You are on page 1of 33

Frege: The Royal Road from Geometry

Author(s): Mark Wilson

Source: Nos, Vol. 26, No. 2 (Jun., 1992), pp. 149-180
Published by: Wiley
Stable URL: .
Accessed: 22/02/2014 22:35

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .

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

Wiley is collaborating with JSTOR to digitize, preserve and extend access to Nos.

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
Frege: TheRoyal Roadfrom Geometry'
Ohio State University

I would like to submit this mode of thinking to philosophers, who otherwise often
limit themselves to mathematicaltrivialities, for a considerationof its principles-
Felix Klein. (Klein 1925)

Ludwig Wittgensteinposed the question,"Whatgroundscan be given in favor of
our presumptionthat the propercontinuationof the "even number"series 2, 4, 6,
8 is 10, ratherthan, e.g., 62?" The tenor of Wittgenstein's discussion suggests
that, however he expected the worries raised by his examples to be dispersed,
their proper"philosophical"solution was not expected to effect the practice of
the working mathematician,anymore than the definitive resolutionof skeptical
doubts about the external world will force new methodology upon subatomic
physics. Within more sophisticatedcontexts, analogousworriesabout the proper
continuationof a mathematicalpractice can assume quite practicaldimensions.
Such cases occur when questions of "the proper setting" for a mathematical
question arise. For example, Leonhard Euler and other early mathematicians
claimed that the "sum" of the series 1 - 1 + 1 - 1 +. . . . ought to be 1/2, an
"howler"for which they are still roundly mocked by writers of the E.T. Bell
school. It is true that the familiarcalculus level approachto infinite summation
pioneeredby A-L. Cauchy does not grantthe Euler series a sum, yet it turnsout
(Hardy 1949) that a variety of generalizednotions of summationexist that fully
ratify the answer "1/2". Viewed retrospectively,these generalized treatments
appear to be "better settings" for the mathematicalcontexts in which Euler
originally employed his series manipulations.It is true that Euler never thought
of these generalized conceptions, but, by the same lights, he never thought of
Cauchy's account either. The Cauchy approachis nowadays "central"in that it
serves as the springboardfor the introductionof the generalizednotions, but this
consideration alone should not force us to evaluate Euler retrospectively as

NOUS 26:2 (1992) 149-180


This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
150 NOUS

making mistakes in the summation of series (by the lights of the generalized
notions, Eulerrarelymade errors).
The fact that "better settings" for old mathematical problems can be
uncoveredafter a long period of mathematicaldigestion raises unsettlingdoubts
about mathematicalcertainty.The mathematicianPhillip Davis, in an interesting
article that asks whether any proposition can be definitively established as
ments are simply definitions, axioms or theorems .... [But] in mathematicsthere is
a long andvitallyimportant
recordof impossibilities
beingbrokenby the introduc-
tionof structural
changes.Meaningin mathematicsderivesnotfromnakedsymbols
but from the relationshipbetween these symbols and the exterior world.... Insofar
as structures
areaddedto primitiveideasto makethemprecise,flexibilityis lost in
the process.In a numberof ways,then,the closerone comesto an assertionof an
absolute"no",the less is the meaningthatcan be assignedto this "no".(Davis
1987,pp. 176-7)

That is, althougha particularapproachto summationcan be renderedprecise

within, say, Principia Mathematica,the final arbitrationof the proper truthvalue
for "1 - 1 + 1 +.... has a sum"will not rest simply with the "nakedsymbols"of
the selected formalization, but will instead turn upon the question of which
formal setting best suits the "meaning"of the original mathematicalideas. The
discovery of the "best account"of original meaning may not emerge until long
after the fact. Davis' worry about the permanenceof mathematicalcertainty(or
impossibility) derives from the fact that, although the truth values of many
statementstransferunscathedinto the revised setting, the truth value of many
others do not. It seems impossible to predict the eventual fate of any claim in
advance, no matter how unshakableit may now seem. Wittgenstein's worries
about the future continuationof "2, 4, 6, . . ." may look as if they may hinge
upon questions whetherhumanbeings can unambiguouslyinternalizealgorithms
possessing infinitely many applications,but Davis' worries depend solely upon
semantical concerns: when can a formalization be established as "optimally
respectingthe innernature"of a certainclass of mathematicalnotions?
"Bettersettings"for old problemscame into their own duringthe nineteenth
century (althoughthe earlier introductionof the negative and complex numbers
can be viewed in this light also). It was discovered that a wide variety of
traditionalmathematicalendeavors could be greatly illuminated if their basic
questions were reset within richer domains of objects than had previously
seemed appropriate.For example, mathematicianshad long struggled with the
ratherintractableformulas for arc length along an ellipse or leminscate. Niels
Abel recognized that it was advantageousto study such functions (or, more
properly,theirinverses) over the complex numbers,even thoughthe "arclength"
of a path stretchingbetween 0 and 2i primafacie sounds like a ridiculousnotion.

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions

BernhardRiemannthen saw that these so-called "elliptic functions"behave in a

quite simple manner if they are viewed as fundamentally defined upon a
doughnut of complex values rather than over the usual complex plane. Their
behaviorhad originally seemed complicatedbecause the core functionshad been
"projected"out of their naturaldomain into the "artificial"setting of real values
only. It is considerationsof this sort that led HermannWeyl to proclaim that
"Riemannsurfaces"(such as the doughnut)representedthe "nativeland, the only
soil in which the functionsgrow and thrive"(Weyl 1955).
The claim here is that the relevant mathematics achieves a simple and
harmoniousclosure only when viewed in the proper-often extended-setting.
One should comparesuch recastings-and the analogy was frequentlydrawnby
many nineteenth century mathematicians-to the completion of physical
explanationsby appealto unseen molecularstructures.When a branchof physics
or mathematicsis "reset"throughan expansion of its ontological territory,the
older explicationsof technicalvocabularyfrequentlycome underreview and are
modified to suit theirreconstitutedsurroundings.We noted that the new explica-
tions often reverse the truth-valuesof apparentlywell establishedpropositions.
Such reversalsdo not troublethe physicist much, but they are worrisometo our
sense of the permanence of mathematics. Davis' concern derives from this
source. When he assertsthat "meaningdoes not derive from naked symbols",he
is maintaining,quite correctly, that the "proper"definition of a mathematical
term should not rest upon the brutefact that earliermathematicianshad decided
that it should be explicated in such-and-suchmanner, but upon whether the
definition suits the realm in which the relevant objects optimally "grow and
thrive". In this respect, Davis' views about mathematicalmeaning resemble
some of the doctrines about the proper meaning of physical terms such as
"water"or "gold" expounded lately by Saul Kripke and Hilary Putnam. The
doctrine that an unsuspected setting secretly determinesthe "truemeaning"of
certain mathematicalvocabularymight well be labeled hidden essentialism, in
tributeto its similaritiesto essentialistclaims about"naturalkind terms".
Insofaras I can tell, most workingmathematicianssubscribe,in theirpractice,
to some form of "hidden essentialism" (although they might be reluctant to
accept this thesis explicitly). Mathematiciansare frequentlyquite scathingabout
colleagues who "don't work in the propersetting".The discovery of an optimal
mathematicalplacementcan representthe highest form of achievement-I have
recently heard it said that, whereas a complete proof of Fermat'stheoremmight
prove mathematically rather insignificant, the older partial results on this
problem (such as E.E. Kummer's)are "deep"in that they reveal hidden aspects
of the integers that become manifest only when they are imbeddedwithin richer
realmsof "idealnumbers".
Despite this de facto insistence upon the "proper"setting of mathematical
notions, the semi-"official"philosophy prevailing among most mathematicians
seems insufficient to ground the distinction between "better" and "worse"

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
152 NOUS

settings. The "official" position, dominant since the start of this century,
maintains that any self-consistent domain is equally worthy of mathematical
investigation; preference for a given domain is justified only by aesthetic
considerations,personal whim or its potential physical applications.Upon such
grounds, how can the mathematicianwho uncovers the "propersetting" for a
traditional problem be distinguished from the mathematician who simply
changes the topic? How can one give teeth to the conceit that certain mathe-
maticalconcepts (e.g., "ellipticfunction")cry out for particular settingsin which
"they grow and thrive"? With only the tools of the "official" philosophy,
preferencefor "bettersettings"simply devolves to becoming a matterof "I like it
better"or "Isn'tthis pretty?".
What one would ideally like, as an excellent pair of papers by Kenneth
Manders(Manders1987 and 1989) has made clear, is a deeper understandingof
mathematicalconcepts that relates the "meanings"of mathematicalconcepts to
the hiddenfactorsthatdrive disciplines into reconstitutedarrangements,regroup-
ings that in some richer way "better respect" the meanings of the concepts
involved. Although a richer account of mathematics'hidden essentialism seems
desirable, the bare mention of the term "meaning",with all of its attendant
vagaries, readily explains why mathematicians,when pressed about their essen-
tialist claims, frequentlybeat a hasty retreatto the limited verities of theirofficial
"any self consistent formalism is equally worthy of study" philosophy. Yet,
lacking a richeraccount of continuityof meaning throughchange of setting, the
very idea of "mathematicalprogress" falls prey to Davis' worry-is every
mathematicalclaim potentially subject to some reconstitutionof domain that
might reverseour currentestimationof its truth-value?
Let us now turn to what was probablythe most notorious case of "domain
extension"arisingwithin the nineteenthcentury.

