPMENT

OF ALGEBRA

2.5

the previous section, this ancient branch of mathematics,with its easily visualizable basic objects, was up to this point predominantly structural. According to
our model of abstractknowledge formation, the next step in the developmentof
geometry should be a transition to an operationalmode of thinking at a higher
level. To attainmore generality,it had to detachitself from concretetrianglesand
pyramids and concentrateon the constructions and transformationsby which
theseprimary objectsare governed.The meansfor this could only come from an
independentlydeveloping algebra: Only algebra . . . would have enabled [the
Greeks] to formulate [certainunsolved geometrical]problemsin termsof operations (Garcia & Piaget, 1989, p. 143). And indeed, the transition from operational to structural thinking in algebra was soon echoed in a substantial step
toward higher level operational thinking in geometry.Descartes(1596-1650)
and Fermat (1601-1665) were thosewho employedsymbolic algebrain geometry for the first time. Geometrical figures and their transformationswere representedthroughthe appropriatecomputationalprocesses.This invention was later
named analytic geometrybecausethe methodof investigationbasedon manipulating algebraic symbols (which later included differential and integral calculus)
becameknown as analysis. One may say that in this way algebra, which once
turned to geometryfor help in reification and verification, and which thus came
to be viewed as a minor appendageto geometry(Kline, 1980, p. 123), could
eventually pay its debt. It reciprocatedwith the means for capturing generality
and conveying operational thought. According to Kline, if Greek algebra expressed through geometry was called geometric, then Descartes invention
should be known as algebraicgeometry.(Unfortunately,this name was given to
anotherbranch of mathematicswhich developedmuch later.) Such terminology
would aptly reflect the symmetry of mutual services renderedby algebra and
geometryto each other.
To our modem eyes, the idea of a variable as any number seemsso obvious
and simple, we can hardly understandwhy it did not appear many centuries
earlier. After all, letterswere used in mathematicsalready in antiquity (e.g., in
Euclids Elements, c. 300 BC.). This fact, however,becomesmuch less surprising when one realizes that a variable as a given imposesfunctional thinking-it
requiresan ability to think simultaneouslyaboutentirefamilies of numbersrather
than aboutany specific quantity.Thus, the introductionof a parameterdemandsa
very sharpchangeof perspectivefor which structuralunderstandingof computational processesis indispensable.
In this context, it is worth mentioning that Vi&es invention was not always
fully appreciatedby historians of mathematics.(Garcia & Piaget credit Jacob
Klein and his book Die griechische Logistic und die Entstehung der Algebra,
1934, with the first attemptat reassessment.)The same mechanismthat concealed the import of Vietes achievementfrom the eyesof historiansoften hides
from todays teachersthe height of the step to be climbed by studentswhen
parametricequations are presentedto them for the first time. Such ignorance

26

SFARD

sometimeshas graveconsequences.I can clearly remembera traumaticexperience of my own. It happenedmany yearsago when I was teachingsimultaneous
linear equationsto two groups of 10thgraders.At thattime I was quite insensitive to the huge conceptualdifferencebetweenequationswith numerical coefficients and equationswith parameters.With my rich and versatileexperienceof
variables and functions, I had come to treat these two kinds of problems as
almost indistinguishable. And, obviously, so did the authorsof the textbookI
was using. They scatteredproblems with parametersall over the chapter,concealing them among otherquestions.They did it without warning. Not knowing
what I was doing, I gave my pupils someof theparametricequationsfor homework. The price of this ill-calculated deedwas high: For a fortnight, I was stuck
doing things I had never planned. My studentswould not let me talk about
anything other than problems with parameters.They could not cope with this
kind of task themselvesand one or two exampleswith adequategeneralexplanations were obviously not enough. After five or six meetingsand two testsdevoted
solely to this topic, my pupils still seemedsomewhatshaky in their understanding. The fact thatthe difficulties I witnessedwerenot somethingparticularto my
students(or, for that matter,to me as a teacher)was quite obvious for several
reasons. First, exactly the same happened quite independently in both my
groups. Second, my colleaguesreportedsimilar experiences.After 2 weeks of
grappling with the difficulties, I could clearly seetheir deeproots. Eventually, I
became aware of the vast conceptual change that occurs during the transition
from the concreteto parametricallygiven problems. I learned it the hard way.
This story is an anecdoterather than a piece of scientifically designed research.More systematicevidencefor the existenceof the problem may be found
in studies by Lee and Wheeler (1989), Booth (1988), and the Assessmentof
Performance Unit (1980). The aforementionedinvestigation by Harper (1987)
shows, along with thepupils tendencyto userhetoricalalgebra,their inability to
use the Vietan kind of variable. Nearly half of the oldestparticipantsin Harpers
experiment,when askedto show thattwo numbersmay always be found if their
sum and differenceare given, preferredthe Diophantantype of argumentto the
Vietan: They chose arbitrary concretenumbers ratherthan lettersfor their givens. This researchis one of the few studiesthatmakesexplicit use of history to
predict studentsbehavior.
There is anotheraspectof the passageto symbolic algebrathat has attracted
the attentionof researchers.More often than not, solving equationsin a rhetorical way was based on reversing computationalprocesses,or undoing what was
done to the unknown quantity (see Example 1 in Figure 1). Much evidencehas
been collected for the particular difficulty of the transition from such a working
backwardtechniqueto the methodinvolving the so-calledpermissibleoperations
on both sides of an equation. A survey of the relevant researchwas given by
Kieran (1992): A major turnaroundmust occur [in algebra]when studentsare
askedto think in terms of the forward operationsthat representthe structureof

THE DEVELOPMENT

OF ALGEBRA

27

the problem rather than in terms of the solving operations [which reversethe
process of computation] (p. 403). This turnaroundcorrespondsto the point in
history where rhetorical algebra gave way to the symbolic. The transition is
problematic becauseit requires this difficult change of perspectivethat has already been mentionedseveraltimes in this article: Operationalthinking must be
replaced by structural.
Peacock and the British School: Dearithmetization
of Algebra
Although soon after Viete symbolic algebrabeganto flourish, someof the most
prominent thinkers voiced their qualms about using it. Newton, for example,
claimed that algebra is the analysis of bunglers in mathematics(Kline, 1980,
p. 124). Once again, historians tend to ascribe thesedoubts to the fact that the
new discipline lacked a logical basis. As in all the previously discussedcases,
however, this impediment did not seem serious enough to prevent widespread
use of the effective analytic method. Indeed, Kline admits that by 1750 the
reluctanceto use algebra had been overcome, although by that time algebra
was a full-grown tree with many branchesbut no roots (p. 125).
Like the other caseswe dealt with, a scrutinizing look into history will reveal
that, along with the concern about internal consistency,doubtsof an ontological
naturepoppedup here and therein various writings. It was the unreifiable notion
of variable that was the core of the problem. This notion, the exactmeaning of
which cannot be easily explainedthrougha rigorous definition, may well be one
of the most problematic in the whole of mathematics-so problematic, in fact,
that doubts permeatethe professional literatureeven today. According to Bell
(1951):
to statefully whata variableis wouldtake a book. And the outcomemight be a
feeling of discouragement,for our attemptsto understandvariableswould lead us
into a morass of doubt concerning the meaningsof the fundamentalconceptsof
mathematics.(p. 101)

And then he immediately added his own imprecise description: Variable is

something which changes. This operational component of change, which resisted all attemptsat reification, is what madethe conceptof variable unacceptable even in the eyes of some 2Oth-centurymathematicians.This is certainly
what botheredFrege (1970, p. 107) who requiredelimination of time because
of its being alien to Analysis. The impossibility of doing it is what eventually
forced him to reject the whole idea by saying The word variable . . . has no
justification in pure Analysis. This is a quite radical opinion with which not
many mathematicianswould agree. Usually doubts about ontological origins
gradually dissolve with persistentuse of a notion. And, indeed, by the end of the
18thcentury mathematicianswere obviously familiar enoughwith algebraand its
techniquesto use them without further ado.

28

SFARD

But not all of them would be able to calm their conscienceby pushing the
ontological questionsaside. In the 1830sand 1840s a debateon the meaning of
algebraand its symbolism was led by prominent British mathematicianssuch as
Augustus de Morgan (1806-l 871) and Sir William Rowan Hamilton (1805%
1865). De Morgans associateGeorge Peacock (1791-1858) is generally regardedas the leading figure in the innovative school of thoughtwhich developed
as a result of this dispute. (Individual contributions of the different mathematicians are presentedin detail by Novy, 1973.)
Until then, algebrahad been regardedas universal arithmetic-a discipline
which specialized in expressingin a general way the rules governing numerical
operations. This interpretationof symbols and symbolic manipulations necessarily restrictedthe scope and force of algebraic laws. The British mathematicians felt an urge to provide algebrawith a sound logical basis thatwould set it
free from such limitations.
The tendencyto broadenthe scopeof conceptsby gradually loosening different restraintson their meaning is one of the typical processesthatcan be traced
throughoutthe history of mathematics.The conceptof negativenumberresulted
from the removal of the embargoon subtractinga number from a smaller one
(this, of course, was done not without reluctance and hesitation, but that is
another story). Similarly, complex number was born when extracting square
roots from negativequantitiesceasedto be seenas a taboo. It is probablythe 19th
centurys gradual reconciliation with the seemingly unacceptableidea of complex numberthatgaveBritish mathematiciansthecourageto claim theemancipation of algebra from the yoke of its original meaning. They had an almost
mystical feeling thatthe laws of algebramust be treatedas completelyuniversal
and that this principle of universality has to be acceptedas superior to any
considerationotherthan the consistencyof the theory.Thus, the fact thatcertain
numerical operations did not comply with some rules was not regardedas a
sufficient reason for restricting the rules. This stancefound its most emphatic
expression in Peacocks principle of permanence:Whatever form is algebraically equivalentto anotherform expressedin generalsymbols, mustcontinue
to be equivalent, whateverthe symbols denote(cited in Novy, 1973, p. 191).
The term form probably means here algebraic expression and algebraically
equivalent means equivalent through symbolic manipulations.
Peacocks immediate conclusion from this principle may be formulated as
follows: If a number does not obey a law, the number ratherthan the law would
be the one to go. A variable should no longer be seenas a generalizednumberbut
must be treatedas a thing in itself, devoid of any externalsense. Variables are
thus mere symbols that denotenothing. They, of course, may be interpretedin
different ways in differentcontextsbut they haveno meaningof their own. From
here Peacock promptly arrived at a completedearithmetizationof algebra. The
meaningof symbols should no longer be expectedto comefrom their nonexistent
designata,but rather,mustbe soughtin theway theformulasaretransformedand