After 1820, geometry under the lead of Jean-VictorPoncelet grew into a much
stranger subject than could be expected simply from an acquaintance with
schoolbook Euclid. By most standardsof comparison,the world of the so-called
"projective geometer" is considerably more bizarre2 than non-Euclidean
geometryper se. Why? The projectiveschool decided that space should contain
many more points and lines than usual. These extra "extensionelements" stem
from two basic sources:To any plane, one adds extra points located along a line
at infinity and also a full complementof "imaginarypoints" whose coordinates
are allowed to be complex numbers.3Most of these imaginarypoints don't even
have the grace to hide away at infinity; some of them hover at ratherpeculiar
The additionof these extra elements would be unproblematicif they merely
served as technical devices to streamline proofs pertaining to the original

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions

Euclidean realm; many philosophershave characterizedthe extension elements

in such a fashion. Paul Benacerraf4writes that:

The points "at infinity"are introducedas a convenience to make simpler and more
elegant the theoryof the things you really care about. (Benacerraf1989b, p. 406)

But the characterization"thingsone really cares about"is wrong; the extended

space displaces Euclideanspace as the primaryfocus of interestin complexified
projectivegeometry. For example, one of its great theoremstells us that on any
cubic surface exactly 27 lines can be inscribed, but the theorem does not
determinewhich of these lines are "real".
Rather than thinking of the extra points as "conveniences",the projective
geometers saw the additionsas revealing the "trueworld"in which geometrical
figures live. Familiarfigures such as circles and spheres have parts that extend
into the unseen portionsof six dimensionalcomplex space, so that when we see
an Euclidean circle, we perceive only a portion of the full figure (nineteenth
centurygeometersliked to claim thatwe see the full shape of geometricalfigures
in the mannerof the shadows in Plato's cave). This way of proceeding is still
prettymuch the normin algebraicgeometrytoday.
This recasting of geometrical ontology representsone of the most striking
manifestationsof "hidden essentialism"to be found in the nineteenth century.
The projective school regarded these changes as "growing organically"-a
phrase drawn from the geometer Jacob Steiner-out of the original world of
Euclideangeometry.The questionposed in the previous section was: how might
we make sense of the claim that such revisions grow "organically"out of prior
practice-how can it be meaningfullyclaimed that geometricalconcepts cry out
for the extended, projective setting? For a variety of reasons we have tended to
avoid seriousengagementwith these sorts of questionin 20th centuryphilosophy
of mathematics.Witness a modem geometry text (Rosenblatt 1963, p. 16)5 on
the topic of the extension elements:

Questions probablyarise immediatelyin your mind: "Whatright have we arbitrarily

to add points to space? And, even if we may be permittedto do so, must we not
differentiatebetween the 'ordinary'points, which we see as dots on the paper, and
the 'ideal points' which are constructsof our imagination?".

So far so good. Here is the answerprovided:

Planes, straightlines, and points are the entities about which certain postulates are
made but which are otherwise undefined. Hence, we can give the name "point"to
anything which conforms to our postulates about points and which suits our
convenience to designate as "point". In short, all our geometry (indeed, all our
mathematics)deals with constructsof our imagination;and an "ideal"point is on
precisely the same footing, from this point of view, as an "ordinary"point.

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
154 NOUS

But it is hard to see from this response why the projective additions are
especially "natural"to geometry; why shouldn't we instead add inaccessible
cardinalsor pineapples to Euclid? Such a weak apology for the extension ele-
ments leaves one wide open to Phillip Davis' kind of worry.One might dress up
Rosenbaum's answer somewhat, but most modem defenses of the extension
elements come down to being simply a matter of "the projective changes are
particularlypretty"or the like.
I don't pretend to have a better answer to such worries myself, but many
nineteenth century mathematiciansthought that they did. A variety of rather
sophisticatedattemptsto rationalizethe changes in geometry were set forth. For
various historical reasons, most of us today are unaware of these discussions,
their importancehaving been pushed out of view by the later debates over non-
Euclideangeometry. But the controversyover the projectiverevisions is philo-
sophicallyjust as interestingas the non-Euclideanchallenge, althoughcompletely
differentin character.The question was not whetherestablishedEuclideantheo-
rems of geometrycould be reversedby an empiricallyconfirmablenon-Euclidean
competitor, but whether internal factors might drive Euclidean geometry to
reconstitute itself within a richer domain of objects in which various older
theorems fail. The projective threat to Euclidean certainty came from within,
ratherthanfrom without.
The central purpose of this paper is to show that Gottlob Frege's celebrated
logicism grew up in the shadow of these debates and that the more obscureparts
of his philosophy-his context principle,for example-can be betterunderstood
when viewed in this geometricalperspective.Frege's context principle,by way
of reminder,is:

[Never] ask for the meaning of a word in isolation, but only in the context of a
proposition... It is enough if the propositiontaken as a whole has a sense; it is this
thatconfers on its partsalso theircontent. (Frege 1884, p. 72)

This is one of those slogans that have deeply influenced a lot of famous later
philosophers,althoughfew of them seem to agree upon what it means!
In any case, the linkage between Frege and geometry is scarcely made
obvious within Frege's betterknown writings.However, Frege did much work in
complexified projective geometry and was undoubtedlyfamiliar with the doc-
trines outlined here. I'll attemptto tell a story here that links the philosophical
traditionsin geometryto Frege's publishedopinions.

Let us first ask what factors led geometers to place Euclidean geometry in its
strange projective setting? We will concentrate upon the complex extension
points (which are the most interesting).The original impetus for adding these
came from the successes of analytic geometry in the fashion of Descartes. Here

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions

"analyticgeometry"means the treatmentof geometricalrelationsthroughthe use

of algebraicmanipulationsupon equationsbased upon numericalcoordinates.At
first blush, the method looks unpleasantlyroundabout.Suppose line L cuts a
coplanarcircle C. The Cartesianmethod asks that our circle be assigned some
equation, say, "x2 + (y-1)2 = 26". This requires C to be calibratedwith two
arbitrarilyselected coordinatelines X and Y. Likewise, line L must also be tied
to X and Y by an equation such as "y=O".The geometrical relationship of
"intersection"is replaced,in the Cartesianscheme, by the algebraic operationof
"substitution".Thus if "y=O"is substitutedinto the equationfor the circle and the
result is solved, we find the intersectionpoints <-5,0> and <5,0>. This algebraic
treatment of "intersection" seems cumbersome for two reasons: (i) The
procedurerelies upon additionalfigures (the lines X and Y) that seem completely
extraneous,from a geometricalpoint of view, to the originalfigures C and L. (ii)
The Cartesiantreatmentof any iteratedseries of geometricalrelationshipsis apt
to culminatein rathermessy algebra.
So much for disadvantagesof analytic geometry; what are its advantages?
The answerreason is that a single Cartesiancalculationcan often cover a multi-
tude of cases whose traditional(often called "synthetic")treatmentwould require
separate proofs. Let us survey quickly two cases that display this Cartesian
economy and, at the same time, introducesome concepts we shall need later.
CASE 1: Let us begin with an intersectingcircle C and line L, as above. We
have already found the two points <-5,0> and <5,0> that representL's inter-
sections with C. Now set up the map 0 from x -* x' along the line defined by the
relation xx'=(+5)2. In traditionalgeometrical terms, 0 representsa "harmonic
division"of the points <x,O>and <xI, O>by the intersections<-5,0> and <5,0>.
A map like 0 is called an involutionbecause it has the propertythat it returnsto
itself after one cycle. Note how 4 "nests" around the two intersectionpoints,
which representO's sole fixed points, i.e., places where a point maps to itself
under0. By a furtheroperation(projectingfrom the top of the circle), this involu-
tion along the line transfersnaturallyto a second involution P intertwiningthe
points aroundthe circle (the algebrathatconstructs P from 4 is rathermessy and
won't be detailed here). The circle map ' turnsout to have a very simple con-
struction:x will map to x' just in case x' lies at the otherend of x's cord in some
bundleof lines emergingfrom a common focus c (in the jargon,c is the "pole"of
the line y=O).c has many prettyproperties,among them this: Find, by the usual
calculus rules, the equations for the circle's tangents at the fixed points. The
intersectionpoint of these two tangentswill, in fact, prove to be our "pole"c.
CASE2: In the foregoing procedure,we tacitly assumed that the substitution
of the line equationinto the circle equationgives real roots, such as the ?5 of our
example. But suppose our line does not intersect the circle, as happens if we
select the line y = -5. Here the Cartesiansubstitutionproceduregives the roots
? W1O, suggesting the ridiculous "points of intersection" <-i410,-5> and
<ial1O,-5>.But let us retain the courage of our calculations and continue our

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
156 NOUS

Figure 1

1 ~~~~~~~~~~~~~~~1

F 2e Pole

Figure 2

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions

construction of involution maps. The line map 0 defined by xx'=-10, still

representsan involution along the line, but now an overlapping map with no
apparentfixed points. Along the circle, the inducedmap v is also an involution,
but now constructedfrom a pole inside the circle. The pole and line retain the
same prettypropertiesas in case 1, including the fact that the "pole"is still the
intersectionof tangentlines to the circle (albeit "imaginary"tangents).Somehow
blind algebraicperseverancehas led us to sound geometric constructionssimilar
to those in the first case, despite the fact that our intermediatecalculationsnow
route throughthe absurd"intersectionpoints",<?l1O,-5>. This is one of those
cases where it seems, to paraphraseEuler, that our pencil has surpassed our
geometricalunderstandingin intelligence.

Figure 3



This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
158 NOUS

Our Cartesiancalculationscover, in a unified setup, two situationsthat, from

a traditionalEuclidean point of view, look completely different (and would
requirecompletely different proofs to establish). And this advantagemaintains
itself quite generally-the algebraic techniques unify Euclidean situations that
were traditionallyregardedas distinct.
Now there were some mathematicians6who believed that the rules of
complex algebra possessed some sort of universal validity as "rules of
thought"-these calculationsalways lead to sound results, in whateverfield they
are applied-even if the formulas generated by the rules seem completely
uninterpretable.On this view, the complex points look as if they were, in
HermannHankel's phrase, "gifts from algebra".I might remarkthat one of the
best philosophers of my acquaintancewas driven to philosophy after having
received this kind of "justification"when he encounteredcomplex numbers in
his engineeringcourses!
In any event, this answerwas commonly rejectedas restingupon a confusion.
CharlotteAngas Scott7wrote:

In teaching the elements of analyticalgeometry we are practicallyforced to allow,

even to encourage,a slipshod identificationof the field of geometrywith the field of
algebra.(Scott 1900a, p. 163)


It is far too much the custom now to rely on the analogy of algebra to justify the
introductionof imaginaries into geometry. Analogy, however, is no justification
unless we first prove the exact correspondenceof the fields of investigation. In
analytical geometry the identificationof the two fields is permissible, and is easily
explained; but in pure geometry any reference to algebra, expressed or implied, is
irrelevantand misleading.(Scott 1900b, p. 307)

That is, formulas like "x2 + (y-1)2 = 26" carry two distinct interpretations:(i)
they are "about"circles and (ii) they are "about"the relationship of various
number pairs <x,y>. The steps in our Cartesiancalculation are quite unexcep-
tionable interpretedover the field of complex numbers,so if geometricalinter-
pretationis eschewed, there is no danger of unsoundness.It is only under the
geometricalinterpretationthat the reasoningappearsunsound.Scott's complaint
is thatthese two interpretationswere often confused8.