THE DEVELOPMENT

OF ALGEBRA

29

combined with each other. These transformationsare subordinateto rules given

by axioms. The axioms themselvesare arbitrary. Thus, Peacock and his colleagues may be regardedas precursorsof the formalist school of thought that
developed fully only several decadeslater when David Hilbert (1862-1943)
generalizedthe idea of semanticallyvoid symbols beyond algebraand applied it
to the whole of mathematics.
The dearithmetizationof algebrais a typical exampleof severinga mathematical idea from its operationalorigins in order to attainfull reification. Things like
that happenedin mathematicsmany times before. The concept of number, for
example, originated in the operationsof counting and measuringbut theseinterpretationshad to be given up if the idea of complex numberwas to be accepted.
Several decadeslater, the mathematicianswho converted the function into a
simple set of orderedpairs sacrificed the algorithmic underpinningsof the concept. One may say, using Garcia and Piagets terminology, that with the dearithmetizationof algebra,the interoperationalstagewas finally attainedwherethe
primary operationswere reified, and the relationshipsbetweenthem, ratherthan
their internal structure,becamethe central object of attention.
The 19th-centurydiscussion betweenBritish mathematicianswas one of the
incentives for a recent study carried out by one of my colleagues and myself
(Linchevski & Sfard, 1991). Fascinated by the fact that mathematiciansthemselves could hardly make up their minds as to the best possible interpretationof
algebra, we decided to turn to todays studentsin an attemptto find out their
implicit beliefs about the meaning of symbolic formulas and manipulations. In
our study many questions, sometimesquite nonroutine and surprising, were
askedregardingsuch conceptsas equivalenceof equationsand inequalities, permissible operations, and solution of an equationor inequality (see examplesin
Figure 3). Before the experiment,we conjecturedthat the studentswould interpret propositional formulas in one of two ways: either as generalizedarithmetic

1. Are the following equations(inequalities)equivalentor not? Explain your

answers.
c.x*+ 1 =o
b.4x- 11 =2x-7
a. 4x* > 9
x*+5=0
(x - 2)2 = 0
2x>3
2. Completethe following sentences:
a. To solve an equationmeans.,............
b. When we solve an equationwe are allowed to executethe following
operations: ......................
c. These operationsare permittedbecause.................
d. When we finish solving the equation,what we get in the end is...................

figure 3. Sample questionsfrom the researchby Linchevski and Sfard (1991)

30

SFARD

expressions,which can only be understoodthroughthe underlying mathematical

operations(primary processes)and throughthe setsof numbersto be substituted
insteadof the letters,or as merestringsof symbols thatdraw their meaningfrom
the manipulations (secondaryprocesses)people use to perform on them. The
result of our investigationwas quite unequivocal:Whateverquestionor problem
was presentedto the students,the answersinvariably showeda clear bias toward
the lattertype of interpretation.The majority of pupils viewed algebraicexpressions as meaninglesssymbols governedby arbitraryestablishedtransformations.
This second approachto algebraicsymbolism deceptivelyreminds us of Peacocks ideas. Thus, on the face of it, our studentsrevealedsurprisingmaturityby
treating algebraic symbols in the way recommendedby mathematiciansthemselves. In fact, however, our pupils conceptions were nothing like the ideas
promotedby Peacock. What the British mathematicianproposedmay seemquite
simple but a closer look disclosesits truly sophisticatednature. Peacocksideas
were generatedby a well-appreciatednecessityto free algebrafrom theburdenof
the initially helpful but now restraining semantic load. They resulted from a
familiarity with the alternativeinterpretationsof algebraand with their disadvantages.To put it anotherway, Peacockssemanticallyemptiedconceptsoriginated
in the consciousdecision of a personwho knew exactlywhat he or she was going
to give up and who was perfectly able to go back to the renouncedmeanings
wheneverappropriate.No suchreturnis possible for thestudentwhoseeyeshave
never been opened to the alternativeoptions. Indeed, our studentsostensibly
structuralthinking did not seembatteredby any operationalunderpinnings and
therewas no reasonto believe thatwhat could be seenwas but an upperlayer of a
sophisticatedconceptualstructure.Their conceptionsappearedone-dimensional
and shallow. Such comprehensionis not very likely to lead to the flexibility in
interpreting variable that is the basis of successful problem solving. This is
probably why a tone of exasperationis usually assumedwhen results of studies
are reportedshowing that studentsalgebraic languageis empty, having only
syntax (Burton, 1988, p. 2).
As was argued by Linchevski and Sfard (1991), it is probably the modem
structuralway of teachingthatmay be partly responsiblefor this situation. In the
discussion of how to teach algebra, I would, therefore,vote for courseswhich
take the historical facts into considerationand compromisethe modem definitions for the sakeof a less advancedbut to the learnermore accessibleoperational approach. (One such course was recently developed in England; see the
National Mathematics Project, 1987, and a brief description of this course in
Kieran, 1992.)
ParaphrasingPicasso, who reportedlyclaimed that one must be able to draw
realistically before becoming an abstractpainter,we may say thatin mathematics
a pupil should be a Platonic realist before turning into a formalist and being able
to deal with pure abstractionbrought into being by stipulation.