Duringthe nineteenthcentury,severalnoteworthyattemptswere made to explain
the successes of the Cartesiancalculationsin exclusively geometricalterms.The
first was due to the great originatorof projectivegeometry, Poncelet. He agreed
that the Cartesian achievements revealed a deep unity in our two cases but
denied that this unity traced to any "universalityof algebra".Rather he gave

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions

4- 5

Figure 5

Figure 6

Figure 7

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
160 NOUS

reasons for supposing that hidden elements existed in Euclidean geometry-

namely, the extension elements. Poncelet picturedthe relationshipof cases (1)
and (2) in cinemagraphicterms as follows: Pull the line of figure 7 upward,
throughthe degeneratemiddle case, to its final position in figure 5. As the line
moves upward,its "pole" is pulled inside the circle and becomes crossed. The
induced involutions on both the line and circle gradually change in character
from that of being nested, as in the first diagram,to being overlapping,as in the
last diagram.The involution's fixed points move together along the line until
they fuse and vanish.
It certainly looks as if some permanentinterconnectionbetween circle and
line, ratherlike a mechanical linkage, persists through these modifications. In
sympathywith this feeling, ratherthanregardingthe fixed points as absentin the
last diagram,why not regardthem as simply invisible9, having been pulled up
into a higher plane by the transformationsleading to figure 5? The intersection
points should be regardedas always governing the interactionsbetween circle
and line, but they can no longer be "seen" when they become imaginary
(althoughthey can always be broughtinto view by dragging the figure back to
diagram7). Poncelet arguedthat in cases when a form persisted,but the accom-
panying objects vanished, one ought to postulate new elements according to a
generalprincipleof so-called continuityor persistenceof form. Thus Poncelet:

Is it not evidentthatif, keepingthe samegiventhings,one can varythe primitive

figureby insensibledegreesby imposingon certainpartsof the figurea continuous
butotherwisearbitrary movement;is it not evidentthatthe propertiesandrelations
foundfor the firstsystemremainapplicableto successivestatesof the system....
[even]whenthoseobjectscease to have a positive,absoluteexistence,a physical
existence?. . . Now this principle,regardedas an axiomby the wisest mathe-
maticians, one can call the principle or law of continuity. (Poncelet 1987,
p. 543 (orderrearranged))

Obviously it will be difficult to articulateany such a principlein a precise way.

But once the requiredobjects are postulated by the "law of persistence of
form", the original success of the algebraic Cartesianmethods can be seen as
stemmingfrom a kind of accident, viz., it so happensthat a partialisomorphism
exists between the proper, expanded set of geometrical objects and algebraic
formula involving complex numbers. But the isomorphism is not completely
exact, for the algebraicformula permit many distinctionsthat have no inherent
geometric significance whatsoever. Poncelet felt that if one directly reasoned
geometrically about the extension elements, without taking an unnecessary
detourthroughthe realm of the complex numbers,one might improve upon the
successes of the Cartesiantechniques while avoiding the clumsy crutch of the
algebraic methods10.Labeling the "hidden elements" of genuine geometry as
"gifts from algebra" is an injustice akin to crediting Christian, rather than
Cyrano, with the authorshipof the letters to Roxana. Poncelet's idea was that
geometry should be pursuedsynthetically,with the imaginarypoints postulated

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions

following the law of continuityin the same spirit as the existence of parallelsis
posited in Euclid. The accidental tie to numericalrealms exploited in analytic
geometryshouldbe avoided as an unnecessarydiversion.
Note that the "principleof continuity" supplies an answer to the dilemma
discussed at the startof this paper.Whateverits shortcomingsin termsof clarity,
the "continuity"principle representsthe new projective elements as the natural
outgrowth of the domain of traditionalgeometry. This "organic"connection
through"continuity"gives one the right, thoughtPoncelet, to claim that the new
extension elements representthe correct extension of Euclidean geometry. The
new elements are added to complete the mechanism that makes Euclidean
geometry work. Accordingly, the "ideal" elements are "just as real" as the
original elements of geometry, even though we can form no visual picture of
what the imaginaryelements look like (of course, nineteenthcenturywriterslike
Frege would have expressed this claim as: "we can form no intuition of the

So much for Poncelet. Between 1847 and 1860, the Germangeometer G.K.C.
von Staudtbroughta new perspectiveto the dilemma of the extension elements
(von Staudt 1847 and 1856). Ratherthan postulatingnew entities in Poncelet's
manner,von Staudtelected to define the extension elements in termsof accepted
Euclideanideas, so thatcomplexified projectivegeometrywould standsimply as
an extension by definitions of the old field. This idea alone would not be novel;
von Staudt's originality lay in the particularstrategy he followed in achieving
this end. Since there are no regular Euclidean objects that can adequately
substitute for the complex points, von Staudt was forced to rely upon some
unexpectedmaterials.In fact, he selected abstractconcepts to supplythe missing
material.His reasoningwent roughlyas follows:
"We are familiar with cases where assertionsabout concepts or relationsare
accordedwith an object-likestatus,as when we say "Personhoodis possessed by
Socrates", ratherthan "Socrates is a person". Consider "personhood"to be a
concept-objectderived from the concept "is a person".From this point of view,
consider the innocent-looking claim: "Line L has the overlapping involution
(defined by xx'=-10) on it". Once again, we have convertedwhat is essentially a
relation ("x maps to x' by the rule xx' =-10") into a "concept-object""the
involutionon L defined by xx'=-10". Recall that Poncelet postulatednew points
to correspondto the missing "fixed points"of such a relationon the groundsthat
the involution relation "persists"along the line althoughits normalfixed points
vanish. But ratherthan following Poncelet and viewing this process as one of
postulation, why not simply define the desired imaginarypoints as the concept-
objects denotedby phraseslike "theinvolutionon L"?"
This is von Staudt's basic idea but, strictly speaking, it doesn't quite work.
We actually need to constructtwo imaginarypoints from every involution along
a line (each involution has two fixed points, albeit occasionally "double"ones).

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
162 NOUS

Von Staudtevades this difficulty by correlatingeach involutionwith two distinct

"senses"-that is, the two "arrow"directions in which the involution can be
taken to act as a function. Such "involutions with a direction" can still be
regarded as "concept-objects", but concept-objects constructed by more
complicatedmeans than simple "personhood"or "involution".
I have adopted the funny hyphenated term "concept-object"to represent
whateverobject it is thatwe plan to extractlogically from an originatingconcept.
Frege, in his practice,invariablyutilizes the concepts' extensions-that is, sets-
as the selected concept-objects.But he also notes thatmore intensional"concept-
objects"might have been used in their stead, a topic to which we will return.I
believe that Frege is best understood if the actual jump into set theory is
postponeduntil the last possible moment.
In an allied manner,ratherthan positing that new points exist where parallel
lines meet, von Staudtproposeddefining a "pointat infinity"directlyin termsof
a concept-object derivable from the relation "x is parallel to L". In this case,
however, too many concept-objectswill be generatedif the criterionof identity
for concept-objects is held to follow the usual identification conditions for
concepts, for such a procedure will create a distinct concept-object for each
concept of the form "parallelto . . . ", even when substituentsL and L' represent
lines parallelto one another.Only one concept-objectought to stem from L and
U; otherwiseunwantedpoints at infinity will be produced.If we follow Frege in
utilizing the extension of "parallelto L" as the appropriateconcept-object,the
problem of an overabundanceof points at infinity vanishes. But there is no
necessity in this choice; a variety of alternativestrategies work with "concept-
objects"more intensionalthanraw sets, yet still producea propercomplementof
points at infinity. Insofar as I can determine, von Staudt's own discussion is
silent upon such questions of intensionality-he simply assumes that a common
concept-objectcan be extractedfrom the "form"common to all parallellines. In
this case, the desired concept-object possesses a familiar name, viz., "the
directionof L", even if someone might wish to debate the question of its exact
"intensional"statusfurther(I'll returnto these issues of "intensionality"later).
Prima facie, the reconstructionof a thought into "concept-object"format
seems as if it ought to be a process falling within the domainof logic. Certainly,
many logic texts of the period included sections on "abstraction"that read as if
they ought to embrace conversions of this sort. But nobody before von Staudt
had tried to define "concrete" objects like points in terms of these abstract
entities. Whereas Poncelet believed that new elements should be postulatedto
meet the demands of "persistenceof form", von Staudt saw that the persisting
forms themselves, reconstituted in concept-object guise, could serve as the
desired extension elements. Complexified geometry is now viewed as a logical
outgrowth of traditionalgeometry. Accordingly, we have obtained a second
explanationof the "naturalness"of the projectiveextensions.
In order to establish the claim that projective geometry is a non-creative
reorganizationof Euclidean geometry, von Staudt needed to supply more than

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions

this initial set of definitions. All of the notions utilized in projective geometry
must be redefined and its theorems shown to be transcriptions,under the
definitions, of regular Euclidean statements (albeit involving concept-objects).
Thus von Staudthad to redefine what a line is (to allow for the imaginaryones)
and then to redefine "lies upon".Armed with such definitions, claims like "two
points always lie upon some line" needed to be proven, even though, in the
originalEuclideanrealm,this propositionis an axiom.