THE DEVELOPMENT

OF ALGEBRA

31

STAGE 3: FROM GALOIS TO BOURBAKI-ALGEBRA

AS A SCIENCE OF ABSTRACT STRUCTURES
first Step: The Emergence of Group Theory
Peacock took the liberty of introducing the componentof arbitrarinessinto algebra. By doing this and by suggestingthatthe axiomatic methodexemptsmathematiciansfrom ontological confinements,he laid a cornerstonefor a new kind of
mathematics,the spirit of which is mocked in the rhyme of dubious didactic
value: Minus times minus equalsplus, thereasonsfor this we neednot discuss.
From now on mathematicianscould invent new mathematicalobjects fearing
nothing and nobody but the laws of logic. The internal consistencyof a new idea
should be their sole concern and no philosophical questionsabout the natureof
the formally defined object or about its relationship with the real world ought to
seem relevant any more. After this ontological breakthrough,introducing new
mathematical objects through axiomatic systems gradually became common
practice. (For a deeperinsight into the process of change which, in fact, was
more complex than may be understoodfrom this concise description, see, e.g.,
Kleiner, 1986.)Algebras bonds with numbersand numerical computationswere
loosenedeven furtherand it gradually turnedinto a scienceof abstractstructures.
Hamiltons invention of quatemionsin the 1850smay be regardedas the first
act of such free creation. His earlier work on complex numbers, which he
presentedsimply as pairs of real numbersruled by formally defined operations,
brought him close to the position held by todays mathematicians:He began to
realize that nothing more than a consistentaxiomatic systemis neededto legitimize the existenceof an abstractobject. (Hamilton was quite aheadof his times
as may be seenin the following statementby his friend John Graves: I havenot
yet any clear view as to the extentto which we are at liberty to createimaginaries, and to endow them with supernaturalproperties,Kleiner, 1987,p. 233.)
He immediately decided to take advantageof this new approachby pushing the
idea of number-like n-tuples a little further. The quatemionsare 4-tuples of real
numbers, subordinatedto a regular type of addition and to a noncommutative
operationof multiplication. Although Hamilton introducedthemto mathematics
on the sole basis of formal definition, he still felt an urge to justify his creationby
pointing out its possible physical applications. Those who came after him soon
freed themselveseven from this kind of consideration. Nevertheless,his step
toward the new kind of mathematicswas so decisive that in the eyes of some
historians all of modem algebra owes its origins to Hamiltons creation of
quatemions (Kline, 1980, p. 295).
Another milestone in the history of abstractalgebrawas the emergenceof the
concept of group. Its origins go back to pre-Hamiltonian times, to the works of
Joseph Louis Lagrange(1736-18 13) and Paolo Ruffini (176% 1822).Both these
mathematicianswere preoccupiedwith one of the central problemsof 18th-and

32

SFARD

19th-centuryalgebra-the question of the possibility of solving equations of

degree5 or more by radicals. Both of them noticed that importantinformation
about the equation may be collected through a study of certain functions of its
roots and throughcounting the number of different values such functions obtain
when the roots are permutedin all possible ways. The notion of permutation
gradually overshadowedthe otherauxiliary conceptsuntil it becamethecenterof
attention.In no time, what was basically a process,an operationof rearranginga
sequenceof entities, came to be treatedas an abstractobject. The first step
toward reification of the concept was made by Augustin Louis Cauchy (17891857)who explicitly talked aboutmanipulatingand combining the permutations
in certain well-defined ways, thus viewing them as inputs to higher level procedures. The operationson permutations,in their turn, were soon to becomethe
central object of inquiry. Evariste Galois (1811-1832) was the one who eventually defined the notion of group, namely,he explicitly declaredhis interestin the
structure imposed on a set of permutationsby the operations which can be
performedon them (the so-called substitutions).The name group, althoughnot
consistentlyused, was introducedby him to denotethis new mathematicalobject
of unprecedentedabstractnessand richness. The English mathematicianArthur
Cayley (182l-1895) took the construction of the concept a step further by
freeing it from any commitmentsto the natureof the basic elementsof a group.
They could be anything: permutations, quatemions, matrices, and so forth.
Thus, Cayley ultimately shiftedthe emphasisfrom the manipulatedentitiesto the
manipulations themselves.
It is noteworthythat althoughLagrange, Ruffini, and Cauchy were only one
small step distant from the idea of group they never actually arrived at it. What
preventedthem from going further was probably the fact that, as far as their
approachto mathematicswas concerned,they still belongedto the preformalist
school of thought.Indeed, the next move involved thekind of changein thebasic
philosophical assumptionsthatwas achievedonly slightly laterby suchwriters as
Hamilton and Peacock. (Although thework of Galois may be regardedas prior to
the emergenceof formalism, it was not recognizeduntil much later.) The decisive step could not be taken by people who felt that certain external factors,
which go beyond mathematicsitself, restricttheir freedomto createnew abstract
beings.
Further Development: The Proliferation of Abstract Beings
After the invention of the concept of group, nothing could stop algebra from
turning into a scienceof abstractstructures.What happenedin mathematicsafter
the first successful attemptsat free creation may be described as a true baby
boom. No longer fetteredby ontological considerationsmathematiciansfelt free
to conjure up new algebraic beings without ever asking about their relationship to the physical world or even about their prospectiveapplications in the
natural sciences. The richness of the new structuresand their links with other