Many 19th century mathematiciansfelt that von Staudt's explication provided

the betterrationalizationof the extension elements, due to its precision and non-
creativecharacter.Frege was almost certainlyamong their number.I'll postpone
my own appraisalof the approachuntil section (xi).
As with Poncelet, these results were not taken to show that the ideal points
were "fictions" or any less "real"than the regular points. The "real"and the
"ideal"form interconnectedpartsof the hiddenstructurethatlies behindthe facts
of Euclideangeometry. It is an accidentalfeatureof the way we happento learn
aboutEuclideangeometry that some of its partsappearto us clothed in intuitive
garb and some do not. As a result, we originally detect the presence of the ideal
elements only through the rigid relationships they induce upon the visible
objects. But anotherclass of beings might directly intuit the imaginaries,rather
thanthe reals.
Such claims have a long traditionin projective geometry, stretching back
before von Staudt's time. It was recognized that if one wanted a mental picture
of the extension elements, one could do so by shifting the "sensuousappearance"
of familiarelements over to the invisible ones (Laguerrewas an early proponent
of such methods-see (Coolidge 1924)). For example, one can use line-like
representationsto paint a useful picture of the real points at infinity. Although
there is no single visual representationwherein all of the elements of geometry
appearsimultaneouslyin intuitive garb, by accepting suitablecompromises,any
desired element can be pictured at the expense of others. Von Staudt's
innovation was to recognize that, as long as complex points were to be
"invisible"anyway, why not allow them to be abstract?Or, more exactly, why
not hold that the distinction between "abstract"and "concrete" object is
somewhatconventional,dependingupon our epistemologicalpoint of entry into
the relevantdomainof knowledge?
It is importantto recognize that Frege frequentlyendorsedtheses of this sort.
In the Grundlagen,Frege claims that two mathematiciansmight fully agree on
the content of projectivetheorems,despite the fact that one "intuits"geometrical
points as "looking like points"and the other as "lookinglike lines" (Frege 1884,
pp. 35-6)1 1. Frege also published some experiments,in the style of Laguerre,
that transferredthe intuitive appearanceof lines to selected imaginary points
(Frege 1873).

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
164 NOUS

Accordingly,one should not be mislead by Frege's frequentlyrepeatedclaim

that i'the only source of geometricaltruthlies in geometricalintuition".By this,
Frege means only that our entire body of geometry knowledge stems from facts
reported to us by "intuition".In agreement with Poncelet, Frege holds that
geometrical data bear no essential relation to any facts about numbers except
through"accidental"isomorphisms.On the other hand, althoughall of thefacts
of geometry must be reportedto us in intuition,it does not follow that all of the
objects proper to geometry must appear in intuitive garb-in particular,the
"invisible"complex points will not. Although the corpus of geometricalfacts is
fixed, their contents may be carved up to reveal unexpectedobjects such as the
imaginarypoints, which logic is able to extractfrom geometricalfact in the guise
of von Staudt'sconcept-objects.
Recognizing thatFrege views geometry in this fashion gives a deeperreading
to assertionslike:

Time and time again we are led by our thoughtbeyond the scope of our imagination,
without therebyforfeiting the supportwe need for our inferences.Even if, as seems
to be the case, it is impossible for men such as we to think without-ideas, it is still
possible for their connection with what we are thinkingof to be entirely superficial,
arbitraryand conventional.(Frege 1884, p. 72)

I suggest that the manner in which the "unimaginable"complex elements

"support"the otherwise uninterpretablesteps in the Cartesianreasoningpattern
discussed above providea nice exemplarof what Frege has in mind.12
On such a reading, Frege intends to reject not only "psychologism"in the
usual sense, but the deeperpresumptionthat definitionalpracticein mathematics
must respect the "intuitivetrappings"of a mathematicalobject, even in a field
such as geometry. Although Frege had not drawn the tone/sense/reference
distinction at the time of the Grundlagen,these geometrical examples suggest
that much of what we might naively regard as the "sense" of a mathematical
assertion would in fact have been assigned by Frege to "tone" (or some other
"non-objective"category). Likewise, although the "sense" of a propositioncan
always be computedas a functionof the "senses"of its componentphrases,such
a decomposition should not be regarded as unique-witness the constant
propositional content shared by an orthodox Euclidean statement and its
reparsingin terms of extension elements. Thus I agree with the commentators
who feel Frege had a "holistic"conception of propositionalcontent, ratherthan
one structurallyconstructedfrom the senses of its componentparts.
Another aspect of von Staudt's approachhad long standing antecedents in
geometry-viz., the tendencyto conceive of objects in theirconcept-like aspects
and vice versa. Thus Julius Plucker noted the following: in suitable (homo-
geneous) coordinates,the equationfor a line looks like Ax + By + Cz =0. If a
point (a, b, c) lies upon the line, it satisfies the equation,i.e., Aa + Bb + Cc = 0.
We normally regard this as a case of the function (frequentlycalled a "form")

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions

determinedby the coefficients A, B and C acting upon the point-object(a, b, c).

Plucker realized that the action could be also seen in reverse: a,b and c are
viewed as coefficients in a function now operatingon the line-object (A, B, C).
That is, the common thought"Aa + Bb + Cc = 0" can be carved up into distinct
functionsand objects, dependingupon one's point of view13.The termsA, B and
C representeither a functionor a line dependingupon how the thoughtis carved
up-in Frege's terminology,whether"(A, B, C)" is regardedin an "unsaturated"
or "saturated"way. This insight of Plucker'sprovedvery importantin unraveling
the mysteries of so-called "duality"in projectivegeometry (I might mentionthat
the purely synthetic work of Jacob Steiner, in rather different ways, also
contributedto the tendency to look for the function-likequalities of geometrical
objects such as points, lines and circles).
In general terms, nineteenth century geometers tended to conceptualize
mathematical entities in terms of the roles they play with respect to their
surroundings.The intersectionsof a circle and line form the basis arounda wide
variety of other geometrical constructions hover (viz., circular and linear
involutions, pole and polar pairs, etc.). When the same clustering of behaviors
occur in the apparent absence of the "fixed points" that generate the
constructions,it is proper, claimed Poncelet, to postulate the missing "inter-
sections".Von Staudtwould have said: The fixed points evaluate the behaviorof
the other constructionsin their neighborhood.If we continue to make identical
evaluations in the absence of regular points, it is permissible to regard the
evaluatingconcept itself as a new "object",for it serves the same basic purposes
that regular objects do. Frege's own discussions of "objecthood"often carry a
While on the subject of von Staudt's contributions,he also worked out an
isomorphismthat aligned a system of coordinatenames such as <6 + i, 9, -2i>
with his expanded set of geometrical points. Here von Staudt achieved a
remarkablegain in clarity-he showed that a mappingto so-called "projective"
coordinates can be set up without relying upon the metrical attributesof the
underlying geometry. In so doing, he was able to not only account for the
successes of analytic geometry (when the complex calculations mirrorgenuine
geometricalrelations),but also for its failures (e.g., why the principleof duality
fails when metrical relations are involved). The realization that coordinate
numbers can be meaningfully assigned to geometry without recourse to any
notion of "magnitude"(= metric)was, I believe, an importantstimulusto Frege's
views of number.

These geometrical considerationsgive us some sense of the freedom to rework
preexisting mathematicsthat Frege wished to defend. This freedom supplies a
way of legitimating the projective revisions without destroying geometry's
claims to permanenceand certainty.Let's now turnto Frege's own work on the
classical numbersystems.

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
166 NOUS

Frege probablycame to his logicism by wonderinghow one might determine

whetheralgebraiccalculationsinvolving complex quantitieswill supply correct
results within a particular field of endeavor-remember our perplexed
philosopher/engineer!Rather than seeing the rules of complex algebra as
unquestionable"leadingprinciples",capableof extractingsoundresultsfrom any
subject matter whatsoever, Frege viewed the general status of complex
calculationsin the terms that Poncelet and von Staudtadvocatedfor geometry-
the manipulationswork only if the field of complex numbers isomorphically
mirrorsstructurespre-existentin the subjectunderinvestigation15.The question
becomes: exactly what sort of structureis requiredfor the valid applicationof
reasoninginvolving complex numbers?
In Frege's time, it was frequentlypresumedthat the notion of real numberis
somehow "abstracted"from the notion of distance along a geometrical line.
Likewise, the notion of complex quantity "abstracts"from vector magnitude
within a two-dimensionalGaussian array.The trouble with "abstraction",as is
familiar from the stock problems of delineating a unique concept by merely
ostendinga groupof objects, is thatit is often unclearwhich featuresare retained
and which abandonedafter the "abstraction".The stock answer presumedthat
some bare notion of "magnitude"was the vital residue left after the abstraction.
As we noted, von Staudtshowed that useful calculationswith complex numbers
work within a geometry that contains no notion of magnitudewhatsoever-so
somethinghad to be wrong in the "abstraction"view.
Frege concluded,in agreementwith several other figures of his time, that real
and complex numbers serve only as a means to mark the places that given
relationsoccupy within largerfamilies of allied relations,families structuredby
relations of "addition"or "composition"and containing selected unit relations,
(By a family of relations,I mean somethinglike the collection;"x is 1 ft from y",
"x is 2 feet from y", etc.). Two relationswithin two differentfields-within the
fields of distance and angle16, say-should be assigned the same real number
only if the two relations correlate under the unique isomorphismbetween the
fields that preserves"addition"and aligns the two unit relations.The important
project,then, is to isolate what sort of structurea field of relationsmust have in
orderthat,e.g., the real numberscan be mappedonto them.
Accordingly, ratherthan regardingr as "abstracted"from concrete distances,
why not view the constructionin the crisp mannerof von Staudt-i.e., r is the
concept-object derived from the concept of two relations falling at the same
place within their respective families? Or, alternatively,a real or complex num-
ber serves as an "evaluator"of position within a family of relations.Such a story
gives a cleaner account of the intellectual processes involved in constructing
numbersystems than is suggested by vague appeals to "abstraction".We noted
that contemporaneouslogic books usually treatedthis as a process within their
ken; Frege instead saw "abstraction"as a hazy perceptionof the precise logical
process involved in converting a thought involving a concept into a thought
involving a concept-object(Frege's codificationof this logical move became his

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions

notorious Axiom V-the axiom that leads to the paradoxes). In my opinion,

Frege's logicism starts (although it doesn't conclude) with this assimilationof
number to von Staudt's treatmentof the extension objects within geometry.
Numbers are the concept-objectslogically induced by the comparativestudy of
families of relations.
It is always important,I think,to keep Frege's uniformapproachto the higher
number systems firmly in mind, for it is only within this wider context that
delimitingthe properrangeof applicationof a numbersystem seems a genuinely
urgent project. Few of Frege's contemporarieswould have puzzled whether
reasoning involving integers alone was likely to prove sound, but it was quite
otherwise with the complex numbers.We noted the rathermysterioussuccesses
of complex reasoningin analytic geometry;such inexplicabletriumphs(coupled
with some equally mysterious failures) occurred across the entire face of
nineteenth century science. We tend to miss the urgency of the general
applicationproblem if we only read the Grundlagen(as opposed to the earlier
writings where an especial concern with complex reasoning is more marked).
Such considerationslead me to presume that Frege must have roughed out his
basic approachto the highernumbersrelativelyearly in his development.


If these suppositionsare correct,then a variety of otherwise mysterious aspects

of Frege's philosophical remarks in the Grundlagen can be clarified. Many
commentatorshave found the discussion leading up to the definition of "natural
number"to be puzzling. Hans Sluga writes:

But thereis somethingsurprisinganddisturbingaboutthe definitionof numberin

termsof extensionsof contextsin the generalcontextof Frege'sthought.He had
originallyreasonedthatnumbersas logicalobjectshadto be definedcontextually
[atthe startof partIV]... Butthe conclusionof thatsectionwasthatthe attempted
definitionwhichfulfilledthatconditioncouldnot legitimatelybe adopted.(Sluga
1980,p. 127)

But if numbers are concept-objects induced in the von Staudt manner over
appropriatefamilies of relations,this confusion about"contextualdefinition"can
be sortedout. That Frege has the von Staudtanalogy in mind is suggestedby the
statementhe providesof his overall objectives:

Ouraimis to constructthecontentof a judgementwhichcanbe takenas anidentity

such thateach side of it is a number.We tirethereforeproposingnot to define
identityspeciallyfor this case, butto use the conceptof identity,takenas already
known,as a meansfor arrivingat thatwhichis to be regardedas beingidentical.
Admittedly,this seemsto be a veryoddkindof definition,to whichlogicianshave
notyet paidenoughattention; butthatit is notaltogetherunheardof, maybe shown
by a few examples.(Frege1884,p. 74)

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
168 NOUS

But what are the "notaltogetherunheardof' examples thatFrege bringsforth?In

fact, his chief example turns out to be von Staudt's definition of the points at
infinity in terms of the relation"x is parallelto y". Unfortunately,Frege chooses
to characterizethe account as providinga definitionfor the object "the direction
of line L" rather than "the point at infinity correspondingto L". There are
reasons why Frege wrote in this fashion, but the unsuspectingreaderis scarcely
alertedto the geometricalsignificanceof the "direction"concept-object.
In general, Frege often makes puzzling allusions to the radical natureof his
"logical"procedures,as when he writes:

Purethought,irrespective of anycontentgivenby thesensesor evenby anintuition

a priori, can, solely fromthe contentthatresultsfromits own constitution,bring
forthjudgmentsthatat firstsightappearto be possibleonly on the basisof some
intuition.Thiscan be comparedwithcondensation, throughwhichit is possibleto
transformthe air thatto a child'sconsciousnessappearsas nothinginto a visible
fluidthatformsdrops.(Frege1879,p . 55)

What does this have to do with his account of numbers? In von Staudt's
treatmentof geometry, our logical ability to reconstitutejudgments in terms of
concept-objectsreally does give the appearanceof pulling invisible points out of
empty air. Assimilating Frege's treatmentof numberto von Staudt's approach,
Frege's radical metaphors begin to seem more appropriate.Certainly, von
Staudt's trick of granting abstractconcept-objects such as "the direction of a
line" with the same mathematical"reality"as points at infinity was widely
regarded,in mid-nineteenthcentury,as a mathematicalgambit "borderingupon
impudence"17.Frege liked the gambit, and used it for his numbers,but in the
numericalcontext the shock value of the move was greatly diminished.Only by
recognizing that Frege's "directions"are really "points at infinity" can one
regainan appreciationof the radicalwork thatthe context principlewas meantto
accomplishwithin mathematics.
The apparentlyconflicting strandsof approvaland disapprovalof "contextual
definitions"found in Frege's thought trace to this: Any clear-cut propertycan
serve as the grounds for extracting a definite concept-object. As such, the
concept-objectcan be expected to possess the full range of propertiesnaturalto
abstractobjects-it must representa "self subsistentobject, subject to the usual
laws of identity"and the other featuresthat Frege so often emphasized.On the
other hand, these abstractobjects do not stand to regularpoints and lines in the
spatial relations needed in projective geometry. As long as we adhere to the
originalmeaningof "lies upon",concept-objectslike "thedirectionof L" or "the
involution 0" will lie upon no lines whatsoever-abstract objects do not possess
spatiallocations. Yet we saw that phraseslike "p lies upon L" can be redefined
so that all the claims of complexified projectivegeometry are validatedfor con-
cept-objects. But these furtherdefinitions must be done piecemeal-it is this
piecemeal quality that replaces what used to be regarded as "contextual

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions

In sum, Frege's puzzling ambivalenceabout"contextualdefinition"should be

understood as a recognition that von Staudt's procedures do justice to the
insights displayed in standard applications of "contextual definition", while
simultaneouslyavoiding the false illusion they engenderof unbridledcreativity
over the originaldomain.
On this reading, Frege's "context principle" can be seen as serving two
purposes:(1) to supportthe claim thatconversionof a thoughtto one involving a
concept-object does not change the underlying thought, despite the different
pictureof the thoughtthat the reorganizationmay convey; (2) to emphasize that
concept-objectsshould be regardedas possessing the same "objective"status as
any other mathematicalobject. One gets a better feel for the radical import of
these claims within von Staudt's original framework, where the "intuitive
contents"of geometrical propositionsseem to be disregardedin a surprisingly
high-handedway. Numbers being ratherremote objects to begin with, Frege's
applicationsof von Staudt'smethods do not trampleupon the claims of expected
arithmeticalcontentin so shamelessa fashion.

These observationsare not sufficient to carryus completely to Frege's logicism.
Although the bare creation of numericalconcept-objectsis sanctionedby pure
logic, this does not entail that their non-trivial behavior-e.g., how the
"addition" relation acts-is determined by logic alone. Such facts will be
determinedby the behaviorof the basic objects that make up the extensions of
the relevant concept-objects.In von Staudt's setting, the non-trivialbehaviorof
the points at infinity-e.g., which lines they lie upon-is determinedby the
behavior of the objects-the parallel lines-lying in the extensions associated
with the points. But notice this: the behavior of the points at infinity is
completely dictatedby the behaviorof a much smallersubset of objects thanare
found in the complete extensions. Suppose we want to know whether three
points at infinity a, b, and c lie upon a common line. Ratherthan worryingabout
all of the parallellines definitionallyassociated with a, select the single line La
passing throughthe origin as representative.We can likewise concentrateupon
the representativesLb and Lc. a, b, and c can now be determinedto be collinear
or not by examining the relationshipsamong La, Lb and Lc. In short, our points
at infinity will behave properly if the selected "representatives"do; we can
ignore what happenswith the rest.
Consider any of the standardsystems of numbers, say the naturalnumbers.
Any system of relations that spawns these numbers as concept-objectscan be
considered as a representative of the collection of numbers. Frege had the
brilliantidea thatone might be able to construct"representatives" for each of the
classical numbersystems within the ontology of pure logic alone. For the natural
numbers,his "representativestructure"is well-known;his approachto the reals
(and, presumably,the unpublishedtreatmentof the complex numbers)follows a
similar pattern. By the observationsjust made, the behavior of these logical

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
170 NOUS

representativesis sufficient to induce the correct behavior on the numerical

concept-objects-we don't have to look at all systems of relationsthat count as
"representatives" of our numbers.
Fromthis point of view, geometry,ratherthanlogic, could have been selected
as the basis for inducing the correct behavior upon Frege's numbers.
Epistemologically, however, it is preferableto utilize a logical representative,
because logic alone fixes how such objects act, whereas both logic and
geometrical intuition are needed to determinehow geometricalentities behave.
The existence of logical exemplars for each of the classical number systems
means that the facts about these numericalconcept-objectscan settled by appeal
to truths of logic alone-no higher source of knowledge is needed to get the
numbersbehavingproperly.Frege discussed what might happenif the imaginary
numberi were defined in a deviantmannerin termsof some concreteobject such
as the moon or a second of time:

Against the particularsense we have proposedto assign to "i", many objections can
of course be brought.By it, we are importinginto arithmeticsomethingquite foreign
to it, namely time. The second stands in absolutely no intrinsic relation to the real
numbers. Propositions proved by the aid of real numbers would become . . .
syntheticjudgments,unless we could find some other proof for them or some other
sense for i. We must at least first make the attemptto show that all propositionsof
arithmeticare analytic.(Frege 1884, p. 112)