THE DEVELOPMENT OF ALGEBRA

33

regions of mathematicsseemedmore important.Mathematicsstoppedbeing the

servantof natural science and from then on was developedfor its own sake.
Three salient traitsof this pure mathematicsin general, and of the new algebra
in particular, were its greatabstractness,its concern for logical foundations, and
its tendencyto split into loosely tied subdisciplines. On theone hand, theabstract
structuresprovided the loftier points of view from which many fields of mathematics, both ancient and modern, [could] be seen as wholes and not as rococo
patchworksof dislocatedspecial problems (Bell, 1951,p. 15). In the language
of Garcia and Piaget (I 989) the trans-operationalstagein algebrawas attained
where the computationalprocessesinvestigatedat the previous stagecould now
be viewed from a much higher vantagepoint. On the otherhand, the passionof
abstraction,sometimesquite furious (Bell, 1951,p. 158) led to the emergence
of ever new structuresand ever new branchesof algebra. The conceptof group
was accompaniedby the notion of invariant and by the theory of matrices, and
followed by the ideasof field, ring, and linear space.Eventually algebramushroomed into a welter of smaller developmentsthat have little relation to each
other or to the original concretefields (Kline, 1972, p. 1157).
No wonder, therefore,that at a certain stagethe necessityof reunification of
algebra and maybe even of the whole of mathematicscould be felt among
mathematicians.Algebra itself provided the means for this endeavor.In his
Erlanger Program (1872), Felix Klein united the different geometries, both
Euclidean and non-Euclidean, into one theory by characterizingeach of them
with the help of a certain group of transformationsand by saying thatfrom now
on geometryshould be treatedas the study of their invariants. Much later, in the
middle of the 20th century,the Bourbaki group setitself an even more ambitious
goal: The whole of mathematicswas to be reducedto threemother structures.
Algebraic structure,with its laws of composition, was proposedas one of them;
the other two were the order structureand the topological structure.Eventually
the theoryof categorieswas developedthatpurportedto unify all the branchesof
mathematics.
Here our journey throughthe history of mathematicsends. As in the previous
sections, I would like to comparethe past developmentsto the experiencesof
those who learn the subject. Not much systematicresearchhas been done that
can provide the relevant data. I can point to only one study that seemsclosely
relatedto the presentsubject. In this investigation, carried out in Israel by Hare1
(1985), a teachingunit on linear spaceswas developedand taughtto secondaryschool students.Classroom observationsled the researcherto theconclusion that
the objects populating vector spacesare not tangible, thus they are not considered by the studentsas objects at all; it is only natural, therefore,that the space
itself is not conceived as a mathematicalobject (p. 64). These findings are
hardly surprising. For the last 100 years mathematiciansthemselveshave not
seemedto have much difficulty with acceptingeven the strangestmathematical
object on the sole basis of the inner coherenceof the resulting systemof con-

34

SFARD

cepts. It was the dismissal of externalcriteria for legitimization of mathematical

ideas that brought logical considerationsto the fore. The intellectual maturity,
however, which gave mathematiciansthe strengthto resist the ontological questions, cannot be expectedfrom beginners. An axiomatic systemis certainly not
enough to convince the mathematica!lyunsophisticatedlearner about the existence of an object thathe or she has no way to see, touch, or just imagine. The
doubtsas to the natureof such objects, as well as to the legitimacy of the very act
of free creation, may impedestudentsunderstandingin exactlythe sameway in
which similar considerationsobstructedthe historical developmentof abstract
structuresuntil the 19th century. Indeed, there is no reasonto assumethat our
studentis more mathematicallymaturethan Lagrange or Cauchy.
When no underlying computationalprocessmay be offeredto makethe introduction of a new mathematicalobject more smooth and natural, the computer
may provide some help. The machinehas almostunlimited power of reification.
The figments of a mathematiciansimagination materializeon the screenso that
it becomesquite natural to treatthem as if they were independentbeings, external to the human mind. The conjecture about the possible influence of the
computeron learning abstractalgebrais now being testedin a studycarriedout in
Israel and in the United Statesby Leron and Dubinsky (1995). Its results will
certainly provide much new information on studentsability to learn advanced
mathematics.If the computerprovesitself a tool for reification, it may even lead
the researchersto the conclusion that the history of algebrawould have taken a
differentcoursehad a powerful numbercruncherbeen availableseveralcenturies
earlier.
CONCLUDING

REMARKS

The history of algebrawas presentedhere as a long sequenceof actsof creation

in which generationsof mathematicalobjects of increasing abstractnesswere
brought into existence.Studentswho learn algebrahaveto recreatetheseobjects
for themselves.Some empirical data have been brought forward to enlighten
severalaspectsof this process.It is not surprisingthatwhat was far from easyfor
mathematiciansinvariably proved to be quite difficult for the learner.
Many exampleshave been provided here to reinforce the thesis that didactic
problems are likely to appear at all those junctions at which mathematicians
themselvesfalteredand askedquestions.For thosewho teach,therefore,familiarity with the history of mathematicsis not just optional; rather,it seemsindispensableto makethem alertto the deeply hiddendifficulties concernedwith new
concepts.The ontological obstaclesare ubiquitous and at the sametime they are
elusive and difficult to detect.Pupils fundamentalproblems with such ideas as
complex number or variable may be overlooked by the teacherbecause the
lattersown implicit beliefs make him or her oblivious to the very possibility of
somebody having a different ontological stance. What helps in concealing on-

THE DEVELOPMENT

OF ALGEBRA

35

tological stumbling blocks is the fact thata studentmay becomequite skillful in

manipulating such mathematicalobjects as number, function, or algebraic expressions even without reifying them.
One importantlesson to be learnedfrom history may be somewhatat variance
with the pedagogicalbeliefs of the modem teacher.The storiesjust told seemto
imply thatthereification thatis neededfor a deepunderstandingof a concept(say
complex number) cannot be expectedbefore some familiarity with secondary
processes(e.g., operationson complex numbers)has been attained,On the other
hand, without the reification theseprocessescannot be truly meaningful. The
surprising pedagogicalconclusion follows from here: Sometimesthe teacherand
the studentsmust put up with the necessityof practicing techniqueseven before
they are fully understood.In light of this, it appearsthatin learning and teachinga
crucial role is played by patience and persistence.Indeed, history has already
shown thatthesemay be thebasic weaponsagainstontologicaldifficulties. Cardan
insisted on using complex numbersregardlessof the mental torturesinvolved
(Kleine, 1980,p. 116).History provedhe was right-mathematicians eventually
reconciled themselvesto the concept. Similarly, todaysstudentshould be disciplined enough to work with algebraictechniquesand manipulateabstractobjects
even if he or she has doubts as to their meaning. The teachershould tamehis or
her impatience when facing deficiencies in learners understanding.If persistently used, the conceptswill eventually becomeeasierto reify-and to accept.
REFERENCES
Assessmentof Performance Unit. (1980). Mathematical development, secondary survey (Rep. No.
1). London: HMSO.
Bell, Eric Temple (1951). Mathematics, queen and servant of science. Redmond, WA: Tempus
Books of Microsoft Press.
Booth, Lesley R. (1988).Childrens difficulties in beginningalgebra.In A.F. Coxford (Ed.), The ideas
ofalgebra, K-12 (pp.20-32). Reston, VA: National Council of Teachersof Mathematics.
Boyer, Carl B. (1985). A history of mathematics (Rev. ed.). Princeton, NJ: Princeton University
Press.
Burton, Martha B. (1988). A linguistic basis for studentditficulties with algebra.For the Learning of
Mathematics, 8 (l), 2-7.
Cardan, Geronimo (1968). The great art. Cambridge, MA: MIT Press.
Clement, John, Lochhead, Jack, & Soloway, Elliot (1979). Translation between symbol systems:
Zsolating a common di$iculty in solving algebra wordproblems (COINS Tech. Rep. No. 7919). Amherst: University of Massachusetts,Departmentof Computer and Information Sciences.
Dreyfus, Tommy (1991). On the statusof visual reasoningin mathematicsand mathematicseducation. In F. Furinghetti (Ed.), Proceedings of the Fifteenth International Conference for the
Psychology of Mathematics Education (Vol. 1, pp. 32-48). Assisi, Italy: IPC, PME 15.
Dubinsky, Ed (1991). Reflective abstractionin advancedmathematicalthinking. In D. Tall (Ed.),
Advanced mathemntical thinking (pp. 95-123). Dordrccht, Holland: Kluwer Academic Press.
Fauvel, John, & Gray, Jeremy. (1987). The history of mathematics: A reader. London: Macmillan.
Frege, Gottlob (1970). What is function? In P. Geach & M. Black (Eds.), Translations from the
philosophical writings of Gottlob Frege. Oxford: Blackwell.

SFARD

36

Garcia, Rolando, & Piaget, Jean (1989). Psychogenesis and the history of science. New York:
Columbia University Press.
Harel, Guershon (1985). Teaching linear algebra in high school. Unpublisheddoctoral dissertation,
Ben-Gurion University, Beer Sheeba,Israel.
Harel, Guershon, & Kaput, James (1991). The role of conceptual entities and their symbols in
building advancedmathematicalconcepts.In D. Tall (Ed.), Advanced mathematical thinking
(pp. 82-94). Dordrecht, Holland: Kluwer Academic Press.
Harper, Eon (1987). Ghosts of Diophantus. Educarional Studies in Mathematics, 28, 75-90.
Hefendehl-Hebeker,Lisa (1991). Negative numbers:Obstacles in their evolution from intuitive to
intellectualconstructs. For rhe Learning of Mathematics,
1 I( I), 26-32.
Kieran, Carolyn (1992). The learning and teaching of school algebra. In D.A. Grouws (Ed.), The
handbook of research on mathematics teaching and learning (pp. 390-419). New York:
Macmillan.
Kleiner, Israel (1986). The evolutionof group theory: A brief survey.Mathematics Magazine, 59(4),
195-215.