Note thatthe main objectionis that, if an intervalof time is selected as a referent

for "i",syntheticfacts abouttime must be cited in orderto show thatthe complex
numbersbehave properly.This specialized reliance on synthetic facts ought to
make one worry whether complex reasoning could serve as adequate"leading
principles" in domains of inquiry that possess no analogue of the required
synthetic facts. Why should one think that Dirchlet's results on the distribution
of primes, obtained using reasoning over the complex numbers, have any
validity, given that no "time"is found among the integers? Or that Clebsch's
complex techniques in elasticity have any validity, given that this branch of
physics involves no temporalprocesses either?Frege put the point this way:

All we need to know [with respect to a system of numbers] is how to handle

logically the content as made sensible in the symbols and, if we wish to apply our
calculus to physics, how to effect the transitionto the phenomena.It is, however, a
mistake to see in such applications the real sense of the propositions; in any
applicationa large part of their generality is always lost, and a particularelement
enters in, which in other applicationsis replacedby otherparticularelements. (Frege
1884, p. 22)


On this reconstructionof Frege's logicism, it is not partof the argumentthat the

naturalnumbers, under analysis, should turn out secretly to represent unique

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions

logical objects. There is little suggestionin Frege of the quasi-Russellian18claim

that the number 3 must be identified with the collection of all 3-membered
concepts in orderto embraceall of 3's potentialapplications.Indeed, this claim
looks plausible only if one is tempted to suppose that 3 represents an
"abstraction"from 3-memberedsets, as Frege assuredlydid not. More generally,
the Frege/von Staudtapproachto definitionalpracticeis rathersophisticatedand
does not obviously succumb to the familiar difficulties scouted in (Benacerraf
This latitude in the treatment of number may surprise many of Frege's
readers,due to such passages as:

When we speak of "the numberone", we indicateby means of the definite article a

definite and unique object of scientific study. There are not divers numbersone, but
only one. (Frege 1884, p. 49)

But Frege's point is simply that in a reconstructionof arithmeticonly a single

number 3 ought to be set up, not that a wide variety of objects couldn't have
been selected as suitable. Looking back at von Staudt, one recognizes that
imaginarypoints could have been equally well defined in terms on involutions
aroundcircles ratherthan along lines, a fact thatFrege explicitly notes19.But one
should not define imaginary points simultaneouslyin terms of both kinds of
involution,for too many imaginarypoints will then be produced.
I stress these facts because it is easy to become confused about two different
methodological strands that are potentially involved in constructions like
Frege's. On the one hand, we have the von Staudt idea that the work of
"concrete"mathematicalentities can be duplicatedby more abstract"concept-
objects". On the other, there is the doctrinethat set theory can be used to free
particularmathematicalobjects from the harness of particular"representational
forms".For example, thereis a sense that,underGauss' approachto congruence,
the objects "3 mod 7" and "10 mod 7"20 are the "same", albeit tied to two
different numerical representations.Richard Dedekind proposed that Gauss'
"congruencenumbers"could be rendered"representationfree" by defining them
as equivalence classes such as {3, 10, 17,. . . ). This practice is ubiquitousin
modem mathematics-the standard definition of "manifold" liberates these
geometrical objects from the fetters of particular coordinate choices by
constructingthe manifold in terms of a complete "atlas"of coordinatecharts.
However laudablethe desire to constructmathematicalobjects in "representation
free" terms may be, it is not the same impulse that motivates von Staudt's
practice.A classic case where the two motivationsrun togethercan be found in
Dedekind's 1871 recasting of Kummer's "ideal numbers"in terms of the sets
algebraistsstill call "ideals".Kummermerely gave rules in "implicitdefinition"
fashion for calculating with his "ideal numbers";one hoped that a generalized
notion might apply in more general contexts. Dedekind found the generalization
by recognizingthat the "persistingforms"of Kummer'smethodcould be treated

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
172 NOUS

as "concept-objects"in von Staudt'smanner.But Dedekind also sought to make

his "ideals"representationfree, which forced them to become infinitaryin a way
they otherwise didn't need to be. A rather bitter debate ensued between
Kroneckerand Dedekind over the virtues and demerits of this "representation
The standardapproachto the points at infinity constructsthem as extensions
consisting of infinitely many lines. This infinitude can be avoided if specific
"representations" of each directionare selected instead, in the mannerdiscussed
several paragraphsback. That is, points at infinity are now defined in terms of
the concept-objects derived from "parallel to L", where the range of L is
restricted to the lines running through some selected origin. Such a
"representationbound" treatmentwould lead to considerableinconvenience in
the later definitions of concepts such as "lies upon", but at least the technique
avoids appeal to infinite totalities. If we wish, the concept-object"the direction
of L" could be credited with whatever intensionality we might like without
necessarily engenderingtoo many points at infinity. It is importantto observe
that, in the Grundlagen,Frege frequentlyexpressed some ambivalenceabouthis
employmentof extensions:

of a concept"is assumed
In [my]definitions thesenseof theexpression"extension
to be known.Thiswayof gettingoverthedifficultycannotbe expectedto meetwith
universalapproval,andmanywill preferothermethodsof removingthe doubtin
question.I attachno decisive importanceeven to bringingin the extensionsof
conceptsat all. (Frege1884,p. 117).

Insofaras I can determine,Frege was willing to concede that logic might afford
other ways of extracting concept-objects from concepts that result in objects
more "intensional"than extensions. Unfortunately,the usual vagaries in the
notion of "intension"make their propercriteriaof identity less than clear. Frege
apparentlyselected sets as the concept-objects utilized in his von Staudt-like
constructions because the conditions of set existence (i.e., Axiom V in the
Grundgesetze)seemed relativelyuncontroversial,even if, from a logical point of
view, extensions constitute objects of secondary importance.His views about
this topic may have changed as his views about the extensionality of
mathematicallanguage22developedon routeto "OnSense and Reference".
In the recent philosophical literature(e.g., Stein 1987), Frege is sometimes
chastised as an example of a mathematicianwho, driven by "bad"philosophical
motives, became committed to an impossible search for the "true essence" of
number, whereas the "good" Dedekind recognized that mathematiciansshould
be concernedonly with the "structures"of numbersystems. There are a variety
of ways in which this appraisalmisrepresentsthe positions of the two writers,
but, at a minimum, it should be recognized that Dedekind, in his demand for
"representationfree" accounts of mathematicalobjects, was somewhat more
"absolutist"about mathematicalconstructionsthan the historicalFrege seems to

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions

have been. I imagine that Frege would have found that, at first blush,
"representationfree" constructions count as aesthetically more pleasing than
their competitors, but that this virtue might pale if the resulting constructions
become overly complicated (as would happen if the "representationfree"
demand forces geometers to include both line and circle involutions within the
definitionof imaginarypoint).
While on this topic, we might examine an objectionto Frege's procedurethat
George Boolos makes in an importantarticleon Frege's approachto numbers:

The well-known distinction that Frege draws. . . between "the direction of line L"
and "the numberbelonging to concept F" is thereforeseriously misleading. We do
not suspect that lines are made of directions, that directions are some of the
ingredientsof lines.. . The principlethatdirectionsof lines are identicaljust in case
the lines are parallel looks, and is, trivial only because we suppose that directions
are... distinct from the things of which lines are made. (Boolos 1990).

In point of fact, in the von Staudt universe some lines-e.g., the lines at
infinity-are "madeof directions".Even worse, it is assumed that concepts can
apply to the very concept-objects manufacturedfrom them. For example, the
original motivation for constructing relation-objects to correspond to the
involutaryrelation"xx' = -10" was to supply two "points"that become mapped
into themselves by the original relation. In the nineteenthcentury, concepts (or
functions) were not regardedas intrinsicallytied to the objects in the domain in
which they are first encountered;otherwiseit makes no sense to maintainthat an
elliptic function's "truehome" is on a Riemannsurface.In this sense, one might
wish to resist the wholely extensionalpoint of view that makes the domain of a
functionconstitutiveof its very identity.
On the other hand, this point of view seems prima facie at odds with the
recognition that the meanings of "line" and "lies upon" must genuinely alter
when the shift to the extendeddomain is made. As is well known, Frege thought
the practiceof using a common symbol "-"to designatethe two "multiplication"
operationsover the reals and the complex numberswas a mistake, encouraging
mathematiciansto believe that they had proved facts for the complex realm
which had really been proved only for the reals. However, as long as the
meaning of the "persistingform" "xx' = -10" is treatedin sufficiently abstract
fashion, the two viewpoints can be reconciled. Frege's strictureson definability
do not entail that one cannot meaningfullyinquire whetherthe triple of objects
<number,number,operation>satisfy the form "xgx' = -10" in both the real and
complex domains.23Although the specific "multiplication"operationsdiffer in
the two domains, we can still study how a particular"form"behaves when we
move from the reals to the complex numbers.Indeed, it would be disastrousfor
Frege's philosophy of mathematicsif it could not constructsome sense in which
the behavior of elliptic functions (or involutions) could be compared over
different domains. After all, much of the genius of nineteenth century

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
174 NOUS

mathematics-and the primarymotivationfor the philosophies of Poncelet and

von Staudt-lay in the realization that the behaviors of certain mathematical
forms become more comprehensible when placed in an extended setting. In
short,the assumption24of which Boolos complainsis deeply imbeddedin much
intuitivemathematicalthought,not simply Frege's.

In summation:I have claimed that Frege's more radical philosophical claims,
especially those expressed in the Grundlagen, come into sharperfocus when
viewed from the perspective of the prior discussions of the extension element
problem within geometry. There are two primary reasons for the increased
vividness of the geometricalsetting:(1) Von Staudt'stechniquesin geometryrun
roughshodover the expected "sensuouscontent"of geometricalpropositions,to
the extent of harnessing erstwhile concepts to the work of concrete objects.
Frege's philosophical pronouncementson definition and context justify these
procedures,but their applicationswithin Frege's own context seem very tame,
largely because we lack a robust sense of the apparent"content"of arithmetical
propositions.(2) Nineteenthcenturygeometryrepresenteda disciplinein internal
tension, a tension arising from the need to admit new elements while retaininga
permanence in subject matter (we want to let geometry enlarge her living
quartersa bit without allowing her to bolt the stable altogether). Von Staudt
found an unusual approach to definitions that (temporarily) resolved these
conflicting demands upon geometry neatly. Frege's own views on definitional
practicewalk the same delicate line as von Staudt's,but, since Frege's numerical
context is a static one, we scarcely appreciatethe methodologicalpitfalls that he
felt constrainedto avoid in his philosophyof mathematics.
The "royal road" of this paper's title is intended somewhat whimsically.
Although virtually every road in nineteenth century mathematicsran directly
through geometry, they also-courtesy of Riemann's grand synthesis-wound
through virtually every other area of mathematics as well. Although I have
concentratedupon geometry here, it should be recognized that closely related
problems of "extension elements" developed elsewhere (e.g., Kummer's "ideal
numbers" mentioned above). Frege was clearly aware of these other
developments and of the deep ties that bound them. Given Frege's rather
unpleasanttendency to not acknowledge positive influences, any reconstruction
of his thoughtprocesses must per force be speculative.For the reasons cited in
the previous paragraph,the geometricalcase remainsthe most salient setting for
understandingthe various philosophical attitudes towards mathematicalexis-
tence that were popularin Frege's era. Accordingly,I have treatedit in isolation
here, althoughFrege would have viewed mattersfrom a richerperspective.
Recognition of such geometricalprecedents,of course, should not be seen as
minimizing the magnitudeof Frege's personalachievements.It is worth stating,
however, that certain modern commentators regularly undervalue the

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions

philosophicalmerits of Frege's rivals in a misguided attemptto glorify Frege's

particularform of genius.
In this respect, we should beware of treatingthe philosophical problems of
arithmetic in the hermetically insulated way that Frege's own writings
encourage.HermannSchubert,whose views on numbersFrege castigatedin an
unfairlyderisive review, had been deeply influenced in his approachto number
by Poncelet's philosophyof geometricalexistence (i.e., with heavy relianceupon
"continuity"or "persistenceof form"). Viewed independentlyof geometry, as
Schubertpresentedthe doctrineto his readers25,this principle looks absolutely
silly. Read in conjunctionwith geometry,"continuity"assumes a ratherprofound
aspect, especially in light of the astonishing "enumerative calculus" that
Schubertdeveloped in algebraicgeometry largely following its guide. The real
purposeof the "principleof continuity",for both Schubertand Poncelet, was to
rationalizethe "resettings"of mathematicalquestions which were vital to their
geometrical investigations. Schubert's work in arithmeticrepresenteda carry-
over of these philosophicalideas, into a setting where the appealsto "continuity"
largely lost theircuttingedge.
Indeed, looking at Schubert'sown geometrical successes, one begins to feel
that Poncelet's explication of the "correctness"of the projective revisions was
probably superior to von Staudt's. Although von Staudt's methods were
undoubtedlymore rigorousthanPoncelet's, rigor is not everything.Von Staudt's
methods are "copycat"proceduresin that one knows which concept-objectsto
extract only after an authoritylike Poncelet has located the important"forms"
that should persist in the extended domain. In contrast, the "principle of
continuity",with all of its vagaries, genuinely affords a discoveryprocedure for
uncovering the proper extended domain. When S.S. Chern writes that "figure
and numberfight for the soul of every geometer"(Chern1980), he is referringto
the way that questions naturalto visual geometry and questions naturalto the
algebra of coordinate geometry interact to drive geometry forward. We
witnessed an early example of this interaction in the Cartesian successes of
section (iii)-the subsequent development of algebraic geometry, in which
Schubertplayed an importantrole, illustratethe continuing influence of these
tensions, leading to abstractionsfar beyond those found in projectivegeometry.
"Principles of continuity" hew closely to the considerations26that force the
geometricalchanges, whereasvon Staudt'smethod submergesthese motivations
as an artifact of the "context of discovery". Despite these limitations, von
Staudt'sapproachwas widely judged in Frege's time to be superiorto Poncelet's
(albeitmore tedious).
Frege's work is commonly characterizedas contributingto the nineteenth
century "rigorization of analysis", a phrase that normally encompasses
investigations of notions of limits, series, the assumption of lower bounds,
notions of derivative and integral, and the like. In fact, Frege's own writings
stand at a remarkabledistance from concerns of these sorts. Rarely does he

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
176 NOUS

comment upon the burning issues within analysis in his time-e.g., the
possibilities for evading the use of Dirchlet's principle within Riemann's work.
Of course, Frege was quite rigorousin his own projects,but this rigor does not
appearharnessedto higher mathematicalends in the mannerof Weierstrassand
other contemporaneousworkerson foundations.On the other hand, the puzzles
of reconstitution of domain, so acutely raised within nineteenth century
geometry,do not hinge upon worriesabout"rigor"per se, but upon questionsof
how unexpectedmodificationsof old definitions and theoremscan be permitted
in mathematics,while retainingsome sense thatthe discipline's subjectmatteris
permanent and immutable. Even though the mathematical task of defining
"naturalnumber"is not, on the face of it, an "extension element" problem, I
submit that if we see Frege's work as conceived in the shadow of the
methodological concerns arising in geometry, we will better understandthe
characterof his philosophyand the mannerin which he conceived his logicism.
Returning once again to the "reconceptualization"issues with which we
began, I believe that, even today, we scarcely understandthe character of
mathematicalconcepts-how it is thatthey can "growand thrive"over time. Too
much concern with set theoretic formalization-whose purpose, after all, is
simply to granta concept a temporaryfoothold on rigor-may have dimmedour
eyes to this fact. Perhapsthe realizationthat one of the founding parentsof set
theoretic rigor was deeply concerned with such issues may serve, as a bit of
carbonateof ammonium, to reawaken us to the deep problems of conceptual
growth within mathematics,even if Frege's particularanswers can no longer be
regardedas wholly satisfactory.

1I would like to thank Tom Ricketts, John Bell, Burton Dreben and, especially, Bill
Demopoulos and Ken Mandersfor helpful comments. Bob Battermandrew the figures.
2 Several good accounts of the projective revolution are: (Coolidge 1940), (Merz 1965), (Nagel
1971). Interestingevaluationsby Frege's contemporariescan be found in (Klein 1925) and (Cayley
By "projective geometry", I mean "geometry of the school founded by Poncelet", a sense in
which the metrical attributesof the expanded geometrical realm are not excluded. A possibly less
misleading term would be the old-fashioned "geometryfrom a higher point of view", a phrase that
properly suggests that Euclidean geometry is not being overthrown or diminished, but simply
3 The two processes, complexification and projective completion, must be performed con-
currently.To avoid undue technicalities, I shall sometimes illustrateprocedures within the affine
partof a projectiveplane.
4 Benacerrafintends this remarkas a gloss on David Hilbert's views, who was unlikely to view
the extension elements of geometry in the mannerdescribed.
5 Historically,"formalist"responses like Rosenbaum'sbecame prevalentamong geometers after
the publication of David Hilbert's Foundations of Geometry. One wonders whether some of the
crankyanimus that Frege displays towardsHilbertin their interchangesmay trace to a sense that the
philosophical concerns common to Poncelet and von Staudt are likely to be plowed under in
Hilbert'sapproach.Insofaras I am aware,however, this issue is not explicitly raisedby either writer.
For an interestinglaterevaluationby a greatgeometerof the Italianschool, cf. (Enriques1929).
6 Chiefly among the English:Peacock, Boole, etc.
7 (Scott 1900b) is one of the best surveys of von Staudt's thought and, unlike the notoriously
turgidoriginal, attemptsto articulatethe philosophy behindthe method.

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions

8 It is possible, of course, to maintainthat geometry simply is a portion of complex analysis, a

piece, in George Birkhoff's phrase, "dressed up in glittering intuitional trappings". An early
advocate of this point of view was Julius Plucker.
9 But when L moves still furtherfrom the center, the points A and B become invisible; yet, for
the sake of continuity,we say they still exist, but are invisible or imaginary"(Russell 1893, p. 23).
10Indeed, in my description of the circle and line cases above, I largely refrained from
presentingthe gory details of the algebraand only supplied the geometric "effect"that the algebraic
manipulationswill give (some of these "effects", of course, representing"imaginary"points and
lines). Poncelet's "synthetic"viewpoint maintains that my algebra-freedescription of "effects" is
the properway to handle the geometry, whereas the fullfledged Cartesiantreatmentof our example
merely introducesmessy complicationsthat trace to the avoidable entanglementof our figures with
the "extraneous"coordinate lines X and Y. Selection of fiducial lines X and Y is needed to map
geometry into the complex numbers,but Poncelet's contention is that such numberassignmentsare
11 Compare: "[S]tereometric ideas can be correctly comprehended only when they are
contemplated purely by the inner power of the imagination, without any means of illustration
whatever"-Jacob Steiner, quoted in (Reye 1898, p. xiii). Echoes of these claims can be found in
12 Von Staudt's discovery, sketched below, that Cartesian reasoning can prove valid in the
absence of a metric illustratesa furtherimplicatureof this quotation.
13 The tangent and cotangent bundles of a manifold exhibit this dual behavior nicely. Such
examples, I believe, are very helpful in unraveling what Frege intended by "the multiple
analyzabilityof a thought".For work in Plucker's vein, see (Frege 1884a).
14 It is worth mentioningthat least one of Frege's stock examples of a non-materialobject shares
some of this evaluative character,vis. the axis of the earth (Frege 1884, p.35). A theorem of rigid
body mechanics says that, at each moment, the earth's movement can be decomposed into a
translation and an instantaneous rotation about an axis, but this axis will usually wander with
respect both to the surface of the earth and to an inertial frame. In short, the laws of mechanics
define an "evaluator"that, at each instant, isolates a spatial line to serve as the present position of
the abstractcontinuant,"axis of the earth".Frege's holistic views of "inertialframe"in (Frege 1885)
are interestingin this respect.
15 Frege published several papers (Frege 1874; 1878) that demonstrated how a variety of
unlikely mathematicaldomains could be coded into complex numbersin such a way that standard
analytic techniquesprove sound over the encoded domain.
16 See (Frege 1874, p. 56). Also:
According to the old conception, length appearsas something material which fills the straight
line between its end points. . . [But on a modem conception] all that mattersis the point of origin
and the end point; whether there is a continuous line between them. . . appears to make no
difference whatsoever; the idea of filling space has been completely lost. . . The concept [of
quantity] has thus gradually freed itself from intuition and made itself independent.This is quite
unobjectionable,especially since its earlierintuitivecharacterwas at bottom mere appearance... If,
as we have shown, we do not find the concept of quantityin intuition but create it ourselves, then
we are justified in trying to formulateits definition so as to permit as manifold an application as
For a good accountof Frege's approachto the reals, cf. (Simons 1987).
17 See (Freudenthal1981). This excellent article is highly recommendedto interestedreaders.
18 Despite tradition,Russell didn't seem to claim uniquenesseither,cf. (Russell 1964, chapterXI).
19(Frege 1873, p. 2). The alternativeapproachto imaginary points that Frege cites is that of
circles and lines combined in pairs,ratherthan a second form of involution.
20 That is, 3 and 10 leave the same remainderon division by 7.
21 For an excellent discussion of the criticisms of Dedekind, see (Edwards 1980; 1987). In this
essay, I have distinguished von Staudt's general contention that abstract"concept-objects"might
serve as adequate replacements from more concrete seeming entities from the more specific
proposal that equivalence classes should serve as the "concept-objects".Insofar as I can determine,
von Staudtis unspecific about the exact natureof the entities he utilizes; the earliest explicit use of
equivalence classes seems to be Dedekind's 1871 reconstructionof Kummer's ideal numbers.One
would be curious whether Frege was aware of Dedekind's work at the time of the Grundlagen,
although Charles Pierce claimed that the equivalence class idea was in the mathematicalair long
before Dedekind (Peirce 1976, p. 130). Much work needs to be done on how the chains of influence