Kleiner, Israel (1987). A sketch of the evolution of (noncommutative)ring theory.LEnseignement

Math&natique,

33, 227-267.

Kleiner, Israel (1988). Thinking the unthinkable: The story of complex numbers. Mathematics
Teacher,

81, 583-591.

Kleiner, Israel (1989). Evolution of the function concept: A brief survey. College
Journal,

Muthemarics

20, 282-300.

Kline, Morris (1972). Mathematical thought from ancient to modern times. New York: Oxford
University Press.
Kline, Morris (1980). Mathematics, the loss of certainty. New York: Oxford University Press.
Lee, Lesley, & Wheeler, David. (1989). The arithmeticconnection.Educational Studies in Mafhematics, 20, 41-54.

Leron, Uri, & Dubinsky, Ed (1995). An abstractalgebra story American

Mathematical

Monthly,

102, 227-242.

Linchevski, Liora, & Sfard, Anna (1991). Rules without reasonsas processeswithout objects-The
case of equations and inequalities. In F. Furinghetti (Ed.), Proceedings of the Fifteenth
International Conference for the Psychology of Mathematics Education (pp. 317-324). Assisi, Italy: IPC, PME 15.
National MathematicsProject. (1987).National Mathematics Project. London: Longman.
Navy, Lubos (1973). Origins of modern algebra. Leyden, Holland: Noordhoff InternationalPublishing.
Sfard, Anna (1987). Two conceptionsof mathematicalnotions: Operational and structural. In J.C.
Bergeron, N. Herscovics, & C. Kieran (Eds.), Proceedings of Eleventh International Conference for the Psychology of Mathematics Education (Vol. 3, 162- 169). Montreal, Canada:
University of Montreal.
Sfard, Anna (1991). On the dual natureof mathematicalconceptions:Reflections on processesand
objects as different sides of the same coin. Educational Studies in Mathematics, 22, l-36.
Sfard, Anna (1992). Operational origins of mathematicalnotions and the quandaryof reificationthe case of function. In G. Hare1& E. Dubinsky (Eds.), The concept offunction: Aspects of
epistemology andpedagogy
(MAA Notes, Vol. 25, pp. 59-84). Washington,DC: Mathematical Association of America.
Soloway, Elliot, Lochhead, Jack, & Clement, John (1982). Does computerprogramming enhance
problem solving ability? Some positive evidenceon algebra word problems. In R.J. Seidel,
R.E. Anderson, &B. Hunter (Eds.), Computer literacy (pp. 171-185). New York: Academic.
Unguru, Sahetai(1975).On theneedto rewritethehistory of Greek mathematics.ArchiveforHistory
of Exact Sciences, 15, 67-l 14.

THE DEVELOPMENT OF ALGEBRA

37

APPENDIX
The Genesis of Mathematical Objects
While glancing every now and then at the long and turbulenthistory of number
systems,I briefly summarizehere one of the possible general scenariosof concept development.This theoreticalmodel bears on the ideas initiated by Piaget
and seemsto be compatiblewith thetheoreticalframeworksproposedrecentlyby
some of his followers (e.g., Dubinsky, 1991;Hare1& Kaput, 1991).It was first
presentedin much greaterdetail in Sfard (1991, 1992).
In all branchesof mathematicsand, in particular, in computationalsciences,
one can clearly distinguish two kinds of components:abstractobjects and computationalprocesses.Numbers are good examplesof the former.The way people
think about themand mentally operateupon themresemblesthemannerin which
they perceive and manipulate physical objects. The abstractobjects, in turn,
serve as inputs and outputs to certain computationalprocedures-this second
ingredient of the mathematicaluniverse.
A closer look at thesetwo separateand ostensibly dissimilar componentswill
reveal an interesting relationship betweenthem. As was explained in detail in
Sfard (1991), abstractobjects and computationalprocesses,as different as they
may seem,arebut oppositesidesof the samecoin-two facetsof thesamething.
In a sense, the abstractobjects arejust an alternativeway of referring to computational processes:Natural and rational numbersare metaphorsfor counting and
measuring, respectively, and the concepts of negative and complex numbers
refer, in fact, to nothing other than the operationof subtractinga number from a
smaller one and to the processof extractinga squareroot of a negativenumber.
Using the terminology introduced in Sfard (1991) I would thus say that any
number (like any other mathematicalconcept, in fact) may be conceived in two
ways: operationally,as a process, and structurally,as an object. (I emphasizethe
word conceived to make it clear that I am talking about the way a person
perceives, thinks, and talks about abstractideas and not about the natureof the
mathematicalentities themselves,whateverthe words the natureof an entity
may mean when consideredindependentlyof the epistemic subject.)
Thus, one may say that rational, irrational, negative, and complex numbers
arejust the more matureincarnations of certain computationalprocesses.When
their historical developmentis scrutinized, it invariably turns out thatno sooner
did the new numbersenter the scenethan a certain nontraditionalcomputational
processbegan to gain recognition. The idea of irrational number stemmedfrom
measuringproceduresthatcould not be encodedas pairs (ratios)of integers.The
notions of negative and complex numbers may be traced back to the work of
Cardan (Am Magnu, 1545) in which the algorithmsfor solving cubic equations
were shown to lead occasionally to such nonroutineproceduresas subtractionsof
a number from a smaller one and extractionof the squareroot from the products

38

SFARD

of such subtractions.The interestingthing is that althoughthe two conceptsdid

not really catch on until nearly threecenturieslater, the persistentreluctanceto
accept them as legitimate objects did not prevent mathematiciansfrom using
them.
The developmentof the number concepthasjust been presentedas a chain of
transitions from operationalto structuralconceptions. As seen more than once,
however,even before the processeswhich engenderedthe new kinds of numbers
were reified, namely turned into full-blown objects, mathematicianswere able
to perform them and even to combine themwith otheroperationsto obtain more
complex computations.I shall say,therefore,thatthe processeswere interiorized
and even condensed: They could be easily performed (so were interiorized)
and they could be referred to as proceduresexecutedinside a black boxsomethingthat no longer had to be describedin full detail when consideredas a
part of a composite process (so they were condensed).This three-component
pattern,interiorization/condensation/reification,seemsto repeatitself at almost
every turning point in the history of mathematicalideas-and in the processof
learning.
From these theoretical reflections mathematicsemergesas a hierarchy of
abstractrealms built in a sequenceof almost identical steps:Time and again,
processesperformedon certain abstractobjectsturn into new objects in order to
serve as inputs to higher level processes.With respectto a given concept, say,
that of negative number, one can distinguish betweenprimary and secondary
processes,those which underlie the concept and those which are applied to it,
respectively.In the caseof negativenumbersthe subtractiona - b, from which
the restriction a 2 b has been removed, is a primary process, whereas the
arithmeticoperationsextendedto all its results are the secondaryprocesses.
It should be mentionedhere that the aforementionedschemeof conceptconstruction is similar in somerespectsto the model proposedby Garcia and Piaget
in their book History of Science and Psychogenesis. Their ideas, like those
presentedhere, arebasedon the assumptionof the cyclic natureof theprocessof
knowledge formation common to historical developmentsand to individual
learning. Like the idea of hierarchical constructionwhere the samenotions are
being conceivedand useddifferently at differentlevels, Garcia and Piagetscycle
of intraoperational, interoperational, and trans-operationalstagesreflects the
changeof perspectivewhich takesplace in the courseof conceptevolution. The
intra- and interoperationalstagesroughly correspondto the phasesof primary
and secondaryprocesses.The transoperationalstageis attainedonly when the
vantagepoint is pushedeven higher and insteadof concentratingon individual
numbersone shifts attentionto the overall structureimposedupon the given set
of objectsby the secondaryoperations.Soon the natureof the elementsin the set
loses its importanceand the structure-imposingoperationsremain theonly object
of interest. Using our language and ideas we may say that such structuresas
groupsor fields, the emergenceof which indicatesthatthe transoperationalstage

THE DEVELOPMENT

OF ALGEBRA

39

with respectto the conceptof number has been attained,are, in a sense,nothing

more than a combination of the most general computationalprocesstreatedas
autonomouswholes, thus already condensed,maybe even reified.
One final remark concerning the difficulty of reification: As an ontological
shift it is an inherently complicatedprocess. At least two seriousreasonsfor its
being very difficult to attain may be mentioned. One of them was called the
vicious circle of reification. The use of this name refers to the fact that the
reification of primary processes(thosewhich underlie the given concept)seems
to be the precondition for the ability to deal with secondaryprocesses(those
which are applied to the given concept),whereasthe latterseem, in turn, to be a
precondition for the former. For example, 3 - 5 must be treatedas a legitimate
mathematicalobject before it can be manipulatedand combined-through secondary processes-with other numbers. On the otherhand, to speakabout such
operationsas 3 - 5 and 1 - 3.5 as numbers,one must be able in advanceto use
them as inputs to the secondaryprocesses.After all, it is the only way in which
one may realize that 3 - 5 and 1 - 3.5 obey the samerules as 2, 5, 12 and the
like, thus behaving like genuine numbers; such a realization is indispensableto
justify and to motivatereification.
The second type of obstaclearises when some semanticconcessionsmust be
madebefore the new abstractobject is fully accepted.For example,to talk about
a squareroot of a negative quantity as a number, people must free themselves
from their deeply rooted conviction that number is somethingwhich expresses
quantity-a result of a measuring procedure. It is thus the very process that
engenderedthe concept of number that must now be given up. In this article, I
observedthis phenomenonof alienation from the primary operationalroots time
and again while surveying the developmentof algebra.

All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and r