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
178 NOUS

might have run in this period (for valuable speculations, see (Freundenthal1981)). Not all uses of
equivalence classes, by the way, seem as philosophically radical as the von Staudtapplications;for
example, if one regardsthe Gaussiancongruence classes as carving new "modulararithmetics"out
the regularintegers. In this kind of case, there is no pressing need to tell a philosophical story a la
von Staudtthat the new domains constructed"have the same reality"as the basic set of objects over
which the construction proceeds. Dedekind clearly conceived all appeals to equivalence classes,
even in von Staudt-likecontexts, in the benign light of the Gaussianparadigm(which, historically,
he was among the first to advance). In essence, the thesis of this paper is that many of the
philosophical concerns in the Grundlagenhave been misunderstoodbecause we now tend to view
all uses of equivalence classes from an uncriticalpoint of view ratherlike Dedekind's.
22 To be precise, Frege might have doubted whether"function"could be properlytreatedin the
set-like way familiar from modern courses on "functionsof a real variable".In the early 1880's it
was more common to presumethat functions, while extensional in their identity conditions, enjoy a
richer, non-set-like charactersuch as they are supplied in the context of "functions of a complex
variable".By the time of the Grundgesetze,"functions of a real variable"was well on its way to
being widely accepted, so Frege may have relaxed his qualms about sets (of course, many logicians
in Frege's time would continued to object to the prevailing "extensionality" among
mathematicians).For a good history of the analysis side of these issues, see (Bottazzini 1980). For a
contemporaneous expression of attitude, cf. (Cayley 1889b). William Demopoulos is currently
investigatingthese difficult matters.
23 Strictly speaking, the "10" and the "-" also shift referentsin the move to the complex realm,
so they should be reassignedto the "tuple"side of this question.
24 Indeed, the sometimes obnoxious restrictionson constructionthat Boolos' stricturesrequire
still representa bone of contentionamong category theorists.
25 Cf. (Schubert 1899). For an interesting perspective on Schubert, see (Kleinman 1976). It is
worth mentioning that various philosophers attemptedto extend the "principleof continuity"into
generalphilosophy of language,e.g. (Cassirer1953).
26 Kenneth Manders, in unpublished work, has given a very deep analysis of the factors that
push the early developments.For geometry's latercareer,see (Dieudonne 1985).

1989a "MathematicalTruth"in Benacerrafand Putnam(eds.), Readings in the Philosophy of
Mathematics,Prentice-Hall,Englewood Cliffs.
1989b "Whatthe NumbersCould Not Be" in Benacerrafand Putnam,op cit.
Boolos, George
1990 "The Standardof Equalityof Numbers",preprint.
Bottazzini, Umberto
1980 TheHigher Calculus, W.V. Egmond (trans.),Springer-Verlag,New York.
1953 Substanceand Function, translatedby W.C. and M.C. Swabey, Dover, New York.
Cayley, Arthur
1889a "PresidentialAddress to the British Association, September1883", in his Collected Papers,
XIII CambridgeUniversityPress, Cambridge.
1889b "Function"in his Collected Papers, op. cit.
1980 "FromTrianglesto Manifolds",AmericanMathematicalMonthly.
Coolidge, J.L.
1924 The Geometryof the ComplexDomain, ClarendonPress, Oxford.
1940 A History of GeometricalMethods,ClarendonPress, Oxford.
Davis, Phillip
1987 "WhenMathematicsSays No" in No Way-the Nature of the Impossible,Phillip J. Davis
and David Park(eds.), W.H. Freeman,New York.

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions

1985 History of Algebraic Geometry,JudithSally (trans.),Wadsworth,Monterey.
1980 "TheGenesis of Ideal Theory",Archivefor the History of the Exact Sciences, 23.
1987 "Dedekind'sInventionof Ideals"in Studies in the History of Mathematics,ed. by Esther
Phillips, MathematicalAssociation of America.
1929 The Historical Developmentof Logic, MacMillan,New York.
Frege, Gottlob
1873 "On a GeometricalRepresentationof ImaginaryForms in the Plane"in (Frege 1984).
1874 "Methodsof Calculationbased on an Extension of the Conceptof Quality"in (Frege 1984).
1879 Begriffsshrift,Stefan Bauer-Mengelberg(trans.),in From Frege to Godel, Jeanvan
Heijenoort(ed.), HarvardUniversityPress, Cambridge1967.
1878 "Lectureon a Way of Conceiving the Shape of a Triangleas a Complex Quantity",in (Frege
1884 Die Grundlagender Arithmetik,Translatedby J.L. Austin as The Foundationsof Arithmetic,
Harper& Brothers,New York 1960.
1884a "Lectureon the Geometryof Pairs of Points in the Plane"in (Frege 1984).
1885 "On the Law of Inertia"in (Frege 1984).
1984 Collected Papers on Mathematics,Logic and Philosophy, Brian McGuiness (ed), Basil
Blackwell, Oxford.
1981 "The Impactof von Staudt'sFoundationsof Geometry"in P. Plaumannand K. Strambach
(ed), Geometry-von Staudt'sPoint of View, D. Reidel, Dordrecht.
1949 Divergent Series, ClarendonPress, Oxford.
Klein, Felix
1925 The Developmentof Mathematicsin the NineteenthCentury,translatedby M. Ackerman,
Math Sci Press, Brookline (no date; originallypublishedin 1925).
1976 "RigorousFormulationof Schubert'sEnumerativeCalculus"in F. Browder(ed.),
MathematicalDevelopmentsArisingfrom Hilbert Problems, AmericanMathematical
Society, Providence.
1987 "Logic and ConceptualRelationshipsin Mathematics"in Logic Colloquium'85, Elsevier
1989 "DomainExtension and the Philosophy of Mathematics",Journalof Philosophy, Vol.
XXXVI, 10.
Merz, John Theodore
1965 A History of EuropeanScientific Thoughtin the NineteenthCentury,Vol. II, Dover,
New York.
Nagel, Ernest
1971 "TheFormationof ModernConceptionsof FormalLogic in the Developmentof Geometry"
in Teleology Revisited, ColumbiaUniversityPress, New York.
Peirce, Charles
1976 TheNew Elementsof Mathematics,Vol. III/1, Mouton,The Hague.
Poncelet, Jean-Victor
1987 Traitedes Proprietes Projectives des Figures, J.J. Gray, trans.,in The History of
Mathematics-AReader,John Fauvel and J.J. Gray (eds.), MacMillanPress, Houndsmill
Reye, Theodore
1898 Lectureson the Geometryof Position, translatedby T.F. Holgate, MacMillan,New York.

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions
180 NOUS

1963 Introductionto Projective Geometryand ModernAlgebra, Addison-Wesley, Reading.
Russell, Bertrand
1964 The Principles of Mathematics,Norton,New York.
Russell, J. W.
1893 An ElementaryTreatiseon Pure Geometry,ClarendonPress, Oxford.
1899 MathematicalEssays, Open Court,Chicago.
Scott, CharlotteAngas
1900a "The Statusof Imaginariesin Pure Geometry",Bul. Amer.Math. Soc. VI.
1900b "Onvon Staudt'sGeometrieder Lage",Math. Gazette.
Simons, Peter
1987 "Frege's Theory of Real Numbers",History and Philosophy of Logic 8.
Sluga, Hans
1980 GottlobFrege, Routledge and Kegan Paul, London.
Stein, Howard
1988 "Logos, Logic and Logistike"in MinnesotaStudies in the Philosophy of Science XI, ed. by
W. Aspray and P. Kitcher,MinnesotaUniversityPress, Minneapolis.
von Staudt,G.K.C.
1847 Die Geometrieder Lage, Erlangen.
1856 Beitrage zur Geometrieder Lage, Nuremberg.
Weyl, Hermann
1955 The Conceptof a RiemannSurface,Addison-Wesley,Reading.

This content downloaded from on Sat, 22 Feb 2014 22:35:14 PM

All use subject to JSTOR Terms and Conditions